Diamond processing for quantum defects

1 Introduction

The interaction between solid-state spin qubits and their host material defines a complex set of synthesis, processing, and fabrication challenges that directly limit yield, stability, and performance (Rodgers et al., 2021; Wolfowicz et al., 2021). A wide variety of materials-related mechanisms degrade qubit performance: nearby surfaces and interfaces, crystallographic defects and strain, background nuclear spins, fluctuating charge traps, and electrostatically active impurities. These mechanisms shorten coherence times, shift optical transitions, or destabilize the desired charge states. Understanding and controlling these mechanisms is therefore a prerequisite for realizing reproducible, high-fidelity quantum devices (Li et al, 2024b).

Diamond is an exceptional material platform, with extreme hardness, very high thermal conductivity, a large bandgap, and high dielectric strength, making it an excellent host for atom-like color centers such as the negatively charged nitrogen-vacancy (NV −) center and group-IV-vacancy (G4V) complexes (e.g., SiV, GeV, SnV, etc.). All of these compounds are generally referred to as XV centers. Producing high-performance, ‘quantum-grade’ diamond places stringent demands on synthesis and processing: deterministic localization of single defects at the atomic scale, control of the local electrostatic and spin environment, and preservation of crystal quality through device fabrication steps. These requirements are compounded by the fact that the highest-quality material is obtained by homoepitaxial growth, which limits integration options.

In isotopically engineered bulk diamond (i.e., reduced ¹³C concentration), single NV center coherence times have been shown to extend to the millisecond range at room temperature, with a landmark report of T 2 ≈1.8 ms in optimized chemical vapor deposited (CVD) material (Balasubramanian et al., 2009). By contrast, NV centers placed within a few nanometers of the diamond surface—typical for many nanoscale sensing applications—routinely suffer substantial coherence degradation. Hahn-echo T 2 times are often reduced into the tens of microseconds range for very shallow centers (depth ≤20 nm), and surface-related electric-field and magnetic noise can dominate decoherence (Sangtawesin et al., 2019). Delta-doped, isotopically purified layers demonstrate that careful growth can preserve long coherence at modest depths, with reported T 2 ≈1.7 ms and T 2 * ≈90 µs for NV − placed within a 100 nm deep isotopically purified CVD layer (Ishikawa et al., 2012; Ohno et al., 2012), showing that depth, isotopic purity, and growth conditions strongly influence spin coherence. Optical coherence (i.e., zero-phonon-line spectral (ZPL) diffusion) is also highly dependent on defect depth and processing. Spectral-diffusion-limited ZPL linewidths of order 1.2 (5) GHz have been reported for certain near-surface growth-incorporated NV centers (Ishikawa et al., 2012), while more recent work has produced near-lifetime-limited optical lines for deeper or well-processed centers (McCullian et al., 2022; Ruf et al., 2019). For G4V centers placed in isotropically grown diamond, careful engineering via strain tuning has pushed T 2 coherence times of SnV centers to an impressive ≈223 µs with near lifetime-limited linewidths (48 MHz) at 4 K (Guo et al., 2023), signaling the vast potential of these G4V defects in quantum applications.

Different quantum applications impose different, and sometimes conflicting, materials requirements. Quantum communication and spin-photon interfacing prioritize isolated emitters with narrow, stable optical lines and reproducible orientation. These requirements favor buried, low-strain centers that couple efficiently to photonic cavities and waveguides. Quantum sensing often prioritizes near-surface ensembles with high defect density and uniform properties to maximize signal, while still requiring long coherence times, stable charge states, and for most applications, operation at ambient temperatures. Scalable qubit architectures demand deterministic placement and high activation yield of individual spin centers at length scales where relevant coupling is strong. In every case, the relevant performance metrics—localization accuracy, charge stability, optical linewidth, and spin coherence—are set by the details of materials synthesis and subsequent processing.

The last decade has seen rapid progress in addressing these materials challenges. Advances in controlled growth synthesis (e.g., isotope engineering and delta-doping) (Ohno et al., 2012; McLellan et al., 2016), low-damage vacancy engineering (e.g., electron/He implantation, laser writing) (Addhya et al., 2024), tailored annealing sequences, and integration with photonic structures (Ding et al., 2024) have substantially improved the optical and spin properties of engineered color centers. At the same time, many groups have developed operational best practices—dedicated gas lines, chamber seasonings, strict handling and contamination protocols—that materially improve reproducibility of ‘quantum-grade’ results. Our aim in this manuscript is to coalesce these developments into a mechanistic, transferable narrative and emphasize the physical reasons behind process choices so that best practices are adoptable across laboratories and device architectures.

This paper is organized around four main themes that map naturally onto the fabrication lifecycle for diamond-based spin qubits (Figure 1). Section 2 addresses pre-synthesis processing, cleaning, and contamination control. Section 3 discusses growth synthesis, with emphasis on plasma enhanced CVD (PECVD) dynamics, isotope and dopant control, and delta-doping strategies. Section 4 treats post-growth processing, including implantation, vacancy engineering, annealing, and overgrowth/capping strategies that determine depth control and defect activation. Finally, Section 5 focuses on band engineering and charge-state stabilization approaches that improve optical stability and spin coherence for sensing and photonic applications. Throughout, we highlight quantitative metrics where available, explain the underlying thermodynamic and kinetic drivers, and point out open problems that must be solved to scale diamond qubit technologies.

2 Pre-synthesis and substrate processing

Successful and deterministic synthesis of single-crystal diamond (SCD) and associated defects requires a well-controlled initial surface state - ideally an atomically-flat, miscut-tuned, and beneficially chemically terminated sp 3 surface. As such, it is critical to understand what exact processing the diamond surface is subjected to and the associated effects on both the surface, bulk crystal, spin-defects, or eventual devices (Rodgers et al., 2024; Schuelke and Grotjohn, 2013; Hicks et al., 2019; Chou et al., 2023).

2.1 Surface polishing

2.1.1 Chemical and/or mechanical polishing

Mechanical polishing (MP), achieved with a fine diamond grit suspended in solution, removes material through a combination of micro-cleavage along crystallographic planes, direct mechanical abrasion, and thermal wear (Tyler et al., 2025; Ou et al., 2025; Pastewka et al., 2011). This effectiveness is limited by surface roughness, morphological anisotropy and the creation of a subsurface damage (SSD) layer that is ultimately detrimental to subsequent material fabrication/processing steps. To overcome these limitations, chemical-mechanical polishing (CMP) is often used as a second surface processing step for applications that demand ultra-smooth surfaces, leveraging chemical action followed by gentle abrasion to achieve extremely low roughness (≤10 nm R q) and low SSD (Doronin et al., 2018, Zong et al., 2016, Yuan et al., 2024). One common approach uses strong oxidants like potassium permanganate (KMnO 4) (Cheng et al., 2005; Chen et al., 2015) to convert the diamond’s hard sp 3 carbon surface into a softer oxidized layer (Xiong et al., 2024), which is then removed due to higher chemical reactivity (Wang et al., 2024b). A variant, the Fenton reaction, employs hydroxyl radicals (OH) generated from H 2 O 2 and an iron catalyst for precise, atomic-scale material removal (Wang et al., 2022; Wang et al., 2024b; Xiong et al., 2024). Other techniques, inspired by the silicon industry, use an alkaline colloidal silica slurry where nanoparticles bind to the diamond surface and are then sheared off, pulling away single carbon atoms (Thomas et al., 2014; Tyler et al., 2025). A typical post-CMP surface map, using atomic force microscopy (AFM), of the <001> surface is shown in Figure 2A. Very clear morphological signs of processing are still present at the substrate surface, i.e., grid polishing streaks, which can vary in presentation based on the exact slurry/grit mixture and rotation/movement settings. Adoption of CMP processes has been shown to significantly extend spin coherence times of near-surface NV centers (Li et al., 2024a; Li et al., 2025; Kubota et al., 2015; Tyler et al., 2025; Yuan et al., 2024).

2.1.2 Sputtering and etching of surfaces

Complementary to MP/CMP techniques, several non-contact ion and dry-chemical techniques, such as ion beam polishing (IBP) (Zhou et al., 2010, Dai et al., 2008), plasma etching (Hicks et al., 2019; Toros et al., 2020; Vivensang et al., 1996), and laser polishing (LP) (Malshe et al., 1999; Moshkani et al., 2025; Mildren et al., 2011), have been leveraged to achieve surface processing and nano-/micro-scale etching. IBP employs energetic ions (e.g., Ar + or O −) to sputter carbon atoms (Hirata et al., 1992), with risks of ion induced defects (i.e., point defects, amorphization, or graphitization) (Pastewka et al., 2011). LP removes material via laser-induced thermal ablation and/or oxidation, offering very high throughput and nanoscale control often leaving undesired graphitic residuum that can introduce thermal stress and cracking, requiring further processing (Moshkani et al., 2025). Reacting ion etching (RIE) combines physical sputtering with chemical reactions in O 2 or H 2 plasmas, typically achieving higher removal rates and serving as a workhorse for SSD mitigation, albeit with tight process-window requirements to avoid roughening, micro-masking, anisotropy, or contamination (Li et al., 2024a).

