Structure of spontaneously formed solid-electrolyte interphase on lithiated graphite determined using small angle neutron scattering JPCC2015

Structure of Spontaneously Formed Solid-Electrolyte Interphase onLithiated Graphite Determined Using Small-Angle NeutronScattering

Robert L. Sacci,*,† Jose Leobardo Banuelos,*,‡,⊥ Gabriel M. Veith,† Ken C. Littrell,§

Yongqiang Q. Cheng,∥ Christoph U. Wildgruber,∥ Lacy L. Jones,∥ Anibal J. Ramirez-Cuesta,∥

Gernot Rother,‡ and Nancy J. Dudney*,†

†Materials Science and Technology Division, ‡Chemical Sciences Division, §High Flux Isotope Reactor, and ∥Spallation NeutronSource, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, United States

*S Supporting Information

ABSTRACT: We address the reactivity of lithiated graphite−anode material for Li-ion batteries with standard organic solventsused in batteries (ethylene carbonate and dimethyl carbonate) by following changes in neutron scattering signals. The reactionproduces a nanosized layer, the solid-electrolyte interphase (SEI), on the graphite particles. We probe the structure andchemistry of the SEI using small-angle neutron scattering (SANS) and inelastic neutron scattering. The SANS results show thatthe SEI fills 20−30 nm sized pores, and inelastic scattering experiments with H/D substitution show that this “chemical” SEI isprimarily organic in nature; that is, it contains a large amount of hydrogen. The graphite−SEI particles show surface fractalscattering characteristic of a rough particle−void interface and are interconnected. The observed changes in the SEI structure andcomposition provide new insight into SEI formation. The chemically formed SEI is complementary and simpler in compositionto the electrochemically formed SEI, which involves a number of different reactions and products that are difficult todeconvolute.

■ INTRODUCTION

One of the fundamental questions confronting battery researchregards the mechanism of the solid-electrolyte interphase’s(SEI) formation.1,2 The physicochemical properties of the SEIfunction to mitigate Li-ion transport and protect the electrolytefrom the highly reducing anode.3 As such, control of the SEI’sformation has important implications in the safety, irreversiblecapacity loss, energy density, and cost of a battery.4 Typically,the SEI is formed during the initial charging of a battery, andwhen formed, it protects the anode from reacting with theelectrolyte, while allowing Li ions to migrate to and from theanode.5 During this initial charging, three general reactionsoccur as depicted in Figure 1a: electrochemical reduction ofelectrolyte (Rxn 1), Li intercalation (Rxn 2), and reduction ofelectrolyte by intercalated Li (Rxn 3). It is unclear which ofthese reactions occur first and how they follow from one

another; that is, SEI formation contains the proverbial “chickenand the egg” problem.The chemistry and structure of the electrochemically formed

SEI have been studied using a wide variety of techniquesincluding infrared spectroscopy, differential scanning calorim-etry, quartz crystal microbalance, and so on.2,3,6,7 These studiessupport the depiction of the SEI being a composite,multilayered structure, with a dense inner layer composed ofLi2CO3 and LiF, and a porous polymeric outer layer.8 Recentdevelopments in novel in situ techniques, such as electro-chemical transmission electron microscopy,9,10 have demon-strated the changes in the structure and thickness of the SEIduring charge and discharge; that is, these studies showed that

Received: January 8, 2015Revised: March 18, 2015Published: March 25, 2015

Article

pubs.acs.org/JPCC

© 2015 American Chemical Society 9816 DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

the SEI is neither chemically nor structurally static. Theoperando small-angle neutron scattering (SANS) study byBridges et al.11 on electrochemical charge/discharge cycling oflithium-intercalated graphite reported that the SEI formedalong the surfaces of a mesoporous framework within the anodematrix; however, because of the complexity of the electro-chemically formed SEI, the scattering data had to be analyzedincluding a relatively large number of chemical species, whichpresented ambiguities in extracting SEI thickness, structure, andso on.Because of the complex nature of SEI formation, we seek to

simplify its chemistry by bypassing the two electrochemicalevents, that is, by decoupling the reactions in Figure 1a. Thiswas done here by first chemically lithiating graphite and thenexposing it to a mixture of ethylene carbonate (EC) anddimethyl carbonate (DMC), solvents typically used in Li-ionbatteries, as depicted in Figure 1b. Because there are no Lisalts present in the solvent, that is, the solvent is free of LiPF6or similar components, the system is relatively simple with onlya few products possible. This approach, therefore, serves as amodel for SEI growth by simplifying the chemistry and allowsus to focus on the evolution of porosity and surface area duringSEI formation.SANS and inelastic neutron scattering (INS) were used to

quantify the structure and composition of the mesoscale SEIformation that occurs on graphite surfaces and within intergrainvoids. SANS is a powerful technique for the study of structuresin the 1−200 nm size range and can provide statisticalinformation about complex systems with a multiscaledisordered structure. Furthermore, because of the highpenetration power of neutrons, studies of processes at buriedinterfaces and in nanoscale pore spaces are feasible.12 As a bulkstructural probe, SANS provides representative microstructural

