Fluoroethylene Carbonate as a Directing Agent in Amorphous Silicon

Dec 18, 2017 - The initial configuration of the electrolyte phase is generated using the Amorphous Cell Builder module as implemented within the Mater...
2 downloads 13 Views 5MB Size
Download PDF
Letter Cite This: Nano Lett. XXXX, XXX, XXX−XXX

pubs.acs.org/NanoLett

Fluoroethylene Carbonate as a Directing Agent in Amorphous Silicon Anodes: Electrolyte Interface Structure Probed by Sum Frequency Vibrational Spectroscopy and Ab Initio Molecular Dynamics Yonatan Horowitz,†,§ Hui-Ling Han,†,§ Fernando A. Soto,‡ Walter T. Ralston,†,§ Perla B. Balbuena,*,‡ and Gabor A. Somorjai*,†,§ †

Department of Chemistry, University of California, Berkeley, California 94720, United States Department of Chemical Engineering, Texas A&M University, College Station, Texas 77843-3122, United States § Materials Sciences Division, Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, California 94720, United States ‡

S Supporting Information *

ABSTRACT: Fluorinated compounds are added to carbonate-based electrolyte solutions in an effort to create a stable solid electrolyte interphase (SEI). The SEI mitigates detrimental electrolyte redox reactions taking place on the anode’s surface upon applying a potential in order to charge (discharge) the lithium (Li) ion battery. The need for a stable SEI is dire when the anode material is silicon as silicon cracks due to its expansion and contraction upon lithiation and delithiation (charge−discharge) cycles, consequently limiting the cyclability of a silicon-based battery. Here we show the molecular structures for ethylene carbonate (EC): fluoroethylene carbonate (FEC) solutions on silicon surfaces by sum frequency generation (SFG) vibrational spectroscopy, which yields vibrational spectra of molecules at interfaces and by ab initio molecular dynamics (AIMD) simulations at open circuit potential. Our AIMD simulations and SFG spectra indicate that both EC and FEC adsorb to the amorphous silicon (a-Si) through their carbonyl group (CO) oxygen atom with no further desorption. We show that FEC additives induce the reorientation of EC molecules to create an ordered, up-right orientation of the electrolytes on the Si surface. We suggest that this might be helpful for Li diffusion under applied potential. Furthermore, FEC becomes the dominant species at the a-Si surface as the FEC concentration increases above 20 wt %. Our finding at open circuit potential can now initiate additive design to not only act as a sacrificial compound but also to produce a better suited SEI for the use of silicon anodes in the Li-ion vehicular industry. KEYWORDS: Lithium ion batteries, additives, nonlinear spectroscopy, simulations of fluoroethylene carbonate (FEC) remarkably improves discharge capacity retention and Coulombic efficiency regardless of the silicon anode form.10−16 However, the property that makes FEC a desirable additive to electrolytes is still puzzling as aside from its reduction to LiF,17 a Li-ion conducting salt, it was also suggested to form fluorinated polymer-like compounds,18 none at all,19 and that fluorine affects the structure of the electrode−electrolyte interface at open circuit potentials.20 In this study, we have assimilated two complementary techniques; sum frequency generation (SFG) vibrational spectroscopy21 and ab initio molecular dynamics (AIMD) simulations22 that reveal the dynamics of an ethylene carbonate (EC) and fluoroethylene carbonate (FEC) mixture solution in contact with amorphous silicon (a-Si, having a native oxide termination)

I

n a battery, electrical energy is generated by conversion of chemical energy via redox reactions at the anode and cathode.1,2 During the charge/discharge cycles, in the presence of Li ions the liquid electrolyte solution (mainly carbonates)3,4 is being consumed on the electrodes surfaces. If continued uninterrupted this will lead to capacity loss and finally render the Li-ion battery useless (cause lifetime shortening).5 The key factor in long-term use (cyclability and stability) of Li-ion batteries is the formation of an insoluble electrically insulating layer that allows lithium ion transport at a reasonable rate while hindering electrolyte consumption on the anode surface, and is termed the solid electrolyte interphase (SEI).6,7 Silicon (Si) was regarded as the next generation anode material for its high theoretical capacity. However, due to the silicon volume expansion as a result of lithiation and delithiation the carbonate-based SEI fails to maintain the Si integrity. In an attempt to increase the capacitance of the Si anode8,9 various compositions of electrolytes3,4 (solvent and lithium salt) were tested. The addition © XXXX American Chemical Society

Received: November 5, 2017 Revised: December 10, 2017 Published: December 18, 2017 A

