1. Introduction

The quantum spin properties of fluorescent nanodiamonds (FNDs) that contain negatively charged nitrogen-vacancy colour centers (NV) can be controlled by applying direct current (DC), microwaves (MW), magnetic fields and light, which enables applications for imaging with high contrast and sensitivity [1–3]. The NV’s spin properties can be coherently manipulated by optically detectable magnetic resonance (ODMR) technique that is applicable for sensing and quantifying the local magnetic fields [3]. In addition, FNDs with NV’s are effective alternatives to standard fluorescent markers as photostable and biocompatible theranostic nanoparticles that can be used for controlled delivery/actuation as imagers and probes [4–6]. Intriguing avenues using FNDs as magnetically sensitive probes in ‘smart’ fiber-optic flexible endoscope systems can be realized [7,8]. Fibers are fabricated mostly using silica glass, due to its availability and excellent optical performance in the visible/near-infrared regions. Several nontrivial multicomponent glasses are accessible e.g.: soda-lime-silicate glass, niobium-lanthanum-borosilicate glass, etc. [9]. In such systems, FNDs can provide accurate physiological sensing capabilities during detection and ablation. To achieve FND-based fiber-optic detection with nanoscale detection volume, the development of practical methods to assemble FNDs as a uniform thin-film/layer on the fiber tip (end face) is essential.

Surface assembly of the FNDs can be achieved by the physical attachment yielding a multilayer of particles or a thin film, and/or the chemical attachment by plating the substrate with linker molecules yielding organized layers. The known physical approaches to immobilize FND thin-films are chemical vapor deposition (CVD) [10] and liquid-phase pulsed laser ablation [11]. Despite the proven usefulness of these methods on various types of flat substrates with subsequent pre- and post- annealing treatments, their potential remains hindered as the scalability and usefulness are limited for coating FNDs on long fiber sections. In addition, the post-annealing treatment may change the specific properties of the FND-fiber interface. On the other hand, the techniques of spin coating, dip-coating, and spray pyrolysis are based on electrostatic or van der Waals interactions that are fragile and may not produce uniform films which severely hinders uniform assembling of FNDs on the fiber tips. Mixing FNDs with an optical adhesive has been recently used for physicochemical attachment to the fiber tip, albeit as a thicker multilayer [12].

The chemical attachment can be thought of as a convenient alternative where a cross-linking reagent can bind one end to the substrate, and the other is ready to bind to FNDs forming the immobilized layer. The previous studies have proven that a variety of nanoparticles can be self-assembled from a solution onto a functionalized glass surface via various cross-linking reagents [13,14], known as ‘the silanization method’. Organofunctional silanes have been commonly used to obtain amine-functionalised layers on nonmetal surfaces for such attachment, whereas the positively charged top amine group of silane is more suitable in attracting negatively charged nanoparticles. The silanization method is facile, straightforward, environment-friendly, relatively inexpensive, and above all it is easily transferable from the laboratory to a mass-production scale.

In this article, we implement the surface chemical approach of the silanization method to covalently anchor carboxylated FNDs on glass surface. Here, we use (3-aminopropyl)triethoxysilane (APTES), arguably the best-known coupling agent to form silanated glass surfaces, and a diamine using carbodiimide as a cross-linker. Silanization method is used in recent works for assembling FNDs on glass surface [15], here this method is used as a new strategy to assemble FNDs onto the tip of fibers which are only a few micrometer in diameters and have different types of glass materials/compositions. We also demonstrate recording of fluorescence and ODMR signals from NV centers in the FNDs anchored to the distal end of single large-core multimode optical fiber (LCF) and imaging fiber bundles (IFB). Signals recorded in the presence of applied magnetic fields are presented. The potential impact of this work is demonstration of a multi-channel sensor capable of instantaneous and simultaneous parallel-in time and magnetic field mapping from individual and multiple pixels of IFB with assembled FNDs.

