1. Introduction

Desirable properties for single-photon sources e.g., in quantum cryptography are in general, a high and stable internal quantum efficiency, a low multi-photon emission probability, a small emission linewidth, a high emission efficiency in a defined direction, and high repetition rates. Single-photon emission using laser excitation has not only been demonstrated at cryogenic temperatures [1] but also at elevated temperatures using quantum dots (QDs) [2], [3], [4], molecules [5], [6] and nitrogen vacancy centres [7] as emission sources.

One advantage of self-assembled QDs is that they can be easily embedded into microcavities that enable directional emission and therefore high count rates at detectors [8], [9]. Hence QDs are very attractive for high data rate quantum communication systems.

Electrical pumping is highly advantageous since this would allow for a miniaturized device that does not require the typically expensive and bulky laser sources used for optical excitation. QDs are very good candidates for electric pumping as they are part of a semiconductor crystal that can be easily engineered into a device structure using existing, well established semiconductor technology. To date, electroluminescence (EL) from single QDs, has only been demonstrated at 4-10 K [10], [11]. Higher temperature (room temperature) electrically pumped single-photon emission has only been demonstrated using single gold nanoparticles [12], however, this system is not epitaxially-grown and suffers from photobleaching and broad emission linewidths. Currently, it is therefore somewhat limited in terms of practical application. In this Letter, we demonstrate a stable, epitaxially-grown single-photon emitter operating up to 80 K, that therefore only requires liquid nitrogen cooling. This system does not suffer from blinking or photo-bleaching effects and emits in the red spectral range (670 nm), close to the detection maximum of silicon single-photon detectors and would therefore be especially suitable for free-space or polymer optical fiber based quantum cryptography applications. This result therefore represents an important step towards the development of a commercial device.

2. Sample growth and experimental setup

The sample structure was grown by metal-organic vapour phase epitaxy using standard sources (trimethylgallium, trimethylindium, trimethylaluminium, arsine, and phosphine) at low pressure (100 mbar) on (100) GaAs substrates oriented by 6° toward the [111]A direction. Si was used for the n-doping and Zn for the p-doping and both doping levels were approximately 5·10¹⁷cm^-3. A n-doped 100 nm buffer layer of GaAs was deposited first. This was followed by 80 nm of n-doped lattice matched Ga_0.51In_0.49P and 200 nm of n-doped AlInP. Then, following the deposition of 50 nm of intrinsic Al_0.55Ga_0.45InP, QDs were grown by depositing 2 monolayers of InP onto 10 nm of intrinsic Al_0.20Ga_0.80InP. The samples used for optical investigations were then capped by a further 10 nm of intrinsic Al_0.20Ga_0.80InP and 50 nm of intrinsic Al_0.55Ga_0.45InP. To complete the p-i-n-structure the remaining 200 nm of AlInP, 50 nm of GaInP and 200 nm of GaAs were p-doped (see Fig. 1).

Aluminium was added to the barrier material in order to provide a higher QD confinement potential. However, as the QD quality degrades at high aluminium contents, the barrier surrounding the QDs is intentionally limited to 20% Al. The higher aluminium containing barriers are kept further away from the QDs in order to maintain QD quality while still achieving higher confinement. Using this approach we could achieve an overall carrier confinement energy of up to approximately 300 meV.

Cr/Zn/Au/Pt/Au-stripes were evaporated on the sample surface to form the p-contact while a Ni/Ge/Au-layer deposited on the back of the sample provided the n-contact. When applying a forward bias, the current spreads out from the contact stripe as it flows towards the n-contact, therefore EL can even be observed 200 µm from the stripe. For single QD investigation, 460 nm diameter polystyrene nanospheres [13] were spin coated onto the sample and the sample was then covered by 50 nm of SiO₂. Finally 40 nm of chromium was deposited to form the nano-aperture shadow mask and the nanospheres were then removed using a lift-off process. The sample was mounted in a He-flow cryostat that can be scanned both horizontally and vertically using two stepper motors, each with an effective spatial resolution of 50 nm. The cryostat is equipped with a heater that enables temperature control from 4 K to 460 K. A current source providing 0.1 mA resolution was used to excite the sample and the luminescence was then transmitted through a 50x microscope objective that could collect the EL emission from an area on the order of a few microns when using a piezo-based actuator for fine adjustment. The EL was dispersed using a 0.75 m spectrometer and detected using either a liquid nitrogen-cooled charge-coupled-device camera when taking spectra, or two APDs, one in each path of a Hanbury-Brown and Twiss-type setup, when recording second-order autocorrelation measurements (SOAMs).

