Unraveling sources of emission heterogeneity in Silicon Vacancy color centers with cryo-cathodoluminescence microscopy

Significance

Color centers are promising quantum optical sources for future quantum information systems. However, it remains unknown how their atomic crystalline structure influences their optical emission. Here, we utilize a unique light-coupled cryo-electron microscope, outfitted with spectrometers for cathodoluminescence, to unveil how atomic-scale crystalline structure influences quantum optical emission. We focus on silicon vacancies in diamond, which are among the most promising color centers for solid-state quantum optical systems. We find grain boundaries created during the diamond crystal growth create distinct pockets of silicon vacancy emission. The expansion of the diamond lattice across grain boundaries modifies the emission, with tensile strain leading to higher emission energy and increased brightness.These findings help inform optimal materials structure for future quantum sources.

Abstract

Diamond color centers have proven to be versatile quantum emitters and exquisite sensors of stress, temperature, electric and magnetic fields, and biochemical processes. Among color centers, the silicon-vacancy (SiV⁻) defect exhibits high brightness, minimal phonon coupling, narrow optical linewidths, and high degrees of photon indistinguishability. Yet the creation of reliable and scalable SiV⁻-based color centers has been hampered by heterogeneous emission, theorized to originate from surface imperfections, crystal lattice strain, defect symmetry, or other lattice impurities. Here, we advance high-resolution cryo-electron microscopy combined with cathodoluminescence spectroscopy and 4D scanning transmission electron microscopy (STEM) to elucidate the structural sources of heterogeneity in SiV⁻ emission from nanodiamond with sub-nanometer-scale resolution. Our diamond nanoparticles are grown directly on TEM membranes from molecular-level seedings, representing the natural formation conditions of color centers in diamond. We show that individual subcrystallites within a single nanodiamond exhibit distinct zero-phonon line (ZPL) energies and differences in brightness that can vary by 0.1 meV in energy and over 70% in brightness. These changes are correlated with the atomic-scale lattice structure. We find that ZPL blue-shifts result from tensile strain, while ZPL red shifts are due to compressive strain. We also find that distinct crystallites host distinct densities of SiV⁻ emitters and that grain boundaries impact SiV⁻ emission significantly. Finally, we interrogate nanodiamonds as small as 40 nm in diameter and show that these diamonds exhibit no spatial change to their ZPL energy. Our work provides a foundation for atomic-scale structure-emission correlation, e.g., of single atomic defects in a range of quantum and two-dimensional materials.

The negatively charged silicon vacancy (SiV⁻) defect in diamond has proven to be a promising optical emitter, owing to its high brightness (1), near transform-limited linewidths (2, 3), and reduced electron–phonon coupling owing to its symmetric geometry and larger mass than the carbon atom (4). These properties also imbue single photons emitted from the SiV⁻s with comparatively high degrees of indistinguishability, promising to drive optical quantum technologies (5). However, the still incomplete understanding of the photo-physics and its interrelation with the diamond atomic structure hampers further progress. For one, the SiV⁻ defect has been associated with various spectrally distinct emission centers ranging from 700 to 800 nm (6). These emitters have been associated with crystal quality within the diamond, and possibly with hydrogen interstitials (7). Additionally, like most point defects in diamond, the SiV⁻ is prone to inhomogeneous broadening due to a number of factors. Stress within the diamond lattice can alter the optical properties of the SiV⁻ defect itself (8). By breaking defect $D_{3d}$ symmetry, diamond-internal strain can change the zero-phonon line (ZPL) energy as well as the splitting of ground and excited states (9). Bulk crystal quality, such as consistent $sp3$ bonding, has also proven important for the reduction of inhomogeneous broadening (10). Shallow SiV⁻ defects in particular are affected by poor crystal surface $sp3$-bonding saturation (11–13). While typically poor photon outcoupling from the optically high-index diamond host matrices can be overcome in diamond nanostructures, dimensionality reduction often exacerbates the intrinsic problem of spectral inhomogeneity (14). Together with the intrinsically low emission quantum efficiency, the trade-off between spectral stability and brightness presents a significant challenge.

