Study of the Initial Stage of Solid Electrolyte Interphase Formation

Aug 22, 2012 - Anthony P. O'Mullane,. § and Salvy P. Russo. ‡,§,*. †. School of Physics, The University of Melbourne, Parkville, Victoria 3010, ...
0 downloads 0 Views 1MB Size
Article pubs.acs.org/JPCC

Study of the Initial Stage of Solid Electrolyte Interphase Formation upon Chemical Reaction of Lithium Metal and N‑Methyl‑N‑PropylPyrrolidinium-Bis(Fluorosulfonyl)Imide Akin Budi,†,‡,§ Andrew Basile,‡,§ George Opletal,§ Anthony F. Hollenkamp,‡ Adam S. Best,‡ Robert J. Rees,‡ Anand I. Bhatt,‡ Anthony P. O’Mullane,§ and Salvy P. Russo‡,§,* †

School of Physics, The University of Melbourne, Parkville, Victoria 3010, Australia CSIRO Energy Technology, P.O. Box 312, Clayton South, Victoria 3169, Australia § School of Applied Sciences, RMIT University, Melbourne, Victoria 3001, Australia ‡

S Supporting Information *

ABSTRACT: Chemical reaction studies of N-methyl-N-propyl-pyrrolidinium-bis(fluorosulfonyl)imide-based ionic liquid with the lithium metal surface were performed using ab initio molecular dynamics (aMD) simulations and X-ray Photoelectron Spectroscopy (XPS). The molecular dynamics simulations showed rapid and spontaneous decomposition of the ionic liquid anion, with subsequent formation of long-lived species such as lithium fluoride. The simulations also revealed the cation to retain its structure by generally moving away from the lithium surface. The XPS experiments showed evidence of decomposition of the anion, consistent with the aMD simulations and also of cation decomposition and it is envisaged that this is due to the longer time scale for the XPS experiment compared to the time scale of the aMD simulation. Overall experimental results confirm the majority of species suggested by the simulation. The rapid chemical decomposition of the ionic liquid was shown to form a solid electrolyte interphase composed of the breakdown products of the ionic liquid components in the absence of an applied voltage.

1. INTRODUCTION The demand on rechargeable battery technology to provide increased specific energy and specific power is being driven by the proliferation in usage of portable electronics and power tools, as well as the development of electric vehicles. Although the current generation of lithium ion technology outperforms alternative technologies, such as NiCd, NiMH, etc., the available power and energy is limited.1,2 Replacement of the carbon or intercalation anode materials with lithium metal can result in a 10-fold increase in specific electrode capacity3 or 20 to 30% increase in specific energy of a typical device.4 Although this has been known for a long time, devices using lithium metal anodes have not been realized due to the significant safety risks associated with lithium metal. Typically, the use of conventional organic solvent based electrolytes, as is done currently in lithium ion technology, results in deposition of lithium metal in a dendritic form during the charging reaction.5,6 The dendritic deposits can eventually form inter electrode connections resulting in short circuits,7 potentially with catastrophic results. In order to alleviate the problems of dendritic deposition of lithium metal, some researchers have advocated the use of ionic liquids (ILs) as a replacement of the organic solvent based electrolytes.8−12 The ability to change the physicochemical properties of ILs by control of choice of cation or anions13−17 is attractive for applications in battery technology. In addition, in light of renewed interest in the Li-Air and Li-Sulfur battery © 2012 American Chemical Society

technologies, an understanding of the interaction of IL electrolytes with Li is important.1,2 Within the vast array of ILs reported, there is a small (but steadily growing) group of compounds that can sustain the reversible deposition and stripping of lithium, on a useful range of substrates.18−23 Indeed, the use of certain classes of ILs as electrolytes, as for example N-methyl-N-propyl-pyrrolidiniumbis(fluorosulfonyl)imide, [C3mPy+][FSI−], has shown that lithium metal can be charged and discharged repetitively up to 800 times.11,12 The increased stability is principally due to the formation of a passivation film on the lithium metal surface termed the Solid Electrolyte Interphase or SEI layer.24 The SEI layer is composed of the decomposition products from the reaction of the negative Li electrode and the electrolyte, which combine into a mainly poly crystalline film on the anode surface.25 Although electrically insulating, the SEI layer is however ionically conducting. Under certain conditions it will allow ingress and egress of Li ions to the lithium metal surface. The SEI functions as a protective barrier partially screening the bulk electrolyte from direct contact with the strongly reducing lithium metal surface potential. The use of organic solvent based electrolytes produces very poor SEI layers leading to the Received: May 11, 2012 Revised: August 20, 2012 Published: August 22, 2012 19789

