Thermal and hydrolytic decomposition mechanisms of organosilicon electrolytes with enhanced thermal stability for lithium-ion batteries

Article abstract of DOI:10.1149/2.0671709jes

The high flammability and thermal instability of conventional carbonate electrolytes limit the safety and performance of lithiumion batteries (LIBs) and other electrochemical energy storage devices. Organosilicon solvents have shown promise due to their reduced flammability and greater chemical stability at high temperatures. A series of organosilicon electrolytes with different functional substituents were studied to understand the structural origins of this enhanced stability. The thermal and hydrolytic stability of organosilicon and carbonate solvents with LiPF₆ was probed by storage at high temperatures and with added water. Quantitative monitoring of organosilicon and carbonate electrolyte decomposition products over time using NMR spectroscopy revealed mechanisms of degradation and led to the discovery of a key PF₅-complex that forms in organosilicon electrolytes to inhibit further salt breakdown. Increased knowledge of specific structural contributions to electrolyte stability informs the development of future electrolyte solvents to enable the safer operation of high-performing lithium-ion batteries.

Full text of DOI:10.1149/2.0671709jes

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1907
0013-4651/2017/164(9)/A1907/11/$37.00 © The Electrochemical Society
Thermal and Hydrolytic Decomposition Mechanisms of
Organosilicon Electrolytes with Enhanced Thermal Stability for
Lithium-Ion Batteries
Sarah L. Guillot,a,∗ Adrian Pen˜a-Hueso,b Monica L. Usrey,b and Robert J. Hamersa,b,z
aDepartment of Chemistry, University of Wisconsin, Madison, Wisconsin 53706, USA
bSilatronix, Inc., Madison, Wisconsin 53704, USA
The high flammability and thermal instability of conventional carbonate electrolytes limit the safety and performance of lithium-
ion batteries (LIBs) and other electrochemical energy storage devices. Organosilicon solvents have shown promise due to their
reduced flammability and greater chemical stability at high temperatures. A series of organosilicon electrolytes with different
functional substituents were studied to understand the structural origins of this enhanced stability. The thermal and hydrolytic
stability of organosilicon and carbonate solvents with LiPF6 was probed by storage at high temperatures and with added water.
Quantitative monitoring of organosilicon and carbonate electrolyte decomposition products over time using NMR spectroscopy
revealed mechanisms of degradation and led to the discovery of a key PF5-complex that forms in organosilicon electrolytes to inhibit
further salt breakdown. Increased knowledge of specific structural contributions to electrolyte stability informs the development of
future electrolyte solvents to enable the safer operation of high-performing lithium-ion batteries.
© 2017 The Electrochemical Society. [DOI: 10.1149/2.0671709jes] All rights reserved.
Manuscript submitted March 27, 2017; revised manuscript received June 5, 2017. Published July 8, 2017. This was Paper 203
presented at the New Orleans, Louisiana, Meeting of the Society, May 28-June 1, 2017.
1 M LiPF6 up to 100◦C.16 Yet, some degradation was observed arising
from hydrolysis of the Si-O bond. More recent studies showed that
increased resistance to hydrolysis could be achieved by introducing
a short alkyl linker (R3Si(CH2)n-OR) and using LiTFSI instead of
LiPF6.15 The prior studies of organosilicon-based electrolytes sug-
gest that the thermal and hydrolytic stability can be improved through
changes in the molecular structure. However, to achieve rational de-
sign of improved molecules it is necessary to have an improved under-
standing of the fundamental pathways of electrolyte thermal decom-
position. In this study, we have used NMR spectroscopy to identify
the molecular-level mechanisms of decomposition of organosilicon
electrolytes. With a silicon headgroup and three-carbon linker as the
base structure, we investigate the impact of glycol or nitrile functional
tail groups and the extent of fluorination of the silicon headgroup on
thermal hydrolysis. We use the decomposition of a standard carbonate
electrolyte under the same experimental conditions as a benchmark
for these results. Our results demonstrate that dramatic improvements
in thermal and hydrolytic stability can be achieved through specific
modifications of the organosilicon molecular structure.
As the ubiquity of portable electronics and electric vehicles in-
creases, the need for high-capacity energy storage devices grows cor-
respondingly. The high energy density of lithium-ion electrochemistry
have established these batteries at the forefront of meeting these grow-
ing energy storage needs.1–4
Despite their promise as energy storage devices, several challenges
remain to fully optimize the performance and safety of LIBs. The
most commonly used electrolyte is composed of lithium hexafluo-
rophosphate (LiPF6) salt and a blend of carbonate solvents, together
with additional additives to stabilize the electrode surfaces. These
solvents, particularly dimethyl carbonate, have high vapor pressures
and low flash points, presenting a flammability hazard.5,6 Studies
have also shown that cycling at elevated temperatures leads to capac-
ity fade and performance degradation.7–10 The thermal instability of
carbonate- and LiPF6-based electrolytes is a significant obstacle to
achieving longer performance lifetimes and safer battery operation.
Despite these drawbacks, LiPF6 is the industry standard lithium salt
because of its low cost, high solubility in organic solvents, and ab-
sence of corrosion when in contact with aluminum, which is typically
used as the cathode current collector.11
Often, the electrolyte degradation at elevated temperatures is ini-
tiated by protic impurities such as trace amounts of water. The water
may come from incomplete drying of the electrolyte or from atmo-
spheric absorption, due to the highly hygroscopic nature of LiPF6.12,13
It is generally accepted that decomposition of LiPF6 initiates with
the formation of lithium fluoride (LiF) and phosphorus pentafluoride
(PF5) followed rapidly by reaction of PF5 with water to form tri-
fluorophosphine oxide (POF3) and hydrofluoric acid (HF).13 These
(shown in Equation 2) can continue to react in the electrolyte with
trace amounts of water and with solvent molecules in a catalytic cycle
that leads to decomposition of the solvent and salt in amounts far
exceeding the amount of water originally present. It is thus important
to design an electrolyte with resistance to decomposition at elevated
temperatures and in the presence of water to enable electrolyte systems
with improved safety and performance.
Organosilicon solvents have shown promise as enhanced stability,
non-flammable solvents for lithium-ion battery electrolytes. Relative
to carbonates, organosilicon-based electrolytes have a wider electro-
chemical window of stability, higher flash points, reduced gas evo-
lution, and comparable viscosities and conductivities.14–16 Previous
studies showed that organosiloxane (R3Si-OR) electrolytes exhibit
good stability, with only minor decomposition of organosiloxane with
Experimental
Chemicals.—Ethylene carbonate and diethyl carbonate were pur-
chased from SoulBrain and stored in an Ar glove boxe immedi-
ately upon receipt. Lithium hexafluorophosphate was purchased from
BASF and dried in a vacuum oven before use. All organosilicon
compounds were received from Silatronix. Added water was from a
NanoPure system (≥ 18.0 Mꢀ-cm).
Thermal stability and NMR analysis.—Samples for NMR anal-
ysis were prepared in an argon-atmosphere glove box. The samples
were filled in FEP 5 mm tube liners (Wilmad LabGlass). The liners
were then heat-sealed and placed in 9” 5 mm standard NMR tubes.