In order to study how SiV⁻ defects behave when close to boundaries, crystal defects, or other potential recombination centers, it is imperative to measure their optical propertiesat their native lengthscales. Most previous studies of SiV⁻s have used optical microscopy with insufficient spatial resolution to delineate the relationships between optical emission and the diamond crystal structure. Scanning probe microscopy (SPM) techniques, such as scanning tunneling microscopy (STM) and Kelvin probe force microscopy (KPFM), have proven useful for identifying and characterizing sub-surface NV⁻ defects in the nearfield (15); however, these techniques rely on defect localization within the near-field of a scanning tip. Optical super-resolution techniques have been used to image defects in bulk diamonds. As such, charge-state depletion (CSD) microscopy (4.1 nm resolution) (16), and stimulated emission depletion (STED) microscopy (5 nm - 2 Å) (17–19) can identify defect emission with impressive spatial resolution. Additionally, scanning transmission electron microscopy (STEM) combined with cathodoluminescence (CL) spectroscopy has also granted nanometer-scale resolution in a sample’s optical properties, revealing heterogeneity in optical emission from a subwavelength volume (20–22). To date, however, no experiments have provided spectroscopic analysis concurrent with the local crystal structure (e.g., identification of crystal stacking faults, twin planes, or nondiamond phases) to explain such heterogeneity. This lack of combined optical, structural, and strain readout has precluded the assessment of structural sources of single-emitter heterogeneity. Here, we combine high-resolution cryo-electron microscopy with cathodoluminescence spectroscopy and 4D STEM to achieve sub-nanometer insight into the correlation between color-center structure, strain, and optical emission. The technique grants sub-nanoscale spatial resolution by virtue of the electron source. However, more importantly, we directly correlate STEM-CL with a multitude of structural and spectroscopic techniques including electron energy loss spectroscopy (EELS), nanobeam diffraction and 4D-STEM, and high-resolution imaging. These combined capabilities allow us to delineate the multiple factors contributing to optical heterogeneity of SiV⁻s in high-quality epitaxially grown chemical vapor deposited (CVD) nanodiamonds. Spectral analysis of spatial CL maps demonstrates that unstable emission centers associated with Si incorporation are spatially located at 2D defects such as at grain boundaries within the nanodiamond. We show that individual sub-crystallites within a single nanodiamond have distinct optical properties, including spectrally shifted ZPLs as well as differences in CL brightness. The changes between crystallites account for heterogeneity in the SiV⁻ emission. We show that a decrease in CL brightness is associated with a ZPL redshift - an effect that occurs in single-nanometer-scale spatial locations, within a single particle. We correlate this reduced brightness and redshift with a 2 to 5% lattice contraction. STEM-EELS indicated that this contraction is due to local SiV-defect implantation and changes in defect density across the nanoparticle. Finally, we interrogate nanodiamonds as small as 40 nm in diameter. In contrast to the larger nanodiamonds studied, we observe complete spatial homogeneity in the ZPL energy within the nanodiamond, and between other small diamonds measured.

Grain Boundaries Support Various SiV⁻ Related Emission Centers

Fig. 1A shows a schematic of the experimental setup. A condensed STEM beam penetrates the nanodiamond and is analyzed via multiple EM methods including Annular Dark Field (ADF) imaging, Convergent Beam Electron Diffraction (CBED), and EELS. All data are collected with a beam voltage of 80 kV, and a temperature of 100 K, for increased diamond stability. In diamond, the STEM beam excites bulk plasmons at 34eV (23), resulting in excited carriers that populate the SiV⁻ defect excited state. We analyze the resulting optical defect emission using CL spectroscopy. Our nanodiamonds are grown by CVD synthesis, providing highly crystalline domains as shown in Fig. 1B (Materials and Methods) and Fig. 1B, ii. We confirm the multi-crystallinity of our nanodiamonds through Selected Area Electrion Diffraction (SAED) analysis (Fig. 1C). We study various sizes of nanodiamonds, including large scale micron-sized multi-crystalline nanodiamonds and sub-50nm diamonds (SI Appendix, Fig. S9).

Fig. 1.