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

dendritic deposition of lithium metal.5,6 However, the use of ionic liquid based electrolytes has been shown to produce more stable SEI layers, which allow repetitive deposition and stripping.11,12 While the SEI layer formation mechanisms are not fully understood, the electrochemical stability of the IL electrolytes at high potentials is seen to be an important factor. To a large extent, the nature of the IL cation and anion determines the electrochemical stability of the electrolyte. The pyrrolidinium and piperidinium cations compared to imidazolium cations have been found to be appreciably more stable for battery applications.8−10 Ionic liquid electrolytes composed from these cations, and an appropriate anion, allow deposition of lithium metal.6−8 Under the extremely reducing conditions imposed by a metallic lithium electrode, the choice seems largely limited to fluorinated anions. Of the range of fluorinated anions, the bis(trifluoromethanesulfonyl)imide, TFSI−, and its simpler analogue, bis(fluorosulfonyl)imide, FSI−, forms ILs with good physical properties as well as providing the required electrochemical stability.11,12 The reversible electrochemistry of lithium metal in these classes of ILs has been characterized in previous studies.11,12,20−23 Although ILs based on the pyrrolidinium cation with TFSI− or FSI− anions appear to be suited for applications in lithium metal rechargeable batteries, little is known regarding the composition of the SEI layer with these IL electrolytes. Previous studies using XPS or other surface science techniques have been performed for the TFSI− based systems.8,26 However, a similar study has not yet been performed on the FSI− based system and recent results from our laboratories have suggested that FSI− based ionic liquids are better suited for lithium metal cycling than the corresponding TFSI− based systems.11,12 Prior to investigating the nature of the SEI created by electrochemical methods, it is important to be able to characterize the underlying surface. The chemical interaction with lithium metal has been postulated previously in our work, although data have not yet been ascertained regarding the specific nature of this interaction. Thus, we have adopted a combined computational and experimental approach to gain better understanding of the mechanism of the chemical interaction to form the SEI layer. To investigate the chemical reaction of [C3mPy+][FSI−] with the lithium metal surface, we have performed ab initio molecular dynamics (aMD) simulations using plane-wave density functional theory (DFT), which is implemented in the Vienna Ab-initio Simulation Package (VASP).27 This type of simulation method has previously been used to study the electronic structure properties of 1-ethyl-3-methyl-imidazolium tetrafluoroborate [emIm+][BF4−] IL on Li (001), (110), (111), and Au (111) surfaces,28,29 and [emIm+][FSI−] on Li (100).30 Here, we focus on the reaction dynamics of the IL/lithium system using aMD simulations. The evolving simulation is then analyzed in terms of fragment population analysis, pairdistribution function, and partial charge analysis. This allows a detailed picture of the initial stages of one of the possible chemical breakdown pathways of IL on the Li (001) surface. Additionally, we present an experimental analysis of the SEI layer using X-ray photoelectron spectroscopy, infrared spectroscopic and X-ray diffraction studies. Finally, we corroborate the two methods to allow an understanding of the reaction pathway for SEI formation between [C3mPy+][FSI−] and lithium metal.

2. COMPUTATIONAL AND EXPERIMENTAL SECTION 2.1. Electronic Structure Calculations. All of the calculations in this study were performed using the planewave density functional theory (DFT) method implemented in Vienna Ab initio Simulation Package (VASP).27 The projector augmented wave (PAW)31 pseudopotentials32 were used with the Perdew-Burke-Ernzenhof (PBE)33 exchange-correlation functional within the generalized gradient approximation (GGA). The bulk lithium metal for the starting configuration was a three-dimensional periodically bounded body-centered cubic (BCC) structure. The initial crystal lattice parameter was 3.51 Å. A full cell relaxation procedure was performed with a planewave cutoff of 550 eV and using a Gamma-centered Monkhorst-Pack34 grid with a reciprocal space k-point setting of 14 × 14 × 14. To obtain minimum energy structures, convergence criteria of 10−6 eV for the electronic energy and 10−4 eV Å−1 for the atomic forces were used. The lithium surface model was prepared by cleaving the bulk crystal in the (001) plane and exposing this to a vacuum spacer with the surface normal defined in the z-direction. Maintaining periodic boundary conditions in both the x- and y-directions models an atomically flat surface of infinite lateral extension. We performed a convergence test in order to determine the optimal amount of layers and vacuum spacing. These were determined to be 7 layers and 20 Å, respectively. In each run, the bottom two layers were constrained to the optimized bulk crystal layer spacing with the remainder allowed to fully relax, while maintaining constant cell volume. During the relaxation procedure, the k-point setting for the [1 × 1] cell was 14 × 14 × 1 (Gamma centered). A [4 × 4] supercell was used to accommodate the ionic liquids, with the corresponding k-point setting of 4 × 4 × 1 (Gamma centered). Microcanonical ensemble (NVE) aMD simulations were performed using VASP with a molecular dynamics time-step of 1 fs. 2.2. Development of the Relaxed Lithium Surface Model. Table 1 provides a comparison of the lattice parameter Table 1. Optimized Bulk Properties of the Bulk Lithium Crystal: Our Plane-Wave DFT Calculations versus Other DFT and Experimental Results property lattice parameter (Å) cohesive energy (eV/atom)

this study, PAW/PBE

PAW/ PW9128

equation of state (EOS) measurements36

3.434

3.45

3.48

1.61

1.62

1.66

and cohesive energy for the bulk lithium crystal calculated in this work, versus other DFT and experimental results.28,35 The crystal was then cleaved along the (001) Miller plane to create a lithium slab. To gauge the effective number of atomic layers needed for the surface model, successive supercell slabs comprising 5 to 14 Li layers with a 15 Å vacuum spacer were tested. Each system was geometry optimized while keeping the bottom two layers fixed, and the surface energy calculated using the expression,28 1 slab 1 σ = (Eopt − NEbulk ) − (E init − NEbulk ) (1) A 2A where A is the surface area, N is the total number of lithium atoms, Ebulk is the total energy per atom for the bulk bcc lattice, Einit is the total energy of the unrelaxed lithium slab, and Eslab opt is 19790