The tubes were transferred to a vacuum manifold using air-free NMR
tube-Schlenk line adapters, and the NMR tubes were carefully flame-
sealed. Flame-sealing the liners at the same time as the NMR tubes
resulted in various FEP decomposition products visible in the 19F-
NMR spectra. These peaks disappeared with the addition of the liner
heat-sealing step before flame-sealing. Heated samples were stored in
an oven at 100◦C. Storage at 100◦C for more than a week causes some
of the carbonate samples to explode, so all samples were wrapped
with aluminum foil for safety. Samples were analyzed with a Bruker
Avance III 500 MHz spectrometer using a CryoProbe Prodigy probe
or BBFO PLUS probe. All spectra were referenced using a substi-
tution reference standard of 1% TMS in CDCl3 and processed with
∗
Electrochemical Society Student Member.
zE-mail: rjhamers@wisc.edu
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1908
Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
temperature and found that the major products observed at elevated
temperature were also observed in smaller amount after storage at
room temperature. Because at room temperature many of the products
are present at low concentrations that yield signals near the detection
limit of NMR, we present here primarily the results at elevated temper-
ature where concentrations can be more accurately determined. Our
results indicate that the same products are formed at both temperatures,
but the use of elevated temperatures to accelerate the decompositions
leads to concentrations more readily detectable by NMR. Similarly,
while an intact battery would typically have water concentrations <30
ppm, the introduction of higher amounts of water accelerates the re-
actions and makes the products more clearly detectable. Thus, the use
of high temperatures and significant amounts of added water provides
a way to accelerate the decomposition reactions without altering the
fundamental decomposition pathways.
Figure 1. Structures of the four organosilicon solvents tested in this study. The
R tail group substituent was either diethylene glycol monomethyl ether 1 or
nitrile 2–4. The X and Y silicon substituents were methyl or fluorine moieties;
the extent of fluorination was increased in the order 2 (non-fluorinated), 3 and
1 (monofluorinated), and 4 (difluorinated).
Organosilicon electrolytes decomposition.—Figure 1 shows the
organosilicon solvents 1–4 in this study and highlights the differ-
ent structural features under investigation. Each electrolyte was for-
mulated as a single solvent with 1 M LiPF6, and the formation
of thermal degradation products over time was monitored by stor-
age at 100◦C and periodic NMR spectroscopy analyses for ∼20
days. While previous work suggested that adding an alkyl spacer
between the Si atom and the glycol group (as in compound 1) reduced
decomposition,14 the influence of the Li-chelating R tail group and
of fluorination of the Si atom on stability have not been investigated.
To assess the importance of these structural features on the thermal
stability of these electrolytes, we compared electrolyte solvents 1
(3-[diethylene glycol monomethyl ether]propylfluorodimethylsilane)
and 3 (3-cyanopropylfluorodimethylsilane) to probe the differences
between glycol vs nitrile tail functional groups respectively. A com-
parison of cyanopropylsilanes bearing trimethyl silicon (2), monoflu-
orodimethyl silicon (3) and difluoromonomethyl silicon (4) groups
was used to determine the impact of silicon fluorination on thermal
decomposition mechanisms.
MestreNova 9.1.0. The calibration was applied by absolute refer-
encing. The decomposition of LiPF6 was quantified by setting the
−
19F-NMR peak integration of PF6 to 6 and subsequently integrating
−
the peaks of all downstream decomposition products of PF6 (flu-
orophosphates and alkyl fluorophosphates), normalizing each peak
integration by the number of fluorines of the given species. Since the
initial concentration of LiPF6 is 1 molar, the concentration of PF6− in
mol% was calculated by dividing 1 M by (1 + sum of integrations of
all PF6 decomposition products) and multiplying by 100. The concen-
tration of any PF6 decomposition species can be calculated in mol%
by dividing the integration of the given species (normalized by num-
ber of fluorines) by (1+ sum of integrations of all PF6 decomposition
products) and multiplying by 100. This method of calculating con-
−
centration assumes that all PF6 decomposition products are visible
in the solution-phase 19F-NMR. The quantification of decomposition
uses the integration of the NMR peaks, which for the fluorine spec-
trum reliably allows a detection limit of a peak that is at least 0.01%
of the LiPF6 peak. Given that the decomposition species LiF is known
to precipitate out of solution, and decomposition species POF3 and
HF could exist in the system as gases, the calculations provide best
estimates of the concentration in mol%. The quantitation may there-
fore slightly underestimate the extent of decomposition; this would
be the case for all samples studied. The vertical axis expansion of
the 19F NMR spectra figures are all reported relative to the spectrum
normalized to the LiPF6 peak.
The storage of carbonate electrolytes at elevated temperatures re-
sults in the appearance of new peaks in the spectra of 1H, 13C, 19F, and
31P-NMR indicative of electrolyte reactions and the formation of new
products.12,13,21 We used 19F and 31P spectra to identify and quantify
−
all degradation products formed from breakdown of the PF6 an-
ion, since all downstream products of PF6− degradation must contain
one of these nuclei. Each product was identified using its chemical
shift, multiplet splitting pattern, and spin-spin coupling constant, and
by correlating splittings and peak areas between 19F and 31P spectra
across many different samples and experimental conditions.
Impact of organosilicon tail group on thermal stability.—Both
Results and Discussion
glycol and nitrile organosilicons were initially formulated with 1 M
−
The goal of this study was to identify how specific molecular
structural changes influenced the decomposition of organosilicon-
based battery electrolytes under conditions of elevated temperature
and added water. The thermal decomposition of the novel organosil-
icon solvents 1–4 shown in Figure 1 has not been investigated pre-
viously. To provide a benchmark for these samples, we utilized the
same conditions to study the decomposition of the industry-standard
electrolyte solvent system diethyl carbonate (DEC)/ethylene carbon-
LiPF6. PF6 appears in organosilicon electrolytes as a doublet in
19F at −71.9 ppm, with a phosphorus-fluorine coupling constant of
710 Hz, and a septet in 31P at −144.1 ppm (JP-F = 710 Hz). Prior
to any thermal treatment and/or intentional addition of water, each
sample was analyzed as-formulated. Table I summarizes the identify-
ing features of each fluorine-containing species in the as-formulated
and thermally treated sample. Note that while fluorophosphates are
reported here as anions, they are likely involved in an equilibrium be-
tween the protonated and deprotonated forms.13 Our results show that
for all common degradation species, the chemical shift of the same
species changes only minimally (<1.5 ppm difference) between the
different organosilicon electrolyte solutions 1–4.
1
ate (EC) (3:7 by volume). We used H-, 13C-, 19F-, and 31P-NMR
spectroscopy to analyze the solution-phase decomposition of carbon-
ate and organosilicon samples. We stored electrolyte samples at 100◦C
for up to 20 days and tested the effects adding 1000 ppm (v/v) excess
water, with control samples having no added water. While the 100◦C
and 1000 ppm water concentrations are well outside normal perfor-
mance limits for conventional lithium ion batteries, these conditions
are relevant to understanding the overall safety profile of electrolytes
and are also relevant to potential use of organosilicon electrolytes in
a wider range of energy storage devices includings high-temperature
Li-CFx primary batteries,17 supercapacitors, and secondary batteries
using anodes such as lithium titanate that do not form solid-electrolyte
interphase layers.18–20 We also performed similar experiments at room
The as-formulated electrolyte solution of 1 has three small 19F
doublets arising from initial decomposition products. The first dou-
blet at −83.2 ppm (JP-F = 928 Hz) has a corresponding 31P triplet with
the same coupling constant. Based on the triplet phosphorus splitting
(indicating a bifluorinated species) and chemical shifts previous re-
ported for this species,12,13,22 we attribute these peaks to PO2F2−. The
third doublet at −89.0 ppm correlates to a quartet in 31P, and we
therefore assign this to the trifluorophosphine oxide POF3. The as-
formulated electrolyte solution of 3 also had two sets of doublets; we
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1909
Table I. NMR identifying features for all fluorine-containing organosilicon degradation products: chemical shifts, coupling constants, and splitting
patterns.
