High-Resolution Overhauser Dynamic Nuclear Polarization Enhanced Proton NMR Spectroscopy at Low Magnetic Fields

Abstract

Dynamic nuclear polarization (DNP) has gained large interest due to its ability to increase signal intensities in nuclear magnetic resonance (NMR) experiments by several orders of magnitude. Currently, DNP is typically used to enhance high-field, solid-state NMR experiments. However, the method is also capable of dramatically increasing the observed signal intensities in solution-state NMR spectroscopy. In this work, we demonstrate the application of Overhauser dynamic nuclear polarization (ODNP) spectroscopy at an NMR frequency of 14.5 MHz (0.35 T) to observe DNP-enhanced high-resolution NMR spectra of small molecules in solutions. Using a compact hybrid magnet with integrated shim coils to improve the magnetic field homogeneity we are able to routinely obtain proton linewidths of less than 4 Hz and enhancement factors > 30. The excellent field resolution allows us to perform chemical-shift resolved ODNP experiments on ethyl crotonate to observe proton J-coupling. Furthermore, recording high-resolution ODNP-enhanced NMR spectra of ethylene glycol allows us to characterize the microwave induced sample heating in-situ, by measuring the separation of the OH and CH₂ proton peaks.

Graphical Abstract

Introduction

Dynamic nuclear polarization (DNP) is a powerful technique for dramatically increasing the sensitivity of nuclear magnetic resonance (NMR) experiments.^1–7 DNP makes use of the large magnetic moment of the electron spin to hyperpolarize surrounding nuclei. Through microwave irradiation, the polarization is transferred from the electron spin to nuclei in the sample through various mechanisms – most commonly the solid effect,^8,9 cross effect,^10,11 and Overhauser effect.^12–16 In liquids, the dominant mechanism for DNP is the Overhauser effect, where hyperpolarization upon microwave irradiation results from a difference in zero quantum (ZQ) and double quantum (DQ) relaxation rates.

Following the classical description of Hausser and Stehlik, the Overhauser effect results in an NMR signal enhancement given by:¹⁷

$$\epsilon =1-E=\xi fs\frac{\left|{\gamma }_e\right|}{{\gamma }_H},$$ (1)

with, $\epsilon$ the amount of polarization transferred, $E$ the enhancement, $\xi$ the coupling factor, $f$ the leakage factor, $s$ the saturation factor, ${\gamma }_e$ the electron gyromagnetic ratio, and ${\gamma }_H$ the proton gyromagnetic ratio.

The leakage factor accounts for nuclear relaxation processes which don’t involve the electron spin. It can be determined experimentally from the longitudinal nuclear relaxation rates with ($T_1)$ and without a radical ($T_{10})$ and is given by:

$$f=1-\frac{T_1}{T_{10}}$$ (2)

The saturation factor is a measure of the extent to which an electron paramagnetic resonance (EPR) transition is saturated. The maximum enhancement is obtained when the leakage factor and saturation factor are both equal to 1. For nitroxides, the saturation factor is further complicated by the electron-nitrogen hyperfine coupling. This results in the maximum possible saturation factor varying from 1/3 to 1 depending on the exchange rate between the ¹⁴N hyperfine lines (or ½ to 1 in the case of ¹⁵N labeled nitroxides).^18,19 The exchange rate is influenced by Heisenberg exchange and ¹⁴N spin relaxation.¹⁹ In low viscosity solvents, at concentrations above 100 mM the nitroxide spectrum collapses into a single line due to Heisenberg exchange and the maximum saturation factor approaches 1.^20–22

The coupling factor contains information about the system dynamics. In a typical measurement, enhancements are measured as a function of the microwave power and the maximum enhancement is calculated from extrapolating to infinite power. By assuming, or experimentally determining the maximum saturation factor, the coupling factor can be determined. The coupling factor can vary from −1 in the case of pure scalar coupling to +0.5 in the case of pure dipolar coupling.²³