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

measurements. We chose CaF2 as the substrate to reduce IR absorption in the far-IR (1600−2100 cm−1). We have deposited the copper current collectors and amorphous silicon anode sequentially via an e-beam deposition machine. The electrochemical cell design,20 presented in Scheme 1, allows the far-IR

anode surface under open circuit potential (OCP) conditions. In this study, we probe pure ethylene carbonate (EC), pure fluoroethylene carbonate (FEC) solutions, and their mixtures on amorphous silicon (a-Si) thin films. We use lithium perchlorate (LiClO4) as salt so to reduce the fluorine sources solely to FEC decomposition. Here we show by SFG vibrational spectroscopy,21 which is a nonlinear optical technique that yields vibrational spectra of molecules at interfaces that upon adding FEC to an EC solution (e.g., EC/FEC mixture) the EC molecules reorient on the silicon anode surface. This reorientation is featured in the SFG spectra as frequency shifts and intensity variations of EC and FEC vibrational peaks. We further explore the intermolecular interactions of the EC/FEC adsorption mechanism by applying AIMD simulations. Our simulations point out that once adsorbed either EC or FEC stick to the a-Si surface through their carbonyl end. On the basis of our SFG spectra and AIMD simulations, we show that as FEC content increases (1) FEC becomes the dominant adsorbate on the a-Si and (2) the EC molecules assume a more perpendicular adsorption angle. Our finding can now initiate additive design to not only act as a sacrificial compound18 but also as a functional reagent influencing the packing electrolyte candidates at the silicon anode/electrolyte to form a SEI that will enable the use of silicon anodes in the Li-ion vehicular industry. Our article is arranged as follows: after a brief description of the experimental and computational methods we examine our results for single component electrolyte solutions. This section is followed by presenting the effect of increased FEC concentration on the EC−amorphous silicon anode interactions. Finally, we address the orientation of EC molecules as a function of FEC concentration. Sum Frequency Generation (SFG) Vibrational Spectroscopy. SFG vibrational spectroscopy has characteristics of nondestructive spectroscopy which allows us to study underworking conditions, the different molecular-surface adsorption structures of electrolyte solutions.23−28 The picosecond laser system consisted of a 1064 nm Nd:YAG pump laser (PL2250, Ekspla) with a repetition rate of 20 Hz and an average pulse power of 25 mJ. A LaserVision optical parametric generator and amplifier system converted the 1064 nm to a visible 532 nm beam and a far-IR beam ranging between 1600 to 2000 cm−1. SFG is achieved when a visible and an infrared beam overlap spatially and temporally on a medium.29 The beam’s orientation in all the SFG experiments was as follows: the angle of the 532 nm beam was 56° and the far-IR beam was 45°, in respect to the surface normal of the sample. We collected the SFG beam by a Hamamatsu photomultiplier tube (Model R922); multiple irises are set along the SFG beam path and several band-pass filters are added to minimize the 532 nm light. We used an SSP (SF, Vis, IR) polarization combination to probe adsorbates with perpendicular (to the surface) dipole moments. For parallel dipole moments, we used an SPS (SF, Vis, IR) polarization combination. The average power of 532 nm and far-IR at 1800 cm−1 on the sample did not exceed 120 μJ, well below the amorphous silicon thin film’s damage threshold. Each SFG spectra is the average of 10 scans, in which each sampling (at 2.5 cm−1 wavenumber steps) is the average of 200 shots. Optical Electrochemical Half-Cell. The optical electrochemical half-cell (ECC-OPTO-STD, EL-CELL), henceforth referred to as eCell, serves to monitor the optical properties of an electrode material in the course of electrochemical

Scheme 1. General Illustration of the Electrochemical HalfCella

a

(a) General illustration of the electrochemical half-cell (eCell) for SFG measurement. (b) The layered amorphous Si anode and its copper current collectors deposited sequentially on a CaF2 window. By depositing the copper rings and amorphous Si directly on the CaF2 window, we enable the visible and IR beams to propagate through the window and the a-Si anode reaching the a-Si/electrolyte interface. The SFG signal then propagates upward (back reflection).