2. Experimental set up

2.1 Chemicals

A carboxylated slurry of FNDs with the average size of 140 nm having 1.5 ppm concentration of NV centres dispersed in deionized (DI) water, were purchased from Adamas nanotechnologies. APTES, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide,HCl (EDAC), ethylenediamine (EDA), toluene, potassium chloride, acetone, and isopropyl alcohol (IPA) were purchased from Sigma-Aldrich in their highest purity form. DI water (18.2 MΩ·cm) was used for the preparation of the solutions, synthesis and cleaning glass substrates/fibers. All other chemicals were used without further purification.

2.2 Types of fibers and glasses

The glass substrates for the FND immobilization were microscopic slides (VWR, Leuven) made of soda-lime glass and cut into the 0.5 × 0.5 cm squares. For LCF, the core was made of niobium-lanthanum-borosilicate glass with refractive index of n_d = 1.7295 (labelled SEL101 or LBS-23), and cladding was made of sodium-potassium-borosilicate glasses (labelled SP27 or CG22) with a refractive index of n_d = 1.5192. For the IFB, the cores were made of a soda-lime-silicate glass (labelled SK222) with a refractive index of n_d = 1.522 while the cladding was made of barium-zirconia-borosilicate glass (labelled Zr3/XV) with n_d = 1.611 [16,17]. For LCF the core/clad diameter (d) was 80/125 μm and for IFB each pixel is d = 3 μm with core/clad d = 586/629 μm, respectively.

2.3 Thin-film covalent assembly of FNDs on APTES-functionalized glass substrates and fibers

To covalently assemble the FNDs on a glass substrate and the fiber’s tip silanization method was followed [13,15,18]. Glass surfaces were sonicated with DI, acetone, and IPA for 20 min each and dried at room temperature before the silanization process. For the silanization, the glass surfaces were hydroxylated with ozone (Osilla UV Ozone cleaner) for 30 min (254 nm output intensity, 20 μW/cm² at a distance of 1 cm). Afterward, the glass substrates and one end of glass fiber tips were immediately immersed in the 0.01% (v/v) solution of APTES in toluene for 30 min under room temperature to crosslink the APTES layer and form the APTES functionalized glass surface. Next, the APTES functionalized glass surfaces were sonicated with toluene for 5 min to remove any physisorbed APTES molecules. To prepare the amine-reactive O-acylisourea, (1 mg/ml) FND-COOH aqueous solution was dispersed in 1:1 (v/v) ratio with 0.3 mg/ml EDAC in 2 mM (pH 6.5) KCL buffer solution for 30 min at room temperature. As-prepared APTES functionalized glass surfaces were immersed in a solution containing amine-reactive O-acylisourea form of FND-COOH for 30 min. Thus, the first cycle to form a single layer of FNDs on the glass substrate and one face of fiber was complete. For increasing the FND-COOH concentrations to form dense coverage minimum of two to a maximum of seven regeneration coating cycles were done and a fresh FND-COOH suspension was used for every regeneration cycle. The FND covered on the glass substrate and one side of fiber ends were regenerated by treatment with dipping in KCl for 2 min, EDA for 30 mins, and DI for 2 min followed by the amine-reactive O-acylisourea steps. Finally, the prepared FND assembled samples were rinsed with DI water, naturally dried and stored in a desiccator until further use.

2.4 Characterizations

Optical microscopy images were acquired in reflection mode using a laboratory microscope equipped with a Motic 3 + digital camera. Fluorescence images were acquired using a home-made confocal microscope. Optical excitation of FNDs were achieved using a 532 nm laser (Sprout G, Lighthouse Photonics) with around 1 mW of power. To illuminate and collect fluorescence from the sample, an oil-immersion objective (100x/1.3, UPLFLN Olympus) was used and the samples were scanned by a piezo-nanopositioning stage (Nano-LP200, Mad City Labs) controlled by a computer with the Qudi software [19]. Detection of the emitted fluorescence was performed by a single photon counting module (SPCM-AQRH-14-FC, Excelitas Technologies). The confocal pinhole placed just before the fiber connected to the SPCM eliminated the fluorescence emitted out of the focal place by spatial filtering.