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

evacuating. SEM images were recorded using a XL30 Field emission SEM (Philips) with a Link ISIS (Oxford Instruments) Energy Dispersive X-ray system (EDX). In order to perform GADDS under an inert atmosphere, each sample was mounted onto a quartz substrate housed in an X-ray diffraction (XRD) sample cell (Bruker) with a 50 μm Kapton window. Sample assembly was performed inside an Ar-filled glovebox. Diffraction patterns were collected using an AXS D8: Discover (Bruker). For XPS measurements, disks were introduced to the Microlab VG310F XPS (Thermo) through use of an Ar atmosphere controlled glovebag (Aldrich). Transfer from the glovebox to glovebag took place by storing a disk within a Schott bottle, which in turn was contained in a heat sealed aluminum laminate pouch (3M Film).

the total energy of the slab with one-side surface relaxation. It was found that a slab of 7 atomic layers in thickness, in which the bottom two layers were constrained to the bulk lattice parameter, was sufficient to obtain a converged system within a practical simulation time while providing enough layers to effectively study surface reactions. In order to accommodate the 2 IL pairs on top of the Li surface, a [4 × 4] supercell was constructed, corresponding to a periodically bounded surface of area 13.7 Å × 13.7 Å and the length of the box was increased by 5 Å perpendicular to the surface in order to incorporate extra vacuum space between the periodic images of the supercell. To model chemical reactions occurring for the IL ions in contact with the lithium metal surface, the fully relaxed [4 × 4] Li (001) surface was decorated with two N-methyl-N-propylPyrrolidinium bis(fluorosulfonyl)imide [C3mPy+][FSI−] ionpairs, that were placed at least 1.8 Å away from the top Li surface and aMD simulations were performed to observe the time evolution of the ensuing reactions. The atomic charges reported here were obtained using the Atoms in Molecules (AIM) method of Bader.36 The AIM method defines atoms by a unique space-partitioning of the total molecular electronic density (that is, the square of the electronic wave function). The atomic subspaces (or basins) are calculated using rigorous topological analysis (maxima, minima, and saddle points). The interatomic boundary surfaces are defined as those for which the normal component of the density gradient is zero. By integrating the electronic density within a basin the partial charge for that atom can be estimated. 2.3. Instrumentation, Synthesis and Sample Preparation. Lithium metal (China Energy Lithium) of 0.33 mm thickness was cleaned using dried hexane (Merck). N-propyl-Nmethyl-pyrrolidinium bis(fluorosulfonyl)imide [C3mPy+][FSI−] (Dai-Ichi Kogyo Seigaku) was dried in vacuo using standard Schlenck line techniques at 100 °C for 12 h prior to storage and handling in a high purity argon glovebox (less than 5 ppm H2O and 1 ppm O2). Dimethylcarbonate (DMC) [Fluka] and hexane were treated with lithium metal in order to decrease both water content and lithium reactive species to negligible levels. Hexane cleansed lithium metal was punched into 10 mm disks and then treated with [C 3 mPy+ ][FSI−] through submersion into 5 mL of [C3mPy+][FSI−] for time periods of 4 h, 7, 12, and 18 days. Upon reaching the desired treatment time with the ionic liquid, disks were then removed and rinsed using DMC. Rinsing involved washing the disks with 10 mL aliquots of DMC twice, for a period of 1 min for each wash step. After washing, both sides of the disk were further rinsed with fresh aliquots of DMC dropwise. Residual DMC was removed via touching each disk to lint-free tissue and left to dry under vacuum for 30 min. All experimental handling and sample preparation was carried out inside the glovebox. 2.4. Surface Characterization. For IR spectroscopy, after contact treatment and washing, the dried disks were loaded and enclosed in an atmosphere controlled Attenuated Total Reflectance (ATR) stage (Specac Golden Gate) in the glovebox. After enclosure, the stage was removed from the glovebox and spectra were obtained using a spectrum 400 FTIR spectrometer (Perkin-Elmer). All spectra were recorded with a 4 cm−1 resolution and 16 scans. For SEM imaging, transfer of treated and washed disks from the glovebox to the SEM was carried out using an Ar environment controlled chamber. Lithium disks were placed into the chamber within the glovebox prior to attaching the chamber to the SEM and

3. RESULTS AND DISCUSSION 3.1. Simulation of the Li (100) Surface and [C3mPy+][FSI−] IL. The structure of the ions comprising the N-propyl-Nmethyl-pyrrolidinium bis(fluorosulfonyl)imide [C3mPy+][FSI−] ionic liquid is shown in Figure 1a. Also shown in Figure 1b is the fully relaxed Li metal (001) surface with infinite lateral extent. For the aMD simulations, the pair of [C3mPy+] cations and the [FSI−] anions were initially placed on the surface. Due to computational limitations, it was not possible to fully decorate the vacuum spacer with more IL pairs. Using aMD, the time evolution of the reaction of [C3mPy+][FSI−] on the Li surface was determined. Figure 2 shows a snapshot of the [C3mPy+][FSI−]/Li surface geometric configuration after a time period of 0, 72, 126, 478, 543, and 694 fs. Shown in Figure 2a is the initial “start” configuration at 0 fs. The first general observation is a tendency for the C3mPy+ cations to move upward away from the Li surface, while FSI− anions react strongly with the top few Li atomic layers. In the trajectory shown, the reaction evolved as follows: first, the formation of the LiF species (Figure 2b), followed by Li2F species (Figure 2c), then LiO species (Figure 2d), and finally Li2O species (Figure 2e). In the final frame taken at 694 fs, shown in Figure 2f, little recognition of the FSI− anion structure remains. The decomposition of the FSI− anion beginning with the cleavage of the S−F bond in closest proximity to the Li (001) surface occurred rapidly (within 50 fs) of the start of the simulation. The resulting fragments of the anion moiety were a negatively charged fluorine radical F·− and F(SO2)2N·. Bader charge analysis indicates that the F radical carries ∼30% of the net charge of the FSI− anion. Within 130 ps, the other FSI− anion had also lost one of its F atoms. The F·− radical bound rapidly with lithium atom of the outermost surface layer, resulting in a long-lived LiF molecule. At this point, the Bader charge of the LiF pair is shown to be slightly negative (about −0.36 e) in this environment (see Table 2), suggesting some solvation by electrons. We occasionally observed that another Li atom became bound to this LiF pair, thus forming a transient Li2F species with marginally positive Bader charge of +0.14 e (the expected formal charge is +1 e). This species is relatively stable but transient, forming at various stages to the end of the simulation. At ∼400 ps, one of the ejected surface Li atoms cleaved and combined with the remaining F− ion from one of the two F(SO2)2N· radicals. This triggered a complete dissociation of the radical, starting with the S−N bond, leaving an SO2 molecule with a Bader charge of −0.2 e and a SO2N· radical with a Bader charge of −2 e. One of the oxygen atoms in the 19791