and visible beam to propagate through the 200 nm thick a-Si film and overlap on the a-Si/electrolyte solution interface in time and space. The generated sum frequency beam (∼480 nm) then reflects through the a-Si toward a photomultiplier tube (PMT). The various electrolyte components were bought from Sigma-Aldrich and used without further purification. The mixtures were prepared from two single solvent electrolyte solution of 1 M lithium perchlorate salt (LiClO4) in ethylene carbonate (EC) and 1 M LiClO4 in fluoroethylene carbonate (FEC) in the desired weight ratio. The mixtures and the eCell assembly (disassembly) were carried out inside a LabStar glovebox (MBraun) under argon atmosphere (water and oxygen levels were both below 0.1 ppm). Ab Initio Molecular Dynamics. Amorphous Silicon. The amorphous silicon structure was generated following an annealing-cooling protocol using the reactive molecular dynamics formalism as previously proposed by Ewing and co-workers30,31 as such we applied the LAMMPS32 protocol for the molecular dynamics (MD) simulations and a NVT (constant number of atoms, volume, and temperature) ensemble with a time step of 0.25 fs. For Si−Si interactions, we employed the ReaxFF potential reported for silicon carbide,33 which allows bond breaking and bond forming events to take place. For the initial structure, we used a bulk silicon crystal with 64 Si atoms, and we modified the lattice parameters such that the final structure reproduced the density reported for a-Si (2.285 g/cm3).34 This structure was then annealed to 4000 K during 150 ps, followed by cooling at a rate of 5 K/ps until it reached 300 K. Finally, an energy minimization was performed using the conjugate gradient algorithm. The a-Si structure was then cleaved to create an a-Si slab and passivated by O atoms to create a SiO0.09 surface slab. Amorphous Silicon-Electrolyte Interaction. The interaction of the electrolyte mixture with the a-Si surface was studied using ab initio molecular dynamics (AIMD) simulations. B

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters The electrolyte mixtures in the system consisted of LiClO4 as the salt dissolved in pure FEC as well as in EC/FEC mixtures with the following weight ratio: 1:1, 6:4 and 8:2. The initial configuration of the electrolyte phase is generated using the Amorphous Cell Builder module as implemented within the Materials Studio application.35 This module uses a configurational bias Monte Carlo (MC) algorithm to add solvent molecules on top of the a-Si surface one by one taking into account interactions with all atoms already positioned inside the three-dimensional (3D) box. Each candidate configuration is subjected to a number of tests such as ring spearing, catenation, and close contacts checks. If the configuration passes the initial checks, the next candidate configuration is chosen until the entire 3D box is packed with the solvent molecules. The result of this sampling method is that low energy configurations are preferred over high energy configurations. The electron−electron and exchange-correlation (XC) interactions were considered by using the generalized-gradient approximation (GGA)36 of the density functional theory (DFT) using the Vienna ab initio simulation package (VASP).37−40 We further applied the projector augmented wave (PAW) pseudopotentials41,42 for pseudopotential treatment of the atoms involved in the simulation. The kinetic cutoff energy for the plane-wave basis set was set to 400 eV, and the Γ point was used for the Brillouin Zone (BZ) sampling. The electronic and ionic degrees of freedom were relaxed using the conjugategradient (CG) method with convergence criteria set to 10−4 and 10−3 for the electronic and ionic convergence, respectively. The NVT ensemble at room temperature is controlled by the Nose−Hoover thermostat.43 The mass of H atoms in the system was replaced with tritium to increase the time step to 1 fs and to increase the time interval in AIMD. The AIMD scheme includes van der Waals force as proposed by BeckeJonson.44 All AIMD simulations were run for 50 ps to obtain accurate statistics of the solvent-surface interactions and solvent orientations. Single Electrolyte Solution of 1 M LiClO4 in Ethylene Carbonate (EC) OR in Fluoroethylene Carbonate (FEC). In this section, we address the stability at open circuit potential (OCP) conditions of either EC or FEC molecules in contact with a-Si by in situ experiments. We examine the single electrolyte solution OCP stability by SFG and AIMD simulations. At OCP, we do not expect that reduction and decomposition of solvent molecules into chemically active radicals or SEI components will take place. We define the single component solution 1.0 M LiClO4 in pure EC, as “pure EC” and 1.0 M LiClO4 in pure FEC, as “pure FEC”. Our AIMD simulations support our experiments and show that EC and FEC molecules adsorb without decomposing except for a rare case in pure FEC solutions which is addressed further along in the text. In Figure 1, we show the resulting SFG spectra at OCP in the carbonyl stretch (CO) frequency region of pure EC and pure FEC. In Figure 1, the pure EC solution (labeled EC) shows two peaks (labeled as a and b and noted in the text as COa and COb, respectively). The EC fundamental CO vibration is sensitive to intermolecular interactions. These can either induce tight packing as in a solid phase or a loose one for example after solvation of Li salts. In solid EC, there are two strong peaks at 1791 and 1829 cm−1 of equal intensity, due to a Fermi resonance between CO stretching and the overtone of the EC ring breathing mode at 895 cm−1.45 The addition of a Li salt causes the ring breathing overtone to redshift and to reduce as a

