Fabrication of quantum emitters in aluminium nitride by Al-ion implantation and thermal annealing

Single-photon emitters (SPEs) within wide-bandgap materials represent an appealing platform for the development of single-photon sources operating at room temperatures. Group III- nitrides have previously been shown to host efficient SPEs which are attributed to deep energy levels within the large bandgap of the material, in a way that is similar to extensively investigated colour centres in diamond. Anti-bunched emission from defect centres within gallium nitride (GaN) and aluminium nitride (AlN) have been recently demonstrated. While such emitters are particularly interesting due to the compatibility of III-nitrides with cleanroom processes, the nature of such defects and the optimal conditions for forming them are not fully understood. Here, we investigate Al implantation on a commercial AlN epilayer through subsequent steps of thermal annealing and confocal microscopy measurements. We observe a fluence-dependent increase in the density of the emitters, resulting in creation of ensembles at the maximum implantation fluence. Annealing at 600 {\deg}C results in the optimal yield in SPEs formation at the maximum fluence, while a significant reduction in SPE density is observed at lower fluences. These findings suggest that the mechanism of vacancy formation plays a key role in the creation of the emitters, and open new perspectives in the defect engineering of SPEs in solid state.


Introduction
Single photon emitters (SPEs) in wide-bandgap semiconductors are promising building blocks for quantum technologies, including quantum sensing, optical quantum computing and quantum communication [1][2][3] .Quantum emitters within solid-state host materials have steadily gained relevance in the last two decades.Alongside single photon emission in quantum dots 4 , the experimental demonstration of anti-bunched emission from colour centres in diamond 5 led to a vibrant scientific field focused on characterization, manipulation, and fabrication of defect systems in wide-bandgap semiconductors [6][7][8] .
With the advancements in ion implantation technology regarding deterministic single-ion doping and nanoscale placement precision [9][10][11][12] , along with the substantial progresses in material synthesis and development in terms of controlled and selective chemical vapour deposition [13][14][15] , it became possible to identify and develop optically-active defects with stable and efficient single photon emission, that in several instances is correlated to their (highly coherent) spin properties.
Aluminium nitride (AlN) is a wide-bandgap semiconductor (E g = 6.03 eV) with a refractive index of ~2.15 at λ = 650 nm.It is well known and employed as a piezoelectric material, a durable ceramic, and the ideal buffer layer for GaN growth 37 , making it an appealing semiconductor for the implementation of high-power electronics and next-generation photonics.Following theoretical studies suggesting the availability of optical dopants with spin properties similar to those of the diamond-based nitrogen-vacancy centre [38][39][40] , the experimental demonstration of single-photon emission from native AlN luminescent defects was recently reported 29,35,36 .A subsequent study also showed the possibility to create luminescent extrinsic defects related via controlled laser-induced damage to the AlN crystal interface with a sapphire substrate 41 .However, the origin of these emitters is still not fully understood.
In this work, we inspect the manufacturability of single-photon emitters in AlN thin films by means of Al ion implantation and subsequent thermal annealing, in order to identify a reliable protocol to produce single-photon emitters from intrinsic point defects.We use Al ion implantation to promote the formation of lattice vacancies without introducing extrinsic defect complexes (i.e.defects related to the introduction of foreign chemical species) into the material.We investigate the correlation between the implantation fluence and the overall photoluminescence (PL) emission intensity from the implanted crystal, leaving out of the study the discrimination between the implanted quantum emitters and the previously present in the material.Furthermore, we assess the role of the subsequent thermal annealing to both activate individual luminescent centres and reduce the radiation-induced background emission.Finally, by means of single-photon confocal photoluminescence characterization, we show that vacancy-related quantum light emitters in AlN can be reliably fabricated by means of ion implantation combined with suitable thermal processing.

