Endogenous Dynamic Nuclear Polarization for Sensitivity Enhancement in Solid-State NMR of Electrode Materials.

Rational design of materials for energy storage systems relies on our ability to probe these materials at various length scales. Solid-state NMR spectroscopy is a powerful approach for gaining chemical and structural insights at the atomic/molecular level, but its low detection sensitivity often limits applicability. This limitation can be overcome by transferring the high polarization of electron spins to the sample of interest in a process called dynamic nuclear polarization (DNP). Here, we employ for the first time metal ion-based DNP to probe pristine and cycled composite battery electrodes. A new and efficient DNP agent, Fe(III), is introduced, yielding lithium signal enhancement up to 180 when substituted in the anode material Li4Ti5O12. In addition for being DNP active, Fe(III) improves the anode performance. Reduction of Fe(III) to Fe(II) upon cycling can be monitored in the loss of DNP activity. We show that the dopant can be reactivated (return to Fe(III)) for DNP by increasing the cycling potential window. Furthermore, we demonstrate that the deleterious effect of carbon additives on the DNP process can be eliminated by using carbon free electrodes, doped with Fe(III) and Mn(II), which provide good electrochemical performance as well as sensitivity in DNP-NMR. We expect that the approach presented here will expand the applicability of DNP for studying materials for frontier challenges in materials chemistry associated with energy and sustainability.

Introduction

The development of new and improved materials for rechargeable batteries requires analytical tools, which can provide insights into the composition and structure of the electrodes, electrolyte, and their interface. In the past years, solid-state NMR (ssNMR) spectroscopy has been shown to be particularly useful, enabling very detailed characterization of the electrochemical and chemical transformations within the bulk of the materials used as electrodes and solid electrolytes as well as the electrode–electrolyte interface/phase.[1−5] In this respect, the main advantages of ssNMR are its high chemical sensitivity, which enables tracking the formation of phases and perturbations to the local environment of the detected nuclei, and the ability to determine proximity between phases and chemical environments. However, as NMR spectroscopy has low detection sensitivity, this information is often limited to phases containing favorable elements, i.e., elements that have highly abundant NMR active isotopes with a high gyromagnetic moment (for example 1H, 19F, and 7Li). To detect isotopes such as 6Li, 13C, and 17O, which are very sensitive probes to their local environment, isotope enrichment, high magnetic fields, and long experiment times are essential.

Magic angle spinning dynamic nuclear polarization (MAS-DNP) can provide an alternative and extremely efficient route for enhancing the detection sensitivity of ssNMR. This is achieved at cryogenic temperatures by transferring the large polarization of unpaired electron spins to their surrounding nuclear spins using microwave (μwave) irradiation.[6,7] Typically, in materials science applications of MAS-DNP, the sample of interest is impregnated or wetted by a solution of nitroxide biradicals, which provide the source of polarization.[8,9] This exogenous approach leads to several orders of magnitude increase in ssNMR sensitivity, thereby alleviating the need for isotope labelling and enables detection of minute components in the sample. In battery materials, the exogenous approach was used successfully to probe the solid electrolyte interphase (SEI) formed on reduced graphene oxide and Si nanowires.[10,11] Nevertheless, the use of exogenous DNP has two main limitations in the study of battery materials: (i) the radical solution may perturb the SEI composition and structure, and (ii) the sensitivity obtained is often limited to the outer layers of the SEI and does not extend to the bulk of the electrodes, especially for micron-sized particles.[12]

We have recently introduced an alternative approach for DNP in inorganic materials, metal ion DNP (MIDNP), based on the use of paramagnetic metal ion dopants.[13,14] We demonstrated up to 104 fold improvement in NMR sensitivity by endogenous DNP from Mn(II) dopants in oxides, including the anode material Li4Ti5O12 (LTO). Such a gain in sensitivity enabled the detection of the 17O nuclei in the bulk of micron-sized particles at natural abundance as low as 0.038%, suggesting that endogenous DNP can become a viable characterization tool for inorganic materials in general and for battery materials in particular.

