A robust, scanning quantum system for nanoscale sensing and imaging
P. Maletinsky, S. Hong, M.S. Grinolds, B. Hausmann, M.D.Lukin, R.-L. Walsworth, M. Loncar, A. Yacoby
AA robust, scanning quantum system for nanoscale sensing and imaging
P. Maletinsky ∗ , S. Hong ∗ , M.S. Grinolds ∗ , B. Hausmann, M.D.Lukin, R.-L. Walsworth,
1, 3
M. Loncar, and A. Yacoby Department of Physics, Harvard University, Cambridge, Massachusetts 02138 USA † School of Engineering and Applied Science, Harvard University, Cambridge, Massachusetts, 02138 USA Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138 USA (Dated: May 23, 2018)Controllable atomic-scale quantum systems hold great potential as sensitive tools for nanoscaleimaging and metrology [1–6]. Possible applications range from nanoscale electric [7] and magneticfield sensing [4–6, 8] to single photon microscopy [1, 2], quantum information processing [9], andbioimaging [10]. At the heart of such schemes is the ability to scan and accurately position a robustsensor within a few nanometers of a sample of interest, while preserving the sensor’s quantumcoherence and readout fidelity. These combined requirements remain a challenge for all existingapproaches that rely on direct grafting of individual solid state quantum systems [4, 11, 12] or singlemolecules [2] onto scanning-probe tips. Here, we demonstrate the fabrication and room temperatureoperation of a robust and isolated atomic-scale quantum sensor for scanning probe microscopy.Specifically, we employ a high-purity, single-crystalline diamond nanopillar probe containing a singleNitrogen-Vacancy (NV) color center. We illustrate the versatility and performance of our scanningNV sensor by conducting quantitative nanoscale magnetic field imaging and near-field single-photonfluorescence quenching microscopy. In both cases, we obtain imaging resolution in the range of 20 nmand sensitivity unprecedented in scanning quantum probe microscopy.
The NV center in diamond is a point-defect that of-fers the potential for sensing and imaging with atomicscale resolution. Sensitive nanoscale detection of vari-ous physical quantities is possible because the NV cen-ter forms a bright and stable single photon source [13]for optical imaging, and possesses a spin-triplet groundstate which offers excellent magnetic [5] and electric [7]field sensing capabilities. The remarkable performance ofthe NV center in such spin-based sensing schemes, is theresult of the long NV spin coherence time [14], combinedwith efficient optical spin preparation and readout [15],all at room temperature. In addition, NV centers can bepositioned within nanometers of a diamond surface [16]and therefore in close proximity of a sample to maxi-mize signal strengths and spatial resolution. In order torealize the full potential of these attractive features, wehave developed a ”scanning NV sensor” (Fig. 1a), whichemploys a diamond nanopillar as the scanning probe,with an individual NV center artificially created withina few nanometers of the pillar tip through ion implan-tation. Long NV spin coherence times ( ≈ µ s) areachieved as our devices are fabricated from high purity,single-crystalline bulk diamond [17]. Furthermore, dia-mond nanopillars are efficient waveguides for the NV flu-orescence band [18], which yields record-high NV signalcollection efficiencies for a scanning NV device.Fig. 1b shows a representative scanning electron micro-scope (SEM) image of a single-crystalline diamond scan-ning probe containing a single NV center. The prepara-tion of such devices is based on recently developed tech- ∗ These authors contributed equally to this work niques in diamond nanofabrication [19], combined withestablished methods for controlled NV creation throughion implantation [20]. Our scanning diamond nanopil-lars have typical diameter ≈
200 nm and length of 1 µ mand are fabricated on few-micron sized diamond plat-forms which can be attached to atomic force microscope(AFM) tips for scanning (Fig. 1b and methods). Ourfabrication procedure (Fig. 1c) allows for highly parallelprocessing as shown in the array of diamond devices de-picted in the SEM image in Fig. 1d. From this array,we select nanopillars that contain single NV centers withhigh photon count rates and long spin coherence timesand mount these single-NV nanopillars onto AFM tipsto yield the finalized scanning probe shown in the SEMpicture in Fig. 1b.To employ the scanning NV sensor and characterizeits basic spin and optical properties, we used a com-bined confocal- and atomic force-microscope as sketchedin Fig. 1a. The setup is equipped with piezo positionersfor the sample and AFM-probe to allow for independentscanning with respect to the optical axis. Optical ad-dressing and readout of the NV center in the tip is per-formed through a long working-distance microscope ob-jective (numerical aperture, NA= 0 . ≈ .