CL spectroscopy of high quality multi-crystalline nanodiamonds. (A) STEM CL Schematic (Bi and ii): TEM of typical Nanodiamond containing SiV⁻s (ii): HRTEM of diamond edge. (C, i): shows a single particle Selected Area Electron Diffraction (SAED) indicating the particle’s multi-crystallinity. (ii): We perform off-axis objective aperture dark field imaging and overlay two dark field images on the bright field TEM image.

First, we study the heterogeneity observed with SiV⁻ emission, by taking STEM-CL point spectra at different positions within a large nanodiamond as shown in Fig. 2. Each acquisition point is 10 nm², significantly better than the spatial resolution provided by traditional confocal optical imaging. Additionally, the energy resolution needed to observe changes in the SiV⁻ electronic structure (0.1 meV) is impossible to obtain with even the most advanced monochromated EELS machines, but well within the limit of our optical-4D STEM. We observe two main classes of emitters inside the same single nanodiamond. The first class of emitters has a spectrum comprised of multiple sharp peaks (pt. 1, pt. 2, and pt. 3). Previous work has shown that narrow emission bands between 700 to 800 nm have been associated with the implantation of silicon into the diamond lattice (6). Additionally, it has been shown that similar narrowband emission is associated with 2d crystal defects, such as twin planes (7). Based on these two studies, our observations indicate that these emission centers are likely SiV- centers associated with imperfect diamond lattices. These emitters are unstable and can only be detected within the first few seconds of electron beam irradiation. The emission bleaching is irreversible; even after a week at room temperature, the emitters did not reappear (SI Appendix, Fig. S1). Second, we identify the split divacancy ($D_{3d}$ symmetry) stable emission, characteristic at 738 nm (pt. 4, pt. 5, and pt. 6). CL brightness, spectral position, and acoustical phonon side bands vary for different point CL spectra, suggesting sub-diffraction limited structural heterogeneity which we interrogate later.

Fig. 2.

Unstable vs stable silicon dopant emission correlates with grain boundaries. (A) CL point spectra taken when a STEM probe at positions pt. 1 to 6, corresponding to the points in the inset BF STEM image, sale bar 200nm. (B, i) TEM image of nanodiamond in Fig. 2, ii three dark field images combined into one image, overlaid on the TEM image (C, i): 2D map of CL counts for CL wavelengths 700 to 800 nm, and (ii): for CL wavelengths 738 $\pm$ 1nm, corresponding in space to the black dashed box in (B, i), scale bars, 200 nm. Dashed lines are used as a guide to the eye to indicate roughly the underlying grain boundaries within the nanodiamond.

The CL intensity is correlated with grain boundaries. Fig. 2B shows a TEM image of the nanodiamond in Fig. 2 overlaid with three dark field images to elucidate the particle grain structure. Grain boundaries occur where two colors meet in these combined images, and white dashed lines serve as a guide to the eye. The CL maps taken in the area of the black box for the unstable and stable emitters (Fig. 2C) reveal that i) the nonstable emitters reside near or within grain boundaries and ii) pockets of intense SiV⁻ emission occur only over a spatial range of a 100 nm, and this intensity correlates with the grain boundaries within the multicrystalline particle. Further data on large nanodiamonds supporting these observations is in (SI Appendix, Fig. S2). We provide direct observation of unstable emission residing near or within grain boundaries, which has been hypothesized previously (7). The observed heterogeneity possibly indicates that silicon incorporation during the CVD growth varies and may be aided by 2D crystal defects, but future studies are needed to confirm silicon defect density.

Sub-Crystallites Exhibit Distinct Optical Properties

We next collect 3D hyperspectral CL maps at higher resolution (5.4 nm) from particles with well-defined grain structure. From the HRTEMs shown in Fig. 3B, we can see that a grain boundary runs across the entire nanoparticle. Specifically, in Fig. 3B, ii and iv, we see a striped pattern cutting the particle in half that is indicative of a microtwinned grain boundary (delineated by the white dashed box) (24, 25). Fast Fourier Transforms (FFTs) of images 1 and 2 confirm the particle contains a twin boundary (B, i and iii) (26). FFTs of the cropped red and blue square regions in images 1 and 2 produce single crystal {110} patterns, where the angle of rotation between these two patterns is measured to be 70.9 $\pm$ 0.5° (SI Appendix, Fig. S17), confirming the microtwinned boundary is a low energy $\mathrm{\Sigma }$3. The $\mathrm{\Sigma }$3 twin boundary preserves the tetrahedral units both in direction and bond length of the diamond structure, and therefore, the crystal structure remains coherent (24).

