Difference-frequency generation in an AlGaAs Bragg-reflection waveguide using an on-chip electrically-pumped quantum dot laser

Second-order ($\chi^{(2)}$) nonlinear optical processes [1, 2] have been widely utilized in quantum computation, communication and metrology [3–11]. In addition, second-harmonic, difference-frequency (DFG) and sum-frequency generation, can be used to prepare and measure ultrafast laser pulses [12, 13], erase spectral distinguishability between single photon emitters [14] and engineer lasers at otherwise hard-to-reach wavelengths [15]. Today, most $\chi^{(2)}$ photonic devices rely on crystals such as BaB₂O₄, LiNbO₃ or KTiOPO₄ that exhibit high nonlinear conversion efficiencies, but are limited in scalability as they cannot directly host active optical components such as lasers and photodetectors.

Photonic circuits in the direct band-gap semiconductor AlGaAs enable integration of active components on a platform with strong $\chi^{(2)}$ nonlinearity. Significant effort has been committed to develop miniaturized, on-chip AlGaAs devices with increasing complexity and richer applications [16–18]. In particular, AlGaAs Bragg-reflection waveguides (BRWs) facilitate phase-matching in the telecom band and exhibit strong mode confinement, which has led to large nonlinear conversion efficiencies up to $2.5 \times 10^{-2}$ % W⁻¹ cm⁻² for DFG [19]. Simple waveguide circuits [20], entangled photon pair sources [21, 22], single photon emitters [23] and on-chip lasers [24] have been demonstrated with BRWs. However, there has been limited development of devices that exploit active components, such as lasers, and the strong second order nonlinearity at the same time [25, 26]. Recent work demonstrates DFG conversion efficiency of up to 169% W⁻¹ cm⁻² by utilizing an internal quantum well laser [27].

In our approach, we demonstrate an AlGaAs BRW with an embedded Al_0.13In_0.47Ga_0.4As quantum dot (QD) laser. This material composition enables the laser to emit at a wavelength around 790 nm at room temperature and increases the gain in comparison to conventional InGaAs QDs [28, 29]. The device is engineered such that the higher-order Bragg mode phase-matches at this wavelength with the fundamental total internal reflection (TIR) mode in the telecom band. The nonlinear waveguide comprises the active gain medium in an edge-emitting cavity, forming our electrically driven pump laser. We determine the threshold current of the internal laser and use the chip temperature to tune its spectral properties, such that it pumps type-2 nonlinear processes. By injecting a telecom signal laser into the waveguide, we generate an additional telecom idler field via DFG. We reconstruct the joint spectral intensity (JSI) by spectrally resolving the generated idler field, and extract a nonlinear conversion efficiency of up to $(0.64 \pm 0.21)$ % W⁻¹ cm⁻². While this is a clear improvement in comparison to previously reported BRWs without internal laser (passive structures), the device requires further engineering to match the results in reference [27].

We fabricate our BRWs starting with an epitaxially grown Al_x Ga$_{1-x}$As stack with varying aluminum concentrations x and layer thicknesses [30]. We subsequently pattern waveguides by electron beam lithography and dry etching in Ar and Cl₂ plasma. In this work, we utilize a 1.48 mm long and 3.5 µm wide waveguide sample with an etch depth of about 4.6 µm. A schematic is depicted in figure 1. The waveguide geometry is designed for tight confinement of the Bragg and TIR modes and this leads to a simulated effective nonlinearity of $d_\textrm{eff} = {56}$ pm V⁻¹ (see supplementary section 1 (available online at stacks.iop.org/JOPT/23/085802/mmedia)). The device is planarized with Benzocyclobuthene (BCB) to allow for depositing electrical contact pads larger than the waveguide width. The layer structure enables modal phase-matching between the TIR mode at ∼1580 nm confined to the core and matching layers, and higher order Bragg modes at ∼790 nm confined by the Bragg-mirror stacks [30, 31]. Two quantum dot layers are embedded at the center of the waveguide core and are surrounded by Al_0.17Ga_0.83As quantum wells to increase carrier accumulation. We dope the upper and lower mirror stacks as well as the substrate with carbon and silicon atoms to create a p–i–n junction. The mechanically cleaved waveguide facets have ∼32% reflectivity and form a cavity surrounding the gain medium. The QD dipoles are oriented to emit horizontally polarized light into the Bragg mode.

