Solid-State Nuclear Magnetic Resonance Studies of Electrochemically Discharged CF_x

Abstract

Electrochemical studies of three types of CF_x (F - Fiber based, C - Petroleum coke based, G - Graphite based) have demonstrated different electrochemical performances types in previous work, with fiber based CF_x delivering superior performance over those based on petroleum coke and graphite. ¹³C and ¹⁹F MAS (Magic Angle Spinning) NMR techniques are employed to identify the atomic/molecular structural factors that might account for differences in electrochemical performance among the different types of CF_x. Small quantitative variations of covalent CF and LiF are noted as a function of discharge and sp³ bonded carbons are detected in discharged F type of CF_x.

1 Introduction

The lithium/carbon monofluoride cathode system is known to have the highest theoretical energy density among solid cathode systems. Li/CF_x batteries[1] have remained attractive because of their very high energy density, long storage life, very good safety record, wide temperature range and very low self-discharge. However its use is restricted to specialized applications where superior performance is required such as aerospace, military and medical applications due to the relatively high cost of CF_x.

Typically x is approximately equal to one for commercial grade Li/CF_x battery material. The electrolyte typically consists of lithium tetrafluoroborate (LiBF₄) in gamma-butyrolactone or lithium hexafluoroasenate (LiAsF₆) in a mixture of propylene carbonate (PC) and dimethoxyethane (DME)[2]. The chemical reaction of the discharging process is summarized below.

$$\mathrm{nLi}\to {\mathrm{nLi}}^{+}+{\mathrm{ne}}^{-}\left(\text{Anode reaction}\right)$$

$${\mathrm{ne}}^{-}+{\left({\mathrm{CF}}_{\mathrm{x}}\right)}_{\mathrm{n}}\to \mathrm{nC}+{\mathrm{nF}}^{-}\left(\text{Cathode reaction}\right)$$

$$\mathrm{nLi}+{\left({\mathrm{CF}}_{\mathrm{x}}\right)}_{\mathrm{n}}\to \mathrm{nLiF}+\mathrm{nC}\left(\text{Overall reaction}\right)$$

The CF_x is converted into elemental carbon which is more conductive than CF_x thereby lowering the internal impedance, improving the voltage regulation and the efficiency. At the same time LiF is formed and precipitates out of the structure during the course of discharge.

Electrochemical studies have shown significant differences between each type of CF_x: prepared from carbon with a fiber morphology, hereafter referred to CF_x - F or F-type; prepared from petroleum coke, hereafter referred to as CF_x – C or C-type; and prepared from graphite, hereafter referred to as CF_x – G, or G-type. Although the fundamentals of the electrochemical discharging process are known, the structure of the CF_x cathode during the discharge and the mechanism of the defluorination process at the molecular level that might be responsible for the electrochemical differences are not known. This study is focused on NMR investigations of these three different types of CF_x materials to determine their structural/chemical differences, discharge mechanism and possible intermediate compound formation during discharge.

2 Experimental

CF_x materials (F, C and G types) were prepared by fluorination of fibrous, amorphous coke and standard graphite at 300 - 600 °C respectively provided by commercial sources. Unfortunately no further details were provided by the vendor. CF_x cathode mixes were prepared with raw CF_x, conductive carbon, and binder materials. Li/CF_x experimental cells were constructed with the three types of CF_x and they were discharged to various depth of discharge (DOD) under a 6 month discharge rate at 37°C.

¹³C MAS NMR (spin 1/2), is widely used, despite its relative paucity in naturally-occurring carbon (approximately 1%). Because of low natural abundance and generally long spin-lattice relaxation time (T₁), spectrum acquisition on samples which have not been experimentally enriched in ¹³C takes a long time. A Varian Inova 500 MHz spectrometer and Doty low C/F background probe were used to perform ¹³C MAS NMR measurements. Samples were packed in 4mm rotors and spun up to 12 - 14 KHz. Tetramethylsilane (TMS) was used as the external reference.

¹⁹F MAS NMR (spin 1/2), is relatively commonly measured, yields strong signals and has a wide chemical shift range. ¹⁹F MAS NMR measurements were conducted with a 300 MHz Varian S Direct Digital Drive NMR spectrometer, (the 300 MHz spectrometer was used because of the availability of higher spinning speeds) at spinning speeds from 32 KHz to 37 KHz in 1.6 mm rotors. A 90_x – 180_y pulse echo sequence was used to minimize the probe background signal. An aqueous solution of lithium trifluoromethylsulfonate was used as an external reference for fluorine at −77.8 ppm relative to the common reference CFCl₃ All the spectral analysis and deconvolution are performed by MestReNova software.