Species
19F ppm, mult (Hz)
31P ppm, mult (Hz)
1
3
−160.6, m (JH-F = 7.3)
−161.3, m (JH-F = 6.3)
-
-
-
4 −
−135.5, m (JH-F = 6.0)
PF6
−72, d (JP-F = 710)
−144, sept (JP-F = 710)
POF3−
−89, d (JP-F = 1066)
−35, quart (JP-F = 1066)
PO2F2
−84, d (JP-F = 930–960)
−20, t (JP-F = 930–960)
PO3F−2
−76, d (JP-F = 916)
−8.29, d (JP-F = 916)
Si(Me3)F
Si(Me2)F2
−156.7, m (JH-F = 7.5)
-
-
-
-
-
-
-
-
-
−130.7, sept (JH-F = 6.3)
−59.3, dd (JP-Feq = 755, JFax-Feq = 55)
−85.5, d (JP-F = 970)
PF5 complex 1: PO2F2→PF5 (equatorial)
PF5 complex 1: PO2F2→PF5
PF5 complex 2: B2→PF5 (equatorial)1
PF5 complex 3: B3→PF5 (equatorial)1
PF5 complex 4: B4→PF5 (equatorial)1
PF5 complex 5: B5→PF5 (equatorial)1
HF
−63, dd (JP-Feq = 778, JFax-Feq = 58)
−60.0, dd (JP-Feq = 747, JFax-Feq = 56)
−60.4, dd (JP-Feq = 742, JFax-Feq = 57)
−60.2, dd (JP-Feq = 749, JFax-Feq = 27)
−182—189, s
1B represents any unknown Lewis base
attribute these features to PO2F2− and POF3 respectively, due to their
correspondence of chemical shift and coupling with the peaks in elec-
trolyte 1. The presence of these fluorophosphates indicates that trace
amounts of water were present in the as-formulated electrolyte and
induced decomposition of minor amounts (<0.4%) of the PF6− before
any thermal treatment or intentional addition of water. The presence
of trace amounts of water in electrolytes in unavoidable and has also
been reported in previous studies of carbonate electrolytes.13,23 The
source of this water may come from both the solvent and the salt,
as LiPF6 is known to be highly hygroscopic and the solvent requires
drying after synthesis.
where Ai is the NMR peak area corresponding to species i and ni
is the number of fluorine atoms of species i. Species p are PF6
and all its fluorophos−phate decomposition products; since the initial
concentration of PF6 is 1 molar, dividing the normalized area of
any species i by this sum gives the concentration of species i in
−
Figure 2 shows the 19F-NMR spectra of electrolytes 1 and 3 after
the addition of 1000 ppm (v/v) water and storage at 100◦C for thre−e
weeks. Electrolyte 1 (Figure 2a) shows an increase of the PO2F2
peak and simultaneous disapp−earance of the POF3 peak, indicating the
successive hydrolysis of PF6 degradation products. The formatio−n
of these fluorophosphates indicates progressive hydrolysis of PF6
via the Reactions 1–3. LiPF6 dissociates to form the reactive Lewis
acid PF5. PF5 reacts with water to form POF3, which reacts further
−
with water to form the fluorophosphate PO2F2 This mechanism is
.
consistent with the known hydrolysis mechanisms of PF6− with water
in carbonate solvents.12,13,23,24 In contra−st, electrolyte 3 (Figure 2b)
shows only minimal growth of the PF6 decomposition peaks, and
even after three weeks of thermal treatment with excess added water,
−
there is still POF3 present that has not been hydrolyzed to PO2F2
.
→
LiPF6 ← LiF + PF5
[1]
PF5 + H2O → POF3 + 2HF
[2]
[3]
POF3 + H2O → HPO2F2 + HF
−
After identifying all PF6 degradation products, we compared the
exte−nt of PF6− decomposition between 1 and 3. The decomposition of
PF6 in the 19F spectra was quantified by integrating the NMR peak
−
areas of all the PF6 decomposition products. To interpret the data
in terms of concentrations of different chemical species, the peak of
each species was normalized to the number of F atoms in the species,
via the equation:
Figure 2. 19F-NMR of decomposition species formed in a) organosilicon gly-
col electrolyte 1, stored at 100◦C for 19 days after adding 1000 ppm of water
and b) organosilicon nitrile electrolyte 3, stored at 100◦C for 22 days after
adding 1000 ppm of water.
A
i
n
i
Concentration of species i (mol/L) = Xi =
[4]
ꢀ
A
n
p
p
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1910
Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
Figure 3. Decomposition of LiPF6 in organosilicon electrolytes stored at
100◦C with (a) no added water and (b) 1000 ppm added water. Organosil-
icon glycol electrolytes (1 + 1 M LiPF6) are light blue square markers (ꢀ)
and organosilicon nitrile electrolytes (3 + 1 M LiPF6) are indigo triangle
markers (ꢁ). Quantification of LiPF6 decomposition was calculated by 100 x
Equation 1.
mole/L. Given the small headspace of the sealed NMR sample tube,
we assume that the loss of fluorophosphate products to the gas phase
is negligible relative to the concentration remaining in solution. There
is no evidence of the complete hydrolysis of fluorophosphates to
form phosphoric acid/phosphate ion in the 31P-NMR (−singlet near 0
ppm),25,26 therefore the complete quantification of PF6 degradation
can be achieved through the 19F spectra.
Figure 4. NMR spectra of 1 + 1 M LiPF6: (a) 19F spectrum immediately
after addition of 1000 ppm water (v/v) (top) and the same sample after storage
at 100◦C for 15 days (bottom). (b) 13C spectrum of the same sample after 19
days 100◦C storage. Thermal decomposition products are highlighted with red
arrows. (c) 13C spectrum of sample after 20 days at room temperature showing
that the same products form as in (b), but at lower concentrations, leading to
NMR signals close to the signal-to-noise limit of the measurement.
−
Figure 3 shows the concentration of PF6 in solvents 1 and 3
as a function of time while stored at 100◦C; data are shown for as-
formulated samples (with no water added, Figure 3a) and samples with
1000 ppm (v/v) added water (Figure 3b). Multiple trials are shown
to demonstrate the reproducibility of thermal decomposition for each
sample. For both organosilicon electrolytes 1 and 3 in Figure 3a,
PF6− shows exceptional thermal stability for the as-formulated sample
(<5% decomposition). For the sample with water added (Figure 3b),
chelate lithium ions, enhancing electrolyte conductivity.27,28 Further-
more, the flammability of these solvents can be tuned by the number
of alkyl ether groups in the glycol chain, with an increasing chain
length corresponding to a higher solvent flashpoint.28 Zhang et al.
showed that organosilicon glycols also have enhanced conductivity
and reduced flammability compared with carbonate solvents.14,15
Figure 4a shows the 19F-NMR spectra of solvent 1 + 1 M LiPF6
+ 1000 ppm (v/v) water, in the case immediately after water addition
and after the samples were stored at 100◦C for two weeks. Figure
4b shows the corresponding 13C spectrum for the latter sample. In
Figure 4a, the as-formulated sample has a primary peak at −160.6
ppm. Based on the silicon satellites and a nonet fluorine-hydrogen
splitting pattern this peak is assigned to solvent 1. After two weeks
of high temperature storage a new septet appears at −130.7 ppm;
this splitting indicates a species coupled to six hydrogens. From the
splitting pattern in 19F and the corresponding triplet that appears con-
currently at −3.7 ppm in the 13C spectrum (Figure 4b), this species is
assigned as difluorodimethylsilane. As the only source of silicon is 1,
−
organosilicon glycol 1 shows 5–10% decomposition of PF6 during
the first 330 hours of storage at 100◦C. In organosilicon nitrile 3, the
decomposition of PF6− is−significantly less, with a concentration loss
of less than 4%. The PF6 decomposition in organosilicon glycol 1
is therefore more sensitive to water concentration than organosilicon
nitrile 3.
