Impact of Phonons on Dephasing of Individual Excitons in Deterministic Quantum Dot Microlenses.

Optimized light-matter coupling in semiconductor nanostructures is a key to understand their optical properties and can be enabled by advanced fabrication techniques. Using in situ electron beam lithography combined with a low-temperature cathodoluminescence imaging, we deterministically fabricate microlenses above selected InAs quantum dots (QDs), achieving their efficient coupling to the external light field. This enables performing four-wave mixing microspectroscopy of single QD excitons, revealing the exciton population and coherence dynamics. We infer the temperature dependence of the dephasing in order to address the impact of phonons on the decoherence of confined excitons. The loss of the coherence over the first picoseconds is associated with the emission of a phonon wave packet, also governing the phonon background in photoluminescence (PL) spectra. Using theory based on the independent boson model, we consistently explain the initial coherence decay, the zero-phonon line fraction, and the line shape of the phonon-assisted PL using realistic quantum dot geometries.

Owing to the technological progress in semiconductor growth, self-assembled quantum dots (QDs) offer nowadays optimal quality of the residing exciton transitions[1] with near unity emission efficiency[2−4] and close to ideal quantum optical properties.[5] Forthcoming applications emerging from combining QDs and nanophotonics, such as quantum light sources in on-chip photonic networks, call for scalability and deterministic QD positioning. In this regard, in situ electron beam lithography[6,7] (EBL) has evolved into a suited technique for the deterministic fabrication of quantum light sources.[8,9] When combined with low-temperature cathodoluminescence imaging, EBL permits sculpturing microlenses above individual QDs, enhancing collection efficiency over a broad spectral range.

Here, using four-wave mixing (FWM) microspectroscopy we reveal coherences of single QDs, deterministically embedded in microlenses realized by in situ EBL. The resulting optical signals in these nanophotonic structures exhibit an enhanced signal-to-noise ratio. This enables us to infer the impact of acoustic phonons on the coherence dynamics of individual QDs. In particular, we report on the exciton zero phonon line (ZPL) dephasing close to the radiative limit in a single QD at 5 K, distinguishing it from its spectral wandering. Phonons are known to play a crucial role in the optical control of QDs.[10,11] With increasing temperature, we observe an increasing impact of phonon-induced dephasing[12] owing to the polaron formation and wave packet emission[13,14] and a broadening of the homogeneous width γ, attributed to a quadratic coupling between carriers and acoustic phonons.[16,17] Single QD microspectroscopy permits associating the measured dephasing during the polaron formation with the spectral shape of phonon-assisted transitions, here accessed via photoluminescence (PL). We thus go beyond the FWM experiments performed on QD ensembles,[1,12] and we consistently explain the initial FWM decay, the ZPL fraction, and the line shape of the phonon-assisted PL using a realistic QD geometry. Additionally, the optical parameters of QDs, i.e., dephasing, lifetime, dipole moments, phonon coupling, are to some extent averaged in QD ensemble measurement due to stochastic distribution of their shapes and alloy composition. This issue is naturally overcome in a single QD spectroscopy carried out here.

When performing FWM on single emitters on a simple planar structure, one is confronted with a huge ratio between the resonant background (typically 106–108 in the field and 105 when assisted by high-quality antireflection coatings[18]) and the induced FWM. We have recently shown that, using suitable photonic nanostructures,[19,20] one can boost the experimental sensitivity by bringing a large amount of the field amplitude to the vicinity of a QD. This enhances its interaction with the excitonic dipole, and hence reduces the required external power constituting the background. Furthermore, using nanophotonic devices, the FWM is efficiently collected by the detection optics, avoiding the total internal reflection affecting planar structures: assuming the gain in the collection efficiency η and an n-time enhancement of the local excitation intensity, the FWM amplitude is increased by , while maintaining the external power of exciting laser pulses .

In contrast to the previous work, the in situ EBL technique overcomes the issue of a low yield of optimally functioning devices, when patterning the sample containing randomly distributed QDs. Microlenses processed with in situ EBL can be defined deterministically, spatially matched to QDs with about 30 nm alignment accuracy,[21] and combined with frequency matching guaranteed by their broadband extraction enhancement spanning across a bandwidth of about 50 nm.