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

trajectory (Figure 4). The results indicate that the average Li−F distance is 1.8 Å, while the average Li−O distance is slightly longer at 1.9 Å. We observed some secondary peaks in the pdf between Li and F, which indicate some coordination between the surface Li atoms and an FSI− F atom further from the surface. It is also associated with some of the LiF species that is drifting around the surface. This behavior is not apparent in the Li−O pdf. Figure 5 shows the time evolution of the molecular population of the breakdown species. In characterizing molecules in an evolving system, a chemical bond is defined to exist between two atoms if their separation is below some cut-off distance. In general, for a multicomponent system, there are numerous cutoff distances between the various combinations of elements. This distance has to be larger than the experimental bond length since there are significant thermally induced bond length fluctuations in the system. In practice, it was found by visual inspection that a single cutoff distance for all of the bonding types was sufficient to avoid producing extra unphysical bonds. This cutoff distance is obtained from the peak of the pdfs (see Figure 4). By counting the number of different elemental bonds to each atom, simple molecules and their abundance can be catalogued. Using this technique, we established that LiF is the first breakdown product formed, followed by Li2F, then LiO and Li2O which are observed at approximately the same time in the aMD simulation. The occupancy of the breakdown products in terms of percentage of simulation time frame (i.e., the amount of time the breakdown products were observed as a fraction of the total aMD simulation time) were are as follows: LiF(76.9%), Li2F(24.8%), LiO(15.3%) and Li2O(14.25%). It is immediately obvious that LiF is a long-lived species, appearing in nearly 77% of the simulation time frame. The Li2F species is relatively transient and only appears in approximately 25% of the simulation time frame, and may indicate its relative instability compared to the LiF species. Interestingly, the LiO and Li2O species appear in roughly the same fraction of simulation time frame. In Figure 5, it was shown that both of these species also appear at roughly the same time. This indicates that the bond between Li and O is not as strong as the bond between Li and F (which is also suggested by the pdf data shown in Figure 4), hence causing a lot of transition between these two structures. 3.2. Experimental Determination of SEI Chemical Components on the Lithium Surface. The chemical SEI formation on a pristine Li surface was experimentally investigated by immersing Li foil in [C3mPy+][FSI−] for various time periods. Figure 6 shows representative SEM images of an unmodified Li metal surface and surfaces which were exposed to [C3mPy+][FSI−] for periods of 4 h, 7, 12, and 18 days. The untreated Li surface (Figure 6a) has a rough morphology due to the rigorous cleaning procedure. After immersion in the IL for 4 h, it can be seen that the surface has become considerably smoother (Figure 6b) which suggests an immediate reaction between Li and the IL facilitated at defect sites on the rough surface. After 7 days, numerous pits have appeared over the surface, sized around 1−2 μm in diameter (Figure 6c). These pits are attributed to corrosion of the reactive Li surface by IL moieties. Significantly, after a 12-day reaction time, a drastic change in the morphology of the Li surface was observed to occur (Figure 6d), which shows large coral like structures over the entire Li surface. An alternate area SEM image (Figure 6e) clearly shows that this deposit grows outwardly from the surface and is quite porous in nature. Over

Figure 1. Structures of the molecules comprising the ionic liquid. (a) (left) N-propyl-N-methyl-Pyrrolidinium [C3mPy+]. (Right) bis(fluorosulfonyl)imide [FSI−] and (b) Model depicting the lithium metal-vacuum interface representing the Li(001) surface plane of infinite lateral extent with 2 pairs of IL on the top surface.

SO2N· radical then reacted with the surface Li atom, forming LiO. Just like the LiF species, the LiO species also has a slightly negative charge, allowing it to capture another Li atom, thus forming Li2O. At ∼650 ps, the remaining F(SO2)2N· radical had lost its F− ion, which combines with a Li atom to form LiF. We observed further breakdown of this radical and subsequent formation of LiO. However, the trajectory time was too short to establish further details of the decomposition. The reaction pathway of the FSI− anion breakdown as determined by the simulation is summarized in Figure 3. It is interesting to note that in our simulation, the C3mPy+ cations did not react with the surface, but instead migrated upward, retaining its molecular structure, we speculate that this may be due the thin (monolayer) coverage of the IL on the surface and that a thicker IL coverage would result in steric restrictions for the cations, which would cause the cations near the Li surface to react with the surface. In order to further quantify the breakdown products of the Li and IL interactions, we calculated the pair distribution functions (pdf) between Li and F, and between Li and O over the whole 19792

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

Figure 2. Snapshots of the simulation trajectory at (a) initial frame, (b) 72 fs, (c) 126 fs, (d) 478 fs, (e) 543 fs, and (f) final frame (694 fs). In (b)− (e), some of the atoms have been highlighted to show the formation of the LiF, Li2F, LiO, and Li2O species. The color scheme used is as follows: Li in green, C in cyan, N in blue, O in red, S in yellow, and F in purple.