information over the entire sample of up to several hundredcubic millimeters. Because neutrons interact with the nucleusand the strength of that interaction does not scale linearly withatomic number, SANS is sensitive to isotopic substitution aswell as to certain light atoms, including the hydrogen isotopedeuterium. We exploit this fact to differentiate between an SEIcomposed primarily of Li-carbonate (hydrogen absent) andpolymeric products (hydrogen-bearing).Attempts to obtain chemical information using optical

spectroscopy were unsuccessful, as the graphite dominatessignals from infrared absorption and Raman spectroscopy;however, INS is extremely sensitive to vibrational modes ofhydrogen in the sample.13 Therefore, if the SEI layer isprimarily organic, then vibrational bands from the layer will bevisible over the graphite in the INS spectrum. This featuremakes INS the only vibrational technique that allows for suchsurface enhancement on black carbons.

■ EXPERIMENTAL SECTION

Synthesis. The graphite source (2 g, MGPA, PredMaterials) and desired molar ratio of stabilized lithium metalpowder (FMC, Inc.) were added to a gastight high-densitypolyethylene (HDPE) container along with 7 g of 5 mmdiameter ZrO2 milling media. The container was then sealed ina stainless-steel (SS) bomb and locked into a SPEX 8000 Mmixer/mill. It was shook vigorously (1725 rpm) for 90 minbefore being brought back into the glovebox. Fresh lithiatedgraphite (0.3 g) was then soaked in deuterated or protiatedEC/DMC solution (3:7 wt %) for 3 days. Excess solvent wasremoved in vacuo for 5 h at 1 × 10−4 Torr. A raw milledgraphite control experiment was performed, and we found noevidence of scattering due to residual EC; that is, EC is fullysublimed under these conditions.

Figure 1. Schematic of the different strategies for forming an SEI along with the types of reactions involved. The electrochemical process (a)incorporates two distinct electrolyte reduction mechanisms: an electron enters the system and reduces an adsorbed molecule/ion (green layer) andintercalated Li exits the lattice and reduces an adsorbed molecule/ion (blue layer). These processes occur parallel and will form a complicatedpassivation layer (cyan). The chemical strategy (b) utilized in this report only allows the chemical reaction of intercalated Li with ethylene carbonateor dimethyl carbonate (Rxn 3).

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9817

SANS Experiments. The sample holder was a CR2023 SScoin cell (depicted in the Supporting Information) as perBridges et al.11 Between 0.15 and 0.3 g of the powdered samplewas packed into an aluminum spacer ring. An SS disk andspring were placed atop the sample, and the cell was

hermetically sealed with a coin cell crimper. All sample coincells, including an empty one for background subtractions, wereloaded on a multisample translation stage of the general-purpose small-angle neutron scattering diffractometer (GP-SANS) located on beamline CG-2 at the High Flux Isotope

Figure 2. (a) Scattering curves of graphite, LiC6, and LiC6 + d-EC. The inset table summarizes the scattering exponent of all samples showing thatthe formed SEI roughens the surface of the graphite particles. (b) SANS curves normalized by Qp(LiCx) highlight changes on the intergrain pore scales.The intensity at Q < 0.01 Å−1 reflects the change in the carbon/SEI system composition in comparison with the surrounding macrospaces. Theregion Q ≈0.014 Å−1 is used to provide a quantitative measure of the amount of polymeric SEI, residual EC, and remaining empty pore space givenin Table 2. At Q > 0.2 Å−1, the increased scattering is due to ∼1 nm size domains within the newly developed phase on the surface of the carbonsubstrate. (c) Inelastic neutron scattering of EC, LiC6 + EC/DMC, and washed LiC6 + EC/DMC. Asterisks (*) indicate major peaks from ascribedto EC18 and daggers (†) correspond to peak locations of poly(ethylene oxide)-type vibrations.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9818

Reactor (HFIR), Oak Ridge National Laboratory.14,15 Scatter-ing intensity as a function of momentum transfer, Q, wascollected for each sample. (Q is related to the d spacing (d),scattering angle (2θ), and radiation wavelength (λ), via Q = 2π/d = 4π sin(θ)/λ.) Measurements were carried out intransmission geometry, and an 8 mm diameter beam size waschosen to avoid scattering from the spring, spacer, and gasketalong the edges of the sample cell. Measurements were carriedout in three instrument configurations: (1) 1.1 m sample todetector distance (SDD), wavelength λ = 4.72 Å; (2) 7.8 mSDD, λ = 4.72 Å; and (3) 19.3 m SDD, λ = 12 Å to obtain awide Q range and sufficient overlap in Q betweenconfigurations. The incoming neutron wavelength was setusing a velocity selector, with a Δλ/λ ≈ 15%. See theSupporting Information for more details.INS Experiments. INS measurements were carried out on

beamline 16B, the Vibrational Spectrometer (VISION) at SNS,ORNL, Oak Ridge, TN. VISION is high-flux high-throughputINS instrument covering a broad energy range from −2 to 1000meV.16 Samples (∼2 g each) were loaded into 8 mm diametervanadium cylindrical canisters sealed with copper gaskets, thencooled to the base temperature (∼5 K). An empty V-canister ofthe same type was measured for background subtraction.Supporting characterizations of the samples using X-ray

diffraction (XRD) and N2 adsorption (Brunauer, Emmet, andTeller (BET)) are described in the Supporting Information.