65 [19]), yielding ahigh collection efficiency even with low NA collection-optics.Fig. 2a shows a confocal scan under green laser illu-mination ( λ exc = 532 nm) of a typical single NV/AFMdevice. The bright photon emission emerging from the a r X i v : . [ c ond - m a t . m e s - h a ll ] A ug AFM opticaladressingdiamond-probe samplestage a mask deposition + top-etchdiamondSiO mask mask deposition + bottom-etchsingle NV 10 (cid:109) mNV b c d ~ m ~2 (cid:109) m (cid:52)(cid:109) m (cid:53)(cid:109) m (cid:53)(cid:109) mdiamond nanopillar (cid:53)(cid:109) m (cid:49)(cid:48)(cid:109) m FIG. 1:
Experimental setup and probe fabrication forthe scanning NV sensor . (a) Schematic of the setup con-sisting of a combined optical and atomic force microscope(AFM). We use a 532 nm laser (green arrows) to address thescanning NV center through its red fluorescence (red arrows).The scanning NV center resides in a diamond nanopillar (in-set) and its proximity to the sample is maintained throughAFM feedback. (b) Scanning electron microscope (SEM) im-age of a single-crystalline diamond nanopillar-probe (false-color coded in red) with a single NV center in its tip (seeFig. 2). (c) Brief depiction of the fabrication process for scan-ning single-crystalline diamond NV sensors. Electron-beamlithography is used to define nanopillars and platforms fromthe top- and bottom-sides of a few micron thin diamond mem-brane. Patterns are then transferred to the diamond by re-active ion etching. (d) SEM image of a finalized array ofdiamond platforms with nanopillars. In all panels, dottedrectangles highlight diamond nanopillars. nanopillar (white circle) originates from a single NV cen-ter, as evidenced by the pronounced dip in the photon-autocorrelation measurement (Fig. 2b) and the charac-teristic signature of optically detected NV electron-spinresonance (ESR) [15] (Fig. 2c), all obtained on differentdevices. Importantly, we found that photon waveguidingthrough the nanopillar dramatically increases excitationand detection efficiencies for NV fluorescence [18]. Forsome devices, maximal NV fluorescence count-rates ex-ceeding 3 · counts per second (cps) were observed forexcitation powers as low as 20 µ W. We thus significantlyincrease fluorescence signal-strength from the single scan-ning NV and at the same time minimize exposure of sam-ples to green excitation light, which is especially relevantfor possible biological or low-temperature applications ofthe scanning NV sensor.Using well-established techniques for coherent NV-spin-manipulation [21], we characterized the spin-coherence time, T , of a single NV center in a diamondnanopillar. Spin-coherence sets the NV sensitivity tomagnetic fields and limits the number of coherent op-erations that can be performed on an NV spin; it is µ mI/I
01 Single emitter τ (ns)0 200 400Single emitter g ( τ ) I/I γ NV B ext
20 40 60 800.40.60.81 τ ( µ s) P ( m s = ) T =33 µ s a bc d FIG. 2:
Single NV centers in scanning diamondnanopillars . (a) Confocal image of red fluorescence froma single-crystalline diamond probe (see side view SEM im-age in 1b). Fluorescence counts are normalized to I =1 . · cps. The encircled bright feature stems from flu-orescence of a single NV center in a nanopillar. (b) Photonautocorrelation measurement (g ( τ )) for NV fluorescence in ascanning nanopillar device. Data with g < . γ NV B NV , where γ NV = 2 . B NV is the magnetic field along the NV axis (here, B NV ≈
16 G).(d) Spin-echo measurement for an NV center on a diamondnanopillar device. The envelope fitted to the characteris-tic NV spin-echo decay (see methods) yields the NV spin-coherence time of T = 33 µ s. Data in panels a-d was eachtaken on different devices with similar properties. therefore an essential figure of merit for applications inmagnetic field imaging [6] and quantum information pro-cessing [9]. Using a Hahn-echo pulse sequence, we mea-sured the characteristic single NV coherence decay [22]shown in Fig. 2d; and from the decay envelope we deducea spin-coherence time of T = 33 µ s. We note that thisT time is consistent [6] with the density of implantedNitrogen ions (3 · − cm − ) and conclude that ourdevice fabrication procedure fully preserves NV spin co-herence. Combining measurements of the T -time withthe fluorescence count-rate and spin-readout contrast ofthe single NV in Fig. 2d, yields an AC magnetic field sen-sitivity [6] of 170 nT/ √ Hz.To demonstrate the resolving power of the scanningNV sensor in magnetic imaging [4, 5], we imaged ananoscale magnetic memory medium consisting of bit-tracks of alternating (out-of-plane) magnetization withvarious bit-sizes. Fig. 3 illustrates our method and re-sults: the scanning NV sensor operated in a modethat imaged contours of constant magnetic field strength( B NV ) along the NV axis through the continuous moni-toring of red NV fluorescence, in the presence of an ESRdriving field of fixed frequency ω RF . We detuned ω RF by δ RF from the bare NV spin transition-frequency, ω NV ,but local magnetic fields due to the sample changed thisdetuning during image acquisition. In particular, whenlocal fields brought the NV’s spin-transition into reso-nance with ω RF , we observed a drop in NV fluorescencerate, which in the image yielded a contour of constant B NV = δ RF /γ NV , with γ NV = 2 . ω NV with δ RF = ±
10 MHz (dark andbright arrows in Fig. 3b). Normalization of the pixel-values in the two data-sets then directly provided a mapof magnetic field contours with positive and negative val-ues of B NV (here, with B NV = ± ≈ ω RF,1 ω RF,2
RF frequency (GHz)
I/I ( % / d i v ) ω RF,1 ω RF,2 I RF,1 /I RF,2 (%): 90 110200 nm 200 nmmodel100 nm 10 nm x" I R F , /I R F , ( % / d i v )
10 nm a bc de
FIG. 3:
Nanoscale magnetic field imaging with thescanning NV sensor . (a) NV magnetic field image oftracks on a magnetic harddrive, highlighted by dashed whitelines added to image. The image shows normalized data I RF , /I RF , (see (b)) and thereby reveals magnetic field linescorresponding to B NV = ± . ω RF , = 2 .