Here, we develop the MIDNP approach further and consider the factors affecting its performance in electrochemically active materials using LTO as a model system. From the materials aspect, we have to consider the effect of the dopants on the electrochemical performance of the materials. Ideally, MIDNP could be employed with dopants that have dual functionality; they improve the electrochemical performance of the electrodes (through increasing their ionic and/or electronic conductivity) as well as provide means for high sensitivity in ssNMR. Alternatively, dopants can be used as structural spies, introduced at a minimal amount solely for sensitivity enhancement. From the methodology aspect, the DNP efficacy in enhancing NMR sensitivity will depend strongly on the materials formulation as electrodes. NMR relaxation properties and μwave distribution across the sample, which are crucial for efficient polarization transfer in DNP, will vary depending on changes in the sample conductivity and disorder during electrochemical cycling. So far, these aspects were not addressed in the context of applications of MAS-DNP in materials science in general and battery materials in particular.

To date, only few paramagnetic dopants were utilized for MIDNP with MAS, these include Mn(II), Gd(III), and Cr(III).[14−19] Thus, expanding the range of dopants that are viable as polarization agents is essential for broadening the applicability of this approach. Here, we first introduce Fe(III) as a new and efficient endogenous agent for DNP. We provide detailed electron paramagnetic resonance (EPR) characterization of Fe(III) in the LTO framework and determine its performance in increasing the sensitivity of lithium NMR. We then consider the effect of iron doping on the electrochemical performance of LTO and identify conditions for its use in electrodes. Next, we determine the factors that are affecting the DNP performance in cycled electrodes. In addition to Fe(III), we consider Mn(II) dopants, which provide high signal enhancement through DNP[14] but at increasing content block Li diffusion.[20] We show that the efficacy of the DNP process depends on the electron and nuclear spin properties and their environment as well as the presence of conductive additives in the sample such as carbon black.[21] Finally, we demonstrate that a carbon free electrode formulation is an excellent route for studies with MIDNP, providing sufficient sensitivity while maintaining good electrochemical performance.

Experimental Section

Synthesis

Fe(III)-doped Li4Ti5O12 (Fe-LTO) powders were prepared via solid-state synthesis. Li2CO3 (Strem Chemicals, Inc. 99.999% purity), TiO2 (Alfa Aeser, 99.9% purity), and Fe2O3 (Strem Chemicals, Inc. 99.995% purity) were used as starting materials. The precursors were mixed at stoichiometric quantities adding 5% excess of the Li source to compensate for Li evaporation at high temperatures. The mixture was milled using high energy ball mill (Spex) for 1 h, and the powder was then pelletized and calcined at 850 ° C for 12 h under air. Fe-LTO was synthesized in five nominal Fe(III) stoichiometries: x = 0.00125, 0.0025, 0.005, 0.01, and 0.02 where x is the amount of Fe(III) per LTO formula. The samples are labelled Fe00125-02. Mn(II)-doped LTO was synthesized as described in earlier work.[13] Mn-doped LTO will be labelled Mn0025, corresponding to Mn stoichiometry of x = 0.0025 per LTO unit.

Powder XRD

Phase purity of all Fe-LTO samples was determined by powder X-ray diffraction measurements on a TTRAX-III Rigaku diffractometer equipped with a rotating Cu anode operating at 50 kV and 200 mA. The 2θ scanning range was 10–120 with a scan rate of 2°/min. Phase analysis was performed using JADE 2010 software.

EPR Spectroscopy

EPR measurements were performed on a Bruker ELEXYS E-580 spectrometer operating at Q-band (35 GHz) fitted with a Q-band resonator (EN-5107-D2). The temperature was controlled by an Oxford Instruments CF935 continuous flow cryostat using liquid helium. Continuous wave (CW) and field-sweep echo-detected (FSED) spectra were acquired at 50 K, and relaxation measurements were performed at 100 K. Electron-nuclear double resonance (ENDOR) analyses were performed around the 7Li Larmor frequency (20.6 MHz) using the Mims sequence[22] π/2-τ-π/2-T-π/2-τ-echo with an RF pulse applied during the time T, at 50 K. The experimental conditions were: π/2 = 10 ns, τ = 100 ns, and tRF=15,000 ns for 1940 scans. EPR spectra were simulated using EASYSPIN simulation package for matlab.[23] Cycled samples were sealed in capillaries using epoxy glue in an argon glovebox.

