Thermodynamics of the Hydrolysis of Lithium Salts: Pathways to the Precipitation of Inorganic SEI Components in Li-Ion Batteries

Li⁺ is strongly coordinated by the PF₆‾ anion in the simulated solvent through electrostatic interactions with the partial negative charges of three fluorine atoms. Following the atomic labels shown in Fig. 1, the interatomic distances are Li1---F3 = 2.007 Å, Li1---F2 = 2.028 Å, Li1---F6 = 2.015 Å being the Li1---P1 distance 2.575 Å.

It is widely known that LiPF₆ undergoes in aprotic solvents two competitive dissociation reactions: ^45–48

$$\begin{eqnarray}&&LiP{F}_{6}\rightleftarrows L{i}^{+}+P{F}_{6}^{-}\end{eqnarray} \tag{ 3 }$$

$$\begin{eqnarray}&&LiP{F}_{6}\rightleftarrows P{F}_{5}+LiF\end{eqnarray} \tag{ 4 }$$

being the first the usual ionic dissociation and the second a Lewis acid/base equilibrium. The calculated Gibbs energy of both reactions at 298 K are positive, i.e.64 kJ mol⁻¹ and 92 kJ mol⁻¹, respectively. The ionic dissociation of reaction 3 is driven by the solvation of Li⁺ ions in dielectric media ⁴⁸ whereas reaction 4 is negligible below 60 °C in consideration to the large values of free reaction energy. This last equilibrium is commonly considered as the onset reaction of the thermal aging of LiPF₆-based electrolytes in presence of traces of water. In fact, above 60 °C the PF₅ molecule easily react with water giving:

$$\begin{eqnarray}&&P{F}_{5}+{H}_{2}O\rightleftarrows PO{F}_{3}+2HF\end{eqnarray} \tag{ 5 }$$

Our calculations suggest for 5 a Gibbs energy of reaction at 298 K of −51 kJ mol⁻¹. In the literature many authors reported that at room temperature 5 is slow, and the consensus suggests a LiPF₆ limited degradation through the formation of LiF: ^45,46

$$\begin{eqnarray}&&LiP{F}_{6}+{H}_{2}O\rightleftarrows PO{F}_{3}+2HF+LiF\end{eqnarray} \tag{ 6 }$$

being our computational prediction for the corresponding Gibbs energy of reactionat 298 K ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E6\right)$ = 41 kJ mol⁻¹.

Besides these known mechanisms, one can also speculate that LiPF₆ can react with water and irreversibly break in fragments leading to the release of LiOH through an Arrhenius acid/base equilibrium in solution:

$$\begin{eqnarray}&&LiP{F}_{6}+{H}_{2}O\rightleftarrows P{F}_{5}+HF+LiOH\end{eqnarray} \tag{ 7 }$$

The formation of HPF₆ is excluded since it is vibrationally unstable and spontaneously evolves in a coordinated adduct PF₅···HF, constituted by two weakly interacting fragments. Overall, the computed thermodynamics of this process is very endergonic (${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E7\right)$ = 186 kJ mol⁻¹ at 298 K) and it is unlikely to occur. Generally speaking, the thermodynamic landscape of the hydrolysis of LiPF₆ in solution outlined from our calculation does not clarify the strong tendency of this solvated salt to easily degrade in presence of traces of water.

A different picture can be drawn by considering the precipitation of solid LiOH(s) and LiF(s). In Fig. 3 the potential energy surface (PES) of the hydrolysis mechanism of lithium hexafluorophosphate considering the precipitation of solids is shown at three different temperatures. These heterogeneous reaction channels show smaller or even negative energy barriers and therefore are thermodynamically feasible in the 248–348 K temperature range:

$$\begin{eqnarray}&&LiP{F}_{6}+{H}_{2}O\rightleftarrows P{F}_{5}+HF+LiOH\left(s\right)\end{eqnarray} \tag{ 7b }$$

$$\begin{eqnarray}&&LiP{F}_{6}+{H}_{2}O\rightleftarrows PO{F}_{3}+2HF+LiF\left(s\right)\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&2LiP{F}_{6}+{H}_{2}O\rightleftarrows 2P{F}_{5}+HF+LiOH\left(s\right)+LiF\left(s\right)\end{eqnarray} \tag{ 9 }$$