766 GHzand ω RF , = 2 .
786 GHz) and collect NV fluorescence counts( I RF , and I RF , , respectively) in synchrony with the RF mod-ulation. (c) Magnetic image obtained with the same methodas in (a), but with a decreased NV-sample distance. Bringingthe NV closer to the sample increases the magnetic field mag-nitude at the NV sensor, and improves the imaging spatialresolution, allowing imaging of ≈
30 nm magnetic bits. (d)Calculated NV response for the experimental situation in (a),assuming a simplified magnetic sample (see SupplementaryInformation) (e) Linecut along the white line indicated in (c)(averaged across 6 adjacent pixels). The scanning NV sen-sor’s ability to spatially resolve magnetic fields is limited bythe local magnetic field gradient, which for the present systemrealization and magnetic harddrive sample leads to a resolu-tion ≈ as such can be utilized for near-field optical imagingbeyond the diffraction and shot-noise limits [2], as wellas for scanning “F¨orster resonance energy transfer” mi-croscopy [1]. We used our scanning NV sensor to demon-strate such near-field optical imaging by performing scan-ning fluorescence-quenching imaging (FQI) on nanoscalemetallic objects. Imaging contrast consisted of the de-tected NV fluorescence in the far-field changing when theNV was in proximity to a metallic object [24]. I/I (%): 80 120400 nm sample−z I/I ∞ ( % / d i v . )
500 nm I/I (%): 84 200100 nm 10 nm a bcd e FIG. 4:
Nanoscale fluorescence quenching imaging(FQI) using the scanning NV sensor . (a) FQI of a sharpmetallic tip on a dielectric substrate. Scanning the diamondpillar over a sharp tip leads to a bright, circular feature dueto sample-topography (see Supplementary Information). Po-sitioning the metallic tip exactly at the location of the NVcenter (red square), however, yields a sharp dip in NV flu-orescence. (b) Zoomed-in image of the red square region in(a); the observed fluorescence quenching dip demonstrates anFQI spatial resolution ≈ ≈
100 nm (horizontal (vertical)tick-spacing is 200 nm (10 %)). (e) Comparison of the NVfluorescence rate as a function of NV-sample distance on (red)and off (blue) a nanowire (see red and blue dots in FQI im-age). The data demonstrates that imaging contrast in FQI isacquired in the optical near-field of the NV and therefore en-ables a breaking of the diffraction-limit. Red and blue curvesare offset by 40 % for clarity. Insets in a and d illustrate theexperimental configuration in FQI.
We determined the scanning-NV FQI spatial resolutionby measuring the point spread function (PSF) of our sin-gle photon microscope. To that end, we fabricated metal-lic tips with <
20 nm diameter (see schematic in Fig. 9aand Supplementary Information), which we imaged bymonitoring the total NV fluorescence rate as we scannedthe sample in direct contact with the pillar (Fig. 9a).The resulting data show signatures of the topographyof the scanning diamond nanopillar (bright ring in theNV fluorescence signal, see Supplementary Informationfor details). More importantly, however, while the sharpmetallic tip scanned the front-end of the diamond probe,we observed a pronounced dip in NV fluorescence (red square in Fig, 9a and zoomed image in Fig. 9b) when themetallic tip was positioned at the location of the NV cen-ter. The Gaussian width (double standard deviation) of25 . ≈
100 nm. Wenote that such sub-diffraction optical imaging is feasiblebecause the scanning NV center forms an atomic-scalelight-source [11], whose optical near-field contains spa-tial frequencies exceeding the inverse optical wavelength.This near-field coupling can be observed by monitoringNV fluorescence intensity as a function of the distancebetween an FQI sample and the NV. Fig. 9e shows a com-parison of such fluorescence “approach-curves” on and offa nanowire, which demonstrates that the imaging signalis acquired within ≈
100 nm of the sample surface (greyshaded area).For all imaging applications demonstrated in this pa-per, spatial resolution is limited by NV-to-sample dis-tance. The biggest uncertainty to this distance is ver-tical straggle in the NV implantation process, which isstill poorly understood [23]. Advances in the controlledcreation of NV centers close to diamond surfaces shouldenable production of stable NV centers as close as 3 nmfrom the nanopillar’s tip [16]. Spatial resolution for scan-ning NV imaging could therefore be further improved byabout one order of magnitude. We note that for mag-netic field imaging, our current ability to resolve indi-vidual magnetic domains already equals the typical per-formance of alternative methods [25, 26], with the addedadvantages of being non-invasive and quantitative.The magnetic field sensitivity we demonstrated withthe scanning NV sensor compares well to the performancerealized previously with single NV centers in ultrapure,bulk diamond samples [5]. Combined with the mechani-cal robustness and durability of our diamond probes (upto several weeks of scanning with the same tip), our re-sults constitute a significant advance in scanning quan-tum probe microscopy and demonstrate the advantage ofour method over alternative approaches [4, 12].The scope of applications of the scanning quantumprobe described here goes far beyond imaging and sens-ing. For example, our nanoscopic, scannable single pho-ton source, could be used to controllably inject singleplasmonic excitations into nanometallic structures [27] atwell-defined locations, which would have broad impactto the field of nano-plasmonics. Additionally, our deviceforms an ideal platform to coherently couple the scan-ning NV spin to other spin systems such as P in Si [28],other NV centers, or carbon-based spin qubits [29], eitherby optical or magnetic coupling. Quantum informationcould thereby be transferred between a stationary qubitand our scanning NV center and from there to other qubitsystems or single photons [30].We thank B. D. Terris and N. Supper from HitachiGST for providing the magnetic recording samples andT.G. Tiecke and J.D. Thompson for providing the sil-ver nanowire sample imaged in our experiments. P.M.acknowledges support from the Swiss National ScienceFoundation, S.H. thanks the Kwanjeong ScholarshipFoundation for funding and M.S.G. is supported throughfellowships from the Department of Defense (NDSEGprogram) and the NSF. This work was supported byNIST and DARPA and in part performed at the Centerfor Nanoscale Systems (CNS), a member of the NationalNanotechnology Infrastructure Network (NNIN), whichis supported by the National Science Foundation underNSF award no.