Experimental methods
The experiments were performed on a 1 μm thick Metal-Organic Chemical Vapour Deposition (MOCVD)-grown AlN epilayer on sapphire purchased from Dowa Electronics Materials Co.The material is a wurtzite-type single crystal and contains native optically-active defects dispersed with low density 36 .The overall procedure adopted for the fabrication and characterization of the samples is highlighted in Figure 1.
In particular, Fig. 1a shows the fabrication process performed on the substrate.The implantation of Al ions induces the formation of lattice vacancies, whose diffusion and recombination are promoted by means of thermal annealing.The resulting effect is the formation of optically-active colour centres.Specifically, the sample was implanted with 60 keV Al -ions with the recently established multi-elemental ion source of the Solid State Physics laboratories of the University of Torino.Several circular regions of ø~1 mm were implanted at different fluences in the 10 12 -10 14 cm -2 range by means of a movable collimating mask.Fig. 1b shows the ion implantation profile for the considered energy as a function of the sample depth along with the corresponding vacancy density profile, suggesting that the implantation-induced colour centres are formed within 100 nm from the surface.Multiple subsequent post-implantation annealing treatments were performed using a tubular oven at 400 °C and subsequent multiple 30 min processes at 600 °C in N 2 atmosphere.The sample was characterised by means of PL confocal microscopy between the annealing treatments, in order to assess the effects of ion implantation and annealing on the formation of optically active centres.The implanted regions were marked by means of high-power laser milling to enable the investigation of the same photoluminescent regions after each of the subsequent treatments performed on the sample.The PL analysis was performed using a custom fibre-coupled single-photon sensitive microscope equipped with a 100x dry objective (0.9 NA).Optical excitation was supplied by a 520 nm laser diode.In all measurements a set of spectral filters defined a detection spectral window in the 550-650 nm range, thus ensuring the removal of the background originating from the chromium line of sapphire 43 and an abrupt cut off in the emitter spectra at the edges of this range.Single-photon emission qualification was performed on isolated photoluminescent spots using a Hanbury-Brown & Twiss (HBT) interferometer implemented by a multimode fibre-fused 50:50 beamsplitter coupled to two independent single photon avalanche photodiodes (SPADs).The assessment of the density of created centres was performed by comparing the number of isolated spots found in the implanted regions by PL mapping 44 , with respect to a reference pristine region.Finally, spectral characterization at room temperature was performed on all of the considered regions using a Princeton Instruments PIXIS spectrometer.

