O ptically active spin defects in wide-band-gap materials have shown great potential for a wide range of emerging technologies, from quantum information processing 1,2 to highresolution sensing of magnetic and electric fields. [3][4][5][6] Color centers in bulk semiconductors such as diamond 7,8 and silicon carbide 9,10 are prime examples that reveal optically detected magnetic resonance (ODMR). Recently, atomic defects in layered van der Waals materials such as hexagonal boron nitride (hBN) are attracting increasing attention as alternative candidates for studying light-matter interaction, nanophotonics, and nanoscale sensing. [11][12][13][14] Atomic defects in hBN are stable in nanosheets as thin as a monolayer and are readily accessible for device integration and top down nanofabrication. 15,16 Furthermore, recent experiments discovered that some defects in hBN could be spin addressable at room temperature. [17][18][19][20][21][22] The negatively charged boron vacancy (V B -) spin defect is the most studied one among these defects. [23][24][25] It has spin s = 1, and its orientation is out of plane. 17 V B -defects can be generated by neutron irradiation, 17,26 ion implantation, 27 femtosecond laser writing, 28 and electron irradiation. 29 Spin defects in thin hBN nanosheets will be useful for quantum sensing and spin optomechanics. 30,31 However, so far, the V B -spin defects have relatively low brightness and ODMR contrast, which limits their sensitivity in quantum sensing. 22 Here we report high-contrast plasmonic-enhanced shallow V B -spin defects in hBN nanosheets for quantum sensing. We fabricate a gold film coplanar microwave waveguide by photolithography to optimize the homogeneity and local intensity of the microwave for spin control. The hBN nanosheets with spin defects are transferred onto the microwave waveguide and are in contact with the gold surface for both plasmonic emission enhancement and spin control. The surface plasmons [32][33][34] provide broadband emission enhancement covering the wide range of the photoluminescence (PL) of V B -defects from 750 to 950 nm. 17 Our method does not require complex nanofabrication or cause adverse effects on quantum sensing. The microwave magnetic field generated by the gold waveguide is parallel to the surface and perpendicular to the orientation of V B -electron spins, which is crucial to achieve high ODMR contrast. With these, we find that the ODMR contrast can reach 46% at room temperature, which is an order of magnitude larger than the highest roomtemperature ODMR contrast of hBN spin defects reported in the previous work. 35 We also observe an up to 17-fold photoluminescence (PL) enhancement of V B -defects due to the gold film. We measure the spin initialization time to be around 100 ns. In addition, we study the laser power and microwave power dependence of the continuous-wave (CW) ODMR and optimize the magnetic field sensitivity to be about 8 T/ Hz μ . Finally, we perform coherent spin control and measure the spin-lattice relaxation time T 1 and the spin coherence time T 2 of V B -spin defects generated by He + ion implantation, which can benefit future works on multipulse sensing protocols. Our results demonstrate the promising potential of shallow hBN spin defects for nanoscale quantum sensing and other quantum technologies.

The results presented in this work are obtained with tapeexfoliated hBN nanosheets which are tens of nanometers thick. We exfoliate hBN flakes onto a Si wafer and then mount the wafer in a home-built ion implanter for doping. We use lowenergy He + ions (200 eV to 3 keV) to implant hBN nanosheets, which creates high-quality V B -defects with average depths ranging from 3 to 30 nm and avoids introducing undesired defects. 36 After ion implantation, the hBN flakes are transferred onto a gold microwave stripline on a sapphire substrate and characterized using an ODMR setup (Figure 1a). CW ODMR measurements are performed to acquire the basic spin properties of the V B -defects generated by He + ions. The defects are excited by a 532 nm laser with an NA = 0.9 objective lens, which also collects the PL of defects. We record integrated photon counts as a function of the applied microwave. We obtain the difference between the photon count rates when the microwave is off (I off ) and when the microwave is on (I on ). The ODMR contrast is then determined by normalizing the PL difference with the PL intensity when the microwave is off: C = (I off -I on )/I off . A positive ODMR contrast C means that the microwave driving decreases the PL intensity. The energy level structure of V B defects in hBN is shown in Figure 1b.