−
In addition to hydrolysis of the PF6 anion, we investigated the
thermal and hydrolytic reaction mechanisms of the solvents them-
selves. Every solution-phase species resulting from solvent reactions
1
can be observed and identified with a combination of H, 19F, and
13C-NMR spectroscopy.
Glycol tail group decomposition mechanisms.—Molecules with
glycol functionalities have been investigated previously as battery
electrolyte solvents due to the ability of the oxygen lone pairs to
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1911
Scheme 1. Proposed mechanism of thermal decomposition of 1 to form di-
fluorodimethylsilane.
formation of difluorodimethylsilane must result from the cleavage
of the silicon-methylene bond in 1 and fluorination of the silicon
fragment. Pentacoordinated silicon species as intermediates in sub-
stitution reactions are well-documented in the literature, particularly
when a source of fluoride is present, due to the Lewis acidic na-
ture of silicon.29,30 Based on this precedent, we propose that difluo-
rodimethylsilane forms via a hypervalent silicon intermediate as de-
picted in Scheme 1. Due to the structural similarity of the by-product
(diethylene glycol propyl methyl ether) to 1, this product would be ex-
pected to appear in the 13C spectra as a series of peaks with chemical
shifts close to those of the main solvent peaks. Indeed, the corre-
sponding 13C NMR spectrum in Figure 4a shows several small peaks
(highlighted by arrows) whose chemical shifts are similar to those ex-
pected for diethylene glycol propyl methyl ether. However, since these
peaks would be similar for most alkyl glycol ether species,31 specific
assignments cannot be determined with certainty. Furthermore, we an-
ticipate that any ethylene glycol propyl methyl ether might continue
to break down further. This is supported by the 13C spectrum (Figure
4b) showing the presence of propene. We therefore simply conclude
that several different alkyl glycol products likely form as breakdown
products of 1.
Scheme 2. Proposed mechanism of thermal decomposition of 1 to form 1,4-
dioxane. B represents conjugate base of H+.
equal (indicating 50% decomposition of the electrolyte solvent), the
single 1,4-dioxane peak represents four equivalent carbons. While
these 13C spectra are not quantitative, the relative concentration of
1,4-dioxane can be estimated by dividing the peak intensity by four.
Even considering this qualitative normalization, it is clear that the
concentration of 1,4-dioxane exceeds that of any other decomposition
product in the 13C spectrum. We hypothesize that either the forma-
tion of 1,4-dioxane from 1 proceeds via several different mechanisms
or that the other products formed with 1,4-dioxane are unstable and
break down further. For example, in Figure 4b, three small septets
appear immediately downfield of the peak for 1 in the 13C spec-
trum. These peaks are consistent with any monofluorodimethyl alkyl
silane product similar to 1, such as the silane by-product formed in
Scheme 2.
Figure 4c shows similar data to 4b, except that the experiment
was performed at room temperature for 21 days with 500 ppm added
water. In this case the same products are observed as in Fig. 4b but
the concentrations are lower, leading to signal levels that are close to
the signal-to-noise limit of the NMR instrument.
The above experiments identified two main decomposition mech-
anisms for the thermal degradation of 1. The primary degradation
pathway is acid-catalyzed cleavage of the glycol tail group to produce
1,4-dioxane, likely by the pathway depicted in Scheme 2. A second,
minor pathway arises from cleavage of the silicon-fluorine bond, pro-
ducing difluorodimethylsilane by the pathway depicted in Scheme 1.
These results suggest that the thermal stability of organosilicons can
be improved by altering the nature of the tail group.
Figure 4b shows the corresponding 13C spectrum for the sample of
1 with added water and then stored at 100◦C; the 13C spectrum shows
a significant new peak at 67.1 ppm. Based on known chemical shifts in
the literature,32 we assign this peak to 1,4-dioxane. A known method
of 1,4-dioxane formation is through acid-catalyzed intramolecular
cyclodehydration.33 Furthermore, glycol species similar to 1 have
been shown to yield 1,4-dioxane in the presence of an acid catalyst.33
The mechanism in that study produces methanol as a side product
when a methyl ether is the starting reagent. However, we find no
evidence of methanol (chemical shift expected at 50 ppm)32 in the
13C spectra of 1. In Scheme 2, we propose that the formation of
1,4-dioxane proceeds via acid-catalyzed intramolecular cyclization of
the glycol tail group, which would transfer the methyl ether to the
silane to give fluoro(3-methoxypropyl)dimethylsilane in addition to
the dioxane. This product results from transfer of the terminal methyl
group to the silane oxygen closest to the silicon.
Nitrile tail group decomposition mechanisms.—Organic nitrile
molecules have previously been investigated for use in battery elec-
trolytes because of their high oxidative stability due to the strong
electron-withdrawing nature of the nitrile substituent.34–36 The elec-
tronegative nitrile group also promotes solvation of the lithium ion.35,37
These promising properties led us to investigate the thermal stability
While the intensity of the 1,4-dioxane peak in Figure 4b appears
to indicate that the concentrations of 1,4-dioxane and 1 are nearly
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1912
Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
Scheme 3. Hydrolysis of 3 to form an organosilicon amide.
that storage at high temperatures leads to the appearance of at least
one new species, which appears as a triplet at 2.72 ppm and peaks
that appear to correspond to a doublet with individual peaks at 8.3
and 8.6 ppm. This peak separation (∼150 Hz) is greater than 1H-1H
coupling constants, which fall within the range of < 1–45 Hz.38,39
Although 19F-1H or 31P-1H coupling constants could fall in this range
40
(typically 115–140 Hz for JP-H and up to 130 Hz for vicinal JF-H41),
the absence of a corresponding peak in the 19F or 31P spectra with
the same coupling constant excludes this possibility. Consequently,
we conclude that the ‘doublet’ is likely a set of two singlets. Primary
amide hydrogens often appear as a set of two singlets due to the
partial double bond nature of the carbon-nitrogen bond, making the
primary hydrogens chemically inequivalent.42 We therefore assign the
two singlets to the hydrogens of a primary amide, and the triplet at 3.2
ppm to a methylene group primary to carbonyl of an amide (Figure
5b inset). The absence of other visible 1H peaks associated with this
amide suggests that they are too similar to the corresponding nitrile
structure of 3, evidence that the amide forms as a direct hydrolysis
of the nitrile group of 3 as shown in Scheme 3. The concentration of
the organosilicon amide is too low to be visible in the 13C spectra.
As Figure 5a shows, no additional fluorine species are seen in 19F
spectrum; we hypothesize that the fluorine peaks of 3 and its amide
analog are too close to resolve due to their structural similarities. De-
spite the small amount of amide that is formed, Scheme 3 is important
in understanding electrolyte decomposition processes in electrolyte 3.
This reaction provides an alternative mechanism for the consumptio−n
of water, which may explain the decreased decomposition of PF6
seen in Figure 3 for the nitrile organosilicon electrolyte. Similar ex-
periments show that 1,4-dioxane and difluorodimethylsilane are also
observed at room temperature but at lower concentrations; this indi-
cates that the pathways outline in Scheme 3 and Scheme 4 are valid at
both temperatures, and the primary effect of elevated temperature is
to accelerates the rates and thereby makes the products more readily
detectable.