In cases where the $T_1$ and $T_{10}$ vary substantially with microwave power due to sample heating it is convenient to express the enhancement by rearranging equation (1):^21,24,25

$$1-E\left(p\right)=C{k}_{\sigma }{T}_1\left(p\right)s\left(p\right)\frac{\left|{\gamma }_e\right|}{{\gamma }_H}$$ (3)

where $C$ is the spin label concentration and $k_{\sigma }$ is the local dipolar cross-relaxation rate. This expression also makes explicit the quantities that depend on microwave power, $p$. In this approach, $k_{\sigma }$ is the quantity to be determined. This description has the advantage of being expressed in terms of variables that are measurable or theoretically calculable. In the classical description given in equation (1), any consideration of sample heating is complicated by the non-linear dependence of the leakage factor. This is simplified in equation (3), where the microwave power dependence of $T_1$ is included in the model. For aqueous samples, where the microwave heating is significant, it is most convenient to express the enhancement using equation (3).

For nitroxides and protons in liquids, the dipolar interaction dominates the coupling factor resulting in a maximum possible enhancement of −330. While the coupling factor is favorable at X-band (9.5 GHz, 0.35 T), e.g. 0.27 in the case of TEMPOL in water,²¹ the coupling factor dramatically decreases at higher microwave frequencies. For example, the coupling factor for the same sample of TEMPOL in water is 0.01 at G-band (260 GHz, 9.2 T).²⁶ This greatly limits the enhancements at high magnetic field strengths and has limited the dissemination of high-resolution proton ODNP-enhanced NMR spectroscopy. However, recent work has shown very high ODNP enhancements on other nuclei, for example ¹³C.^27,28 Additionally, high-resolution ODNP has been performed at X-band on ¹⁹F nuclei.²⁹ Nuclei such as ¹³C and ¹⁹F typically show a larger chemical shift dispersion compared to protons, and high-resolution NMR experiments are significantly less challenging.

In this work, we take advantage of the large coupling factor at X-band and therefore an efficient ODNP effect, to greatly improve the signal-to-noise in a ¹H NMR experiment while still recording NMR spectra with high chemical shift resolution. Currently, most ODNP systems are home-built and rely upon commercial X-band electromagnets designed for EPR spectroscopy. These magnets do not provide the homogeneity necessary for high-resolution NMR measurements. In addition, standard EPR magnets do not incorporate shim coils, a necessity for high-resolution NMR experiments. While shim coils have been the standard in NMR spectrometers since the 1960s, they are not standard in an X-band ODNP spectrometer. More recently, various shim methods have been incorporated into compact NMR spectrometers for the purpose of improving resolution in mobile systems.^30–35

Our work has been performed using a compact hybrid magnet made from a permanent magnet with additional sweep coils. The ability to perform NMR spectroscopy in a compact permanent magnet has numerous advantages.^36–39 Permanent magnets can be used in highly mobile applications for on-site NMR experiments.⁴⁰ They are cost-effective compared to superconducting high-field magnets and they eliminate the requirement for cryogens.⁴¹ The challenges associated with permanent magnets are the strong temperature induced magnetic field drifts and the overall lower magnetic field strength compared to superconducting magnets, resulting in a reduced NMR sensitivity due to a decreased Boltzmann factor and spectral resolution. This is especially challenging for complicated multiplet structures and strong coupling effects can further complicate spectra.

It was first proposed that ODNP could be used for enhancement of weak NMR signals and the study of molecular distances and associations.^42,43 The first experimental application of ODNP spectroscopy was to study molecular dynamics where interpretation of the coupling factor can provide the translational and rotational correlation times of molecules.^44–46 Examples are the molecular dynamics of benzene, toluene, and ethers.⁴⁷ As NMR spectroscopy moved to higher magnetic field strengths, numerous studies used flow and shuttle systems to hyperpolarize at low magnetic fields and quickly transfer the sample to high magnetic fields to perform high-resolution NMR experiments.^48–50 For the application of enhancing NMR sensitivity, it was also shown that crude oil can be hyperpolarized using the endogenous radicals inherent to the samples.⁵¹ In this context ODNP has also been applied to NMR relaxation dispersion experiments.^52,53