Fig. 3.

Sub-crystallites exhibit distinct optical properties at deep subwavelength spatial volumes. (A, i): Bright field TEM image with 0° alpha, 29° beta goniometer stage tilts (ii): Bright field TEM image with 0 alpha, 0 beta goniometer stage tilts, scalebars 50 nm (B, i and iii): FFTs of the cropped TEMs in (B, ii and iv), corresponding to red squares 1 and 2 shown in (A, i). (C) point spectra with the electron probe at circles in aii (corresponding in color) (D, i): 2D map of BF STEM counts (ii): 2D map of CL counts for CL wavelengths 741 $\pm$ 4 nm, (iii): 2D map of central wavelength of SiV⁻ emission, (iv): 2D map of FWHM of SiV⁻ emission. 2D maps correspond to the black dashed box in (A, ii). Scale bars 50 nm. (E, i): TEM image. (ii and iii): cropped TEM images corresponding to red boxes in (E, i). (F) point spectra take at circles in (E, i) (corresponding in color), inset is a zoom in on ZPLs of point spectra. (G, i): 2D map of BF STEM counts (ii): 2D map of CL counts for CL wavelengths 738 $\pm$ 3 nm, (iii): 2D map of central wavelength of SiV⁻ emission (iv): 2D map of FWHM of SiV⁻ emission. 2D maps correspond to the black dashed box in (E, i).

We observe discrete spectral heterogeneity in CL emission depending on which subcrystallite we probe. Fig. 3C reveals that point spectra display red-shifted SiV⁻ emission, with a 2nm wavelength shift across the boundary, concurrent with a reduction in CL intensity by a factor of two. We extract spatial maps of the total CL intensity at 741 nm, the central wavelength, and the FWHM of the ZPL from Lorentzian fits to the 3D hyperspectral maps. The results are shown in Fig. 3D, i–iv, together with the Bright Field STEM intensity. All maps cover the area of the black dashed rectangle in aii. Fig. 3D, ii confirms that the twinned boundary separates two regions of distinct optical properties with a discrete reduction in the SiV⁻ emission intensity by about 40% from the top to the bottom crystallite. Remarkably, this drop is consistent across the entire length of the boundary and occurs within 1 pixel length (5.4 nm) across the boundary. We note that the reported excited charge carrier diffusion lengths in nanodiamond CL experiments exceed 5nm (20, 27), suggesting that the microtwinned grain boundary acts as a barrier to carrier diffusion.

Beyond a drop in CL intensity, we also see a sharp, discrete change in central wavelength across the particle grain boundary (D, iii), where the top crystallite’s emission is centered around 741.3 nm, and the bottom crystallite’s emission is centered around 743.3 nm. These changes in ZPL energy are, however, not correlated with any significant change in the FWHM of the 738 nm emission, suggesting that the mechanism inducing red-shifting causes no emission linewidth broadening (Fig. 3D, iv).

Confirming that spectral inhomogeneity is caused by sub-crystallites, we include similar data from another nanodiamond (Fig. 3E, i). This diamond has two microtwinned grain boundaries instead of one. These boundaries are shown in the HRTEM images in Fig. 3E, ii and iii, where the HRTEMs correspond in space to the red squares in ei. Positioning a STEM probe at points 1,2 and 3 (denoted in ei and corresponding in color) results in three distinct, spectrally shifted emission profiles (pt. 1 at 738.8 nm, pt. 2 at 738 nm, and pt. 3 at 737.3 nm). The 3D hyperspectral mapping is shown in G, i–iv; the approximate crystal structure is indicated by the white dashed lines, which separate the nanodiamond into three separate crystallites, each spectrally shifted from the other. We again observe that a majority of CL emission is being produced by one crystallite compared to the rest of the particle. These results are consistent across multiple particles and sizes (SI Appendix, Figs. S9–S15). We conclude that the differences between individual crystallite domains are a dominant contributor to inhomogeneous broadening in SiV⁻ defect ensembles in nanodiamonds. In fact, in diamonds less than 50 nm in diameter, we see no spatial inhomogeneity of the SiV⁻ ZPL energy (SI Appendix, Fig. S9). Here, the average FWHM of SiV⁻ emission is 4 nm, and since single SiV⁻ FWHMs range from 2 to 5 nm (28), these smaller nanodiamonds may contain unperturbed emitters, indicating that small nanodiamonds could be a promising route to synthesizing indistinguishable photon sources. However, further studies, particularly single-emitter studies, are needed for confirmation.