■ RESULTS

Lithiation and Solvent Reaction. High-energy ball-milling was used to increase the surface area17 and providethe energy to promote Li intercalation.18 Electron microscopeimages provided in the Supporting Information (Figure S1)show that the ball-milling procedure produced high-aspect-ratiolithiated carbon grains of ∼3 μm in thickness and between 8and 18 μm in diameter.18 BET measurements show that thetotal surface area in the initially smooth nonporous graphitechanges from 1.1 m2 g−1, to ∼88 m2 g−1, with 10 m2 g−1 of thesurface area corresponding to milling-formed micropores(Figure S2 in the Supporting Information). These substantialincreases in surface area and microporosity indicate that thegrain surface roughness increased. The packing fraction of thecarbon particles is ∼0.5; that is, approximately half the samplevolume consists of macropores and nanopores. The macroporedimensions are on the same micron length scale as the carbonparticles and not directly observable by the SANS technique;however, nanopores, which are voids formed by the rough-surface contacts of the particle grains, are amenable to SANSstructural characterization. Therefore, the structural changescaused by ball-milling and subsequent exposure to solvent maybe directly probed by SANS.After exposure to and reaction with the carbonate solution

and subsequent vacuum removal of excess solution, thelithiated graphite showed a mass increase compared to theinitial LiCx mass. LiC12 gained significantly more mass thanLiC6 in both the deuterated and protiated samples; however,the mass increases could not be fully assigned to SEI; that is,some unreacted EC was still present. (See the SupportingInformation.) LiC6 showed a large shift in the (002) diffractionpeak from 22 to 25°, as shown in the XRD patterns (Figure S3in the Supporting Information), which is sensitive to the Li−Cstoichiometry. This indicates that a fraction of the intercalatedLi reacted with the solvent and exited the graphite galleries.The reaction with LiC12 also caused a color change (purple to

black), and (002) peaks from LiC24 (26°) and raw graphiteappeared (27°). The change in mass, coupled to the decrease inthe gallery spacing, suggest the formation of an SEI. The puregraphite under identical solution exposure and removal showedno mass change, and the SANS signal showed no change (seethe Supporting Information), indicating both no reactionstaking place and complete solvent removal.

SANS Characterization. The SANS plot in Figure 2ashows that when graphite is lithiated, the scattering intensity,I(Q), decreases. We report I(Q) as a function of momentumtransfer, defined as Q = (4π sin θ)/(λ), where λ is the neutronwavelength and θ is half the scattering angle. The size of theobserved features scale approximately as 2.5/Q; that is, largefeatures appear at small Q, and vice versa. In Figure 2a, theSANS signal of the graphites originates from scattering by thegraphite−air interface and the nanoscale pore spaces betweenthe milled particles. For a two-phase system, I(Q) isproportional to the squared difference in the neutron coherentscattering length density (NSLD) of the phases, (Δρ*)2,calculated from the sum of the coherent scattering lengths of allof the atoms in a given phase. Therefore, changes in I(Q) canbe correlated with a change in NSLD of one phase or both.Carbon has a high NSLD, while voids have a NSLD of zero, sovoids in carbon scatter strongly. Because the scattering lengthof naturally abundant Li is negative, lithiation of graphite leadsto a lower scattering contrast with voids and thus a lowerscattering intensity. When the LiCx is reacted with the EC/DMC mixture, the intensity decreases over the Q range of0.005 to 0.4 Å−1, which is due to a reduction of NSLD contrastbetween lithiated carbon and solvent-filled pores. Deuteratingthe solvent further decreases the scattering intensity becausethe SEI phase has an NSLD closer to that of carbon.The lithiated carbon (phase C), the reaction product phase

(phase SEI), and the void spaces (phase 0) give a total scatteredintensity, which is weighed by the square of the NSLDdifference of the cross-terms, (ρi,j*)

2, and a volume fractionfactor, cn, such that

ρ ρ ρ= * + * + *I c I c I c I( ) ( ) ( )total 1 C,02

C,0 2 C,SEI2

C,SEI 3 SEI,02

SEI,0

(1)