 

Fig. 1. (Color online) Structure of the sample. EL is only seen near the contacted Au-stripe.

Download Full Size | PDF

3. Low temperature electroluminescence

In Fig. 2(a)–(c) we present EL emitted at 4 K from three different QDs labelled QD A, QD B and QD C. At voltages below approximately 2.3 V, the line labelled 1 typically dominates the EL spectra with a typical linewidth of 0.2 meV. Note that this regime could not be investigated in the case of QD C as the EL intensity was too low at such voltages. Increasing the voltage to above approximately 2.5 V, and thereby also increasing the injection current, the overall EL intensity increases and an additional line (labelled 2) rapidly emerges. As the energy difference between these lines varies between 1.5 and 3 meV, and is therefore distinctly smaller than that of the typical exciton-biexciton binding energy of approx. 4-6 meV in this system [14], we conclude that lines 1 and 2 originate from a neutral and a negatively charged exciton complex, respectively. We propose that the charged exciton complex tends to dominate at high injection currents, as the electron quasi Fermi-level approaches the QD electron levels with increasing electric field. This in turn leads to an enhanced probability of electron capture, and consequently to the negatively charged exciton complexes becoming increasingly likely. A similar behaviour has also been observed from charge tunable QDs [15]. At low injection currents when a small electric field is applied, and the quasi Fermi-level is energetically far away from the QD, we assume that QD electron and hole capture have approximately the same probability and therefore that neutral exciton complexes tend to form in the dots.

As the intensity increase with applied current has an exponent of around 2 for the case of the neutral exciton complex and about 3 for the charged exciton complex, we tentatively assign peak 1 to biexciton emission and peak 2 to negatively charged biexciton (XX^-) emission. This assumption is supported by previous measurements that show that bright exciton state emission is strongly suppressed at low temperatures in such InP QDs as a result of a considerably lower-lying dark exciton state [16] and therefore that biexciton emission is enhanced. It is also reasonable that peak 2 has the larger exponent as XX^- formation requires five carriers and is also more probable with increasing injection current. Exponents of 1.1 and 2, for exciton and biexciton emission, have been reported for InGaAs electrically pumped QDs [10], [17].

 

Fig. 2. (Color online) Spectra obtained from (a) QD A, (b) QD B, (c) QD C all at 4 K. (d) Spectra obtained from QD C at 40 K, 60 K and 80 K.

Download Full Size | PDF

4. Electroluminescence at elevated temperatures

To achieve improved operation at elevated temperatures, we have used Al_xGa_1-xInP as the barrier material as this material provides a higher confinement potential than the normally used GaInP material. In Fig. 2(d) we present EL from QD C at 40 K, 60 K and 80 K. Due to the thermionic activation of carriers that are frozen out at 4 K, especially at the interfaces, the voltage required to achieve a given current decreases with temperature and therefore neutral transitions become more probable. The higher carrier mobility also implies that slightly less current is needed to observe EL emission. At 80 K almost no emission from peak 2 is observed, while peak 1 is dominant and exhibits a linewidth of 0.8 meV. Although another recombination feature emerges, labelled 3 in Fig. 2(d), this emission is thought to originate from neighbouring dots and will not be considered further. If we now compare the intensities of the EL lines as a function of temperature, we observe that the intensity of QD C at 60 K and 12 mA increases by a factor of 4 compared to that at 40 K and 12.7 mA. Even at 80 K, the intensity only decreases slightly by a factor of 1.5 at 11.9 mA.