2.2 Damage removal and strain release protocols

The nature and effect of SSD and surface morphology has been extensively investigated (Rodgers et al., 2024; Sangtawesin et al., 2019) and comprehensive processing guidelines have been developed, most critically the idea of SSD removal via Inductively Coupled Plasma Reactive Ion Etching (ICP-RIE) methods (Hicks et al., 2019; Ruf et al., 2019; Rose et al., 2018; Sangtawesin et al., 2019; Tyler et al., 2025). Broadly speaking, this approach allows for a controlled etch of the diamond surface, isotropically removing a predetermined thickness (i.e., typically, etch depth should be at least ∼3–5 times the smallest polishing particle diameter/size) of the compromised material, leaving behind only damage-free underlying single crystal.

The choice of plasma chemistry is key in minimizing any undesired side-effects. Argon-only physical etches (i.e., sputtering or ion milling) are effective, but often result in micromasking due to the physical nature of the etch process (Hicks et al., 2019). Chlorine-based processes may be deleterious to the properties of near-surface NV centers (Tyler et al., 2025), but chlorine is critical in minimizing micro masking when performing isotropic etches (Hicks et al., 2019). Finally, oxygen-based plasmas are commonly used and result in an effective removal of diamond (Luo et al., 2021). However, they are anisotropic in their nature, and if left unchecked will result in a roughened surface. Given that ICP-RIE is a corrective step that prioritizes surface quality over efficiency (Chu et al., 2014), a combination of all three processes (of Ar / Cl 2 and O 2 plasmas), intermixed with regular pump/purge steps, have demonstrated best reproducibility and effectiveness (Appel et al., 2016; Chu et al., 2014; Heupel et al., 2020.; Hicks et al., 2019; Lee et al., 2008; Maletinsky et al., 2012; Sangtawesin et al., 2019). These processes have yielded root-mean-square (RMS) roughness values of 0.30 (15) nm (around the lattice constant of diamond, i.e. 0.357 nm) over large areas (400 μm²) as exemplified in Figure 2B, while minimizing undesired contamination and their associated effect on NV center T 2 coherence lifetimes (Heupel et al., 2020; Ruf et al., 2019; Sangtawesin et al., 2019; Tyler et al., 2025).

Recent efforts have focused on self-limiting approaches that promise atomic layer etching (ALE) capabilities. This is achieved by employing either self-limiting chemistry or carefully tuned biasing to remove only surface adsorbent-weakened surface layers (Michaels et al., 2023). These techniques promise unparalleled capabilities in terms of precise depth etching and surface smoothening (Gerritsen et al., 2022) with non contact methods while paving the way for in-situ (i.e., during processing) tuned surface termination states.

2.3 Thermal processing

Crystal strain fields are introduced during post-growth processing steps such as laser cutting, grinding/polishing, and ion implantation. Like SSD, strain inhomogeneity is highly detrimental to the properties of spin qubits, often resulting in broadened and shifted optical and spin transitions, degraded coherence, and even quenched fluorescence. Therefore, diamond materials often not only need to be smooth and damage-free but also possess a homogeneous, low-strain crystal lattice. High-temperature annealing is a primary method used to address these challenges (Meng et al., 2008).

High temperature (HT) or high-pressure, high-temperature (HPHT) annealing relieves residual stress in diamond via thermally activated plastic deformation (i.e., dislocation slip). However, diamond exhibits strong resistance to this mechanism. Plastic relaxation is unlikely at ≤850 °C and remains difficult even near 1,100 °C (diamond hardness ≈22 GPa and yield strength ≈7 GPa) (Michler et al., 1999). As such, relaxation by dislocation mobilization is only expected at very high temperatures sufficient to overcome this exceptional mechanical strength (Masuya et al., 2017; Chen et al., 2024). Interestingly, this process can also reduce biaxial compressive stress. Experiments show that a 20 µm-thick diamond layer under compression exhibits a decrease in residual stress upon stepwise annealing, consistent with the reported plastic-deformation threshold of 900 °C–1,100 °C (Schreck and Peter, 2025). As such, HPHT anneals at 1,800 °C–2,500 °C at pressures of ≥5 GPa are typically performed to help reduce strain while avoiding graphitization (Chen et al., 2024; Meng et al., 2008).

A more detailed explanation of varying thermally induced atomic dynamics that are associated with such treatments are presented further in Section 4.

2.4 Wet chemical processing, i.e., cleaning

Although often overlooked, a fundamental understanding of both basic and more involved cleaning processes is key. These approaches are varied, specific, and purposeful, but all aim to minimize undesired contamination, either in preparation for and/or cleaning up after specific processing steps.

2.4.1 Pre and post-CMP cleaning

Drawing inspiration from the microelectronics industry, a widely reported protocol shown to be effective for pre-polish preparation and minimizing CMP residues consists of a Standard Clean-1 (RCA1) followed by immediate ultrasonic DI rinse (Hsieh et al., 2023). RCA1, a H 2 O 2:NH 4 OH:H 2 O (1:1:5 ratio, with 30% strength peroxide) mixture removes particles, metallic impurities, and improves wettability while the ultrasonic de-ionized water (DI) rinse is employed to dislodge remaining particulates. Post-CMP, RCA1 is frequently paired with hydrofluoric acid (HF) to dissolve silica-based slurry remnants and further reduce organic contamination, yielding debris-free diamond surfaces (Thomas et al., 2014). RCA2 can also be added to remove organic impurities, for this, samples are immersed for 10 min in a 70 °C solution consisting of HCl:H 2 O 2:H 2 O (1:1:6) and then liberally rinsed in DI water.

2.4.2 Post- etch and anneal cleaning

Surface defects and contaminants on single crystal diamond are known to reduce photonic and electronic performance, and quench near-surface NV coherence. Wet acid oxidation serves the dual purpose of removing residues via thorough oxidation and imposing an oxygen-rich termination (i.e., – OH, C = O, – COOH) (Sangtawesin et al., 2019). Typically, a ‘tri-acid’ mixture of H 2 SO 4:HClO 4:HNO 3 at 1:1:1 ratios at highest stable molalities is refluxed at temperatures surpassing the nitric acid’s (sometimes even perchloric acid’s) boiling point ≈120 °C to serve as an aggressive sp 2/amorphous-carbon stripper (Codreanu et al., 2025; Jean Brown et al., 2019). Under these conditions, powerful oxidizing radicals such as NO 2 + and ClO 4 −, from HNO 3 and HCLO 4, respectively are able to break C = C bonds in graphitic and amorphous carbon, effectively oxidizing the carbon to CO 2 and related oxides, leaving nitrates and perchlorates in solution. Usage of a non-perchloric mixture (H 2 SO 4:HNO 3) is also possible for surface removal of most metal contaminants (that result from the various processing steps or undesired contact), while ensuring minimal sp 2 over-etching. In parallel, ‘piranha’ mixtures HSO 4:H 2 O 2 can also be used for cleaning and organic carbon removal purposes, leaving the final surface state of diamond mostly hydroxylated (C – OH) (Lo et al., 2025) with minimal effects of near surface defect coherence times (Tyler et al., 2025).

2.5 In-situ (in growth reactor) substrate processing

2.5.1 Interfacial contaminant degassing

Due to the multitude of processing the diamond surface undergoes before homoepitaxial overgrowth (including being exposed to a nitrogen rich atmospheric environment), an undesired layer of surface and sub-surface adsorbed nitrogen rich region can inherently form. This, if unchecked, can get effectively doped into the material via subsequent overgrowth. Figure 3 shows a secondary ion mass spectroscopy (SIMS) spectra highlighting such a scenario. We note two nitrogen peaks: i. a desired ¹⁵N doping peak around the mid point of the isotopically enriched homoepitaxial diamond overgrowth of 101 (2) nm (determined via ¹²C:¹³C SIMS spectra), and ii. the aforementioned undesired ¹⁴N peak at the substrate interface that has been doped into the material. To remedy this, as well as to desorb any other volatile contaminants and oxygen bearing species, prior to plasma ignition, the reactor is evacuated to high vacuum ≈2×10^–6 Torr, and the substrate is then heated to temperatures ≥500 °C for a few hours.