Deuteration modifies the contrast at two of these interfaces,namely, the carbon−SEI interface and the void−SEI interface.Analyzing the ratio (ρC,SEI* /ρSEI,0* )2 for both the protiated anddeuterated cases clearly illustrates the influence of deuterationon the interfacial scattering in Itotal. For the protiated case, thisratio is ∼15.2, whereas for the deuterated case it is ∼0.38. Thisdifference by a factor of 40 tells us that structural details of thecarbon−SEI product interface are better obtained from theprotiated sample, while the details of the void−SEI productinterface are better obtained from the deuterated sample.In the range 0.02 to 0.1 Å−1, the signal decays as I(Q) ∝ Q−p

so that the scattering law is given by the slope of the SANScurve in the log−log plot. The scattering exponents for theunreacted carbons are ca. 3.5. When the carbons react with d-EC:DMC, p decreases to ca. 3.0; however, reaction with h-EC:DMC produces only a small decrease in the scatteringexponent (inset in Figure 2a). The decrease in the scatteringexponent from 3.5 to 3.0 indicates that the carbon surfaceexhibits surface fractal scattering associated with a roughsurface, and the subsequent SEI product coats the carbonsurface with an even rougher, nearly mass-fractal-type structure(p < 3 corresponds to mass fractals).19,20 The observeddecreases in scattering intensity and scattering exponent upon

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9819

solvent hydrogen isotope substitution show that the SEIcontains hydrogen, and intimately binds to the carbon to makean interface with the remaining void space, which is muchrougher than that of the initial lithiated carbon.The Porod plot in Figure 2b highlights subtle structural

changes at the intergrain pore scales of the LiCx−solventreaction. Here we normalize the I(Q) to the scatteringexponent, Qp(LiCx), of pristine LiCx−air system. Within themid-Q range (0.005 to 0.1 Å−1), the scattering is slightlyincreased by natural hydrogen-bearing SEI and significantlydecreases with the deuterated SEI layer, regardless of LiCx

stoichiometry. At Q > 0.2 Å−1, the scattering increases for bothprotiated and deuterated SEI, indicating the formation ofnanodomains along the SEI−LixC6 interface. The scatteringincreases due to overall sample composition changes involvingdeuterated SEI for Q < 0.01 Å−1; however, because this part ofthe SANS signal is convoluted with some degree of multiplescattering, quantitative comparisons between the samples arenot made in this region.INS Characterization. Figure 2c shows the INS spectra of

pure EC, LiC6, and LiC6 + EC/DMC, and LiC6 + EC/DMCafter being washed three times with DMC and dried for 24 h.Note that the LiC6 background contributes little to the overallspectral intensities. That is, INS is more sensitive toward ananometer-sized layer of hydrogen-bearing compounds than to2 g of carbon.13,21 As expected from the XRD results, EC isclearly present in the SEI layers of the unwashed samples,which is evidenced by the presence of large peaks at 110, 210,and 235 cm−1 in both neat EC and unwashed samples, as wellas the minor peaks 525, 680, and 880 cm−1.22 Careful washingof the reacted carbon with DMC removes the unreacted EC, asseen with the removal of the major peak at 110 cm−1 and allminor peaks. The washing allows for new peaks at 92 and 150cm−1 to be resolved. These peaks are similar to those ofpoly(ethylene oxide) (PEO),23 which is expected if the SEI ismore polymeric in nature. The spectrum of the washed samplealso contains original EC peaks at 210 and 235 cm−1 that areblue-shifted by 4 cm−1 as well as broadened; however, thesepeaks can also be assigned to an oxygenated polymer, especiallyin the absence of the minor EC peaks.22

■ DISCUSSION

The electrochemical reduction of EC in DMC in the presenceof Li-salts has been well-studied using electrochemical quartzcrystal microscopy and infrared and Raman spectroscopy.7 Thereduction potential of EC, as written in eqs 2−4, is ∼1.2 Vversus Li/Li+; therefore, its reduction is thought to be the majorcontributor of the SEI.2,8,24 Zhuang et al.6,25 compared theinfrared adsorption of synthesized Li2(CH2OCO2)2 with theelectrochemical reduction of EC, and found strong evidence ofit being the primary product. If the intercalated Li has access tosurface sites, it will act as a reducing agent (E = 0.05 V vs Li/Li+) on EC to form carbonate, ethylene gas, and intermediateradical species that can polymerize via7,24,26

+ →− −(CH ) CO e CH CH OCO2 2 3 2 2 2 (2)

· + → +− − −CH CH OCO e (CH ) (g) CO2 2 2 2 2 3

2(3)

· →− −2 CH CH OCO (CH CH OCO )2 2 2 2 2 2 2

2(4)

DMC is known to reduce to form lithium methoxide andLi2CO3; however, because EC is a major component of the SEI,we assume its reaction is more favorable.24 Given that the XRD

diffraction profiles are indicative of the Li−C stoichiometry, wecan approximate the extent of the reaction; that is, the reactionbetween LiC6/12 and EC is of the form

+ +

→ + − +

⎜ ⎟

⎜ ⎟

⎛⎝

⎞⎠

⎛⎝

⎞⎠

x

x x

2LiC1

2(CH ) CO

LiC1

2Li CO Li (CH CH CO )

6 2 2 3

12 2 3 2 2 2 2 2

(5)

+ +

→ + − +

⎜ ⎟

⎜ ⎟

⎛⎝

⎞⎠

⎛⎝

⎞⎠

x

x x

2LiC1

2(CH ) CO

LiC1

2Li CO Li (CH CH CO )

12 2 2 3

24 2 3 2 2 2 2 2

(6)

where 0 ≤ x ≤ 0.5 (disregarding ethylene gas production).Therefore, we expect the surface layer to be a mixture ofLi2CO3 and Li2(CH2CH2CO3)2, with the latter representingthe polymeric or hydrogen-bearing component of the SEI. TheNSLDs of these compounds and the graphite species are givenin Table 1. The NSLD difference between deuterated andprotiated EC, along with XRD results, was used to calculate therelative amounts of carbonate and polymer comprising the SEI.