In recent years, the main application for X-band ODNP spectroscopy has been the measurement of hydration dynamics on bio-macromolecular surfaces.⁵⁴ This method has shown promise in elucidating the role of biological water which is now recognized as an active constituent in many bio-molecular processes.^55–58

The primary motivation for this work is to demonstrate that even at low magnetic field strengths (e.g. 0.35 T corresponding an 1H Larmor frequency of 14.5 MHz) chemical shift information can be extracted from ODNP enhanced NMR spectroscopy. This method offers not only an improved sensitivity, but also raises the possibility of extracting dynamical information from different chemical sites.

Materials and Methods

Chemicals:

4-Oxo-2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPONE), toluene, ethyl crotonate, acetylsalicylic acid (Aspirin), 4-hydroxy-2,2,6,6-tetramethylpiperidine 1-oxyl (TEMPOL), were purchased from Sigma-Aldrich. Ethylene glycol was purchased from Honeywell Fluka. Deuterated chloroform (DLM-7TB-50S) was purchased from Cambridge Isotope Laboratories, Inc. All chemicals were used without further purification.

Sample preparation:

The nitroxide radical concentration was 10 mM for all samples. For samples with low dielectric losses (ethyl crotonate, toluene, and d-chloroform) 8 μL was loaded into a 0.98 mm ID, 1.00 mm OD quartz capillary (Hampton Research, HR6–146). For samples with high dielectric loss (ethylene glycol) 3 μL was loaded into a 0.60 ID, 0.84 OD quartz capillary (VitroCom, CV6084-Q-100). For experiments on ethylene glycol, TEMPOL was used as the polarizing agent. All other experiments were performed using TEMPONE.

ODNP Spectrometer:

An ODNP spectrometer requires a high-power microwave source, a microwave resonator with integrated NMR coil, an NMR spectrometer, and a magnet. The microwave source is home-built with a maximum output power of 10 W over a frequency range of 9.3 to 10.5 GHz. A home-built, dielectric resonator operating in the TE₀₁₁ mode at a frequency of 9.38 GHz with integrated NMR coil was used in all experiments. The unloaded quality factor Q of the resonator is 6900, while the loaded Q remains above 4000. While the maximum output power of the microwave source is 10 W, we measured 1.5 dB losses due to cables and connectors. All microwave powers given throughout the article correspond to the estimated microwave power levels at the position of the sample. The microwave source has both a forward (Tx) and reverse (Rx) power monitor. The detection is based on using a directional coupler and zero-biased Schottky diodes (Herotek, DZM series). The diode signal is analyzed by the embedded microcontroller and the microwave power is disabled automatically if the reflected power exceeds a predefined value. The resonator is coupled by means of an iris coupled waveguide. To optimize coupling to the resonator, the reflected power from the cavity was monitored and minimized by adjusting an iris screw.

NMR experiments were performed using a Kea2 spectrometer (Magritek), with an external RF amplifier (MiniCircuits, model ZHL-32A+). Due to the limited RF power (approximately 2 W), the observed NMR pulse length for a 90°-pulse was about 30 μs.

A home-built hybrid magnet consisting of NdFeB permanent magnet (B₀ = 350 mT) and a solenoid sweep coil, with a sweep range of +/− 40 mT was used for all experiments. The magnet included an electric shim system to correct for small inhomogeneities of the magnetic field at the position of the sample. Shim coils are fabricated from printed circuit boards mounted to the face of the magnet poles and included the zonal correction coils Z1 and Z2 and the tesseral correction coils X and Y. The physical dimensions of the coils were determined following the procedure outlined by Anderson.⁵⁹ Bench-top power supplies (Matsusada, model R4K-80) were used to drive the current of the shim system. Although the power supplies used here are switching power supplies, we have no indication that the observed linewidth is negatively influenced by the nature of the power supply. The native linewidth of a water sample without shimming the magnet was 110 Hz (8 ppm). With shim-coils engaged, it was possible to achieve better than 4 Hz (0.3 ppm) linewidth (see Figure S1). A detailed description of the instrumentation including a detailed discussion of the magnet design and the shim system will be subject to a separate publication.