2.5.2 Hydrogen plasma etching

A brief in-situ hydrogen plasma etching removes near-surface non-diamond carbon (Hu et al., 2023). Atomic hydrogen, generated by H 2 dissociation in the plasma, passivates surface dangling bonds on sp 3 carbon, yielding an H-terminated surface that resists graphitization, while also enhancing chemical removal of sp 2 carbon through volatile species formation (Battaile et al., 1999; Jiang et al., 2019). In plasma enhanced CVD (PECVD), short H 2 plasma exposures typically preserve or improve morphology and can enhance near-surface charge transport via the negative electron affinity effect (Ristein et al., 2006). The effective etch rate decreases with time consistent with rapid removal of a defect-rich top layer converging to a steady-state rate for bulk diamond often ≤100 nm h⁻¹ under multi kW plasma powers (Ivanov et al., 2015).

2.5.3 Mixed plasma etching

Studies have shown that introducing oxygen into the H 2 plasma increases the availability of atomic O and OH radicals, which more efficiently remove sp 2 carbon and defective sp 3 sites by forming volatile CO and CO 2, often at rates comparable to diamond growth at set conditions (Naamoun et al., 2012; Boussadi et al., 2018; Tallaire et al., 2004; Naamoun et al., 2015). In practice, while significantly higher than H 2 only etching, the instantaneous etch rate in H 2 O 2 plasmas is not constant but decreases in a quasi-exponential manner with etching time, where the initially high rate is associated with the rapid removal of any defective, damaged, or otherwise less-than-pristene sp 3 surface layers (Naamoun et al., 2013) due to the aggressive and anisotropic nature of oxygen-based etching. As this damaged/rough subsurface layer is progressively removed, the density and size of etch figures are markedly reduced and the etch front reaches material of higher crystalline quality, where the process eventually enters a steady-state regime governed by the intrinsic etch rate of high-quality diamond. Generally, the mixed plasma etch rate rises with O 2 fraction, however this enhanced reactivity comes at the cost of increased global and local roughness ( R q ) as a result of the formation of etch pits, frequently rectangular or pyramidal due to anisotropic etching of defects and differing crystallographic faces (Langer et al., 2021). Consequently, H 2 O 2 mixtures must be used judiciously and with an optimized etch duration, gas flow ratio, and plasma powers such that any desired damaged or contamination layer is removed without excessively thinning the substrate, degrading the final epitaxial and surface finish quality, or operating in unstable gas flow and plasma regimes, i.e., O 2:H 2 flow >5%.

3 Quantum-grade diamond and spin-defect synthesis

3.1 Diamond growth

Microwave plasma enhanced chemical vapor deposition (PECVD) remains the method of choice for high-quality, homoepitaxial diamond growth since it offers precise tuning of the material: isotopic control through precursor selection (e.g., ¹²C enrichment), unwanted impurity reduction via ultra-high vacuum practices and in-line purification of source gases, control of growth morphology by tuning in-situ parameters, and deterministic dopant placement through delta-doping (i.e., time-gating of precursor flows) (Ashfold et al., 2020; Gicquel et al., 2001; Ohno et al., 2012; Schwander and Partes, 2011). To ensure an ideal lattice environment around color centers and maintain long spin-coherence times, crystallographic defects during growth must be mitigated via carefully considered substrate preparation and pre-growth processing as discussed in Section 2.5, while ensuring a well-controlled growth regime (Achard et al., 2026; Achard et al., 2005; Fischer et al., 2023; Tokuda et al., 2012).

PECVD diamond growth is driven by a hydrogen-rich, low-pressure plasma that dissociates H 2 into atomic hydrogen and activates a carbon feedstock (typically CH 4) into reactive CH x radicals; the CH x species supply lattice carbon while atomic hydrogen stabilizes the sp 3 surface and preferentially etches non-diamond (sp 2) carbon, biasing net incorporation toward tetrahedral bonding (Ashfold et al., 2020; Schwander and Partes, 2011; Silva et al., 2009). High-quality material commonly grows in a step-flow mode on substrates with sufficient atomic step density (Figure 2C): in the Burton–Cabrera–Frank picture adatoms diffuse across terraces and attach at step edges so growth proceeds by lateral step propagation rather than by repeated two-dimensional nucleation (Burton et al., 1951; Tokuda et al., 2012). Both ex-situ substrate miscut and in-situ growth parameters set the terrace thermodynamics that determine whether arriving carbon atoms reach step edges or nucleate islands. Too small a miscut (low step density) promotes island nucleation and pyramidal hillocks (Figure 4b), while excessive miscut causes step-bunching and roughness. In parallel, analogous islanding results from high CH 4 (adatom clustering) (Ahmad Reshi et al., 2019; Tokuda et al., 2012), low temperature (reduced adatom mobility) (Achard et al., 2005), or high chamber pressure (increased adatom density) (Silva et al., 2009). Although optimal windows vary, monocrystalline step-flow is typically achieved at low CH 4:H 2 ratios (0.005% to 0.025%) with temperature and pressure tuned to the specific reactor and substrate (Achard et al., 2026; Achard et al., 2005; Tokuda et al., 2012). Note that commercial substrates (Figure 2C) often specify broad miscut tolerances (e.g., ±3◦), wider than the narrow windows recommended by growth studies (Meynell et al., 2020; Six Element, 2026).

Finally, the crystallographic orientation of the growth surface strongly influences kinetics and final morphology. The (001) surface (square symmetry) commonly used for quantum-grade films supports four-fold symmetric step propagation along <110> directions (Achard et al., 2026). The (111) surface, by contrast, is more densely packed and favors triangular terraces and can be grown with a more uniform, layer-by-layer character (Achard et al., 2026; Tokuda et al., 2012). Because step structure and terrace kinetics modulate how dopants (e.g., substitutional nitrogen) incorporate, step-flow on miscut (001) can reduce random nucleation sites and thereby limit inhomogeneous dopant trapping, while (111) growth can give different incorporation energetics at kink/step sites. Practically, this often results in 111 faces presenting a significantly higher nitrogen incorporation rate than 001 (Ishiwata et al., 2017). However, this effect can be leveraged for NV − centers, as (111)-oriented films can preferentially align NV axes perpendicular to the surface (uniform orientation along <111>) (Ishiwata et al., 2017; Lesik et al., 2014; Michl et al., 2014; Osterkamp et al., 2019; Ozawa et al., 2017; Tahara et al., 2015), leading to direct advantages in ensemble-based sensing and spin-addressability (Doherty et al., 2013) while minimizing post synthesis processing.

3.2 Spin-defect synthesis

To incorporate color centers of interest such as NV and G4V (e.g., SiV, GeV, SnV, etc.) for quantum applications, two dominant approaches are widely-used and discussed in this section: ion implantation (Bradac et al., 2019; Lagomarsino et al., 2021; Meijer et al., 2005; Smith et al., 2019) and in-situ doping during growth with subsequent activation processing (Ohno et al., 2012; Ohno et al., 2014; Sedov et al., 2022; Ralchenko et al., 2019; Sedov et al., 2018).

3.2.1 Ion implantation

Ion implantation is a post-growth technique in which atoms or molecules are accelerated, electromagnetically focused, and implanted into the diamond lattice at targeted sites. The technique’s flexibility makes implantation attractive for deterministic placement of a variety of dopants.

The physical picture of ion implantation is governed by the energetics of an ion losing its energy as it traverses the diamond lattice. These processes are routinely modeled with Monte Carlo codes (e.g., SRIM/TRIM (Ziegler et al., 2010)), which predict implantation range, longitudinal and lateral straggle, and damage density as functions of ion species, energy, angle, and target density (Smith et al., 2019; Orwa et al., 2000). At high kinetic energy (i.e., ≥100 keV), the ion primarily undergoes electronic stopping (ionizing the target and exciting electrons). As it slows, nuclear stopping (elastic collisions with carbon nuclei) becomes dominant. Nuclear collisions displace carbon atoms from their lattice sites, producing Frenkel pairs (vacancy and interstitial) and because the nuclear scattering cross-section grows as the ion loses energy, the majority of atomic displacements occur near the end of the ion range (the Bragg peak). Low-energy implants in the keV range place this dopant Bragg peak at the near-surface region (tens of nm) with small straggle, while higher-energy implants penetrate much deeper but with proportionally larger depth spread. When ion beams align with major crystal axes, channeling allows ions to penetrate deeper and broadens implantation profiles; it is mitigated by tilting the substrate, using molecular ions, or pre-amorphizing the surface (Toyli et al., 2010; Ito et al., 2017). Because implantation generates vacancies through collisional cascades, careful control of ion energy and fluence (typically ≤1×10¹⁴ cm⁻²) is essential to produce the desired dopant–vacancy complexes without amorphization or spin-degrading damage (Meijer et al., 2005; Lagomarsino et al., 2021). This resulting non-equilibrium population of vacancies and interstitials is distributed along the ion tracks. Careful annealing is required to pair vacancies with dopants and to repair lattice damage and is discussed in detail in Section 4.

