Resolution of Lithium Deposition versus Intercalation of Graphite Anodes in Lithium Ion Batteries: An In Situ Electron Paramagnetic Resonance Study

Abstract In situ electrochemical electron paramagnetic resonance (EPR) spectroscopy is used to understand the mixed lithiation/deposition behavior on graphite anodes during the charging process. The conductivity, degree of lithiation, and the deposition process of the graphite are reflected by the EPR spectroscopic quality factor, the spin density, and the EPR spectral change, respectively. Classical over‐charging (normally associated with potentials ≤0 V vs. Li+/Li) are not required for Li metal deposition onto the graphite anode: Li deposition initiates at ca. +0.04 V (vs. Li+/Li) when the scan rate is lowered to 0.04 mV s−1. The inhibition of Li deposition by vinylene carbonate (VC) additive is highlighted by the EPR results during cycling, attributed to a more mechanically flexible and polymeric SEI layer with higher ionic conductivity. A safe cut‐off potential limit of +0.05 V for the anode is suggested for high rate cycling, confirmed by the EPR response over prolonged cycling.


Introduction
Li ion batteries (LIBs) play an increasingly important role in the energy system, so the development of an understanding of their degradation mechanisms is apre-requisite to extending battery life. [1] Graphite,w ith at heoretical capacity of 372 mAh g À1 ,iswidely used as the anode material in LIBs.Li metal deposition onto graphite can cause serious safety and degradation problems,a nd time-resolved monitoring of this process is challenging. [2] In common with other cell degradation processes,L id eposition can be mitigated by additives within the electrolyte,t he capacity ratio between the anode and cathode,t he temperature,a nd the charging/discharging rate. [2a-c, 3] Although it is normally assumed that the local potential difference at the anode/electrolyte interface has to fall below 0V (on the Li + /Li scale) to overcome the overpotential associated with Li nucleation and growth, conditions most likely to be met during fast charging,r ecent work from Manthiram and co-workers indicates that underpotential deposition of Li can occur during lithiation of graphite (at ca. + 0.1 Vvs. Li + /Li), triggered by dissolved Mn ions from the cathode material. [4] Va rious non-destructive real time techniques have been used to clarify the lithiation process and the failure mechanisms. [5] In situ nuclear magnetic resonance (NMR) spectroscopy can reveal structural phase changes during lithiation via the 7 Li frequency shift, [5b,d] even with early stage lithium graphite intercalation compounds (e.g. Li 1/12 C 6 ). [5b] Electron paramagnetic resonance (EPR) spectroscopy is ap owerful alternative technique which can be harnessed to provide real time information on the chemical changes within cells, through its sensitivity to the density and environment of unpaired electronic spins.I ns itu EPR studies of electrochemical cells have av ery long pedigree,b ut the technique has been surprisingly underused as am eans to interrogate batteries under operando conditions. [6] This is all the more surprising given the insights made into LIB chemistry through the application of NMR spectroscopy. [5d,e,7] Of the notable operando EPR studies of LIBs to date,o ne report describes the Li-rich layered oxide Li 2 Ru 0.75 Sn 0.25 O 3 as the positive electrode,w hich has shown that its high capacity (> 270 mAh g À1 )i sd ue to the reversible formation of (O 2 ) nÀ species. [6c] However,the number of real-time EPR studies of lithiation/de-lithiation of carbon based anode materials is still limited. [8] Thef irst in situ EPR study of ac arbon-based material in aLIB was reported in the mid-90s, [8a] and studies on lithiation of carbons attributed different EPR signals to "Pauli spins" (electrons at the Fermi level) and "Curie spins" (localised electrons) due to lithiation at ordered and disordered structures,r espectively. [8c, 9] Wandt et al. demonstrated the power of EPR to enable time-resolved and quantitative study of Li plating on ag raphitic electrode,d istinguishing between Li intercalation and deposition behaviour, in operando studies at À20 8 8Cu sing al ithium iron phosphate counter/reference electrode. [6i] In this work, aconcentric geometry,three-electrode in situ EPR cell was designed with am etallic Li cathode and graphitic anode to study (de-)lithiation of the LIB anode at room temperature.W eu se this approach to assess the formation of the solid electrolyte interface (SEI) layer on the 1 st potential cycle,w hich is based on the change in microwave skin depth via its effect on the conductivity of whole cell. We further report quantitative analyses of the EPR spin density changes during multiple lithiation/delithiation cycles of the graphite anode and correlate these data with electrochemical measurements on the same cell. Theparallel anode processes,Liintercalation and deposition, are resolved by spectral simulation:irreversible deposition on the graphite is seen to begin at higher potentials (! + 0.04 V vs.Li/Li + )even at low scan rates (0.04 mV s À1 ). Theinhibition of the anode degradation by the VC electrolyte additive is confirmed by the decrease in the Li signal during long-term cycling.