Chemical Determination of the SEI. INS is sensitive tovibrational modes involving hydrogen and has no selectionrules.13 Therefore, if the layer were primarily organic, thenbands from the layer would be visible even above a 2 g carbonbackground.13,20 The major result from the INS spectra (givenin Figure 2c) is that the SEI is polymeric in nature. The red-shifting (5 cm−1) and broadening of the original EC peaks at110 and 210 cm−1 may result from either EC trapped withinthe SEI or the SEI itself;22 however, if said peaks of theextensively washed sample were from EC, then major vibrationsat 680 and 880 cm−1, corresponding to H2C scissoring andasymmetric stretching, respectively, would be expected to bepresent.22 These peaks were not found, and the shape of thespectra more closely resembles a linear alkyl-oxide chain,23

which supports the polymeric SEI interpretation; however, caremust be taken as the hydrogen-containing vibrational bands inLi2(EC)2 overlap with those of EC, which again, overlap tothose of an extensive polymer like PEO. (See the SupportingInformation.) This is because vibrations associated withethylene groups (−CH2−CH2−) tend to appear around 100and 200 cm−1. In this regard, it is difficult to come to adefinitive chemical identification based solely on INS; however,hydrogen isotopic substitution in the SANS measurementshelps to discern between these components.

SEI Filling Pores. Filling of the close-contact intergrainvoids by the SEI is observed as a decrease in I(Q) near Q =0.014 Å−1 (arrow in Figure 2a). The SANS signal is

Table 1. Neutron Scattering Length Densities of LithiatedCarbons and Deuterated (d)/Protiated EC Used in ThisStudy, Along with Possible Lithium Carbonate andPolymeric SEI Products

Phase NSLD, 10−6 Å−2 Phase NSLD, 10−6 Å−2

C6 7.36 EC 2.02

LiC6 6.4 d-EC 5.78

LiC12 6.86 Li2(EC)2 1.4

Li2CO3 3.48 d-Li2(EC)2 4.25

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9820

proportional to the square of the NSLD difference between theLi−graphite substrate and the contents of the pore space, that is

ρ

ρ ρ

= = × Δ *

= × * − *

−I Q A m

A m

( 0.014 Å ) / ( )

/ ( )

1C

2

C LiC pore2

x

x x (7)

where A is a constant determined from the pristine samplescattering and mCx

is the carbon mass of the LixC6 substrate

used for normalization. Because the NSLD of the deuterated Li-EC species is close to that of LixC6 (Table 1), Δρ* is expectedto be small for a primarily (deuterated) organic SEI. Thescattering signals resulting from deuterated components in theintergrain pore spaces will show a greater decrease in SANSintensity compared with the protiated samples. The largedifference between the scattering from the deuterated andprotiated products and the INS spectra support the fact thatSEI is primarily organic in nature. In other words, the formationof Li2CO3 is less favorable or kinetically slower than polymerformation.The XRD results showed some residual EC; therefore, we

treat the NSLD of the nanopore space, ρpore* , as an averageNSLD of three phases, an organic SEI phase, solid EC, andempty space, such that

ρ ρ ρ ρ* = * + * + *c c cpore 1 SEI 2 solidEC 3 empty (8)

where the cn are normalized volume fractions, such that ∑cn =1. The deuterated samples showed more mass gain than theprotiated samples, which corresponds to the increased amountof EC detected within the sample by XRD (Figure 2S in theSupporting Information). Because the XRD of the LiCx + EC/DMC samples showed no signals of a Li2CO3 phase, weconstrained the fitting of the pore volume fraction in eq 7 tothe normalized scattering intensity using eq 8. The results areprovided in Table 2. In the protiated samples, LiC6 showedmore SEI growth than LiC12, which suggests that the drivingforce for EC reduction is greater for LiC6 than LiC12. This isexpected due to the greater amount of Li present.We follow the changes in the pore volume and specific

surface areas (SSAs) of the particles within the nanopores bymodeling the deuterated samples’ scattering data usingPRINSAS,27 a program that fits a polydisperse distribution ofspheres to the scattering data, as illustrated in Figure 3a−d anddiscussed in detail in the Supporting Information. The model isintended for two-phase systems, for example, carbon and air(Figure 3b), and the SEI forms a third phase (Figure 3c);therefore, only qualitative information can be obtained bycomparison of the scattering signals of the LiCx−SEI with thepristine LiCx. Given that the NSLD of the deuterated

