263 GHz EPR of polarizing agentsEPR characterization (giso, and relaxation times T1e and T2e) of the PAs in various solvents was performed with a quasi-optical EPR spectrometer (Bruker ElexSys E780) operating at magnetic fields around 9.4 Tesla. The output power of the microwave (MW) bridge is produced by an amplifier multiplier chain (AMC) and amounts to about 100 mW. Three types of resonators were used depending on the specific requirements of the experiments: a cylindrical TE011-mode resonator for sample volumes 10 – 30 nL (Bruker BioSpin, model E9501610), a cylindrical TE012-mode resonator with sample volumes up to ~50 nL (Bruker BioSpin, model E9501510) and a non-resonant probe for larger sample volumes ( ~ 1–4 µL, Bruker BioSpin, model E9501310). CW EPR experiments of deoxygenated samples for T2e measurements were acquired with the non-resonant probe. CW EPR measurements for g-factor and pulsed EPR experiments for T1e determination were performed with the cylindrical resonators. All experiments were recorded using quadrature detection to enable signal phase correction. Experimental parameters for CW EPR using resonator: MW power ≈ 0.2–0.5 mW (depending on solvent), modulation frequency (MF) = 100 kHz, modulation field amplitude (MA) = 0.01 – 0.05 mT, number of scans (NS) = 1 – 10. Parameters for T2e measurements (Supplementary Fig. 17 and Supplementary Table 5) in the non-resonant probe: MW power ≈ 0.5–5 mW, MF = 100 kHz, MA = 0.1–0.15 mT, NS = 1. The conversion time was 81.92 ms in all CW EPR measurements.T2e of the PAs were extracted from the peak-to-peak CW EPR line widths, as explained in Supplementary Discussion 1. Measurements are shown in Supplementary Fig. 17. To determine the resonance frequency of the PAs, CW EPR spectra were recorded along with g-standards in the same EPR capillary. g-standards were a carbon fiber (giso = 2.0064452), encapsulated 14N in C60 (N@C60, giso = 2.0022, calibrated against the carbon fiber) or deuterated BDPA (BDPA-d27, giso = 2.002553). The determined isotropic giso values of the PAs in various solvents are reported in Table 1. These values were required to calculate the MW frequencies of the gyrotron for DNP pumping.To determine the electron spin-lattice relaxation time (T1e) at 263 GHz, Free Induction Decay (FID)-detected inversion recovery (IR) experiments were performed (Supplementary Fig. 18). Typical parameters of the pulse experiment in liquid solution were: inversion π-pulse = 160 – 200 ns, detection π/2 pulse ≈ 80 – 100 ns, τ ≈ 4 ns – 3.1 μs, dead time = 130 – 150 ns, detection bandwidth (BW) = 200 MHz, shots per point (SPP) = 50, shot repetition time (SRT) = 8 – 15 ms, number of scans (NS) = 4 –5. T1e was extracted by fitting of the recovery time traces as illustrated Supplementary Fig. 18 and explained in Supplementary Discussions 1, the extracted values are reported in Supplementary Table 6.PA concentrations were 1–50 mM and verified by comparison with standards at 9 GHz CW EPR (Bruker E500T). For 263 GHz experiments in the non-resonant probe, samples were degassed by freeze-pump-thaw cycles (3–5) in quartz tubes (I.D. = 1.6 mm, O.D. = 2 mm, Wilmad) and sealed with a flame. Samples for experiments in resonators were filled in quartz capillaries with I.D. = 0.20 mm and O.D. = 0.33 mm (Wilmad). For T1e experiments in resonators and under oxygen exclusion, sample stock solutions were first degassed and transferred to a glove box (MBraun-Unilab Plus, N2 atmosphere, O2 and H2O content ≤0.1 ppm). EPR capillaries were filled inside the box, sealed with sealing rubber (Critoseal®) and stored in a flask under nitrogen atmosphere prior transfer into the resonator. Sealed tubes were inserted into the resonator that was flushed with nitrogen or helium gas. Contamination of samples by oxygen was monitored over minutes up to hours as reported in Supplementary Fig. 18.Saturation factor and microwave field strength B