Figure 1c shows measured high-contrast ODMR spectra of V B -defects at room temperature. Strikingly, with strong microwave driving, we find that these defects can exhibit up to 46% ODMR contrast, which is 1 order of magnitude higher than the best contrast of hBN spin defects reported previously. 35 This value is even larger than the ODMR contrast of diamond nitrogen-vacancy centers at room temperature. 7 The ODMR contrast can readily reach around 20% without significant power broadening using low-power microwave driving. We also observe a similar high contrast with a thick hBN flake on the gold stripline which does not show PL enhancement (Supporting Information, Figure S7). It is noted that such a high ODMR contrast is due to the good alignment of the microwave magnetic driving field and the strong driving field strength. Since the hBN flakes are placed directly on the gold stripline, the driving magnetic field is parallel to its surface and perpendicular to the V B -spins, which yields the maximum ODMR contrast. In comparison, for a hBN flake placed inside a Ω-shaped waveguide, we observe a 6fold lower ODMR contrast with the same microwave power (Supporting Information, Figure S6). The decrease of the ODMR contrast is due to the misalignment of the driving microwave magnetic field, which is nearly out of plane inside a Ω-shaped ring.

Without an external magnetic field, the measured ODMR spectrum shows two resonances at ν 1 and ν 2 centered around ν 0 (Figure 1c). The results agree well with a double Lorentzian model. Here ν 0 is determined by the zero-field splitting (ZFS), ν 0 = D gs /h = 3.47 GHz, where h is the Planck constant. And the splitting between ν 1 and ν 2 is due to the nonzero off-axial ZFS parameter E gs /h = 50 MHz. 17 Figure 1d presents ODMR spectra in different external static magnetic fields. We use a permanent magnet to apply a static magnetic field perpendicular to the nanosheet surface. A translation stage is used to change the position of the magnet and tune the magnetic field strength. With an external static magnetic field B, ν 1 and ν 2 will be split further owing to the Zeeman effect,

, where g = 2 is the Landeǵ -factor. The splitting (ν 2 -ν 1 ) is 560 MHz (346 MHz) at 9.8 mT (5.9 mT).

It is highly desirable to create spin defects as close to the surface as possible without degrading the spin properties for nanoscale quantum sensing. This can decrease the ultimate distance between a sample and the sensor, which can significantly improve the signal. In Figure 2, we study the formation of shallow V B -defects by using He + ions. First, we use the Stopping and Range of Ions in Matter (SRIM) software to calculate the depths and densities of vacancies created with different ion energies from 200 eV to 3 keV (Figure 2a,b). The most probable depth is 3.5, 6.4, 15, and 25 nm when the He + ion energy is 300 eV, 600 eV, 1.5 keV, and 2.5 keV, respectively. Then, we perform the CW ODMR measurements on the samples with different doping depths. Here we use weak microwave driving to avoid the power broadening. Therefore, we can extract the nature linewidth of V B -defects (Figure 2c). All of the ODMR spectra display similar line widths as well as the contrasts, indicating the hBN spin properties are nearly the same at different doping depths.