Zoom In Zoom Out Reset image size

We use a pulsed voltage supply to characterize the QD laser with 0.4% duty cycle (8 kHz repetition rate, 500 ns pulse width) to reduce the thermal load on the sample while still achieving peak currents above the lasing threshold. This is necessary, as parts of the gold layer are too thin to withstand the electrical power necessary for continuous wave lasing (see supplementary section 2). We increase the peak voltage of the pulse generator and measure the current and generated optical power. This way, we determine the lasing threshold of $(618 \pm 4)$ mA, which corresponds to a current density of $(11.9 \pm 0.5)$ kA cm⁻², as can be seen in figure 2. The insets depict the spectrum of the laser below and above the threshold current, respectively. This result of the lasing threshold is comparable to similar integrated lasers in BRWs [25]. However, optimized structures can offer significantly lower lasing thresholds down to a few tens of milliamperes [24, 27].

Zoom In Zoom Out Reset image size

In addition to the introduction of an internal QD laser, we engineer our BRWs to enable second-order nonlinear optical processes. The frequency conversion obeys energy and momentum conservation

$$\begin{align} \omega_\textrm{p}& = \omega_\textrm{s}+\omega_\textrm{i} \end{align} \tag{ 1 }$$

$$\begin{align} k_\textrm{p}(\omega_\textrm{p})& = k_\textrm{s}(\omega_\textrm{s})+k_\textrm{i}(\omega_\textrm{i}), \end{align} \tag{ 2 }$$

where ω_j and $k_j(\omega_j) = n_j\omega_j/c$ are the frequency and wavenumber with $j\in\{\textrm{p, s, i}\}$ corresponding to pump, signal and idler and c is the speed of light. The wavenumbers are directly proportional to the effective refractive indices n_j of the corresponding modes. Thus, we engineer the waveguide geometry to yield matched effective refractive indices at the desired wavelengths [32].

We characterize the phase-matching by measuring the laser spectrum and second harmonic generation while tuning the temperature (see supplementary sections 3 and 4). Due to the resulting change in the material bandgap and refractive index, the central wavelength of the laser can be shifted. This way, we determine a working point at 3.5 ^∘C and 792.9 nm for the pump laser, at which we can investigate DFG.

We utilize the setup shown in figure 3(a), where we use a half-wave plate and a polarizer to control the power and polarization of the seed laser, which is coupled into the waveguide with a 100× microscope objective. The generated idler photons are then collected with a 0.68 NA aspheric lens and coupled to a spectrometer with a high-sensitivity InGaAs CCD. We use polarization, longpass, and bandpass filters to suppress the pump and seed lasers.

Zoom In Zoom Out Reset image size

We characterize the DFG process with the telecom seed laser set to 1614 nm and a normalized power of 909 µW. By turning on the internal QD laser as the pump, we stimulate the corresponding DFG idler emission at 1558 nm as can be seen in figure 3(b). We tune the seed laser wavelength from 1530 nm to 1640 nm to measure the JSI—the correlation of signal (seed) and idler (DFG) wavelengths—which is enabled by the broad phase-matching of the BRW. The resulting JSI is shown in figure 3(c). Close to degeneracy ($\lambda_\textrm{s}\approx\lambda_\textrm{i}$), the bandpass filter can no longer suppress the seed and we therefore cannot record the DFG spectrum, as the laser oversaturates the detector. Narrow bandwidth filters at different central wavelengths would be necessary to measure the DFG in this region. Thus, we potentially miss the highest conversion efficiency, as our samples are designed for degeneracy.

Finally, we investigate the nonlinear conversion efficiency of our BRWs. We use the same settings as before, but fix the seed laser wavelength while tuning its power. This way, we measure the power dependency of the integrated DFG spectrum, as shown in figure 3(d). By adding a linear fit, we extract the slope b in units of Hz mW⁻¹ and the offset, which corresponds to the constant background (i.e. photoluminescence, dark counts and stray light). Note that we have chosen to tune the seed power since photoluminescent background light is not promoted by telecom wavelengths and is thus a constant in our measurement [33]. We determine the conversion efficiency

$$\begin{equation} \eta_\textrm{conv} = \frac{1}{\eta}\frac{P_\textrm{i}}{P_\textrm{s}P_\textrm{p}L^2} = \frac{1}{\eta}\frac{b E_\gamma}{P_\textrm{p}L^2}, \end{equation} \tag{ 3 }$$

where P_p, P_s and P_i are the normalized powers of the pump, signal and idler beams within the device (see supplementary section 5), respectively. The waveguide length is L = 1.48 mm and E_γ is the energy of the frequency converted photons. We normalize the result for the total loss in our experimental setup $\eta = {-54.94}$ dB for telecom light (see supplementary section 6). The normalized internal QD laser power is set to $P_\textrm{p} = {2.436}\$µW and we thus extract a conversion efficiency of ($0.64 \pm 0.21$)% W⁻¹ cm⁻². For comparison, we perform DFG in the same waveguide using an externally coupled pump laser and present the results in table 1 (more results can be found in supplementary section 7).