The entire ODNP system was held at a temperature of 34 °C. This temperature was chosen because the magnet temperature reaches 30 °C with the solenoid sweep coils engaged. The magnet temperature is then slightly elevated to decrease the influence from fluctuations in room temperature. To cool the sample, dry air was continuously flowed through the resonator at a rate of 5 L/min.

ODNP Experiments:

The magnetic field strength was adjusted using the solenoid coils of the hybrid magnet assembly to yield maximum ODNP enhancement, corresponding to a proton NMR frequency of 14.24 MHz. Experiments were performed using continuous wave (cw) microwave radiation. The FID was acquired with 16384 points for a total acquisition period of 1.64 s. For aspirin measurements, the repetition time was 5 s. For all other measurements, a repetition time of 20 s was used.

Data Processing:

All spectra were processed similarly. For spectra of toluene and ethyl crotonate, the FID was apodized with an exponential decay corresponding to a 2 Hz Lorentzian linewidth. For spectra of aspirin and ethylene glycol a 10 Hz apodization function was used. The FID was zero-filled to 4 times the original length and Fourier transformed. Spectra were referenced based on known chemical shift values taken from the literature.

For ODNP-enhanced spectra, where the NMR signal is inverted, the spectra are phased positively. To automatically phase the spectra, a phase angle was calculated from taking the arctangent of ratio between the sum of imaginary and real components. zeroth order phase was sufficient to properly phase all spectra.

All spectra except for aspirin were acquired in a single shot. To account for any drift in magnetic field over the course of signal averaging, the individual scans were overlaid by maximizing the FFT cross-correlation of each scan with a reference scan. This method shifts the spectra in discrete points and calculates the cross-correlation function between the two spectra for each step. The maximum correlation corresponds to the optimum overlap (maximum of the cross-correlation function) between the spectra. A number of methods exist for alignment of NMR spectra.^60,61 The cross-correlation method was chosen because of its simplicity and robustness. All processing was performed by in-house python scripts. All linewidths given throughout the text are measured as full width at half maximum (FWHM).

Results and Discussion

In the following section we present ODNP enhanced NMR spectra of different small molecules to demonstrate several applications.

Toluene.

In Figure 1 we show ODNP spectra of a sample of toluene with 10 mM TEMPONE with and without microwave irradiation. Two peaks can be observed at chemical shifts of 2.11 and 7.01 ppm corresponding to resonances of the methyl group (labeled A) and aromatic (labeled B) protons, respectively. The spectrum is referenced using the known chemical shift of the toluene methyl peak at 2.11 ppm. While no internal reference was used for this spectrum, the distance between the methyl and aromatic peaks agrees well with literature value of 4.91–5.02 ppm for toluene in benzene.⁶²

Figure 1.

Left: Single shot ODNP spectra of toluene with 10 mM TEMPONE, with and without microwave irradiation. The enhanced spectrum in red has been phased positively. Spectra have been offset for clarity. Right: Enhancement curves as a function of microwave power for toluene for the methyl (A, orange) and aromatic (B, green) protons.

From the spectrum a linewidth of 3.9 Hz (0.27 ppm) was obtained for the methyl protons. The Lorentzian character of the lineshape for the methyl peak was characterized by calculating the ratio of the 50% linewidth to the 0.55% linewidth. The ratio of these linewidths was calculated to be 20, which is larger than the expected value of 13.5 for a purely Lorentzian lineshape.⁶³ The broader baseline NMR linewidth indicates the presence of higher order field inhomogeneities in the system.⁶⁴ We find that a significant contribution to our observed linewidth is not a result of static magnetic field homogeneity, but rather small fluctuations in the magnetic field on a timescale relevant to the FID. The fluctuations are attributed to coupling of the magnet to the environment and improved shielding of the magnet would be necessary to obtain narrower linewidths. Another source of line broadening is the temperature induced field drift of the NdFeB permanent magnets. We determined the temperature coefficient of the magnet to be 662 +/− 11 ppm/°C (see Figure S2). Therefore, a temperature change of 1 mK during the acquisition of the FID will result in significant line broadening and (phase) distortion of the NMR signal. These small temperature fluctuations become significant when the acquisition time of the FID exceeds 100 ms. Since we generally observe a linewidth well below 14 Hz (1 ppm) we estimate the temperature fluctuations during acquisition of the FID to be well below 1 mK.