Lateral resolution is ensured via a variety of techniques. Focused ion beam (FIB) tools, for example, can provide beam diameters down to 20 nm–100 nm (Delegan et al., 2023) and can be operated with single-ion counting for deterministic implantation into lithographically pre-fabricated structures (Adshead et al., 2025). Masked (broad-beam) implantation uses lithographic apertures to pattern large arrays but offers less flexibility for single-ion control (Spinicelli et al., 2011; Toyli et al., 2010; Scarabelli et al., 2016). For ultimate positional precision, AFM-tip collimation approaches have been demonstrated in which ions are funneled through a nanoscale aperture at the AFM tip; although serial and slow, these methods can achieve placement precision at the few-to tens-of-nanometers scale (Pezzagna et al., 2010; Smith et al., 2019).

3.2.2 In-situ doping

To avoid the issues that arise due to lattice damage introduced by ion implantation, in-situ doping during CVD growth offers a powerful alternative to produce depth-confined color centers. During growth, a dopant precursor is introduced transiently into the plasma environment so that incorporation occurs into a narrow epitaxial layer (nanometer-scale resolution). Near-surface centers created in this method have been demonstrated to have superior optical (Sipahigil et al., 2014) and spin properties (Ohno et al., 2012), and by combining delta-doping with post-growth vacancy creation techniques such as ion/electron irradiation or local laser writing, it is possible to achieve three-dimensional localization, building on in-situ doping (Ohno et al., 2014; McLellan et al., 2016; Chen et al., 2017; Addhya et al., 2024; Myers et al., 2014).

N 2 is the precursor of choice for nitrogen delta-doping, but its high dissociation energy produces low incorporation efficiency under typical MPCVD conditions. Alternatives such as N 2 O and NH 3 increase the effective nitrogen supply but carry risks: the extra oxygen in N 2 O can oxidize or etch the surface and induce microstructural changes if not tightly controlled (Su et al., 2012; Ashfold et al., 2020), while NH 3 perturbs the hydrogen chemistry of the plasma and may produce non-ideal growth morphologies (Truscott et al., 2016; Martyanov et al., 2025). For group-IV delta-doping (SiV, GeV, SnV, etc.) commonly silane (SiH 4) or organometallic/tin-halide precursors (e.g., SnCl 4) are used. Metal-halide precursors (for example, SnCl 4 or Pb-halides) introduce corrosion-capable by-products (e.g., HCl, halide residues) and can lead to metal/halide deposition on chamber walls or components. Such residues increase the risk of unintended incorporation into the film or tool, which may seed unwanted color-centers or paramagnetic defects and degrade spin coherence. Rigorous chamber maintenance, dedicated delivery systems, and thorough residual-gas monitoring are therefore essential when using such compounds (Bradac et al., 2019; Ashfold et al., 2020; Guo et al., 2025).

After growth, converting these substitutional dopant atoms into optically-active spin qubits typically requires further processing and co-localization of vacancies. As CVD growth generates relatively few mobile vacancies at the temperatures used for homoepitaxy, most recipes rely on a secondary vacancy-creation step. Electron irradiation (bulk (Ohno et al., 2012), or TEM (Mclellan et al., 2016)), light-ion implantation (C, He) (Ohno et al., 2014), or ultrafast laser writing (Klink et al., 2025; Cheng et al., 2025) have been used to generate controlled vacancy populations while minimizing additional impurity incorporation. Thermal annealing after vacancy creation is essential to form the desired centers as discussed in detail in Section 4. It has been shown that combining delta-doping with low-damage vacancy generation followed by optimized annealing yields deterministically localized near-surface centers with narrow spectral diffusion and improved coherence compared to centers produced by high-energy ion implantation schemes (Ishikawa et al., 2012; Mclellan et al., 2016; Ohno et al., 2014; Smith et al., 2019; Bradac et al., 2019).

3.2.3 Shallow defect implantation and overgrowth

Both defect-incorporation routes discussed have drawbacks. Delta-doping with certain precursors (e.g., Sn-bearing gases) can leave residues in the CVD tool and contaminate subsequent runs, while ion-implantation often produces qubits with reduced performance and limited depth control because of implantation energy and damage. A hybrid route that preserves depth control while minimizing ion-induced lattice damage is shallow ion implantation followed by epitaxial overgrowth (Rugar et al., 2020). In this approach ions are implanted at very low energies (typically ≤20 keV) so the collision cascade is confined to the upper few nanometers. A thin, high-quality CVD cap is then grown to set the final emitter depth. The overgrowth buries the implanted species and, under appropriate hydrogen-rich growth conditions, provides a gentle, quasi-in-situ anneal that repairs much of the implantation damage (see Section 4). This decouples dopant placement from deep, high-energy implantation while avoiding many of the contamination risks associated with reactive dopant precursors (Smith et al., 2019; Lesik et al., 2013; Schröder et al., 2017).

Experimental reports show clear benefits from the shallow-implantation and overgrowth route. Near-surface NV and group-IV centers produced this way exhibit narrower optical lines, reduced spectral diffusion, and longer T 2 times compared with centers created by higher-energy implantation followed by blanket annealing (Rugar et al., 2020). Recent demonstrations of site-controlled group-IV emitters embedded under high-quality overgrowth caps (Lukin et al., 2020) further highlights the method’s suitability for photonic integration, reporting substantially improved spectral stability and device compatibility (Bradac et al., 2019; Schröder et al., 2017).

3.3 Diamond synthesis on non-standard substrates

Recent work has moved beyond growth on pristine bulk substrates to exploit pre-growth processing: patterned or modified substrates can improve implantation/doping yield, enable membrane-like architectures (Guo et al., 2024; Moriceau et al., 2012; Basso et al., 2024), and simplify heterogeneous integration (Katsumi et al., 2025; Mandal et al., 2021).

Templated growth (mask-defined or seeded nucleation) confines CVD deposition to apertures and directly defines nanopillar arrays (Sartori et al., 2019) or resonator geometries (Zhang et al., 2016; Zhang et al., 2023a), reducing the need for post-growth patterning. In practice, small variations in mask thickness, aperture profile, or local gas flow cause anisotropic edge growth because adatoms at feature rims see different step densities and incorporation kinetics than those on wide terraces. The result is skewed sidewalls and imperfect facet control. Liftoff and mask removal add risks: residual mask material, redeposited species, or microcontamination at the mask–diamond interface act as heterogeneous nucleation centers during overgrowth, producing roughness and scattering losses that degrade optical performance and create long-lived chamber contaminants (Zhang et al., 2016; Zhang et al., 2023a; Shimoni et al., 2014). For reliable templated growth, reactor hydrodynamics, local chemistry, and mask processing must therefore be explicitly tuned and monitored (Ashfold et al., 1994; Schwander et al., 2011).

Smart-cut or ion-cut techniques use high-dose implantation (commonly He or H, e.g., He ≥5×10¹⁵ cm⁻²) to create a buried, highly defective layer that forms a hetero-stack. Post-implantation annealing at 1,200 °C coalesces this damage into a sharp graphitic tether (Guo et al., 2024; Guo et al., 2021) (see Section 4 for annealing dynamics). Such a buried layer modifies local thermalization, can perturb the hydrogen-mediated etch/deposit balance, and seeds nucleation sites that increase island density relative to bulk substrates. Consequently, membrane-ready diamonds require tailored pre-growth cleaning (to avoid sub-surface etching as shown in Figure 4a), reduced thermal budgets or adjusted gas chemistries, and a narrower growth window to maintain smooth step-flow (Figure 2C) growth.

Growth-parameter windows likewise shift. We observe that maintaining step-flow morphology on smart-cut membranes often requires lower substrate temperatures than used for bulk homoepitaxy (typically 50 °C–100 °C lower). Lowering the temperature below the membrane window promotes three-dimensional islanding and strong pyramidal faceting, while raising it above the window increases step-edge proliferation and roughening via the formation of non-epitaxial crystallites. Figure 4 shows examples of common defects seen on smart-cut substrates as a result of these effects. The temperature trends observed on smart-cut substrates are consistent with surface-kinetic arguments: the surface diffusion length of carbon adatoms and the hydrogen etch/deposition balance change rapidly with local temperature, and subsurface damage or reduced thermal conductivity modifies the local thermal boundary conditions so that a nominal heater set-point produces different adatom kinetics on processed substrates compared to bulk. This complexity is furthered by in-situ pyrometry failing to compare substrate types due to varying surface reflectivity properties. As such, a universal and easily calibrated thermometry approach would be highly desirable. As already mentioned in previous sections, ensuring a consistent miscut profile is critical to avoid mixing confounding variables.

