Selective NMR observation of the SEI–metal interface by dynamic nuclear polarisation from lithium metal

While lithium metal represents the ultimate high-energy-density battery anode material, its use is limited by dendrite formation and associated safety risks, motivating studies of the solid–electrolyte interphase layer that forms on the lithium, which is key in controlling lithium metal deposition. Dynamic nuclear polarisation enhanced NMR can provide important structural information; however, typical exogenous dynamic nuclear polarisation experiments, in which organic radicals are added to the sample, require cryogenic sample cooling and are not selective for the interface between the metal and the solid–electrolyte interphase. Here we instead exploit the conduction electrons of lithium metal to achieve an order of magnitude hyperpolarisation at room temperature. We enhance the 7Li, 1H and 19F NMR spectra of solid–electrolyte interphase species selectively, revealing their chemical nature and spatial distribution. These experiments pave the way for more ambitious room temperature in situ dynamic nuclear polarisation studies of batteries and the selective enhancement of metal–solid interfaces in a wider range of systems.


Supplementary
: Schematic of the Overhauser Effect. The four energy levels of a two-spin system, containing one electron spin ( e ) and one nuclear spin ( n ), where α and β represent upand down-spin states, respectively. The electron states represent a single point in reciprocal space within the band structures shown in Fig. 1b in the main text. The splitting of the electron and nuclear spin levels are denoted by 0e and 0n , respectively. On microwave irradiation at the CESR frequency (μw, green wavy lines), electron spins are flipped, populating the upper levels. These electrons then relax to their ground states (red straight lines), but can also cross relax by simultaneously flipping a nuclear spin (blue straight line). Zero quantum cross-relaxation (α e β n → β e α n ) dominates because the dominant electron-nuclear interaction is Fermi contact, which is isotropic and cannot drive double quantum crossrelaxation (α e α n → β e β n , not shown). [1][2][3] Since the cross-relaxation is asymmetric, the net effect is to overpopulate the nuclear spin up state ( n = α), resulting in positive hyperpolarisation of the NMR signal.
Supplementary Figure 2: Effect of microwave power on peak position. 7 Li NMR spectrum of microstructural lithium metal (sample B) without microwave irradiation (μw OFF) and as a function of microwave power (7.4, 8.7, 11.0, 13.2 and 15.6 W), recorded at 14.1 T, 10 kHz MAS and a sample temperature of ~100 K, using a Hahn echo pulse sequence and a recycle delay of 0.1 s. Spinning sidebands are marked with an asterisk. The spectra in b) have been rescaled to all have the same intensity, to allow easier comparison of the average chemical shifts.
Supplementary Figure 3: DNP enhancement of Li metal at 100 K. 7 Li NMR spectrum of microstructural lithium metal (sample A), with and without 15.6 W of microwave irradiation at 395.29 GHz (μw ON/OFF), recorded at 14.1045 T, 12.5 kHz MAS and a sample temperature of ~100 K, using a Hahn echo pulse sequence and a recycle delay of 0.25 s. Spinning sidebands are marked with an asterisk.
Supplementary Figure 4: EPR of Li metal as a function of temperature. The X-band (ν = 9.376393 GHz) EPR spectra of lithium metal as a function of temperature, showing the expected Dysonian lineshape. The microwave frequency changed slightly as a function of temperature, so the magnetic field axes of the low temperature spectra have been rescaled to the 290 K frequency to allow comparison of the spectra, and the intensities have been scaled to the same maximum. The peak absorbance (derivative = 0) occurs at B0 = 3345.268 G, corresponding to g = 2.0026.