Results and Discussion
Emitter formation upon ion implantation and annealing.The effects of Al ion implantation on the formation of colour centres in AlN were investigated by PL confocal microscopy mapping as a function of implantation fluence and post-implantation annealing temperature.Figure 2 shows the PL maps (20x20 µm 2 area, 2 mW excitation power) of the unimplanted region and of the regions implanted at 1x10 12 cm -2 , 1x10 13 cm -2 and 1x10 14 cm -2 fluences after each of the above-mentioned thermal treatments.The maps acquired prior to any subsequent annealing (condition referred to as "as implanted" in the following) are presented in the top row of Fig. 2a-d.The pristine region (first column) exhibits a low density of single emitters (~0.075 μm -2 ), corresponding to individual bright spots (~50 kcps emission intensity) with respect to the background originating from the surrounding region (~2 kcps).The single-photon emission from individual spots was verified by HBT interferometry on a set of ~10 emitters for all each region.The characterization of an exemplary emitter (corresponding to the emission spot circled in white and labelled as "C1" in Fig. 2a) is shown in Fig. 3.The second-order auto-correlation histogram acquired via HBT interferometry at different optical excitation powers (Fig. 3a) highlight the occurrence of non-classical emission originating from an individual source (i.e.g (2) (t=0)<0.5).The curves were fitted according to a model based on a three-level system that allows for the presence of a metastable shelving state 2,45 : Under vanishing optical excitation power, the decay rate through the shelving state of the three level system is negligible with respect to that of the radiative transition.The excited state lifetime was estimated as (6.7±0.2) ns by a linear regression of the λ 1 parameter versus the optical excitation power (Fig. 3b).The emission intensity at saturation was determined (Fig. 3c) to be (110±5) kcps at an optical power of (3.6±0.3)mW.The spectral signature of the emitter consisted of a broad band in the 550-650 nm, probably extending outside the range pass-band of the filtering optics.
The region implanted at 1x10 12 cm -2 fluence (Fig. 2b) did not exhibit any significant increase in the density of emitters (~0.08 μm -2 ) with respect to the unimplanted region.In contrast, the map acquired from the region implanted at 1x10 13 cm -2 (Fig. 2c) shows an increase in the emitter density (~0.37 μm -2 ), each displaying comparable photon count rates (i.e. 60 kcps under 1 mW excitation optical power) with respect to the pristine region.This observation is indicative of the fact that ion implantation is responsible for the formation of optically active defects 45 , as supported by the further increase in the overall PL intensity (~100 kcps) observed upon ion implantation at 1x10 14 cm -2 (Fig. 2d).In this latter case, the substantial PL increase prevents the identification of individual emitters.
Following the PL characterization of the as-implanted sample, the above-listed thermal processing steps were carried out.Figs.2d-h and Figs.2i-l report the PL maps (20x20 µm 2 area) of the unimplanted region and of the regions implanted at 1x10 12 -1x10 14 cm -2 fluences after 400 °C and 600 °C treatments, respectively.The same trend of the emitter density on the ion fluence, as highlighted prior to the thermal processing for all the irradiation conditions, was observed.A first annealing at 400 °C did not significantly modify the emitter density in the pristine sample (Fig. 2e), while it resulted in an increase (0.1-0.2 µm 2 ) in the areal density of emitters in the regions implanted at 1x10 12 (Fig. 2f) and 1x10 13 (Fig. 2g) cm -2 fluences, although the latter was accompanied by a noticeable increase in the background PL emission.Notably, none of the native centres disappeared as a result of the 400 °C process.
The region implanted at 1x10 14 cm -2 fluence (Fig. 2h) exhibited a three-fold increase in the PL intensity, indicating the formation of a large density of optically active intrinsic defects, whose overall emission prevented in this case the identification of individual emitters.The subsequent treatment at 600 °C did not alter the initial density of emitters in the pristine region (Fig. 2i) with respect to the previous treatments.Conversely, in the case of the second thermal process the 1x10 13 -1x10 14 cm -2 ion fluences (Figs.2j-l) this higher temperature process suppressed the background PL emission that was observed after the 400 °C treatment, thus resulting in a distribution of single emitters with a higher visibility and with a 0.20-0.25 µm -2 areal density, respectively.

Figure 3. Single-photon emission parameters from an individual centre in the unimplanted region prior to any thermal processing.
The emitter is circled in white and labelled as "C1" in Fig. 2a: a

