Deterministic fabrication of GaAs-quantum-dot micropillar single-photon sources

TL;DR

This work demonstrates a scalable, deterministic fabrication workflow for integrating droplet-etched GaAs QDs into micropillar cavities, achieving unity QD positioning yield across 74 devices and precise spectral alignment within low-Q cavities. Under p-shell excitation, the emission exhibits biexponential decay and notable intensity fluctuations dominated by charge noise, which are partially mitigated by low-power above-band LED stabilization, doubling the source efficiency to about $9\%$. Hong-Ou-Mandel measurements reveal limited photon indistinguishability (raw visibilities around $\sim\!14\%$, rising to $\sim\!21\%$ with stabilization) due to slow relaxation dynamics and residual dephasing, with a coherence time of $\tau_c\approx219$ ps. The study also tests cylindrical rings around micropillars to suppress background modes, predicting up to a $4\times$ gain in collection efficiency, while experiments show a more modest $\sim\!1.3\times$ improvement, underscoring fabrication sensitivity. Overall, the deterministic approach delivers reliable, high-yield devices and provides detailed insight into charge noise and relaxation dynamics as critical factors for optimizing SPS performance.

Abstract

This study investigates the performance of droplet-etched GaAs quantum dots (QDs) integrated into micropillar structures using a deterministic fabrication technique. We demonstrate a unity QD positioning yield across 74 devices and consistent device performance. Under p-shell excitation, the QD decay dynamics within the micropillars exhibit biexponential behavior, accompanied by intensity fluctuations limiting the source efficiency to < 4.5%. Charge stabilization via low-power above-band LED excitation effectively reduces these fluctuations, doubling the source efficiency to $\sim$ 9%. Moreover, we introduce suppression of radiation modes by introducing cylindrical rings theoretically predicted to boost the collection efficiency by a factor of 4. Experimentally, only a modest improvement is obtained, underscoring the influence of even minor fabrication imperfections for this advanced design. Our findings demonstrate the reliability of our deterministic fabrication approach in producing high-yield, uniform devices, while offering detailed insights into the influence of charge noise and complex relaxation dynamics on the performance.

Figures (5)

• Figure 1: Simulation of micropillar structures. (a) Collection efficiency as a function of wavelength of a planar structure for different numerical apertures (NAs). The vertical dashed lines indicate the pre-selected QDs wavelength range (775 to 785 nm). Inset: Purcell factor of the planar structure as a function of wavelength. (b) Collection efficiency (NA = 0.82) and Purcell factor as a function of micropillar diameter at the central wavelength of the QD distribution (778 nm). Purple dashed: Collection efficiency of the planar structure in (a).
• Figure 2: Deterministic fabrication of single-QD micropillar structures. (a) Photoluminescence image of the QDs in a planar structure and (b) image of alignment markers taken with a confocal imaging setup. (c) Processed image of the combined QDs and markers showing the extracted location of the alignment markers and pre-selected QDs. (d) Scanning electron microscope image showing the fabricated micropillar structures containing a single QD. (e) Example of photoluminescence spectrum of one of the pre-selected QDs before and after structuring, showing the spectral red-shift of the emission peak. Inset: Histogram of the spectral shift distribution. (f) Reflectance intensity of five different devices, showing the QDs spectrally positioned within the cavity mode. Inset: Zoomed-in plot of the QDs peaks.
• Figure 3: Single-photon emission properties of QDs in a micropillar structure. (a) Second-order correlation of a device's emission under p-shell pulse excitation. (b) Time-resolved photoluminescence measurements of representative QDs in planar and micropillar devices (P1 ($D = 1.58$$\mu$m) and P2 ($D = 1.76$$\mu$m)) under p-shell excitation. IRF is the instrument response function of the SNSPD measured with a pulsed laser at the wavelength of the QD emission. The measurements were taken with a 20 ps time bin. (c) Histogram of decay times of various QDs in planar and micropillar devices. The quoted decay times are obtained from a mono and bi-exponential function fit for the planar and micropillar samples, respectively. (d) Detected count rate as a function of pump power for a planar and the P1 device. The right axis shows the source efficiency at the first lens obtained after accounting for the setup detection efficiency. Solid curves represent fits to the saturation curve $C_{0} (1 - \text{exp}(-P / P_{0}))$. Inset: the photoluminescence spectrum of the measured devices.
• Figure 4: Charge stabilization of QD under p-shell excitation with additional above-band LED excitation. (a) Photoluminescence map as a function of time for device P1, showing the intensity fluctuations of the emitter under p-shell excitation (left) and the stabilization of the emitter under additional low-power above-band LED excitation (right). For each frame (1 s), the PL spectrum was acquired for 0.25 s at pump power $P= 10$ nW. (b) Detected count rate and source efficiency as a function of pump power for the planar and micropillar device P1. Solid curves represent fits to the saturation curve $C_{0} (1 -\exp(-P / P_{0}))$. Inset: Detected count rate as a function of time (integration time = 100 ms) for the measurement with and without LED, normalized to the average count of each measurement. (c, d) and (e, f) Hong-Ou-Mandel two-photon interference experiment coincidence histograms taken at $P=2P_{0}$ for measurements with and without LED, respectively (bin width = 100 ps). The visibility is obtained by integrating the data over the full 12.5 ns window. Solid lines correspond to fit using a double-sided mono-exponential function. For clarity, a temporal offset of 4 ns was added between the parallel (HOM$_{\parallel}$) and orthogonal (HOM$_{\perp}$) polarization measurements. The uncertainty on the raw visibilities was calculated from the assumption of a Poissonian distribution of the total counts of each peak.
• Figure 5: Investigation of the effect of adding 2 rings around a micropillar. (a) Simulated collection efficiency as a function of ring thicknesses and air gap. (b) Collection efficiency enhancement produced by the rings over a bare micropillar for the region defined by the black dashed box in (a). The contour lines indicate the experimentally measured enhancement factor. (c) Scanning electron microscope image of the fabricated rings around the micropillar, targeting the highlighted region in (a). (d) Detected count rate and source efficiency as a function of pump power for micropillars with and without rings under p-shell excitation, with additional LED excitation for charge stabilization. The rings were fabricated with a r$_1$ = 0.6 $\mu$m fixed airgap.