Atomic force microscopy (AFM) was performed for FND-coated glass samples using a NanoSurf FlexAFM C3000 atomic force microscope with Accurion Nano 20 vibration isolation stage. Four spots on the sample (size: 4 × 4 μm) were scanned to ensure representative areas. All scans were done in the tapping mode. AFM images were leveled by mean plane subtraction, corrected for horizontal scars and aligned rows with the median method, using the Gwyddion package [20]. Grains on the image were masked and their area surface was calculated using Mark Grains by Threshold tool and statistical functions. The degree of surface coverage with FNDs was determined as the ratio of the all grain surface area to the total surface area of the scan.

ODMR signals were recorded using a home-made wide-field microscopy set-up with MW frequency around 2.87 GHz, delivered through a planar-ring inductor or microstrip line. The glass substrate and FND coated fiber’s tip were positioned near (0.5 mm) the MW inductor or microstrip line [7]. Figure 1 shows the schematic representation of the ODMR set-up. The FND’s were continuously excited by a focused 530 nm LED with optical power below 70 mW. LED was preferred over laser due to it’s compactness, high-enough output power, good stability and the lack of interference patterns (or speckles).The beam was focused by a microscope objective (40x, NA = 0.60, WD = 4 mm Olympus) and for the fibers on the side without FND coating. The emitted red fluorescence light (∼ 600 - 800 nm) from the glass substrate and the fiber’s non-coated side was transmitted by the same confocal lens system as the green exciting beam and spectrally filtered by the dichroic mirror, optical long pass filter and recorded by a CMOS camera. For magnetic field-induced ODMR signals, an external permanent magnet was placed at different positions near the microscope to create a static magnetic field with a varied strength between B = 0 to 2.5 mT. Further, the ODMR data was instantaneously collected from various pixels of fibers tip using an in-house developed Python script and processed in MATLAB [21] environment.

 

Fig. 1. Schematic representation of the ODMR set-up. The sketch represents the assembly of the optical imaging system used for collecting the ODMR signals through the fibers.

Download Full Size | PDF

3. Results and discussion

3.1 Thin-film assembly of FNDs on glass substrates

The thin-film assembly was carried out using carbodiimide as a mediator between carboxylated FNDs and EDA on APTES-functionalized glass surfaces. Figure 2 shows microscopy images of polycrystalline FNDs thin-film assembled by directed covalent deposition on a glass substrate with two Fig. 2(a-c) and seven Fig. 2(e-f) regeneration (repetition) cycles.

 

Fig. 2. 140 nm sized FND covalently assembled (two regeneration cycles) in thin film form on glass substrates: (a) Wide field microscopy images, (b) Fluorescence image, (c) AFM image FND assembled on glass substrates (d) averaged ODMR spectrum as shown in inset (red circle area) (e) Wide field optical microscope image and (f) AFM image for seven regeneration cycles.

Download Full Size | PDF

Figure 2(a) shows the image of the FNDs assembled glass substrate with two regeneration cycles obtained from the wide-field microscope. The two regeneration cycles allowed to form a denser coverage on the glass substrate. The image showed that the substrate was covered with FNDs over the aminosilane treated glass surface and evenly spread over a large area up to few millimeter scale (Fig. S1). No cracking on the films was observed after drying them but it was noticed that the FND films were soft and prone to tweezer scratching while handling the substrates during immobilization. Figure 2(b) shows the fluorescence map acquired by scanning the FND-covered surface. It demonstrates that FNDs uniformly assembled over the whole substrate emit fluorescence intensity at 18 kc/s photon count level while having the background level of about 8-10 kc/s rate. The measured signal was emitted from the NV centers in FND assembled on the glass substrate, as the glass substrate did not fluoresce. Figure 2(a and b) depicts that the thin film assembly represents the Volmer-Weber island-type mode as reported in [15,22], which is characteristic for the particle-particle cohesive force that is stronger than the particle-surface adhesive force [23]. Metwalli et al. [24], examined the island formation processes for APTES by changing the silanization conditions such as temperature, concentration, solvents, hydration, reaction time, and the pre- and post-treatments on glass substrates. It was found that the island-type mode of thin-film assembly, besides the strong particle-particle cohesive force, can also be attributed to the non-uniform multilayer organosilane deposition (island formation) during silanization of the substrates [24].