being the corresponding Gibbs energies of reaction at 298 K 35, –117 and −68 kJ mol⁻¹, respectively. On passing, one can note that the precipitation of LiOH(s) as a first hydrolysis products decrease the Gibbs energy of reaction from 186 (7) to 35 kJ mol⁻¹ (7b) at 298 K: this energy barrier is therefore small enough to start the LiPF₆ hydrolysis at room temperature observed experimentally. ^45,46 Overall, this thermodynamic path nicely matches with the experimental evidence in model studies ^45,46 as well as the reported presence of oxofluorides, LiOH(s) and LiF(s) in the composition of passivation films grown over negative electrodes in aprotic LIBs. ²⁷

Zoom In Zoom Out Reset image size

The summary of the ${{\rm{\Delta }}}_{r}{G}_{T}^{o}$ values for reactions involved in the hydrolysis of LiPF₆ are reported in the SI, Table SII.

Li⁺ cation is coordinated by FSI⁻ anion in simulated aprotic solvent via electrostatic interactions with partial negative charges of O2 and O4 atoms (see Fig. 1). Following the atomic labels depicted in Fig. 1, the interatomic distances Li1---O₂ and Li1---O₄ are identical (1.88 Å) and the value of angle S2-N1-S1 is 122°. The two fluorine atoms and computed structure of LiFSI are in syn configuration.

In Fig. 4 the potential energy surface (PES) of the hydrolysis mechanism of LiFSI is shown.

Zoom In Zoom Out Reset image size

This reaction pathway starts with the acid/base equilibrium between LiFSI and water (10–11):

$$\begin{eqnarray}&&LiFSI+{H}_{2}O\rightleftarrows HFSI+LiOH\left(s\right)\end{eqnarray} \tag{ 10 }$$

$$\begin{eqnarray}&&LiFSI+{H}_{2}O\rightleftarrows HFSI+LiOH\left(sol\right)\end{eqnarray} \tag{ 11 }$$

The precipitation of LiOH drives the thermodynamics, being the computed Gibbs reaction energy to give solid lithium hydroxide (10) 13 kJ mol⁻¹ at 298 K, whereas the reaction to the solvated LiOH(sol) (11) has a ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E11\right)$ = 164 kJ mol⁻¹ at 298 K.This different thermodynamic landscape suggests that LiFSI can react with water leading to the precipitation of lithium hydroxide and forming HFSI, especially for temperatures below 298 K.

The structure of HFSI is shown in the SI, Fig. S1 (available online at stacks.iop.org/JES/168/100514/mmedia); it can react again through reaction (12):

$$\begin{eqnarray}&&HFSI\rightleftarrows FN{O}_{4}{S}_{2}+HF\end{eqnarray} \tag{ 12 }$$

This reaction is endergonic and consists in the neutral loss of hydrofluoric acid from HFSI. The main product of the 12 reaction is the molecular fragment FNO₄S₂ (see SI, Fig. S1). The minimal energy barrier to climb to activate this process is as high as 43–57 kJ mol⁻¹, in respect to the reagent molecule (HFSI). This value is high enough to expect a limited reactivity through this channel, even smaller if large activation barriers occur. Therefore, in comparison to LiPF₆ hydrolysis, LiFSI is less prone to release HF. On the other hand, the equilibrium of the overall hydrolysis shifts to the products side including in the reactive pathway also reaction (13):

$$\begin{eqnarray}&&LiFSI+HF\rightleftarrows HFSI+LiF\left(s\right)\end{eqnarray} \tag{ 13 }$$

This reaction is strongly exergonic with a computed ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E13\right)$ = −90 kJ mol⁻¹ at 298 K and likely drives a shift of the overall equilibrium in favour of the hydrolysis products. It is to be noted that the precipitation of LiF(s) results in the release of another HFSI molecule that can react again, through reaction (12), thus opening the door to a self-feeding process.

Turning to the effect of temperature, the hydrolysis of LiFSI is enhanced at low temperature being favoured the precipitation of solid products.

The summary of the ${{\rm{\Delta }}}_{r}{G}_{T}^{o}$ values for reactions involved in the hydrolysis of LiFSI are reported in the SI, Table SIII.