In this study, we have chosen a radical concentration of 10 mM to prevent significant line broadening due to paramagnetic induced relaxation. However, if larger enhancements are desired, concentrations of 20 mM or 50 mM TEMPONE can be used, at the cost of slightly larger linewidths. Concentrations of 100 mM and above would result in noticeable broadening of the NMR lines (see Figure S3). At this point, the reduced nuclear relaxation time T2 becomes the dominant factor limiting the spectral resolution not the stability of the external magnetic field. Thus, for cases with highly limited sensitivity, it is possible to compromise on linewidth for increased enhancements. We also note that while for some experiments a microwave power of 7.1 W was used, we do not observe lineshape distortions caused by dielectric heating.

The observed ODNP enhancement ($E$) for toluene as a function of microwave power is shown in Figure 1 (right). Extrapolating to infinite power, we report a maximum enhancement $E_{max}$ of −50 +/− 1 and −45 +/− 1 for ring and methyl protons, respectively. Different enhancements for ring and methyl protons have been observed experimentally^22,65,66 and predicted by MD simulations.⁶⁷ The difference in enhancements is attributed to (1) a larger leakage factor for ring protons caused by a greater T₁ for ring protons in the absence of radical⁶⁵ and (2) a slightly larger coupling factor expected for ring protons due to differences in the average number density for each type of proton around the nitroxide radical.⁶⁷ While this effect was already observed at X-band, this is the first time that this effect can be studied with clear chemical shift resolution between the two peaks.

Differences in enhancement factors for different molecular sites can potentially complicate quantitative ODNP enhanced NMR experiments. This presents a challenge for reaction monitoring and metabolomics studies, however, if the enhancements for different sites are reproducible, it should be possible to calibrate for different enhancements and recover the quantitative information. Our work provides a method to study the enhancements for different molecular sites. Theoretical studies can also provide insight into the molecular basis for enhancements.^26,67,68 More experimental and theoretical work is required to address the feasibility of quantitative ODNP enhanced NMR.

In general, our calculated values for $E_{max}$ appear to be lower than values reported in the literature.⁶⁵ This is primarily attributed to lower microwave induced sample heating. The dielectric resonator used in this study has a higher conversion factor and a lower electric field at the sample position. Additionally, our experiments use ¹⁴N TEMPONE radical, which will have a lower saturation factor than ¹⁵N TEMPONE. Another consideration is none of our samples were degassed to remove the effect of paramagnetic oxygen. Degassing the sample would increase $T_{10}$ and result in a larger enhancement. Finally, we know from electromagnetic simulations that the microwave induced magnetic field inside the resonator, that the B₁ shows some inhomogeneity across the sample (data not shown). This will result in a lower overall enhancement factors, since regions with lower B₁ intensities still contribute to the spectrum but exhibit lower overall enhancement, while the signal recorded without microwave radiation (off-signal) will be constant across the length of the sample. This will lead to non-uniform enhancement values across the sample volume. However, for ODNP measurements aimed at hydration dynamics this is not a problem, since measurements are performed at different microwave power levels and the $E_{max}$ values is calculated as the asymptotic value assuming $B_1\to \infty$.

