Precision sub-monolayer manipulation of diamond surface chemistry using laser direct write oxidation in air

Diamond is an extreme material of substantial interest for advancing capability in quantum computing and sensing, optomechanics, electrochemistry, biosensors, electronics, and photonics.^1–4 In many of these applications, the chemical composition of the surface affects critical properties such as the stability of near-surface qubits and quantum sensors, surface conductivity and electrochemical potential, to name just a few.^5–8 Hydrogen and oxygen termination represent important stable preparations of the surface, with each providing contrasting behavior in surface properties, including the electrical conductivity, electron affinity, and hydrophobicity.⁹ Hence, manipulation and control of the chemical state of the surface is of paramount practical importance in many diamond applications.

Methods available for manipulating the surface termination include chemical or plasma, mechanical polishing,¹⁰ heating, and laser processing, whereas processes for conversion between H and O termination are well established,^10,11 only a few enable the sub-monolayer (ML) precision required to controllably engineer the O and H surface coverage. Partial layer oxidation has been demonstrated, for example, by exposure of an H-terminated surface to ozone for 3–5 min, which provided an oxygen coverage of 0.1–0.2 ML.^12,13 Methods that are able to adequately induce an a priori specified partial fraction of the top layer are not yet established. Many applications require patterning involving lithography, which is time-consuming and may adversely affect the properties of the masked regions.

Though lasers provide a highly practical technique for direct-write patterning, they are not known for monolayer or sub-monolayer precision. Laser ablation is the primary established technique for optically machining wide-bandgap materials like diamond, in which case, a depth resolution of nanometers is possible, and, in the case of diamond, this is usually accompanied by sp² contamination.¹⁴ However, diamond has another direct optical material removal mechanism that occurs for laser fluences below the ablation threshold. This process, which appears to be distinctive to diamond, is most easily observed in the case of pulsed deep UV lasers in air^15,16 on all major crystal planes.¹⁷ The surface removal shows no threshold and is an atom-by-atom process with a per-pulse removal rate typically in the range 10⁻⁵–10⁻² ML/pulse.^15,16 Etched surfaces are clean, with no graphite formed on the surface, and are terminated with O functional groups.¹⁸ The process is suppressed under vacuum, suggesting that the primary action of the UV is a photochemical effect involving environmental oxygen or other oxidants such as water.^19,20 In contrast to ablation, the etching rate is a simple function of laser intensity, which, in the case of nanosecond deep UV laser pulses, increases with approximately the square of the laser intensity.^16,17,21 The atom-by-atom nature of the process enables sub-monolayer z-resolution,²² whereas the multi-photon dependence enables sub-diffraction limited lateral resolution.¹⁶ For multi-layer etching, emergent nano-structures form with orientation and morphology determined by laser polarization with respect to the crystallographic directions.^17,23 It has also recently been shown that etching leads to the formation of nitrogen-vacancies near the surface.²⁴ Thus, the features of this laser etch process offer a unique and versatile method for manipulating the chemical, morphological, and quantum properties of diamond surfaces. To date, the mechanism for the etch process has been assumed to rely on the ejection of CO functional groups from the surface^17,22,25 and has only been characterized for O-terminated diamond starting surfaces.

One prominent area that highlights the critical role of the surface monolayer is in transfer-doped conducting surfaces on diamond. H-terminated diamond surfaces have negative electron affinity, so that an electron exchange between the valence band maximum and empty conduction band of certain dopants (adsorbates) in contact with the surface results in the formation of a hole accumulation layer (2D hole gas) and a p-type surface conductivity.^8,26 This conduction route is attractive for the fabrication of surface electronic structures such as field effect transistors,^4,27–31 as it avoids the limitations inherent in impurity doping of diamond.^32–34 The role of O terminations and other defect sites among the otherwise H-terminated surface create potential barriers to carrier conductivity,³ an effect that has been used to enhance the stability of 2D hole-gas transistors used in sensing applications.¹² 

In this work, we show that the two-photon etching process oxidizes H-terminated (100) diamond surfaces with sub-monolayer precision, and we measure the etching rate for the top layer. Rather than using a direct measure of depth, such as an atomic force microscope that would have a large uncertainty for sub-nm depths, we use the conductivity as a function of UV dose to characterize the altered chemical state of the surface. Using doses calibrated for etch depths up to 1.0 ML, it is found that the surface resistance increases by over 7 orders of magnitude, and that the evolution of electrical properties agrees well with a percolation model for carrier transport. From the comparison of the model with the experiment, we deduce that the H-terminated surface etches at an almost identical rate to the O-terminated, revealing a highly practical method for converting the surface termination and manipulating the surface chemistry of diamond with unprecedented precision. The similar etch rates of the H- and O-terminated surfaces are unexpected and require adaption of current models for the carbon ejection mechanism to include H functional-group precursors.