MAS-DNP NMR Spectroscopy

Experiments were carried out on a Bruker 9.4 T Avance Neo spectrometer equipped with a sweep coil and a 263 GHz gyrotron system. A 3.2 mm triple resonance low-temperature (LT)–DNP probe was used for the experiments at an MAS of 10 kHz. All experiments were performed around 100 K. The Fe-LTO samples were inserted into 3.2 mm sapphire MAS-DNP rotors and sealed with a Teflon plug. The cycled samples were packed into the rotors in the glovebox under an argon atmosphere, sealed with a Teflon plug, and quickly inserted into the magnet for measurements. Several cycled samples were diluted with predried KBr by manual mixing with a mortar and pestle.

6,7Li spectra were acquired with a short train of saturation pulses followed by a relaxation delay and excitation. 50 and 150 pulses were used for saturation separated 1 and 1.5 ms with radio frequencies of 63 and 67 kHz for saturation and excitation, for 7Li and 6Li, respectively. T1 relaxation and DNP build up times were measured using a saturation recovery sequence.

Cell Assembly and Electrochemistry

Electrochemical tests were performed using 2032 coin cells (Hohsen Corp.) assembled in an Ar-filled glovebox, with water and oxygen contents less than 0.5 ppm. LTO anodes were studied in half-cell configuration (cycled versus Li).

Carbon-free (CF) film electrodes were prepared by mixing 95 wt % active material with 5 wt % polyvinylidene fluoride binder (PVDF) in NMP using a mortar and pestle. The slurry was then casted on Al foil by the bar-coating method. The resulting electrodes, with thicknesses in the range 50–90 μm, were roll-pressed at varying pressures resulting in electrode thicknesses in the range 10–35 μm. The electrodes were punched into disks (14 × 14 mm2), dried in a vacuum oven overnight at 100 °C, and inserted into the glovebox. All cells were prepared with a borosilicate glass fiber separator, which was soaked with approximately eight drops of 1 M LiPF6 in a 1:1 weight ratio of ethylene carbonate (EC) and dimethyl carbonate (DMC) solution.

Galvanostatic cycling tests were conducted at room temperature using Biologic VMP3 or BCS-805 potentiostats. The cells were cycled in the voltage range of 0.9–2.5 V at a rate of C/20. Cyclic voltammetry (CV) tests were performed on a Biologic BCS-805 cycler with a scan rate of 0.1 mV s–1.

Following cycling, the coin cells were disassembled in the glovebox, the doped LTO electrodes were extracted and rinsed with DMC to remove electrolyte residues and dried overnight under vacuum. The CF electrodes were then carefully scraped from the Al foil with a plastic spatula for MAS-DNP experiments.

Mössbauer Spectroscopy

Zero-field 57Fe spectroscopy was carried out in transmission mode at room temperature as discussed previously (see ref (24) and references therein). In a typical spectrum, on the order of 5 × 106 counts per channel were recorded resulting in a signal-to-noise ratio in excess of 10:1.

Results and Discussion

Fe-LTO Characterization

Fe-doped LTO samples were synthesized with various concentrations of Fe dopant via the solid-state route. Powder X-ray diffraction patterns (Figure S1) confirm the formation of the Fd3̅m spinel LTO phase with a minimal amount of Li2TiO3 impurity phase (3–4%). Li2TiO3 is a common impurity in solid-state synthesis of LTO, and it has a minor effect on the electrochemistry.[25,26]

Continuous-wave EPR (cw EPR) spectra were acquired for all Fe-LTO samples at 34.2 GHz (Q-band) and are plotted in Figure a. The spectra confirm the presence of Fe ions in all samples. Spectral broadening is observed with increasing nominal Fe content due to the increase in electron interactions (note that the cw spectrum of the lowest Fe content displayed a different line shape, most likely due to signal saturation). The changes in the peak to peak width as a function of the Fe content along with the electron longitudinal relaxation time are plotted in Figure b. The monotonic increase in spectral broadening along with shortening of the electron relaxation time is a strong indication that the Fe content in the lattice is increasing.