Results and Discussion
Theinsitu EPR cell consisted of athree-electrode system in aq uartz tube (diameter 2mm; Figure 1). Ad etailed description of the cell fabrication is given in the Experimental section (Supporting Information, Figure S1-S3).
Thecell is positioned in the EPR resonator such that only the exposed graphite layer is in the sensitive part of the cavity; the Li electrode is not in the active part of the cavity (as proven by the absence of ab ulk Li signal on initial set up). Thec oncentric geometry of the in situ EPR cell gives an electrochemical performance which is close to that of the corresponding coin cell ( Figure S2) without compromising on the spectroscopic sensitivity.T hus it enables study of the phase transformations of the lithiated graphite anode,and of the Li deposition process,o nc ycling under realistic conditions.
TheEPR spectroscopy of conducting materials is sensitive to microwave skin depth effects, [10] this has been observed in Li x C 6 compounds as af unction of lithiation. [6i,11] Thee nhancement in conductivity on lithiation decreases the EPR resonator quality (or Q)f actor and hence the EPR signal intensity.T om onitor and account for this we have followed the approach outlined by Wandt et al., [6i] through use of aM nO reference sample external to the cell (detailed description in Experimental section 1.3 in SI;F igure S4a). TheM nO signal intensity as af unction of the applied potential ( Figure S4b) reflects the changes in the resonator Q, [6i] decreasing on sweeping to lower WE potentials, correlating with the increased lithiation and hence conductivity of Li x C 6 ( Figure 2). Hence,t he maxima and minima in the MnO signal intensity ( Figure 2) correlate with the discharged and fully charged states,respectively,ofthegraphitic electrode.W eh ave followed this process over multiple potential cycles ( Figure 2). On the first cycle,d ecreasing from ah igh starting potential, the MnO signal intensity increased slightly at around 1.3 V, and began to decrease again at around 0.65 V. Since the pristine material starts from adischarged state (also shown by the graphite EPR signal;see below), this shallow maximum in the MnO signal during the 1 st charging process cannot be due to de-lithiation of the graphite.Hence,itseems likely that it can be attributed to the solid electrolyte interface (SEI) formation, which has poor electronic conductivity. [12] It is also useful to exploit the evolution of Q to reveal the improved reversibility of the cell due to the VC additive.The SEI interface takes more time to form without the VC,w hich is deduced from the slower rate of change in the background response.