Table 2. Percent Composition of Products of LixC6 + EC/DMC and the Mass Gaina

LiC6 LiC12 high stage C6 mass gain % c1 SEI c2 solid EC c3 empty

LiC6 + EC/DMC 19 81 0 0 11 0.419 0.581

LiC12 + EC/DMC 0 100 0 0 48 0.172 0.828

LiC6 + d-EC/d-DMC 28 72 0 0 52 0.419 0.123 0.458

LiC12 + d-EC/d-DMC 100 0 0 161 0.172 0.472 0.356aDeuterated samples display large EC diffraction peaks that agree with the increase in mass gain %. The volume fractions of phases 1, 2, and 3 thatgive the average pore NSLD, ρpore* , are displayed. The SEI fraction, c1, is lower for LiC12 than LiC6, corresponding to less SEI formation.

Figure 3. (a−d) Regions that are probed by SANS. The SEI causes agglomeration of the carbon particles and decreases the empty volume within thenanopores. PRINSAS software26 fits the empty space in the intergrain voids with a distribution of spheres to produce volume-normalized pore sizedistributions (e) and the specific surface area (SSA) (f) of the initial LiC6 and LiC12 compared with those reacted with d-EC/DMC. The SSA valuesfor r near 0.4 nm agree with the BET−obtained SSA.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9821

electrolyte is close to the value of the carbon, we treat the SEI/carbon as a single phase, and aim to identify the pore scales atwhich changes (SEI formation) occur. As previously discussed,the material filling the intergrain pores consists mainly ofLi2(EC)2 along with some unreacted solid EC. Using theseassumptions, we obtained pore volume distributions and SSAfor the samples, as shown in Figure 3e and f, respectively. TheSSA calculated for pristine LiCx was ∼40 m

2 g−1, which is closeto the N2−BET value of milled graphite (88 m2 g−1). Gasadsorption gives a higher value because the probe (N2

molecule) is smaller than the minimum r in the calculation,and nitrogen sorption may assess finer surface features thanSANS, thereby yielding a higher total surface area.The apparent pore size distribution changes drastically after

exposure of the carbon substrate to d-EC/DMC. Thedistributions of both the total pore spaces and SSA decreaseby one to two orders of magnitude in the reacted samples, withthe largest observed relative porosity decrease at about r = 10−15 nm, showing that SEI fills the voids. A new peak forms inthe pore size distribution between 1 and 2 nm, which is on thelength scale of the SEI nanodomains within the SEI phase. Thisfeature is indicative of the SEI layer, causing increased particleroughness on this length scale. The largest decrease in porevolume distribution and SSA is seen with the LiC12 system,which shows that the voids of LiC12 become filled to a greaterextent than LiC6; however, only 20% of the filling phase ispolymeric (Li2(EC)2) in LiC12, whereas LiC6 contains 77% ofpolymeric (Li2(EC)2) as the filling phase (Table 2); therefore,it is inferred that the pores of the LiC12 samples are filled withSEI and unreacted EC. Again, this coincides with LiC6 beingmore reactive than LiC12. For r > 50 nm, LiC6 shows littlechange and LiC12 shows an increase in the surface area. Anincrease in the SSA agrees with the observation that thesurfaces become rougher (as observed by the change in thescattering exponent, Figure 2a) after exposure to the electro-lyte. For a more quantitative assessment, additional datacollected at lower Q-values and with less multiple scatteringthan the present experiment are necessary.The SANS results are in good agreement with various post

mortem analyses of the SEI formation in Li-ion batteries byelectron microscopy,8,24 and are qualitatively similar to theSANS results of Bridges et al.11 A number of in situexperiments show that the SEI can be thicker, upwards of100 nm before “collapsing”, when formed at high overpotentialsand in the presence of excess electrolyte.9,10,28,29 A dense andthick SEI is desirable, as it would provide better anodeprotection without sacrificing Li-transport rates.24

■ CONCLUSIONS

We have shown that neutron scattering techniques can be usedeffectively to investigate the SEI that forms on battery materials.While the SEI is a complicated system, we have attempted tosimplify the system by chemically lithiating graphite, increasingits surface area, and then exposing it to EC/DMC solventmixture to construct the SEI piecewise. Ball milling thegraphites produced a nanotextured interface; solvent filling andSEI growth along this interface was readily observed usingSANS. By coupling SANS and INS, and assuming that the SEIwas formed from the sole reaction with EC to produce acombination of Li2CO3 and an organic salt (e.g., Li2(EC)2), wefind that the SEI is larger on LiC6 than on LiC12, and that littlelithium carbonate is formed. We showed that the organiccomponent of the SEI had vibrational features similar to PEO,

and that EC binds strongly with the surface of lithiated graphitedue to the formation of high surface area SEI. Further analysisof the SANS data showed that the Li−graphite surfaces becamerougher, exhibiting nearly mass fractal scattering due to the SEIstructure that contained 1 to 2 nm sized domains. Theformation of SEI seems to block access of solvent molecules tothe pore surfaces, as unfilled nanoscale spaces remained in thereacted samples.