1e
The effective saturation factor for TEMPONE-15N-d16 (c ≈ 10 – 25 mM) in CCl4 and the MW field strength B1e at the sample position were determined experimentally from the observed signal enhancements combined with the measured T1e and T2e relaxation times, as explained in the following.A saturation factor seff ≈ 0.3 was calculated from the Overhauser equation (eq. 2), using the signal enhancement of CCl4 (\(\varepsilon=120,\) Fig. 1f) and other Overhauser parameters from the literature (ξ ≈ -0.17, f ≈ 0.98)25. A 25% error was estimated for seff based on the experimental error of \(\varepsilon\) and literature data.The MW field strength B1e is related to seff through eq. 15 (Supplementary equations 2), which is valid for a two-spin system such as TEMPONE-15N-d16, where one electron with S = ½ couples to one 15N nucleus with I = ½. Based on the range of measured T1e and T2e (Supplementary Figs. 17–18 and Supplementary Tables 5 – 6), seff was calculated and plotted as a function of B1e (Supplementary Fig. 20). The B1e value that delivered the observed seff amounts to B1e ≈ 0.045–0.065 mT (at PMW ≈ 40 W). This was found consistent with the estimates from electromagnetic field simulations (Supplementary Fig. 3 and Table 3). Because \({B}_{1{{{{{\rm{e}}}}}}}\) is experimentally estimated from \({s}_{{{\mbox{eff}}}}\) based on an NMR measurement, this value represents an average \({B}_{1{{{{{\rm{e}}}}}}}\) over the NMR sample. The method of paramagnetic shift suppression as a function of MW power proposed in ref. 54 is suitable for high PA concentration (c ≈ 100 mM) and s close to unity54, however, it turned out difficult at our experimental conditions with low PA concentrations.DNP/NMR setupThe DNP/NMR setup illustrated in Fig. 1a consists of three main components: 1) a custom-designed gyrotron as a MW source; 2) a commercial 9.4 Tesla magnet and NMR spectrometer; 3) a DNP/NMR probe. The tunable gyrotron (Bruker/CPI, 4.8 T) produces microwaves (second harmonic) at 263.3 ± 0.1 GHz at ≥ 20 W and 263.3 ± 0.25 GHz at ≤ 10 W. Specifications of the gyrotron frequency stability over time as well as the power-frequency profile as a function of hardware parameters are reported in Supplementary Note 2, Supplementary Table 1, and Supplementary Fig. 6. The linearly-polarized MW is transmitted as a free-space Gaussian beam (TEM00-mode) to the DNP/NMR spectrometer via a corrugated waveguide (I.D. = 19.3 mm, Bruker BioSpin/CPI, mode HE11). The waveguide consists of two segments separated by a quasi-optical bench (Bridge 12). The bench controls the power and polarizations of the MW. The bench also includes a mechanical shutter (Vincent Associates, 35 mm aperture, switch on/off time of 3 ms) and a water-cooled Teflon® absorber orthogonal to the MW pathway to absorb the reflected MW when the shutter is closed.The NMR spectrometer consists of a commercial wide-bore 9.4 T magnet equipped with an NMR console (Bruker BioSpin Avance Neo). A liquid NMR probe (Bruker, base frequency for 13C: 100.4149 MHz; for 1H: 399.3090 MHz), with a normal coil configuration (inner heteronucleus coil, shared with 2H channel; outer 1H coil), was adapted here to permit MW irradiation of the sample. A corrugated waveguide (O.D. = 7.6 mm, Thomas Keating Ltd) was inserted into the central pillar of the NMR probe and coupled to the larger corrugated waveguide from the gyrotron (I.D. = 19.3 mm) via a waveguide taper (Thomas Keating Ltd). At the end of the corrugated waveguide in the probe, the 263 GHz beam is transmitted through free-space onto the sample over a combination of four mirrors and across the