Microlenses with a height of 0.35 and 2 μm diameter have been etched, as described in detail in ref (9) so as to create the enhancement of the excitation and detection mode field exactly at the QD plane, located between the underneath Bragg reflector and the microlens surface;[8] see inset in Figure a. We use InAs QDs grown by metalloorganic chemical vapor deposition.[23] As a result of the high light-extraction efficiency[22] of around 30% in our devices, we routinely note a bright QD photoluminescence, with a spectrally integrated count rate of 200 kHz below the PL saturation (not shown). For comparison, up to 1 kHz PL count rates are typically observed in our setup for high-quality QDs embedded in planar samples.

Figure 1

(a) Spectrally resolved FWM amplitude generated by a few QD excitons embedded in a lens structure. The selected QD trion (GX±) is labeled with the green ★. The horizontal bar indicates a neutral exciton–biexciton system (GXB) in a QD located at the lens periphery. Inset: Calculated distribution of the near-field intensity for the QD-lens structure. The semiconductor–air interface is shown by the solid black line, and the distributed Bragg reflector starts below the dashed line. (b) Spectrally integrated FWM amplitude of a trion in the target QD as a function of the pulse area of . The blue line shows the fit to the expected |sin(θ1/2)| amplitude dependence of the FWM.

For the FWM spectroscopy we use radio frequency acousto-optic deflectors providing frequency shifts of Ω1,2,3 for , respectively. FWM signals, which in the lowest order are proportional to (* denoting the complex conjugate), are then detected by performing optical heterodyning. To select the required heterodyne beat component, we interfere the reflected field with a frequency-shifted reference field . Spectrally resolved interferograms are recorded by a CCD camera[18,20] and analyzed by spectral interferometry to retrieve amplitude and phase of the signal. In Figure a we show an exemplary two-pulse FWM spectrum (driving with and and detecting at the heterodyne frequency 2Ω2 – Ω1), over a range of 21 meV displaying several excitonic transitions. Owing to a strongly improved excitation and collection compared to planar structures,[24] we here achieve a gain in the signal-to-noise ratio of the measured FWM amplitude by 2 orders of magnitude. Exciton–biexciton pairs can be identified (an example is denoted with a green bar above 1375 meV) by employing FWM polarization and delay selection rules[25] (not shown). A similar FWM signal was found on another investigated microlens, pointing toward the deterministic character and high quality of the in situ EBL nanoprocessing platform.

In the following, we study the dominating transition at 1360.4 meV labeled as ★ in Figure a. This is to minimize the time required to perform the following FWM sequences and thus to avoid drifts. We observe no FWM at τ12 < 0, showing that it stems from a charged exciton (trion) transition GX±.[25] To illustrate the enhanced in-coupling of offered by the microlenses, in Figure b we present the FWM amplitude as a function of the pulse area , where P1 represents the intensity of , while P2 is fixed to 1.5 μW. The FWM amplitude displays a Rabi rotation following the expected |sin(θ1/2)| dependence[26] with the first maximum at μW1/2, corresponding to a pulse area of θ1 = π/2. For higher intensities the FWM signal deviates from this behavior.

To measure the exciton density lifetime T1, we employed the three-pulse FWM, where the signal is detected at Ω3 + Ω2 – Ω1, as a function of the second delay τ23, displayed in Figure . From its exponential decay we determine the lifetime[19,20] as T1 = 347 ± 12 ps at T = 5 K. Such a rather short lifetime, compared to about 1 ns typically observed[27] in these structures, is attributed to the selectivity of the FWM technique favoring particularly bright QDs with a high dipole moment and thus displaying fast population decay dynamics (less intense transitions in Figure a are expected to exhibit longer T1). Additionally, the radiative lifetime is slightly shortened due to a Purcell effect of the microlens.[9]

Figure 2

(Top) Three-pulse sequence employed to measure the trion population dynamics. and , having a delay of τ12 = 20 ps, create the trion population and are jointly advanced in time, such that the FWM triggered by probes the population decay via the τ23 dependence. (Bottom) Measurement yielding the exciton lifetime T1 = 347 ± 12 ps. The noise level is indicated by open circles.