Figure 1

(a) CW-EPR spectra of the Fe-LTO samples measured on Q-band at 50 K. (b) Dependence of Fe(III) electron relaxation (red) and peak width (grey, from CW spectra) in Fe-LTO on the dopant concentration. (c) FSED spectrum (grey) of Fe0025 (grey) acquired on Q band at 50 K and a simulation (red) of Fe(III) spectrum with S = 5/2, g = 1.99, and ZFS parameters of D = 2300, E = 400 MHz with strains of 2300 and 300 MHz, respectively.

The field sweep echo-detected (FSED) EPR spectrum (Figure c) suggests that Fe is found in high-spin d5 configuration. The spectrum was fitted with a single-Fe species with electron spin S = 5/2 broadened by zero-field splitting (ZFS). The 7Li Mims electron nuclear double resonance (ENDOR) experiment was carried out at Q-band on the Fe0025 sample to support the incorporation of Fe(III) dopants in the bulk of the spinel LTO phase. As shown in Figure , the ENDOR spectrum displays well-defined and sharp resonances that can be fitted with two dipolar Pake patterns for Fe-Li distances of 2.71 and 3.7 Å corresponding to dipolar coupling strengths of 1.55 and 0.61 MHz, respectively. For the simulations, we used the point dipole approximation and assumed that there is no significant isotropic contribution to the couplings. This assumption was supported by fast magic angle spinning (MAS) 6Li spectra acquired at room temperature for three of the doped samples (Figure S2), which reveal spectral broadening upon doping with only negligible Fermi contact shifts.

Figure 2

Bottom: Li Mims ENDOR spectrum measured at Q-band at 50 K for the Fe0025 sample. Top: simulation of two possible Li sites contributing to the ENDOR spectrum.

These EPR results support the conclusion that the LTO lattice is doped with an increasing amount of Fe(III). Fe(III) EPR spectra commonly display several resonances (at g = 4, 6) as its ZFS interaction is typically larger than the low fields commonly used for EPR. The appearance of a resonance at g = 1.99 only, along with the relatively low ZFS strength, suggest that the Fe(III) environment is relatively symmetric.[27] The ENDOR spectrum can provide further insights into the doping site. As the ionic radii of high-spin Fe(III) (S = 5/2) are 0.65 Å in Oh and 0.49 Å in Td coordination,[28] Fe(III) in the LTO framework can potentially replace Li+ in Oh and Td sites (0.76 and 0.59 Å, respectively) or Ti4+ in Oh sites (0.605 Å). Li et al. observed an increase in the LTO cell parameter with Fe doping and determined that Fe(III) replaces Ti ions.[29] Taking into account the two dominant Fe-Li distances in the ENDOR pattern along with the probability of having such Fe-Li pairs in the spinel structure at the different sites, suggests that Fe(III) is predominantly found in Oh sites. Nevertheless, as the surrounding Li ions around the two sites have similar distances, we cannot rule out that some Fe(III) also replaces Li on Td sites (see also Figure S3 for additional simulations).

DNP on Fe-LTO Powders

Next, we consider the performance of Fe(III) as a polarization agent for MAS-DNP. First, the optimal field position, providing maximal enhancement in sensitivity, was determined. This was achieved by acquiring sweep profiles, measuring the signal intensity of 6,7Li resonances as a function of the magnetic field during μwave irradiation. Sweep profiles for the Fe005 sample are plotted in Figure a. These revealed positive and negative enhancements for the two nuclei. The enhancement lobes are separated by roughly twice the nuclear Larmor frequency for both 6,7Li, suggesting that the solid effect, in which a single Fe(III) dopant transfers polarization to its surrounding spins, is dominant at this concentration.