For sensing applications, the PL brightness is an important factor that directly affects the sensitivity. Former theoretical studies indicate that the near-infrared optical transition of V B defects is not an electric-dipole-allowed transition and is hence relatively dark. 24,25 In this context, improving their brightness is a crucial task. Here we utilize surface plasmons of a metallic film [32][33][34] to enhance the brightness of the V B -. Surface plasmons are collective oscillations of coupled eletromagnetic waves and free electrons on metallic surfaces. They have large localized electric and magnetic fields which can speed up both radiative and non-radiative decays. We chose plasmonic enhancement because it can cover the whole broad PL spectral range of V B -defects. In addition, this method can utilize the metallic surface of our microwave waveguide and does not require complex nanofabrication. Our microwave waveguide is made of a 300 nm thick gold film prepared by electron-beam physical vapor deposition on top of a sapphire wafer. The width of the center microstrip is 50 μm. The hBN flakes with V B -defects are transferred onto both the gold film and the sapphire substrate for comparison (Figure 3a). V B -defects on both gold and sapphire surfaces display broad PL emission spectra around 810 nm (Figure 3b). Remarkably, the V B defects on the gold film show an order of magnitude higher PL intensity than those on the sapphire substrate under the same laser excitation. Parts c and d of Figure 3 present the PL intensities of these two samples and their PL ratio at different laser excitation powers. The experimental data in Figure 3c are fit to I = I sat /(1 + P sat /P laser ), where I is the PL intensity of the V B -, I sat is the saturation PL intensity, P laser is the excitation laser power, and P sat is the saturation laser power. On a gold film, the V B -defects show up to 17-fold enhancement of PL intensities under low-power excitation. We also observe a strong modification of the saturation behavior (Figure 3c). The laser saturation power P sat is reduced by a factor of 5, and the saturation PL count rate I sat is increased by around 3.5 times. These indicate that a gold film can improve the quantum efficiency of V B -defects significantly. We also study the effect of the separation between the gold film and V B -defects on the brightness enhancement. We transfer hBN flakes with different thickness onto the gold micostrip, so that we can get various distances between the V B defects and the gold film. The thickness of hBN nanosheets is measured by an atomic force microscope (AFM). The PL intensities are measured before and after transfer. Here we use 600 eV He + ions to generate shallow V B -with a most probable depth of 6.4 nm. Therefore, the average separation between the gold film and V B -defects is equal to the thickness of hBN nanosheets subtracted by 6.4 nm. As a result, we observe a strong thickness dependence of the PL enhancement. The highest enhancement is obtained when the hBN flake thickness is around 32 nm (Figure 3e). Our result is consistent with the former result on plasmonic enhancement of quantum dots on a gold surface. 33 When a hBN nanosheet is too thin, the nonradiative decay dominates. There is also little brightness enhancement when the hBN nanosheet is too thick because the surface plasmonic modes decay exponentially away from the surface. Thus, there is an optimal thickness for plasmonic enhancement. In addition, we characterize the electron spin initialization time of V B -defects on both gold and sapphire surfaces as a function of the power of the 532 nm excitation laser (see the Supporting Information for more details). We find that the required spin initialization time can be about 100 ns under high-power excitation (Figure 3f). Thus, the groundstate recovery time should be less than 100 ns, which is much shorter than the former theoretical prediction. 25 Our results may be helpful for future theoretical works on V B -spin defects. The gold film reduces the spin initialization time (Figure 3f), which also speeds up quantum sensing.

Sensitivity is the most important parameter to determine the performance of a sensor. To measure an external static magnetic field with spin defects, a common way is to use CW ODMR to detect the Zeeman shifts of the spin sublevels caused by the magnetic field. 5 The precision to determine the magnetic field is directly affected by the photon count rate R,

where μ B is the Bohr magneton. In this expression, A is a numerical parameter related to the specific line shape function.

For a Lorentzian profile, A ≈ 0.77. The values of C, Δν, and R are further related to the microwave power and laser power 37 (see the Supporting Information). To improve the detection sensitivity, it is crucial to increase the count rate and contrast as high as possible without significant power broadening of the linewidth.

Here we perform a group of CW ODMR measurements with various microwave powers and laser powers to find the

Letter optimal conditions for magnetic field sensing (Figure 4). Figure 4a presents three typical CW ODMR spectra at different microwave powers. Under low-power 40 mW microwave driving, we obtain an ODMR contrast of ∼10%. Such a low microwave power does not induce significant spectral power broadening. A natural linewidth can be extracted as ∼110 MHz. With an increasing microwave power, we first observe a significant improvement of the contrast without much spectral broadening. When we increase the microwave power further, the linewidth broadening becomes severe. Parts b-d of Figure 4 present the quantitative measurements of the ODMR contrast, linewidth, and sensitivity as functions of the microwave power. The experimental results fit well with theoretical models, as discussed in the Supporting Information. Here we perform the experiments at two different laser powers (1 and 5 mW). The 5 mW laser excitation gives a broader linewidth but a higher saturation ODMR contrast compared to those with 1 mW laser excitation. As a result, if we increase the microwave power, the sensitivity is first improved owing to the increase of the ODMR contrast and then becomes worse when the spectral power broadening dominates. The best sensitivity that we have achieved is about 8 T/ Hz μ . This sensitivity is 10 times better than the former result with hBN spin defects (our system also has a better spatial resolution). 22 This sensitivity will be enough for studying many interesting phenomena in magnetic materials. For example, the magnetic field generated by a monolayer CrI 3 (a 2D van der Waals magnet) is on the order of 200 μT. 6 Finally, we perform pulsed ODMR measurements to determine the spin-lattice relaxation time T 1 and the spin coherence time T 2 of the shallow V B -defects generated by ion implantation. T 1 and T 2 of hBN spin defects have only been measured for neutron irradiated samples before. 19,26 It will be useful to know their values for our shallow spin defects created by ion implantation. In addition, pulsed ODMR measurements are inevitable steps for realizing more complex sensing protocols. A pulsed ODMR measurement consists of optical initialization of the ground state, coherent manipulation of the spin state with microwave pulses, and optical readout of the final spin state. Here we add an external magnetic field of 13 mT to split two branches of the V B -spin sublevels. m s = 1, 0 states are used as the two-level spin system to carry out the spin coherent control. Figure 5a shows the Rabi oscillation as a function of the microwave power. The data is fit using