NMR coils (Fig. 1a, b), irradiating the sample tube from the side. The mirrors (Thomas Keating Ltd) were specifically designed to extend the electromagnetic beam waist over a sample area of about 80 mm2 (Supplementary Figs. 2, 3). To allow for penetration of 263 GHz MW, the liquid sample is confined into a thin layer of thickness d ≈ 25–75 μm, formed by two concentric quartz tubes. The I.D. of the outer tubes was 4.2065 ± 0.0065 mm (Wilmad Labglass 528(or 535)-PP-7QTZ) and the O.D. of the inner tube was ~4.059 – 4.158 mm (Hilgenberg). This combined with the effective irradiated length of the tube (length ~20 mm, Supplementary Figs. 2 and 3) results in an effective volume of \(\lesssim\) 20 µL. This setup allows for tuning d according to the MW absorption coefficients of various solvents and also for preserving the cylindrical arrangement of the sample for optimal shimming. Experimental characterizations of MW beam shape and polarization are given in Supplementary Fig. 2.Sample temperature is controlled using a flow of cold nitrogen gas, which is passed through a Dewar containing liquid nitrogen, at constant flow rates (1000–1400 liters per hour). As the shortest MW irradiation time implemented in this study is on the order of seconds (on the order of T1n), slow sample spinning at 20 Hz (a standard capability of a Bruker NMR probe) is introduced to improve homogeneity of the MW irradiation experienced by the sample, as well as to reduce the temperature gradient in the sample. Sample spinning might produce side bands, however so far, we have hardly observed spinning sidebands likely because the volume is still restricted to ~20 μL and the overall SNR is in most cases too low. More details about sample characterization and temperature are given in the next sections.Characterization of the microwave path in the DNP probeBeam alignment and shape at the sample were characterized by imaging the beam spot on a liquid crystal sheet (Edmund Optics Ltd.). For this, the probe was connected to the gyrotron and the last (M4) mirror was replaced with a camera. Liquid crystal sheet that changes color upon heating was placed at the sample position between the NMR coils, and the gyrotron was set to operate at low power \({P}_{{{{{{\rm{MW}}}}}}}\approx 5-10\) W. The observed beam spot indicated that the beam is aligned with the sample and can be expanded over the accessible sample space between the coils, as intended by design (Supplementary Fig. 2b–d).The MW beam polarization in the probe was examined as well. A polarizing wire grid was mounted inside the probe replacing the last mirror (M4) (Supplementary Fig. 2f). Using a vector network analyzer (VNA) as a low frequency source (9.7 GHz) that was fed into the quasi-optical bridge of the 263 GHz EPR spectrometer (Supplementary Fig. 2e), low-power 263 GHz MW irradiation was generated and directed to the probe. The reflected signal from the wire grid was down-converted (263.3 GHz to 9.7 GHz) and read out by the VNA as the S11 (reflection) parameter. A metal plate, placed at the entrance of the probe, was used as a reference for the S11 maximum. The observed S11 against the grid rotation angle shows that the polarization is preserved throughout the probe and B1e is orthogonally oriented to the probe axis at the sample position, and thus to the external B0 field in the NMR magnet (Supplementary Fig. 2g). Power losses over the entire transmission line, including mirrors, were measured to be ~0.8 dB at mirror 3 (replaced by a reflector) and ~6.5 dB (single path including the NMR coils, and the coil support quartz tubes, but without the sample tube assembly) at mirror 4. These values are close to electromagnetic field calculations (following section), which predict losses of ~0.7 