Aspirin:

The feasibility of detecting small molecules in solution using ODNP-enhanced NMR spectroscopy was studied using aspirin in deuterated chloroform. The NMR spectrum of 2.6 % (w/v) aspirin (corresponding to 144 mM) in d-chloroform with 10 mM TEMPONE is shown in Figure 2. We detect distinct peaks with baseline resolution for the methyl (2.35 ppm) and ring (7.14–8.13 ppm) protons, which are in excellent agreement with values reported in the literature.⁶⁹ The carboxylic acid proton is a broad peak barely visible at 11 ppm. The NMR spectrum was acquired as 16 individual traces over a period of about 1.6 minutes and the spectra were aligned using the cross-correlation method described above before summing over all transients. However, signal averaging using the cross-correlation method results in slight line broadening due to imperfect performance of the cross-correlation method. In general, the method requires a sufficient signal-to-noise ratio in a single-shot experiment to produce satisfactory results. Because no NMR signal was observed without microwave power, it was not possible to obtain a spectrum without microwave radiation, demonstrating the limitations of the current technology. After signal averaging we observe a linewidth of 9.4 Hz (0.66 ppm) for the methyl proton line.

Figure 2:

ODNP enhanced NMR spectrum of 2.6 % (w/v) aspirin and 10 mM TEMPONE in d-chloroform. Spectrum is referenced to the resonance corresponding to the methyl group (A) at 2.35 ppm. The spectrum has been phased positively.

Because no signal was observed without microwave radiation, we can only estimate the observed enhancement. Based on the aspirin concentration and the observed signal-to-noise ratio, we estimate an absolute enhancement of about 30–40.

Sample heating in general is a severe problem in ODNP experiments. However, in the case of Aspirin in d-chloroform, the low dielectric losses of the solvent ($ϵ\text{'}\text{'}$= 0.79 ⁷⁰) makes it a good candidate for ODNP-enhanced NMR spectroscopy. While in many cases sample heating is unwanted, the reduced viscosity and therefore faster solvent/solute dynamics lead to higher enhancements at higher temperatures.⁷¹ For certain applications including process NMR and reaction monitoring, where ODNP could be used to detect small molecules in solution, microwave heating would be tolerable.

Ethyl Crotonate:

pEthyl crotonate is a small molecule with a broad range of chemical shifts from different functional groups. It also exhibits large J-couplings which makes it an ideal candidate for demonstrating high-resolution ODNP enhanced NMR.

The ODNP spectrum for neat ethyl crotonate is shown in Figure 3. As can be observed from the spectrum, we can assign peaks for CH₃, CH₂, and alkene protons. The two methyl groups at 1.28 and 1.88 ppm can be distinguished, although their peaks are overlapped. The methylene quartet at 4.18 ppm is well-resolved and displays the correct splitting and peak intensities. The enhancement obtained by integrating over the entire spectrum was found to be −32 at a microwave power of 7.1 W. We have assigned different regions of the spectrum to different protons on ethyl crotonate as shown in Figure 3 (left). The J-couplings for protons A and B (see Figure 3, left) are clearly resolved and are 7.1 and 6.8 Hz, respectively. These values are in excellent agreement with the accepted value of 7.0 Hz for the J-couplings for both protons.⁶³

Figure 3:

Left: ODNP-enhanced NMR spectrum of neat ethyl crotonate with and without microwave irradiation (microwave power of 7.1 W). The ODNP-enhanced spectrum has been phased positively. Spectra have been offset for clarity. Right: ODNP enhancement for distinct regions of the ethyl crotonate spectrum as a function of microwave power.

The high resolution observed in our experiments allows us to determine enhancement factors for individual proton sites in ethyl crotonate. The maximum enhancement for regions A+E, B, and C+D were found to be −43 +/− 2, −43 +/− 2, and −47 +/ 2, respectively (see Figure 3, right). While the difference in enhancements is well outside the error region, further experiments are necessary to confirm this.

