Dynamic Nuclear Polarization of Silicon Carbide Micro- and Nanoparticles

Abstract

Two dominant crystalline phases of silicon carbide (SiC): α-SiC and β-SiC, differing in size and chemical composition, were investigated regarding their potential for dynamic nuclear polarization (DNP). ²⁹Si nuclei in α-SiC micro- and nanoparticles with sizes ranging from 650 nm to 2.2 μm and minimal oxidation were successfully hyperpolarized without the use of free radicals, while β-SiC samples did not display appreciable degrees of polarization under the same polarization conditions. Long T₁ relaxation times in α-SiC of up to 1600 s (∼27 min) were recorded for the ²⁹Si nuclei after 1 h of polarization at a temperature of 4 K. Interestingly, these promising α-SiC particles allowed for direct hyperpolarization of both ²⁹Si and ¹³C nuclei, resulting in comparably strong signal amplifications. Moreover, the T₁ relaxation time of ¹³C nuclei in 750 nm-sized α-SiC particles was over 33 min, which far exceeds T₁ times of conventional ¹³C DNP probes with values in the order of 1–2 min. The present work demonstrates the feasibility of DNP on SiC micro- and nanoparticles and highlights their potential as hyperpolarized magnetic resonance imaging agents.

Introduction

Magnetic resonance imaging (MRI) is a noninvasive, radiation-free diagnostic imaging tool giving exquisite soft tissue contrast and high anatomical resolution. However, its application in molecular imaging has often been restricted due to its inherently low sensitivity.^1,2 Conventional MRI modalities usually detect high natural abundance nuclei, such as the protons in water or lipid molecules, to obtain sufficient signal intensity in spite of the low degree of polarization under thermal equilibrium.³ Contrast agents are frequently used to improve measurement sensitivities by shortening T₁ or T₂ relaxation times of water protons in close contact or proximity to the agent.^4,5 However, accumulation of gadolinium-based contrast agents after (repeated) administration gives concerns regarding toxicity issues.^6,7 A different approach is to perform imaging of the contrast agent or labeled particle itself instead of its effect on water protons, and use the high proton signal of water as an anatomical reference. Particles containing different nuclei can be observed without issues regarding the high proton background signal of water, and in principle, any nucleus with a nonzero net spin number, such as ¹³C, ¹⁵N, ¹⁹F, ²⁹Si, or ³¹P, can be imaged with MRI.⁸⁻¹⁰ Unfortunately, direct detection of nuclei other than protons remains difficult due to low gyromagnetic ratios, low natural abundancies, and low concentrations of particles of interest, leading to long acquisition times and low spatial resolutions.

An interesting alternative strategy that alleviates the sensitivity problem of MRI is hyperpolarization of the spins of interest, which increases nuclear spin polarization far beyond thermal equilibrium.¹¹ Dynamic nuclear polarization (DNP) is a technique that utilizes the much higher spin polarization of unpaired electrons to hyperpolarize nearby nuclear spin elements by transferring the electron polarization under microwave irradiation. The most widespread DNP method in MRI is dissolution DNP, which results in up to 10 000-fold signal amplification compared to typical thermal polarization.¹ Dissolution DNP has been used to investigate metabolic pathways and to study enzymatic activities. Small hyperpolarized ¹³C-labeled substrates, such as 1-¹³C-pyruvic acid or 1-¹³C-fumarate, have been successfully used to detect aberrant metabolism in cell suspensions, animal models, and human tumors.¹²⁻¹⁵ However, the hyperpolarized state is relatively short-lived, and the polarization will restore to thermal equilibrium with the relaxation time constant T₁. The majority of hyperpolarized ¹³C-labeled molecules have a T₁ in the order of 1 min or less.¹⁶ Therefore, when used as tracers or substrates, imaging is restricted to visualizing fast biological processes in areas or organs that can easily be reached by the tracer. Slower biological processes, such as protein-cell binding, tracer diffusion, and metabolism in tissues, cannot be studied. Moreover, these short lifetimes cause significant signal decay during quality control of the hyperpolarized substances, intravenous administration, and transport to the tissue of interest, impeding general application in the clinic.

