Polarization study of single color centers in aluminum nitride

Color centers (CCs) are emissive point structures in a crystal, consisting of combinations of impurities, vacancies, and lattice defects, which, at a high density, can give a crystal color. When a single center is isolated, the photon statistics often demonstrate anti-bunching. CCs with internal energy levels embedded deep within the bandgap of their semiconductor host can emit at room temperature. The most investigated example is the negatively charged nitrogen vacancy (NV⁻) complex in a diamond,¹ which shows great promise as a room-temperature light source for quantum communications,² bio-compatible quantum-sensor,³ long-lived spin memory,⁴ and as a platform for tests of fundamental quantum physics.⁵ In recent years, CCs have been found in a large range of other wide-bandgap materials, including silicon carbide⁶ and hexagonal boron nitride.^7,8

Isolated CCs in the group III-nitrides have been less well investigated, with a small number of works reporting their presence in gallium nitride (GaN)^9–11 and aluminum nitride (AlN),^12–14 emitting in the visible and near infrared. The prevalence of these semiconductors in high power electronics and solid-state lighting makes this an especially promising platform to investigate, with an established route to cost-effective and commercial-scale fabrication, and epitaxial material available at low cost. AlN is a wurzite semiconductor with a direct bandgap of 6.2 eV and a high refractive index of 2.14 at 600 nm.¹⁵ Its wide transparency window, second- and third-order optical non-linearity¹⁶ have motivated the development of waveguide-coupled devices^17–19 including cavity structures.²⁰ This processing technology could be adapted to integrate CCs in photonic integrated circuits for applications in quantum technologies.

In previous publications, density functional theory has been used to predict emission from several CCs in AlN, namely, the anti-site nitrogen vacancy complex (N_AlV_N) at 712 nm,²¹ the divacancy (V_AlV_N) at 867 nm,²¹ and negative vacancy (⁠ $V N −$⁠) at 443–517 nm.²² The conditions to form these CCs are not well understood, but one report has shown that CCs in GaN are correlated with the density of threading dislocations but are not correlated spatially,²³ which may point to a common cause. Other works have pointed to the polarity of the crystal playing a role in the epitaxy of material containing CCs in GaN.²⁴ However, the small number of experimental reports on isolated CCs in AlN means that there is little evidence CCs in GaN and AlN are related, even if they share some optical characteristics and exist in similar semiconductors.

Here, we report our studies on isolated CCs in a commercially sourced AlN-on-sapphire sample that has a convenient density of CCs with zero-phonon lines clustered around to 620 nm. To learn more about the AlN CC's physical origin, we investigate the ensemble distribution of absorption dipoles. We find that, despite the crystalline nature of the film, there is no preferred absorption dipole orientation direction for the CCs. A detailed study of the temporal, spectral, and photo-physical properties of a small number of CCs provides an insight into their internal energy levels and physical structure.

The sample measured in this work is a single crystal, 1 μm thick epi-layer of AlN grown via metal-organic chemical vapor deposition on a [0001] plane sapphire substrate. In confocal scan maps, CCs appeared as diffraction limited spots, Fig. 1(a). A confocal microscope was used to excite and collect light from the CCs. The sample was excited with a wavelength of $λ exc =$ 532 nm using a DPSS laser. The excitation polarization was purified with an additional linear polarizer. A 0.9 numerical aperture objective focused the excitation laser onto the sample and collected the resultant fluorescence. The fluorescence between 550 and 650 nm was coupled into a SMF28 fiber and subsequently measured with an SPCM-AQRH silicon avalanche photo diode (APD) from Excelitas. The beam was translated laterally on the sample's surface by a mirror galvanometer and 4f imaging system, while the focal depth was controlled by a piezoelectric actuator.