Figure 3

(a) DNP sweep profiles acquired for Fe005 with build-up times of 10 and 30 s for 7Li and 6Li, respectively. (b) 6,7Li spectra acquired with/without μwave at the optimal field position (marked with an arrow on the sweep profile in (a)). (c) Enhancement factors for 6,7Li as a function of Fe content in LTO.

Enhancement factors were determined for the two nuclei by setting the field to the position with the highest enhancement (marked by arrow in Figure a). Signal enhancement (εon/off obtained from the ratio between the integrated signal intensity with and without μwave) as a function of the Fe(III) content measured at steady-state conditions is shown in Figure c (along with representative spectra in Figure b). Excellent gains in sensitivity reflected by the enhancement factors were achieved for the two nuclei, with εon/off values of 51 ± 5 and 181 ± 18 for 7Li and 6Li, respectively, in the Fe005 sample. These values exceed the best enhancement we obtained previously with Mn(II) dopants in LTO.[14] Most likely, this increase is due to the fact that in Fe(III) the electron spin polarization of the central transition (between electron spin states with ms = 1/2, −1/2) is not split between several transitions by nuclear hyperfine couplings as is the case for Mn(II). Furthermore, the enhancement from Fe(III) has a weaker concentration dependence compared with Mn(II). These differences will be investigated in more detail in future work.

It is well accepted that a simple comparison of the spectra with and without irradiation is not sufficient to evaluate the gain in sensitivity.[30] A more accurate estimation of the gain in sensitivity should take into account the signal loss due to paramagnetic quenching (a phenomenon in which part of the NMR signal is lost in the presence of a paramagnetic dopant due to significant shortening in relaxation time and/or spectral broadening). For Fe(III), we observe minimal signal quenching at the low Fe content (less than 10% quenching), with significant drop in the 7Li signal with increasing dopant content and only minimal change in signal quenching for 6Li at the higher content (Figure S4). Nevertheless, taking into account paramagnetic quenching effects lead to minimal change in the enhancement factors (defined as εq,Figure c) below Fe content of x = 0.01. A thorough calculation of the absolute sensitivity (taking into account the temperature in DNP experiments and the change in nuclear relaxation times due to doping) is provided in the Supporting Information (Figure S5).

Finally, we consider whether the gain in sensitivity is achieved uniformly across the LTO lattice (outside the quenching sphere). In our previous study of Mn-LTO, we observed the formation of a broad resonance increasing in relative contribution with Mn content. We assigned this resonance to Li sites in close proximity to Mn dopants, broadened by dipolar interactions with the dopant and short transverse relaxation. While for low Mn content (x < 0.01, with x Mn in LTO), uniform enhancement was obtained across the spectrum, at higher Mn content, higher enhancement was obtained for the broad component. In the Fe(III)-doped samples, the growth of the broad resonance is also observed with increasing dopant content (Figure S6a,b). However, uniform enhancement is achieved for all spectra independent of the Fe content up to Fe content of x = 0.02 (Figure S6c).

Electrochemistry of Fe-Doped LTO

We now turn to consider the effect of the dopant on the electrochemical performance of LTO. Fe doping was shown to have a positive synergistic effect in a composite of Fe-doped LTO with reduced graphene oxide used as a hybrid supercapacitor.[31] It was also demonstrated to increase charge transfer properties of LTO used as an anode when doped with high Fe content (x = 0.2), which resulted in improved voltage profiles,[32] higher capacity, and rate performance.[29] Here, we compared the performance of the undoped LTO with that of Fe0025 and Fe02 in galvanostatic cycling (Figure ). As observed previously,[29,32] higher capacity was achieved consistently through 20 cycles in the Fe-doped samples. While the undoped LTO capacity was 123 mAh/g on the first discharge, with Fe02 electrodes 140 mAh/g was achieved. We note that for both samples this capacity is lower than the theoretical one of 175 mAh/g. The reduced capacity is most likely a result of the large micron-sized particles obtained in the solid-state synthesis.[13] Due to the low conductivity of LTO reduced particle size is preferred for achieving optimal performance.[33,34]

Figure 4

Electrochemical performance of undoped and doped LTO samples tested in half cells versus Li with galvanostatic cycling at C/20.