In summary, by changing the glycol tail group to the nitrile, the
cleavage of the silicon-carbon bond is eliminated. The nitrile tail
group reacts with water to form the corresponding amide, resulting
Figure 5. NMR spectra of 3 + 1 M LiPF6: (a) In the 19F spectrum, no thermal
decomposition products are seen when comparing the solution immediately
after adding 1000 ppm water (top) the same solution after storage at 100◦C for
22 days (bottom). (b) In the 1H spectrum, Inset shows vertical and horizontal
expansion of peaks in 2.5–3.5 ppm and 7.5–9.0 ppm regions, evidence of
hydrolysis of 3 to organosilicon amide after storage at 100◦C for 10 days.
−
in less PF6 hydrolysis than in the glycol electrolyte. These results
show that the nature of the tail group of these organosilicon solvents
significantly impacts the extent and mechanisms of electrolyte decom-
position, and suggest that further variation of the solvent structure can
effect additional tuning of the thermal properties.
Impact of organosilicon fluorination on thermal stability.—Fluo-
rination of organic molecules such as carbonates, esters, and ethers
have been shown to stabilize the highest-occupied molecular orbital
(HOMO) levels of these compounds, resulting in greater stability
against oxidation.43,44
The effect of fluorination of the silicon atom on solvent stability
was investigated by comparing the stability of the non-fluorinated sol-
vent 2 with the mono-fluorinated compound 3 and the difluorinated
solvent 4. The fluorinated electrolytes 3 and 4 both show the same fluo-
rophosphate decomposition products observed for the non-fluorinated
molecule (3, Figure 2b). Figure 6 shows the mole % PF6 remaining
in all three nitrile-based electrolytes.3–5 All three nitrile-based elec-
trolytes show <5% PF6 decomposition even after more than 400 hours
stored at 100◦C. Furthermore, the comparison of Figures 6a and 6b
show that electrolyte systems with solvents 2–4 did not show any
significant change upon intentional addition of −1000 ppm (v/v) water.
Thus, we conclude that the protection of PF6 against degradation
of organosilicon electrolytes with nitrile tail groups, beginning with a
comparison of 3 with the analogous glycol molecule 1.
Figure 5a shows the 19F-NMR spectrum of the as-formulated sam-
ple of 3 + 1 M LiPF6 and the same sample after storage at 100◦C for
18 days; Figure 5b shows the corresponding 1H-NMR spectra. A com-
parison of the as-formulated and heat-treated 19F spectra in Figure 5a
shows that storage at elevated temperatures does not lead to any new
peaks. Even when magnified, the spectrum does not show evidence for
any new products. The absence of detectable difluorodimethylsilane
formation shows that the organosilicon nitrile 3 resists the silicon-
methylene bond cleavage suffered by glycol groups (Scheme 2 ).
When the Li-chelating group is changed from glycol in 1 to nitrile in
3, the silicon-methylene bond therefore becomes less susceptible to
cleavage at high temperatures.
While the 19F spectra show that 3 appears to be completely resis-
tant to degradation, the corresponding 1H spectra in Figure 5b show
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1913
Scheme 4. Proposed mechanism of thermal fluorination of 2 via pentacoordinated siliconate intermediates.
via thermal hydrolysis (Equation 4–6) occurs for organosilicon nitrile
solvents regardless of the extent of silicon fluorination.
the absence of detectable silicon-methylene bond cleavage. Figure 8
shows the same chemical shift region for electrolyte 2. Since 2 is
non-fluorinated, there should be no peaks in this region. However,
even in the as-formulated sample in Figure 8a, there is a significant
nonet peak at −161.3 ppm and a small peak at −156.7 ppm. The
peak at −161.3 has silicon satellites and the identical line shape and
chemical shift as 3 and was therefore assigned as such, indicating that
the mono-fluorinated analog of 2 (which is species 3) is an impurity
found at low levels in the pristine solution of 2. The small peak at
Figure 7 shows the 19F-NMR spectra of difluorinated electrolyte
4 before and after storage at 100◦C for 19 days. These spectra are
focused on the organosilane region of the chemical shift range. The
main peak is a sextet at −135.5 ppm (Figure 7a inset) with silicon
satellites and split by the methyl and methylene groups adjacent to
the silicon. This peak has the expected splitting and chemical shift
for the solvent 4 fluorine peak. Identical to 3 in Figure 5, no new
peaks are observed when 4 is stored at high temperature, indicating
Figure 6. Decomposition of LiPF6 in organosilicon electrolytes stored at
100◦C with (a) no added water and (b) 1000 ppm added water. Non-fluorinated
organosilicon nitrile electrolytes (2 + 1 M LiPF6) are red triangle markers (ꢂ),
monofluorinated organosilicon nitrile electrolytes (3 + 1 M LiPF6) are indigo
triangle markers (ꢁ), and difluorinated organosilicon nitrile electrolytes (4 +
1 M LiPF6) are yellow triangle markers (ꢃ). Quantification of LiPF6 decom-
position was calculated by 100 x Equation 1.
Figure 7. 19F-NMR spectra (magnified by 5000x) of (a) an as-formulated
solution of 4 + 1 M LiPF6 and (b) the same solution after storage at 100◦C for
20 days. Inset in (a) shows expansion of the main peak, representative of both
spectra.
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1914
Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
Figure 9. 19F-NMR spectrum of PF5 complexes formed in (a) 3 + 1 M LiPF6
stored at 100◦C for 22 days and (b) 1 + 1 M LiPF6 stored at 100◦C for 15
Figure 8. 19F-NMR spectra of (a) an as-formulated solution of 2 + 1 M LiPF6
and (b) the same solution after storage at 100◦C for 10 days. The inset in (a)
shows the expansion of the nonet peak at −161 ppm, repesentative of both
spectra. (c) shows fluorotrimethylsilane concentration vs. Time for samples
stored at 100◦C and samples stored at room temperature.
days. The B groups represent different unidentified Lewis base species.
the silicon-fluorine bond (∼135 kcal/mol) relative to silicon-carbon
(∼70 kcal/mol).46
From the investigation of varying fluorination of the organosili-
con solvents, we observe the greater thermal stability of fluorinated
electrolytes 3 and 4 over the non-fluorinated organosilicon electrolyte
2. There is no apparent thermal stability difference in the mono-
fluorinated and di-fluorinated species under our experimental condi-
tions.
−156.7 ppm in Figure 8a could not initially be identified, but after the
growth of this species at high temperatures (Figure 8b) it is clearly
split into a multiplet by nine coupled hydrogens and has silicon satel-
lites. From the chemical shift and multiplicity we assign this species
as fluorotrimethylsilane, an assignment that is confirmed by NMR
reports in the literature.45 In addition to an increase in the presence of
fluorotrimethylsilane, storage at 100◦C results in an increase in fluo-
rination of 2 (as evidenced by growth of the peak assigned to 3) and
appearance of a small amount of dimethyldifluorosilane at −130.5
ppm. Formation of the trimethylfluorosilane molecule can also be de-
tected at room temperature, although at lower concentrations. Figure
8c shows that at room temperature and at 100◦C, the concentration
of fluorotrimethylsilane grows linearly with time over the duration of
these experiments.
The presence of fluorotrimethylsilane indicates cleavage of the
silicon-methylene bond in 2, while formation of 3 from the fluori-
nation of 2 must result from breaking a silicon-methyl bond. Diflu-
orodimethylsilane is a ‘secondary’ decomposition product and could
occur either through fluorination of fluorotrimethylsilane or fluori-
nation of solvent 3 resulting in silicon-methylene bond cleavage.
However, the absence of difluorodimethylsilane in the samples of
3 suggests that it forms solely from fluorotrimethylsilane. Scheme 4
shows the proposed reactions to form the three thermal decomposition
species of 3, proceeding via penta-coordinated siliconate intermedi-
ates as proposed in Scheme 1. The driving force for the formation of
fluorinated organosilicons might be the greatly increased stability of
Stabilization of PF5 in organosilicon electrolytes.—In addition to
the peaks assigned to the decomposition species described above, in
every organosilicon solvent stored at 100◦C, a series of doublet of
doublets appears in the chemical shift range −58 ppm to −65 ppm.