Previous reports of CCs in c-plane AlN have all reported linear absorption dipoles. Motivated to determine whether these absorption dipoles are aligned to the crystal directions, we have developed an automated procedure to measure the polarizations of a large number of CCs. Conventionally, the in-plane orientation angle when maximum absorption occurs, $θ abs$⁠, for a given CC is found by continuous rotation of the laser polarization θ while monitoring the emission intensity. This results in a detected intensity variation, $I ( θ ) = a + b cos 2 ( θ − θ abs )$ for a given CC located in some initial scan area. η is the visibility of the variation in $I ( θ )$⁠. High visibility is consistent with a single linear absorption dipole. Figure 2(a) illustrates one CC with $η = 0.92$⁠. Measurements of 47 CCs are presented in Fig. 2(b), showing the majority of CCs surveyed have a high degree of polarization, indicating absorption to a single dipole. However, applying this technique to an ensemble of CCs identified in a scan area naturally pre-selects CCs aligned to the laser polarization of the initial scan. A different approach is required to uniformly sample the ensemble.

Figures 3(a)–3(c) illustrates three scan maps measured over the same area at laser excitation angles θ = 0°, 60°, and 120°, without polarization filtering in the collection path. It is evident that some CCs are preferentially excited at one laser polarization, and entirely suppressed in the other scans. To obtain the full picture of the CC locations, the three scan maps in 3(a)–3(c) are combined with the positive-root of their quadrature sum, shown in 3(d). The CC density is one CC per 11 μm².

$D abs →$ is parameterized by its magnitude $r = | D abs → |$ and its angle $θ abs$⁠.

We have verified that this method is accurate in measuring the polarization in regions of high intensity, such as where CCs are located, by comparison to subsequent laser polarization scans on individual CCs. However, in the area between CCs, the low intensity background contains uncorrelated noise in the $I 0 ° → , I 60 ° →$⁠, and $I 120 ° →$ maps resulting from counting noise in each map. We, therefore, weight each point in the angle map, Fig. 3(e), using the corresponding point in the intensity map, Fig. 3(d). The weighting function, w(x, y), takes the form $w ( x , y ) = ( I ( x , y ) − I Min ) / ( I Max − I Min )$⁠, where I(x, y) is the intensity at each location, $I Max$ the maximum and $I Min$ the minimum on the intensity map in 3(d). The resulting angle map in 3(e) shows polarization as a color, with intensity at each pixel weighted by the intensity in 3(d). White indicates a low intensity. This procedure provides an efficient and automated method of assessing every CC within an ensemble.

In our optical system, the birefringence of a dichroic beam splitter results in a small reduction in the purity of the laser polarization at some angles. The maximum visibility achievable by the microscope is 99.8%. When the effect of the birefringence was greatest, the visibility was 94.8%, which has no impact on the conclusions of our study.

With this information, we can look at the statistics of CC absorption angles over a 2025 μm² area. A histogram of weighted pixel angle, shown in Fig. 3(f), indicates that the absorption dipole angles of CCs are uniformly distributed. This is in contrast to many other solid-state emitters; for instance, the neutral exciton in InAs quantum dots usually has two dipoles along crystal axes in the plane of the sample²⁵ and the NV⁻ in diamond has four possible orientations that can be readily observed in [111] oriented samples.²⁶ We anticipate that scanning samples featuring CCs with high η and a limited number of dipole orientations would yield distinct peaks at angles corresponding to the crystal axis of the sample.

The lack of any preferred direction in the in-plane component of absorption dipoles in AlN CCs points to a more complex origin than the diamond NV⁻. In future, sampling the ensemble distribution of emission dipole orientations could yield further insight into the physical origin of these CCs, or could be applied to other material systems. Identification of all CCs in these samples, not just those aligned to the laser in a single scan, could be useful in correlative microscopy techniques where transmission electron microscopy and cathodoluminescence is compared to PL maps to investigate the link between threading dislocations in AlN and CCs.

We acknowledge financial support provided by EPSRC via Grant No. EP/T017813/1 and the European Union's H2020 Marie Curie ITN project LasIonDef (Grant No. 956387). Device processing was carried out in the cleanroom of the ERDF-funded Institute for Compound Semiconductors (ICS) at Cardiff University.