Emitter CL Intensity Variations at the Nanoscale

We analyze the ZPL energy and brightness along linescans (Fig. 4A, i–iv) for four representative nanodiamonds. Remarkably, for most particles, the intensity and ZPL energy are linearly correlated as shown in ci-iv (additional nanodiamonds are explored in SI Appendix, Figs. S7 and S8). The consistent trend between emitter brightness and ZPL wavelength suggests that the same crystal perturbation is affecting both the energy and the brightness of the SiV⁻ defect. Note that we explore and account for sources of brightness heterogeneity in SI Appendix, Figs. S3–S6, including excitation efficiency, thickness, electron–phonon coupling, crystal orientation, local dielectric environment, defect density, nonradiative relaxation pathways, or defect charge state blinking (22, 29, 30). Our results strongly suggest that the change in emitter brightness is due to either a modification of the defect’s quantum yield (QY), or a change in defect density.

Fig. 4.

ZPL energy and brightness correlations reveal multiple findings. (A, i–iv) four TEM images of example nanodiamonds delineated by white dashed lines (B, i–iv) profile along red arrow of both the 738 $\pm$ 3 nm CL counts in red (Left axis) as well as ZPL central wavelength in black (Right axis). (C, i–iv) scatter plots of 738 $\pm$ 3 nm CL counts vs ZPL central wavelength at each pixel of the hyperspectral map pixels in the red box. (D, i–iv) profiles along the red and orange yellow arrows (corresponding in color), of the ratio of 738 $\pm$ 3 nm CL counts to that of the STEM camera intensity. Bins are created parallel to the boundary (perpendicular to the red/orange arrow), and error bars represent a SD of the bin. Scale bars 100 nm.

CL intensity (normalized by STEM intensity as a proxy for crystal thickness, see SI Appendix, Fig. S3 for thickness measurements) scans across the grain boundary suggests that the changes in CL brightness can be ascribed to local differences in defect incorporation density, or intrinsic QY of the SiVs. The corresponding data across the same boundary but toward the surface shows no significant changes, even for the pixel closest to the surface. Indeed, we see similar spatially invariant brightness in small nanodiamonds ($<$50 nm), where a step size of 2.9 nm is used between pixels, ensuring that our probe interacts with the first few nm of nanodiamond material (SI Appendix, Fig. S9). The same trends of grain-dependent brightness are seen in a majority of particles studied and can be found in SI Appendix, Figs. S7 and S8. Shallow defect incorporation and the defect’s subsequent interaction with the surface has been the focus of recent studies, where it has been shown that surface treatment can improve SiV⁻ optical properties, such as narrower linewidths and brighter emission (11). Shallow defects are important for device deployment, due to their optical addressability and enhanced coupling to nanophotonic structures (11, 31). These results could indicate near-surface SiVs exhibit unchanged CL brightness, but further studies are needed to confirm these results are not the result of a large diffusion length.

Nanoscale Strain Mapping

We employ 4D STEM analysis to deduce whether the strain states of individual sub-crystallites are responsible for the CL changes. 4D STEM maps lattice strain at nanometer length-scales by measuring changes in Bragg disk positions across multiple convergent beam electron diffraction (CBED) patterns, with strain resolutions down to 6 $\times$${10}^{-4}$ (32). In Fig. 5B, i and ii, we show the 738 $\pm$ 3 nm CL counts, and the ZPL central wavelength, respectively, corresponding in space to the dashed red box in A. As observed in previous particles, we can clearly see that the crystal domain boundary indicated by the black dashed line separates the particle’s optical properties, both in brightness and in ZPL energy.

Fig. 5.