■ ASSOCIATED CONTENT

*S Supporting InformationDetails of the synthesis of the lithiated graphite, SANS and INSexperiment, and SANS modeling. SEM image of the milledcarbons along with particle size distribution (S1); N2 and CO2

gas adsorption isotherm curves obtained from N2(g) andCO2(g) demonstrating micropore formation from milling (S2);XRD and representative Rietveld refinement of productsshowing the amount of material reacted (S3); SANS of carbonsalong with control reaction between graphite and EC/DMC(S4); and calculated volume fractions of the SEI and carbons(S5). This material is available free of charge via the Internet athttp://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Authors*E-mail: [email protected]. Tel: +1 865 241 5135 (R.L.S.)*E-mail: [email protected]. Tel: +44 (0) 1235 445923(J.L.B.)*Email: [email protected]. Tel: +1 865 576 4874 (N.J.D.)

Present Address⊥J.L.B.: ISIS Facility R3 UG15, STFC Rutherford AppletonLaboratory, Harwell, Didcot, Oxon, OX11 0QX, U.K.

Author ContributionsThe manuscript was written through contributions of allauthors. All authors have given approval to the final version ofthe manuscript and contributed equally.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The experiments and authors (R.L.S., J.L.B., G.R., and N.J.D.)were supported as part of the Fluid Interface Reactions,Structures and Transport (FIRST) Center, an Energy FrontierResearch Center funded by the U.S. Department of Energy,Office of Science, Office of Basic Energy Sciences (BES-DOE),and as part of a user proposal by Oak Ridge NationalLaboratory’s Spallation Neutron Source (Y.Q.C., C.U.W.,L.L.J., A.J.R.-C.) and High Flux Isotope Reactor (K.C.L.),which are sponsored by the Scientific User Facilities Division,BES-DOE. Additional experimental support for G.M.V. wasprovided by Materials Science and Engineering Division of theU.S. BES-DOE.

■ ABBREVIATIONS

SANS, small-angle neutron scattering; SEI, solid-electrolyteinterphase; EC, ethylene carbonate; DMC, dimethyl carbonate;PEO, poly(ethylene oxide); INS, inelastic neutron scattering;SSA, specific surface area

■ REFERENCES

(1) Armand, M.; Tarascon, J.-M. Building Better Batteries. Nature2008, 451, 652−657.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9822

(2) Aurbach, D. Electrode-Solution Interactions in Li-Ion Batteries: aShort Summary and New Insights. J. Power Sources 2003, 199, 497−503.(3) Aurbach, D.; Ein-Eli, Y.; Chusid, O.; Carmeli, Y.; Babai, M.;Yamin, H. The Correlation Between the Surface-Chemistry and thePerformance of Li-Carbon Intercalation Anodes for RechargeableRocking-Chair Type Batteries. J. Electrochem. Soc. 1994, 141, 603−611.(4) Aurbach, D.; Zinigrad, E.; Cohen, Y.; Teller, H. A Short Reviewof Failure Mechanisms of Lithium Metal and Lithiated GraphiteAnodes in Liquid Electrolyte Solutions. Solid State Ionics 2002, 148,405−416.(5) Peled, E. In Li Batteries; Gabano, J. P., Ed.; Academic Press:London, 1983.(6) Zhuang, G. V.; Yang, H.; Blizanac, B.; Ross, P. N. A Study ofElectrochemical Reduction of Ethylene and Propylene CarbonateElectrolytes on Graphite Using ATR-FTIR Spectroscopy. Electrochem.Solid-State Lett. 2005, 8, A441−A445.(7) Moshkovich, M.; Gofer, Y.; Aurbach, D. Investigation of theElectrochemical Windows of Aprotic Alkali Metal (Li, Na, K) SaltSolutions. J. Electrochem. Soc. 2001, 148, E155−E167.(8) Lithium-Ion Batteries: Solid-Electrolyte Interphase; Balbuena, P. B.,Wang, Y., Eds.; Imperial College Press: London, 2004.(9) Sacci, R. L.; Dudney, N. J.; More, K. L.; Parent, L. R.; Arslan, I.;Browning, N. D.; Unocic, R. R. Direct Visualization of Initial SEIMorphology and Growth Kinetics During Lithium Deposition by InSitu Electrochemical Transmission Electron Microscopy. Chem.Commun. 2014, 50, 2104−2107.(10) Sacci, R. L.; Black, J. M.; Balke, N.; Dudney, N. J.; More, K. L.;Unocic, R. R. Nanoscale Imaging of Fundamental Li BatteryChemistry: Solid-Electrolyte Interphase Formation and PreferentialGrowth of Lithium Metal Nanoclusters. Nano Lett. 2015, 15 (3),2011−2018.(11) Bridges, C. A.; Sun, X. G.; Zhao, J.; Paranthaman, M. P.; Dai, S.In Situ Observation of Solid Electrolyte Interphase Formation inOrdered Mesoporous Hard Carbon by Small-Angle NeutronScattering. J. Phys. Chem. C 2012, 116, 7701−7711.(12) Feigin, L. A.; Svergin, D. J. Structure Analysis by Small-Angle X-Ray and Neutron Scattering; Plenum Press: New York, 1987.(13) Mitchell, P. H. C.; Parker, S. F.; Ramirez-Cuesta, A. J.;Tomkinson, J. Vibrational Spectroscopy with Neutrons: With Applicationin Chemistry Biology and Materials Science; World Scientific: Singapore,2005.(14) Berry, K. D.; Bailey, K. M.; Beal, J.; Diawara, Y.; Funk, L.; Hicks,J. S.; Jones, A. B.; Littrell, K. C.; Pingali, S. V.; Summers, P. R.; et al.Characterization of the Neutron Detector Upgrade to the GP-SANSand Bio-SANS Instruments at HFIR. Nucl. Instrum. Methods Phys. Res.,Sect. A 2012, 693, 179−185.(15) Wignall, G. D.; Littrell, K. C.; Heller, W. T.; Diawara, Y.;Melnichenko, Y. B.; Bailey, K. M.; Lynn, G. W.; Myles, D. A.; Urban,V. S.; Buchanan, M. V.; et al. The 40 m General Purpose Small-AngleNeutron Scattering Instrument at Oak Ridge National Laboratory. J.Appl. Crystallogr. 2012, 45, 990−998.(16) Seeger, P. A.; Daemen, L. L.; Larese, J. Z. Resolution ofVISION, a Crystal-Analyzer Spectrometer. Nucl. Instrum. MethodsPhys. Res., Sect. A 2009, 604, 719−728.(17) Lu, M.; Tian, Y.; Yang, Y. A Comparison of ElectrochemicalPerformance of Natural Graphite Sulfurized by Ball-Milling and Heat-Treating as an Anode for Lithium Ion Batteries. Electrochim. Acta2009, 54, 6792−6796.(18) Sacci, R. L.; Adamczyk, L. A.; Veith, G. M.; Dudney, N. J. DrySynthesis of Lithium Intercalated Graphite Powder and Fiber. J.Electrochem. Soc. 2014, 131, A614−A619.(19) Bale, H. D.; Schmidt, P. W. Small-Angle X-Ray-ScatteringInvestigation of Submicroscopic Porosity with Fractal Properties. Phys.Rev. Lett. 1984, 53, 596−599.(20) Teixeira, J. Small-Angle Scattering by Fractal Systems. J. Appl.Crystallogr. 1988, 21, 781−785.(21) Anjos, D. M. Carbon Paper on Quinones Adsorbed on Nano-Onions. Carbon 2013, 52, 150−157.