Ethylene glycol:

Microwave induced sample heating is a concern for ODNP experiments where significant microwave power must be used to achieve large enhancements. Given the excellent resolution that we are able to achieve in our experiments we can quantify the microwave induced sample heating by measuring the NMR spectrum of ethylene glycol as a function of the microwave power. Ethylene glycol is frequently used in NMR spectroscopy as a sensitive probe of the sample temperature.^63,72 The chemical shift of the OH group is sensitive to the intermolecular hydrogen bonding which results in a shielding effect as the temperature is increased.

NMR spectra for ethylene glycol as a function of microwave power are shown in Figure 4 (left). The OH and CH₂ peaks occur at a chemical shift of ~5 ppm and 3.42 ppm, respectively. As the microwave power is increased, the separation between the two NMR lines decreases. In addition, the individual peaks become narrower. The observed ODNP enhancements as a function of microwave power are shown in Figure 4 (right). The ODNP effect saturates at a microwave power of 0.7 W, beyond this, significant sample heating occurs resulting in larger enhancement factors. This fact is well-known and documented throughout the literature.^{21,22,71,73,74} While the region at 0.7 W is attributed to the saturation of the ODNP effect, dividing by the $T_1$ as a function of power to account for the temperature dependence of $T_1$ would be required to determine this unambiguously. In addition, we observe different enhancement factors for the OH and CH₂ protons of −15 +/− 1 and −11 +/− 1, respectively. The difference in enhancements is attributed to hydrogen bond formation between the OH of ethylene glycol and the nitroxide moiety of the radical.

Figure 4:

Left: ODNP-enhanced NMR spectra of ethylene glycol with 10 mM TEMPOL as a function of microwave power. All spectra have been phased positively and offset for clarity. The dashed line at position A is added as a guide for the eye. Right: ODNP enhancement E as a function of microwave power for ethylene glycol. The inset shows an expanded region for a microwave power of 0 to 0.8 W.

To calculate the temperature as a function of the applied microwave power, the NMR spectra were fitted to the sum of two Lorentzians. The temperature is related to the distance between the two maxima $\mathrm{\Delta }\delta$ by:⁷⁵

$$T\left(K\right)=465.9-102.24\text{Δ}\delta$$ (3)

The chemical shift difference $\mathrm{\Delta }\delta$ and the calculated sample temperature as a function of the applied microwave power are shown in Figure 5. The temperature without microwave power agrees exactly with the measured temperature of the magnet at 34 °C. While the sample is cooled using a constant stream of air entering the magnet at room temperature, it warms up to the magnet temperature, while it is traveling to the sample. It should be noted that with less than 1 W of microwave power, a value at which we are able to saturate the ODNP enhancement of the ethylene glycol sample, the microwave induced heating is less than 10 °C. For this particular sample, there would be no reason to exceed 1 W of microwave power to study the hydration dynamics.

Figure 5:

Chemical shift separation between OH and CH₂ protons for ethylene glycol (green) and corresponding sample temperature (orange).

At microwave powers greater than 2 W, significant sample heating occurs due to the large ohmic losses (microwave absorption) of ethylene glycol. A summary of the dielectric properties of samples studied here is given in Table S1. The value of $\epsilon \text{'}\text{'}$ for ethylene glycol is 6.58 at X-band.^76,77 Other samples presented here have significantly lower ${\epsilon }^{\text{'}\text{'}}$ values and therefore show significantly less heating. At a maximum microwave power of 7.1 W at the position of the sample, we observe a sample temperature of 78.5 +/− 1.0 °C for ethylene glycol.

While the separation between the two peaks becomes smaller at higher sample temperatures, two effects aid in resolving the peaks. The peaks become narrower at higher temperatures due to a decrease in the viscosity of the ethylene glycol, and the observed enhancements are larger at increased temperatures which yields better signal-to-noise.

The enhancement curve for ethylene glycol shown in Figure 4 (right) can be explained by considering the viscosity and microwave absorption of the sample. The relatively high viscosity of ethylene glycol results in a maximum value of ${\epsilon }^{\text{'}\text{'}}$ at ~1.5 GHz. As the sample temperature increases, a decrease in viscosity shifts the maximum for ${\epsilon }^{\text{'}\text{'}}$ to higher frequencies, resulting in increased ohmic losses at higher temperatures. In addition, the lower viscosity at higher temperatures results in dramatically increased enhancements. A decrease in viscosity will have two effects, an increased $T_{10}$ and increased coupling factor, both of which will result in a larger enhancement.

