19 F and 31 P Solid-State NMR Characterization of a Pyridine Pentaﬂuorophosphate-Derived Solid-Electrolyte Interphase

Solid-state NMR spectroscopy was used to characterize the solid-electrolyte interphase (SEI) formed on the graphite elec- trode surface from the electrolyte additive pyridine pentaﬂuorophosphate (PPF). The PPF-derived SEI was prepared in a full Li(Ni 0.4 Mn 0.4 Co 0.2 )O 2 /graphite cell, rather than in a coin cell or using a bench-top synthesis approach. 19 F and 19 F → 31 P cross- polarization measurements provide direct evidence that F atoms and P–F bonds are present at the graphite surface. By comparing rinsed and unrinsed samples, an insoluble SEI component was differentiated from residual LiPF 6 from the dried electrolyte. The SEI species contains F and P chemical environments that are similar to, but distinct from, the additive starting material and the lithium hexaﬂurophosphate (LiPF 6 ) electrolyte salt used in this work. The results are consistent with the previously proposed formation of a dilithium 4,4’–bipyridine-N,N’–bis(pentaﬂuorophosphate) salt. LiF was also observed and is attributed to decomposition of the PPF-derived surface layer and/or the LiPF 6 electrolyte salt. ©

There is an increasing demand for longer-lasting and higher energy lithium-ion cells. Sacrificial electrolyte additives are a practical route to forming passive solid-electrolyte interphase (SEI) layers that limit electrolyte decomposition during cell storage and operation, which in turn can lead to improved cell performance and lifetime. [1][2][3] In recent years, significant steps have been made toward understanding the underlying chemistry of several electrolyte additives, including organic carbonate additives, [4][5][6][7][8][9] sulfur-containing additives, 6,10,11 and additives containing a Lewis acid-base adduct. 6,[12][13][14] Pyridine pentafluorophosphate (PPF, Figure 1a) belongs to this last category and has been shown to improve charge capacity retention after cycling at high temperature and high cell voltage.
By combining several experimental techniques with density functional theory (DFT) calculations, a proposed pathway for SEI formation on the negative electrode from PPF has recently been proposed ( Figure 1b). 13 It was suggested that during the formation cycle, dilithium 4,4'-bipyridine-N,N'-bis(pentafluorophosphate) salt, herein abbreviated as Li 2 (PPF) 2 , is produced and incorporated into the negative electrode SEI. However, the direct characterization of the PPF-derived SEI component(s) remains a significant challenge. This work explores the use of 19 F and 31 P solid-state nuclear magnetic resonance spectroscopy (ssNMR) as a complementary method to examine the chemical environment of the F and P atoms incorporated into the SEI.
Whereas the sensitivity of ssNMR may be correctly anticipated to pose experimental challenges, the PPF additive is appealing because it is rich in both F and P atoms. These atoms are both desirable for the practical reasons of the high natural abundance of the NMR-active 19 F and 31 P isotopes (∼100%), their relatively high gyromagnetic ratios (γ 19F = 40.052 MHz T −1 , γ 31P = 17.235 MHz T −1 ), and because they are both spin 1 2 nuclides. Moreover, ssNMR has already been used extensively in lithium-ion cell research, for example to study the local environments of lattice atoms (Li, Al) in a wide range of electrode materials [15][16][17][18][19][20][21][22][23][24][25][26][27] or of Al atoms in inorganic coatings. 28 ssNMR has also been applied to the study of carbonate-based electrolyte decomposition and of SEI composition. 7,[29][30][31][32][33][34][35][36][37][38][39] With these motivations, this work describes the use of 19 F and 31 P ssNMR to study the PPF-derived SEI, as formed in full Li(Ni x Mn y Co z )O 2 /graphite lithium-ion pouch cells. Figure 2. A typical cell voltage (solid black) and current (dashed red) profile for the formation protocol used in this study. The NMC442/graphite pouch cells were filled with 1M LiPF 6 in 3:7 EC/EMC with 10% PPF and maintained at 40 • C. The majority of cell capacity between 2.0 and 2.8 V is from PPF reduction. 13 The cells were immediately transferred to an argon-filled glove box, opened, and allowed to dry. The graphite was removed from the copper substrate by gentle scraping with a stainless steel razor.
For the identification of NMR resonance positions, a 'control' electrolyte, LiPF 6 complexed with EC, was prepared from a 1 M LiPF 6 solution in 3:7 EC/EMC without any additive, and then dried under vacuum. During this procedure the majority of the EMC evaporated and a solid, whose composition was near 3 M LiPF 6 in EC, was obtained.