¹³C or ²⁹Si nuclei in diamond and silicon particles have been demonstrated to display remarkably long T₁ relaxation times of up to several hours¹⁷⁻²¹ and alleviate the short-lifetime issues associated with ¹³C-labeled small molecules. Since interactions between nearby spin elements result in relaxation, spatial isolation of spin elements beyond the distance of spin interaction in a solid matrix can impede relaxation processes, leading to dramatically increased T₁ times. For instance, diamonds have been shown to display spin-lattice relaxation times of as long as 3–4 h.¹⁷¹³C nuclei in micro/nanosized diamonds have been successfully hyperpolarized, showing enhanced DNP signals. However, T₁ times in nanodiamonds with particle sizes ranging from 30 to 500 nm are still relatively short, usually in the order of 5 min or less,²² though these particle sizes are most interesting for biomedical applications.

Aptekar and Cassidy et al. elegantly demonstrated extended T₁ relaxation times of up to 5 h for hyperpolarized ²⁹Si nuclei in silicon microparticles.^19,23 Interestingly, the presence of unpaired electrons in surface defects allows for hyperpolarization without the addition of exogenous radicals.^24,25 Whiting et al. employed hyperpolarized silicon particles with sizes ranging from 20 nm to 2 μm in the visualization of tumorous tissue.²⁶ Their investigations revealed a size dependence for the recorded T₁ relaxation times, with larger particle sizes tending to display longer T₁ times and stronger DNP enhancements.^27,28 However, the relatively large size and limited solubility of these silicon particles impede studies involving blood circulation.²⁹

In recent years, silicon carbide (SiC) received increased attention as a viable material in biomedicine (e.g., in device fabrication, for implants, coatings, and scaffolds) due to its high biocompatibility and low toxicity.³⁰ SiC may however also have potential for use in hyperpolarization MRI. The divacancy defects in SiC, much like the diamond nitrogen-vacancy center, can be polarized through optically pumped dynamic nuclear polarization.^31,32 Studies by Hartman et al. demonstrate long ²⁹Si and ¹³C T₁ relaxation times of up to 7 h for high-purity bulk silicon carbide.³³ Importantly, the Larmor frequencies of ²⁹Si and especially ¹³C are both within the tuning ranges of commercial multinuclear MRI systems, building on existing imaging infrastructure. SiC therefore presents an interesting, yet hardly explored material for sensitivity-enhanced MRI.

In this work, we study the potential of various SiC samples with different crystal structures (α-SiC and β-SiC), sizes (micro- and nanoparticles), and surface groups for DNP, to evaluate the feasibility of SiC as hyperpolarized MRI agents.

Materials and Methods

Silicon carbide (SiC) with two different crystalline phases, β-SiC and α-SiC, were obtained from four commercial sources. All information below about SiC, such as (average) particle sizes, color, and preparation methods are according to supplier specifications. β-SiC with particle size ranging from 100 to 1200 nm (β-SiC_{100–1200,ST}, gray powder) and α-SiC with particle size ranging from 200 to 1200 nm (α-SiC_{200–1200,ST}, light gray powder) were purchased from STREM Chemicals, Inc. β-SiC samples with average particle sizes (APS) of 18 nm (β-SiC_18,US, gray powder) and 80 nm (β-SiC_80,US, gray powder), as well as one α-SiC with APS of 800 nm (α-SiC_800,US, light gray powder) were purchased from US Research Nanomaterials, Inc. β-SiC_18,US and β-SiC_80,US were prepared by laser synthesis and plasma chemical vapor deposition (CVD), respectively. α-SiC with a size range of 5–250 nm (α-SiC_5–250,PC), prepared via a hot plasma jet process, was purchased from PlasmaChem. α-SiC samples with APS of 650 nm (α-SiC_650,Ab), 750 nm (α-SiC_750,Ab), and 2.2 μm (α-SiC_2200,Ab), prepared via the Acheson process, followed by ball milling, were purchased from ABCR GmbH (Germany). Purities of all SiC samples were >99%; no further purification was carried out on the samples unless stated otherwise. Table S1 summarizes the primary information of all samples provided by the manufacturer, and information about crystalline phase as determined here.