Supplementary Figure 5: Field sweeps of Li metal DNP at 100 K and 300 K. The enhancement of the integrated intensity for the metal 7 Li NMR signal of microstructural lithium as a function of the B0 field, measured at 100 K (sample B, Fig. 1b, main text) and 300 K (sample G), with microwave irradiation at 395.29 GHz and 15.6 W, recorded at 12.5 kHz MAS with a recycle delay of 0.25 s.    , and 19 F (i-l) NMR spectra of lithium microstructures produced by cycling with the LP30 (sample F, top half) and LP30+FEC (sample G, bottom half) electrolytes, recorded with and without microwave irradiation (μw ON/OFF; 11.0 W for a-i, 15.6 W for j-l). All spectra were recorded at 12.5 kHz MAS, 14.1 T and room temperature, unless otherwise stated; spinning sidebands are marked with asterisks. The direct spectra were recorded with a Hahn echo pulse sequence and the CP spectra were recorded with the 7 Li carrier at 0 ppm and contact times of 1 ms and 0.1 ms for 7 Li→ 1 H and 7 Li→ 19 F respectively. For the direct 1 H and 19 F experiments, the difference between the spectra recorded with and without microwave irradiation is also shown. The recycle delays were: 1s (ad, e, f, h, j, l); 3 s (i, k); or 45 s (g). The deconvolution parameters for (b) and (d) are shown in Supplementary Table 3, and those for (e) and (g) in Supplementary Table 4.
Supplementary Figure 14: DNP-enhanced spectra of diamagnetic Li at 100 K. 7 Li NMR spectra a) and deconvolutions b) of microstructural lithium deposited with the FEC electrolyte (sample E) with and without microwave irradiation, recorded at 105 K, 14.1 T and 12.5 kHz MAS, with a 1 s recycle delay and a Hahn echo pulse sequence. Similar results are seen as for the room temperature spectra ( Supplementary Figure 15: Deconvoluted diamagnetic 7 Li NMR spectra for sample D. 7 Li NMR spectra with and without microwave irradiation for microstructural lithium deposited with the LP30 electrolyte (sample D, Fig. 3b, main text), recorded at room temperature, 14.1 T and 12.5 kHz MAS with a 1 s recycle delay and a Hahn echo pulse sequence.
Supplementary Figure 16: Deconvoluted diamagnetic 7 Li NMR spectra for sample E. 7 Li NMR spectra with and without microwave irradiation for microstructural lithium deposited with the FEC containing electrolyte (sample E, Fig. 3d, main text), recorded at room temperature, 14.1 T and 12.5 kHz MAS with a 1 s recycle delay and a Hahn echo pulse sequence.
Supplementary Figure 17: Deconvoluted 1 H NMR spectra for sample D. 1 7 Li{ 19 F} REDOR spectra of microstructural lithium produced with the LP30+FEC electrolyte (sample E), recorded at 14.1 T and 12.5 kHz MAS, with (S1) and without (S0) 5 rotor periods of recoupling. a) Spectra recorded with and without microwave irradiation, using a 10 s recycle delay and 32 scans. b) Spectra recorded with two different 19 F carrier frequencies, with microwave irradiation, a 1 s recycle delay and 160 scans.
Although the observed dephasing is minor, the fact that it disappears when the 19 F carrier is misset indicates that it is a real effect.   Fig. 3a-d, main text), and for sample E at 105 K ( Supplementary Fig. 14). See Supplementary Fig. 15 and Supplementary Fig. 16.  This signal could not be seen in the microwave off spectrum, and the intensity in the difference spectrum is the same as in the microwave on spectrum.
Supplementary   Fig. 3e,g, main text) of microstructural lithium deposited using the LP30 and LP30 + FEC electrolytes (samples D and E). The deconvoluted spectra are shown in Supplementary Fig. 17 and Supplementary Fig. 18. The enhancements are given by the ratio of the microwave off and on spectra; if the signal could not be easily distinguished in the microwave-off spectrum, the microwave-off intensity was calculated by subtracting the deconvoluted intensity in the difference spectrum from that of the microwave-on spectrum. This signal could not be seen in the microwave-off spectrum, and the intensity in the difference spectrum is the same as in the microwave-on spectrum within error.

Supplementary Note 1: Skin Depth Calculation
The skin depth for penetration of electromagnetic radiation into a metal is given by: where the resistivity of the metal (92.8 nΩ m for Li metal at 293 K 7 ), 0 is the vacuum permeability, is the relative permeability (1.4 for Li metal 7 ) and ν is the frequency of the radiation.
For the radiofrequency and microwave radiation used at 14.1 T in this study, this yields the skin depths shown in Supplementary Table 8. At lower temperatures the resistivity decreases, resulting in a shallower skin depth.