Solid-state NMR spectroscopy.-For the ssNMR studies, the material was mixed in a 1:1 ratio, by volume, with ground SiO 2 (Aldrich, 99.9%) and packed into the NMR rotors (see below). The SiO 2 was added for experimental considerations, specifically because the large electrical conductivity of the graphite was found to interfere with the ability to spin the rotors. 31 The mixing of the sample with an insulating and inert material is a common approach to overcome this practical restriction (SiO 2 , Al 2 O 3 and Teflon are each widely used for this purpose).
The 19 F solid-state NMR spectra were acquired on Bruker Avance (16.4T) and Bruker Avance DSX (9.4T) spectrometers with Larmor frequencies of 658.394 MHz and 376.6 MHz, respectively. The chemical shift scale was referenced externally to Teflon at −123.2 ppm as the secondary reference. 40 The spectra in the high field magnet were acquired with either direct excitation (90 degree pulse length of 3.25 μs, Figures 3b-3d) or Hahn-echo sequence 90-T-180, with 3.25 and 6.5 μs for the 90 degree and 180 degree pulses, respectively, and T delays of 86 μs (Figures 3a, 3e, and 3f). Samples were spun at 22 kHz around the magic angle (MAS) in rotors of 2.5 mm diameters. Up to 128 scans were accumulated with a 5s delay. Use of the Hahn-Echo sequence removed the substantial 19 F probe head background. The 19 F MAS NMR spectra of the rinsed and unrinsed samples ( Figure 4) in the 9.4T magnet were also acquired with a Hahn-echo sequence using 6.25 and 12.5 μs for the 90 degree and 180 degree pulses, respectively, and T delays of 52 μs. The samples were spun at 19 kHz accumulating 4,096 scans with a 1s delay in rotors of 2.5 mm diameter.
The 31 P NMR experiments were performed on a Bruker Avance NMR spectrometer with a 9.4 T magnet (162.0 MHz 31 P Larmor frequency) and using a H-F/C-P probe head for rotors of 2.5 mm diameter. The 31 P NMR chemical shift scale was referenced externally against NH 4 H 2 PO 3 at 0.81ppm as a secondary reference. 19 F→ 31 P cross-polarization (CP) conditions, such as rf powers, CP contact time (5ms) and 19 F TPPM decoupling times were optimized on PPF. Samples were spun at 10.0 kHz. Scan repetition times were 5s, except for sample LiPF 6 , where 480 s were used. The number of transients accumulated ranged between 4 (PPF), to 10,240 (unrinsed graphite sample).  19 F MAS ssNMR spectra of a) the unrinsed, treated graphite, b) the LiPF 6 starting material, c) LiPF 6 complexed with EC, d) LiPO 2 F 2 , e) the PPF starting material, and f) LiF. In b-f, the isotropic peaks are identified. All other peaks are spinning sidebands. In the top trace, the spinning sideband is marked by "ss". These spectra were taken at 658.4 MHz Larmor frequency.

Results and Discussion
Like many sacrificial additives, it is thought that PPF is electrochemically reduced at the negative electrode surface as the cell is charged for the first time. This is supported by the presence of a peak in the differential capacity (dQ/dV) plot, where the area of the peak is directly proportional to the amount of additive in the cell. 13 One valuable observation from this experiment was that the electrochemical reduction of PPF is a two-electron process. As discussed above in the Methods section and shown in Figure 2, the faradaic charge measured in this work agrees with a two-electron reduction. Perhaps less obvious is that this also demonstrates that PPF will seemingly reduce in any quantity, which is a behavior that is unique, to the authors' knowledge, to the PPF/PBF-type additives. Other additives that have been studied in such detail, such as prop-1-ene-1,3-sultone (PES), vinylene carbonate (VC) and fluoroethylene carbonate (FEC), instead show a self-limiting behavior, where the graphite electrode is effectively passivated and further additive reduction ceases. 6 The above observation therefore raises two possible interpretations: i) the PPF reduction product is electrically conductive and capable of sustaining interfacial charge transfer, or ii) the PPF reduction product is soluble. It is known that the semicarbonate product shown in Figure 1b will be soluble, 41,42 whereas the solubility of Li 2 (PPF) 2 is difficult to predict from its structure alone. From X-ray photoelectron spectroscopy measurements, it is known that PPF reduction deposits some amount of P-and F-containing species onto the graphite surface that is resistant to rinsing. 13 A signal from the underlying graphite material is not present in the XPS spectra reported in Ref. 13. Given that XPS is only sensitive to the first few nanometers of a sample's surface (typically ∼5-10 nm), this is therefore evidence that the Pand F-containing species is at least this thick. Thus, it is likely that the PPF-derived SEI has low solubility (i.e., the solution may have become saturated) or is insoluble in the electrolyte.