Li⁺ cation is coordinated by the TFSI⁻ anion by coulombic interactions between the positively charged metal ion and the negatively charged sulphonyl groups: the distances between Li1-O1 and L1-O2 are both 1.60 Å (see Fig. 1). Both trifluoromethyl groups are in the anti configuration. In Fig. 5 the potential energy surface (PES) of the LiTFSI hydrolysis process is displayed.

Zoom In Zoom Out Reset image size

Also in this case, the hydrolysis begins with the acid/base equilibrium between LiTFSI and water:

$$\begin{eqnarray}&&LiTFSI+{H}_{2}O\rightleftarrows HTFSI+LiOH\left(s\right)\end{eqnarray} \tag{ 14 }$$

$$\begin{eqnarray}&&LiTFSI+{H}_{2}O\rightleftarrows HTFSI+LiOH\left(sol\right)\end{eqnarray} \tag{ 15 }$$

Like LiPF₆ and LiFSI, the precipitation of solid lithium hydroxide leads to a Gibbs energy of reaction 14 of 13 kJ mol⁻¹ at 298 K, approximately 150 kJ mol⁻¹ smaller compared to the formation of solvated LiOH(sol).

The precipitation of lithium hydroxide occurs in parallel to the formation of the solvated acid HTFSI, being its structure shown in the SI, Fig. S2. The proton is coordinated by the electron pair on the nitrogen atom.

The release of hydrofluoric acid starting from HTFSI through reaction 16 is an endergonic reaction:

$$\begin{eqnarray}&&HTFSI\,\rightleftarrows \,{C}_{2}{F}_{5}N{O}_{4}{S}_{2}+HF\end{eqnarray} \tag{ 16 }$$

The abstraction of a fluorine atom from −CF₃ is thermodynamically expensive due to the strength of the covalent C–F bond. The structure of the fragment molecule C₂F₅NO₄S₂ released in reaction 16 is shown in the SI, Fig. S2. Being the energy barrier to release HF as high as 63 kJ mol⁻¹ at room temperature in respect to HTFSI, this path is unlikely to occur also in consideration of the additional energy costs due to the possible kinetic activation energies and reaction 15 (i.e. 13 kJ mol⁻¹ at 298 K). Overall HF release by hydrolysis is likely strongly limited in the case of LiTFSI, in comparison to LiFSI and LiPF₆.

Again, the thermodynamics of the overall hydrolysis is shifted towards the products by the strongly exergonic character of reaction 17:

$$\begin{eqnarray}&&LiTFSI+HF\rightleftarrows HTFSI+LiF(s)\end{eqnarray} \tag{ 17 }$$

The computed Gibbs energy of reaction 17 is −91kJ mol⁻¹, leading to a final product blend constituted by LiF(s)+LiOH(s)+HTFSI(sol)+C₂F₅NO₄S₂(sol). Also in this case, the final HTFSI molecule can react again through reactions 16–17 opening also in this case a self-feeding reaction channel.

The summary of the ${{\rm{\Delta }}}_{r}{G}_{T}^{o}$ values for reactions involved in the hydrolysis of LiTFSI are reported in the SI, Table SIV.

Li⁺ is coordinated to FTFSI⁻ anion trough electrostatic interaction between Li cation and partial negative charges on oxygen atoms (see Fig. 1). The distance between Li1 and O4 is 1.86 Å and the distances between Li1 and O2 is 1.88 Å.This ionic couple is asymmetric due to the different substitution of sulphur atoms: S2 atom is linked to two oxygens and one fluorine atom (F1) while S1 is connected totwo oxygens and one CF₃ group. In Fig. 6 the potential energy surface (PES) for the LiFTFSI hydrolysis mechanism is shown.

Zoom In Zoom Out Reset image size

The hydrolysis of LiFTFSI starts from the acid/base equilibrium between LiFTFSI and water, as described in reactions 18 and 19:

$$\begin{eqnarray}&&LiFTFSI+{H}_{2}O\rightleftarrows HFTFSI+LiOH\left(s\right)\end{eqnarray} \tag{ 18 }$$

$$\begin{eqnarray}&&LiFTFSI+{H}_{2}O\rightleftarrows HFTFSI+LiOH\left(sol\right)\end{eqnarray} \tag{ 19 }$$

As expected, the precipitation of lithium hydroxide drives the thermodynamics with a Gibbs reaction energy for 18 of 15 kJ mol⁻¹ at 298 K; on the contrary ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E19\right),$ where LiOH is solvated, is 166 kJ mol⁻¹.