X-ray diffraction measurements were used to determine the lattice parameters of the investigated SiC samples. SiC powder samples were characterized by X-ray diffraction (XRD) on a PANalytical Philips X’Pert MPD diffractometer at room temperature using Cu Kα.

Scanning electron microscopy (SEM) images were recorded on a Zeiss Merlin HR-SEM spectrometer. Dilute suspensions of SiC powder in water were sonicated for 30 min, after which a droplet of the suspension was placed onto a silicon platform. Particle size distributions were determined with ImageJ.

Attenuated total reflectance (ATR)–FT-IR spectroscopy (Bruker Alpha) was used to analyze the surface groups of SiC samples.

Continuous-wave electron paramagnetic resonance (EPR) spectra were recorded at room temperature using an X-band EMX spectrometer (Bruker Biospin GmbH, Germany). The magnetic field was swept from 3100 to 3600 Gauss with 1024 points, at a modulation frequency of 100 kHz, 60 ms conversion time, and 164 ms time constant. The g-value was determined from the zero-crossing in the spectrum. Samples sizes between 0.5 and 2.0 mg were loaded into a glass capillary (OD = 780 μm), closed off with wax. This sample was placed inside the EPR tube. Using these small sizes resulted in minimum retuning between samples and comparable Q-factors, allowing semiquantitative evaluation of the electron density.

Dynamic nuclear polarization (DNP) was performed on a 3.38 T in-house-built polarizer, comprising a pumped bath cryostat,³⁴ an iSpin NMR spectrometer (Spincore Technologies, Florida), and a 94–96 GHz microwave source (Virginia Diodes, Inc., Virginia). A fixed ²⁹Si or ¹³C tuned Alderman–Grant coil made to fit the sample was inductively coupled to an un-tuned saddle coil connected to the spectrometer. For ²⁹Si DNP measurements, about 60 mg of sample powder was tightly packed and enclosed in a poly(tetrafluoroethylene) (PTFE) cup. Microwave irradiation with frequency modulation was performed with a triangular-shaped modulation from 94.6500 to 94.7375 GHz, with a modulation frequency of 500 Hz. Spectra were processed using the ssNake software package,³⁵ using a 500 Hz line broadening prior to Fourier transform. DNP buildup curves were recorded with a series of pulse-acquire experiments with a repetition time (TR) of 60 s and a low flip angle of ∼15°. The DNP enhancement value of ²⁹Si in the solid state was obtained by comparing the integrated signal intensity of the hyperpolarized sample with the thermal equilibrium signal. Thermal polarization was measured using the same setup as used for DNP experiments.

Magnetic resonance imaging and T₁ relaxation measurements were performed on an 11.7 T Bruker BioSpec MRI system using an in-house built three turn solenoid coil tuned to ²⁹Si (∼99.3 MHz). After polarization in the DNP setup, samples were taken out and transferred as a solid into the scanner. T₁ measurements were recorded using a pulse-acquire scheme with a TR of 60 or 120 s and flip angle of approximately 18°. Imaging was performed using a single-shot multiecho RARE (rapid acquisition with relaxation enhancement) sequence with 8 (sagittal) complete images recorded using a single 90° excitation with multiple refocusing. The single-slice images (10 mm thickness) existed of 32 × 32 pixels over a field of view of 64 × 64 mm². With an echo spacing of 2.02 ms and a RARE factor of 32, every 64.5 ms a complete image was acquired, repeated 8 times resulting in spin echo images at 2.02, 66.5, 131.1, 195.7, 260.2, 324.8, 389.3, and 453.9 ms. With this single-shot acquisition, all hyperpolarized magnetization was used at once.