Additional insight into the effect of the dopant was obtained from cyclic voltammetry (CV) measurements (Figure ). In these tests, a higher Fe content was used (x = 0.1), which provided lower capacity than that obtained with the lower dopant content. Nevertheless, compared to the undoped LTO, with high Fe content, a second reduction peak is clearly observed at 1.78 V on the cathodic scan of the Fe01 sample. Surprisingly, this process does not appear on the anodic scan and does not appear on the second cathodic scan suggesting it is not reversible within the voltage window of 0.9–2.5 V versus Li. Furthermore, the number of electrons contributing to this reduction peak is proportional to the amount of Fe in the sample (see the Supporting Information). We have also tested the reversibility of this reduction process in a broader electrochemical window (Figure b). Expanding the anodic scan to 4.45 V leads to the appearance of an oxidation peak at about 4 V followed by the reappearance of the reduction process on the second cathodic scan.

Figure 5

Cyclic voltammetry measurements performed on undoped and Fe-doped LTO versus Li in a voltage range of (a) 0.9–2.5 V and (b) 0.9–4.45 V using a rate of 0.1 mV/s.

The correlation of the current with the Fe content suggests that this peak is due to reduction of Fe(III) + e → Fe(II), which is reversed when expanding the voltage range for oxidation. To confirm this, EPR measurements were performed on the Fe-doped samples following cycling in the two voltage windows (Figure S7a). While no signal was observed following cycling in the limited range, the Fe(III) resonance reappears following charging to higher potentials. ENDOR spectrum acquired from this sample (Figure S7b) was identical to the spectrum acquired from the uncycled powder, suggesting that with the extended cycling window, the local environment of Fe(III) remains the same as in the pristine sample.

Additional support for the effect of cycling on the Fe dopants was provided by Mössbauer spectroscopy. To this end, 57Fe(III)-doped LTO with x = 0.05 was prepared and examined before and after cycling in the limited potential range versus lithium. The Mössbauer spectrum of the pristine sample (Figure a) displays a single broad line with a width of 0.74 mm/s and an isomer shift of 0.277 +/– 0.004 mm/s. The theoretical line width in 57Fe spectra is 0.194 mm/s, while the experimentally observed width for a well-defined crystallographic site is approximately 0.24–0.26 mm/s. The broadening observed in the spectra shown in Figure clearly indicates a distribution of hyperfine parameters due to crystallographic inhomogeneity of the occupied iron sites in these samples. Due to the broadening, we cannot draw conclusions on the oxidation state or coordination of the Fe. However, the broadening may be expected since Fe is found in several coordination environments due to the random distribution of Ti and Li ions in Oh sites in the spinel structure, resulting in a distribution of hyperfine parameters (IS and QS) for 57Fe. The spectrum obtained for a sample cycled in the limited voltage range versus Li is significantly different (Figure b). At 97.8 K, there seems to be two Fe sites. One has an IS of 1.047 ± 0.008 and a QS of 2.482 mm/s, and the second site has an IS of 0.406 ± 0.14 mm/s. Again, the line broadening masks information related to the oxidation state of Fe, but we can conclude that Fe ions undergo a significant change compared to their initial state.

Figure 6

Mössbauer spectra (grey) and fits (color) acquired from Fe enriched Fe005 (a) before cycling and (b) after multiple cycles versus Li at 100 and 97.8 K, respectively.