We now turn to the coherence dynamics as a function of temperature, to determine the impact of the phonon interaction on the exciton dephasing. The FWM transient, generated in our time-averaged and multirepetition heterodyne experiment, is emitted after the arrival of . It is expected to exhibit a Gaussian echo,[19,24] owing to a Gaussian spectral wandering of standard deviation σ, with the maximum at t = τ12 and temporal full width at half-maximum (fwhm) of . To retrieve σ, we apply the pulse sequence depicted in Figure a; namely, we keep τ12 = 110 ps fixed, while scanning the delay τ2R between and . As such, the temporal sensitivity of the experiment S(t) (green curve centered around ), originating from the finite spectral resolution of the spectrometer, is scanned through a broad FWM transient. A measurement of S(t) is given in the Supporting Information Figure S1. The FWM integrated overlap between S(t) and the echo plotted against τ2R in Figure a indeed reproduces a Gaussian form, with the expected maximum at τ12 = 110 ps and σ = 8.2 μeV. The measured inhomogeneous broadening (fwhm) is plotted versus temperature in the inset, where we find that it varies only marginally within the investigated temperature range.

Figure 3

(a) (Top) Two-pulse sequence applied to probe the echo profile. (Bottom) Integrated FWM amplitude versus τ2R at 5 K revealing the Gaussian echo with a temporal width yielding σ; the theoretical fit is given by the solid line. Inset: Inhomogeneous broadening retrieved from the echo temporal width for different temperatures. (b) (Top) Two-pulse sequence applied to probe the trion dephasing. (Bottom) Measured FWM amplitude as a function of the delay τ12, yielding coherence dynamics for different temperatures; theoretical fits as solid lines. Inset: Dephasing time T2 as a function of temperature.

To extract the ZPL dephasing rate γ = 1/T2, we measure the time-integrated FWM amplitude as a function of τ12. For a fixed τ2R, the echo moves through S(t) when varying τ12. This has previously been compensated by correcting the signal in the time domain by S(t).[19,20,24] However, for sufficiently large τ12, the echo is generated at times not accessible via S(t), such that the signal cannot be retrieved via spectral interference. Here, this issue is overcome by simultaneously increasing τ2R toward positive times when increasing τ12, such that S(t) probes the same time portion of the echo for every τ12, as depicted in Figure b. We initially set τ2R = −70 ps to ensure the time ordering between and FWM, as required to perform spectral interferometry. The resulting FWM amplitude for several temperatures is presented in Figure b.

After the echo has developed for τ12 > 150 ps, the decay of the signal is given by a single exponential, yielding the dephasing time T2. At low temperature the latter reaches T2 ≈ 1.3T1, close to the radiative limit (T2 = 2T1), in spite of the significant inhomogeneous broadening. As shown in the inset, with increasing temperature, T2 shortens rapidly consistently with previous measurements on ensembles[12] and more recent complementary approaches employing photon-correlation techniques.[27] The dominant term in the electron–phonon coupling in semiconductors is linear in the lattice displacement; that is, it is linear in the phonon creation and annihilation operators. For the present case of a QD excited at the lowest exciton transition, which represents a two-level system, this reduces to the independent boson model.[15] This model, for the 3D acoustic phonon density of states, provides a band of phonon-assisted transitions and an unbroadened ZPL. The finite width of the ZPL and its temperature dependence are explained by phonon processes, which are of second order in the phonon operators that may originate from virtual transitions to higher excitonic states[16] or phonon anharmonicities,[17] which are not included in our model.

Figure b also reveals a pronounced decrease of the initial FWM amplitude at τ12 = 0 with increasing temperature, such that the signal cannot be measured beyond T = 35 K. This decrease is attributed to the phonon-induced dephasing caused by the linear coupling to phonons, which dominates the short-time behavior of the signal. To analyze this effect, we measure the coherence dynamics on a picosecond time scale. The results are shown in Figure . The linear coupling describes the fact that the equilibrium positions of the lattice ions in the presence of an exciton are different from their values in the absence of an exciton; that is, a polaron is formed. When an exciton is abruptly generated by a femtosecond pulse, the formation of the polaron is accompanied by the emission of a phonon wave packet,[14,28] traveling through the QD volume and then through the surrounding lattice with a sound velocity of about 5 nm/ps, as illustrated in Figure a. Once the wavepacket has left the QD, the phonon-assisted transitions have dephased and are not further contributing to the FWM. In the τ12-dependence of the FWM signal this manifests itself as a fast decay on a time scale of about 2 ps,[29] clearly revealed in Figure b.

Figure 4

(a) Cartoon depicting propagation of a phonon packet from a QD after its excitation with a short, femtosecond pulse. (b) Two-pulse time-integrated FWM amplitudes for initial delays τ12. The FWM amplitudes at different temperatures (see legend) are normalized at τ12 = 0 to show the phonon-induced dephasing. (c) PL spectra for different temperatures. Solid lines: experimental data; dashed lines: theoretical curves. (d) Final FWM values after the initial decay (red cirles) as a function of temperature along with the theoretical calculation (blue line) (cf. panel b). Additional temperatures for (b) and (c) are shown in Supporting Information Figure S2. Z2 estimated from temperature-dependent PL spectra are given by gold crosses.