dB and ~4.6 dB at M3 and M4 (assuming a 563 mm long wave guide), respectively.Electromagnetic field simulationsElectromagnetic field simulations were performed to predict the B1e distribution over the sample using a TLM (Transmission Line Matrix) Time-Domain Solver of CST Microwave Studio® software (Dassault Systèmes). The 3D model of the probe consisted of a short, corrugated waveguide (75 mm length), four MW mirrors, two NMR coils, support and the sample-assembly quartz tubes, as well as a specific sample solvent. We used a linearly polarized Gaussian beam (TEM00-mode) at the entrance of the corrugated waveguide with an input power of 10 W and in the frequency range of 260.0-263.6 GHz. The MW polarization at the input MW of the probe was set to reproduce the beam polarization delivered by the gyrotron.The thickness d of the sample layer was modeled as set in the experiment for the solvent type (\(d \, \approx \, 75\) µm and \(d \, \approx \, 25\) µm, for CCl4 and H2O, respectively). The values of the complex dielectric constant for quartz and CCl4 at 263 GHz were extrapolated from low-frequency data (CST Microwave Studio® material library and https://cem.co.en/microwave-chemistry/solvent-choice, respectively) assuming the Debye model for frequency dependence. The resulting real parts \({\epsilon }_{{{\mbox{r}}}}\) and corresponding tan δ values were 3.74, 2.17 and 0.00037, 0.00041 for quartz and CCl4, respectively. For H2O, the high-frequency (264 GHz) experimental values \({\epsilon }_{{{{{{\rm{r}}}}}}}=5.36\) and tan δ = 1.20, tabulated for T = 30 o C, were employed55. Extrapolation of dielectric constants by CST Microwave Studio® was validated by calculating the absorption coefficient \({{{{{\rm{\alpha }}}}}}\) and comparison with literature values for propagation of a Gaussian beam (\({P}_{{{{{{\rm{MW}}}}}}}=10\,{{{{{\rm{W}}}}}},\,\nu \, \approx \, 263.3\) GHz) through H2O, CHCl3, and CCl4. Damping of the power density along the propagation axis of the beam was fitted to a Lambert-Beer law (\(I={I}_{0}\exp \left(-\alpha t\right)\)) to obtain \(\alpha\). Extracted values of \(\alpha\) are listed in Table 2 and are consistent with literature data55,56,57. The number of mesh cells was in the range of 6.3 – 6.7×107 depending on the sample and tube properties (size and permittivity). To simplify the model, lossy metals were simulated as perfect electrical conductors (PEC). Simulations were performed using GPU-acceleration. The results from the simulations are displayed in Supplementary Fig. 3.Samples preparation for DNPAll solvents were purchased and used as received. Chloroform-13C, 2,6-di-tert-butyl-α-(3,5-di-tert-butyl-4-oxo-2,5-cyclohexadien-1-ylidene)-p-tolyloxy (galvinoxyl), 4-Oxo-2,2,6,6-tetramethylpiperidine-d16,1-15N-1-oxyl (TEMPONE-15N-d16), ethyl acetoacetate (EAA), ethyl acetoacetate-1,2,3,4-13C4, chlorobenzene-13C6, bromobenzene-13C6, iodobenzene-13C6, phenylacetaldehyde, ST034307, α,α,α-trifluorotoluene, diethyl fluoromalonate, decafluoropentane, and flutamide were obtained from Sigma-Aldrich. Amiodarone hydrochloride, sodium diatrizoate and, mitotane were obtained from Fisher Scientific. Trans-2-hexenyl acetate was obtained from TCI chemicals. 