In the case of water, enhancement curves which vary linearly at high powers have been shown.²¹ This is caused by a similar, but not as apparent heating effect. For water, which has much lower viscosity than ethylene glycol, the maximum for ${\epsilon }^{\text{'}\text{'}}$ occurs at ~18 GHz (Figure S5).⁷⁷ Near room temperature, the viscosity for water is low enough to still yield large enhancements. In addition, the heat capacity for water is greater than ethylene glycol, which would result in lower temperature heating for an aqueous sample for the same amount of dissipated heat. Therefore, we expect that for our instrument, in most experiments to extraction molecular dynamics information, a microwave power of < 1 W will provide sufficient saturation of the EPR spectrum while minimizing the effects from microwave heating.

Conclusions

In this work, we demonstrate the ability of ODNP to enhance the sensitivity of NMR experiments at low fields (0.35 T) while still retaining chemical shift information. These measurements are made possible by improvements in magnet design and the incorporation of electrical shims to increase the field homogeneity. We have achieved a linewidth as low as 3.9 Hz (0.27 ppm) in the case of the methyl peak for toluene. Furthermore, it is possible to resolve the J-couplings of ethyl crotonate. We envision a number of useful applications for chemical shift resolved ODNP experiments at low field.

To this point, the main application of X-band ODNP is to determine hydration dynamics of biological water on bio-macromolecular surfaces. Our experiments demonstrate the ability to extract dynamical information from more complicated systems (e.g. water/solute mixtures). The nitroxide spin probe can be used not only as a sensitive probe of water dynamics, but also the dynamics of other small molecules in solution. It is therefore possible, to study the translational correlation times for multiple distinct solutes in solution if their chemical shifts can be resolved.

One example would be porous materials such as Nafion, where it has been shown that channel water and bulk water exhibit different chemical shifts.⁷⁸ It may be possible to simultaneously measure the coupling factors for water in these different environments.

Another promising application of ODNP involves process NMR and reaction monitoring. In this application, ODNP would primarily provide a sensitivity gain to low field NMR experiments. Numerous samples with endogenous radicals can be used such as crude oil.^79,80 Alternatively, exogenous radicals can be used either by adding them to the solution or immobilized on a gel.^81,82

While all experiments in this work were performed using 10 mM radical concentration, there are several compromises that can be made to either achieve higher enhancements or better chemical shift resolution. For systems with highly limited sensitivity, the nitroxide concentration can be increased to > 50 mM to improve enhancements at the expense of line broadening. In addition, if sample heating is a concern, the microwave power can be limited to levels < 2 W, in which case enhancements greater than 10 can still be readily obtained.

We expect that further improvements in the linewidth are possible by adding additional electric shims to correct for higher order field inhomogeneities, especially higher order tesseral terms. Additionally, a frequency lock could be introduced to correct for magnetic field drifts, which can be implemented as a separate channel of the NMR spectrometer to correct the spectrometer frequency. Another possibility is to use a separate single-board NMR spectrometer to acquire the resonance frequency of a reference sample and use the sweep coil (or an additional B₀ lock coil) to adjust the external magnetic field. With these additional measures, linewidths determined by the nuclear relaxation time T₂ should be within reach. In the future, we expect the introduction of 2D NMR experiments will further increase the resolution and capabilities of ODNP enhanced NMR spectroscopy at low fields.

Abbreviations

ODNP

Overhauser Dynamic Nuclear Polarization

DNP

Dynamic Nuclear Polarization

NMR

Nuclear Magnetic Resonance

EPR

Electron Paramagnetic Resonance

TEMPONE

4-oxo-TEMPO

TEMPOL

4-hydroxy TEMPO

FID

Free Induction Decay

cw

continuous wave