Figure 1. We present the SFG spectra of pure EC, FEC, and their mixtures at open circuit potential probed with two distinct polarizations: at SSP (left), sensitive to perpendicular adsorbates in reference to the surface normal, and SPS (right), sensitive to parallel adsorbates in reference to the surface normal. For clarification, we assign a and b to EC carbonyl group vibrations. For the FEC carbonyl group, we note c and d.

result of the primary solvation of the lithium salt by EC.46 In this study, the ring breathing overtone is at 1820 cm−1, therefore we suggest that 1820 cm−1 vibration can be assigned to a “loose” liquid like phase. Hence, we attribute the COa peak at 1780 cm−1 to strong intermolecular interactions with tight packing, resembling solidlike EC intermolecular interactions.45 We assign the second peak COb at 1820 cm−1 to liquid-like (loose) EC intermolecular interactions. In Figure 1, the pure FEC spectrum (labeled FEC) shows two vibrations (labeled as c at 1840 cm−1 and d at 1855 cm−1, and referred in text as COc and COd respectively) that we associate with the carbonyl stretch modes of the FEC molecule. Our first-principles calculations predict that EC and FEC molecules adsorb at the a-Si surface mainly through their carbonyl group (see Figure 2). Our data also suggest that, overall, both the EC and FEC adsorbed species retain their molecular characteristics (i.e., bond angles and bond distances). As we have mentioned above, in a rare case, in the pure FEC model, the FEC molecule loses its fluorine atom and its oxygen atom (from the ethylene group) causing its closed-ring structure to undergo a ring-opening reaction. As previously reported in the literature, the addition of FEC to the electrolyte may lead to an increase in LiF and polycarbonates.47 In the pure FEC model, three FEC molecules are firmly adsorbed at the a-Si surface while one FEC molecule gets closer to the surface (Figure 2a). The time evolution of the SiO bond length is shown in Figure S7. The SiOcarbonyl average bond length is ≈2.00 Å during the simulation while the CO average bond distance (labeled C1O2 in the Supporting Information document) of the FEC molecule that undergoes a ring opening is 1.34 Å. The short bond length suggests a strong interaction between the Si and O atoms. The data shows that the adsorption takes place within the first 5 ps of simulation time. We have also addressed whether the silicon surface oxygen atoms would have a repulsive effect on the solvent molecules. To test this, we ran an AIMD simulation for the pure FEC C

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters

Figure 2. Snapshots at approximately 50 ps of simulation time for the models with (a) 20 wt % FEC, (b) 40 wt % FEC, (c) 50 wt % FEC, and (d) pure FEC. The color code representing the various atoms is as follows: yellow, Si; red, O; gray, C; white, H, and dark blue spheres represent F atoms. For clarity, some solvent molecules were removed from the snapshots. Solvent molecules and salt not interacting with the a-Si surface are shown in a line display style.

atomic force microscopy (AFM) carried out on a-Si samples that were soaked a 1 M LiClO4/EC/FEC 7:3 (w/w) solution for a week. No reduction or decomposition products were observed and the data is presented in the Supporting Information Figures S4−S6. In Figure 1, we present the SFG vibrational spectra of three mixtures, in wt/wt ratio of EC/FEC: 8:2 (20 wt % FEC), 7:3 (30 wt % FEC), and 1:1 (50 wt % FEC) electrolyte solutions. The addition of FEC to EC yields new vibrational peaks that we associate with FEC (Figure 1 noted as c, d). For EC (COa), we observed a slight blue-shift in frequency and an amplitude increase due to reorientation by FEC since no redox electrochemical processes take place at OCP. The physical properties (peak frequency, amplitude, and peak width) that we have derived from fitting our SFG spectra are summarized in Table S1. In Figure 1 (left) for pure-EC, the COa vibration that corresponds to its vertical dipole component (at SSP polarization) is located at 1778 cm−1. Upon adding 20 wt % FEC the COa frequency redshifts to 1773 cm−1; increasing amounts of FEC (30 wt %) continue to shift the frequency up to a maximum of 1792 cm−1 at 50 wt % FEC content. The amplitude of the COa component at SSP polarization drops from 1.8 (pure EC) to 0.8 when adding 20 wt % FEC. However, as the FEC content increases so does the amplitude reaching 2.0 at 50 wt % FEC content. We attribute the COa vibration to the packing degree of EC molecules (i.e., intermolecular EC−EC interactions).45