The final value after the initial drop FWMdrop is plotted as a function of temperature in Figure d. With increasing temperature the phonon coupling becomes more effective, and, accordingly, already at T = 30 K the coherence decays by a factor of 5 during the first 5 ps. This explains why the initial value of the signal seen in Figure b, where this initial decay is not resolved, rapidly decays with increasing temperatures.

In the spectral domain, the initial decay is associated with a broad phonon background around the ZPL.[29,30] This is seen in the PL spectra taken from the same QD, shown in Figure c. For low temperatures the background is asymmetric, reflecting the dominance of phonon emission processes over absorption processes, while at higher temperatures, when the thermal occupation of the involved phonons becomes much larger than one, the phonon background becomes symmetric.

For the theoretical modeling of the signals we employ the standard model of a QD coupled to acoustic phonons via the pure dephasing mechanism,[10,11,29−31] which has been proven to successfully describe a variety of optical phenomena in single QDs and QD ensembles. For this model exact analytical formulas for linear and nonlinear optical signals after excitation with an arbitrary series of short laser pulses can be obtained within a generating function formalism.[32,33] To be specific, we model the QD trion transition as a two-level system, which is coupled via deformation potential coupling to longitudinal acoustic (LA) phonons with a linear dispersion relation. Assuming an approximately harmonic confinement potential, we take Gaussian-shaped wave functions for electrons and holes with electron localization lengths a in the in-plane direction and a in the out-of-plane direction; that is, we model a lens-shaped QD and treat the exciton wave function as a product of electron and hole wave functions.

The results of our calculations are shown as solid lines in Figure b and as dashed lines in Figure c. We use electron localization lengths a = 8 nm and a = 1 nm (the respective hole localization lengths are scaled by 0.87) and assume GaAs parameters. We find an excellent agreement between theory and experiment for both the FWM signals and the PL spectra over the whole range of temperatures, however for an increased phonon coupling strength. Specifically, to achieve agreement with both the FWM signal and phonon background in PL simultaneously, we had to increase the phonon coupling constant by 50% using increased deformation potentials De = 10.5 eV and Dh = −5.25 eV, with respect to the parameters De = 7.0 eV and Dh = −3.5 eV, which have previously been used to quantitatively describe various optical signals from QD structures.[10,34,35] A more detailed discussion of the role of the parameters for the FWM signals and PL spectra can be found in the Supporting Information. While we do not have a definite explanation, we note that similarly higher values for the deformation potentials can also be found in the literature,[36] and an increased deformation potential has also been used to explain mobilities in a 2DEG.[37−39] One explanation could be that in the present sample there are additional mechanisms, such as piezoelectric coupling, which are usually negligible but contribute here, e.g., because of an increased spatial separation of electron and hole wave functions,[28] leading to an effective increase of the coupling described by larger deformation potentials.

Finally, it is instructive to compare the FWM initial decay with the ZPL weight in PL[12] (Z), defined as the fraction of ZPL in the total absorption spectrum. Within the formalism describing our experiment we obtain FWMdrop ∝ Z2. We approximate Z as the PL ratio between the measured ZPL and the entire PL, including the phonon background. In spite of finite spectral resolution and significant σ, for all considered temperatures we obtain close agreement between FWMdrop and Z2 independently estimated from PL, as shown in Figure d.

In summary, we have employed in situ EBL to deterministically embed QDs within microlenses, providing a convenient nanophotonic platform to perform coherent nonlinear spectroscopy of individual QDs. Microlens structures enabled efficient penetration across the dielectric boundary and tight focus of the light field around the QD, which has been exploited to perform FWM microspectroscopy. We have measured and modeled the role of acoustic phonons on the coherence of single QD excitons, in particular corroborating signatures of single phonon wave packet emission in FWM and PL. Our fundamental studies, aiming to understand the complex interplay between charges and lattice vibrations, are at the heart of condensed matter optics. They are relevant for a large class of individual emitters in solids, such as epitaxial and colloidal QDs or color centers in diamond, or emerging QD-like emitters in transition-metal dichalcogenides.[40−42] Our findings are also pertinent for ultrafast nonlinear nanophotonics, optomechanics, and phonon transport in nanostructured devices.