Table 2. Average Bader Charges for the Reaction Productsa partial charge [e] species

72 fs

126 fs

478 fs

543 fs

average

LiF Li2F LiO Li2O

−0.34

−0.36 0.14

−0.58

−0.53

−0.45 0.14 −0.91 −0.42

−0.91 −0.42

a

Only species that are persistent for at least 10 fs with a bond distance of less than 1.9 A have been considered.

the 12 day period, the data show that corrosion of the lithium surface occurs by interaction with the IL and a subsequent roughening of the surface is observed. Remarkably however, allowing the reaction time to continue to 18 days shows a much smoother surface film with the observation of larger deposits, ca. 2 − 3 μm in diameter (Figure 6f). The leveling out of the film over the 12 to 18 day time period occurs over the entire surface as determined from low magnification images (shown in the Supporting Information). This suggests that the interaction of the Li surface with [C3mPy+][FSI−] is a highly dynamic process. This data corroborates well with the simulation where the calculations show disordering of the pristine (100) surface due to breakdown of the anion (Figure 2). Presumably if the disordering suggested by the simulations continues for significantly longer time scales then a substantial change in the surface morphology may be expected as observed in the SEM images. Illustrated in Figure 7a is the time dependent FT-IR spectra recorded of the Li surface. By comparison with the spectrum presented for the pure ionic liquid, it is readily seen that many of the bands arising from the IL-modified lithium surfaces are associated with ionic liquid retained by the SEI layer. This is consistent with the results of XPS studies by Howlett et al..8,26 After 4 h reaction time, the spectrum is mainly composed of

Figure 3. A reaction schematic of FSI’s breakdown process. The reaction proceeds from top to bottom. The F and O atoms combine with Li to form either LiF or LiO, which subsequently form Li2F or Li2O transiently. See text for details. 19793

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

Figure 4. Pair distribution functions of (a) Li−F and (b) Li−O. The details around the first peak (associated with bonding distance) are highlighted in the inset. The thin line shows the raw data, and the thick line shows a local regression fit to the data.

Figure 6. SEM (secondary electron) images of a Li surface after immersion in [C3mPy+][FSI−] for: (a) 0 h; (b) 4 h; (c) 7 days; (d) 12 days; (e) 12 days alternate area and (f) 18 days. The scale bar in each image is 10 μm.

Figure 5. Time evolution of the number of major breakdown products: LiF, Li2F, LiO, and Li2O.

bands due to the ionic liquid, at greatly reduced intensity, as well as a broad band at around 1500 cm−1 that is seen in the reference spectrum for untreated lithium (Figure 7a). Some bands, such as those associated with the ring mode vibrations of the pyrrolidinium cation (886, 905, 939, and 970 cm−1)26 are barely detectable when compared to the neat IL. In the samples taken subsequently, the spectrum changes only slightly, with small variations in band intensity. This suggests that the IL reacts immediately with the Li surface. However, this result can be explained using three reasons: (1) the simulation is only showing one possible breakdown pathway, not an ensemble as would be probed using much longer time scales probed using FT-IR; (2) the time scale of the simulation is not long enough to allow the cations to interact with the Li surface; and (3) the simulation only considers 2 IL pairs due to resource limitations (in a real system, there will be many IL pairs, providing steric restrictions which disallow the cation from drifting away from

Figure 7. (a) FT-IR spectra of [C3mPy+][FSI−], Li metal and a Li surface after immersion in [C3mPy+][FSI] for 4 h, 7, 12, and 18 days; (b) X-ray diffraction patterns of quartz substrate, Li metal and each Li surface after immersion in [C3mPy+][FSI−] for 4 h, 7, 12, and 18 days.

19794

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

the surface). Significantly, the band at 827 cm−1 for the S−F stretching mode37 observed in the neat IL was no longer present suggesting cleavage of the S−F bond of the FSI− anion. It can also be seen that there is a gradual increase in intensity in the band at 1286 cm−1 which can be assigned to the SO stretching mode37 and indicates decomposition of the FSI− anion.38 These data support the FSI− anion breakdown mechanism proposed by the simulation. The appearance of bands in the 18-day sample at 1728 and 1755 cm−1 can be attributed to stretching modes of CH2 bands. A prominent peak at 3677 cm−1 was observed in some samples and is attributed to the formation of LiOH. 7 Given that the FT-IR data appear to be influenced by liquid species that adhere to the Li surface, X-ray diffraction (XRD) analysis was used to obtain information on any solid (crystalline) products of the reaction between lithium and the ionic liquid. Diffraction patterns recorded at times up to 18 days (Figure 7b with relevant data summarized in Table 3) indicate that Li reacts with [C3mPy+][FSI−], but that the bulk of the products are largely amorphous.