Figure 9 shows the 19F-NMR spectrum of 3 + 1 M LiPF6 that has
been stored at 100◦C for 22 days. Two sets of doublet of doublets ap-
pear with large coupling constants of 750–780 Hz and small coupling
constants of 55–59 Hz. The values of the large coupling constants
are in a range characteristic of o−ctahedral fluorine-phosphorus species
(the coupling constant of PF6 is 710 Hz).47 Previous studies on
PF5 complexed to a sixth electron donor ligand has shown the same
splitting with similar coupling constants and chemical shifts.22 The
doublet of doublet lineshape comes from the equatorial fluorines, with
a large phosphorus-equatorial fluorines splitting and a smaller split-
ting between the inequivalent axial and equatorial fluorines. The axial
fluorines would have high multiplicity (doublet of quintets) and are
often too far into the baseline to be observed; only one set of doublet of
quintets can be seen in the spectrum in Figure 9. Based on the similar
species reported in literature, we assign these peaks as two different
PF5 complexes. PF5 complexes were observed in all organosilicon
electrolytes 1–4 when stored at 100◦C for over 10 days, regardless
of added water. Organosilicon nitriles 2–4 have two species as seen
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1915
Scheme 5. Reversible sequestration of PF5 by a Lewis base (B).
in Figure 9, while organosilicon glycol 1 forms at least four dis-
tinct complexes. The chemical shifts and coupling constants of these
species, labeled ‘PF5 Complex X’, are described in Table I. In the ni-
triles 2–4, we observe formation of a broadened doublet (−85.2 ppm)
−
near the PO2F2 doublet (−83.6 ppm) at precisely the same rate
and quantity as the PF5 Complex 1 peaks at −59.3 ppm (S7). Based
on this species, we assigned PF5 Complex 1 to PF5-PO2F2−. The
PF5-PO(OR)F2 complex has been proposed as a reactive int−ermediate
in carbonate electrolytes but never observed.12 PF5-PO2F2 has also
been observed in the synthesis of POF4− in CHF3 solution, but only at
temperatures of −140◦C.22 In the case of organosilicon–based elec-
trolytes, we hypothesize that the solvent stabilizes the PF5 complexes,
which were visible in solution throughout the length of this study.
Computational studies by Tasaki et al. indicated that the nature of
the electrolyte solvent can greatly impact PF5 complex formation and
stabilization, supporting our hypothesis.48 The identity of the electron
donor for the second PF5 complex in Figure 11 (‘B’) is still unknown.
A number of electron-donating species that occur naturally in the
electrolyte could form complexes with PF5 including H2O, ethers,
fluorophosphates, and nitrogen-containing species.21,22,49,50
Figure 10. (a) 19F-NMR and (b) 31P-NMR spectra of decomposition species
formed in carbonate electrolytes stored at 100◦C.
The strong Lewis acid PF5 has been hypothesized as the
main initiator of electrolyte degradation (polymerization, decom-
position, and fluorination) (Equation 1).24,51–53 Based on this the-
ory, studies have investigated the addition of Lewis bases such
as tris(2,2,2-trifluoroethyl)phosphite, pyridine, hexamethylphospho-
ramide (HMPA), hexamethoxycyclotriphosphazene (HMOPA), or
dimethyl acetamide with the intention of inhibiting electrolyte degra-
dation. The sequestration of PF5 with these bases proved to be success-
ful at significantly enhancing electrolyte thermal stability.16,21,51,54 Fol-
lowing the same mechanism postulated by Li et al. for the reversible se-
questration of PF5 by pyridine, HMPA, and HMOPA, Scheme 5 shows
our proposed mechanism of the reversible formation of a generic PF5
complex in organosilicon electrolytes. The spontaneous sequestration
of PF5 in o−rganosilicons appear to inhibit the reactivity of PF5 and
protect PF6 fr−om further degradation, resulting in the high thermal
stability of PF6 in organosilicon electrolytes as shown by Figures 3
and 6.
carbonate electrolyte samples after storage at 100◦C for 17 days. The
identifying details of the carbonate decomposition species observed
in the 19F and 31P-NMR are described in Table II, including chemical
shifts, coupling constants, and splitting patterns.
After the high-temperature storage, in addition to an increase in
PO2F2− , a similar 19F doublet forms at −85.9 ppm (JP-F = 1006
Hz). The corresponding 31P peaks (identified by the corresponding
coupling constant) are a triplet of triplets with splitting from two
fluorines but also coupling with hydrogens (JP-H = 9.7 Hz), consis-
tent with the alkyl fluorophosphate PO(OCH2CH3)F2. The remaining
three doublets in the 19F spectra are also doublets in 31P, indicat-
ing monofluorinated species. The 31P doublet with no hydrogen cou-
pling is assigned to PO3F2−, while we assign the double of triplets to
PO2(OCH2CH3)F− and the doublet of quintets to PO(OCH2CH3)2F.
The corresponding peaks of these fluorophosphates in 19F are identi-
fied by corresponding phosphorus-fluorine coupling constants. While
it is difficult to assign the peaks to all the different possible fluo-
Carbonate electrolytes decomposition.—As these organosilicon
electrolytes are novel, we needed to benchmark our thermal stability
studies against a well-known electrolyte system. Towards this end we
investigated the thermal stability of the binary carbonate electrolyte
EC/DEC (3:7, v:v) 1 M LiPF6 during storage of NMR samples at
100◦C. The high degree of decomposition in carbonate electrolytes
was clear because many carbonate sample NMR tubes broke when
stored at 100◦C, due to increased internal pressure in the sealed tubes
from the gaseous decomposition products being generated. Figure
10 shows the 19F and 31P NMR spectra of some of the unbroken
rophosphates with complete confidence, the correlation of 19F and 31
P
coupling constants and the splitting patterns provide strong evidence
for the assignments stated above. Furthermore, the chemical shifts
and coupling constants of PF6−, PO2F2−, PO3F2−, PO(OCH2CH3)2F,
PO(OCH2CH3)F2 have been reported previously and our observations
agree with these known and accepted values, accounting for small
chemical shift and coupling constant variation due to differences in
carbonate composition.12,13,24,55
−
The successive formation of fluorophosphates PO2F2 and
PO3F2− is consistent with the known and accepted hydrolysis
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

A1916
Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
Table II. NMR identifying features for all fluorine-containing EC/DEC degradation products: chemical shifts, coupling constants, and splitting
patterns.