Electrochemistry and DNP of Carbon Free Electrodes

Finally, we consider the effect of electrochemical cycling on the DNP performance. In a typical electrode formulation, carbon black is added to increase the electrical conductivity of the electrode. Conductive carbon particles were shown to have a deleterious effect on the enhancement obtained in DNP caused by μwave absorption, sample heating, and general shortening of relaxation times.[21] To avoid these effects, we opted for a carbon free (CF) electrode formulation, which is enabled by calendaring the electrode films.[35] Results from galvanostatic cycling of CF LTO electrodes are shown in Figure , comparing the performance of commercial LTO with that made via solid-state synthesis, with and without dopants. Despite having no carbon additive, commercial LTO electrodes performed very well, achieving about 160 mAh/g with excellent capacity retention. Electrodes made with LTO prepared via the solid-state route achieved lower capacity (as was also observed with the standard electrode formulation, Figure ), but after a short decay in the first few cycles, steady-state capacity of about 125 mAh/g was observed. Similar performance was obtained from CF electrodes prepared with Fe0025 powder and Mn0025. We expect that further improvement in the electrochemical performance can be achieved by optimizing the preparation process, ball milling the powder, and applying several rounds of calendaring. Nevertheless, for testing the MIDNP approach following cycling, the achieved electrochemical performance is sufficient.

Figure 7

Electrochemical performance of CF electrodes tested with galvanostatic cycling at 20 mA/g for (a) commercial LTO powder, (b) synthesized undoped LTO, (c) Mn0025, and (d) Fe0025 samples.

The enhancement factors obtained for the different samples are given in Figure . Both Fe0025 and Mn0025 were tested as these low dopant contents were shown to provide high enhancement factors (Figure c and ref (14)). First, to demonstrate the deleterious effect of the carbon black additive, we compared the enhancement achieved in pristine Fe0025 powder to that obtained when this powder is thoroughly mixed with 5% weight of C65 (carbon black additive). As expected, for both 6Li and 7Li, the enhancement drops significantly, from 180 to 9 and from 45 to 4, respectively. Next, we tested the enhancement in pristine CF films. Surprisingly, the process of casting the doped powders as CF electrodes itself already led to some drop in enhancement: from 180 to 53 for 6Li and 50 to 5.2 for 7Li for the Fe0025-CF samples with a similar trend observed for the Mn0025-CF samples. We note, however, that the enhancement for 6Li in pristine CF-Mn0025 films may be underestimated since the spectra were not acquired at steady-state conditions due to the very long relaxation time of 6Li in uncycled samples.

Figure 8

6,7Li enhancement factors measured for (a) Fe- and (b) Mn-doped samples with different formulations before and after electrochemical cycling.

The drop in enhancement in uncycled electrodes may be due to shortening of the nuclear longitudinal relaxation times (T1,) as well as residual pieces of Al foil used as current collector (the CF electrodes were scraped from the foil for measurements). The nuclear relaxation, T1,, was measured for the different samples, and for the pristine Fe0025-CF, it decreased from 23.4 ± 0.4 to 18.7 ± 0.5 s for 7Li and from 1942 ± 278 to 1785 ± 535 s for 6Li (for 6Li, the low sensitivity without DNP enhancement leads to large errors). Thus, the drop in enhancement after casting the powders may be explained, at least in part, by shortening of the nuclear relaxation, which limits the extent that the polarization can spread across the sample and between the nuclei through the process of spin diffusion.[36] The shortening of relaxation is possibly a result of the presence of the polymer binder, PVDF, which contains 1H and 19F that can increase relaxation through dipolar couplings to the sample and/or increased disorder due to the hot roll-press treatment.

Upon cycling these CF electrodes versus lithium, no enhancement was observed for 6Li and 7Li in the Fe0025-CF sample, see also the field sweep profile in Figure , further supporting the irreversible reduction of Fe(III) to Fe(II) in the 0.9–2.5 V window. Fe(II) is most likely not suitable for MIDNP due to its nonfavorable EPR properties, spin 2 with high ZFS and short relaxation times. For the cycled Mn0025-CF sample, another decrease in enhancement was observed, from 55 to 17.5 for 6Li and 8.9 to 1.5 for 7Li. Here again, this decrease can be due to shortening of the nuclear relaxation times as well as increase in sample conductivity. Nevertheless, the enhancement can be significantly improved by mixing the cycled Mn0025-CF with KBr, leading to enhancement factors of 53 for 6Li and 4.5 for 7Li.