TheE PR spectrum of the pristine graphite gives aw eak signal with aD ysonian lineshape centred at g = 2.015, indicative of the existence of mobile electrons from intrinsic defects in the graphite (peak-to-peak linewidth ca. 20 G; Figure 3a). [13] Thes pin density of the pristine graphite was calculated to be 3.79 10 17 spins g À1 (see Experimental section 1.4, SI). EPR spectra were recorded every ca. 6mins while sweeping the applied potential at 0.1 mV s À1 starting from 1V to 5mV( vs.L i/Li + )i nL P57 (1 ML iPF 6 in 3: 7 ethylene carbonate/ethyl methyl carbonate (EC/EMC)) with 2% VC additive.T his spectrum changes little between OCV and + 0.6 Vo nd ecreasing the potential in the initial cycle ( Figure 3a). Specifically,t here are no noticeable changes in the EPR spectrum as we sweep through ca. 1V ,that is,where the first peak is seen in the external MnO standard signal that monitors the resonator Q-factor.Hence,the above deduction that the change in the MnO signal must be associated with the formation of the SEI at the graphite surface.However, we do not observe any free radical intermediates which are thought (1) Exposed and (2) insulatedC uwire (diameter 0.5 mm) act as the current collector for the working electrode (WE); (3) graphite anode (length 1.5 cm, mass loading ca. 0.4 mg cm À2 , thickness5 0-100 mm) coated onto the exposedp art of the WE current collector (1);(4) separator (Celgard 2325, thickness25mm), preventing short circuit between the graphite layer and the Li;(5) Li metal layer as counter electrode (CE, length 3cm);(6) twined, exposed Al wire (diameter 0.1 mm) and (7) insulated Al wire (diameter 0.1 mm) as the current collector for the CE;( 8) Li deposited onto (9) exposed Cu wire (diameter 0.2 mm) as the reference electrode (RE);the RE is placed in the middle of the exposed part of the graphite WE.
to be generated by single electron reduction of EC or the VC additive: [14] these are presumably very short lived and may be detectable by spin trapping methods.S ignificant changes in the EPR signal are observed at potentials below ca. + 0.6 Von the first sweep.Between + 0.55 Vand + 0.42 V, asignal grows rapidly in intensity with decreasing potential, accompanied by adecrease in g-value to 2.006 with adecrease in linewidth to 3G (Figures 3b,4e,f). On further decrease of the potential from + 0.4 Vt o+ 0.005 V, the linewidth (2.5 G) and g-value (2.006) stabilized, while the signal intensity increased rapidly (Figure 3c). Thesignal could be fitted with asingle Dysonian function ( Figure S5 in SI), [6i, 8a] and its behavior is consistent with the formation of Li x C 6 phases on charging of the cell by decreasing the potential. These changes in EPR signal were found to be reversible on increasing the potential from + 0.005 to + 1V (Figure 3d-f), as the cell is discharged. On loss of the Li x C 6 signal at higher potentials on the first cycle, an ew,n arrow and very weak signal is apparent (Figure 3d). This is attributed to the formation of Li metal and will be discussed below. [6i, 15] Figure 4s ummarizes the in situ EPR characterization of the graphite anode with 2% VC additive in LP57, including

Angewandte Chemie
Research Articles the spin density (S;a fter calibration accounting for the changing resonator Q factor), the first derivative of S with respect to potential (dS/dV), the linewidth and the g-value, charge (Q el )c alculated from integration of the current recorded in the electrochemical experiment. We have correlated the EPR signal intensity (which is am easure of spin density;F igure 4d)with the total charge passed as measured by electrochemical measurements (Figure 4c)o nt he same cell. Theshapes of the curves (spin density vs.charge), or their first derivatives (comparison with current data) track remarkably closely.Asexpected, the absolute EPR spin density is smaller than the charge injected, because EPR can only detect unpaired spin density which, in these conducting materials,are due to electrons at the Fermi level. [8c, 9] EPR is aq uantitative technique for spin density,h ere reflecting the dynamic lithiation/de-lithiation processes.T he Li content in Li x C 6 can be calculated from the charge number ( Figure S2d): the data for the second cycle are more reliable because the first cycle involves irreversible SEI formation. The"gas type" stages (! LiC 72 ), formed at potentials higher than + 0.42 V, are less readily detected by other techniques such as NMR, XRD,R aman spectroscopies due to their lower sensitivities. [5b,c,16] In contrast, this is the regime in which we observe rapid changes in g and linewidth by EPR (Figures 3a nd 4). Based on the charge passed, over the potential range 1.0 Vto 0.42 V, we observe changes in the EPR response corresponding to an average formulation of LiC 102 (based on the 2 nd CV, Figure S2d). This highlights one advantage of EPR:itis very sensitive to unpaired electrons and is blind to diamagnetic materials which will mask the spectral response in other techniques.These early stages (above + 0.4 V) are thought to relate to the shallow insertion of lithium into the graphite edge.T he EPR spin density increases slowly,with aconstant linewidth and g-value,u ntil + 0.2 Vw hich corresponds to LiC 32 based on the charge trace (Figure 4b), corresponding to the formation of the stage 4( LiC 36 ). [16] Thes pin density ( Figure 4d)increased more rapidly from + 200 mV to + 5mV (corresponding to LiC 6.9 ), reflecting the phase transformation of dilute stage 4t od ense stage 2/1. [5b] Although we can correlate the changes in the EPR spectra with different lithiation stages,wedonot observe these as discrete regions: this may be due to the limited data points over the relatively narrow potential range (200 mV to 5mV). or possibly to the co-existence of distinct phases as the Li + penetrates further into the electrode. [5f] Thed e-lithiation process became faster when the potential increased from 5mVt o0 .4 V, and the complete de-lithiation occurred at around 0.8 Va st he linewidth and the gv alue returned back to the state of the pristine graphite.T he in situ EPR data from the second potential cycle are very similar to those from the first, indicating areversible process ( Figure S6).