studies.8,26 Crucially, the detection of LiF is consistent with the predictions of the aMD calculations. As shown before, the LiF phase is determined to be the most stable compound formed from the decomposition of the FSI− anion on the Li surface. Analysis of the high resolution XPS spectra recorded as a function of immersion time of Li in [C3mPy+][FSI−] allows the percentage surface abundance of breakdown products of the IL to be determined and is shown in Figure 8. Typical XPS spectra

Table 3. 2θ and Corresponding hkl Values for Lithium and Lithium Fluoride Crystal Planes, and Observed 2θ and hkl Values for Custom Quartz Substrate

Figure 8. Percentage abundance summary (%) determined from the region spectra for the Li surface as a function of immersion time in [C3mPy+][FSI−] using the N, F, O, Li, and C 1s spectra and S 2p spectrum.

quartz substrate (*)

pristine lithium (Li)

lithium fluoride (LiF)

2θ (deg)

hkl

2θ (deg)

hkl

2θ (deg)

hkl

39.599 55.027 68.512 80.131 87.665

[1,0,2] [2,0,2] [3,0,1] [2,1,3]

36.365 52.355 65.512 77.271 88.526

[1,1,0] [2,0,0] [2,1,1] [2,2,0] [3,1,0]

38.461 44.709 65.123 78.241 82.471

[1,1,1] [2,0,0] [2,2,0] [3,1,1] [2,2,2]

obtained for a sample of Li metal after reaction with [C3mPy+][FSI−] are shown in the Supporting Information. The XPS data obtained and proposed assignments are summarized in Table 4. As stated above, the time scale for determination of species (over 18 days) is in significant excess of that used for the simulation studies. Furthermore, only the stable species formed can be detected via these experiments. However, the experimental results show good correlation with the simulated results. The simulated reactions of the cation and anion are based on the pristine lithium surface where no defects or additional species are present. However, the experimental XPS spectra (Table 4) show that for a battery grade Li foil the spectra have peaks due to F, Li, and O. On the basis of literature reports on previous XPS studies of Li compounds, we attribute these to the formation of LiF and LiOH in the 0 h sample. Clearly, this is in contrast to the simulated lithium slab, which assumes none of these impurities are present. The LiOH present on the surface is most likely due to the DMC washing procedure to remove IL. The residual amount of LiF was also confirmed by both XRF analysis and detection by a fluoride ion selective electrode after dissolution of Li in an aqueous solution. We attribute this LiF to the manufacturing process of the battery grade Li foil. After a reaction period of 4 h, peaks which can be attributed to LiF,26 LiOH,26 and a SO2F39,40 species as well as species containing N+ and N− (assigned to surface confined [C3mPy+] and [FSI−]) are present on the surface. Similar results are observed over 7 to 18 day samples, albeit with varying peak intensities, suggesting that the quantity of these species changes during the course of the chemical reaction. Turning first to the LiF peaks, it was found that after 7 days of exposure to IL that there was an increase in the surface coverage of fluoride species by nearly 1.8 times. The F 1s peak position at 686 eV is unperturbed, suggesting growth of pure LiF on the surface.26 The significant increase cannot be due to the trace impurities observed in the native foil and is thus attributed to reaction between the IL and the Li metal.

Hence, after 4 h of contact, only the reflection at 2θ = 65° ((211) crystal plane) is present. The persistence of the (211) reflection is perhaps an indication of the preference of the reaction to proceed at specific crystalline planes. Another factor that appears to have a significant bearing on the relative strength of responses in Figure 7b is the intensity and incident angle of X-ray radiation. This is best illustrated in the 12-day pattern, where the diffraction intensity is greatest as is penetration through to the substrate (quartz). We note that the apparent tendency for preferred orientation in most of the data (varying intensity ratios for the reflections of each phase) is due to the combination of two factors: (i) the GADDS technique surveys a relatively small area of the sample, through the 50 μm window; (ii) the samples were not rotated during measurement. Beyond the 4 h point, the only new (crystalline) phase that is detected on the lithium surface is lithium fluoride (LiF), the peaks for which are seen in the 12-day sample. While the presence of this compound has often been inferred from compositional analyses (e.g., XPS), its direct detection in ILtreated lithium samples has not been reported to date. XRD data demonstrated that a large proportion of the Li sample after exposure to the IL consists of amorphous breakdown products as the intensity of reflections of Li at 2θ = 36.4, 52.4, 65.5 and 77.3° associated with the (110), (200), (211), and (220) planes respectively, decreased in intensity after contact with the IL. Significantly however, the emergence of peaks at 2θ = 44.7 and 78.2° confirmed the presence of LiF which can be assigned to the (200) and (311) planes. Overall, the apparent preponderance of amorphous material from the reaction of lithium and ionic liquid is in line with the findings from previous similar 19795

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

observed over the 4 h to 18 days samples to LiOH surface bound species. This is not as contradictory to the simulation as first appears. The formation of Li2O, according to the predicted simulation mechanism, occurs via formation of a transitory LiO radical. It is possible that the LiO radical formed abstracts a hydrogen atom from the cation, which is present at the surface to form the stable LiOH. This hypothesis is supported by the IR data which show a perturbation of the cation ring modes that could occur if an H atom is abstracted from the cation. However, this mechanism is currently beyond the scope of our simulation, and is an area for further research. Next turning to the S 2p spectra, the presence of a broad S 2p peak at 169 eV suggests the presence of oxidized sulfur species.8,26 Previous studies have shown that SO2F− species are observed at 169.5 eV in the S 2p XPS spectra.39,40 On the basis of these reports as well as the simulated reaction mechanism, which predicts an SO2F species as a degradation product, we can assign the S spectra to a surface bound SO2F species. Finally, the N 1s scan showed two peaks at 398 and 400 eV which have been assigned previously to the presence of N− and N+ containing species, respectively.26 We attribute these peaks to the surface confined [C3mPy+] cation and the FSI− anion.