Results and Discussion

Characterization of SiC

Particle size, crystal phase, and surface groups of SiC may affect DNP enhancement and T₁ relaxation time. All tested SiC samples were therefore characterized in detail to confirm the manufacturer-provided properties. XRD was carried out first to confirm the crystal phase of the SiC samples, since crystal structure and defects form a source of electrons that can be used for DNP. Figure 1a shows the XRD patterns of five representative SiC samples. The bottom curve (blue) shows diffraction peaks at 35.6, 41.5, 59.8, 71.8, and 75.4°, which could, respectively, be indexed as the (111), (200), (220), (311), and (222) reflections and well match with a β-SiC phase.^36,37 This sample has a cubic crystal structure and 3C-SiC is the only cubic polytype. The pink curve (SiC_80,US) exhibits an additional weak diffraction peak at 34.1°, usually indicating stacking faults within the crystals.³⁸ The remaining samples (Figure 1a, green, orange, and red curve) feature diffraction peaks at 34.1, 41.5, and 65.5° that are in agreement with an α-SiC phase.³⁹ XRD results of the α-SiC samples are consistent with the 6H-SiC polytype; however, other possible polytypes such as 4H-SiC may also exist.⁴⁰ Notably, α-SiC can consist of many different polytypes due to the different stacking sequences of carbon and silicon atoms in its crystal structure. XRD alone is not enough to distinguish all possible polytypes in SiC samples, further characterization, for example by low-temperature photoluminescence (LTPL) can help to understand the structure of SiC in more detail.⁴¹ At present, XRD measurements mainly display the differences of crystal phase in these SiC samples. Therefore, the five samples are assigned as β-SiC_{100–1200,ST}, β-SiC_80,US, α-SiC_5–250,PC, α-SiC_{200–1200,ST}, and α-SiC_2200,Ab, according to their dominant crystal phases, as well as their particle sizes as provided by the suppliers. Detailed sample information can be found in Figure S1 and Table S1.

Figure 1.

SiC characterization: (a) XRD and (b) FT-IR spectra of representative SiC samples.

SiC samples were further studied using FT-IR spectroscopy to determine the nature of their surface functionality, which may affect hyperpolarization buildup and relaxation times. All β-SiC samples, such as β-SiC_18,US, β-SiC_80,US, and β-SiC_{100–1200,ST} (Figures 1b and S2a), display two intense absorption peaks at 1076 and 785 cm^–1, indicating Si–O–Si bonds (1076 and 467 cm^–1) and Si–C stretching vibrations.^42,43 For α-SiC samples, only α-SiC_5–250,PC shows both significant Si–O–Si and Si–C absorption peaks. The remaining α-SiC samples primarily reveal intense Si–C bonds (770 cm^–1) and only contain a very weak Si–O–Si band at about 1100 cm^–1 (Figure 1b, orange and red curves; Figure S2, pink curve), indicating there are no significant silicon oxide contributions in these α-SiC samples. In addition, the SiC samples are very stable over time, as IR spectroscopy remains unaltered even after 2 years of storage without exposure to air (Figure S2b).

Finally, the particle size distribution of each sample as provided by the suppliers was assessed by scanning electron microscopy (SEM) imaging (representative samples in Figure 2). According to the size provided by the suppliers, sample α-SiC_5–250,PC with a size ranging from 5 to 250 nm has a mean size of 133 nm. The mean sizes of α-SiC_650,Ab (Figure 2b) and α-SiC_2200,Ab (Figure 2d) are found to be 712 nm and 3.6 μm, respectively. The mean size of sample α-SiC_800,US (Figure 2b) is 921 nm and SEM images of other SiC samples are shown in Figure S3. The particle sizes of these samples, as determined by SEM, are not fully consistent with the sizes provided by the suppliers, and we therefore primarily consider size trends, rather than precise particle sizes for the following studies. The morphology of SiC nanoparticles in samples β-SiC_18,US and β-SiC_80,US is spherical. In sample α-SiC_5–250,PC a mixture of differently shaped particles is observed, such as cubic, hexagonal, as well as spherical, presumably due to the hot plasma jet synthesis process. The observed morphologies of other submicron and micron-sized SiC particles are irregularly shaped fragments (Figure 2).

Figure 2.