Results from an equivalent experiment in the absence of the VC additive in LP57 are summarized in Figure S7  graphite anode (i.e.Li x C 6 ), linewidth and g value,isshown in Figure S7 and S8, and was similar to the results presented above (Figure 3, 4), that is,w ith VC additive.B yc ontrast, as tronger metallic Li signal was observed from the graphite anode after cycling ( Figure S7 d, e), which is attributed to the irreversible formation of am uch greater extent of "dead" lithium. [2b, 3f,6i, 8a] Then otable difference compared with the case of VC presence is the irreversibility of the charging process in the non-VC case (reflected by the higher residual spin density after de-lithiation over different cycles from Figure S9), indicating that the VC additive acts to improve the electrochemical reversibility.The EPR signal of graphite after the 1 st discharge process appeared significantly different (vide infra) with the strong new signal indicating the irreversible formation of degradation products generated during cycling.
As noted above,o nr epeated charge/discharge cycling ar esidual EPR signal is found in the discharged state which increases in intensity on cycling,and the lineshape simulation confirmed the existence of two different components.T his signal is significantly sharper than that attributed to Li x C 6 , with alinewidth of ca. 1.25 G, and is characteristic of Li metal ( Figure S10). [6h, 17] Hence,i ti sc oncluded that Li metal is deposited on the graphite at low potential even though our vertex potential is + 0.005 V, as can be easily distinguished by two different EPR signals shown in Figure S11. This signal is not observable at low potentials (in the presence of VC) because the weaker Li signal is masked by the much more intense Li x C 6 resonance ( Figure 3). TheL im etal signal has ac haracteristic Dysonian lineshape ( Figure S12), which can be analyzed to estimate the particle size from the EPR linewidth and lineshape,specifically the A/B ratio (the peakto-trough amplitude). [6h, 15, 17] Thel ineshape is due to microwave skin depth effects:particles much smaller than the skin depth give essentially isotropic signals,w hereas bulk Li typically gives A/B of ca. 8a tX -band and room temperature. [6h,i, 15, 17] Forour cell the A/B ratio of the weak Li metal peak after de-lithiation, in the presence of the VC additive ( Figure S11a) was 1.75 after the first cycle,i ncreasing to 2.3 after the second cycle.T he increase in A/B therefore reflects the increase in size of the Li deposits for which, following Pifersa nalysis, [15] we estimate the mean diameters to be ca. 1.2 mma nd ca. 1.6 mma fter the first and second de-lithiation cycles,r espectively,a ssuming as kin depth of 1.1 mma t am icrowave frequency of 9.8 GHz. [17] Increasing the vertex potential to + 25 mV led to no significant "dead" Li formation over the first three cycles,which suggests that this vertex is as afe threshold for further long cycling (Experimental section 1.5;F igure S13).