. We observe an oscillation with two Rabi frequencies, which are in the tens of megahertz range. The exponential decay of one oscillation component gives the spin-dephasing time of T 2 * = 120 ns. To gain more insight into the spin properties of the V B -defects at different depths, we measure the spin-lattice relaxation times T 1 and spin-spin relaxation times T 2 of the V B -defects created with different ion implantation energies. The pulse sequences are shown as insets in the left panels of Figure 5b,c. By fitting the results, T 1 and T 2 are obtained as ∼17 and ∼1.1 μs, respectively. As shown in Figure 5b,c, both T 1 and T 2 are independent of the ion energy, indicating the spin properties of the V B -defects are nearly independent of the depth. The depth-independent T 2 suggests that the coherence times are not limited by the surface. They are likely dominated by nuclear spin noise but may also be limited by the inhomogeneity, as we use an ensemble of spin defects instead of a single spin defect. 38 In conclusion, we have realized a significant improvement of the ODMR contrast and the brightness of hBN V B -spin defects. We observe a record-high ODMR contrast of 46%, which is 1 order of magnitude higher than the best former result with hBN. We use low-energy He + ion implantation to create V B -defects as shallow as 3 nm in hBN nanosheets. Moreover, both CW and pulsed ODMR measurements display that their spin properties are nearly independent of the ion implantation energy. This result confirms the feasibility to create high-quality V B -defects proximal to the hBN surface. In addition, we utilize the gold film surface plasmon to enhance the brightness of V B -defects and obtain an up to 17-fold enhancement of the PL intensity. We also explore the effects of laser power and microwave power on the CW ODMR contrast and linewidth. With these, we achieve a CW ODMR sensitivity around 8 T/ Hz μ . We expect that the PL can be enhanced further with plasmonic nanoantennas, 39 and the magnetic field sensitivity can be improved by using more powerful pulsed sensing protocols. In addition, during the preparation of this manuscript, we became aware of a related work 40 that reported a PL enhancement of hBN spin defects with a photonic cavity by a factor of ∼6 at high NA collection and an ODMR contrast of about 5%. The surface plasmon may be combined with a photonic cavity to obtain better results in the future. Our work strongly supports the promising potential of V B -defects as a nanoscale sensor in a 2D material platform. F Afterward, the PL was coupled into a 125-um single-mode fiber, and guided to a single photon counter (Excelitas, SPCM-AQRH) or a spectrometer (Andor shamrock SR-500).

Optically Detected Magnetic Resonance Measurements: ODMR measurements were performed using the confocal system described above. A gold microwave coplanar waveguide (Figure S1) was patterned using standard photolithography techniques and electron beam evaporation of 300 nm Au on sapphire. A 10 nm Ti layer is deposited on sapphire before Au deposition. The root-mean-square (rms) roughness of the gold film is measured to be 1.

Experimentally, we have successfully generated V - B defects with an ion energy as low as 200 eV, corresponding to a depth around 2.5 nm. When the dose density is fixed at 5×10 13 cm -2 for implantation, the PL intensity increases when we increase the ion energy (Fig. S2(a)), implying that ions with higher energies have a larger probability to break the B-N bonds and create V - B defects.