Figure 9

7Li DNP sweep profiles acquired with a build-up time of 10 s for Fe-doped samples before and after galvanostatic cycling. The sample cycled in a broad voltage range was held at 4.45 V on charge until the current dropped to the noise level.

Finally, when Fe0025-CF films were cycled in the broader voltage range, some of the initial enhancement could be recovered: with enhancement factors of 11 for 6Li and 4.4 for 7Li. With the extended voltage range also the overall shape of the 7Li field sweep profile was regained (Figure ), suggesting that indeed the iron oxidation was at least partially reversed. Surprisingly, for the cycled sample there was higher asymmetry in the enhancement obtained at the positive and negative lobes of the sweep profile. This may suggest that the distribution of recovered Fe(III) ions is not uniform across the particles and that some of the LTO particles remain unpolarized (see the Supporting Information for further discussion).

As mentioned, the drop in enhancement following cycling can be due to shortening of the electron and nuclear relaxation times. After cycling, we can expect an increase in disorder in the electrode as a result of the repeated Li intercalation and deintercalation. Furthermore, upon lithiation, Ti(IV) ions, which are diamagnetic, are reduced to paramagnetic Ti(III). If delithiation is not completely reversible, the presence of Ti(III) ions can complicate the DNP process by shortening the electron and nuclear relaxation times as well as interfering DNP pathways from the Ti(III) and Mn(II) or Fe(III) ions. Furthermore, the presence of Ti(III) can result in increased electrical conductivity and sample heating. Diluting the sample with KBr decreases the heating effect as well as improves the μwave distribution in the sample resulting in higher enhancement.[37]

Conclusions

We have shown that Fe(III) dopants in highly symmetric environments such as in the spinel structure can be used as efficient polarization agent for MAS-DNP. Compared to Mn(II), Fe(III) is advantageous as its EPR line does not spread over several hyperfine transitions, and despite having shorter electron relaxation times, it provides higher sensitivity and uniform enhancement across the lattice at a broad range of dopant concentrations (9.5–150 mM tested here).

Ideally, for the implementation of MIDNP in the study of inorganic materials, one would like to choose dopants with multiple functionalities, namely, dopants that improve the material properties and provide a route for high sensitivity NMR. We have shown that, for the anode material LTO, Fe(III) has a beneficial effect on the performance. However, it is irreversibly active electrochemically in the standard operation range of LTO, with conversion of Fe(III) to Fe(II) upon discharging LTO. Nevertheless, we have demonstrated that Fe can be reactivated for MIDNP by the reverse process of Fe(II) to Fe(III) when the oxidation potential is increased.

We expect that the MIDNP approach, in particular with the high sensitivity gained from Fe(III) dopants, can be utilized in ssNMR studies for characterizing thin coatings deposited on the LTO surface. Thin coatings are a viable approach for increasing the rate performance of LTO and providing a protection layer preventing surface reactions.[38,39] Furthermore, as titanates are of interest as anode materials for Na-ion batteries,[40,41] Mn(II) and Fe(III) MIDNP can potentially be used to gain high sensitivity in ssNMR studies of this family of materials. In the broader context, the addition of Fe(III) to the family of MIDNP active dopants increases the range of inorganic materials that can be probed with high sensitivity ssNMR.

Finally, we have also performed the first investigation of the performance of the MIDNP approach for sensitivity enhancement on cycled electrodes. We have shown that electrode formulations with limited to no carbon are preferable and these can be obtained with minimal effect on the cycling performance in terms of capacity retention and discharge/charge potential. This approach will benefit other NMR- and DNP-based studies, including those making use of exogenous polarization sources such as nitroxides as they will minimize the effect of sample heating and μwave absorption by the carbon additive.