Engineering diamond interfaces free of dark spins

TL;DR

The study addresses the pervasive problem of surface dark spins limiting NV-based nanoscale sensing. It introduces TiO2 passivation via atomic layer deposition, revealing two growth regimes and a substantial reduction in surface spin density that enhances near-surface NV coherence. A 2D dark-spin bath model, combined with ab initio band-structure analysis of the diamond-TiO2 interface, provides a cohesive explanation for how passivation suppresses surface noise while maintaining functionalization potential. The approach is broadly transferable to other solid-state qubits and bio-sensing contexts, offering a scalable route to higher-sensitivity NV EPR/NMR measurements.

Abstract

Nitrogen-vacancy (NV) centers in diamond are extensively utilized as quantum sensors for imaging fields at the nanoscale. The ultra-high sensitivity of NV magnetometers has enabled the detection and spectroscopy of individual electron spins, with potentially far-reaching applications in condensed matter physics, spintronics, and molecular biology. However, the surfaces of these diamond sensors naturally contain electron spins, which create a background signal that can be hard to differentiate from the signal of the target spins. In this study, we develop a surface modification approach that eliminates the unwanted signal of these so-called dark electron spins. Our surface passivation technique, based on coating diamond surfaces with a thin titanium oxide (TiO$_2$) layer, reduces the dark spin density. The observed reduction in dark spin density aligns with our findings on the electronic structure of the diamond-TiO$_2$ interface. The reduction, from a typical value of $2,000$~$μ$m$^{-2}$ to a value below that set by the detection limit of our NV sensors ($200$~$μ$m$^{-2}$), results in a two-fold increase in Hahn-echo coherence time of near surface NV centers. Furthermore, we derive a comprehensive spin model that connects dark spin relaxation with NV coherence, providing additional insights into the mechanisms behind the observed spin dynamics. Our findings are directly transferable to other quantum platforms, including nanoscale solid state qubits and superconducting qubits.

Figures (23)

• Figure 1: Layout and characterization of ALD-coated TiO2 films on diamond. (a) Schematic of TiO_2 heterostructures on diamond with implanted near-surface NV centers. (b) Ellipsometry measurements of film growth with varying numbers of ALD cycles to determine the resulting film thicknesses. Data is fit to the Nilsen equation describing the delayed nucleation on the diamond surface allowing for the density of nucleation sites to be estimated. (c) Illustration of the stages of the island growth model for site selective nucleation of ALD TiO_2 on diamond surfaces. (d) XPS carbon 1s spectra for varying number of ALD cycles. After 300 cycles only adventitious carbon incorporated into the TiO_2 film is measureable. (e) XPS oxygen 1s spectra for varying number of ALD cycles. Peaks are normalized to the Ti(IV) peak.
• Figure 2: DEER spectroscopy measurement of surface spins. (a) Pulse sequence for DEER measurement. The phase of the last $\pi/2$ pulse is denoted as $\phi$ to indicate the $\pm x$ phases used for variance detection normalization. (b) DEER measurements for Sample 1. The surface spins are probed for a total free precession time $t = 1.6~\mu s$ and a surface spin $\pi$ pulse $\sim$ 120 ns is determined from DEER Rabi measurements. (c) DEER contrast and (d) NV Hahn-echo coherence versus the number of ALD growth cycles for all samples. Single-NV data (Sample 0) were acquired at a magnetic field $B\approx 400~\mathrm{G}$ with $t=4\,\mu\mathrm{s}$ for the DEER measurement, whereas ensemble data (Samples 1--3) were acquired at $B\approx 200~\mathrm{G}$ with $t=1.6\,\mu\mathrm{s}$.
• Figure 3: Dark spin relaxation characterization. (a) Pulse sequence for dark spin $T_1$ measurement including laser (green) and microwave (blue) pulses. Laser "P" ("R") block denotes polarization (readout) pulse. (b) Dark spin $T_1$ as a function of the number of TiO_2 ALD growth cycles for the three NV ensemble diamonds. For all measurements, $t = 1.6 \, \mu s$. (c) Time traces of the $T_1$ measurement for Sample 1 up to 75 ALD cycles. Past this coating, the $T_1$ for Sample 1 is undetectable. Black curves represent the fits of the form given in Appendix \ref{['SI:DarkSpinT1']}.
• Figure 4: Model fitting for normalized DEER coherence measurement. (a) Computed $\log( -\log(\langle F(t) \rangle ))$ curve dependencies based on changing values of dark spin relaxation rate ($\gamma$, top), dark spin density ($\sigma$, middle) and NV-bath separation ($d$, bottom). (b) Sample 1 and (c) Sample 2, $\log( -\log(\langle F(t) \rangle ))$ curves with model fits. Solid black lines indicate model fit assuming a 2D layer of spins at the diamond surface which agrees well with the data. Dark gray dashed line for Sample 2, 300 cycle data indicates model fit assuming a 3D layer of spins which poorly matches the curve profile. Light gray dotted lines indicate lines with slope 2 and 2/3 respectively. The curves are artificially offset vertically for clarity. (d) Extracted fit parameters ($\gamma$ top, $\sigma$ middle, $d$ bottom) for the three NV ensemble diamonds.
• Figure 5: (a) Surface model of diamond C (100) terminated with with C-O-C, C-H, C-O, and C-OH groups. (b). Heterostructure of C(100):TiO$_2$ (101). (c) Band alignment of diamond and TiO$_2$ from DFT. The $E_{CBM}$ (peach color) and $E_{VBM}$ (blue color) represents the conduction and valence band edges respectively. The Conduction band offset (CBO) is 2.35 eV, and the Valence band offset (VBO) is 0.25 eV.