SiV⁻ optical properties correlate with strain at the nanoscale. (A) TEM image (B) 2D hyperspectral maps taken at the red dashed box in (A), (i) Central wavelength of lorentz fit, (ii): 738 $\pm$ 3 nm summed intensity, (iii): CL point spectra taken along black line in (B, i). (C) 4D STEM data taken at goniometer stage tilts 19.4°to 4.3° (i): Virtual DF image, (ii): Twin boundary crystal classification, (iii and iv): CBED patterns from spots in i, corresponding in color. (D) 4D STEM data taken at goniometer stage tilts $-$9.8° 0.6° (i): Virtual DF STEM image, (ii): CBED pattern take at spot in (i), and (iii–vi): ${ϵ}_{\mathit{xx}}$ strain, ${ϵ}_{\mathit{yy}}$ strain, ${ϵ}_{\mathit{xy}}$ shear, rotation. (E, i–iii): ZPL wavelength vs ${ϵ}_{\mathit{xx}}$ strain, ${ϵ}_{\mathit{yy}}$ strain, and ${ϵ}_{\mathit{xy}}$ shear. (iii and iv): ZPL intensity normalized by STEM counts vs ${ϵ}_{\mathit{xx}}$ strain, ${ϵ}_{\mathit{yy}}$ strain, and ${ϵ}_{\mathit{xy}}$ shear, for CL data taken along black line in (B, i) that crosses the middle boundary of the particle.

Multiple 4D STEM datasets were taken of the entire particle; a virtual dark field STEM image is produced from this dataset and shown in (C, i), which delineates where strain can be analyzed. We orient the particle along the $110$ zone axis and can clearly observe that the central boundary is a $\mathrm{\Sigma }$3 twin. We classify each CBED pattern in the dataset to produce an image of the two grains and their boundary, shown in (C, ii), where the corresponding representative CBED patterns are shown in C, iii and iv. Because the crystals are rotated by 70.5° with respect to each other along the $\{110\}$ zone, we re-orient the particle along $\{211\}$ to perform comparative strain analysis of the entire particle. We perform strain analysis of the ${ϵ}_{\mathit{xx}}$ and ${ϵ}_{\mathit{yy}}$ strain, the ${ϵ}_{\mathit{xy}}$ shear, and $\theta$ rotation, show in Fig. 5D, iii and vi). The middle twin boundary separates the crystallite into distinctly strained regions, where the top right crystallite is expanded and rotated compared to the bottom left (Fig. 5D, iii–vi specifically).

Using 4D STEM-CL enables the correlation of the ZPL intensity and energy with the local strain of the nanodiamond, allowing us to draw two trends. First, we identify a blue shift of the ZPL and an increase in CL intensity, with a positive ${ϵ}_{\mathit{xx}}$ and ${ϵ}_{\mathit{yy}}$ strain (i.e. a lattice expansion) across the middle boundary (Fig. 5E, i and ii and iv and v). This trend across domain boundaries is consistent across multiple particles (shown in SI Appendix, Figs. S19–S21). Second, with a positive change in ${ϵ}_{\mathit{xy}}$ shear, we observe a red shifting ZPL, and a decreasing intensity of the SiV (Fig. 5E, iii and vi) and is again consistent in multiple particles (shown in SI Appendix, Fig. S19). The $\overrightarrow{x},\overrightarrow{y}$ basis is oriented along the twin boundary, as shown in Fig. 5D, vi.

Taken together, the trend we observe across domain boundaries is due to a change in the dopant incorporation rate of the different crystallites. Defect incorporation (specifically nitrogen and boron defects) into diamond can be facet-dependent (33). Intuitively, a large change in the defect density within a crystal lattice will lead to large changes in the lattice’s stress state (34). In a diamond lattice, changes in lattice stress states have been experimentally measured across grain boundaries via Raman spectroscopy of the optical phonon (35), where a grain boundary has been shown to separate regions of distinct Raman frequencies. Silicon defects in diamond expand the lattice due to Si’s larger diameter, which we confirm using molecular dynamics simulations (SI Appendix, Fig. S6) (36). Such a mechanism explains the trend observed across domain boundaries, where lattice contraction produced less intense CL, due to a smaller SiV⁻ defect density. This mechanism would also explain why we observe shifts in intensity and energy, even across a $\mathrm{\Sigma }$3 twin boundary (such as the one identified in Fig. 3, or Fig. 5. An ideal $\mathrm{\Sigma }$3 remains coherent, and therefore should induce little lattice distortion, which suggests that varying defect incorporation rates across the $\mathrm{\Sigma }$3 boundary is the most plausible source of strain in the particle.