We characterized the surface electrical properties for 10 laser-processed samples for calibrated doses up to 1.0 ML. The samples were (100)-oriented diamonds (Element Six, CVD optical grade) with less than 0.1 ppm nitrogen and RMS roughness of 1 nm. The RMS roughness of the samples was measured with an optical surface profiler with a 50× objective and a sampling area of 50 × 50 μm². The diamonds were cleaned using aqua regia and piranha solutions before H-termination.

H-termination was carried out using a plasma-enhanced chemical vapor deposition system from SEKI (AX 5250M). The microwave power, H₂ flow, substrate temperature, and chamber pressure were 900 W, 200  sccm, 600 °C, and 30 Torr, respectively, and the processing time was 15 min. The resistance and I–V characteristics were measured between contact pairs spaced by 15 μm (2.5 times beam diameters) laid down using a liftoff process. A negative resist (AZ nLof2020) was coated on the diamond surface, and the contact patterns were transformed onto the resist using UV lithography. The patterns were etched using AZ326 developer onto the resist prior to deposition of metallic contacts comprising of a stack of Pd, Ti, and Au layers of thicknesses 10, 20, and 100 nm, respectively. Finally, the unwanted metal areas were removed by immersing the sample in N-methyl-2-pyrrolidone.

An in-house built laser direct write system operating at the wavelength of 266 nm was used for laser etching. A Q-switched pulsed frequency-doubled Nd:YAG laser (Bright Solutions) at 532 nm with a repetition rate of 35 kHz and 3 W of maximum average power was used as input to a second-harmonic generator in order to produce 1.1 ns UV pulses. The laser polarization was aligned to the ⟨110⟩ crystal direction and focused using 10 cm focal length objective to a spot with a diameter of 6 μm.

The etch rate at the beam center was calibrated by etching multiple deep spots on an O-terminated surface, with etch time ranging from 1 to 20 s. The spot depth and profile were measured with an optical surface profiler (Veeco NT9800) operating in the phase-shifting interferometry mode using a 532 nm laser source and 50×/0.55 NA objective to give a lateral resolution of 500 nm and vertical resolution of 0.1 nm (≈1 ML), which easily resolved the micrometer-scale etched profile. The duration and laser fluence were selected to constrain the maximum depth to less than 600 nm to ensure the etch rate remains in a linear range.¹⁷ The per-pulse etch rate r was determined from the line of best fit to the data and is expressed in ML/pulse assuming each ML is 0.089 nm (the diamond lattice constant, 0.35 nm, divided by 4). The r values for two laser fluences of 0.5 and 2 J cm⁻² were 1.0 ± 0.1 × 10⁻³ and 1.0 ± 0.1 × 10⁻⁴ ML/pulse, respectively (Fig. 1).

Though we observe a lower ablation threshold for H-terminated surfaces when using much higher fluences than we use for etching (4 J cm⁻² compared to 16 J cm⁻² for O-terminated surfaces), we have found anecdotally that etching proceeds with the H-terminated surface without any incubation effects, providing a first indication that an H top layer etches at similar rate to an O-terminated layer. We start the investigation with the ansatz that the H-terminated layer etches at the same rate as the O-terminated layer and show below that this yields excellent agreement with the evolution of electrical properties as the top monolayer is laser etched.

Laser durations as short as 6 ms, limited by the speed of the mechanical shutter, provide minimum equivalent doses as small as 0.02 ML when using the fluence 0.5 J cm⁻². Even smaller doses can be readily delivered by using lower laser power settings. To provide lines of varying depth up to 1 ML, etches were performed by translating the sample with respect to a focused laser spot. For a translation speed of v, beam diameter D, and laser repetition rate f_rep, each location in the beam path is exposed to f_repD/v number of pulses so that the number of monolayers removed from the surface is $z = r f rep D / v$⁠. To obtain a range of z values, v was varied between 20 and 100 μm s⁻¹ for a fixed fluence.

Each electrode pair was electrically isolated and defined also using the laser etch direct-write process, but with the higher laser fluence (2 J cm⁻²) and a translation speed of 10 μm s⁻¹. These settings remove approximately 17 ML, ample to ensure complete electrical isolation of the resistors from the surrounding surface. The strip between the two contacts, as shown in the diagram of Fig. 2(a), was etched using a single laser pass with a fluence of 0.5 J cm⁻². The locations of the isolation and resistor tracks are indicated in the microscope image of Fig. 2(b). An initial surface resistance of 2.5 ± 0.5 k $Ω$ was measured for all samples. After laser exposure, measurements were carried out after a period of 48 h to ensure that changes in conductivity were not attributable to the short-term laser-induced removal of the atmospheric adsorbates essential to forming the 2D hole gas.

