As shown in the inset in Fig. 1a, four site-controlled pyramidal InGaAs QDs evenly separated by 450 nm are deterministically placed along the x-axis in a modified L7 PhC cavity (hole pitch a = 225 nm) and are arranged symmetric with respect to the cavity center. Figures 1a and 1b present an artistic illustration of our experimental observation of the spatial superposition of the QD excitonic states in an extended cavity photonic state of the 1st order cavity mode (CM1). The total electrical field intensity distribution patterns of the 1st order cavity mode (CM1) calculated using the 3D finite-difference time-domain (FDTD) method is overlapped with the L7 cavity region in Fig. 1 (see also Figure S2 in Supplementary Information for the x- and y-polarized electric field distribution patterns of the fundamental cavity mode (CM0) and CM1). CM1 exhibits a spatially extended pattern with two spatial lobes along the x-axis of the L7 cavity, which are also artistically represented by two yellow clouds. We observe that the photogenerated QD2 exciton (blue balloon) can be coupled with the left or right lobes by controlling the detuning between exciton and CM1 energies as illustrated in Fig. 1. Such spatially distributed coupling of QD2 excitons with two lobes of CM1 can lead to photon emission (blue wavy arrows) at different locations of the L7 cavity.
We first excite four QDs simultaneously using a large laser spot (≈ 3 µm) with high power (5 µW) for each device with various air hole radii. In this case, the saturated QD transitions exhibit a broadband emission. All CMs are sufficiently pumped and can be easily identified (see Figure S3 in Supplementary Information). Specifically, we identify and carefully examine two 4QD-L7 devices, denoted as device 1 and 2, with PhC hole radii r1 = 32 nm and r2 = 31 nm, respectively. These two devices exhibit good spectral overlap between QD and CM1. At a temperature T of 10 K, the µPL spectra of devices 1 and 2 are measured with a small laser spot (≈ 1µm) at the center of the cavity as shown in the bottom spectrum of Fig. 2a and 2b respectively. Then we subsequently scan the focused laser spot across QD1 to QD4 along the x-axis of the L7 cavity. The variation in the exciton emission intensity can be used to identify the specific QDs contributing to each emission peak in the µPL spectrum. The results of PL scanning are summarized as intensity bars on top of each exciton peak. For device 1, four exciton lines (X1-X4) are identified in proximity to CM1 and identified to be corresponding to the same QD. For device 2, eight exciton lines are observed and associated with QD1, QD2, and QD3 while QD4 is too weak to be measured. Note that some exciton emissions are contributed from different QDs simultaneously. For example, as shown in Fig. 2b, the exciton line at ≈ 980.46 nm is contributed by both QD1 and QD2, while the exciton line at ≈ 988. 34 nm is contributed by QD2, QD3, and QD4 (contribution from QD4 is weak but is visible with scale that saturates QD2 and QD3 as shown in Figure S4 in the Supplementary Information). Note that for exciton line at ≈ 988.34 nm, the high intensity area roughly corresponding to the QD1 position is from the saturation of CM1 intensity. For simplicity, we associate these two exciton peaks to QD2 which dominates the emission compared to the contribution from other QDs. CM0 is also observed at the longer wavelength side of the µPL spectra (≈ 12 meV below the CM1) for both devices. The cavity Q-factor of device 1 is measured to be ≈ 12,000 for CM1 and ≈ 9,000 for CM0. For device 2, the Q-factor is measured to be ≈ 11,000 for CM1 and ≈ 7,000 for CM0. The observed Q-factor in our study is comparable to the recently reported value for single site-controlled pyramid QD in an L7 PhC cavity operating at the onset of the strong coupling regime [52]. Notably, these Q-factors are more than twice of previously reported values of similar devices operating in the intermediate coupling regime (Q ≈ 4,500) [51] or weak coupling regime (Q < 3,500) [46, 54, 55]. Such improvement in cavity Q-factor is achieved by reducing the pyramid nominal size to approximately 200 nm which leads to red shifting the QD emission energy to ≈ 1.24 eV (≈ 1,000 nm). At this photon energy, the reduced absorption losses from the Urbach tails of GaAs can be achieved, leading to an improved cavity Q-factor [52]. As shown in Figure S5 in Supplementary Information, the second-order photon correlation of the cavity emission which is tuned in between QD2-X2 and QD1-X1 exhibits photon antibunching with g2(0) ≈ 0.9. The large g2(0) value suggests that each QD emits photon individually through the cavity decay channel, and no cooperative spontaneous emission (superradiance) is observed. This lack of superradiance is likely due to the relatively large pure dephasing process, non-resonant pumping, and possible different coupling efficiency of QDs with the CM1.