4 Post-synthesis processing

To realize electrostatically stable spin-defect color centers with high optical and spin coherence, post-synthesis thermal and chemical treatments are indispensable. Regardless of whether the defects are introduced by ion implantation, in-situ doping, or irradiation, these processes inevitably generate lattice damage and modify the local defect/dopant populations. The subsequent thermal evolution of these species such as their diffusion, recombination, and complex formation, determines both the yield and quality of the resulting color centers.

In the context of heavily ion bombarded regions, where the damage creates a critical threshold of point-defects D c (Fairchild et al., 2016) above 1×10²² vacancies/cm³ to 9×10²² vacancies/cm³, dislocation slips and mobile vacancies coalesce around damage-defined nucleation points, resulting into a regional sp 3 to sp 2 transition (i.e., formation of a graphitic carbon layer) (Guo et al., 2021). If well controlled, this can be leveraged for interesting nanofabrication processes as graphitic carbon presents a vast set of opportunities for both wet and dry chemical processing. In parallel, regions with damage below D c can partially recrystallize to diamond. At 1,260 °C, the interface between damaged and undamaged zones becomes significantly sharper as implantation induces large local pressures (from density changes during transformation). Finally, stress control remains critical to prevent cracking of the preserved layer (Fairchild et al., 2016).

In diamond, the mobility of vacancies and hydrogen is strongly temperature-dependent, and their interactions govern the activation and charge-state stabilization of common quantum defects such as the nitrogen-vacancy (NV −) and group-IV vacancy complexes (SiV, GeV, SnV, PbV). Controlled annealing drives vacancy diffusion, but excessive temperatures can lead to defect clustering or charge-state conversion. Similarly, hydrogen and interstitial carbon play a dual role: it can passivate unwanted non-radiative centers and surface traps, yet overabundance or uncontrolled diffusion can neutralize desired color centers by converting XV − to XVH (e.g., NV − + H + → NVH) or annihilating it to X C (X dopant in a carbon substitutional site).

Post-synthesis processing therefore serves a dual purpose: it heals and activates, but must be engineered to avoid overcompensation and charge instability. The following subsections examine the thermal kinetics of XV defects in diamond, their stability, and potential mechanisms that explain reduced yield, all in the context of achieving reproducible and robust color centers suitable for quantum technologies.

4.1 Carbon interstitials and monovacancies

Lattice damage in the form of vacancies and interstitials (Frenkel pairs), produced by ion implantation or electron irradiation (Campbell and Mainwood, 2000), is central to spin-defect formation, but can lead to irreversible amorphization/graphitization above vacancy densities of order ≥1×10²² cm⁻³ (Hickey et al., 2009). Frenkel pairs are generated when incident particles deposit energies ≥7 eV (Mainwood et al., 1994; Deák et al., 2014), the bond-breaking cost for a carbon sp 3 lattice site (Mainwood and Stoneham, 1997). Upon thermalization, the interstitial carbon C i (sometimes denoted I 0) mobilizes at relatively low temperatures ≈400 °C (Hunt et al., 2000; Goss et al., 2004b; Allers et al., 1998; Iakoubovskii et al., 2003) with an activation energy of ≈1.68 eV and a preferred <100> migration direction (Weigel et al., 1973), whereas the neutral vacancy (GR1), mobilizes at 600 °C with a diffusion rate of 2 nm s⁻² (Onoda et al., 2017; Alsid et al., 2019; Räcke et al., 2021), at an activation energy of 2.2 (2) eV (Hunt et al., 2000; Davies et al., 2001; Pu et al., 2001; Iakoubovskii et al., 2003; Goss et al., 2004b; Iakoubovskii et al., 2005; Goss et al., 2005). With 1,000 °C annealing, it is observed that the amount of vacancies drops by ≈90% (Onoda et al., 2017). It is commonly accepted that vacancies annihilate at the crystal surface. However, DFT results suggests that they could also become trapped in a 5 eV well just below the surface (Lvova et al., 2017) due to the slight lattice reconstruction. In contrast, charged monovacancies, such as ND1, are reported to be immobile and unable to create clusters or other complexes (Alsid et al., 2019; Davies, 1977; Fávaro de Oliveira et al., 2017; Lühmann et al., 2019; Yamamoto et al., 2013; Goss et al., 2004b; Deák et al., 2014), highlighting the need to account for their rich electronic level structure (Deák et al., 2014).

Each individual crystal defect species, such as C i and V C, can be modeled to migrate within the lattice using the Arrenhius Equation 1 where ν 0 describes the attempt frequency of the lattice (typically on the order of diamond’s Debye frequency, 10 THz). Γ is the rate at which the defect in question translates within the lattice, and E a is the necessary activation energy to achieve this. Calculated activation energy barriers for a given temperature and lifetime are given in Table 1. It is useful to cross-reference activation energies listed for each defect to provide intuition about movement and activation temperatures necessary.

Γ = ν 0 e − E a k b T (1)

TABLE 1. Calculated E a barrier (given in eV) to achieve migration within the time and temperatures for ν 0 = 10 THz, e.g., the activation energy that would be necessary to only move a species at a temperature of 1,600 °C at a rate of one translation every 30 min would be ≈6.02 eV.

4.2 XV center formation

XV centers, X (atomic) + V (vacancy) clusters, in diamond can be formed by numerous channels. A key difference between the NV center and the G4V centers is that both the Nitrogen and associate vacancy sit on substiutional sites (Figure 5a), where the G4V centers adopt a split-vacancy configuration, with the G4 atom sitting between two carbon sites that are unoccupied (Figure 5c). Additionally, we note the relevant quantization axes for the B field projection to allow for degeneracy breakage via Zeeman effect as they relate to the energy diagrams presented in Figures 5b,d for the N V − (Delegan et al., 2023; Gali, 2019) and G4V (Guo et al., 2023; Bradac et al., 2019; Thiering and Gali, 2018; Meesala et al., 2018) centers.

Depending on doping and exact X dopant size (e.g., N is smaller than Sn), irradiation, or annealing procedures a few competing mechanisms might be at play. The following mechanisms (Goss et al., 2004b; Goss et al., 2005; Deák et al., 2014; Thiering and Gali, 2015; Onoda et al., 2017; Gali, 2019; Chakravarthi et al., 2020; Wahl et al., 2020; Wahl et al., 2024; Liu et al., 2024; Pugliese et al., 2025; Baba et al., 2025) have been proposed: for formation (Equations 2, 3), disassociation (Equations 4, 5) and competition (Equations 6–8).

X s + E break at X → X V + C i (2)

X s + E move V + V → X V (3)

X V + E break and move V → X s (4)

X V + C i → X (5)

X V + C i → X V C i (6)

X V + V → X V V (7)

V + V → V V (8)

All of these mechanisms are intimately dependent upon the charge state of the constituents, we focus on NV −, SiV −, GeV −, and SnV −. Goss et al. (Goss et al., 2005) investigated the stability of X and their XV formations. From their work, Sn is the only thermodynamically preferred dopant in the XV configuration compared to its substitutional location (Sn C). As such, it requires no activation energy to combine with a nearby vacancy to create the split-vacancy configuration. In contrast, upon meeting a vacancy N, Si, and Ge will require an activation energy to surpass the barrier necessary to favorably form (N, Si, or Ge)V-centers. ultimately, they will form an XV center that has a lower overall energy than the input, as such they will subsequently release 3.5, 4.4, or 4.9 eV respectively (Goss et al., 2005; Gali, 2019). The barrier necessary to overcome the XV binding energy while allowing the vacancy to migrate away is ≥5 eV for all the aforementioned defects, hinting at their thermal stability post-formation. Table 2 shows that this allows for excellent thermal stability, requiring temperatures ≥1,400 °C (2,200 °C) for NV (SnV) to dissociate. For a sense of representative migration rates at temperature, see Table 1.

TABLE 2. Formation and disassociation energies of XV centers in eV.

Fundamentally, the routes highlighted in Equation 2 and Equation 3 provide a source of vacancies for substitutional dopants. It has been argued theoretically (Deák et al., 2014; Gali, 2019) and shown experimentally (Wahl et al., 2020; Wahl et al., 2024) that the former mechanism drives the formation of XV centers. The intuition is that at a substitutional dopant site, the presence of one unsatisfied sp 3 bond lowers the vacancy formation energy, increasing the likelihood of vacancy creation nearby. This trend is supported by β − emission studies showing that immediately after implantation, approximately 20% of Ge atoms (Wahl et al., 2024) and 41% of Sn atoms (Wahl et al., 2020) occupy bond-centered sites, suggesting they are in the split-vacancy configuration.