Table 4. Summary of XPS Data and Assignments for Battery Grade Li Metal and Li Metal after 4 h, 7 Days, 12 Days and 18 Days Contact Time with [C3mPy+][FSI−] element

peak position/eV

assignment

pristine lithium foil (hexane polished) 54.73 LiF or LiOH 684.88 LiF 531.65 LiOH lithium foil after 4 h contact with [C3mPy+][FSI−] Li (1s) 54.28 LiF or LiOH F (1s) 683.70 LiF and SO2F 686.96 O (1s) 531.89 LiOH N (1s) 398.6 N− 401.51 N+ S (2p) 169.06 SO2F lithium foil after 7 days contact with [C3mPy+][FSI−] Li (1s) 54.31 LiF or LiOH F (1s) 683.83 LiF and SO2F 687.04 O (1s) 531.7 LiOH N (1s) 398.58 N− 401.56 N+ S (2p) 168.97 SO2F lithium foil after 12 days contact with [C3mPy+][FSI−] Li (1s) 54.38 LiF or LiOH F (1s) 683.74 LiF and SO2F 686.97 O (1s) 531.89 LiOH N (1s) 398.63 N− 401.6 N+ S (2p) 169.19 SO2F lithium foil after 18 days contact with [C3mPy+][FSI−] Li (1s) 54.1 LiF or LiOH F (1s) 684.2 LiF and SO2F 687.2 O (1s) 530.86 LiOH N (1s) 398.87 N− 401.61 N+ S (2p) 166.46 SO2F

Li (1s) F (1s) O (1s)

4. CONCLUSIONS A combination of plane-wave density functional theory calculations and XPS experimental data was employed to yield insight into the possible breakdown products arising from spontaneous reactions of the [C3mPy+][FSI−] IL in contact with planar lithium metal (100) surface. From these results, we are now able for the first time to not only identify the species present in an SEI formed between an IL and a Li surface, but also follow their formation process. Our data suggest that the following steps occur in the breakdown of the IL; The S−N bond in [FSI−] is broken, due to strong attraction of oxygen atoms to the lithium surface. The unstable SO2F· radical rapidly loses its F ion to the lithium surface, resulting in the formation of a long-lived LiF species and a stable SO2 molecule which promptly drifts away from the surface; at a later time, the remaining SO2N·− radical loses its oxygen atom to the lithium surface forming LiO and eventually Li2O. XPS data also suggest that the cation also breaks down, contributing an H atom to the LiO species to form a long-lived LiOH species.



Interestingly, one of the main species suggested by the reaction simulation is formation of LiF with very high occupancy rates of 76.9%. Thus the XPS data are in agreement with the simulation and it can be seen that this stable species appears to be present over the 18 days of reaction employed for the experiment. The peak position for the Li 1s peak for pristine Li occurred at 55 eV, which is consistent with LiF as reported previously,26 however it was found after exposure to the IL that the peak shifted to a lower binding energy of 54 eV (Table 4) which suggests the additional presence of either LiOH or Li2O on the surface.26 However, in contrast to the simulated mechanism, we find no evidence for Li2O in the XRD patterns. It should be stated that the occupancy rate for Li2O is only 14.2%. Furthermore, in the XPS O (1s) spectra over the entirety of the reaction period, we only observe a peak at 531 eV. Previous literature reports place the O peak for Li2O at 528 eV;26 however, our XPS spectra show no evidence for any peak at this position. However, the literature reports also place an O peak for LiOH at 531 eV.26 Furthermore, evidence for LiOH was also observed using IR spectroscopy for reaction of Li with [C3mPy+][FSI−]. Thus, we attribute the O peak at 531 eV

ASSOCIATED CONTENT

S Supporting Information *

Scanning Electron Microscope (SEM) images and additional XPS spectra of Li surface after immersion in [C3mPy+][FSI−] ionic liquid. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors acknowledge the Australian Government Department of Innovation, Industry, Science and Research (DIISR) for an International Science Linkages (ISL) Competitive Grant (CG110204) which provided partial funding for use in conducting this work; the Australian National Computational 19796

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797

The Journal of Physical Chemistry C

Article

(29) Valencia, H.; Kohyama, M.; Tanaka, S.; Matsumoto, H. J. Chem. Phys. 2009, 131, 244705. (30) Valencia, H.; Kohyama, M.; Tanaka, S.; Matsumoto, H. J. Phys. Chem. C 2012, 116, 8493−8509. (31) Blöchl, P. E. Phys. Rev. B 1994, 50, 17953−17979. (32) Kresse, G.; Joubert, D. Phys. Rev. B 1999-I, 59, 1758−1775. (33) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868. (34) Monkhorst, H. J.; Pack, J. D. Phys. Rev. B 1976, 13, 5188−5192. (35) Anderson, M. S.; Swenson, C. A. Phys. Rev. B 1985, 31, 668− 680. (36) Bader, R. F. W. Atoms in MoleculesA Quantum Theory; Oxford University Press: Oxford, 1990. (37) Lustig, M.; Bumgardner, C. L.; Johnson, F. A.; Ruff, J. K. Inorg. Chem. 1964, 3, 1165. (38) Balducci, A.; Schmuck, M.; Kern, W.; Rupp, B.; Passerini, S.; Winter, M. ECS Trans. 2008, 11, 109. (39) Trudeau, M. L.; Rochefort, A.; Fréchette, M. F. IEEE Annual ReportConference on Electrical Insulation and Dielectric Phenomena, San Francisco, October 20−23, 1996. (40) Herlem, M.; Mathieu, C.; Herlem, D. Fuel 2004, 83, 1665.