A linear polarizer and λ/2 waveplate are used to obtain the polarization-resolved µPL spectra. The degree of linear polarization (DOLP) of QD emission is determined using the expression: \(\:DOLP=\left[\right({I}_{y}-{I}_{x})/({I}_{y}+{I}_{x}\left)\right]\), where \(\:{I}_{x}\) and \(\:{I}_{y}\) are the intensity of the QD emissions in the x- and y-polarized directions. Note that CM0 and CM1 are dominant in y-polarization and exhibit large positive DOLP values due to the geometry of L7 cavity. The tuning of QD exciton lines across the CM1 is achieved by adjusting sample temperatures. The temperature-dependent polarization-resolved µPL and the corresponding DOLP of devices 1 and 2 are shown in Figs. 3 and 4 (see Figure S6a and S9a for the accompanying spectral data). For device 1 as shown in Fig. 3b, excitonic transition X2 is tuned through CM1 from T = 6 K to 46 K, reaching a maximum DOLP of approximately 0.96 around T = 34 K when X2 is in resonance with CM1. Other exciton lines exhibit lower but positive DOLP values at all temperatures. For device 2 as shown in Fig. 4b, excitonic transitions QD1-X1, QD2-X2, and QD3-X2 are tuned through CM1 subsequently by adjusting temperatures from T = 6 K to 48 K. Large positive DOLP is observed when they are in resonance with CM1, and other exciton lines exhibit lower but positive DOLP at all temperatures (see also Figure S6b and S9b for the accompanying DOLP spectra). We define the exciton-CM detuning as \(\:{\delta\:}_{X}={E}_{X}-{E}_{CM}\), which represents the relative energy difference between exciton emission (\(\:{E}_{X}\)) and CM (\(\:{E}_{CM}\)). DOLP as a function of \(\:{\delta\:}_{X}\) can serve as an indicator of the QD-cavity coupling which is summarized in Fig. 5b for device 1 and Fig. 6b for device 2. As a result, the DOLP are positive with values > 0.5 for all QDs, due to both phonon scattering and pure dephasing [51, 54, 55]. The co-polarization of all QDs with CM1 suggests they can be simultaneously coupled with the same CM1 for device 1 and 2.
The exciton emissions and CM1 are then resolved spatially on the vertical axis of the CCD. To achieve this measurement, the sample is oriented such that the axis of the PL image corresponding to the x-axis of the L7 cavity is parallel to the spectrometer slit. The far-field image of the higher-order CMs, extending along the x-axis of the cavity can thus be spatially mapped by a proper choice of the lens focal length for focusing the beam on the spectrometer slit. However, the position of four QDs in the studied sample cannot be effectively identified using this method as a result of the limited spatial resolution \(\:{\epsilon\:}\) of the objective with numerical aperature (N.A.) due to the long emission wavelength \(\:\lambda\:\) (\(\:{\epsilon\:}\propto\:\frac{\lambda\:}{N.A.}\)) and low SNR due to the relativly weak exciton emission intensities. In Fig. 5a, the spectrally resolved far-field image of the spatial distribution of X2, X3, and CM1 of device 1 is shown at varying temperatures (see Figure S7 of Supplementary Information for full temperature range data). Note that the vertical axis corresponds to the enlarged spatial distribution along the x-axis of the cavity. A saturated logarithm color scale is used to facilitate the readibility of excitons. From the finite-difference time-domain (FDTD) calculated E-field intensity pattern, CM0 shows a single spatial lobe and CM1 exhibits two spatially extended lobes (Figure S2). With increasing temperatures, X2 is tuned across CM1 and X3 is tuned away from CM1. The vertical positions of QD excitons are extracted by Gaussian fitting their line profile and their relative values to the center of CM1 as a function of detuning \(\:{\delta\:}_{X}\) are then summarized in Fig. 5c for y- and x-polarized components. Interestingly, the y-polarized components of QD excitons tend to coalesce with the lower (upper) lobe of the CM1 when they are brought in resonance with the CM1 from the positive (negative) detuning side. Overall, this spatial behavior exhibits a shape of spatially avoided crossing within a detuning range of approximately ± 3 meV, where the DOLP of QD exhibits large positive values.