To get more insight into the film growth mechanism, AFM imaging was performed. Figure 2(c) presents the AFM image of FNDs deposited on the glass substrate which demonstrates about 50% coverage. The disordered stack-like assembly of FND aggregates re-confirms the Volmer-Weber mode (island-type) of the surface thin-film assembly [15,22]. Full description of the AFM data processing and the determination of the coverage degree can be found in the experimental section. After uniform fluorescence signals from FND layers on glass substrates were identified, we recorded ODMR signals associated with NV⁻ color centers in diamond using a wide-field detection set-up (as shown in Fig. 1). Figure 2(d) shows the signal averaged over a selected area of 0.05 mm² of the FND-film on the glass substrate. The ODMR spectra consist of two branches characteristic for the NV diamond zero-field splitting of 2870 MHz at room temperature [25] and corresponding to MW transitions between states m_S = 0 and m_S = ±1 spin states of NV⁻, which are degenerate in zero magnetic field, B, but split when B ≠ 0, which enables magnetometric applications. The fluorescence intensity is normalised to the off-resonance value to obtain the ODMR contrast spectrum. The contrast of the ODMR signal determined as the percentage ratio between the maximum and minimum fluorescence intensity for a selected area of FND-film has reached the value of ∼1%. Some background is coming from glass, and that reduces the contrast. In addition, the contrast was lower than ∼4% value observed for the 140 nm FND powder at B = 0 (not shown here), due to the thin-layer assembly of the FNDs. When assembling FNDs on the fiber’s tip, higher signal contrast values were required which was achieved by deposition of four to seven regeneration cycles to get denser coverage of FND.

Figure 2(e and f) shows the wide-field optical and AFM images acquired for the substrate subjected to seven regeneration cycles for growth and density comparison. With more deposition cycles, the formed FND films were spread all over the surface of the glass much denser than with two cycles at approximately 87% of coverage. The film morphology still looks like Volmer-Weber mode assembly, albeit a densely covered one. With more cycles, amine groups on the islands have higher exposure to the environment than the rest of the surface which results in a higher probability of FND attachment. The orientation of APTES molecules on the glass surface plays an important role for the interfacing of FNDs. Different orientations of APTES molecules on glass substrate and thermal treatments affect the surface silanol coverage, and thus the organization of FNDs on APTES-functionalized glass substrate. In our case, at higher deposition cycles observed surface morphology depends on the size distribution of initial FNDs utilized along with the formed APTES molecules, although we could see that the FNDs assembled are densely covered throughout the glass substrate with higher deposition cycles.

3.2 Thin-film assembly of FNDs on the fiber end-face

For achieving covalent assembly of FND-films on the end facets of a LCF and IFB, we firstly gently sonicated sections of the fragile glass fibers to achieve a clean surface. Next, we placed the fibers horizontally on the tray of an ozone cleaner to hydroxylate the entire fiber section. Then the complete deposition procedure was skillfully performed with only one end of the fiber dipped in the solutions (as described in the Experimental Set-up Section). For that purpose, a small amount of each solution was pipetted into Eppendorf tubes that allowed us to place fibers vertically such that only one end of the fiber was exposed for assembling the FNDs.

The number of coating cycles was decided based on the diameter of the fiber core(s). For LCF we deposited FNDs with four cycles while seven cycles were performed for the IFB, which required a denser coverage to record signals from an individual, much smaller pixel (core). The FND films obtained at fibers’ end-face were observed to be very similar to those for the glass substrate. Figure 3 shows the optical microscope images of the FND coated (right side) and uncoated (left side) sides of LCF and IFB. It can be seen that the degree and evenness of fiber coverage are close to those for the FND thin-film coating on the glass substrates. While we have not measured the film thickness on the fiber end-faces, the same two cycles deposition procedure carried out on a glass substrate resulted in a film thickness of less than 200 nm (measured using Bruker DektakXT stylus profilometer), which corresponds to a single, not fully covered layer of 140 nm FND particles. Similar film-formation mechanism has been observed in Ref. [15]. Surface roughness after four and seven regeneration cycles has been shown in Fig. 2(c and f), respectively. It matches the film thickness measured using the stylus profilometer. Since the film is not uniform, only an effective thickness can be estimated and in the case of fibers shown in Fig. 3 we estimate it to be around 200-300 nm.