Electron micrographs of representative SiC involving 200 particles in calculation: (a) α-SiC_5–250,PC, scale bar, 200 nm; (b) α-SiC_650,Ab, scale bar, 1 μm; (c) α-SiC_800,US, scale bar, 1 μm; and (d) α-SiC_2200,Ab, scale bar, 3 μm.

EPR Analysis of SiC

The concentration of defects in the SiC particles determines the number of free electrons that can be directly hyperpolarized. EPR was performed to determine the nature and density of these defects and to obtain the microwave frequency for maximum polarization transfer. Figure 3a shows semiquantitative EPR spectra of selected SiC samples. Interestingly, the densities of free electrons are higher in α-SiC than in β-SiC samples. Notably, the similarly sized samples α-SiC_{200–1200,ST} and β-SiC_{100–1200,ST}, which only differ in crystal phase, displays a ∼2.5-fold difference in free electron density. The g-values of α-SiC_{200–1200,ST} and β-SiC_{100–1200,ST} are 2.0024 and 2.0025, respectively, which are typical of C-related defects (carbon vacancies).^44,45 It is worth mentioning that a different surface composition, such as a silicon oxide layer, may result in decreased densities of free electrons. For example, FT-IR analysis on sample α-SiC_5–250,PC reveals strong contributions by Si–O–Si bonds (see Figure 1b), though this sample exhibits only a relatively weak and shifted EPR signal (g-value ∼2.0573). All other α-SiC particles, such as α-SiC_800,US and α-SiC_2200,Ab, display only minimal contributions of silicon oxides (see Figure S2), but show significant densities of free electrons, at g-values between 2.0024 and 2.0026 (see Figure 3b). Interestingly, all β-SiC samples display strong silicon oxide signals in FT-IR (Figures 1b and S2), but none of these samples contain satisfactory concentrations of free electrons, similar to α-SiC_5–250,PC. Therefore, the free electrons in SiC may originate primarily from divacancies in the crystalline sites, rather than from surface defects. These EPR results indicate that α-SiC without an oxide layer has more potential to be hyperpolarized than β-SiC.

Figure 3.

Semiquantitative EPR spectra of (a) α-SiC_{200–1200,ST} and β-SiC_{100–1200,ST} and (b) β-SiC_80,US, α-SiC_5–250,PC, α-SiC_800,US.

Hyperpolarization of ²⁹Si

DNP profiles against microwave spectra were measured relatively fast, by recording the ²⁹Si NMR signal after only 60 s of irradiation (i.e., not after complete polarization). All α-SiC samples, except for α-SiC_5–250,PC, display obvious DNP signals (see Figure S4). The DNP performance of these α-SiC samples is in good agreement with the aforementioned EPR results, relating high defect densities to strong DNP signals. Figure 4 shows the DNP buildup profile of α-SiC_750,Ab with a maximum enhancement of ∼34 times at 4 K, by comparing the integrals and correcting for the different flip angle used. Interestingly, the enhanced signal has a narrower linewidth than the thermal signal (1.7 and 2.6 kHz full width at half-maximum, respectively; see Figure S5), which suggests that not all ²⁹Si nuclei in the particle have been polarized. It is, however, also found that all β-SiC samples show poor DNP signals, revealing no significant differences in signal intensities when samples are either hyperpolarized or in thermal equilibrium. Detailed information about SiC samples and their DNP performance can be found in Table S1.

Figure 4.

(a) ²⁹Si DNP buildup curve and (b) signal enhancement of α-SiC_750,Ab at 4 K.