In contrast, QD excitons in x-polarized directions do not show such spatial features as shown in the lower panel of Fig. 5c. Note that, in QW systems, the spatial features of the excitonic state from its superposition with the photonic state cannot be distinguished because of the extended feature of QW excitons. In our presented QD system, fortunately the localized QD excitons show spatial coalescence with lobes of the extended photonic states at small detuning (not superimposed with photonic states at large detuning). It reveals how the subwavelength-confined excitonic states can be spatially deviated from its original state as a result of interaction with a spatially cavity extended photonic state. Note that in the polaritonic picture, such deviated excitonic state can be regarded as polaritons with more excitonic components (lower polaritonic arm at negative detuning or upper polaritonic arm at positive detuning). Rabi spectral splitting is not resolved in this case due to the much stronger CM1 emission intensity compared to the QD exciton lines.
To further investigate the effect of QD positions, as shown in Fig. 6, a similar measurement is conducted on device 2 where QD excitons can be identified and associated with different QDs (see Figure S10 in Supplementary Information for full temperature range data). Again CM1 exhibits upper and lower spatial lobes along the x-axis of the cavity but CM0 shows only a single lobe (see Figure S10). With increasing temperatures, QD2-X2 is tuned across the CM1 and QD1-X1 is tuned out of resonance with the CM1. QD2-X1, QD3-X1, and QD3-X2 are tuned towards the CM1 while QD2-X3, QD1-X2, and QD1-X3 are tuned away from the CM1. Vertical positions of y-polarized components of two center QDs (QD2 and QD3) exhibit spatially avoided crossing with the center of the CM1 within approximately ± 3 meV detuning range, corresponding to large positive DOLP values, although QD3 is only at the positive detuning side in the measured temperature range. Interestingly, excitons from side QD (such as QD1) do not show such phenomena. Again, the avoided crossing is not observed in the x-polarized component of all QD excitons as shown in the lower panel of Fig. 6c. A recent study [53] on the DOLP proposes a theory of Fano-like [56–57] quantum interferences between QD decay channels when it is coupled to the CM0 of a L3 cavity: direct QD decay into x- and y-polarized free space modes (FMs) and QD decay into y-polarized FMs through cavity decay channels. In our device, inspired by Ref. [53], with four QDs and two lobes of CM1, for QDj (j = 1, 2, 3, and 4), the total QD emission rate into the y-polarized free space mode mediated by lobe k (k = 1, 2) is given by
\(\:{W}_{yjk}=\frac{\kappa\:+{\gamma\:}_{y}}{2}-Re{[{\left(\frac{\kappa\:-{\gamma\:}_{y}}{2}-i{\delta\:}\right)}^{2}-\left(2\left|{g}_{jk}\right|-i{\chi\:}_{jk}\sqrt{{\gamma\:}_{y}\kappa\:}{e}^{-i{\varphi\:}_{jk}}\right)\times\:\left(2\left|{g}_{jk}\right|-i{\chi\:}_{jk}\sqrt{{\gamma\:}_{y}\kappa\:}{e}^{i{\varphi\:}_{jk}}\right)]}^{1/2}\) | (1.1) |
where \(\:\kappa\:\) and \(\:{\gamma\:}_{y}\) are the cavity decay rate and direct QD decay rate into y-polarized FMs. \(\:{g}_{jk}\) is the coupling strength of QDj with lobe k. \(\:{\chi\:}_{jk}\) represents the spatial overlap of the direct emission patterns of QDj and field patterns of lobe k. \(\:{\delta\:}\) is the QD-CM detuning. \(\:{\varphi\:}_{jk}\) is the relative phase difference between decay channels which depends on the QD positions with respect to the lobe. In this framework, observation of detuning dependent spatial features of QD2, for instance, can be due to the dominance of \(\:{W}_{y21}\) over \(\:{W}_{y22}\) at the positive \(\:{\delta\:}\) and vice versa for negavie \(\:{\delta\:}\). Then in the small detuning region, we could phenomenologically use \(\:{W}_{y21}/{W}_{y22}\) to mimic the spatial feature of QD2 excitons which is shown as the blue curve in Fig. 6b, with \(\:\kappa\:\) = 100 µeV, \(\:{\gamma\:}_{y}\) = 0.7 µeV, \(\:{\chi\:}_{21}\) = 0.9 and \(\:{\chi\:}_{22}\) = 0.2. Assume \(\:{g}_{21}={g}_{22}={g}_{2}\) for simplicity, the blue curve with \(\:{g}_{2}=45\:{\mu\:}\text{e}\text{V}\) mimic the QD2 behavior well. Similar analysis can be done on device 1 as shown as the red curve in Fig. 5b with