Supplementary Note 2: Microwave-induced Sample Heating
In order to investigate the degree of sample heating induced by microwave irradiation at room temperature, the 79 Br spectra were recorded for a microstructural lithium sample mixed with KBr and for pure KBr, both with and without microwave irradiation ( Supplementary Fig. 33). The 79 Br resonance in KBr has a temperature dependence of −0.025 ppm/K. 8 Firstly, under MAS the microstructural lithium sample has a higher sample temperature than pure KBr by ~4 K, which could be due to eddy currents in the metal. Under microwave irradiation the heating for each sample is similar with the temperature increasing by ~35 K (Supplementary Table 9), although the nominal temperature increase as measured by the sample thermocouple was only 8 K. The linewidth is greater for the microstructural lithium sample, however, suggesting a broader distribution of sample temperatures; in particular, on microwave irradiation the resonance spreads to even lower frequencies, suggesting there could be hot spots in the sample which experience greater heating. It is challenging to determine the extent of the temperature variations within the sample since there is also an inherent contribution to the linewidth of the 79 Br signal, particularly for the heterogeneous lithium sample, but by comparing the full-width at halfmaximum of the resonances, the microstructural lithium sample may experience a temperature range of ~20 K. The sample heating appears to be more severe at room temperature than at 100 K, where sample heating was approximately 5 K for a pure KBr sample.

Electrochemistry
Copper metal disks (MTI) were soaked in concentrated acetic acid for 10 minutes for oxide removal. The acetic acid residue was removed with dry nitrogen flow and the disks were dried at 100 °C under vacuum overnight. The lithium metal disks (LTS research, 99.95%) were used without any pretreatment. Printed LiNi0.8Mn0.1Co0.1O2 (NMC811) electrodes were fabricated in Argonne National Laboratory (A-C020, made by CAMP facility) using the following materials: 90 wt % NMC811 (Targray), 5 wt % conductive carbon (Timcal C45), and 5 wt % PVDF binder (Solvay 5130). They were dried overnight at 120 °C under dynamic vacuum. Galvanostatic electrodeposition was performed using a Biologic MPG2 battery cycler with EC-Lab software. For all the samples except C, the constant currents shown in Supplementary Table 7 were applied in a single direction for a capacity of 4 mAh cm −2 (e.g. Supplementary Fig. 30). For sample B, constant current was applied at 0.033 mA cm −2 for 60 hours (2 mAh cm −2 ) followed by 12 cycles of stripping for 1 hour and plating for 5 hours (Supplementary Fig. 31), to increase the amount of the SEI formed. For sample C, galvanostatic charging was performed at 0.455 mA cm −2 until the voltage reached 4.5 V, followed by holding at 4.5 V for 8 hours; over both steps, a total charge of 1.76 mAh cm −2 was transferred ( Supplementary Fig. 32). The coin cells were disassembled in an Ar atmosphere glovebox and the microstructures scraped off gently with a razorblade without rinsing. The sample was then diluted by ~5× by mass with dry KBr by lightly hand-grinding with an agate mortar and pestle to improve microwave penetration and allow the metallic samples to be more easily spun; this does not have a significant effect on the microstructural morphology (see SEM images, Supplementary Fig. 28). The sample was finally packed into a 3.2 mm outer diameter (2.2 mm inner diameter, 28 μl volume) sapphire rotor.

NMR Spectroscopy
For the NMR experiments performed at 14.1 T, radiofrequency (rf) powers were used of 45 kHz for 7 Li, 110 kHz for 1 H and 85 kHz for 19 F. For the CP experiments, rf powers during contact were used of ~45 kHz for 7 Li , ~85 kHz for 1 H and ~70 kHz for 19 F. The experiments at 9.4 T used a 7 Li rf power of 100 kHz. 1 H NMR spectra were referenced to adamantane at 1.81 ppm, 7 Li spectra to LiF at −1 ppm and 19 F spectra to LiF at −203 ppm, all at room temperature. The gyrotron microwave powers were measured using a calorimeter half-way along the waveguide, while the klystron power was measured at the output using a directional coupler then scaled to account for microwave reflection. Spectra were deconvoluted using dmfit software. 9