In the absence of VC in the LP57, there are two notable changes,b oth of which indicate more rapid deposition of Li metal on the anode.F irst, the signal attributed to Li metal is much more intense after ag iven number of cycles.F or example,after the first discharge sweep,the Li metal signal is an order of magnitude more intense (referenced to the intrinsic graphite defect spectrum) in the absence of VC ( Figure S11b and Figure S12). Second, we find ab igger A/B % 4.5/5.5 (1 st /2 nd cycle,r espectively) indicating at hicker deposit, with estimated diameters of 2.3 mma nd 2.5 mm. Hence,the evidence suggests the VC additive plays asignifi-cant role in suppressing the irreversible formation of Li metal deposits,w hich is in accord with the conclusions drawn by Dahn and co-workers. [3f] Given that the signal from the Li metal deposit is much stronger (with respect to that from Li x C 6 )i nt he absence of VC,i ti sp ossible to monitor it over the lower potential range (Figure 5a,black curve). Thesignal cannot be detected on the initial charging sweep but (as above) it is apparent on the first discharging sweep,a nd it then increases in intensity after fully discharging on repeated cycling. On the first discharge,adecrease in the signal assigned to Li metal is seen, but at as lower rate of loss than the Li x C 6 signal. Hence,t he Li metal deposition is only partially reversible.O nt he second charging process,f urther deposition of Li metal starts when the potential is below ca. + 0.1 V: the signal intensity change indicates that approximately one-quarter of the Li deposited during the second cycle is "dead", although this analysis does not consider the effects of changing particle size,vide infra (the corresponding EPR signal showing aresponse due to Li x C 6 and Li 0 is shown in Figure S11 b). When the scan rate is lowered to 0.04 mV s À1 (Figure 5a,r ed curve), the deposition potential further decreases to ca. + 0.05 V: for aN ernstian redox process,t his potential limit would correspond to the reduction of 14 %o f the lithium ions in the vicinity of the electrode.The A/B ratio of % 2/2.9 after full de-lithiation during the 1 st and 2 nd cycles, respectively,g ives estimated particle sizes of 1.5 mma nd 1.8 mm, respectively.T he electrochemical stripping of deposited Li begins at al ower onset potential (e.g. < 0.3 V), with aquick decrease of the Li signal. Further slow diminishing of the Li should be due to the side reaction of "dead" Li with the electrolyte.T he initial increase in the Li signal during discharge may be related to the skin effect. Thes tripping of Li reduces the particle size,asdemonstrated by apeak in the A/B value near the lower vertex potential ( Figure S14). Therefore we can speculate that the EPR signal intensity continues to increase because more of the Li is becoming detectable as the particle size decreases and this initially outweighs the decrease in the total amount of Li. [6l] Our results indicate that classical "over-charging" (potentials 0V vs.L i + /Li)i sn ot required for Li metal deposition to occur. Heterogeneous deposition of Li, rather than homogeneous deposition on the graphite,m ay be more favorable. [2c] Thep otential of the onset of Li deposition is currently debated: [3g] although av alue of À150 mV has been reported for deposition using ap yrolytic graphite sample, [5g] anumber of reports suggest deposition can occur at potentials > 0V . [3g,h] Forexample,the onset of Li deposition on agraphite anode was reported to be at + 48 mV when the graphite/Li coin cell system was charged at C/100 currents at room temperature, [18] From at hermodynamic perspective,t his is reasonable,a st he equilibrium potential, rather than the onset, of Li deposition is 0V .F rom ak inetic perspective, deposition at potentials > 0Vis more surprising since metal deposition normally requires an overpotential associated with the phase formation process.O nt he other hand, underpotential deposition of metals is also known, with recent works suggesting this phenomenon occurs in Li + ion cells. [4] Other factors may be at play,i ncluding the structural heterogeneity of polycrystalline graphite,w hich could cause local potential/current deviations.H ence it is possible that, while the net potential applied to the anode is > 0V ,there is some local fluctuation in potential due to particle-particle conductivity.T he properties of the SEI interface will also affect the Li deposition process:L id eposition occurs under the SEI layer and relies on the Li + distribution determined by the chemical composition and the physical microstructure of the SEI