Deák et al. argue from DFT calculations that vacancy creation by long-range vacancy mobilization is energetically disfavored for NV formation and instead promotes higher-order complexes (Deák et al., 2014). By contrast, Onoda et al. provide direct experimental evidence of vacancy migration as detailed in Equation 3, where NV centers were observed microns away from ion-bombarded damage sites, a clear signature that mobile vacancies can travel and pair with substitutional nitrogen (Onoda et al., 2017). The most consistent interpretation is that both pathways operate, but with different probabilities where creating a vacancy adjacent to a substitutional nitrogen (the ‘local break-make’ route as shown in Equation 2) is energetically preferred. Deák estimates an effective energy discount of ≈5.0 eV at a N C site (Deák et al., 2014; Gali, 2019), so it is likely the dominant mechanism under many conditions, while vacancy diffusion contributes where sufficient thermal budget or damage cascades exist. Critically, one must account for the formation of higher-order complexes when accounting for those formation routes. Resolving the relative importance of these routes for NV center formation will require more targeted experiments (e.g., β − emission or spatially resolved implantation/anneal studies) that directly track impurity and vacancy sites before and after annealing.

With sufficient energetic input, pathway Equation 2 allows for paired XV + C i defects. At modest temperatures of ≥400 °C, interstitial carbon C i becomes mobile and can recombine with the XV (Equation 5), migrate away, or form bound complexes following Equation 4 (Chen et al., 2025; Cheng et al., 2025). Experimentally, a 300 °C anneal reduces Ge bond-centered fractions by 25% in 10 min (Wahl et al., 2024), and a 920 °C anneal further lowers Ge bond-centered sites from 20% to 6% to 9% while Sn bond-centered fractions drop from 41% to 32% (Wahl et al., 2020).

Although XV complexes form deep potential wells, they can collapse to substitutional X C via carbon recombination as highlighted in Equation 5 due to the fact that C i is highly mobile (activation E a ≈1.68 eV) and recombination is energetically favorable for smaller impurities (N, Si, Ge) (Goss et al., 2005). Consequently, many freshly formed XVs are annihilated by immediate C i recombination, limiting yields to a few percent in practice (Pugliese et al., 2025). In contrast, SnV is unusually stable. It generates large local strain that favors SnV – C i complexes over simple recombination, with an estimated dissociation energy ≈2 eV with experimental validation from laser annealing ( Cheng et al., 2025). This suggests SnV complex formation will benefit greatly from annealing at relevant temperatures (see Table 1).

Further high-temperature annealing can instead produce higher-order complexes (e.g., XV – V (Kuate Defo et al., 2019)) when mobile vacancies are trapped near dopant sites. DFT indicates these complexes are substantially more stable (by ≈4.5 (5) eV) than simpler configurations and require very high temperatures (≥1,600 °C) to dissociate (Wahl et al., 2020; Görlitz et al., 2022; Narita et al., 2023). Such extreme thermal budgets, however, necessitate deep dopant placement ≥1 µm to mitigate the effects of surface roughening and etching (Wang et al., 2021; Narita et al., 2023; Wang et al., 2024a).

In summary, XV centers preferentially form during ion implantation or electron irradiation because the dopant site already lacks one sp 3 bond, lowering the formation energy. Smaller XV centers (e.g., N, Si, Ge) are more susceptible to annihilation through interstitial carbon recombination, whereas SnV’s greater thermodynamic stability and local strain favor the formation of higher-order complexes with migrating vacancies or interstitial carbon. Although seldom examined experimentally, XV centers are predicted to exhibit elevated rates of XV–V pairing due to these interactions.

4.3 Competitive defect pathways

Thus far we have emphasized the structural routes to form XV centers. However, achieving a reliable, optically active spin qubit also requires an electrostatically stable, low-noise local environment. Residual defects—divacancies, vacancy clusters, and hydrogenated complexes—act as charge traps, fluctuating charge/noise sources, and unwanted recombination or passivation sites that destabilize the desired charge state and shorten coherence. Any cohesive defect-formation and activation recipe must therefore pair precise dopant/vacancy placement with targeted strategies for vacancy management, hydrogen mitigation, and band-engineering (annealing schedules, passivation/overgrowth, and Fermi-level control) to produce reproducible, charge-stable qubits.

Vacancy clusters form via the pathway in Equation 8 when mobile vacancies meet and bind exothermically. The divacancy VV (optical absorption band near 450 nm–480 nm) is a particularly important case. Its formation is energetically favorable (releasing 4.2 eV) so it can form more readily than an NV complex in damaged regions (Deák et al., 2014; Pu et al., 2001). Divacancies and larger vacancy aggregates compete with NV and G4V formation (i.e., they act as alternative acceptors). They are often paramagnetic and thus a major source of charge trapping and spin noise that degrades coherence (Pezzagna et al., 2010; Ishikawa et al., 2012; Yamamoto et al., 2013; Haque and Sumaiya, 2017; Capelli et al., 2019; Santonocito et al., 2024; Deák et al., 2014; Thiering and Gali, 2018). Vacancy clustering can also occur in the vicinity of an existing XV center because the local dopant environment traps migrating vacancies, further increasing local decoherence channels (Kuate Defo et al., 2019).

Practically, these deleterious aggregates are addressed thermally. Dissociation and reduction of small vacancy clusters are commonly achieved with high-temperature anneals in the 1,000 °C–1,200 °C window, which improves optical and spin metrics (Yamamoto et al., 2013; Alsid et al., 2019; Santonocito et al., 2024). Even higher thermal budgets (≥1,700 °C) have been used to mobilize substitutional nitrogen and promote formation of more benign (relatively speaking) aggregates such as NVN (i.e., H3 centers), further reducing spin noise in some recipes (Wong et al., 2022). For broader context on nitrogen aggregation and annealing pathways see (Davies et al., 1992; Davies and Alan, 1993; Mainwood, 1994; Iakoubovskii et al., 2003; Jones et al., 2015).

Suppressing the formation of divacancies and larger vacancy clusters by introducing controlled doping in either p-type (Fávaro de Oliveira et al., 2017) or n-type (Lühmann et al., 2019) diamond is also a possibility. This strategy stabilizes the lattice by charging and immobilizing vacancies, while also increasing the availability of free carriers (particularly electrons) that can be trapped at XV sites without competition from other vacancy-based acceptors. Vacancies themselves are known to contribute to spectral diffusion and optical instability (Li et al., 2024a), while divacancies introduce charge instability (Trusheim et al., 2019; Görlitz et al., 2022). Thus, doping-assisted vacancy control offers a promising route toward charge-stable, optically coherent color centers. Importantly, this approach aligns with the vacancy formation mechanism described in Equation 2, while simultaneously suppressing competing vacancy complexes. Interestingly, despite their detrimental role, divacancies have also been proposed as a possible precursor pathway for XV center formation (Baba et al., 2025). Further theoretical studies examining the interplay between VV and X species (Qiu et al., 2023) are needed to determine whether such a route could rival the dominant mechanism of Equation 2.

During PECVD growth, hydrogen is inevitably incorporated into the near-surface region of diamond (see Section 3), and it can exist in multiple forms (molecular-like H 2 * or interstitial H i). The dominant speciation remains debated. Interstitial hydrogen is particularly important because it readily passivates dopants and XV centers (e.g., XV + H i → XVH) and, once bound, can be difficult to remove—breaking some hydrogenated complexes has been reported to require very high thermal budgets ≥1,600 °C) (Goss et al., 2002; Czelej et al., 2018; Nie et al., 2021; Findler et al., 2024; Ghafari Cherati et al., 2025). For comparison, the H–X binding strength is comparable to a C – H bond (Glover et al., 2004). Making things more complex is the fact that charged hydrogen species are extremely mobile with positively charged hydrogen having a near-zero migration barrier, making if effectively mobile at room temperature, whereas neutral and negatively charged hydrogen exhibit migration barriers on the order of 2.0 eV and 2.4 eV, respectively. These values are comparable to neutral vacancy migration (Mehandru and Anderson, 1994; Connell et al., 1998). Because hydrogen in its various charge states is mobile across the relevant processing temperature range, active mitigation (degassing/anneals, careful growth chemistry, and overgrowth/passivation strategies discussed above) is required to avoid dopant passivation and charge-state quenching of spin defects.