1-fluoro-4-iodobenzene was obtained from Fluka. Fluorobenzene was obtained from EGA. Fluorobenzene-13C6 was obtained from Cambridge Isotope Laboratories. For all DNP experiments, commercial 5.0 mm O.D. quartz tubes were used (Wilmad Labglass). Solutions of polarizing agent ( ~35–75 μL) were degassed within the NMR tube by at least five freeze-pump-thaw cycles. Samples were then transferred to a nitrogen filled glovebox (MBraun) where a quartz insert ~4.1 mm O.D. (Hilgenberg) was inserted and the tubes were capped using airtight caps made in-house. EPR spectra were then recorded at 9 GHz (X-band, Bruker Elexsys E500T, Elexsys high sensitivity probe), line intensities and width were compared to standards to verify the PA concentration and check deoxygenation. Typical 13C and 19F DNP samples were prepared with c(PA) ≈ 10 – 100 mM and c(target) ≈ 200–500 mM. The p-cymene samples were prepared by dissolving the PA in neat p-cymene. In the case of flutamide, we used a target concentration of ~10 mM and ~330 mM of DMSO to account for the poor solubility of the compound. Radical concentrations were optimized for the best compromise of signal enhancement and NMR line width of the target molecule.Overhauser DNP experimentsBefore MW irradiation, the sample temperature was stabilized at reduced temperature (210–270 K) under a flow of cold nitrogen gas. The effective sample temperature was verified based on the chemical shift difference of the solvent signals, as determined in a separate experiment, using a sample containing no radical, but otherwise identical composition. Depending on the solvent, the temperature dependence of the chemical shift difference was monitored over a range of 70–90 K without MW irradiation and compared to the chemical shift difference during CW MW irradiation at different MW power (Supplementary Fig. 4a–b) or varying MW irradiation time (Supplementary Fig. 4c–e). From the comparison, temperature settings for the specific solvents were obtained. For DNP we applied MW irradiation on-resonant with the PA low field line (Fig. 1c). By adjusting the gun current, the cavity temperature, the collector current and the cathode voltage of the gyrotron (Supplementary Note 2), the MW frequency was first set at the resonance condition estimated from the isotropic g-factor of the PA and then kept fixed (Table 1).Magnetic field homogeneity was optimized by shimming, and the lock-field ( ± 1.1 mT) swept to record a DNP enhancement field profile. The lock-field was then set to the maximum of the profile, corresponding to the low-field EPR line of the PA. All NMR spectra were recorded without field locking. The NMR probe was tuned and matched to nuclei under investigation. All experiments were at least duplicated under similar experimental conditions. From this, the error in enhancement was estimated to be ~10%. In cases where the signal-to-noise ratio of the Boltzmann spectrum was poor, an increased uncertainty of ~15% was assumed.Pulse-acquire experiments were performed with 10.5 μs (~41 W, 25.5 kHz) and 14.7 μs (~21 W, 17.01 kHz) high power 90° pulses for 13C and 1H, respectively, whereas 1H decoupling was performed with WALTZ16 using a nominal 90 μs pulse ( ~0.58 W, 2.78 kHz). If not noted otherwise, 1D 13C DNP experiments were performed with 1H decoupling during acquisition but without 1H pre-saturation.19F pulse acquire experiments were carried out by retuning the 1H NMR coil to the resonance frequency of 19F ( ~375 MHz) with a 17 µs high power pulse ( ~21 W, 14.71 kHz). All other experimental details were similar to 13C 1D DNP experiments.DNP enhancements were calculated as the ratio of the absolute integral of the signal in the absence (Boltzmann signal) and presence (DNP) of MW irradiation. Spectra were processed with the same parameters (exponential line broadening, zero-filling, zero- and first-order phase correction). When the number of scans between these two experiments was not the same, the absolute integrals were multiplied/divided accordingly. For samples, in which the Boltzmann signal was too weak to be recorded in the DNP-tube, the Boltzmann spectrum was collected in a 5 mm regular NMR tube, and intensities were scaled by the volumetric ratio.The recycle delay (RD) was set to ~1.2–5 times T1n and no dependence of the enhancement on RD was