 

Fig. 3. Microscope images without coatings (left side) and with thin film covalent FND coating (right side) on LCF (four regeneration cycles) and IFB (seven regeneration cycles). The LCF has core diameter (d) 80 μm and cladding diameter d = 125 μm. The IFB has core d = 586 μm and clad d = 629 μm with individual pixel d = 3 μm.

Download Full Size | PDF

The surface properties and structure of mono-aminosilane treated glass surfaces for different types of glasses (silica, soda-lime, borosilicate, and synthesized fused silica) were examined by Metwalli et al. in Ref. [24] (and the citations therein). Their results proved that regardless of the exact glass chemistries, the silanization with APTES produced roughly a monolayer of silane coatings. Our results, confirmed by optical images, prove that the silane-based coatings can be utilized on many types of glass material and with multicomponent types of glasses as substrates and can be used to evenly immobilize FNDs.

3.3 ODMR detection through the fiber

For recording the ODMR signals remotely, the changes in the fluorescence intensity were detected and collected through the non-coated side of the fiber, which we call as proximal end. The lengths of the LCF and IFB sections were 3.4 cm and 1.5 cm, respectively. The NV centers were optically excited and pumped into the m_S= 0 spin sublevel by the green light and a spin-dependent fluorescence light from the NV⁻ center was collected through the high NA microscope objective on the proximal end of the fiber. On the other hand, the MWs were delivered directly to the distal end of the fiber (fiber tip) that was coated with FNDs.

A resonant MW field drives transitions between the m_S= 0 and m_S= ±1 spin sublevels of the NV⁻ spin triplet ground state modulating the observed fluorescence level. Figures 4 and 5 illustrate the ODMR spectra acquired without and with externally applied magnetic field. The magnetic field originated from a permanent magnet and its strength was controlled by adjusting the relative distance from the fiber (using a translation stage) in-plane with the end-face. The magnetic field value at the sample surface is measured using a Gauss meter.

 

Fig. 4. ODMR recorded through the LCF without and with applied magnetic fields B = 0, 0.5, 1, 1.5, 2, and 2.5 mT (inset: inside red circle area). MW field is applied at FND coated side and the ODMR signals are recorded from non-coated side (far end) of the fiber. Vertical line indicates the symmetry frequency for the zero field. The visible broadening of a signal is caused by an increase in the applied magnetic field.

Download Full Size | PDF

 

Fig. 5. ODMR recorded through IFB without and with applied magnetic fields B = 0 and 2.5 mT. MW field is applied at FND coated side and the ODMR signals are recorded from non-coated side (far end) of the fiber. Vertical line indicates the symmetry frequency for the zero field. The visible broadening of a signal is caused by an increase in the applied magnetic field. (a) average signal recorded (inside the red circle area) and (b) signal recorded from individual core 3 μm, as shown in the respective insets.

Download Full Size | PDF

Figure 4 corresponding to LCF shows that the ODMR signal exhibits a drop in the luminescence intensity at resonance frequency of 2870 MHz (the zero-field splitting). The inset shows the non-coated side of the LCF employed for this measurement , having a single core and cladding with the core diameter being 80 μm. The ODMR signals were collected from the core of the fiber (area within the red circle). The magnetic fields were varied between B = 0 to 2.5 mT. With applied magnetic field we observed an increase in the splitting of the ODMR peaks associated with the separation of the m_S= ±1 spin states caused by the Zeeman effect. In addition to the splitting, the broadening of the resonance and its contrast reduction is seen which we associate with different Zeeman splitting in the randomly oriented FNDs, resulting in a continuous range of possible values of field projections on the NV⁻ axis. The overall shape of observed magnetic resonance spectra, collected without and with magnetic field, is consistent with previously reported ODMR signals [26]. By differentiation of the ODMR resonance shapes with respect to the MW frequency, one can retrieve values of the local magnetic field intensity which is relevant for sensorics and magnetometry as shown in Ref. [7]. The magnetic field sensitivity can be estimated using the observed ODMR fluorescence contrast and the photon flux coming from the fiber core. For the LFC, this results in the magnetic sensitivity of approximately 15 μT/$\sqrt {Hz} $ at the bias field of B = 1 mT, while the temperature sensitivity (at B = 0) is around 3.8 K/$\sqrt {Hz} $.