ECS − † [email protected][1] S. Sekatskii and V. Letokhov, JETP Letters , 319(1996).[2] J. Michaelis, C. Hettich, J. Mlynek, and V. Sandoghdar,Nature , 325 (2000).[3] B. M. Chernobrod and G. P. Berman, J. Appl. Phys. ,014903 (2005).[4] G. Balasubramanian, I. Chan, R. Kolesov, M. Al-Hmoud,J. Tisler, C. Shin, C. Kim, A. Wojcik, A. Hemmer,P.R. amd Krueger, T. Hanke, A. Leitenstorfer, R. Brats-chitsch, F. Jelezko, and J. Wrachtrup, Nature , 606(2008).[5] J. R. Maze, P. L. Stanwix, J. S. Hodges, S. Hong, J. M.Taylor, P. Cappellaro, L. Jiang, M. V. Gurudev Dutt,E. Togan, A. S. Zibrov, A. Yacoby, R. L. Walsworth,and M. D. Lukin, Nature , 644 (2008).[6] J. Taylor, P. Cappellaro, L. Childress, L. Jiang, D. Bud-ker, P. Hemmer, A. Yacoby, R. Walsworth, andM. Lukin, Nature Physics , 810 (2008).[7] F. Dolde, H. Fedder, M. W. Doherty, T. N¨obauer,F. Rempp, G. Balasubramanian, T. Wolf, F. Reinhard,L. C. L. Hollenberg, F. Jelezko, and J. Wrachtrup, Na-ture Physics , 459 (2011).[8] C. L. Degen, Appl. Phys. Lett. , 243111 (2008).[9] P. Neumann, R. Kolesov, B. Naydenov, J. Beck,F. Rempp, M. Steiner, V. Jacques, G. Balasubramanian,M. L. Markham, D. J. Twitchen, S. Pezzagna, J. Meijer,J. Twamley, F. Jelezko, and J. Wrachtrup, Nat Phys ,249 (2010).[10] L. P. McGuinness, Y. Yan, A. Stacey, D. A. Simp-son, L. T. Hall, D. Maclaurin, S. Prawer, P. Mulvaney,J. Wrachtrup, F. Caruso, R. E. Scholten, and L. C. L.Hollenberg, Nat Nano , 358 (2011). [11] S. Kuhn, C. Hettich, C. Schmitt, J. Poizat, and V. San-doghdar, Journal of Microscopy , 2 (2001).[12] A. Cuche, A. Drezet, J.-F. Roch, F. Treussart, andS. Huant, Journal of Nanophotonics , 043506 (2010).[13] C. Kurtsiefer, S. Mayer, P. Zarda, and H. Weinfurter,Phys. Rev. Lett. , 290 (2000).[14] G. Balasubramanian, P. Neumann, D. Twitchen,M. Markham, R. Kolesov, N. Mizuochi, J. Isoya,J. Achard, J. Beck, J. Tissler, V. Jacques, P. Hemmer,F. Jelezko, and J. Wrachtrup, Nature Materials , 383(2009).[15] A. Gruber, A. Drabenstedt, C. Tietz, L. Fleury,J. Wrachtrup, and C. Borczyskowski, Science , 2012(1997).[16] C. Bradac, T. Gaebel, N. Naidoo, M. J. Sellars,J. Twamley, L. J. Brown, A. S. Barnard, T. Plakhotnik,A. V. Zvyagin, and J. R. Rabeau, Nature Nanotechnol-ogy , 345 (2010).[17] P. L. Stanwix, L. M. Pham, J. R. Maze, D. Le Sage, T. K.Yeung, P. Cappellaro, P. R. Hemmer, A. Yacoby, M. D.Lukin, and R. L. Walsworth, Phys. Rev. B , 201201(2010).[18] T. M. Babinec, B. J. M. Hausmann, M. Khan, Y. Zhang,J. R. Maze, P. R. Hemmer, and M. Loncar, Nature Nan-otechnology , 195 (2010).[19] B. J. Hausmann, M. Khan, Y. Zhang, T. M. Babinec,K. Martinick, M. McCutcheon, P. R. Hemmer, andM. Loncar, Diamond and Related Materials , 621(2010), proceedings of Diamond 2009, The 20th Euro-pean Conference on Diamond, Diamond-Like Materials,Carbon Nanotubes and Nitrides, Part 1.[20] R. Kalish, C. Uzan-Saguy, B. Philosoph, V. Richter,J. P. Lagrange, E. Gheeraert, A. Deneuville, and A. T.Collins, Diamond and Related Materials , 516 (1997).[21] F. Jelezko, T. Gaebel, I. Popa, A. Gruber, andJ. Wrachtrup, Phys. Rev. Lett. , 076401 (2004).[22] E. V. Oort and M. Glasbeek, Chemical Physics , 131(1990).[23] M. S. Grinolds, P. Maletinsky, S. Hong, M. D. Lukin,R. L. Walsworth, and A. Yacoby, Nature Physics , 0(2011).