) Second order auto-correlation chronograms. The missing datapoints corresponds to the backflash peaks of the detection system, which were removed to enhance the readability of the graph. b) Linear regression of the λ 1 parameter as a function of the excitation power allowing the estimation of the reported spontaneous emission lifetime value. c) Saturation curve for the emitter, after background subtraction. d) PL emission spectrum.
The nature of single-photon emitters in ion implanted and annealed AlN was investigated with the same characterization procedure adopted for the unprocessed substrate in Fig. 3. Figure 4 shows the single-photon emission properties of an individual emitter located in the region processed by Al implantation at 10 13 cm -2 fluence and 600 °C annealing (the centre is labelled as "C2" in Fig. 2k).The measurement of g (2) (t=0)<0.5 in the HBT chronograms confirmed the occurrence of single-photon emission (Fig. 4a), which was modelled with the same three-level system model adopted for the "C1" centre (see Eq. 1).The analysis resulted in an estimation of the radiative lifetime of the "C2" centre as (3.39 ± 0.2) ns.The emission intensity and excitation power at saturation parameters (Fig. 4c) were determined to be (119±9) kcps and (0.28±0.09) mW, respectively.While the emission intensity at saturation is comparable, the centre's lifetime and excitation power in saturation conditions are somewhat different with respect to the "C1" centre.This could be explained by considering the different spectral signature of the "C2" centre (Fig. 4d), which exhibits a large band with a broad emission peak around 570 nm.Such a feature is not observed for the "C1" emitter.The two emitters could therefore be attributed to different defective complexes, both being compatible with what was previously reported in the scientific literature for unimplanted AlN 36 .A further discussion on the spectral variability of the emitters investigated in this sample is reported in the following section.In order to quantify the occurrence of single emitters following each fabrication step, the respective areal density is shown in Fig. 5 for the unimplanted sample, and the regions implanted at 1x10 12 -1x10 14 cm -2 .The maximal density of emitters (0.4 µm -2 ) is found in the 1x10 13 cm -2 ion fluence prior to any annealing treatment.The 400 °C annealing results in a moderate increase in the density (up to 0.2 µm -2 ) for the 1x10 12 cm -2 fluence only, while inducing a significant background PL at higher ion fluences.Finally, the 600 °C results in a slight decrease in the density (0.15 µm -2 ) for 1x10 12 cm -2 ion fluence with respect to the previous treatment, and a significant decrease (0.2 µm -2 ) for the 1x10 13 cm -2 ion fluence with respect to the unprocessed sample.Ultimately, the 600 °C annealing with the highest implantation fluence (1x10 14 cm -2 ) results in the formation of individual emitters.We note that for several combinations of ion fluence and annealing conditions we are unable to quantify the density of individual emitters because the high levels of PL coming from the whole regions, suggesting the formation of ensembles of centres not individually resolvable.These combinations have not been included in Figure 5.

Spectral features of AlN emitters.
The spectra reported in Figs.3d, 4d are in good agreement with what has been reported in literature, but not yet attributed to any specific defective configuration [36].Figure 6 shows a statistical analysis of the spectral signature of a set of 58 PL spectra acquired from individual emitters.All the emitters are characterised by large (i.e.>20 nm) spectral bands as in the case of the exemplary emitters shown in Figs.3-4, which is centred around a peak wavelength.Figure 6a shows the statistical distribution of the central peak wavelength as a function of the annealing temperature, independently of the ion implantation fluence at which the region containing each individual emitter was processed.Conversely, Fig. 6b highlights the peak wavelength occurrence as a function of the implantation fluence, independently of the post-implantation temperature treatment.In this latter case, no centres from the region implanted at 1x10 13 cm -2 and 1x10 14 cm -2 and subsequently annealed at 400 °C could be identified due to the intense background PL.In all cases, the peak central wavelength displayed a large variability in the 560-620 nm range, with the majority of the centres emitting between 570 nm and 600 nm.No specific trend in the central wavelength distribution was observed neither as a function of the implantation fluence nor the annealing temperature, besides the lack of emission centred at 620 nm following any of the considered thermal processes.The variability in the distribution of the central peak wavelengths observed in the considered emitters might be indicative of the presence of different classes of emitters, or alternatively to local lattice effects on the electronic states of the point defects.This latter interpretation could be justified by the piezoelectric properties of AlN, as well as by local strains at the interface between the 1 µm thick AlN layer and the underlying sapphire substrate.To provide further context, a spectral analysis of the sample ensemble emission acquired from the region implanted at 1x10 14 cm-2 ion fluence and processed by 400 °C annealing.In this case, corresponding to the PL map shown in Fig. 2h, the broad PL spectra in the 560-640 nm range are reminiscent of the individual PL peaks observed at the single-photon emitter level, suggesting that the ensemble emission is a convolution of the multiple spectral components recorded in the histograms in Fig. 6a-b.This result confirms the role of ion implantation in the formation of optically-active intrinsic defects in AlN.The observation of ensemble emission with increasing intensity at the highest considered ion fluences further corroborates this interpretation since the overall PL emission increases with the density of ion-induced radiation damage in the host crystal 23 .Finally, an additional insight in the role of the thermal annealing was obtained by performing an analysis on an individual emitter located in the unimplanted region (centre "C3" in Fig. 7a), following each thermal treatment.The "C3" centre exhibited a broad emission band centred at ~590 nm, which was left unchanged by the subsequent annealing steps.This observation indicates that the isolated defect was still optically active after annealing at temperatures as high as 600 °C and that its emission spectrum was not altered by the thermal process.