Analogous to the case of LCF, Fig. 5 shows the ODMR spectra for the IFB without (B = 0) and with applied magnetic field B = 2.5 mT. The insets show the wide-field fluorescence image observed on the proximal side of the IFB and the red circles mark areas from which ODMR spectra were extracted. In Fig. 5(a), the red circle encompasses area of 0.03 mm² which includes several individual fibers (pixels). Here, the corresponding ODMR signal represents spatially averaged fluorescence inside the circle of an individual fiber core with no contribution from the spacing between the individual cores (claddings). Figure 5(b) depicts a spectrum obtained from an individual fiber in the bundle that exhibits decreased signal-to-noise ratio. The main difference between the signal amplitudes in Fig. 4 and Fig. 5 comes from the significant variation in the corresponding diameters, i.e. areas from which the ODMR signals are recorded. While the individual fiber in the bundle has a 3 μm core size, the LCF had a much larger core of 80 μm diameter and therefore, the fluorescence level was higher due to much larger number of contributing FNDs per core. Additionally, fluorescence contrast is lowered by the autofluorescence of the core glass used for the bundle preparation, which is why short fiber section has been used. However, the spectra shape, position, and splitting between lines are all consistent with the measurements performed on glass substrates and show that there is good randomization of FND orientations.

Observation of high quality ODMR signals through the fibers was possible thanks to the high numerical apertures of around 0.5. To achieve such high numerical apertures for the fibers, careful selection of the glass components and glass fabrication skills are crucial [17]. These results demonstrate that the fluorescence and ODMR signals can be effectively collected through the fiber enabling endoscopic applications in which the thin-film assembled with FNDs can be used as a single or multi-channel sensing probe.

3.4 Multi-channel readout for high-spatial resolution

Successful covalent assembly of uniform FND coatings on optical fibers and detection of the ODMR spectra through the individual fibers of the IFB enables construction of a multi-channel NV sensing platform which can be used for temperature or magnetic-field mapping with a high spatial resolution governed by the spacing of pixels (fibers) in the bundle. Below, we demonstrate the parallel readout of ODMR signals (Fig. 6) that can be used to perform wide-field sensing. Briefly, to achieve good mapping of magnetic fields (or other modality) at high spatial resolution and sensitivity, it is essential to have an imaging system that collects many photons, i.e. a sensor which is capable of recording most of the incoming photons during a single exposure. Since NV ensembles exhibit a fluorescence contrast typically of just a few percent, for the IFB ODMR-based magnetometry, we used a sensitive industrial CMOS camera (IDS UI-3240CP-NIR-GL). The camera’s global shutter mode allows all the pixel rows to reset and then be exposed simultaneously, which offers a ‘freeze frame’ style of capturing images. With such a sensor type, a sub-micrometer spatial resolution and a magnetic field sensitivity of up to 7.9 μT per single pixel and frame can be achieved [27], provided the sensor receives enough fluorescence light. The parallel readout of multiple pixels enables simultaneous mapping of the magnetic field by precision ODMR detection in each fiber core. Figure 6(a) depicts a single MW-frequency fluorescence image. For each of the fibers an ODMR spectrum can be retrieved from a stack of such images taken while sweeping the MW frequency.

 

Fig. 6. (a) The x and y pixel positions (+) of all the bright spots identified in a circular region-of-interest taken from a single exposure image of the camera. Randomly selected bright spots (red circles) within the selected region-of-interest. (b) Reconstructed ODMR spectra for the randomly selected bright spots.