2.1 Deconvolution calibrations

The accuracy of deconvolution techniques and associated software was tested with known physical mixtures of raw CF_x - G and LiF. Four different mixtures of LiF/CF_x - G (LiF by weight - 10%, 20%, 50%, 75%) were investigated.

2.2 Slow (Greatbatch) discharged cells

Li/CF_x cells were electrochemically discharged for 6 months. The cell discharge voltage profiles and cell internal resistance vs. the DOD were obtained. Cathode materials were extracted at different DOD levels (Table 1) and washed thoroughly with DME to remove residual electrolyte salt.

Table 1.

DoD levels of Greatbatch discharged samples.

CFx – F%	CFx – C%	CFx – G%
1.2	1.2	1.2
7.5	7.9	7.8
22.2	23.0	23.3
45.2	46.5	45.4
67.2	69.8	67.9
89.3	92.1	86.7
97.0	99.4	95.6

Powder X-ray diffraction (XRD) measurements of discharged CFx samples were performed using a Shimadzu X-ray diffractometer XRD-6000, working with a CuKα radiation. SEM images were obtained by using LEO 1455VPSE Scanning Electron Microscope. ¹⁹F and ¹³C studies were conducted for a significant period of time (1 - 3 months) after sample preparation. Apart from MAS NMR measurements, nine samples (three of each type) were picked among 21 samples to perform ⁷Li and ¹⁹F T₁ measurements with inversion recovery pulse sequence.

2.3 Fast (Hunter) discharged cells

A focus of this study was to detect and investigate any short-lived intermediate species during the discharge process. NMR measurements were done in a short period of time (1-2 hours) after cell disassembly and washing with DME to improve the probability of detecting any metastable compounds. Six Li/CF_x batteries (two of each type) were electrochemically discharged at the Hunter lab with an Arbin battery cycler with a particular discharge protocol described below.

Cells were discharged at room temperature under 2.93 mA constant current according to Table 2. After the long discharge, cells were equilibrated at open circuit and discharged again for a short period of time (1 minute) at higher rate (75 mA) just before cell dis-assembly and NMR experiments. One particular cell (type C) with an estimated DoD of 80% could not sustain the short discharge at 75mA, due to the cell having exceeded the voltage limit set for the procedure. A possible reason for this is mentioned later. Therefore it was discharged for 3 minutes at 25mA instead of 1 minute at 75mA to acquire the target DoD.

Table 2.

Battery discharge protocol at Hunter.

CFx Battery Type	Discharged Time (hrs)	Target % DoD
F/C/G	309.12	46
F/C/G	537.60	80

Each battery was dissembled inside the argon filled glove-box. The cathode was extracted carefully and half of it was left alone and labeled as un-washed while the other half was rinsed with DME three times for 15 minutes each (labeled as washed). ¹³C and ¹⁹F MAS NMR studies of un-washed samples were performed immediately after the disassembly. Washed samples were left inside of the glove box antechamber for several hours to dry completely before NMR studies.

3 Results and Discussion

High speed MAS at 32 - 37 KHz greatly reduces the homonuclear dipolar coupling broadening of ¹⁹F spectra, shown in Fig.1 for the case of mixtures of CF_x and LiF. Sidebands were clearly identifiable and most of the features were easily distinguishable with no significant overlap. In discharged materials, four main features are prominent to different degrees depending on the sample that was studied. The most intense peak in all series at low discharge levels is around −184 to −187 ppm and was clearly larger for correspondingly lower level discharged samples, lower for higher level discharged samples, and assigned to covalent CF groups[3]. A semi-ionic CF peak is found around −170 ppm. The broad peak around −113 to −116 ppm was assigned to CF₂ groups located at the edge of graphite layers[3, 4]. A relatively sharp Teflon™ (present as a binder) peak occurs around −121 ppm. Upon discharging, the LiF resonance appears around −204 ppm[2]. Very slight changes of the positions of sidebands correspond to slight spinning speed variations. However this had a negligible effect on deconvolution and interpretation of the data.

Figure 1.