Species in EC/DEC
19F ppm, mult (Hz)
31P ppm, mult (Hz)
−
PF6
−74.1, d (JP-F = 708)
−89.5, d (JP-F = 1067)
−84.8, d (JP-F = 946)
−77.0, d (JP-F = 924)
−85.9, d (JP-F = 1006)
−83.1, d (JP-F = 962)
−80.7, d (JP-F = 941)
−189, s
−144.3, sept (JP-F = 708)
POF3−
-
PO2F2
−19.3, t (JP-F = 947)
PO3F−2
−9.0, d (JP-F = 924)
PO(OCH2CH3)F2
PO(OCH2CH3)2F
PO2(OCH2CH3)F−
HF
−20.6, tt (JP-F = 1006)
−10.1, dp (JP-F = 961)
−9.6, dt (JP-F = 941)
-
-
-
CH3CH2F
−212.4, tq (JH-F = 47.1, 26.5)
−226.3, tt (JH-F = 47.3, 29.3)
RCH2CH2F1
1R is an unknown fragment
−
mechanism of PF6 with water (Equation 4–6), as seen in the
organosilicon electrolyte 1–4. The alkyl fluorophosphates that ap-
pear (PO(OCH2CH3)2F, PO(OCH2CH3)F2, PO2(OCH2CH3)F− are
evidence of the reaction of fluorophosphates with the carbonate sol-
vents, as previously reported.12,24,55 The mechanism proposed by Cam-
pion et al. for the reaction of a linear carbonate with POF3 produces RF
and CO2, where R is the alkyl group on the carbonate.12 Both CO2 and
CH3CH2F were seen in the decomposition of the binary EC/DEC elec-
trolyte (S1). CH3CH2F is also seen in the 19F spectrum in Figure 10a
at −212 ppm. These species suggest the decomposition of carbonates
in our system follows the mechanisms shown in Scheme 6. Another
alkyl fluoride appears at −226 ppm in the 19F spectrum, with the triplet
of quartet splitting pattern consistent with an ethylene fluoride sub-
stituent (RCH2CH2F). Wilken et al. observed (PO(F2)(OCH2CH2F))
as a decomposition product resulting from the reaction of ethylene
carbonate with POF3. While the RCH2CH2F peak in Figure 10 would
be consistent with this species, the lack of a corresponding doublet for
the phosphorus fluorides (POF2) suggests that if this pathway occurs it
is not a major method of decomposition under our experimental condi-
tions. The identity of RCH2CH2F and its exact formation mechanism
from carbonate breakdown remains unknown. The fluorophosphates
and alkyl fluoride decomposition species identified confirm that these
carbonate samples decompose via the mechanisms proposed in previ-
ous studies.23,55
decomposed after 450 hours of elevated temperature storage. When
water was added, the decomposition increased to 40% (Figure 11).
These results indicate that water contributes to carbonate electrolyte
decomposition, as further indicated by the observed fluorophosphate
decomposition species and reaction mechanism described above. The
carbonate sample with no water added most likely contains trace water
impurities from incomplete drying of the electrolyte. Figure 11 also
shows a comparison of the decomposition of LiPF6 in carbonates and
organosilicons. In both electrolytes 1 and 3, the decomposition of
LiPF6 is at most 15% and seems to reach a plateau over time. By
contrast, in carbonate electrolytes, LiPF6 rapidly decomposes by 30–
40%, and appears to continue to degrade, although we were unable to
test the degradation for longer experiment times due to the carbonate
The role of water in the decomposition of carbonate electrolytes
was investigated by preparing samples with 1000 ppm of added water
and control samples with no water added. Both samples were stored
at 100◦C. There were no differences in the types of decomposition
species produced when excess water was added. However, extent of
decomposition was increased slightly with the addition of water. In the
sample with no added water, 30% of the initial concentration of LiPF6
Figure 11. Decomposition of LiPF6 in carbonate and organosilicon elec-
trolytes at 100◦C with (a) no added water and (b) 1000 ppm added water.
Carbonate electrolytes (3:7 EC:DEC + 1 M LiPF6) are circle markers (●),
organosilicon glycol electrolytes (1 + 1 M LiPF6) are square markers (ꢀ) and
organosilicon nitrile electrolytes (3 + 1 M LiPF6) are triangle markers (ꢁ).
Quantification of LiPF6 decomposition was calculated by 100 x Equation 1.
Scheme 6. Proposed mechanism for the reaction of diethyl carbonate with
POF3 to produce the observed alkyl fluorophosphates PO(OCH2CH3)2F,
PO(OCH2CH3)F2, PO2(OCH2CH3)F− .
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 164 (9) A1907-A1917 (2017)
A1917
samples breaking. These data show the significant increase in thermal
stability of the lithium salt provided by organosilicon solvents over
carbonates.
10. P. Ramadass, B. Haran, R. White, and B. N. Popov, J. Power Sources, 112 (2), 606
(2002).
11. S. S. Zhang and T. R. Jow, J. Power Sources, 109 (2), 458 (2002).
12. C. L. Campion, W. Li, and B. L. Lucht, J. Electrochem. Soc., 152 (12), A2327
(2005).
13. A. V. Plakhotnyk, L. Ernst, and R. Schmutzler, J. Fluorine Chem., 126 (1), 27 (2005).
14. L. Zhang, Z. Zhang, S. Harring, M. Straughan, R. Butorac, Z. Chen, L. Lyons,
K. Amine, and R. West, J. Mater. Chem., 18 (31), 3713 (2008).
15. L. Zhang, L. Lyons, J. Newhouse, Z. Zhang, M. Straughan, Z. Chen, K. Amine,
R. J. Hamers, and R. West, J. Mat. Chem., 20 (38), 8224 (2010).
16. X. Chen, M. Usrey, A. Pen˜a-Hueso, R. West, and R. J. Hamers, J. Power Sources,
241, 311 (2013).
17. M. L. Usrey, X. Chen, J. A. Pena Hueso, R. C. West, and R. J. Hamers,
Lithium/carbon monofluoride batteries with organosilicon electrolytes, US Pat.
#8,486,569, (2013).
18. J. C. Burns, N. N. Sinha, G. Jain, H. Ye, C. M. VanElzen, E. Scott, A. Xiao,
W. M. Lamanna, and J. R. Dahn, J. Electrochem. Soc., 161 (3), A247 (2014).
19. C. R. Fell, L. Y. Sun, P. B. Hallac, B. Metz, and B. Sisk, J. Electrochem. Soc., 162
(9), A1916 (2015).
20. I. Belharouak, G. M. Koenig, T. Tan, H. Yumoto, N. Ota, and K. Amine, J. Elec-
trochem. Soc., 159 (8), A1165 (2012).
Conclusions
The standard carbonate electrolytes used in lithium-ion batteries
have poor thermal stability. In this study, organosilicon solvents were
shown to be stable against reaction with LiPF6, even when stored
at high temperatures. Unlike carbonate solvents, the organosilicon
decomposition mechanisms indicate that the salt and solvent degra-
dations are decoupled, which leads in part to the enhanced stability of
LiPF6 in organosilicon solvents. Furthermore, the formation of PF5
complexes in organosilicon solvents prevents further LiPF6 degrada-
tion. The addition of water to organosilicon electrolytes produces only
a slight increase in the onset and extent of decomposition of LiPF6 but
does not introduce new types of decomposition products. Investigation
of the specific impact of different organosilicon structures showed that
a glycol tail group is a point of intramolecular attack, while a nitrile
substituent is stable against breakdown even when stored at high tem-
peratures. Fluorination of the silicon also leads to decreased solvent
reactions and improved thermal stability.
Due to species produced at trace concentrations in carbonate and
organosilicon electrolytes at high temperatures, and the production of
gas-phase degradation species which were not analyzed in this study,
a complete mechanistic and kinetic picture of thermal degradation
in organosilicon electrolytes could not be achieved. Nevertheless, we
have uncovered many important mechanisms in the decomposition
processes for these novel electrolyte solvents. These studies can aid in
the future design of organosilicon electrolytes with exceptional ther-
mal stability and improved performance in battery cells. Organosilicon
molecules represent a promising new class of thermally stable battery
electrolyte solvents whose thermal stability can be tuned by the nature
of the molecular substituents.
21. W. Li, C. Campion, B. L. Lucht, B. Ravdel, J. DiCarlo, and K. M. Abraham, J.
Electrochem. Soc., 152 (7), A1361 (2005).
22. K. O. Christe, D. A. Dixon, G. J. Schrobilgen, and W. W. Wilson, J. Am. Chem. Soc.,
119 (17), 3918 (1997).