We also observe some nanodiamonds with CL ZPL strain correlations within single crystallites of nanodiamonds, which have the opposite correlation to those found across grain boundaries (SI Appendix, Figs. S19 and S20). We attribute trends within a crystallite to result from changes of the QY of the SiV⁻ defect itself. The QY of the SiV⁻ defect is estimated to be 5% (37), indicating that a vast majority of energy is not radiated into the ZPL channel. Although the SiV⁻ QY is consistently low, the values of QY can vary between individual emitters (38), which has often been attributed to defect implantation crystal defects, bulk structural defects, or nondiamond phases and graphitic bonding (11, 39, 40). Notably, our results exclude these factors as the main contributors to CL intensity changes, as we see no decreases in CL emission approaching the grain boundaries, and no change in diamond bonding across or within crystallites (SI Appendix, Figs. S22–S24). Therefore, QY changes within a single crystal lattice are due to the local emitter atomic structure; future work combining atomic electron tomography with CL can enable precise insights into this phenomenon.

In summary, our results show that grain boundaries within nanodiamonds promote largely heterogeneous emission from defects associated with silicon dopants in diamond. Individual subcrystallites within a single nanodiamond have differing SiV⁻ optical properties with up to 2 nm spectral shifts and 70% brightness changes and are likely the largest contributor to inhomogeneous broadening of SiV⁻ ensembles. Changes found between sub-crystallites largely overshadow any changes within a single domain. We find that in a large majority of particles, ZPL energy and intensity are positively correlated and are spatially correlated with large, static strains within the diamond lattice, that permanently shift the ZPL of the emission, without altering the electron–phonon coupling or homogeneous broadening of the emission. The ZPL intensity can change by 50% or more, and we propose this change occurs due to multiple mechanisms, including the presence of defect density gradients within single nanodiamonds, and possible changes in the defect’s emission pathways. These findings elucidate the structural sources of heterogeneity of SiV⁻ optical emission, and can inform materials design and synthesis of future quantum sources and sensors.

Materials and Methods

PECVD nanodiamonds were grown directly onto SiO2 TEM grids from diamondoid seeding solutions. Diamondoids were isolated and purified from petroleum extracts using previously reported protocols (41). Mass spectrometry, gas chromatography, and NMR spectroscopy were used to ensure diamondoid purity. Diamondoid molecules were then selectively functionalized with POCl2 via multistep synthesis (42) to enable covalent attachment of diamondoid self-assembled monolayers on SiO2 surfaces. PECVD growths were carried out using a Seki PECVD system with a 2.45-GHz direct plasma microwave source. A piece of silicon wafer was put into the chamber during growth to act as the source of silicon for SiV centers. CL data were taken with an FEI Titan aberration-corrected environmental transmission electron microscope (TEM), using the Gatan Vulcan cathodoluminescence in-situ holder, at 80 kV of accelerating voltage. All CL datasets were taken at a temperature of 100 K. The beam voltage and sample temperature were chosen to improve the diamond lattice stability; at 300 keV and room temperature, the diamond lattice would degrade. A majority of the 3D hyperspectral CL maps were taken with 100 Pa of current, and 5 s of dwell time (see SI Appendix Info for exact parameters), with a 14 mrad convergence angle. Dark field images were taken at 80 kV, with both on axis and off axis objective aperture configurations. 4D STEM datasets were acquired with Gatan’s Oneview camera at 300 kV and room temperature with a current of 30 Pa, a camera length of 480 mm, and a convergence angle of 1.8 mrad. Data analysis was performed in Python, utilizing multiple common packages, such as numpy, scipy, and matplotlib. Non-negative matrix factorization and general data loading and viewing was done with hyperspy (43). 4D STEM analysis, including virtual dark field images as well as local strain mapping was performed with py4dstem (44).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.