(1 M LiClO4) model in contact with a SiO2 (001) surface slab (shown in Figure S8). After 2.8 ps of simulation time, one FEC molecule is adsorbed at the SiO2 (001) surface slab via its O atom of the CO group. The Si−O bond length (2.07 Å) is comparable to the Si−O bond distances found in the amorphous Si models. Therefore, we conclude that the amount of O content at the surface of the amorphous Si slab will not change the adsorption mode of the solvent molecules. The short bond length and fast adsorption time combined with a narrow range adsorption angle (refer to The Effect of FEC on Ethylene Carbonate Adsorption Angle) support our claim that once an FEC molecule is adsorbed it does not desorb with adsorption energies in the presence of lithium ions, that is, −29.7 and −26.2 kcal/mol for EC and FEC, respectively.48 Our AIMD simulations show that for the EC molecules the described above interactions of the FEC and Si atoms is similar to EC in all the other three models (mixtures) investigated. Ethylene Carbonate and Fluoroethylene Carbonate Electrolyte Mixtures. In this section, we examine how the addition of FEC affects EC interactions with neighboring EC molecules (EC packing). We analyze the carbonyl stretch associated with a crystalline-like packing of EC (COa) in SFG spectra of the three EC/FEC mixtures along with their dynamics obtained from our AIMD simulations. As with the pure EC and pure FEC solutions, we do not expect the EC/FEC mixtures to decompose at OCP. Our hypothesis was verified by ex situ X-ray photoelectron spectroscopy (XPS) measurements, scanning electron microscopy (SEM), and D

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters Therefore, in accordance with our SFG fitting we speculate that FEC induces tighter packing on the adsorbed EC layer as the EC COa frequency shifts to higher values (becomes more energetic) and the corresponding COa amplitude increases. We further assume that this tight packing takes place in the plane normal to the a-Si surface. If this reorientation mainly occurs when the EC molecules are upright then the parallel component (SPS) amplitude should decrease while its frequency should not shift much. Indeed, our COa (SPS) parallel component amplitude decreases as more FEC is added to the EC solution while its frequency remains constant (relative to our SFG resolution). The result is shown on the right side of Figure 1 (SPS). It is worth noting that the interpretation of the SFG amplitude is not straightforward. The SFG amplitude is governed in large part by the number of the adsorbed molecules (surface coverage) and their bond angles in respect to the surface normal. Even more, when interpreting the contribution of the perpendicular and parallel components to the SFG signal, a previous SFG study has shown that in a EC/diethyl carbonate (DEC) or EC/dimethyl carbonate (DMC) electrolyte mixtures the surface composition is indifferent to its bulk environment and that EC when compared to the linear esters (diethyl carbonate and dimethyl carbonate) is the dominating molecule adsorbed on the electrode surface.25 We know that both EC and FEC have similar adsorption energies to the amorphous silicon, as explicit density functional theory (DFT) calculations have shown that the FEC surface adsorption energy is only slightly lower compared to that of EC on Si clusters and that the fluorine orientation in respect to the silicon surface has little effect on the adsorption energy.48 Therefore, we turn to our AIMD simulations in order to reveal the dynamics of the electrolyte at the interface. In the 20 wt % FEC model, after 50 ps of simulation time, five EC molecules are attached to the a-Si surface through their carbonyl group (Figure 2a). In models with higher content of FEC molecules, the surface coverage of the FEC molecules increases while the surface coverage of the EC molecules decreases (Figure 2b). At 50 wt % content, FEC becomes the dominant adsorbate leading to a 2:1 FEC/EC surface ratio (Figure 2c). Hence, by our simulations we can suggest that the realignment of EC by FEC to a more up-right position causes an amplitude increase of the COa vibration in the SSP polarization. The frequency shift and amplitude increase of the COa vibration (crystalline-like EC intermolecular interactions) around 1780 cm−1 as a function of the EC/FEC ratio matches reported literature values for ratios of EC/FEC which produce beneficial SEI. The EC/FEC ratio is only beneficial until a certain level.49 It seems that adding FEC above 50% to electrolyte mixtures has a marginally incremental to detrimental effect on the SEI properties. Researchers have suggested that this occurs due to LiF precipitation that coats the anode surface with a rigid film which results in a decrease of the cohesion and flexibility of the SEI layer. The Effect of FEC on Ethylene Carbonate Adsorption Angle. In this section, we first present our AIMD simulations of the EC adsorption angle followed by our SFG orientation analysis. In previous sections, we have reported that as the FEC bulk concentration increases so does their surface concentration. At a 50% EC/FEC weight ratio, FEC becomes the dominant species found at the a-Si surface with a 2:1 FEC/EC surface adsorption ratio. These adsorption ratios have a direct

consequence on the EC alignment that will be discussed next. The total simulation time was approximately 50 ps, where we took measurements at every femtosecond (50,000 measurements) for each EC adsorbed at the a-Si to create the timedependent histogram graphs shown in Figure 3. Each count on