¹⁹F MAS NMR of manually mixed LiF/CF_x mixtures. The percentages given above refer to the fraction of LiF in the mixture. Asterisks denote spinning sidebands.

3.1 Deconvolution calibrations

Right sidebands are not shown in the spectra, (Fig. 1) for the sake of clarity but they have been included in the deconvolution shown in Fig. 2. LiF weight percentages were calculated from deconvolutions. Table 3 reveals that NMR measurements/deconvolution were very consistent with molar-weight-calculated values, although there is some uncertainty in deconvoluting the ionic and covalent CF features due to overlap (their ratios should not change in these spectroscopic standards).

Figure 2.

¹⁹F MAS NMR deconvolution of manually mixed LiF/CF_x mixtures. Numbers are molar percentages which correspond to weight percentage values in Table 3.

Table 3.

Comparison of weight percentage of LiF in known mixtures of CF_x – G and LiF with spectroscopic determination.

Percent from known mixture	Percent from NMR deconvolution
10.0	10.3
20.0	22.1
50.0	51.4
75.0	77.2

3.2 Slow (Greatbatch) discharged batteries

CFx experimental cells were discharged under 6 month rate at 37°C. The cell discharge voltage profiles are shown in Fig. 3. CFx G cells show a more sloped discharge voltage profile than CFx C and CFx F cells. At beginning of life, CFx G cells exhibit higher cell voltages than that of CFx C and CFx F cells. The CFx G cell voltages gradually decrease and cross over at around 20% to 35% DOD resulting in lower cell voltages than that of CFx C and CFx F cells at later DODs. The voltage profile of CFx C cells is slightly higher than that of CFx F cells throughout discharge. Both CFx C and CFx F cells exhibit a voltage plateau at around 2.8V until ~60% DOD and then gradually decrease. The discharge voltages from all three types of CFx cells merge to 2.3V at approximately 95% DOD.

Figure 3.

Discharge voltage profiles of CF_x F, CF_x G, and CF_x C cells under 6 month discharge rate.

Figure 4 shows the relationship of the cell internal resistance (Rdc) vs. DOD. As expected, all cells exhibit decreased Rdc when they were discharged from BOL to ~30% DOD due to the formation of carbon as one of the products. However, in the 2^nd half of cell discharge, the cell Rdc starts to increase. For CFx C, the cells have the lowest Rdc from BOL to ~35% DOD. Beyond 35% DOD, CFx C cells develop higher internal resistance relative to the cells from CFx G and CFx F. Beyond 80% DOD, CFx C cells exhibit a significant Rdc increase, about 7 times that of the other two CFx types. This could be related to the cell's response to the 75 mA pulse discharge described in the experimental section. Among the three types of CFx, CFx F cells exhibit the best stability with the least Rdc increase throughout discharge.

Figure 4.

CF_x type effect on cell internal Rdc change vs. %DOD.

All three types of CFx materials were electrochemically discharged to form carbon and LiF as the only detectable products as indicated by the XRD spectra shown in Fig 5. The CFx peaks at 13° and 42° are observed to deplete with increasing lithiation coincident with the increase in the prominence of the LiF peaks observed at 38°, 45°, 65° and 79°. A diffraction peak near 25°, which increases continuously with increasing lithiation, is assigned to graphite. The sharp LiF peaks indicate the crystalline nature of the LiF product. Further supporting evidence of the crystalline nature of the LiF is shown in the SEM images where cubic shaped crystals are observed in the lithiated samples, as shown in the case of CF_x-C in Fig. 6.

Figure 5.

X-ray diffractograms of CF_xC at three different DoD values.

Figure 6.

SEM image of CFx C at 1.2% DOD and CFx C at 69.8% DOD.

¹⁹F NMR spectra for very slightly discharged materials (1.2% DoD) are shown in Fig. 7, G type has the maximum covalent CF content and lower semi-ionic CF. The sum of the both covalent CF and semi-ionic CF in the 1.2% discharged F, C and G samples is about 81%, 84% and 89% respectively of the total fluorine. The spectral intensity of the Teflon™ peak was excluded from the deconvolutions based on the assumption that it is chemically inert throughout discharge. As expected CF and LiF peaks dominate the lower DoD levels and higher DoD levels, respectively as shown in Fig. 8 for the F series. G series shows the highest covalent CF throughout the whole range of discharge followed by C and F as shown in the deconvolutions displayed in Fig.9. Some kind of structural re-arrangement could be responsible for the small increase of covalent CF in first few DoD levels (around 7 - 8%), although this value is close to the deconvolution uncertainty of these two species as described earlier in the context of the manually mixed standards. The CF₂ content appears to decrease gradually upon discharging. In all three series, the semi-ionic CF is consumed faster than covalent CF in the electrochemically discharged cells, contrary to what was reported in chemically lithiated materials[2] (which is a comparatively very rapid process). Also no CF₃ groups around −50 to −70 ppm were noted as observed in previous results[2] for chemically lithiated CF_x.