a coupling strength of 45 µeV. Such coupling strength aligns well with similar multi-site-controlled QD systems [58]. It is interesting to note that QD2 and QD3, which are symmetric with respect to the center of the cavity, exhibit the same spatial shape as shown in Fig. 6c. It implies the existence of the symmetry breaking of two spatial lobes of CM1 possibly due to the fabrication disorder of PhC membrane or perturbation induced by the pyramid. This leads to additional strongly-confined localized modes of CM1 spatial pattern [59–60], an additional contributions to the above framework which needs to be further investigated. To further confirm these observations solely occuring with CM1, we extend our measurement to device 3 (PhC hole radii r3 = 36 nm), where we investigate the interaction of QDs with CM0 (see Section VII in Supplementary Information). Unlike CM1, CM0 is not spatially extended and shows a single lobe. Excitons from one of the central QD (QD3-X2 and QD3-X3) are tuned away from CM0 with increasing temperature. No avoided crossing is observed within a negative detuning range of 4 meV. It implies that the observed spatial avoided crossing is correlated with the extended feature of CM1.
Based on our experimental observation, we propose the concept of a functional spatially-distributed single-photon source using site-controlled InGaAs QDs embedded in a L7 PhC cavity, evanescently coupled with two waveguides (WGs) along the x-axis as shown in Fig. 7. The two side WGs are terminated close to two lobes of the CM1 to facilitate single-photon extraction. As shown in Fig. 7a, detuning of excitons with CM1 can be controlled for QD2 to allow single-photon propagation either to the left (negative detuning) or right (positive detuning) using evescent field coupling. Such operation can also be expressed in the logical bases, as shown in the upper left inset of Fig. 7a. With positive detuning, the output state will be |0L1R>; with negative detuning, the output state is |1L0R>. Here we define the positive detuning as logic 0 and negative detuninng as logic 1, and similarly define left-propagating photon as 0 and right-propagating photon as 1. To construct a truth table for the outputs, we retrieve the PL intensity of exciton emission when it is tuned close to CM1 from prior spatially- and spectrally-resolved PL measurments. Specifically, the positive and negative detuning values used are + 1.5027 meV and − 0.5857 meV respectively, and the resulting truth table is shown in lower right inset of Fig. 7a. We evaluate the output logic using the expression: \(\:F=\left(1/2\right)Tr\left(\frac{{M}_{exp}{{M}_{ideal}}^{T}}{{{{M}_{ideal}M}_{ideal}}^{T}}\right)\), which calculate the fidelity between the measured truth table \(\:{M}_{exp}\) and the ideal truth table \(\:{M}_{ideal}\), and obtain an estimated fidelity of 70.76 ± 1.87% for the output states [61]. Such logic operation is suitable for building single-photon sources capable of switching operations which control photon propagating along different directions. Particularly, using optical pumping from remote quantum dots [62–64] and dynamic Stark-based ultrafast control of detuning [65], it can open doors for possible low-photon-number all-optical switching [66–70] which are crucial components in linear optical and photonic quantum computers.
Based on the symmetry breaking of the CM1 lobes, Fig. 7b shows a L7 PhC cavity with two central QDs (QD2 and QD3) simulatanously embedded. When the CM1 detuning energy falls between the QD2 and QD3 excitons, single-photons from QD2 and QD3 (with different colors) can propagate to the L and R directions independently. Such behavior facilitates a tunable single-photon source for path encoding. By tuning CM1 to the middle of QD2 and QD3 (lower right inset of Fig. 7b), the single-photons can be emitted through the left or right waveguides, encoded as |L > or |R>. Due to the different energies of the QD2 and QD3 emission (ω2 and ω3), the resulting state can be expressed as |1ω2, L> or |1ω3, R>. Such a two-color tunable single-photon source behaves as a photon switch with different path encodings, which provides possibilities for spatially multiplexed quantum communications [71–73], quantum Boson sampling [74–77], and programmable information processing [78–86].