Figure 3. Time-dependence histogram of the ethylene carbonate (EC) carbonyl group orientation relative to the normal of the surface for (a) 20 wt % FEC, (b) 40 wt % FEC, and (c) 50 wt % FEC content.

the histogram bar represents the number of times the carbonyl groups had an orientation between 0 and 90° relative to the normal of the surface. Here, we define the surface as the XY plane of the 3D simulation box. We see first that with 20 wt % FEC the histogram in Figure 3a shows a large number of events (events being the EC’s carbonyl group orientation) centering at a median value of ≈59°. With the increase of FEC content to 40%, the median value of the events shown in the histogram in Figure 3b is ≈68°. However, a closer look at the angle distribution statistics shows a bimodal distribution. That is, during the simulation the most common carbonyl group orientation is 75° relative to the surface’s normal but a more vertical 35° orientation is abundant as well. An increase in the frequency of EC alignment is observed but still accompanied by less ordered ECs. As we have stated before at 50 wt % FEC content, our simulations show that FEC molecules are the dominant species found at the a-Si surface with a 1:2 EC/FEC ratio. According to the adsorption angle distribution histogram shown in Figure 3c, this dominance translates to a more perpendicular orientation of the adsorbed EC molecule relative to the surface normal (≈28°). Thus, from our AIMD simulations, the adsorption angle of EC is shown to take on a more upright position with increasing FEC concentration. By polarizing the laser beams, that is measuring SFG spectra under SSP and SPS (SF, Vis, IR) polarization combinations, E

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Nano Letters

■

we can deduce the orientation angle of the carbonyl stretch of EC and FEC. We explain the SFG mathematical model fitting in the Supporting Information section. In brief, by fitting SFG data with eq S2 we can deduce the second-order of the surface nonlinear susceptibility, |?(2)| from the SFG signal amplitude at given bond frequencies. We then take the amplitude ratio at the two polarizations in conjunction with the CO bond molecular coordinates (eqs S3−S4). This leads to a CO bond angle assuming a Gaussian angular distribution (Figure S2). Note that the angle is defined in respect to the surface normal (Z-axis in Figure S2, left). When we take into account the standard deviation of the raw SFG data, shown in Figure S3, which is in the range of 10−15% of the amplitude, along with a 5−10% percent uncertainty, depending on the vibrational mode, due to the fitting according to eq S2 we calculate an ∼±6° uncertainty to the orientation angle. According to our analysis (Supporting Information), we show that in the case of pure EC, ∼60% of the ethylene carbonate molecules lie in a ∼45° angle to the a-Si anode surface and ∼40% with an ∼80° angle. Therefore, in the case of pure EC we have concluded that EC molecules lie in parallel to the a-Si anode surface. The EC orientation angle changes when mixed with FEC molecules. When adding 20 wt % FEC, we deduce that the majority (∼90%) of the EC molecules obtain a more upright position with their carbonyl group oxygen facing the a-Si anode substrate, having a ∼25° angle in respect to the a-Si surface normal. Nevertheless, we have detected that some of the EC molecules (∼10%) are still lying in parallel to the a-Si surface with an ∼80° angle in respect to the surface normal. While nominally the AIMD CO bond angle values differ from the SFG ones, we do notice that the general trend of the EC carbonyl bond assuming a more upright position as the content of FEC increases follows both the theory and experiments. In conclusion, sum frequency generation vibrational spectroscopy is a powerful tool to investigate interfacial/two-dimensional systems at long-term steady-state conditions. Ab initio molecular dynamics is a potent technique to observe short-term perturbations in defined systems on the molecular level. By combining SFG and AIMD, we were able to clarify the interaction of EC and FEC at various electrolyte mixtures in contact with an amorphous silicon anode, having a native oxide termination. Our results indicate that the global behavior of the electrolyte mixture on the silicon surface under OCP conditions is as follows: first, EC and FEC molecules adsorbed through their carbonyl group to the a-Si surface, and once the adsorption takes place the SiO bond is strong enough that the molecules do not desorb as the system evolves. Second, as the FEC content is increased it becomes the dominant species at the a-Si surface; third, with more FEC content the EC molecules assume a more perpendicular adsorption angle in respect to a-Si surface plain. Our finding can help in determining EC/FEC electrolyte ratios better suited for the use of silicon anodes in the Li-ion vehicular industry.