Download Full Size | PDF

Spectra were recorded by sweeping the MW frequency over a 125 MHz range, centered at 2870 GHz, in 1 MHz steps. The camera captured image for each MW frequency at a rate of 64 frames (frequencies) per second, with a single-frame exposure time of 200 ms. Exposure of the current image and readout/transfer of the previous image were performed simultaneously. A few (3 to 5) sweep repetitions and denser FNDs assembly were applied to achieve a typical contrast of 1% with sufficient signal-to-noise ratio per pixel resulting in the entire measurement time of around 3 minutes. The captured images were appended in the PC memory to form a 3D volumetric data giving the luminescence value for each pixel coordinates of the camera and MW frequency value. The recorded image frames (typically a few hundreds) were processed using in-house developed MATLAB [21] routines. Each of the individual fluorescent fiber cores is automatically identified by the algorithm from a user-selected region-of-interest or from the full field-of-view. Such identification is visible in Fig. 6(a) by small blue markings in many core centers in the centrally located circular region of the image. The ODMR signals were subsequently extracted by averaging the pixel intensities within individual cores of the fiber. As an example, zero-field ODMR spectra for eight arbitrarily selected cores (encircled in red and numbered) are plotted in Fig. 6(b). Such signals from a single measurement, allow for a truly instantaneous parallel spatial mapping of magnetic field or temperature using the IFB.

The demonstrated multi-channel ODMR readout Fig. 6 (Fig. S2 and S3) is fast, reliable, sensitive, and accurate due to the remarkable NV spin properties and nanoscale size of the FNDs. This opens up several opportunities not only for nano-thermometry and nano-magnetometry, but finds marvellous opportunities in the field of biomedicine, especially for endoscopy, or even for collecting information from far-away places where most of the sensors might not reach. The described sensor may easily be adapted to measure also the temperature and pressure changes (physical stress) via the exact shape of the ODMR signals [28,29]. Such information can be collected through the fibers. These FND-fiber based multichannel sensor should not age for much longer time scales due to the NV centers photostability [30]. We expect our multi-channel sensor to provide improved long-term stability over other sensors that show a noticeable decrease in sensitivity over a time frame of multiple months of use. In addition, this multi-channel sensor can enable magnetic mapping at high spatial resolution. Work on this current multichannel sensor is limited to high MW powers, spin relaxations, size, agglomeration, number of NV centers, whereas they do not show any severe constraint on the FND based fiber applications in materials sciences or device physics. We attempt to improve the magnetic signals by increasing the NA of the fibers to collect more light and are working on possibly increasing the working length of such a device.

4. Conclusions

In conclusion, we show the surface structure of thin film assembled FNDs on amino silane treated glass substrate and fiber tips using confocal, fluorescence and AFM microscopy. Even though the coatings were prepared on several glass types with different composition, the chemical method implemented allowed us to assemble FNDs uniformly over small to large glass surface areas. From the microscopic images we observe that the FNDs immobilized have a Volmer Weber type of growth mode that is dependent on the formed silane structure on the glass surfaces. AFM images show 56% and 87% degree of coverage with increasing regeneration cycles from two to seven. The formed FND layers are thinly covered and this is reflected in the ODMR signals showing relatively low contrast values, around 1%, limited by the glass autofluorescence. Further, we have successfully demonstrated recording of ODMR signals through the short sections of high-NA fibers, with the optical excitation and detection from the non-coated side, and only MWs delivered to the FND-coated end. The ODMR spectra shapes are consistent with previous reports on FND suspensions and powders. The ODMR contrast here was limited by the glass autofluorescence and, hence, short fiber sections have been used. However, we expect novel glass types to be developed with better optical properties, enabling the true endoscopic use of such a sensing platform.

Moreover, we have demonstrated a multichannel sensor with instantaneous readout capability for the parallel in-time mapping and to readout information from multiple pixels with high spatial resolution. We envision that this multi-channel sensor can be extended to include instant quantitative ODMR’s based on the shape of the signals in bio-diagnostic applications.