5. Autocorrelation measurements

We have also performed SOAMs, in order to prove that the EL is emitted by a single QD. The results are shown in Fig. 3(a) for QD B at 4 K and in Fig. 3(b) for QD C at 80 K. A pronounced antibunching dip at τ=0 of 0.41 (0.43) with respect to the Poisson-normalized level is observed for QD B at 4 K (QD C at 80 K). This is below the threshold of 0.5 expected for the case of two independent single-photon emitters and therefore the emission can clearly be attributed to a single QD. Nevertheless, for an optimal source, the antibunching should approach 0 at τ=0. The observed deviation from this value is mainly caused by two effects, the limited temporal resolution of the experimental setup (instrument response function (IRF)) and unwanted background luminescence. The IRF was evaluated to be approximately 500 ps from an SOAM of the pulsed femtosecond laser source (pulse width ≈150 fs). The influence of the temporal resolution on the antibunching dip was determined by convoluting the IRF of the form C·exp(-|τ/0.5ns|) with the expected SOAM function: g ⁽²⁾(t)=1-A·exp(-|τ/τ_c|), where C is a normalization factor, A is the value of the antibunching dip and τ_c is the antibunching dip time constant. The convoluted function is then fitted to the experimental data, and the antibunching value including the background contribution g ⁽²⁾ _b (0) can be determined, where 0 represents the time for simultaneous photon detection events on both APDs. We found that the antibunching including background is reduced to g ⁽²⁾ _b (0)=0.15±0.05 (g ⁽²⁾ _b (0)=0.25±0.05) for QD B at 4K (for QD C at 80 K). This clearly demonstrates EL single-photon emission from QD C up to 80 K. The signal intensity being a statistical mixture of regulated single photons and a small background of photons with Poisson statistics was approx. 21500 cts/sec for QD B at 4K (4600 cts/sec for QD C at 80 K). This corresponds to a single photon detection rate of 19800 cts/sec for QD B at 4K (4000 cts/sec for QD C at 80 K), when taking into account the values of g ⁽²⁾ _b (0) following Ref. [18]. When considering the efficiency of the setup which is approximately 22 %, we estimate that the single photon collection rate into the first lense is 90 kHz for QD B at 4 K (18 kHz for QD C at 80 K). From the rise of the antibunching dip in SOAMs we can estimate a system recovery time of 1 ns. The optimum single photon emission rate is therefore 1 GHz. Consequently, the collection efficiency is 0.9·10^-4 for QD B at 4 K (1.8·10^-5 for QD C at 80 K). This low collection efficiency could be potentially increased by more than one order of magnitude by using a planar resonator structure [11], [9], or even higher by using pillar cavities, and by reducing the thickness of the 200 nm GaAs top layer that absorbs approximately 40 % of the light emitted by the QDs.

To investigate the influence of the background on the antibunching dip, we carefully examined the corresponding EL spectra and evaluated the background under the portion of the relevant emission lines that were sent to the avalanche photodiodes (APDs). This background contribution is shaded red in Fig. 3(c), (d). The antibunching value without the background contribution can be obtained by considering that g ⁽²⁾(0) without background is given by [19];

were the signal to noise is represented by ρ. A value for ρ=S/(S+B), of 0.94 (0.88) for QD B (QD C) was then calculated, where S is the signal and B is the background. Using this approach a value for g ⁽²⁾(0) of 0.04 (0.03) for QD B (QD C) was obtained. This implies, within the errors, that the deviation from a single-photon emission signature can be almost entirely accounted for by temporal resolution and background and clearly demonstrates that even at 80 K, these QDs are capable of providing electrically pumped single-photon emission. We would also like to point out, that achieving higher temperatures was mainly limited by background luminescence that strongly increases above 80 K and not by the signal intensity. We would like to point out that in PL measurements of a similar sample, single dot emission at a reasonable signal level could be observed up to 150 K as background signal was not a problem in that case.

The background problem should be very much reduced by truly single-dot selection, e.g. using (oxide) apertures [20], [21] to confine the current path and/or a planar resonator structure. This together with pulsed electrical injection, whose feasibility was demonstrated in Ref. [10], would then potentially enable the use of thermoelectric cooling and provide the first practical commercial device.

 

Fig. 3. (Color online) (a) SOAM of QD B at 4 K. Red line: Convolution of the IRF with the expected g⁽²⁾-function. Inset: IRF which was used for the convolution. (b) Same as (a) but for QD C at 80 K. (c) Spectra of QD B at the same temperature and current as in (a). The vertical lines represent the part of the spectrum that was sent to the APDs for the SOAM. The dashed horizontal line indicates the signal height of the background. The red area is therefore assumed to be the background contribution. (d) Same as (c), but for QD C at 80 K.

Download Full Size | PDF

6. Conclusion

In summary, we have presented single-photon EL from single QDs up to a temperature of 80 K. Single QD EL was observed and the single-photon emission character was confirmed via SOAMs. The deviation from the perfect single-photon emission signature could be almost entirely accounted for by the temporal resolution of the experimental setup and background luminescence. Consequently,we have demonstrated that epitaxial QDs are capable of providing single-photon EL emission up to at least 80 K.