[24] B. C. Buchler, T. Kalkbrenner, C. Hettich, and V. San-doghdar, Phys. Rev. Lett. , 063003 (2005).[25] F. Wolny, T. M¨uhl, U. Weissker, K. Lipert, J. Schu-mann, A. Leonhardt, and B. Bchner, Nanotechnology , 435501 (2010).[26] T. Kohashi, M. Konoto, and K. Koike, Journal of Elec-tron Microscopy , 1 (2009).[27] O. Mollet, A. Cuche, A. Drezet, and S. Huant, Diamondand Related Materials , 995 (2011).[28] B. E. Kane, Nature , 133 (1998).[29] P. Recher and B. Trauzettel, Nanotechnology , 302001(2010).[30] E. Togan, Y. Chu, A. S. Trifonov, L. Jiang, J. Maze,L. Childress, A. S. Dutt, M. V. G. AU Sorensen, P. R.Hemmer, A. S. Zibrov, and M. D. Lukin, Nature ,730 (2010).[31] C. Lee, E. Gu, M. Dawson, I. Friel, and G. Scarsbrook,Diamond and Related Materials , 1292 (2008).[32] L. Childress, M. Gurudev Dutt, J. Taylor, A. Zibrov,F. Jelezko, J. Wrachtrup, P. Hemmer, and M. Lukin,Science , 281 (2006).[33] R. J. Epstein, F. M. Mendoza, Y. K. Kato, and D. D.Awschalom, Nat Phys , 94 (2005). [34] T. Gaebel, M. Domhan, I. Popa, C. Wittmann, P. Neu-mann, F. Jelezko, J. Rabeau, N. Stavrias, A. Greentree,S. Prawer, et al. , Nature Physics , 408 (2006).[35] G. D. Fuchs, V. V. Dobrovitski, R. Hanson, A. Batra,C. D. Weis, T. Schenkel, and D. D. Awschalom, Phys.Rev. Lett. , 117601 (2008).[36] J. F. Ziegler, The Stopping and Range of Ions in Matter (2010).[37] D. M. Toyli, C. D. Weis, G. D. Fuchs, T. Schenkel, andD. D. Awschalom, Nano Letters , 3168 (2010).[38] K. Karrai and I. Tiemann, Phys. Rev. B , 13174(2000).[39] K. H. Drexhage (Elsevier, 1974) pp. 163 – 192, 192a,193–232. MethodsDiamond tip fabrication.
Devices were fabricated from a sample of high pu-rity, single crystalline diamond (Element Six, electronicgrade) of 50 µ m thickness. We implanted the samplewith atomic nitrogen at an energy and density of 6 keVand 3 · cm − , respectively. Subsequent annealing at800 ◦ C for two hours yielded a shallow layer of NV cen-ters of density ( ≈
10 NVs/ µ m ), and depth ≈
10 nm. Wethen etched the sample from the back-side to a thickness ≈ µ m with reactive ion etching (RIE, Unaxis shuttle-line), using a combined ArCl [31] and O [19] process.On the thin diamond membrane, we fabricated an arrayof diamond nanopillars on the top-side by using electron-beam lithography and RIE as described in [19]. Next, weperformed a second lithography step on the back-side ofthe diamond slab, which defined platforms to hold the di-amond nanopillars. A final RIE process transferred theresist-pattern to the sample, and fully cut through thediamond membrane to yield in the structure shown inFig. 1d.To mount a pre-selected diamond platform on an AFMtip, we employed a focussed ion beam (FIB) system (ZeissNVision 40) which was equipped with a nanomanipula-tor (Omniprobe AutoProbe 300) and ion-assisted metaldeposition. We employed tungsten deposition to fuse adiamond platform to a quartz AFM tip and then usedFIB cutting to release the diamond platform from thebulk. With a properly aligned FIB, this process does notcontaminate the scanning diamond nanopillar, and yieldsa NV/AFM probe as shown in Fig. 1b. Combined confocal and atomic-force microscope.