Table 1. Highest measured conversion efficiencies for different measurement settings, using the external and internal laser.

Type	Normalized power	λp (nm)	λs (nm)	λi (nm)	b (Hz mW−1)	$\eta_\textrm{conv}~(\%\,\textrm{W}^{-1}\,\textrm{cm}^{-2})$
Internal laser	2.436 µW	792.9	1626	1548	$5.70 \pm 155$	$0.64 \pm 0.21$
External laser	2.687 mW	792.9	1619	1554	$409 \pm 8$	$0.041 \pm 0.006$
Passive sample	2.88 mW	767.3	1510	1560	$(12.34 \pm 0.002) \times 10^3$	$0.72 \pm 0.12$

As these results are fully normalized, we expect the same for both types of measurements with the active device. The factor of ${\sim}16$ difference in efficiency can be explained by the internal laser coupling more efficiently into the correct Bragg mode. Previous measurements [34], have shown that about 4%–8% of the external pump power is coupled into this mode. This is a significant advantage to using the on-chip laser where the gain medium emits at an anti-node of the optical mode.

We also perform DFG using a similar BRW, but without the QD layer or doping and with comparable phase-matching wavelengths. For this, we use the external pump laser set to 2.88 mW normalized power at 767.3 nm and the seed laser set to 1510 nm in the same experimental setup, which yields a conversion efficiency of $(0.72 \pm 0.12)$% W⁻¹ cm⁻² for a DFG emission at 1560 nm. This value is about a factor of 18 higher than for the active sample when using the external pump. As the conversion efficiency is normalized for all losses and the modes for both devices are very similar, we explain this by a reduction of the effective nonlinearity in the active device due to its layer structure and aluminum concentrations. The type-2 process is the most efficient one in our previous structures—typically about twice as strong as type-0 and type-1. In the active BRW, however, its signal strength has been decreased by about a factor of 10 in comparison to the other types. The cause are thickness variations of the AlGaAs layers along the length of the waveguide due to imperfections in the MBE growth process. This leads to phase-matching at only certain parts of the waveguide, thus reducing the conversion efficiency. We expect that optimization of the device [30] could increase the conversion efficiency by at least an order of magnitude. Furthermore, the losses within our active waveguide of −4.458 dB mm⁻¹ for telecom light are much higher than typical values in our passive waveguides of about −2 dB mm⁻¹. These additional losses are caused by the QDs and doping of the sample. In addition, adding a reflective coating to the waveguide facets would improve the laser cavity and decrease the lasing threshold. Combined with thicker gold contacts, this could enable DC pumping.

In conclusion, we have demonstrated an integrated QD laser embedded in an AlGaAs BRW structure. By controlling the temperature of the chip, the internal laser wavelength can reach the nonlinear phase-matching condition, allowing for electrically pumped nonlinear processes. We have determined the nonlinear conversion efficiency of this sample via DFG, and made a comparison by externally pumping the process in this device and a passive structure. Even though our BRW does not reach a conversion efficiency as high as in devices based on a quantum well laser [27], a comparison with our passive structures shows the potential for improvement. In principal, QD lasers allow for a lower threshold current [35], a higher spectral range and a large modulation bandwidth in comparison to quantum wells systems [36]. This flexibility can be important for advanced integrated photonic circuits. Furthermore, the tighter confinement of the modes in our device leads to a higher simulated effective nonlinearity of $d_\textrm{eff} = 56$ pm V⁻¹. Since DFG uses the same principle and nonlinear effects as spontaneous parametric down-conversion our work presented here marks another step towards fully integrated active and nonlinear photonic circuits.

The authors acknowledge funding by the Austrian Science Fund (FWF) under Projects Q3 (IGUANA), project I2065 and the Special Research Program (SFB) BeyondC Project No. F7114, the DFG Project No. SCHN1376/2-1, the EU H2020 quantum flagship program UNIQORN (Grant No. 820474) and the State of Bavaria. We thank B Pressl for laboratory assistance.

Data underlying the results presented in this paper are available in [37] as found here www.osapublishing.org/submit/review/data-availability-policy.cfm.

Conceptualization, AS, RJC, SF, GW; Formal analysis, AS, MG; Methodology, AS, RJC, SF; Investigation, AS, MG; Resources, HT, HS, MK, SH, CS; Software, AS, MG, RJC; Supervision, RJC, SF, GW; Writing—original draft, AS, RJC; Writing—review and editing, All Authors; Funding acquisition, CS, GW.