Figure 7.

¹⁹F MAS NMR Spectra of Greatbatch discharged nearly starting materials (F, C, G). Asterisks denote spinning sidebands.

Figure 8.

¹⁹F MAS NMR Spectra of Greatbatch discharged F Series. Asterisks denote spinning sidebands.

Figure 9.

¹⁹F MAS NMR deconvolution of Greatbatch discharged F, C and G series.

¹³C MAS NMR spectra were broader with lower SNR (Signal to Noise Ratio) than ¹⁹F MAS NMR spectra although each sample was run for at least 48 hours. ¹³C MAS NMR via cross-polarization by fluorine would help to reduce the relaxation time considerably. However cross-polarization data are not accurately quantifiable as the resonances are excited unequally. This is especially the case as the structural fluorine is reduced by lithiation. Two clear peaks were prominent despite the low SNR. The one around 82 to 90 ppm is assigned to CF[5] as shown in Fig. 10. The graphitic carbon is present around 125 ppm[6] while CF₂ and Teflon™ peaks are overlapped heavily with the broad graphite carbon peak. As in the case of the ¹⁹F results, the intensity of the Teflon™ peak is not included in the ¹³C deconvolutions which are displayed in Fig. 11.

Figure 10.

¹³C MAS NMR Spectra of Greatbatch discharged F series.

Figure 11.

¹³C MAS NMR deconvolution of Greatbatch discharged F, C and G Series. The spectral results of C at 99.4% DoD are not included due to the low SNR characteristics of that sample. Although CF intensity appears in the ¹⁹F spectra for the ~ 89% DoD samples, it does not show up distinctly in the ¹³C deconvolutions due to large overlap.

In the ¹³C MAS NMR spectra the CF peak resonates at 84 - 88 ppm. The ¹³C NMR chemical shift of a carbon in a semi-ionic/ionic interaction with fluorine is near 89 ppm. Therefore this peak is not distinguishable from the covalent resonance. There are three closely spaced main features around −125 to −108 ppm assigned to bulk graphite, Teflon™ and components that are generally attributed to some type of CF₂ groups on the edge of the graphite planes. However, unlike the CF₂ edge groups seen in the ¹⁹F MAS NMR data, this resonance disappears upon complete discharge. A complication, however, is the overlap of the broad graphite resonance, which of course becomes a severe problem for deconvolution at successively higher DoD. It is therefore concluded that if this is indeed representative of CF₂ groups, they are bound differently to the bulk CF, and are more accessible (e.g. surface vs. bulk sites) to the lithium than those seen in the fluorine data. However it is noted that even the ¹⁹F data show decreasing CF₂ content with increasing DoD.

An sp³ bonded carbon species around 55 ppm and another around 65 ppm, also assigned to sp³ bonded carbon, appears in the middle stage of the electrochemically discharge samples. The F series shows the maximum sp³ bonded carbon amount which increases steadily from the middle to the end of discharge. The increase in bulk graphite content with increased discharge is also a consistent trend. Overall, the C series has the lowest CF consumption for a given nominal DoD while F and G maintain comparable amounts.

In order to check for possible differences in LiF formation between CF_x types, spin-lattice relaxation (T₁) measurements were performed. LiF is the only compound that appears in the ⁷Li spectra as shown in Fig. 12. There is a clear increase of T₁ in all the series with increasing DoD. ⁷Li T₁ of C and G series increase more slowly in comparison to the F series as shown in Fig. 13. One possible interpretation is that LiF precipitates out of the C/CF_x structure of the F series more readily as it more rapidly approaches bulk-like characteristics (i.e. longer T₁). Similar behavior of increasing T₁ with increasing DoD was observed in the ¹⁹F measurements as shown in Fig. 14. Finely dispersed nanoscopic LiF that is formed in the early stages of discharge is generally characterized by shorter T₁ values compared to its bulk macroscopic phase due to structural defects. The more rapid exclusion of LiF may be an explanation of the superior performance of the F series material, since LiF is an insulator. For completeness, a ⁷Li NMR spectrum of a nearly completely discharged sample is displayed in Fig. 12, and the spectrum is identical in appearance to that of bulk LiF. Debye-Scherrer analysis of LiF XRD peak widths showed an increase of crystallite size from low DoD values (< 10%) which appeared to level off by 40% DoD. There was no conclusive difference observed between the different forms of CFx, indicating that in the LiF product formed from these lithiated CFx samples, NMR, as a short-range structural probe, is more sensitive to lattice strains and defects than XRD