Following cell formation, the graphite material was removed as described above and packed into rotors for ssNMR characterization. Given that PPF contains five times as many fluorine atoms as phosphorus and because 19 F has the larger gyromagnetic ratio, 19 F ssNMR spectra were presumed to have a stronger signal and were therefore measured first. Measurements from the treated graphite electrode sample are compared with the LiPF 6 salt, LiPF 6 complexed with EC, LiPO 2 F 2 , the PPF starting material, and LiF in Figure 3. The isotropic chemical shift of the LiPF 6 salt is −81 ± 1 ppm. The as-received PPF additive contained a center peak at −63.3 ± 0.8 ppm and an additional, smaller peak at −77.8 ± 0.8 ppm, each split by 19 F-31 P J-coupling of about 750-800 Hz, which can be removed by 31 P decoupling (spectra not shown). Additionally the 19 F ssNMR spectra of LiPO 2 F 2 and LiPF 6 complexed with EC were measured and their shifts are reported in Table I.
Given the proximity of the peak at −77.8 ppm to the LiPF 6 chemical shift, it is attributed to a small amount of pyridinium hexafluorophosphate impurity, [Py-H][PF 6 ]. A few weight percent of this impurity is consistent with the chemical analysis provided by the supplier. The spectrum measured from the treated graphite material (Figure 3a) also contains a peak centered at −75 ± 3 ppm, which resonates near the frequency of LiPF 6 complexed with EC ( Figure  3c) but is distinctly offset from that of the starting materials. This upper part of the major peak has a width of approximately twice that of LiPF 6 but resides on a broad base, which might encompass resonances from other species (for example PPF, LiPF 6 ) or be broadened by its vicinity to conducting graphite. This major peak shows a spin- ning sideband around −111 ppm. In addition the spectrum displays features at −152 ppm and at −205 ppm. The latter peak is assigned to LiF (δ LiF = −203 ppm 34,43 ). The origin of the peak at −152 ppm was not conclusively identified. Based on general 19 F chemical shift tables, highly substituted aryl fluorides such as benzenes, pyridines, styrenes and pyrimidines, resonate around this frequency range. 44 It is therefore possible that some C-H fluorination of the pyridine occurred, similar to the reaction described by Fier and Hartwig, which involves mixing pyridine with a fluorinated Lewis acid (AgF 2 ) under mild conditions. 45 However, this tentative assignment could not be confirmed in the present work.
The measured frequencies are spread over a wide frequency range, making it technically difficult to excite them simultaneously and consequently rendering their intensities non-quantitative. Nevertheless, these results may indicate the successful measurement of the PPFderived SEI component, produced during the formation protocol. Moreover, the presence of a LiF signal in Figure 3a is attributable to the decomposition of the electrolyte salt and/or of the PPF-derived species. However, it must first be considered whether the treated sample contains some amount of LiPF 6 from the dried electrolyte. Therefore, an identical sample was prepared with the additional step of gently, but thoroughly, rinsing the electrode with EMC prior to the removal of the graphite from the copper substrate. Removing dried electrolyte in this manner is commonly performed before surface analysis by XPS. 46,47 Experimentally concentrating on the major peak and acquiring the spectra at a lower magnetic field strength, Figure 4 compares the 19 F NMR spectra of the rinsed and unrinsed graphite samples (a and b, respectively) and a difference plot between the two (c). This difference spectrum is presented to show the line shape distinctions between the two experimental spectra and is constructed by scaling the relative intensities such that most of the signals cancel but no negative traces appear. The results demonstrate that the unrinsed spectrum consists of two superimposed signals centered at approximately the same chemical shift. However, the difference in their respective widths confirms that these are overlapping but distinct peaks. Since the sharp component shown in the lower most plot is not present in the 19 F NMR spectrum of the rinsed sample, the difference trace characterizes some of the material as a soluble F-containing species that was removed from the sample surface by the rinsing procedure. The peak corresponding to this soluble material is sharper than that of the insoluble species, which may indicate greater 19 F mobility.