We employed a homebuilt microscope combining op-tical (confocal) imaging and AFM. The optical micro-scope is based on a long working-distance microscopeobjective (Mitutoyo ULWD HR NIR 100x, 0.7NA). The AFM was tuning-fork based, controlled using commer-cial electronics (Attocube ASC500) and mounted using ahome-built AFM-head. Both the sample and the AFM-head were fixed on 3-axis coarse and fine positioning units(Attocube ANPxyz101 and ANSxyz100, respectively) toallow positioning of the diamond tip with respect to thefixed optical axis and subsequent scanning of the samplewith respect to the diamond probe.Optical excitation of the NV center was performed bya diode-pumped solid-state laser (LaserGlow LRS-0532-PFM-00100-01) at a wavelength of 532 nm. Pulsed exci-tation for coherent NV spin manipulation used a double-pass acousto-optical modulator AOM setup (Isomet,AOM 1250C-848). ESR was driven with a microwavegenerator (Rhode Schwartz, SMB100A) and amplifier(MiniCircuits, ZHL-42W). Both the AOM and mi-crowave source were timed using a computer-controlledtrigger-card (Spincore, PulseBlasterESR-PRO-400).
Fit to spin-echo data.
To obtain the NV T -time form the spin-echo mea-surement presented in Fig. 2d, we fitted the data to asum of gaussian peaks, modulated by a decay envelope ∝ exp[ − ( τ /T ) n ], i.e., we employed the fit-function [32]exp[ − ( τ /T ) n ] (cid:88) j exp[ − (( τ − jτ rev ) /T dec ) ] . (1)Taking T , n, τ rev and T dec as free fitting parameters,we found T = 33 . µ s, n = 1 . τ rev = 16 µ s and T dec =4 . µ s for the data shown in Fig. 2d.The following supplementary material is divided intofive sections. Each section provides background infor-mation related to specific topics of the main text. Thesections are not built upon each other and can be readindependently. Section provides details for the model-calculation used to simulate the NV magnetic image inFig.3d. In section , we discuss limitations to NV mag-netic imaging if the NV sensor is in close proximity toa strongly magnetized sample. Experimental limitationsto the achievable NV-to-sample distance are discussed insection . The fabrication of the sharp metallic tips em-ployed in FQI (Fig.4a) is detailed in section . Finally,section contains a description and simple model for thetopographic features observed in the FQI image in Fig.4a. Supplementary Information
The following supplementary material is divided intofive sections. Each section provides background infor-mation related to specific topics of the main text. Thesections are not built upon each other and can be readindependently. Section provides details for the model-calculation used to simulate the NV magnetic image inFig.3d. In section , we discuss limitations to NV mag-netic imaging if the NV sensor is in close proximity toa strongly magnetized sample. Experimental limitationsto the achievable NV-to-sample distance are discussed insection . The fabrication of the sharp metallic tips em-ployed in FQI (Fig.4a) is detailed in section . Finally,section contains a description and simple model for thetopographic features observed in the FQI image in Fig.4a.
S1. Simulation of magnetic images
In order to reproduce the magnetic images obtainedwith the scanning NV sensor, we performed a model-calculation of the local magnetic fields in proximity to thehard-disc sample we imaged in our experiment. The mag-netic domains were approximated by an array of current-loops in the sample-plane as illustrated in Fig. 5a. Wechose the sizes of the loops to match the nominal size ofthe magnetic bits on the sample (bit-with 200 nm andbit-length 125 nm and 50 nm for the tracks in the figure)and set the current to 1 mA (corresponding to a densityof ≈ . ), which we foundto yield the best qualitative match to the magnetic fieldstrengths observed in the experiment. We then appliedBiot-Savart’s law to this current-distribution to obtainthe magnetic field distribution in the half-plane abovethe sample.Fig. 5b shows the resulting magnetic field projectiononto the NV center at a scan height of 50 nm abovethe current loops. The NV direction was experimentallydetermined to be along the ([011]) crystalline directionof the diamond nanopillar (in a coordinate-system where x − , y − and z − correspond to the horizontal-, verticaland out-of plane directions in Fig. 5b), by monitoring theNV-ESR response to an externally applied magnetic field(using 3-axis Helmholtz-coils). We then allowed for slightvariations of the NV orientation due to alignment errorsbetween the diamond crystallographic axes and the scandirections to find the NV orientation that reproduced ourexperimental data best. With this procedure, we foundan NV orientation ( √ φ ) , √ φ ) , / √
5, with φ = π / . ±
10 MHz from the bare ESRfrequency, all in accordance with our original experimen-tal parameters.
S2. NV magnetometry in close proximity to astrongly magnetized sample
The presence of a strong magnetic field B ⊥ , transverseto the NV axis leads to a reduction of contrast in opticallydetected ESR and moreover reduces the overall fluores-cence intensity of the NV center [33]. These effects resultfrom a mixing of the NV spin-levels in the optical groundand excited states of the NV center in the presence of B ⊥ .Such mixing on one hand allows for spin non-conservingoptical transitions and on the other hand suppresses thespin-dependance in shelving from the NV excited state(triplet) to the metastable NV singlet state. Both, spin-conservation under optical excitation and spin-dependantshelving are responsible for the non-zero contrast in opti-cally detected ESR of NV centers [34] and consequently,their suppression with transverse magnetic fields explainsthe disappearance of NV magnetometry features whenclosely approaching a strongly magnetized sample.Fig. 6a shows the raw NV fluorescence counts observedwhen scanning an NV in a diamond nanopillar in closeproximity (estimated 10 −
20 nm distance between NVand sample surface) to the sample. Dark features ap-pear when the NV is scanned over magnetic bits thatenhance B ⊥ , while the inverse happens when B ⊥ is re-duced (or the longitudinal field B NV enhanced) by localfields. This mode of bit-imaging allows for spatial res-olutions ≈ −
30 nm (Fig. 6c). At the same time, amagnetic image recorded with the technique describedin the main text shows no appreciable imaging contrast(Fig. 6b). Only exceedingly long integration times on theorder of hours allowed us to reveal weak magnetic fea-tures with dimensions on the order of 20 nm (Fig. 6d).The rates of the two effects which lead to a disap-pearance of ESR contrast, i.e. spin-flip optical transi-tions and shelving of m s = 0 electronic states into themetastable singlet, scale approximately as (cid:16) B ⊥ D GS − D ES (cid:17) and (cid:16) B ⊥ D ES (cid:17) , respectively, with D GS(ES) the ground-(excited-) state zero-field spin-splitting of 2 .