Infrastructure (NCI) facility for their continued support and providing access to high-performance computing resources. A.B. acknowledges CSIRO for the provision of an Office of the Chief Executive (OCE) PhD scholarship. The authors thank Prof. Johan Du Plessis of RMIT's Microscopy and Microanalysis Facility for assistance with the XPS measurements.



REFERENCES

(1) Bruce, P. G.; Hardwick, L. J.; Abraham, K. M. MRS Bull. 2011, 36, 506−512. (2) Ji, X.; Nazar, L. F. J. Mater. Chem. 2010, 20, 9821−9826. (3) Based on comparison of the theoretical specific capacity of the lithium metal (3860 mAhg−1) and the LiC6 graphite (372 mAhg−1) negative electrodes, respectively. (4) Based on the material composition by weight percentage in the commercial standard 18650-type Li-ion cells, see e.g., Johnson, B. A.; White, R. E. J. Power Sources 1998, 70, 48−54. (5) Lopez, C. M.; Vaughey, J. T.; Dees, D. W. J. Electrochem. Soc. 2009, 156 (9), A726−A729. (6) Crowther, O.; West, A. C. J. Electrochem. Soc. 2008, 155 (11), A806−A811. (7) Lane, G. H.; Bayley, P. M.; Clare, B. R.; Best, A. S.; MacFarlane, D. R.; Forsyth, M.; Hollenkamp, A. F. J. Phys. Chem. C 2010, 114, 21775. (8) Howlett, P. C.; MacFarlane, D. R.; Hollenkamp, A. F. Electrochem. Solid State Lett. 2004, 7, A97−A101. (9) Sakaebe, H.; Matsumoto, H. Electrochem. Commun. 2003, 5, 594− 598. (10) Zhou, Q.; Henderson, W. A.; Appetecchi, G. B.; Montanino, M.; Passerini, S. J. Phys. Chem. B 2008, 112, 13577−13850. (11) Bhatt, A. I.; Best, A. S.; Huang, J.; Hollenkamp, A. F. J. Electrochem. Soc. 2010, 157, A66−A74. (12) Best, A. S.; Bhatt, A. I.; Hollenkamp, A. F. J. Electrochem. Soc. 2010, 157, A903−A911. (13) Ohno, H., Ed. Electrochemical Aspects of Ionic Liquids; J. Wiley: Hoboken, NJ, 2005, 392 pp. (14) Welton, T. Chem. Rev. 1999, 99, 2071−2083. (15) Silvester, D. S.; Compton, R. G. Z. Phys. Chem. 2006, 220, 1247−1274. (16) Hapiot, P.; Lagrost, C. Chem. Rev. 2008, 108, 2238−2264. (17) Barrosse-Antle, L. M.; Bond, A. M.; Compton, R. G.; Mahony, A. M. O.; Rogers, E. I.; Silvester, D. S. Chem. Asian J. 2010, 5, 202− 230. (18) Gorodetsky, B.; Ramnial, T.; Branda, N. R.; Clyburne, J. A. C. Chem Commun. 2004, 1972−1973. (19) Sakaebe, H.; Matsumoto, H.; Tatsumi, K. Electrochim. Acta 2007, 53, 1048−1054. (20) Ishikawa, M.; Sugimoto, T.; Kikuta, M.; Ishiko, E.; Kono, M. J. Power Sources 2006, 162, 658−662. (21) Sugimoto, T.; Atsumi, Y.; Kikuta, M.; Ishiko, E.; Kono, M.; Ishikawa, M. J. Power Sources 2009, 189, 802−813. (22) Saint, J.; Best, A. S.; Hollenkamp, A. F.; Kerr, J.; Shin, J.-H.; Doeff, M. M. J. Electrochem. Soc. 2008, 155, A172−A180. (23) Appetecchi, G. B.; Montanino, M.; Balducci, A.; Lux, S. F.; Winter, M.; Passerini, S. J. Power Sources 2009, 192, 599−605. (24) Peled, E. J. Electrochem. Soc.: Electrochem. Sci. Tech. 1979, 2047− 2051. (25) Balbuena, P., Wang, Y., Eds. Lithium-Ion Batteries: SolidElectrolyte Interphase; Imperial College Press: London, 2004; p 420. (26) Howlett, P. C.; Brack, N.; Hollenkamp, A. F.; Forsyth, M.; MacFarlane, D. R. J. Electrochem. Soc. 2006, 153, A595. (27) Kresse, G.; Hafner, J. VASP The Guide; University of Vienna: Vienna, 2007, [http://cms.mpi.univie.ac.at/vasp/]; Kresse, G.; Furthmuller, J. Comput. Mater. Sci. 1996, 6, 15. Kresse, G.; Hafner, J. Phys. Rev. B 1993, 47, RC558. Kresse, G.; Furthmuller, J. Phys. Rev. B 1996, 54, 11169. (28) Valencia, H.; Kohyama, M.; Tanaka, S.; Matsumoto, H. Phys. Rev. B 2008, 78, 205402. 19797

dx.doi.org/10.1021/jp304581g | J. Phys. Chem. C 2012, 116, 19789−19797