On the basis of its chemical shift, the peak corresponding to the soluble component is not attributable to PPF or to LiPF 6 , but rather to   19 F atoms in a chemical environment that is intermediate between the two. The 19 F NMR spectrum of LiPF 6 complexed with EC (Table I) was indeed shifted relative to the pure salt, broadened and split into a doublet by J-coupling to 31 P. The average position of these two peaks is a good match for the soluble component shown in Figure  4c. Therefore, it is concluded that this soluble component was dried electrolyte, whereas the insoluble component observed in Figure 4b is attributed to a low solubility or insoluble PPF-derived SEI. That is, it is possible that this peak corresponds to the proposed Li 2 (PPF) 2 species. Such an environment is consistent with the proposed structure for the PPF-derived SEI component shown in Figure 1b. 13 To further characterize the SEI, 31 P ssNMR spectra were measured. Attempts to obtain a 31 P NMR signal from the treated graphite, by direct excitation or a Hahn echo method, were unsuccessful, possibly due to its lower concentration, a vicinity to the conducting graphite and the lower signal-to-noise of the 31 P nuclide, relative to 19 F. However, 19 F→ 31 P CP/MAS NMR spectra could be collected, as shown in Figure 5 and summarized in Table II, from the treated graphite, PPF starting material, and LiPF 6 salt. The use of this technique provides two major benefits: i) a significant increase in signal, due to the higher gyromagnetic ratio of 19 F, and ii) direct evidence of P-F bonds. The sample spectrum contains a sharp feature at −147 ± 3 ppm and a shoulder at approx. −160 ppm. Following the successful measurement of these spectra, it was considered whether the signal from the treated graphite arises from a P-containing component in the SEI or from residual LiPF 6 .
Figures 5a and 5b also compares 19 F→ 31 P CP/MAS spectra measured from the unrinsed and rinsed graphite electrode. As observed for the 19 F spectra above, the results indicate the presence of a sharper peak, corresponding to a soluble compound, and a broader underlying peak, corresponding to an insoluble species. The nature of the soluble component is unknown, but could be residual PPF, [Py-H][PF 6 ], or attributable to LiPF 6 that is complexed with residual EC from the electrolyte solution. This is consistent with the observations from the 19 F spectra. It is acknowledged that the signal-to-noise ratio is very poor, especially for the rinsed sample, despite the long acquisition times.
Thus, the 19 F→ 31 P spectrum alone could be considered inconclusive. However, when considered as a complement to the much stronger 19 F measurements, it is reasonable to conclude that the peak in Figure  5b is real, rather than a spurious signal. Because the spectrum of the rinsed material in Figure 5b was measured using a 19 F→ 31 P crosspolarization protocol, the presence of any peak must correspond to a compound that contains a P-F bond. Therefore, this result supports that there is a P-F bond in the PPF-derived SEI. Therefore, the results are consistent with the proposed SEI component in Ref. 13 and shown in Figure 1b.

Conclusions
In summary, this work utilized 19 F MAS and 19 F→ 31 P CP MAS to directly measure the chemical environment of the graphite electrode SEI, as formed electrochemically on the negative electrode surface of a full NMC442/graphite pouch cell in a full lithium-ion cell that contained the PPF additive. By repeating measurements on rinsed and unrinsed samples, the PPF-derived SEI could be distinguished from LiPF 6 in the dried electrolyte. The 19 F and 19 F→ 31 P ssNMR spectra provide direct evidence that F atoms and P-F bonds are present in the PPF-derived SEI. It can be concluded that the chemical environments of the F and P atoms are generally similar to, but distinct from, the respective environments of these atoms in the PPF and LiPF 6 starting materials. The results also provide evidence of the formation of LiF by-product, which may be attributable to decomposition of the PPFderived SEI and/or of the LiPF 6 electrolyte salt.
Overall, while the ssNMR spectra measured in this work are insufficient to independently identify the composition of the SEI, due to low signal-to-noise, especially for the 91 F→ 31 P measurements, the results are consistent with the previously proposed chemical structure of the PPF-derived species shown in Figure 1b. In general, the results in this work confirm previous recommendations that a large variety of experimental techniques, coupled with computational methods such as density functional theory calculations, are required to study the underlying chemical pathways of SEI formation in lithium-ion cells. Thus, the present work demonstrates that, for certain additives, ssNMR may be a valuable complement for studying the underlying chemical pathways by which electrolyte additives improve the lifetime and performance of lithium-ion cells.