To demonstrate the impact of hydrogen passivation on XV activation, we annealed post-CVD diamond in high vacuum (HV, ≤1×10^–6 Torr) over the range 400 °C–850 °C and monitored outgassing with a residual-gas analyzer (RGA). A pronounced H 2 desorption signal appears beginning near 500 °C, consistent with reported activation energies for neutral hydrogen (Figure 6a). We then compared two otherwise-identical samples containing a ¹⁵N delta-doped layer (2 nm thick at 10 nm depth): one was degassed prior to electron irradiation (Figure 6c) and the other was not (Figure 6b). Post 850 °C thermal activation under similar vacuum conditions for 2 h, the degassed sample showed a clear increase in ¹⁵NV⁻ yield, consistent with effective mitigation of NVH formation by hydrogen removal. Additionally, the observed defects were identified as ¹⁵NV⁻ (and distinguished from substrate-residing ¹⁴NV⁻) using pulsed optically detected magnetic resonance (ODMR) measurements (Figures 6d,e). Using an inverted double-Gaussian fit, we extract a hyperfine splitting of 2.85 (5) MHz and 3.09 (1) MHz MHz before and after degassing, respectively. The temperatures explored are expected to mobilize H 2 and H * species. A more wholistic process, targeting H − would require similar annealing at temperatures ≥1,000 °C.

In summary, judicious thermal processing is essential to produce high-yield, optically active XV centers while preserving surface integrity. Approximate activation windows are: ≥400 °C (C i mobilization), ≥600 °C (neutral-vacancy mobility), ≥1,200 °C (dissociation of small vacancy clusters/improvement of optical metrics), ≥1,600 °C (onset of dissociation for some deep aggregates such as NV −/Type-II SnV behavior), and ≥2,100 °C regimes where only the most robust XV complexes remain (e.g., SnV −). Annealing in atmosphere ≥400 °C raises practical challenges. The sp 3 lattice is susceptible to surface graphitization in the presence of oxygen, so treatments are typically run in UHV or reducing environments (UHV is preferred when hydrogen removal is required to avoid passivation) and require careful selection of furnace/crucible materials because common ceramics and metals can react with diamond or process gases at high temperatures (≥1,300 °C) (Goss et al., 2002; Komarek et al., 1963).

Because very high temperatures bring additional risks (i.e., surface etching, tool contamination), alternative routes such as rapid thermal annealing or localized laser annealing - both of which provide large instantaneous activation energies with reduced cumulative time at peak temperature - are attractive and have shown promise for activating XVs while limiting graphitization and diffusion (Chen et al., 2017; Addhya et al., 2024; Cheng et al., 2025). Thermal recipes must be co-designed with depth placement (deep implants for extreme thermal budgets) and protective caps/overgrowth. Finally, band-engineering (donor control, passivation, gates) is critical in order to realize stable XV − centers suitable for quantum applications (Narita et al., 2023; Wang et al., 2024a).

5 Band engineering and charge stability

Band engineering via electrostatic gating, co-doping, surface termination, or targeted optical charging provides a primary toolkit for stabilizing the desired charge state of color centers and for suppressing the local charge-noise that degrades spin coherence. Physically, these techniques operate by changing the local electrostatic potential: contacting or gating an interface with a different work function drives charge transfer until equilibrium, producing a built-in potential that bends the bands, and redistributing carriers near the surface. The spatial extent of this space charge or depletion region is understood by Poisson’s equation. For a uniform dopant density we expect this region to scale approximately as Equation 9, with ε, Δ ϕ, and q N describing the dielectric constant, effective surface potential, and charge density, respectively. This illustrates that the range over which a gate or surface termination alters the charge availability depends strongly on the free-carrier density. Surface terminations and adsorbates introduce surface states that can pin the Fermi level and/or induce transfer doping. In practice, tailoring the near-defect electrostatics is achieved by combining one or more of these knobs: set the bulk Fermi level by (co-)doping, applying local gates or buried delta-doping to tune the local potential, and tailor the surface chemistry that minimizes surface traps and charge fluctuations. Theoretical work (Gali, 2019; Thiering and Gali, 2018) provides guidelines for electronic band positions that enable the formation of the negatively charged XV center: NV − requires a Fermi level near the substitutional nitrogen center, while G4Vs are typically lower, around 2 eV–2.5 eV above the valence band. Realizing the latter has proven to be a challenge.

W ∼ 2 ε Δ ϕ / q N (9)

For NV centers these principles set straightforward constraints on achievable NV − yield. The neutrally charged substitutional nitrogen (i.e., N C, often called P1) acts as the dominant donor (Kuate Defo et al., 2021) in many quantum-grade diamond samples and thus largely determines the bulk Fermi level. The fraction of negatively-charge NV centers depends on whether the Fermi level lies sufficiently high relative to the NV – / 0 transition level (Weber et al., 2010). Introducing additional electron donors (i.e., raising the local Fermi level by electrostatic gating or buried n-type layers) increases the NV − fraction, whereas acceptor doping (e.g., boron, vacancy) or surface transfer that depletes electrons pushes centers toward the neutral NV 0 state. Importantly, simply increasing donor concentration is not always desirable. High P1 density raises the spin-bath noise and shortens coherence times, so band engineering must balance charge stability against magnetic noise. For near-surface devices, the combined approach - careful background doping, local delta-doping or gates to raise the local chemical potential, and robust surface passivation to eliminate fluctuating traps - has proven effective at producing stable, optically bright NV − ensembles and single emitters (Collins, 2002; Doherty et al., 2013).

In parallel, group-IV vacancy (G4V) centers have emerged as attractive quantum emitters because of their narrow optical lines (low spectral diffusion due to their inversion symmetry) and as such, favorable photonic integration. However, questions about charge stability remain. Both theory and experiment indicate a simple physical origin. The electronic levels relevant for optical excitation and ionization lie relatively close to the host bands, so photo-excitation or environmental charge fluctuations readily change the center’s charge state (Thiering and Gali, 2018; Görlitz et al., 2022). First-principles studies report charge-transition/formation levels for many group-IV centers on the order of a few electronvolts above the valence-band maximum (i.e., ≈2.4 eV in some calculations) placing these centers near the energetic range of other acceptor-like defects and making them susceptible to ionization and hole capture (Thiering and Gali, 2018). Experimentally, this manifests as emitter blinking, spectral jumps, and reduced photostability under typical excitation conditions (Görlitz et al., 2022). Practically, stabilizing group-IV centers therefore requires the same band-engineering toolbox used for NV: raise the local Fermi level (i.e., co-doping, buried n-type layers, electrostatic gates), eliminate nearby charge traps (i.e., anneal/overgrow/passivate), and minimize surface/adsorbate-induced transfer doping—measures that reduce photo-ionization probability and suppress spectral diffusion for photonic and sensing applications.

5.1 Surface termination

Surface engineering is an effective lever to control the charge state of near-surface color centers. Hydrogen termination (H-terminated) produces a negative electron affinity and, in ambient, commonly leads to adsorbate-mediated hole accumulation that pins the Fermi level near the valence band—conditions that tend to favor neutral charge states and deplete electrons required for NV − or other negatively charged centers (Maier et al., 2000; Kaviani et al., 2014; McCloskey et al., 2024). In contrast, oxygenation (O-termination) yields a positive electron affinity, suppresses surface conductivity, and thereby helps preserve electron availability in the near-surface region, making it easier to stabilize negatively charged G4V and NV centers (Baumann and Nemanich, 1998; Zhang et al., 2023b; Corte et al., 2022).

The details of the termination recipe are critical. Thermal oxidation, tri-acid cleans, and oxygen plasma treatments produce different chemistries, reconstructions, and amounts of residual sp 2 carbon, and these differences strongly affect spin noise, charge stability, and spectral diffusion. Depth-resolved XPS combined with polarization-dependent X-ray absorption and coherence measurements show that some oxygen treatments induce sub-nanometer sp 2 reordering that creates electrically active traps and increases spectral instability, while other oxidation protocols yield superior spin metrics (Sangtawesin et al., 2019; Vidrio et al., 2024; Vidrio et al., 2026). Thus, although O-termination is often preferred for stabilizing negative charge states, its application must be carefully chosen and verified for each workflow (Rodgers et al., 2021).

Recently, alternative terminations such as fluorination (Weber et al., 2010; Rodgers et al., 2024; Thake and Jenkins, 2025; Pershin et al.,2025; Sadzak and Krueger, 2025), controlled nitrogenation (Kawai et al., 2019), and oxide thin-films via atomic layer deposition (ALD), have been explored to decouple emitters from fluctuating surface chemistry while preserving charge stability (Rodgers et al., 2021). Surface fluorination has been reported to increase the fraction of negatively charged NV centers relative to both H- and O-terminated surfaces (Cui and Hu, 2013). However, many of these studies focus on charge-state statistics and photoluminescence yield rather than detailed spin-coherence metrics, so the impact of fluorination on quantum performance (i.e., T 2, spectral diffusion, etc.) requires further, more systematic characterization. Nitrogen-based terminations and nitridation approaches have also been proposed and tested. Under well-controlled conditions they can yield optically stable NV − centers while reducing the risk of oxygen-driven graphitization (Kaviani et al., 2014; Chou et al., 2017; Kawai et al., 2019), but incomplete removal of residual amine or NH x fragments can introduce process contaminants that complicate coherence measurements.