layer. [2b,c] TheV Ca dditive helps to impart the polymeric SEI layer with increased mechanical flexibility and increased ionic conductivity. [3f, 14b,19] Consequently,the current distribution on the graphite anode tends toward homogeneity and Li + intercalation into graphite is more uniform. TheSEI structure,w hich contains various inorganic Li-based salts without the VC additive,issusceptible to breaking or cracking due to the mechanical stress induced during the lithiation/delithiation process.T his leads to an inhomogeneous distribution of the current density over the anode surface.T hus partial overcharging can occur, which causes subsequent Li deposition. Li deposition can occur readily as the de-solvated ions under the SEI interface are as hort distance from the graphite interface.T he scan rate,0 .1 mV s À1 ,c orresponds to ahigh Crate (3C), and is also areason for the high potential Li deposition. Li deposition on the graphite anode is only partially reversible,t hat is,s ome "dead" lithium remained after each cycle,a nd there is as ignificant accumulation of "dead" Li after the second cycle for the cell in the absence of VC.T his indicates that the VC additive can suppress the formation of metallic Li 0 on the graphite anode and thus improve the electrochemical reversibility of the system. Theprevention of Li formation by the VC additive is also reflected in the long-time cycling at ah igher scan rate of 2mVs À1 from 1Vto 0.05 V, as shown in Figure 5b.Nosignal due to "dead Li" is seen during the first 4cycles with the contribution of VC additive,only asmall amount of "dead" Li is generated, which increased only slightly during further cycling. In the absence of VC,arapid increase of "dead" Li over the initial cycles was detected on the graphite anode, followed by aw eak rise and approximately linear increase with further cycling.The large increase of Li at the beginning of cycling might be related to the slowly-formed inhomogeneous SEI interface as described above,l eading to rapid deposition. Thes ubsequent linear increase after 10 cycles is possibly due to the poor mechanical strength of the SEI without VC additive,a nd the associated stress during lithiation/de-lithiation, which is likely to crack the SEI layer. Figure 5. a) the EPR intensity of metallic Li 0 deposition at graphite anode during the first two cycles with VC (blue) and without VC additive (black) at 0.1 mVs À1 and without VC at lower scan rate of 0.04 mVs À1 ;b )Li 0 formation on the graphite surface during cycling from 0.05 Vto1V at 2mVs À1 .The EPR signal intensity shown in (a) and (b) was normalised to the signal intensity of the pristine graphite. The missing data in (a) on the first charging cycle for all three curves, and the lower potentialr ange for the VC result (blue curve), was due to the swampingofthe weak Li 0 signal by the Li x C 6 at the low potentialr ange.

Conclusion
In situ EPR spectroscopy has been used to understand the electrochemical behavior, including lithiation and the Li deposition, of the graphite anode upon voltammetric cycling in ab espoke three-electrode EPR cell operated at room temperature.The EPR resonator quality factor,monitored by an external MnO standard, reflects the conductivity of the lithiated graphite and suggests that the SEI layer formation started at around 1.3 V. TheE PR spectra with an arrowing process at higher potential (! 0.42 V) is correlated to the ("gas type") stages,w hich are not readily detected by other techniques (e.g. NMR is typically limited to mM concentrations of spins,w hile mMc oncentrations are detectable by EPR [20] ). Thel ithiated graphite (Li x C 6 , x < 6/100) shows as ingle Dysonian lineshape with ac onstant linewidth (2.5 G) and g value (2.006). Thes pin density calculated by the EPR spectra synchronize with the evolution of the charge during cycling,a nd the first derivative of the spin density summarizes the phase transformation of lithiation process. TheD ysonian lineshape simulation helps to separate the contribution from the Li x C 6 and the Li deposition, assisted by the much smaller linewidth of the latter, being around 1.2 G. Further analysis reveals that the onset of Li deposition of the in situ cell occurs at higher potentials:for example,ca. + 0.1 V for ascan rate of 0.1 mV s À1 and ca. + 0.05 Vfor 0.04 mV s À1 , which is mainly due to the inhomogeneous SEI layer formation. TheVCadditive has asignificant inhibitory effect on the Li deposition during prolonged cycling, which is attributed to increased mechanical flexibility of the polymeric SEI layer compared to that formed under non-VC conditions.