23. B. Vortmann, S. Nowak, and C. Engelhard, Anal. Chem., 85 (6), 3433 (2013).
24. C. L. Campion, W. Li, W. B. Euler, B. L. Lucht, B. Ravdel, J. F. DiCarlo,
R. Gitzendanner, and K. M. Abraham, Electrochem. Solid-State Lett., 7 (7), A194
(2004).
25. Y. Aihara, A. Sonai, M. Hattori, and K. Hayamizu, J. Phys. Chem. B, 110 (49), 24999
(2006).
26. S. Cosgrove, L. Rogers, C. M. Hewage, and J. P. G. Malthouse, Biochem., 46 (39),
11205 (2007).
27. S. Fang, G. Wang, L. Qu, D. Luo, L. Yang, and S.-i. Hirano, J. Mat. Chem. A, 3 (42),
21159 (2015).
28. Y. Liu, S. Fang, D. Luo, L. Yang, and S.-i. Hirano, J. Electrochem. Soc., 163 (9),
A1951 (2016).
29. R. Pietschnig, F. Belaj, and J. J. Tirre´e, Organometallics, 23 (21), 4897 (2004).
30. J. L. Brefort, R. J. P. Corriu, C. Guerin, B. J. L. Henner, and W. W. C. Wong Chi Man,
Organometallics, 9 (7), 2080 (1990).
31. M. Barrelle, C. Be´guin, and S. Tessier, Org. Magn. Resonance, 19 (2), 102 (1982).
32. H. E. Gottlieb, V. Kotlyar, and A. Nudelman, J. Org. Chem., 62 (21), 7512 (1997).
33. A. Wali, S. M. Pillai, S. Unnikrishnan, and S. Satish, J. Mol. Catal. A: Chem., 109
(2), 149 (1996).
Acknowledgments
34. R. Rohan, T.-C. Kuo, J.-H. Lin, Y.-C. Hsu, C.-C. Li, and J.-T. Lee, J. Phys. Chem. C,
120 (12), 6450 (2016).
The Bruker Avance III 500 MHz spectrometer was supported by
a generous gift from Paul J. and Margaret M. Bender. This mate-
rial is based upon work supported by the Wisconsin State Economic
Engagement and Development Program and the National Science
Foundation Graduate Research Fellowship Program under grant No.
DGE-1256259. Any opinions, findings, and conclusions or recom-
mendations expressed in this material are those of the authors and
do not necessarily reflect the views of the National Science Founda-
tion. Robert Hamers is a co-founder of Silatronix and has a financial
interest in the outcome of this work.
35. M. Ue, Y. Sasaki, Y. Tanaka, and M. Morita, in Electrolytes for Lithium Ion Batteries,
T. R. Jow, O. Borodin, K. Xu, and M. Ue, Editors, p. 93, Springer Science, New York
(2014).
36. Y. Liu, S. Fang, P. Shi, D. Luo, L. Yang, and S.-i. Hirano, J. Power Sources, 331, 445
(2016).
37. Q. Ma and B. K. Mandal, J. Electrochem. Soc., 162 (7), A1276 (2015).
38. P. S. Pregosin, NMR in Organometallic Chemistry, John Wiley & Sons, Weinheim,
Germany (2012).
39. J. Cazes, Ewing’s Analytical Instrumentation Handbook, CRC Press, Boca Raton,
FL, USA (2004).
40. R. M. Silverstein, F. X. Webster, D. J. Kiemle, and D. L. Bryce, Spectrometric Iden-
tification of Organic Compounds, John Wiley & Sons (2014).
41. S. Berger, S. Braun, and H.-O. Kalinowski, NMR Spectroscopy of the Non-Metallic
Elements, Wiley, Chichester, West Sussex, UK (1997).
42. S. A. Richards and J. C. Hollerton, Essential Practical NMR for Organic Chemistry,
in, p. 216, John Wiley & Sons, ltd (2011).
References
43. T. Achiha, T. Nakajima, Y. Ohzawa, M. Koh, A. Yamauchi, M. Kagawa, and
1. R. J. Brodd, in Lithium-ion Batteries: Science and Technologies, M. Yoshio,
R. J. Brodd, and A. Kozawa, eds., p. 1, Springer New York, NY, USA (2010).
2. D. Choi, W. Wang, and Z. Yang, in Lithium-ion Batteries: Advanced Materials and
Technologies, X. Yuan, H. Liu, and J. Zhang, eds., p. 1, CRC, Boca Raton, FL, USA
(2016).
3. G. Pistoia (ed.), Lithium-ion Batteries: Advances and Applications, Elsevier
(2013).
4. C. Cao, Y. Wu, and T. v. Ree, in Lithium-ion batteries: Fundamental and Applications,
Y. Wu, ed. p. 526, CRC, Boca Raton, FL, USA (2015).
5. B. Wang, X. Yuan, S. Zhao, Y. Wu, and T. van Ree, in Lithium-ion Batteries: Funda-
mentals and Applications, Y. Wu, Editor, p. 273, CRC Press, Boca Raton, FL, USA
(2015).
6. A. Lex-Balducci, W. Henderson, and S. Passerini, in Lithium-Ion Batteries: Advanced
Materials and Technologies, X. Yuan, H. Liu, and J. Zhang, ed., p. 147, CRC Press,
Boca Raton, FL, USA (2016).
7. J. R. Belt, C. D. Ho, T. J. Miller, M. A. Habib, and T. Q. Duong, J. Power Sources,
142 (1–2), 354 (2005).
8. J. Shim, R. Kostecki, T. Richardson, X. Song, and K. A. Striebel, J. Power Sources,
H. Aoyama, J. Electrochem. Soc., 156 (6), A483 (2009).
44. T. Achiha, T. Nakajima, Y. Ohzawa, M. Koh, A. Yamauchi, M. Kagawa, and
H. Aoyama, J. Electrochem. Soc., 157 (6), A707 (2010).
45. B. Suvorov, Russ. J. Gen. Chem., 76 (9), 1401 (2006).
46. R. J. H. Voorhoeve, Organohalosilanes: Precursors to Silicones, Elsevier Publishing
Company, Amsterdam, Netherlands (1967).
47. W. R. Dolbier, Guide to Fluorine NMR for Organic Chemists, John Wiley & Sons
(2009).
48. K. Tasaki, K. Kanda, S. Nakamura, and M. Ue, J. Electrochem. Soc., 150 (12), A1628
(2003).
49. A. F. Janzen, X. Ou, and M. G. Sowa, J. Fluorine Chem., 83 (1), 27 (1997).
50. G. Ceccarelli and F. Andruzzi, Makromol. Chem., 180 (5), 1371 (1979).
51. S. S. Zhang, K. Xu, and T. R. Jow, Electrochem. Solid-State Lett., 5 (9), A206 (2002).
52. S. E. Sloop, J. K. Pugh, S. Wang, J. B. Kerr, and K. Kinoshita, Electrochem. Solid-
state Lett., 4(4), A42 (2001).
53. S. E. Sloop, J. B. Kerr, and K. Kinoshita, J. Power Sources, 119–121, 330
(2003).
54. W. Li and B. L. Lucht, J. Power Sources, 168 (1), 258 (2007).
55. S. Wilken, M. Treskow, J. Scheers, P. Johansson, and P. Jacobsson, RSC Advances, 3
(37), 16359 (2013).
112 (1), 222 (2002).
9. D. Aurbach, B. Markovsky, A. Rodkin, E. Levi, Y. S. Cohen, H. J. Kim, and
M. Schmidt, Electrochimica Acta, 47 (27), 4291 (2002).
Downloaded on 2017-07-08 to IP 80.82.77.83 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).