The structure of the acid HFTFSI is shown in the SI, Fig. S3. Following the general scheme already outlined for the FSI and the TFSI anions, HFTFSI can break in fragments to release hydrofluoric acid following the reaction 20:

$$\begin{eqnarray}&&HFTFSI\rightleftarrows C{F}_{3}N{O}_{4}{S}_{2}+HF\end{eqnarray} \tag{ 20 }$$

The Gibbs reaction energy of 20 is 46 kJ mol⁻¹ at 298 K, a value very close to reaction 12. This value is expected since the HF release originates from the cleavage of the S–F bond, like in reaction 12, differently from the more expensive break of the C–F bond in reaction 16 (see above). The other reaction product of 20 is the fragment molecule CF₃NO₄S₂: its molecular structure is shown in the SI, Fig. S3.

Overall, the combination of reaction 18 and 20 leads to an energy barrier of 61 kJ mol⁻¹ (thermodynamic limit), without considering the kinetic activation energies. This threshold matches the similar path of the FSI-hydrolysis (i.e. reactions 10–12), whereas it is 15 kJ mol⁻¹ smaller compared to the combined path of the HF release (14–16) from LiTFSI. Once formed, the HF molecule can easily react with LiFTFSI to give solid LiF(s) following reaction 21:

$$\begin{eqnarray}&&LiFTFSI+HF\rightleftarrows HFTFSI+LiF\left(s\right)\end{eqnarray} \tag{ 21 }$$

This reaction is exergonic and ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\left(E21\right)$ = −89 kJ mol⁻¹ at 298 K and leads to the release of HFTFSI, that can feed back reaction 20 and thus possibly opens the door to a chain reactivity.

The summary of the ${{\rm{\Delta }}}_{r}{G}_{T}^{o}$ values for reactions involved in the hydrolysis of LiFTFSI are reported in the SI, Table SV.

The comparison among the various hydrolysis mechanisms here proposed allow to identify three main common reaction steps:

• 1.  
  The LiOH(s) precipitation;
• 2.  
  The HF release;
• 3.  
  The LiF(s) precipitation.

The comparison of the Gibbs reaction energies at 298 K is shown in the Fig. 7 for these three main reaction steps.

Zoom In Zoom Out Reset image size

All imide-based salts show an energy barrier below 15 kJ mol⁻¹ at room temperature for the formation of LiOH(s). On the other hand, the HF release is strongly endergonic for all lithium imides being the corresponding ${{\rm{\Delta }}}_{r}{G}_{298K}^{o}\gt$ 50 kJ mol⁻¹. The final hydrolysis step, that also feedback the process (see above) is the LiF(s) precipitation that is in all cases strongly exergonic, therefore driving the overall hydrolysis. In this respected LiPF₆ is a compromise, as in reaction 7b the precipitation of LiOH(s) and the HF release are simultaneous: the corresponding Gibbs energy of reaction is larger compared to all the LiOH(s) from imides, whereas it is smaller compared to HF release from imides.

The different thermodynamic landscape of the hydrolysis of imides compared to LiPF₆ necessarily impact the chemical nature of the native passivation layer over electrodes grown in aprotic lithium batteries. In fact, the inorganic phases formed during the hydrolysis can accumulate over the electrodes starting the accumulation of the SEI even in open circuit conditions. Overall, the use of imide-based lithium salts likely leads to LiOH(s)-richer SEI layer over electrodes but likely hinders a massive accumulation of LiF(s) due to the strongly endergonic HF release steps, especially in the case of LiTFSI. A SEI richer in LiOH(s) and poorer in LiF(s) is expected to facilitate the lithium ionic mobility from and into the electrode active material thank to the larger ionic conductivities. ²⁷ On passing it is important to underline that SEI grown on negative electrodes or passivation films deposited on positive electrodes are the results of the complex interplay between the chemically driven salt hydrolysis (that is the goal of this study), the electrochemical degradation of the electrolyte constituents (solvents, additives, salts), and the spontaneous reactivity of soluble and insoluble chemical species (electrolyte components as well as byproducts originating from degradations) with the reduced/oxidized active material in the electrodes. Thus, salt hydrolysis plays a role that must be taken into account in the view of the larger complexity of the entire battery chemistry.