In addition, we study the dependence of the PL intensity on the He + dose. We fix the ion energy and tune the implantation duration to change the dose (Fig. S2(b)). The PL intensity first increases with the increasing implantation dose, and then becomes saturated at a high dose. The implantation with 600 eV He + ions has a lower saturation dose and PL intensity than those with 2.5 keV He + ions. This is because the implantation depth of low-energy ions is smaller than that of high-energy ions.

(a) (b)

We also characterize the electron spin initialization time as a function of the power of the 532 nm excitation laser. Figure S3 presents time-resolved photon count rates beginning with different spin states. The initial state is a mixture state at thermal equilibrium when the laser has been off for a long time (much longer than T 1 ). When the laser pulse is turned on, the electron spins are getting polarized and the photon count rate increases. To prepare the polarized state as a reference, we apply a 3-µs-length laser pulse to polarize electron spins to the m s = 0 state. The laser is turned off for 100 ns (which is much shorter than T 1 ) before being turned on again for recording the reference photon count rate. The difference between the count rates of these two states is calculated and then fitted exponentially to determine the spin initialization time. We find that the required spin initialization time is on the order of 100 ns. The gold film reduces the spin initialization time, which also speeds up quantum sensing.

Figure S3: Time-resolved photon counts starting from different spin states. The laser power is set as 200 µW.

In the main text, we reported the PL enhancement due to the gold film. Here we compare the brightness of the V - B on different substrates (Fig. S4). We use the same photolithography method to fabricate a 300-nm thick silver microstrip on top of the sapphire wafer. Then we transfer the ion implanted hBN flakes onto both the silver and sapphire substrates. On the silver film, we also observed a comparable brightness enhancement (up to 16 times) as the gold film. Besides, we also compare the brightness of V - B on SiO 2 and indium tin oxide (ITO). No brightness enhancement is observed on either SiO 2 or ITO substrates.

Fit to the ODMR lineshape: Here we use a double Lorentzian model to fit the ODMR spectra, and determine the ODMR contrasts and lineshapes. The intensity of the ODMR spectrum as a function of the microwave frequency ν can be written as 2

where R is the photon count rate, C i (i=1,2) is the ODMR contrast associated to the dip of the PL intensity, ν i (i=1,2) is the associated central resonance frequency and ∆ν i (i=1,2)

is the associated linewidth (FWHM). An external magnetic field will induce a shift of the central frequency ν i (i=1,2) through the Zeeman effect.

To measure an external dc magnetic field, we can directly evaluate the Zeeman shifts in the CW ODMR spectrum. The sensitivity in this is determined by the PL intensity, contrast and linewidth, which reads as

where h is the Planck constant, g is the Landé g-factor and µ B is the Bohr magneton. In this expression, A is a numerical parameter related to the specific lineshape function. For a

Lorentzian profile that we used above, A ≈ 0.77.

The effect of the microwave power on ODMR contrasts and linewidths: In the main text, we show the microwave power P M W dependence of the ODMR contrast and linewidth. We fit the ODMR signal to a double lorentzian model at different microwave powers P M W . Then the contrasts and linewidths are fit to 3 The effect of the laser power on ODMR contrasts and linewidths: Here we also perform a group of ODMR experiments to determine the optimal laser power for getting a high magnetic-field sensitivity (Fig. S5). When we increase the laser power from 200 µW to 10 mW, the ODMR contrast first increases and then decreases after reaching the maxima.

Meanwhile, the ODMR linewidth keeps increasing. According to Eq. 2, the sensitivity is proportional to ∆ν C √ R . Thus, besides the contrast and linewidth, the excitation laser power affects the sensitivity also through the PL count rate R. When we increase the laser power, the PL intesity also significantly increases before reaching the saturation power. As a result, we find that 5 mW laser in our case is around the optimal condition.

The effect of a different microwave alignment: We also designed a Ω-shaped microwave waveguide to apply the microwave (Figure S6(a)). The magnetic field of the driving microwave is nearly perpendicular to the hBN flake at the center of the Ω ring. Under 1 mW laser excitation and 2.5 W microwave driving, the CW ODMR contrast is around 7.5%, which is about 6 times smaller than what we observed when microwave field is aligned correctly (e.g. when the hBN flake is on a straight-stripe waveguide in Fig. S1).