■

Letter

AUTHOR INFORMATION

Corresponding Authors

*(G.A.S.) E-mail: [email protected]. *(P.B.B.) E-mail: [email protected]. ORCID

Yonatan Horowitz: 0000-0002-7150-9264 Walter T. Ralston: 0000-0002-7632-7304 Perla B. Balbuena: 0000-0002-2358-3910 Gabor A. Somorjai: 0000-0002-8478-2761 Author Contributions

All authors have given approval to the final version of the manuscript. Y.H., H.-L.H., and F.A.S. contributed equally. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Freedom CAR and Vehicle Technologies of the U.S. Department of Energy under Contract No. DE-AC02 O5CH1123. P.B.B. and F.A.S. acknowledge financial support from the Assistant Secretary for Energy Efficiency and Renewable Energy, Office of Vehicle Technologies of the U.S. Department of Energy under Contract No. DE-EE0007766, of the Advanced Batteries Materials Research (BMR) Program. The SFG instrumentation was purchased with funding from the Director, Office of Basic Energy Sciences, Materials Science and Engineering Division of the U.S. Department of Energy. P.B.B. and F.A.S. acknowledge highperformance computational resources at Texas A&M University (College Station) and the Texas Advanced Computing Center (TACC).

■

REFERENCES

(1) Goodenough, J. B.; Park, K. S. J. Am. Chem. Soc. 2013, 135, 1167−1176. (2) Kraytsberg, A.; Ein-Eli, Y. Adv. Energy Mater. 2012, 2, 922−939. (3) Xu, K. Chem. Rev. 2004, 104, 4303−4417. (4) Xu, K. Chem. Rev. 2014, 114, 11503−618. (5) Xu, K.; von Cresce, A. J. Mater. Chem. 2011, 21, 9849. (6) Peled, E. J. Electrochem. Soc. 1979, 126, 2047−2051. (7) Gauthier, M.; Carney, T. J.; Grimaud, A.; Giordano, L.; Pour, N.; Chang, H.-H.; Fenning, D. P.; Lux, S. F.; Paschos, O.; Bauer, C.; Maglia, F.; Lupart, S.; Lamp, P.; Shao-Horn, Y. J. Phys. Chem. Lett. 2015, 6, 4653−4672. (8) Zhao, J.; Lu, Z. D.; Wang, H. T.; Liu, W.; Lee, H. W.; Yan, K.; Zhuo, D.; Lin, D. C.; Liu, N.; Cui, Y. J. Am. Chem. Soc. 2015, 137, 8372−8375. (9) Goldshtein, K.; Freedman, K.; Schneier, D.; Burstein, L.; Ezersky, V.; Peled, E.; Golodnitsky, D. J. Electrochem. Soc. 2015, 162, A1072− A1079. (10) Choi, N.-S.; Yew, K. H.; Lee, K. Y.; Sung, M.; Kim, H.; Kim, S.S. J. Power Sources 2006, 161, 1254−1259. (11) Nakai, H.; Kubota, T.; Kita, A.; Kawashima, A. J. Electrochem. Soc. 2011, 158, A798. (12) Elazari, R.; Salitra, G.; Gershinsky, G.; Garsuch, A.; Panchenko, A.; Aurbach, D. J. Electrochem. Soc. 2012, 159, A1440−A1445. (13) Nie, M.; Abraham, D. P.; Chen, Y.; Bose, A.; Lucht, B. L. J. Phys. Chem. C 2013, 117, 13403−13412. (14) Lin, Y.-M.; Klavetter, K. C.; Abel, P. R.; Davy, N. C.; Snider, J. L.; Heller, A.; Mullins, C. B. Chem. Commun. 2012, 48, 7268−7270. (15) Etacheri, V.; Haik, O.; Goffer, Y.; Roberts, G. A.; Stefan, I. C.; Fasching, R.; Aurbach, D. Langmuir 2012, 28, 965−76. (16) Dalavi, S.; Guduru, P.; Lucht, B. L. J. Electrochem. Soc. 2012, 159, A642.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.nanolett.7b04688. Thorough explanation on the SFG process and how one calculates the adsorption angle of an adsorbate. XPS, SEM, AFM and AIMD simulations. (PDF) F

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX

Letter

Nano Letters (17) Burns, J. C.; Kassam, A.; Sinha, N. N.; Downie, L. E.; Solnickova, L.; Way, B. M.; Dahn, J. R. J. Electrochem. Soc. 2013, 160, A1451−A1456. (18) Shkrob, I. A.; Wishart, J. F.; Abraham, D. P. J. Phys. Chem. C 2015, 119, 14954−14964. (19) Michan, A. L.; Parimalam, B. S.; Leskes, M.; Kerber, R. N.; Yoon, T.; Grey, C. P.; Lucht, B. L. Chem. Mater. 2016, 28, 8149. (20) Horowitz, Y.; Han, H.-L.; Ralston, W. T.; de Araujo, J. R.; Kreidler, E.; Brooks, C.; Somorjai, G. A. Adv. Energy Mater. 2017, 7, 1602060. (21) Shen, Y. R. Nature 1989, 337, 519−525. (22) Marx, D.; Hutter, J. Ab initio molecular dynamics: Theory and implementation. In Modern Methods and Algorithms of Quantum Chemistry; Grotendorst, J., Ed.; John von Neumann Institute for Computing: Julich, 2000; Vol. 1, pp 301−449. (23) Liu, H.; Tong, Y.; Kuwata, N.; Osawa, M.; Kawamura, J.; Ye, S. J. Phys. Chem. C 2009, 113, 20531−20534. (24) Mukherjee, P.; Lagutchev, A.; Dlott, D. D. J. Electrochem. Soc. 2012, 159, A244−A252. (25) Yu, L.; Liu, H.; Wang, Y.; Kuwata, N.; Osawa, M.; Kawamura, J.; Ye, S. Angew. Chem., Int. Ed. 2013, 52, 5753−6. (26) Nicolau, B. G.; García-Rey, N.; Dryzhakov, B.; Dlott, D. D. J. Phys. Chem. C 2015, 119, 10227−10233. (27) Liu, W.-C.; Melaet, G.; Ralston, W. T.; Alayoglu, S.; Horowitz, Y.; Ye, R.; Hurlburt, T.; Mao, B.; Crumlin, E.; Salmeron, M.; Somorjai, G. A. Catal. Lett. 2016, 146, 1574−1580. (28) Alves Dalla Corte, D.; Caillon, G.; Jordy, C.; Chazalviel, J.-N.; Rosso, M.; Ozanam, F. Advanced Energy Materials 2016, 6, 1501768. (29) Bain, C. D. J. Chem. Soc., Faraday Trans. 1995, 91, 1281−1296. (30) Ewing, C. S.; Bhavsar, S.; Veser, G.; McCarthy, J. J.; Johnson, J. K. Langmuir 2014, 30, 5133−5141. (31) Ewing, C. S.; Hartmann, M. J.; Martin, K. R.; Musto, A. M.; Padinjarekutt, S. J.; Weiss, E. M.; Veser, G.; McCarthy, J. J.; Johnson, J. K.; Lambrecht, D. S. J. Phys. Chem. C 2015, 119, 2503−2512. (32) Aktulga, H. M.; Fogarty, J. C.; Pandit, S. A.; Grama, A. Y. Parallel Computing 2012, 38, 245−259. (33) Newsome, D. A.; Sengupta, D.; Foroutan, H.; Russo, M. F.; van Duin, A. C. T. J. Phys. Chem. C 2012, 116, 16111−16121. (34) Custer, J. S.; Thompson, M. O.; Jacobson, D. C.; Poate, J. M.; Roorda, S.; Sinke, W. C.; Spaepen, F. Appl. Phys. Lett. 1994, 64, 437− 439. (35) Akkermans, R. L. C.; Spenley, N. A.; Robertson, S. H. Mol. Simul. 2013, 39, 1153−1164. (36) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868. (37) Kresse, G.; Furthmüller, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 11169−11186. (38) Kresse, G.; Furthmüller, J. Comput. Mater. Sci. 1996, 6, 15−50. (39) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558−561. (40) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 49, 14251−14269. (41) Blöchl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 17953−17979. (42) Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758−1775. (43) Nosé, S. Mol. Phys. 1984, 52, 255−268. (44) Grimme, S.; Ehrlich, S.; Goerigk, L. J. Comput. Chem. 2011, 32, 1456−1465. (45) Fortunato, B.; Mirone, P.; Fini, G. Spectrochimica Acta Part A: Molecular Spectroscopy 1971, 27, 1917−1927. (46) Song, S. W.; Zhuang, G. V.; Ross, P. N. J. Electrochem. Soc. 2004, 151, A1162. (47) Kim, J. S.; Byun, D.; Lee, J. K. Curr. Appl. Phys. 2014, 14, 596− 602. (48) Ma, Y.; Balbuena, P. B. J. Electrochem. Soc. 2014, 161, E3097− E3109. (49) Zhao, X.; Zhuang, Q. C.; Xu, S. D.; Xu, Y. X.; Shi, Y. L.; Zhang, X. X. Int. J. Electrochem Sc 2015, 10, 2515−2534. G

DOI: 10.1021/acs.nanolett.7b04688 Nano Lett. XXXX, XXX, XXX−XXX