Thin dielectric overlayers deposited by ALD (e.g., Al 2 O 3, SiO 2, TiO 2) offer a complementary strategy. A conformal nm-scale capping overlayer can passivate dangling bonds, supply fixed charge to tailor band-bending, and physically separate adsorbates from near-surface emitters (Xie et al., 2022; Yu et al., 2025). These films modify the near-surface electrostatics (via fixed charges and work-function differences) and can reduce surface electronic traps, but they also increase the emitter-to-environment separation and therefore reduce near-field sensing sensitivity if the cap is too thick. Practical implementation requires careful control of deposition temperature, film stoichiometry, and post-deposition curing. ALD precursors and process residues can introduce charge traps, magnetic contaminants, or intrinsic defects (e.g., oxygen vacancies in binary oxides) that degrade performance if not properly managed. Functionalized caps can also provide application-specific capabilities (e.g., biotin–PEG–silane layers for biomolecular immobilization), but these add further chemical complexity that must be vetted for quantum-grade performance (Chi-Durán et al., 2025).

Regardless of the chosen approach, interface engineering is essential for device integration and remains challenging to develop reproducibly. Characterizing the exact chemical state, reconstruction and sub-nanometer disorder at these interfaces is difficult with routine tools, which complicates process transfer between labs and makes reproducible, quantum-grade surface recipes difficult to establish (Sangtawesin et al., 2019; Vidrio et al., 2024; Vidrio et al., 2026).

5.2 Electrostatic fields and doping

Electronic doping in diamond—through substitutional donors and acceptors that raise or lower the Fermi level—represents a direct route to tailor charge stability and band bending, but a comprehensive discussion of doping mechanisms is beyond the scope of this paper. A detailed overview of dopant formation energies, charge transition levels, and incorporation mechanisms can be found in prior reviews (Thiering and Gali, 2018; Deák et al., 2014).

Representative dopants that have been explored in the context of electronic and quantum applications: boron for p-type doping (Rose et al., 2018; Volpe et al., 2009; Bilgin et al., 2024; Czelej et al.,2017), nitrogen for n-type behavior and deep donor states (Farrer, 1969; Heremans et al., 2009; Weber et al., 2010; Deák et al., 2014; Ferrari et al., 2018; Bilgin et al., 2024), and phosphorus for shallow donor doping (Goss et al., 2004b; Kalish, 2001; Czelej et al., 2018). Co-doping strategies using phosphorus in combination with other donors or acceptors have been proposed to mitigate compensation effects and enhance charge-state control (Lühmann et al., 2019; Geng et al., 2023). More recently, sulfur has emerged as a potential deep donor candidate (Ghafari Cherati et al., 2025). Among these, boron doping remains the most experimentally mature approach, offering reliable p-type conductivity but also introducing acceptor levels that can act as charge drains, driving NV − → NV 0 transitions, a key trade-off for quantum applications (Rose et al., 2018).

In parallel with static doping strategies, the direct application of electrostatic fields presents a powerful route to dynamically bend the bands near the diamond surface and thereby modulate the local charge state of color centers (Grotz et al., 2012; Lozovoi et al., 2022; Karaveli et al., 2016). In one seminal study, an electrolyte-gated diamond device enabled controlled shifting of the Fermi level at the surface, switching ensembles of NV centers between NV 0 and NV − states via gate voltage–induced band bending (Grotz et al., 2012). For example, planar Schottky or gate structures on hydrogen-terminated diamond have demonstrated high-speed switching of the charge state of a single nitrogen-vacancy center (NV) from NV + → NV − → NV 0 on timescales of 10-100ns by exploiting field-induced depletion or accumulation and shifting the Fermi level relative to the NV center charge transition levels. Importantly, this form of band-bending control is different from the conventional Stark effect. Instead of shifting the energy levels of the defect under an applied local electric field, it modifies the carrier distribution and Fermi level in the host lattice and thereby changes the defect’s equilibrium charge state.

From a device-engineering perspective, diamond gating architectures have been implemented via Schottky or ohmic contacts on hydrogen-terminated or oxidized diamond surfaces (Karaveli et al., 2016; Grotz et al.,2012), buried doped layers (Chaudhuri et al., 2022), and metal–insulator–semiconductor (MIS) structures that exploit hydrogen or oxygen terminations in conjunction with electrode stacks (Lozovoi et al., 2022). These architectures mirror high-voltage wide-bandgap transistor designs (e.g., hydrogen-terminated diamond in heterostructures with graphene or other two-dimensional materials) (Sasama et al., 2022; Yuan et al., 2025; Qi et al., 2024; Lo et al., 2025). Integrating these gating strategies with spin-defect systems enables charge-state preparation and reconfigurable electrostatic control of defect qubits, offering routes to dynamically tuned quantum sensors and photonic interfaces.

Metal contacts on diamond remain a key materials challenge due to its wide range of potential operation (often outliving the contacts), wide band gap structure and deep dopant levels (Araujo et al., 2021). For p-type diamond, reliable ohmic contacts have been formed using carbide-forming metal stacks and annealing (Manifold et al., 2021). Lateral p-type diamond Schottky-barrier diode using (Hoff et al., 1996; Manifold et al., 2021) of 30 nm–30 nm–100 nm with annealing. On hydrogen-terminated diamond, Pt / Au (Zhang et al., 2019) showed nearly ohmic behavior via negative barrier (Tsugawa et al., 2010). Schottky contacts on diamond are typically reserved for O-terminated surfaces with Au, Zr, or Al, being the most common (Gaowei et al., 2012, Tsugawa et al., 2010, Araujo et al., 2021).

6 Conclusion and outlook

As device-integrated color centers mature toward technological deployment, a sharper materials science foundation is required to turn promising demonstrations into robust, scalable technology. In this review we catalogued processing levers across substrate preparation (Section 2), growth synthesis (Section 3), post-growth activation and healing (Section 4), and band-engineering approaches for charge stabilization (Section 5). Different applications place different weights on metrics such as coherence, spectral stability, depth, and yield, so progress must be judged against the target use case (quantum sensing vs. communication vs. integrated photonics). Looking forward, three coordinated thrusts are especially warranted.

First, we must develop new methods to leverage the rich diffusion and atomic dynamics that govern defect formation and evolution. This includes: synthesis and anneal recipes that actively steer vacancy/interstitial kinetics rather than passively tolerate them. In this vein, localized rapid thermal protocols (e.g., laser annealing or RTA) that supply high activation energy without long-range diffusion. Other opportunities include the use of theoretically guided synthesis approaches (Marcks et al., 2024) and other more nascent autonomously guided techniques that leverage robotics, Bayesian-driven optimization, and more complex ML models to drive the materials synthesis directly via in-situ metrology, as inspired from synthesis of other material systems (Szymanski et al., 2023; Tom et al., 2024). As well as experiment-theory campaigns to measure and model activation barriers and reaction pathways relevant to XV formation and aggregation while ensuring high host material quality.

Second, we need far better, quantum-grade surface characterization that allows for detailed investigations of both chemical termination and structural state (i.e., reconstruction, to differentiate ordered and disordered surface terminations). Routine metrologies must go beyond bulk spectroscopies, quantify sp 2 fractions, and identify electrically active surface states that pin the Fermi level. Advances in ALD (Jones et al., 2023)/atomic-layer etching (Michaels et al., 2023) metrology, X-ray crystallography (Delegan et al., 2023) and subsurface probes (Dontschuk et al., 2023), and in-situ analytical techniques will be central here. Critically, these tools must be validated by the quantum metrics that matter, such as T 2, spectral diffusion, and charge-state statistics.

Third, a wide and reliable toolkit for surface passivation and functionalization is required—and it must work for both principal growth orientations. Practical device engineering demands passivation chemistries and thin caps that preserve or enhance NV − and G4V performance on both (001) and (111) faces, along with scalable recipes for ALD caps and selective terminations (e.g., fluorination/nitridation), as well as functionalization (Rodgers et al., 2021).

Finally, industry-grade progress will require community coordination: agreed material specifications (miscut tolerances, surface finish), reproducible process flows, and shared benchmarks for spin/optical performance. Progress in instrumentation (high-resolution, depth-resolved chemical/structural probes), theory-enabled synthesis, and autonomous optimization together can close the loop from fundamental understanding to scalable device/sensor/materials fabrication. When these pieces advance in concert, the field will be positioned to deliver deterministic, high-fidelity quantum emitters in scalable and manufacturable diamond platforms.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. AI used for spell checking/grammar.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Jean-Philippe Tetienne, RMIT University, Australia

Reviewed by: Shingo Sotoma, Kyoto Institute of Technology, Japan

Thitinun Gas-osoth, University of Phayao, Thailand