Spin-Lattice Relaxation Times of ²⁹Si

According to previous descriptions of SiC, the source of unpaired electrons in SiC are paramagnetic centers in the lattice.³² The ²⁹Si nuclei closest to these paramagnetic centers are polarized first, after which transfer of polarization to nuclear spins at further distance occurs through spin diffusion. Therefore, DNP buildup times and T₁ relaxation times may also depend on particle size or rather crystallite size. In addition to the average particle size of each SiC sample, as determined by SEM, the mean crystallite size of each sample was calculated from the XRD data, using the Scherrer equation (see Table S1).^24,46T₁ relaxation times of α-SiC samples with sufficient DNP signal were recorded at a temperature of 4 K after 1 h of DNP. β-SiC samples and α-SiC_5–250,PC, which display poor polarization enhancements, are not further investigated. Table 1 summarizes the T₁ relaxation times of the α-SiC samples, demonstrating long T₁ relaxation times, with some samples displaying T₁ times exceeding 20 min, comparable to earlier reported values for similarly sized silicon particles.^19,24,26,29 Larger α-SiC particles, such as α-SiC_2200,ABCR with an average particle size of 3.6 μm (from SEM) and a mean crystallite size of 40 nm (from XRD), display considerably longer T₁ times (∼22 min) in comparison to smaller particles, such as α-SiC_650,ABCR with an average size of 712 nm and a mean crystallite size of 20 nm (T₁ ∼ 9 min). T₁ relaxation times and DNP buildup times are not strictly dependent on SiC particle size, as α-SiC_800,US is not the largest particle, but does result in the longest T₁ relaxation time of over 26 min.

Table 1. Sample Information and T₁ Relaxation Time of α-SiC Samples.

sample	mean sizea	mean crystallite size (nm)b	29Si spin-lattice relaxation time T1 (s)
α-SiC650,Ab	721 nm	20.3	535
α-SiC750,Ab	817 nm	28.9	660
α-SiC200–1200,ST	755 nm	30.9	961
α-SiC800,US	921 nm	27.3	1600
α-SiC2200,Ab	3.6 μm	40.2	1315

^aMean size as determined by SEM.

^bMean crystallite size as determined from XRD, using the Scherrer equation.

Hyperpolarization of ¹³C

Considering the structure of SiC with both covalently bound Si and C atoms, the promising ²⁹Si hyperpolarization results prompted us to also investigate hyperpolarization of the ¹³C spin species in this material. Based on the ²⁹Si DNP enhancement results, α-SiC_750,Ab was selected for this evaluation. As shown in Figure 5a, the polarization of ¹³C results in a strong signal amplification, in particular when considering the over fourfold lower natural abundance of ¹³C (1.1%) compared to ²⁹Si (4.7%). The polarization buildup time (Figure 5a) for ¹³C is approximately 13 min when fitted with a monoexponential buildup curve. However, the buildup curve systematically deviated from the monoexponential fit, suggesting the presence of two or more ¹³C pools with different time constants as also reported earlier.⁴⁷ Fitting the data with a bi-exponential function provided a better fit, with buildup time constants of 6.5 and 107 min. When after DNP the signal decay was measured with repeated excitations with low flip angle, one single apparent ¹³C T₁ relaxation time in α-SiC_750,Ab of approximately 33 min (Figure 5b) was observed, which was considerably longer than the ²⁹Si T₁ relaxation time of the same particle.

Figure 5.

¹³C DNP buildup profile (a) and signal sampling during T₁ relaxation (b) of α-SiC_750,Ab in the magnetic field of the polarizer (3.38 T).