87 GHz and1 .
425 GHz [35], respectively. Given that D GS ≈ D ES ,the scaling of the two mechanisms with B ⊥ will be verysimilar. The characteristic scale of D ES ( D GS /
2) for thedisappearance of ESR contrast thus allows us to esti-mate B ⊥ close to the sample to be B ⊥ ≈ D ES /γ NV ≈
514 Gauss. We note however that this simple argumentlikely gives and over-estimation of B ⊥ as smaller valuescan already significantly effect ESR contrast and NV flu-orescence intensity due to the complex dynamics of NVspin pumping. Indeed, strong reductions of NV fluores-cence rates for B ⊥ less than 100 G have been observed
200 nm Currents B NV (Gauss): −52 50200 nmMagnetic field along NV axis I norm (%): 80 120200 nmNV response a b c FIG. 5:
Simulation of NV response to bits of a magnetic memory . (a) Current distribution used to simulate themagnetic bits imaged in this work. Red (blue) loops indicate currents of 1 mA in the (counter-)clockwise direction. (b) Magneticfield generated by the current-distribution in (a), projected on the NV axis at a height of 50 nm above the current loops. TheNV axis was tilted by 37 ◦ out of the scan plane ([011] crystalline direction) with an in-plane component as illustrated bythe blue arrow. (c) NV magnetometry response obtained from the magnetic field distribution in (b), assuming a LorentzianNV-ESR response and RF detunings as in the original experiment (see text).
100 nm x " I norm (%): 80 130100 nm x" I no r m ( % / d i v . )
100 nmsample−z
I/I ∞ ( % / d i v . )
500 nm I norm (%): 97104100 nm ab cd e
FIG. 6:
Quenching of NV fluorescence and ESR contrast in hard-disc imaging . (a) Total NV fluorescence I norm as afunction of sample position for an NV in close proximity to the hard-disc sample. I was normalized to the average fluorescenceintensity of I ≈ I norm shows a periodicity of ≈
64 nm, indicating a bit-width of 32 nm. (d)Fluorescence approach curve on the magnetic memory medium. NV fluorescence I was normalized to the fluorescence rate I ∞ = 27 (cid:48)
000 cps when the NV center was far from the sample. In contact with the magnetic sample, NV fluorescence wasreduced by almost a factor of two compared to the NV counts far from the sample. (e) Magnetic imaging with the same NVsensor: Even in close contact to the sample, NV magnetic imaging using ESR is still possible, albeit with a strongly reducedESR contrast and signal to noise ratio compared to the data shown in the main text. Data in (e) was acquired over 180 minutes,the smallest resolvable magnetic domains (top third of image) have a width of ≈
20 nm. in the past [33]. Transverse magnetic fields on this orderwere consistent with the largest on-axis magnetic fieldsobserved on our experiments as well as with the calcula-tions of magnetic field profiles presented in Sect. (for theparameters used in Fig. 5, we obtain maximal values of B ⊥ ≈