Conclusions
In this paper we performed a systematic investigation on the role of ion implantation in the formation of single-photon emitting intrinsic defects in AlN.We demonstrated that the implantation of 60 keV Al -ions increases the density of individual colour centres and the background emission intensity was correlated with the ion fluence.The study of which of these optically active quantum emitters corresponded to an already present centre and which to an implanted one was out of the scope of this work but would require a specific investigation of each emitter level to separately analyse the properties of the implanted defects.The highest formation yield was achieved upon irradiation at 1x10 13 cm -2 fluence, without any subsequent thermal treatment of the sample.Higher ion fluences resulted in the formation of emitter ensembles that resulted in a limited capability to isolate single colour centres.A spectral analysis of the emitters revealed that all the point defects analysed in the ion implantation region have similar features with respect to those found in pristine AlN, thus highlighting that Al-ion implantation results in the formation of intrinsic defects.The structural and spectral stability of individual colour centres following thermal processes up to 600 °C was also demonstrated.These results show a viable pathway, based on industry-compatible ion implantation technique that in perspective can be potentially scaled to the single-ion delivery level 9,11 , for the fabrication of quantum emitters in chip-integrable AlN platform.

Figure 1 .
Figure 1.Schematic overview of the experimental work.a) Scheme of the fabrication protocol adopted in the present work.Following ion implantation, the annealing-induced thermal diffusion of vacancies and/or impurities within AlN results in the formation of stable, optically-active emitters within AlN. b) Vacancy density profile (blue) and the implanted Al ion distribution (red) associated with 60 keV Al implantation, calculated via SRIM Monte Carlo code 42 .

Figure 2 .
Figure 2. PL confocal mapping of the same processed regions (20x20 µm 2 scan area) upon different thermal annealing steps under 520 nm laser excitation and collection window between 550-650 nm.The colour scale encodes the range 0-70 kcps for all the considered PL scans, except for the region implanted at 1x10 14 cm -2 fluence and annealed at 400 °C, due to the high overall emission intensity of the latter.

Figure 4 .
Figure 4. Single-photon emission parameters from an individual centre in the region implanted at 1013 cm -2 fluence after the 600°C thermal processing.The emitter is circled in white and labelled as "C2" in Fig.2a: a) Second order correlation measurements and lifetime extracted from the power dependence of the λ 1 fitting parameter.The missing data points corresponds to the backflash peaks of the detection system, which were removed to enhance the readability of the graph.b) Linear regression of the λ 1 parameter as a function of the excitation power.c) Saturation curve of the emitter, after background subtraction.d) PL spectrum with spectral filtering to remove light outside the 550-650 nm range.

Figure 5 .
Figure 5. Boxchart representation of the areal distribution of single-photon emitters in AlN implanted at different Al-fluences following different thermal processes, namely: untreated, 400 °C annealing, and 600 °C annealing.

Figure 6 .
Figure 6.: Statistical distribution of the central wavelength of the main PL emission peaks from a set of individual emitters in AlN.a) distribution of the emitters at different annealing temperatures; b) distribution for different ion implantation fluences.c) PL spectra of the background PL emission acquired from the region implanted at the highest ion fluence (1x10 14 cm -2 ) and annealed at 400°C, corresponding to the PL map in Fig. 2h.

Figure 7 .
Figure 7. Spectral comparison of the same single emitter (labelled as "C3" in the PL maps isolated in the unimplanted region after each post-implantation process): a) pristine material; b) following 400 °C thermal annealing; c) following 600 °C thermal annealing.