Figure 12.

⁷Li MAS NMR spectrum of Greatbatch 99.4% discharged F sample.

Figure 13.

⁷Li MAS NMR T₁ measurements of Greatbatch discharged batteries. The straight lines are guides to the eye.

Figure 14.

¹⁹F MAS NMR T₁ measurements of Greatbatch discharged batteries. The straight lines are guides to the eye.

3.3 Hunter discharged batteries

All the samples were run under comparable conditions described for the previous set. Although washed samples were rinsed with DME, significant amount of LiAsF₆ was present in the samples as indicated by the feature at −69.5 ppm overlapped with left sidebands of LiF (slightly overlapped) and covalent CF (heavily overlapped) as shown in Fig. 15. Some samples were run at slightly lower spinning speed (32 KHz) due to mechanical issues and hence slight differences in the positions of sidebands are observed in the spectra. One clear observation is the LiF contents of Hunter discharged batteries are higher than those of Greatbatch discharged batteries especially in F and G series. That is, the actual LiF content is higher than the amount prescribed by the nominal DoD value. The reason for this could be the inhomogeneity of the sample in the different parts of the cathode which experienced different discharge rates. The ¹⁹F deconvolutions are displayed in Fig. 16. It appears that all of semi-ionic CF is consumed at the higher DoDs which is consistent with the results for the slower discharged samples.

Figure 15.

¹⁹F MAS NMR Spectra of Hunter and Greatbatch discharged batteries. Asterisks denote spinning sidebands.

Figure 16.

MAS NMR Spectra of Hunter discharged batteries for F, C and G series.

¹³C MAS NMR spectra reveal more subtle but important details. Three representative sets of spectra are displayed in Figs. 17-19, for washed and unwashed samples. The unwashed samples are characterized by large residual solvent peaks, despite the post-rinse vacuum treatment. The spectral intensity ratios of the two main peaks, Graphite and CF, are similar in the washed and unwashed samples. Only F series consistently shows both features assigned to sp³ bonded carbon at 55 and 65-68 ppm as in Greatbatch slow discharged samples. C Series shows two sp³ bonded carbon peaks in the same region. However at 80% DoD the peak around 65 −68ppm was absent. Also the peak around 55ppm is not as strong as it is in the 46% DoD sample. The G Series is quite different. At 46%, no peaks are observed around 55 and 65-68ppm. However a small peak around 65ppm is observed at 80% DoD.

Figure 17.

MAS NMR Spectra of Hunter 80% discharged F series.

Figure 19.

¹³C MAS NMR Spectra of Hunter 46% discharged C series.

4 Conclusions

¹⁹F and ¹³C NMR of electrochemically lithiated CF_x of different morphologies reveal largely similar behavior of LiF formation. Unlike chemically lithiated samples[2] the semi-ionic CF consumption is faster than covalent CF in electrochemically discharged cells. Samples undergoing relatively fast discharge exhibit higher LiF contents at the corresponding DoD values, most likely due to inhomogeneity. An important observation was that no metastable compounds that can be detectable by ¹⁹F or ¹³C NMR were found in the fast discharged samples. ¹³C MAS NMR data also showed some similar behavior of carbon formation. However one significant observation was the consistent appearance of sp³ bonded carbon in the F series at higher DoD values. Given the superior performance of the F series at higher DoD values, it is surmised that there is a structural rearrangement of the carbon lattice involving sp³ bonded carbon which stabilizes the material at high DoD levels. Supporting evidence for the stability of F-type CF_x comes from both ⁷Li and ¹⁹F T₁ measurements which suggest that LiF formed upon discharge is excluded from the structure of the material more readily than in the C and G series.

Figure 18.

¹³C MAS NMR Spectra of Hunter 46% discharged G series.