200 Gauss for an NV-to-sample distance of 20 nm).
S3. Limitations to NV-sample distance
As mentioned in the main text, NV-sample distance isan essential parameter for the performance of our micro-scope as it determines the overall resolving power withwhich weak magnetic targets can be imaged. We identi-fied three critical parameters that can affect NV-sampledistance: • Implantation-depth of NV centers in the diamond
FIG. 7:
Contamination of diamond tips . (a) AFM image of the end of a scanning diamond nanopillar after contaminationduring scanning. The image was acquired by scanning the diamond nanopillar over a sharp diamond tip as shown in Fig. 8.(b) AFM Image of the same nanopillar as in (a) after cleaning of the pillar’s end-face by repeated “scratching” over the sharpdiamond tip. nanopillarsThe depth of the NV centers created using ion im-plantation can be controlled by the energy of theions used for NV creation. However, the stoppingof ions in matter is a random process [36] and thedepth of the created NV centers therefore not per-fectly well-defined. This straggle in ion implanta-tion poses an intrinsic uncertainty to the distancebetween the scanning NV and the end of the di-amond nanopillar. For implantation energies of6 keV (with nominal implantation-depths of 10 nm)as used in this work, NV straggle has recentlybeen shown to be as large as 10 −
20 nm [23, 37].We note that since straggle in NV implantation ishard to circumvent it is essential for the future todevelop techniques to precisely pre-determine thedepth of a given sensing NV in a diamond nanopil-lar. This could be performed using recently de-veloped nanoscale imaging methods for NV cen-ters [23], or by scanning the NV sensor over a well-defined magnetic field source. • Contamination of scanning diamond nanopillarsDuring scanning-operation, the scanning diamondnanopillar can gather contamination from the sam-ple or environment. An example for such a con-taminated diamond-tip is shown in the AFM im-age shown in Fig. 7a (which was acquired with thescanning protocol employed in Fig.4, using the asharp diamond tip as shown in Fig. 8). Such con-tamination can artificially increase the distance ofthe scanning NV center to the sample by several10 ’s of nm (see Fig. 7a). To undo contamination ofthe diamond-tip after excessive scanning over dirty samples, we developed a “tip-cleaning technique”that allowed us to revert a contaminated tip to itsinitial, clean state (as illustrated by the transitionfrom Fig. 7a to b). Tip cleaning is performed byrepeated scanning of the diamond nanopillar overa sharp diamond tip (Fig. 7a) in the absence ofAFM feedback. Such feedback-free scanning canpartly remove contamination from the diamond pil-lar, which after repeated operation leads to a cleandevice as the one shown in Fig. 7b.We note that with proper sample-cleaning, controlover environmental conditions and occasional “tip-cleaning” runs, adverse effects of tip-contaminationcan be essentially eliminated. This, together withthe excellent photo-stability of NV centers, thenallows for long-term operation of the scanning NVsensor. • AFM controlProper AFM control is necessary to assure closeproximity of the NV center to the sample surface.It has been shown in the past that bad mounting orimproper AFM feedback control can lead to AFMtip-sample distances in excess of 20 nm [38]. Care-ful mounting of AFM tips and proper setup andtuning of AFM feedback (here provided by an At-tocube ASC500 controller) was therefore essentialto observe, for instance, the FQI features discussedin Fig.4 of the main text.0
FIG. 8:
Sharp diamond tip for FQI . Image of a sharpdiamond tip similar to the one used for the experiments pre-sented in Fig.4a of the main text. Typical tip-radii are on theorder of 10nm.
S4. Fabrication of sharp diamond tips
For the experiment presented in Fig.4a of the maintext, we fabricated sharp diamond tips which were metalcoated for FQI. Diamond tip fabrication was based on thenanofabrication techniques [19] that we already employedfor the production of the scanning diamond nanopillarspresented in Fig.1. A type Ib diamond (Element six)was patterned with circular etch-masks (flowable oxide,FOx XR, Dow Corning) of 100 nm diameter. Here, inorder to obtain sharp diamond tips instead of cylindri-cal diamond nanopillars, we modified the RIE etchingrecipe we had previously used: While we kept the (oxy-gen) etching chemistry identical to pillar fabrication, wesignificantly increased the etching time, such as to com-pletely erode the etch mask on the diamond substrate.As a result, the etched diamond structures acquired theform of sharp tips as shown in the representative SEMimage in Fig. 8. Typical tip-radii were in the range of10 nm and tip lengths were on the order of 200 nm.For FQI, we then coated the sharp diamond tips witha thin metallic layer using thermal metal evaporation.To avoid oxidation of the metal, we chose gold as thequenching metal and used a chrome adhesion layer be-tween the gold and the diamond. For the tips employed in this work, we deposited 5 nm of gold and 5 nm ofchrome.
S5. Explanation of FQI features
The features observed in Fig.4a of the main text weregoverned by direct fluorescence quenching through metal-lic objects (as highlighted by the red square in the figure)and by a confluence of the distance-dependance of theNV fluorescence with topographic features on the sam-ple (bright, ring-shaped feature in the figure). Whenapproaching the NV to the (metallic) sample, the to-tal NV fluorescence collected in the far-field through thepillar changed as shown in the measurement in Fig.4e(blue curve) and the corresponding data for the FQI sam-ple shown in Fig. 9b. This well-known [39] variation ofNV fluorescence is a result of the variable electromag-netic density of states in the vicinity of a dielectric in-terface which influences the NV radiative lifetime as wellas the total effective laser excitation intensity imping-ing on the NV center. During our scanning experiments,the topography causes the mean distance between thescanning NV center in the nanopillar and the metallicsubstrate to vary, which in turn causes variations in thecollected NV fluorescence rate. Assuming to first orderthat the metallic tip does not itself affect NV fluores-cence (so long as it is not placed in direct contact to theNV center as in the “red-square region”), one can un-derstand most features observed in Fig.4c as a pure ef-fect of topography. Based on this principle, in Fig. 9 wereconstruct the FQI image from a measurement of sam-ple topography (a) and an independently acquired fluo-rescence “approach-curve” (b). The reconstructed FQIimage (Fig. 9c) was obtained by taking the value of theAFM z-displacement for each point in the scan and look-ing up the corresponding fluorescence-rate obtained inthe approach-curve. The resulting image shows strikingsimilarity with the actually measured FQI image (Fig. 9e;same data as Fig.4a) and confirms the validity of our ex-planation.1 z (nm): 0 264100 nm AFM data I=I(z)0 100050100150 sample−z
I/I ∞ ( % ) I/I (%): 70 140100 nminferred NV fluorescence I/I (%): 84 200100 nmmeasured NV fluorescence a b c d FIG. 9: