Although electrolyte decomposition is a key issue for the stability of Li-ion batteries and has been intensively investigated in the past, a common understanding of the concepts and involved processes is still missing. In this article, we present an overview on our results obtained with a surface science approach and discuss the implications for the stability window of Li-ion electrolytes under consideration of calculated oxidation potentials from the literature. We find LiCoO2 valence band–solvent highest occupied molecular orbital offsets that are in agreement with expectations based on ionization potentials, polarization effects, and solvent–salt interactions. In agreement with thermodynamic considerations, our data show that surface layer formation on pristine electrodes occurs inside the electrochemical window as defined by the measured oxidation and reduction potentials, which can be attributed to electrode surface interactions. The results demonstrate that the simple energy level approach commonly used to evaluate the stability window of Li-ion electrolytes has very limited applicability. The perspectives for further investigations of the electronic structure of Li-ion cathode–liquid electrolyte interfaces are discussed.
Electrode–electrolyte interfaces in Li-ion batteries are characterized by a multitude of side reactions, which lead to electrode corrosion and electrolyte decomposition.1,2 For practical electrodes, reaction rates are limited by coatings and/or surface layers, which are applied during material preparation3,4 or formed during the first operation cycles.5,6 Nevertheless, side reactions are ongoing and strongly influence the cycle life of the battery cells, which is currently a major issue for the implementation of high energy electrode materials (see, e.g., Refs. 7–12).
At the cathode–electrolyte interface (CEI), electrolyte decomposition is strongly related to oxidation reactions at high potentials (high charge states). The oxidized electrolyte species (organic solvent, additive, or salt) undergo irreversible reactions forming different types of products, ultimately resulting in the growth of surface layers and the presence of gas.
Information on the side reactions is commonly gained by the evaluation of the reaction products by surface and gas analysis,13,14 and fundamental concepts and reaction paths have been proposed.15–18 Side reactions are aggravated by the presence of additional (parasitic) energy levels, which reduce the electrochemical stability window (see Ref. 19). However, a common, comprehensive understanding of the concepts and involved processes is still missing (see, e.g., Ref. 20).
The lack of fundamental understanding also extends to other practical electrochemical systems. The reasons for this are analytical difficulties caused by the multitude and high complexity of the involved reactions and interfaces as well as the lack of a consistent, yet practical conceptual framework used across the different disciplines.
Within Li-ion battery research, the commonly used and accepted model to assess electrolyte decomposition considers the kinetics of electron transfer on the basis of the electronic structure of the electrode in relation to the electronic levels of the electrolyte15 (see Fig. 1). In its simple (and most used) form, this model considers outer sphere charge transfer without solvation effects or any other interaction between the molecular species in the electrolyte. Other, more sophisticated theoretical approaches include solvation effects by means of free energy calculations21–23 or surface-induced reactions,16,24,25 respectively.
In this article, we present an overview on our experimental results on the electronic structure of LiCoO2 –solvent interfaces. The data were obtained using a surface science approach,27,28 allowing the determination of the electronic structure of the interface including effects from the electrochemical double layer. The results are linked to literature data of oxidation potentials for solvent–salt complexes, which demonstrate that the presence of salt significantly reduces the oxidation stability compared to the pure solvent phase. As a consequence, interfacial energy level diagrams are proposed, which contain electronic states in the electrolyte phase arising from the coordination of solvent molecules with salt anions. Different aspects of the interface formation of LiCoO2 electrodes have been published previously, especially in Refs. 29–32. Here, an integrated overview with key insights regarding the electronic structure and reactivity is given based on Ref. 26, and the perspectives for further work are discussed.
II. FUNDAMENTAL CONSIDERATIONS
The positions of the energy levels in the electrolyte are fundamental to the concept of the electrochemical stability window and will be considered in Subsections II A and II B. Other important aspects are the role of subsequent reactions for oxidation potentials and electrolyte stability, the influence of surface interactions, and double layer effects, which are discussed together with the data particularly in Secs. IV, V B, and V D.
A. Oxidation potential
There are several approaches for the evaluation of energy levels in the electrolyte, which are mainly derived from thermodynamical considerations and address different aspects of the energy level structure. The electrochemical potential of redox electrons is commonly established via a Born-type of cycle, including the ionization potential of the isolated species as well as the solvation energies of the reduced and oxidized species.33 On the other hand, the distribution of electronic states around the redox potential is treated according to the Marcus model, which states that the electronic levels of reduced and oxidized species are separated due to solvent reorganization and are subject to statistical broadening due to the fluctuations of the solvent shell.33,34 Finally, the oxidation potential of the organic solvent is typically evaluated via the ionization potential of the isolated solvent molecule diminished by the difference in free energies of solvation between the compound and its cation. This last approach has been found to be consistent with experimental data in the past35–37 and is also currently applied to battery electrolytes with reasonable results.21–23
Nevertheless, none of these models presents by itself a detailed view of the electronic states referenced to the vacuum level, which would be needed for an improved conceptual understanding of oxidation potentials. Toward this end, it is useful to consider the evolution of the electronic structure from single particles (gaseous molecules) to the condensed phase and relate the oxidation potential to the resulting energy level structure. Figure 2 illustrates this procedure for the simple case of solvent molecules forming a solvent phase. Starting from the ionization potential of a gaseous molecule, both relevant polarization effects and thermal broading need to be included. In addition, the interaction of the molecules in the solvent phase is expected to have an influence, which should remain small, however (a few tens of eV). The consideration of the different effects results in an energy level distribution from which the oxidation potential can be specified by the position of the high energy tail of the highest occupied molecular orbital (HOMO) states. Note that the presence of salt is expected to shift the solvent HOMO states to higher energy due to the electrostatic interaction with salt anions, as indicated by theoretical calculations and experimental data.23 This shift, which is essential for the understanding of oxidation potentials of practical electrolytes, is discussed in more detail in Sec. V.
It should be noted that the value for the polarization contribution depends on the time scale for charge transfer. For electrochemical systems, the rate of electron transfer is assumed to be on the order of 10−15 s.33 On this time scale, no reorganization can occur (only electronic polarization), in agreement with the Marcus model. This reflects the situation that is also encountered during photoemission.38,39 Reorganization effects, which can be substantial,40 are indirectly included via the broadening of the states, however, so that the oxidation potential as discussed above depends on polarization effects on all time scales.
B. Electrochemical stability window
In the discussion of battery electrolytes, it is useful to first briefly recall the specific situation of organic electrolytes (see also Ref. 30). In contrast to water, organic solvents are aprotic and oxidation (reduction) results in unstable compounds, which undergo decomposition (irreversible reactions). Therefore, the stability window is given by oxidation (reduction) potentials related to the HOMO (LUMO) levels of the electrolyte, as discussed above, and not by a pair of redox-couples (H+/H2 and O2/H2O), which dispose of energy levels in the HOMO–LUMO gap (see Fig. 1). As a result, the electrochemical stability window is larger for organic solvents and intrinsically coupled to the electronic structure of the electrolyte phase.
At this point, it should be noted that due to its nature described above, the electrochemical stability window of organic solvents does not constitute a region of thermodynamic stability, in contrast to water. The large electrochemical stability window can therefore be attributed to a high kinetic inhibition rather than to a high thermodynamic stability, which is much lower. Suitable reaction partners can further lower the stability so that surface films can be formed and also maintained inside the stability window. Furthermore, the stability window may be significantly lowered by surface reactions and catalytic effects, and it may be shifted due to double layer effects.
Figure 3 illustrates the different electronic and thermodynamic levels in the electrolyte obtained from the literature and calculations.30 The distance between the Li+/Li redox level and a given energy level indicates the (positive) potential value vs Li+/Li. It can be seen that the theoretical and experimental oxidation potentials using inert electrodes are significantly lower (vs Li+/Li) than the expected from the HOMO position. In addition, it is demonstrated that the formation of lithium oxide (Li2O) and lithium carbonate (Li2CO3) is possible inside the electrochemical stability window.
III. EXPERIMENTAL APPROACH
The data on the electronic structure presented in this paper are derived from interface experiments using photoelectron spectroscopy (XPS) or synchrotron based XPS with soft x rays (SXPS). Such experiments consist of stepwise condensation (or deposition, respectively) of overlayer phases consisting of solvent or inorganic compounds with intermediate analysis, resulting in a spectral series with growing overlayer thickness.28,41,42 Evaluation of the evolution of the core level and valence band signatures, of (coordinated) binding energy shifts, and of the evolution of the work function allows the determination of energy level offsets, band bending, and surface/interface dipole layer formation (see Fig. 4). Most data regarding offsets have been previously reported and are summarized in Ref. 29, where the original references can also be found.
Additional data have been obtained by emersion experiments.28 In such experiments, the electrode is immersed in a solvent or electrolyte and afterward analyzed by photoemission. This approach is well suited for the investigation of the surface layer formation by chemical reactions in the presence of salt.
Both interface and emersion experiments have been conducted using thin film LiCoO2 as the electrode material deposited by sputter deposition.43 This thin film material is fully lithiated and has a potential of 3 V vs Li+/Li. More details on the material and the different approaches can be found in Ref. 27 and 43–46.
IV. CHEMICAL STRUCTURE OF CATHODE–ELECTROLYTE INTERFACES
In our adsorption experiments, solvent decomposition and surface layer formation can be observed for all solvents during the first steps. At somewhat higher exposures, the reactions stop and a layer of condensed solvent is formed, demonstrating that the reaction layer passivates the surface.
Figure 5 illustrates the situation with a layer of condensed carbonate solvent covering the reaction layer after the extended exposure. The thickness and the composition of the reaction layer depend on the type of solvent.47–49 For carbonate solvents, sub-nm surface layers are formed, which contain Li-semicarbonate (ROCO2Li), Li-alkoxide (ROLi), Li-carbonate (Li2CO3), and Li-oxide (Li2O).
The results demonstrate that lithium from the electrode reacts with the solvent. In this process, both Li-ions and electrons are transferred to the solvent phase, and the reaction product forms a deposit on the surface. Considering the potential of the electrode material (3 V vs Li/Li+), this reaction occurs well within the electrochemical stability window of carbonate solvents, indicating that surface interactions and/or the availability of Li+-ions have reduced the stability. This view is supported by the low amount of reaction products required to passivate the surface and by experiments with different coating materials (Co3O4, ZrO2, and LixZryOz), demonstrating the dependence of reactivity on surface chemistry.
In the case of practical battery electrolytes, the surface layer contains products from salt decomposition; in addition, a subsurface corrosion layer is formed (not shown). In the case of a LiPF6 electrolyte, the top layer also contains lithium fluoride and lithium fluorophosphates, and the corrosion layer consists of cobalt oxide (Co3O4) and/or cobalt oxy-hydroxide.13,32
Overall, it can be stated that the behavior of cathodes in contact with battery electrolytes is typical for systems with spontaneous surface passivation, as observed for some metals, or also for poisoning of catalyst surfaces. In the present case, it can be presumed that the Li-containing surface compounds block further transport of lithium across the interface and/or cover reactive sites, respectively, such as point defects and grain boundaries.
V. ELECTRONIC STRUCTURE OF CATHODE–ELECTROLYTE INTERFACES
A. Energy level offsets for different solvent phases
In this subsection, offsets between the valence band maximum of the electrode and the HOMO onset of different solvents are discussed (VB–HOMO offsets). The use of the valence band maximum position for the discussion of the electrolyte stability (instead of the Fermi level, which is more commonly used) has the advantage that it addresses directly the states that are available for electron transfer; in addition, the resulting offset is less dependent on the charge state of the electrode. Nevertheless, it should be noted that also VB–HOMO offsets are subject to the influence of the charge state of the electrode, which affects changes in the double layer and in the electronic structure of the electrode itself.
The offsets reported here have been obtained from the valence band spectra of solvent layers condensed on thin film LiCoO2-electrodes and include double layer effects caused by transfer of mobile Li-ions to the electrolyte and dipole potentials due to orientation or chemisorption. The formation of a double layer is evidenced by band bending in the LiCoO2, which is observed upon the formation of the solvent layer. However, due to the absence of any salt in the solvent phase, somewhat different electrostatic drops at the interface are expected than for practical interfaces, which are discussed in Subsection V D in conjunction with band diagrams.
For the investigated organic solvents, the offsets range from 3 eV to 5 eV (Table I). Considering that the ionization potential of LiCoO2 is close to 5 eV and neglecting any double layer effects, these offsets result in onsets of the energy levels in the electrolyte at energies between 8 eV and 10 eV with respect to the vacuum level. This is in principal agreement with the positioning of the energy levels presented in Fig. 2 and the high ionization potentials of the solvent molecules, as obtained by theoretical calculations (8–11 eV).21,23
|Solvent .||EVB–HOMO (eV) .||eVbb (eV) .|
|Solvent .||EVB–HOMO (eV) .||eVbb (eV) .|
Interestingly, the order of the offsets observed in the experiments dimethylsulfoxide (DMSO) > diethyl carbonate (DEC) > ethylene carbonate (EC) does not correspond to the order as obtained from ionization potentials from the literature (DMSO < DEC < EC5,21), indicating an influence of other effects, such as differences in polarization or interface dipole effects.
The high offsets demonstrate that no significant outer sphere oxidation should occur for pure, condensed phases of the investigated organic solvents at fully discharged lithium cobalt oxide (LCO) electrodes. Exemplary band diagrams and considerations with regard to the presence of salt and higher charge states are given in Subsection V D.
B. Interface states and surface chemistry
The bonding between molecules and surfaces of oxides (chemisorption) leads to additional electronic states (interface states), which can be observed in valence spectra.50 Common for metal oxide surfaces is the formation of dative bonds (acid–base interactions) such as the interaction of the HOMO of the molecule with unoccupied orbitals of the metal cation.50,51 Due to the lowering of the energy, the formation of such bonds leads to states with higher binding energies than the binding energies of the occupied orbitals involved.
Figure 6 shows the valence band difference spectra obtained at different steps of DEC adsorption.31 At high exposure, the difference spectrum reflects the electronic structure of the condensed phase and allows the determination of the onset of the HOMO states. At low exposure, on the other hand, the spectral signature contains additional contributions due to interface states and reaction products. In the present case, we attribute the additional feature at 4.5 eV to an interface state as the valence band states of the foremost reaction products are located to lower binding energies.
The binding energy of the additional feature is lower than the binding energy of the HOMO, indicating that not the HOMO but the LUMO of DEC is involved. Consequently, we attribute the feature to an interface state, which derives from the interaction of O2p orbitals of surface oxygen with the LUMO of DEC, as shown in Fig. 7. This interaction is synonymous with the formation of a dative bond between the surface oxygen and the carbonate carbon, constituting a nucleophilic attack on the carbonate carbon, which was already proposed by others.16
In the proposed interaction, the surface oxygen atoms act as a Lewis-base site, and calculations support the basic character of the LiCoO2 surface.52,53 Nevertheless, by way of adsorption of molecular probes, we also found evidence for the presence of surface sites with Lewis-acid character54 likely related to cobalt surface atoms. Thus, overall indications are found that both surface sites with the acid and base character are present. These sites may open reaction paths for both solvent reduction and oxidation, which is further discussed in Sec. VI.
C. Valence band offsets of surface layers
As discussed previously, cathode material surfaces are usually covered with surface layers. These layers may be native reaction layers, coatings applied during the production process, or surface layers formed during the first cycles. Such CEI layers passivate the surface and protect both the electrode and electrolyte from fast degradation. Only in the case of cracking, pristine surfaces will be exposed to the electrolyte, which also subsequently undergo passivation.
Typically, inorganic compounds cover the surface of the electrode and strongly contribute to the passivation of the cathodes, i.e., to a significant decrease in electrolyte oxidation. For a full picture of the reactivity of practical cathode–electrolyte interfaces, the presence of these compounds and their effect on the electronic interface properties has to be considered.
Table II gives the LiCoO2 VB–overlayer VB offsets (VB-offsets) for different compounds, as obtained from interface experiments. Included are compounds present in the natural CEI as well as compounds that are applied as coatings. All compounds are wide bandgap materials with low electronic conductivity but differ considerably in the value of their bandgaps (LiF: 13 eV55 and LiPON: 4–6 eV56,57).
|Overlayer .||ΔEVB (eV) .||eVbb (eV) .|
|Overlayer .||ΔEVB (eV) .||eVbb (eV) .|
The VB-offsets cover a wide range between 1.3 eV (LiPON) and 4.5 eV (LiF) with an order that reflects the value of the bandgap (LiPON < LiPO ≤ Li2O < LiF). This observation is in agreement with the notion that materials with a higher bandgap have a higher ionization potential and that interface dipole potentials remain moderate (for a more detailed discussion, see Ref. 29).
In addition, for the interfaces with the overlayers, band bending in the LiCoO2 surface is observed, which is attributed to Li-ion transfer and the formation of electrochemical equilibrium across the interface.
The high offsets for all overlayer compounds demonstrate that electron transfer from the overlayer to LiCoO2 as required for electrolyte oxidation is strongly inhibited. Furthermore, implications are discussed in Subsection V D.
D. Band diagrams
The results obtained from the surface science experiments allow us to establish band diagrams derived from experimental data.30 In the following, these band diagrams and their implications for the energy level structure of the electrolyte will be discussed.
Figure 8 reveals the experimentally derived band diagram of LiCoO2-solvent interfaces. Shown is the situation of the pristine surface in contact with the pure DEC phase without consideration of any surface or reaction layer. For the other solvents, similar band diagrams are obtained but with different values for band bending and VB–HOMO offsets (see Table I).
Next to band bending in the LiCoO2, the interface is characterized by the presence of a dipolar layer, as indicated by our data and illustrated by the discontinuous profile of the vacuum level close to the interface. In fact, the interface exhibits quite a complex double layer formation, but overall, the impact on the VB–HOMO offset remains low.
The space charge layer formation as indicated by band bending is attributed to the transfer of Li-ions to the electrolyte in the process of obtaining electrochemical equilibrium. As a consequence, the negative space charge remains in the LiCoO2, leading to downward band bending and positive, partially or fully solvated Li-ions are located in the solvent phase at the inner and outer Helmholtz planes. In this context, the high value of band bending is caused in the initial absence of Li-ions in the solvent phase, resulting in the transfer of a high amount of Li+-ions. The dipolar layer, on the other hand, is attributed to the orientation of solvent molecules at the interface.
Due to the comparatively low value of the dipole potential at the interface, the onset of energy level distribution in the electrolyte with respect to the vacuum level can be expressed by the sum of VB–HOMO offset and ionization potential of the LiCoO2, which yields 9 eV for DEC, as indicated in Fig. 8. This is lower than the values of around 11 eV, which are typically quoted for the ionization and oxidation potentials, respectively, for linear carbonates.5,23 The discrepancy is attributed to broadening and polarization effects, as shown in Fig. 2. For EC, a lower value (8.3 eV) is also obtained.
With these onset values with respect to the vacuum level, the expected values for the oxidation potential vs Li+/Li are calculated to 7.5 V and 6.8 eV for DEC and EC, respectively. Experimentally, values as high as 6.7 eV and 6.2 eV are reported for oxidation potentials of linear carbonates and EC measured with inert electrodes,59 respectively, depending on the type of salt (see Ref. 23). This observation is in agreement with theoretical investigations, which indicate that interaction between salt anions and solvent plays a significant role in solvent oxidation and significantly reduces the oxidation potentials.23 Typical values range from 0.5 to 1 eV depending on the type of solvent and anion, resulting overall in a reasonable agreement between the energy level onset positions and measured oxidation potentials.
The single effects that lead to the decrease in the oxidation potential and their influence on the electronic levels of anion–salt complexes are difficult to assess. The theoretical calculations show that H- and F-transfer occurs, but also electrostatic interaction between anion and the solvent will be involved.
In order to discuss interface and material related issues, additional band diagrams are shown in Figs. 9–11. These band diagrams illustrate the effect of the presence of salt (Fig. 9), of polarization (Fig. 10), and of the presence of a passivation layer (Fig. 11).
The presence of salt anions is expected to shift the HOMO levels of the solvent upward, and this explains the decrease in the oxidation potential from an electronic level point of view (see also Fig. 2). Figure 9 shows a LiCoO2 electrode, as discussed previously, including solvent energy levels that shifted according to the reduction in oxidation potential, as calculated in Ref. 23. It should be noted that to our knowledge, the electronic structure of the electrolyte (solvent and salt) has not yet been simulated or measured and that, especially, the influence of local solvation effects is not clear.
In addition, the presence of Li+-ions in the electrolyte is expected to change the electrostatic potential drop in the Helmholtz layer. For a battery electrolyte with a concentration of 1M Li+-ions, the total electrostatic potential drop across the interface is expected to be decreased compared to our experiment due to the higher Li+-ion chemical potential in the electrolyte. Due to the relative insensitivity of the electrode potential to concentration changes (59 mV/decade), this decrease should be comparatively small (<0.5 eV), which is verified by emersion experiments using different concentrations of salt (LiCl in DEC). In the case of fully lithiated LCO, this decrease is reflected by a lower band bending (Fig. 9), while for charged electrodes, the drop in the Helmholtz layer is modified accordingly.
Figure 10 proposes the band diagram for fully charged LCO with a pristine surface at first contact with a battery electrolyte, i.e., in the absence of surface layers. In reality, this situation can occur due to the cracking of the cathode material, exposing the pristine surface to the electrolyte. In the charged state, the Fermi level of the LCO material is located at lower energy (at higher binding energy) inside the valence band, resulting in a metallic-like behavior of the material. As a consequence, the electrostatic potential drop at the interface now occurs in the Helmholtz layer. For the fully charged state at a potential of 4.2 V vs Li+/Li, the electronic states in the electrolyte with the highest energy derived from solvent molecules that are part of a solvation shell are still located at 2.5–3.0 eV below the Fermi level of the cathode, in agreement with the high oxidation potential at inert electrodes previously discussed. This is in line with the notion that surface induced processes are involved when decomposition is observed at significantly lower potentials, which is discussed in more detail in Sec. VI.
For real cathodes, the surface is covered with a surface layer or coating. This situation is illustrated in Fig. 11, which depicts a fully lithiated LCO electrode with an overlayer of a high bandgap material such as LiF in contact with an electrolyte. The thickness and Li+-ion concentration of the overlayer are assumed to be high enough so that there is no electrostatic potential drop across the overlayer. Note that the HOMO states of the carbonate solvent (here DEC) are drawn according to their position relative to the vacuum level as previously determined. This approach is reasonable as no significant dipole layer contributions at LiF–solvent interfaces are expected.
The presence of surface layers such as LiF strongly inhibits the oxidative electrolyte decomposition, demonstrating that surface interaction and/or electron transfer is significantly reduced. In the case of a thick overlayer as shown, electron tunneling is prohibited, and solvent oxidation can only proceed via electron injection into the valence band of the overlayer. Although this may theoretically occur if the HOMO states are located above the valence band maximum of the overlayer as shown, electron injection into the cathode material is hardly possible due to the high VB-offset at the interface with the cathode material. Note that for a charged cathode, it is expected that the offset between the Fermi level of the electrode and the valence band of the overlayer is decreased by 1.2 eV, comparable to the cathode–solvent interface (Fig. 10). In this situation, the strong blocking property of the interface with the cathode is lost for overlayer materials with a significantly lower bandgap such as Li2O.
In principle, the band diagrams as shown should also be qualitatively valid for other layered cathode materials such as Li(Ni,Mn,Co)O2 (NMC) or Li(Ni,Co,Al)O2 (NCA). It is stressed, however, that significant differences exist, which have to be considered in a more detailed discussion, such as a different energetic position of the O1s states16 and differences in surface chemistry.19 Another factor to be considered is the change of electronic structure upon delithiation.19,28,60
VI. DECOMPOSITION PATHWAYS
The results concerning the surface chemistry of oxide cathodes by us and others (see, e.g., Refs. 16, 18, and 19) highlight the significance of surface interactions for electron transfer and electrode reactivity. In addition, the availability of Li+-ions is essential for the surface layer formation, and our results show that it also likely participates in the initiation of the decomposition process.61 In effect, these processes enable electrolyte decomposition reactions inside the electrochemical stability window, which have previously not been widely discussed. It appears that chemical processes and mixed reactions such as the solvent reduction and coupled Li+-ion transfer as observed in our experiments account for a larger part of reactions than commonly assumed.
On the basis of the results, initial reaction and subsequent decomposition pathways can be proposed for EC and DEC on pristine LCO surfaces, as shown in Fig. 12 for DEC.32 After the nucleophilic attack of the surface oxygen on the carbonyl carbon and subsequent cleavage of the DEC molecule, the remaining fragment of the chemisorbed carbonate is reduced and Li-ion transfer takes place. Subsequently, the carbonate undergoes decarboxylation resulting in a hydrocarbon radical, which initiates polymerization. Multiple reduction processes result in the formation of Li2O and Li2CO3, which effectively cover the catalytic sites and passivate the surface.
It should be noted that the reaction mechanism as discussed above involves the reduction of the solvent and is likely to be more prominent for discharged cathodes. For oxidative decomposition on oxide cathodes, deprotonation of the solvent and dissociative solvent adsorption, respectively, is believed to act as a trigger.19,25 In addition, for this mechanism, the surface oxygen with its nucleophilic nature plays a significant role.
Considering the previous results, the main function of such inorganic passivation films is likely to act as effective barriers for electrolyte species and prevent electrocatalytic processes at the cathode surface. At high electrode potentials and/or high temperature, they may prevent, in addition, outer sphere processes by blocking electron transport. It should be noted, however, that in real systems the long time effect of coatings is questionable due to the cracking of the active material particles.
In our investigations of cathode–solvent interfaces, we find valence band offsets, which are in agreement with expectations based on ionization potentials of the single phases and polarization effects. Under the consideration of solvent–salt interactions, these results are also in line with oxidation potentials, which were experimentally determined using inert electrodes.
In agreement with thermodynamic considerations, our data show that surface layer formation on pristine electrodes occurs inside the electrochemical window, as defined by measured oxidation and reduction potentials. The high reactivity is attributed to surface interactions, specifically to surface oxygen initiating solvent decomposition via nucleophilic attack and the presence of Li+-ions.
Our results support the view that the simple energy level approach using the position of the interaction-free solvent HOMO levels, which is often used to evaluate the oxidative stability of Li-ion electrolytes, is questionable. In fact, the evaluation of the oxidative stability by means of energy level diagrams also requires the consideration of additional electronic levels due to interaction with the salt ions and the electrode surface. Specifically, our results highlight the significance of ultra-thin, non-catalytic passivation layers made from compounds with low surface basicity.
While electronic energy levels are often invoked to discuss oxidative electrolyte decomposition, the electronic structure of electrode–electrolyte interfaces is hardly known. Clearly, more research using both theoretical and experimental approaches is required in order to shed more light on the electronic structure and its origin.
From the fundamental point of view, the evaluation of this structure must take into account the energy levels in the electrode, the levels in the electrolyte, any interactions that occur at the electrode surface, and the electrostatic potential drop across the interface. This certainly applies to cathode–electrolyte interfaces in Li-ion batteries but is also generally applicable to other electrochemical interfaces such as in catalysis.
So far, fundamental activities with respect to Li-ion electrolyte decomposition have focused on the calculation of thermodynamic oxidation potentials of the electrolyte. More recently, surface interactions have also been evaluated by calculations, and chemisorption processes have been demonstrated. Both theoretical approaches yield useful insights into relevant physical and chemical processes but fall short to capture the full complexity of cathode–electrolyte interfaces.
In order to come to meaningful results for practical systems, the different theoretical approaches need to be more closely linked to experimental investigations and have to include additional aspects related to charge transfer kinetics. For theoretical calculations, this means that additional configurations have to be considered to capture fluctuations in the surrounding media as well as ionic and double layer effects, resulting in a distribution of electronic states at the interface, which can be compared to experimental data.
On the other hand, experimental investigations need to become more sophisticated in order to obtain a more refined picture of the electronic structure and related charge transfer processes. Ultimately, this requires the combination of surface science experiments as presented and operando investigations using near-ambient pressure XPS62,63 as well as other electron spectroscopic methods (see Ref. 64).
Concerning surface science experiments, the next step is the investigation of salt effects as well as that of charging in order to probe more realistic conditions. For thermally stable salts such as lithium bis(trifluoromethanesulfonyl)imide (LiTFSI), the effect of salt can be simply investigated by co-deposition. Experiments using charged, layered oxide cathode materials prove to be more difficult. As these materials are thermodynamically unstable and cannot be easily prepared, charging the fully lithiated thin film material is the more promising route. Using an in situ solid state cell, this can be accomplished without contamination of the pristine surface, as we have recently shown.65 Such experiments also open up the possibility to perform operando investigations of electrode surfaces as a function of potential and temperature. Perspectively, it should also be possible to perform such experiments on particle materials (see, e.g., approach in Ref. 66), opening up the possibility to work also on commercial battery materials.
Fruitful discussions with Wolfram Jaegermann and Bernd Kaiser are gratefully acknowledged.