Transfer of α-SiC Samples to 11.7 T

The α-SiC_750,Ab sample was selected for transfer to an 11.7 T preclinical MR system. Conventionally, small glassy hyperpolarized molecules with free radicals are dissolved and ejected out of the DNP system with a burst of hot buffer, diluting the solution and increasing the distance between radicals and nuclear spins. Here, our samples could be taken out of the DNP setup as a solid without diluting and transferred to the 11.7 T system in approximately 1 min. For ²⁹Si of α-SiC (6H-SiC polytype) particles the timescale of the relaxation process is independent of temperature and magnetic field.³³ The samples were not actively heated and the temperature was estimated between 100 and 200 K during measurement. Since magnetization typically diminishes rapidly in the absence of a magnetic field, the hyperpolarized SiC nanoparticles were transferred immediately to the 11.7 T MRI system. In a single-shot pulse-acquire spectrum (Figure 6A) at 11.7 T, we observed two ²⁹Si resonances, which were not seen at the low magnetic field (3.38 T) of the polarizer due to a limited spectral resolution and relatively inhomogeneous field. In earlier studies, magic angle spinning NMR spectroscopy on 6H-polytype SiC revealed three ²⁹Si resonances in SiC at 4.7 T at −14, −21, and −25 ppm.^33,48 We only observe two resonances, approximately 11 ppm apart, suggesting that only the two outer resonances of the three are efficiently polarized. This is in agreement with the MAS–NMR-based findings reporting that as soon as any impurities are present, the outer two resonance T₁ times are shorter than the center ²⁹Si peak of 6H-polytype SiC, which can have T₁ times of up to ∼5.5 h at room temperature at 11.7 T.³³ Determining the T₁ relaxation time of this sample is not trivial. Knowing the exact effective flip angle is important to correct the apparent signal decrease for the radiofrequency (RF)-induced decay. Using two experiments at different TR values and an estimated flip angle of 15°, we estimate the T₁ time between 62 and 158 min (Figure 6B). However, as the T₁ time for the two TR settings does not converge to a single value, nearly all signal decay is governed by the used flip angle, instead of T₁ relaxation. If we only fit a flip angle to these curves, neglecting T₁ relaxation, the fits converge to an effective flip angle between 17.5 and 18.1°. With this actual flip angle and repetition times, we could not properly assess T₁, but can safely assume it is longer than 63 min, the calculated value with a 15° flip and 60 s TR.

Figure 6.

(a) Hyperpolarized ²⁹Si spectra and (b) T₁ fit to the signal decay after DNP of α-SiC_750,Ab with two repetition times at 11.7 T.

MR Imaging

In a separate experiment, the hyperpolarized α-SiC_750,Ab sample was transferred to the 11.7 T preclinical MR system to illustrate its potential for imaging the ²⁹Si nuclei. A single slice in sagittal orientation through an Eppendorf tube with the sample produced eight echoes with ²⁹Si signal (Figure 7). The ²⁹Si axial images with six echo times can also be seen in Figure S6.

Figure 7.

²⁹Si MR imaging of hyperpolarized α-SiC_750,Ab at 11.7 T. Echoes 5–8 are depicted at a fourfold smaller scale to illustrate T₂ signal decay with echo number. In the image of echo 7, the size of the Eppendorf tube is indicated.

Fitting monoexponential curves to the signal decay of the center 18 voxels in the multiecho images resulted in an average T₂ relaxation time of 50 ms (SD 13 ms). The multiecho images can be used to calculate a single (weighted) sum image, or after pixel fitting of relaxation decays used to reconstruct images at any given echo time between 0 and approximately 500 ms with improved SNR.

Conclusions

Hyperpolarization of silicon carbide is demonstrated for micro- and nanoparticles with different crystal phases, particle sizes, and surface groups. α-SiC micro- and nanoparticles show strong DNP signal enhancements due to their high densities of free electrons. In contrast, β-SiC particles consistently showed poor DNP enhancements. Both ²⁹Si and ¹³C nuclei in α-SiC particles were successfully polarized in the absence of external radicals, revealing long T₁ relaxation times and strong signal amplifications, enabling background-free MR imaging over prolonged periods of time. The Larmor frequencies of ²⁹Si and ¹³C are both within the tuning ranges of commercial multinuclear MRI systems, greatly facilitating clinical translation. SiC displays high biocompatibility and minimal (cyto)toxicity,^49,50 is chemically inert, and can retain hyperpolarization for several hours, making SiC particles promising for use as high-sensitivity imaging agents for MRI. Further studies are focused on the preparation and selection of uniformly sized SiC nanoparticles, investigation of their biostability and biocompatibility, and modification of their surface with functional groups/ligands for targeted imaging and theranostic applications.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.1c07156.

• NMR properties of all SiC samples, XRD diffraction patterns, IR spectra, SEM images, DNP buildup curves, ²⁹Si NMR spectra of hyperpolarized SiC particles and in thermal equilibrium, and ²⁹Si axial imaging of hyperpolarized SiC particles (PDF)

The authors declare no competing financial interest.