observed. For measurements with H2O as a solvent, a longer RD of max. 30 s was used to account for sample heating.Two-dimensional NMR/DNP experimentsDNP spectra of all 2D correlation measurements were collected under CW MW irradiation at output power of \({P}_{{{\mbox{MW}}}} \, \approx \, 40\) W. The 13C TOCSY experiments were performed a slightly modified version of a standard pulse sequence provided in the TopSpin software package (dipsi2ph), which was modified by including broad band decoupling on 1H during acquisition, and with incorporation of an additional 180o pulse on 1H during the t1 period (Fig. 3b). FLOPSY-16 and DIPSI-2 spin locks, both with ω1/2π = 10 kHz and zero-quantum filters of δ1 = 2 ms and δ2 = 3 ms58 are used to account for the different chemical shift dispersion of EAA ( ~17.7 kHz) and iodobenzene-13C6 ( ~4.5 kHz). Carrier frequencies of the spin-lock were set at the center of the chemical shift ranges.13C-TOCSY spectra of EAA were collected on a CCl4 solution containing 500 mM target molecule and 25 mM TEMPONE-15N-d16. Spectra were collected with \(\tau\)mix = 9.4 ms spin-lock, 8 (OE-DNP) and 16 (Boltzmann) scans, 16 dummy scans, 6 s recovery delay, 32768 × 128 complex points, zero-filled to two times the number of data points, and processed with a shifted sine-bell filtering in both dimensions. The 13C-TOCSY spectra of iodobenzene-13C6 were collected on a cyclohexane solution containing 500 mM target molecule and 25 mM TEMPONE-15N-d16. Spectra were collected with \(\tau\)mix = 34.5 ms spin-lock, DIPSI-2 spin lock, 8 scans, 16 dummy scans, 6 s recovery delay, 4096 × 256 points, zero-filled to four times the number of data points, and processed with a shifted sine-bell filtering in both dimensions. Corresponding 1D spectra of EAA were collected with 32 scans, 32768 points, 30 s recovery delay, and processed with 2 Hz exponential broadening. 1D spectra of iodobenzene-13C6 were collected with 8 scans, 16384 points, 30 s recovery delay, and processed with 2 Hz exponential broadening.The DNP-enhanced 13C-INADEQUATE spectrum of neat p-cymene (6.4 M) with 100 mM TEMPONE-15N-d16 as PA was obtained as a summation of five spectra, each recorded with 512 scans and 16 dummy scans. Each spectrum was collected with 3 s recovery delay, 2048 × 64 points, zero-filled to four times the number of data points, and processed with a shifted sine-bell filtering in both dimensions. Inter pulse delay for the double quantum evolution was set to \(\tau\) = 5.55 ms and optimized for observing 13C homonuclear scalar-coupling with JCC = 45 Hz. To keep track of the sample condition, 1D pulse-acquire measurements were performed after each of the five 2D spectra (Supplementary Note 4, Supplementary Fig. 15).The DNP-enhanced 13C-INADEQUATE spectrum of 1-fluoro-4-iodobenzene (natural 13C abundance) was collected on a cyclohexane solution containing 1.5 M target molecule and 25 mM TEMPONE-15N-d16 and with 288 scans. The reference Boltzmann spectrum was collected with the same target and PA concentration as the DNP measurement in a regular NMR tube with total volume of ~0.6 mL and with 128 scans. In both cases, spectra were recorded with 16 dummy scans, 6 s recovery delay, 2048 × 128 points, zero-filled to four times the number of data points, and processed with a shifted sine-bell filtering in both dimensions. Inter-pulse delay for the double quantum evolution was set to \(\tau\) = 4.5 ms and optimized for observing 13C homonuclear scalar-coupling with JCC = 55 Hz. WALTZ-16 and WALTZ-64 broadband heteronuclear decoupling59 (ω1/2π = 2.78 kHz) were implemented for the 13C-TOCSY and 13C-INADEQUATE experiments, respectively.
Overhauser enhanced liquid state nuclear magnetic resonance spectroscopy in one and two dimensions