SEM images of the complete and partially removed H-termination layer are shown in Figs. 2(c)–2(e). For the SEM voltage used (secondary electron, 2 kV), the brightness of the SEM image provides some indication of the extent of surface oxidation. The deeply etched isolation tracks appear darker in the SEM image than the unetched regions, whereas lightly etched surfaces (0.2 ML of carbon removed) appear brighter [Fig. 2(d)]. As the number of carbon layers removed approaches 0.35 ML, darkened regions appear in the center of the etched zone [Fig. 2(e)] where the etch is deepest. This points to a transition in surface chemistry and conductivity associated with surface oxidization. Though there is a correspondence between the SEM image and the etch depth, the relationship is expected to be complex due to the detailed properties of surface charging and secondary electron emission on etch depth. Hence, it is unlikely that the width and uniformity of the SEM-imaged track closely reflects the etch profile.

The surface resistance as a function of z is shown in Fig. 3(a), measured using a 10 V bias. A small increase in resistance is observed for z values up to just above 0.2 ML, whereas for z > 0.3 ML, the exponential rise of the curve changes character and increases from 90 k $Ω$ to the limit of the measurement (1 × 10⁸ k $Ω$ at 0.50 ML). Around a transition point, shown as between regions I and II in Fig. 3(a), the I–V characteristics of the resistors also shift from linear (ohmic) to nonlinear when the resistance exceeds 100 k $Ω$ [Fig. 3(b)]. The examples shown in Fig. 3(b) show a linear characteristic is obtained for z = 0.26 ML at a resistance of 9 k $Ω$ and a nonlinear characteristic for z = 0.35 ML and 900 k $Ω$⁠.

The rapid change in electrical characteristics for z < 0.5 ML is analyzed through a percolation concept, which deals with connectivity in inhomogeneous systems.³⁵ For the present 2D-hole-gas system, spatially varying H-terminated (conducting) and O-terminated (insulating) regions are expected due to the irregular H termination and other surface inhomogeneities.^3,36 Following Refs. 37 and 38, a transition from classical geometric percolation to quantum tunneling percolation occurs when the fractional density of conducting regions, p, is lower than a critical value, p_c. For 2D systems, p_c has a universal value of 0.676. The transition from classical percolation to quantum percolation is shown conceptually in Fig. 4. The conductivity scales with $( p − p c ) 1.3$ and exp(− $γ p − 0.5$⁠) in each of these respective regimes, where γ is a fit parameter of magnitude approximately 10². By proposing that the density of conducting regions is equal to the H coverage (⁠ $p = 1 − z$⁠), fits to the resistance experimental data (using $p c = 0.7$ and γ = 68) on either side of the transition point at z = 0.3 are obtained as shown in Fig. 3(a). Good agreement is obtained apart from the point at z = 0, for which a higher resistance is obtained than the theoretical prediction. Nevertheless, the observed transition in the two curve fits is consistent with a transition from geometric to tunneling percolation near $1 − z = p c$⁠, while the I–V characteristics also confirm a transition between 0.65 and 0.74, which envelopes the p_c value. This good agreement between the experimental threshold and the universal value indicates that z is close to the oxygen coverage, and, therefore, the top-monolayer etch rate is similar to that for an O-terminated surface. Possible explanations for the poor agreement for the data point at z = 0 may be due to incomplete H-termination or surface roughness. We note that choosing a critical threshold slightly lower at $p c = 0.676$ [dashed line in Fig. 3(a)] though improving the fit at z = 0 does not reduce the overall fit error.

From the perspective of the mechanism for UV etching, the similar etch rates of the H- and O-terminated surfaces are surprising since the two-photon etching mechanisms proposed to date have been based on ejection of precursor carbonyl groups.^17,22,25 There is evidence, however, that other carbon–oxygen containing functional groups (such as the bridge-bonded ether) eject from the surface as a result of the observations that UV laser etching proceeds even for surfaces that clearly support a range of O functional groups. The results here show that the H-terminations are etched at a similar rate to the O-terminated surface. It seems unlikely that the UV pulses eject just H atoms without displacing a carbon since we would expect that the rates for carbon-containing molecules and H-atoms should not be the same. Though -CH ejection remains to be experimentally verified, the results here indicate that the carbon ejection rate is insensitive to the type of termination. The insensitivity of the etching to termination indicates that the action of the laser pulses is more directly coupled to the sub-surface states associated with carbon back bonds to the lattice.

In conclusion, we have demonstrated an in-air direct-write laser technique that enables the oxidation of H-terminated diamond surfaces with sub-monolayer precision (0.02 ML, exposure time limited). A resistance increase by over 7 orders of magnitude and a transition from linear ohmic to nonlinear resistance for etch doses corresponding to up to 0.5 ML are consistent with a model involving geometric and tunneling percolation. The H-terminated top monolayer is deduced to etch at a rate similar to O-terminated surface, a result that challenges current models for carbon ejection based on O functional groups. The results reveal a highly practical technique for manipulating the diamond surface chemistry with sub-monolayer precision that may benefit applications such as diamond electronics and diamond quantum science that rely sensitively on oxygen coverage.

This material is based on research sponsored by the Australian Research Council Grant (No. LP200301594) and the U.S. Air Force Office of Scientific Research Grant (No. FA2386-21-1-4030).