(22) Fortunato, B.; Mirone, P.; Fini, G. Infrared and Raman Spectraand Vibrational Assignment of Ethylene Carbonate. Spectrochim. Acta1971, 27A, 1917−1927.(23) Kearley, G. J.; Johansson, P.; Delaplane, R. G.; Lindgren, J.Structure, Vibrational-Dynamics and First-Principles Study of Diglymeas a Model System for Poly(Ethyleneoxide). Solid State Ionics 2002,147, 237−242.(24) Peled, E.; Golodnitsky, D. In Lithium-Ion Batteries: Solid-Electrolyte Interphase; Imperial College Press: London, 2004.(25) Zhuang, G. V.; Xu, K.; Yang, H.; Jow, T. R.; Ross, P. N. LithiumEthylene Dicarbonate Identified as the Primary Product of Chemicaland Electrochemical Reduction of EC in 1.2 M LiPF6/EC:EMCElectrolyte. J. Phys. Chem. B 2005, 109, 17567−17573.(26) Zhang, X.; Kostecki, R.; Richardson, T. J.; Pugh, J. K.; Ross, P.N. Electrochemical and Infrared Studies of the Reduction of OrganicCarbonates. J. Electrochem. Soc. 2001, 148, A1341−A1345.(27) Hinde, A. L. PRINSAS−A Windows-Based Computer Programfor the Processing and Interpretation of Small-angle Scattering DataTailored to the Analysis of Sedimentary Rocks. J. Appl. Crystallogr.2004, 37, 1020−1024.(28) Unocic, R. R.; Sun, X.-G.; Sacci, R. L.; Adamczyk, L. A.; Alsem,D. H.; Dai, S.; Dudney, N. J.; More, K. L. Direct Visualization of SolidElectrolyte Interphase Formation in Lithium-Ion Batteries with In SituElectrochemical Transmission Electron Microscopy. Microsc. Micro-anal. 2014, 20, 1029−1037.(29) Zeng, Z.; Liang, W.-I.; Liao, H. G.; Chu, Y. H.; Zheng, H.Visualization of Electrode−Electrolyte Interfaces in LiPF6/EC/DECElectrolyte for Lithium Ion Batteries via In Situ TEM. Nano Lett. 2014,14, 1745−1750.

The Journal of Physical Chemistry C Article

DOI: 10.1021/acs.jpcc.5b00215J. Phys. Chem. C 2015, 119, 9816−9823

9823