Electrolyte Solvation and Ionic Association II. Acetonitrile-Lithium ...

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
0013-4651/2012/159(9)/A1489/12/$28.00 © The Electrochemical Society
Electrolyte Solvation and Ionic Association
II. Acetonitrile-Lithium Salt Mixtures: Highly Dissociated Salts
Daniel M. Seo, a,∗ Oleg Borodin, b Sang-Don Han, a,∗ Paul D. Boyle, c
and Wesley A. Henderson a,∗∗,z
a Ionic Liquids & Electrolytes for Energy Technologies (ILEET) Laboratory, Department of Chemical & Biomolecular
Engineering, North Carolina State University, Raleigh, North Carolina 27695, USA
b Electrochemistry Branch, Sensor & Electron Devices Directorate, U.S. Army Research Laboratory, Adelphi,
Maryland 20783, USA
c X-ray Structural Facility, Department of Chemistry, North Carolina State University, Raleigh,
North Carolina 27695, USA
A1489
The electrolyte solution structure for acetonitrile (AN)-lithium salt mixtures has been examined for highly dissociated salts. Phase
diagrams are reported for (AN) n -LiN(SO 2 CF 3 ) 2 (LiTFSI) and -LiPF 6 electrolytes. Single crystal structures and Raman spectroscopy
have been utilized to provide information regarding the solvate species present in the solid-state and liquid phases, as well as
the average solvation number variation with salt concentration. Molecular dynamics (MD) simulations of the mixtures have been
correlated with the experimental data to provide additional insight into the molecular-level interactions. Quantum chemistry (QC)
calculations were performed on (AN) n -Li-(anion) m clusters to validate the ability of the developed many-body polarizable force field
(used for the simulations) to accurately describe cluster stability (ionic association). The combination of these techniques provides
tremendous insight into the solution structure within these electrolyte mixtures.
© 2012 The Electrochemical Society. [DOI: 10.1149/2.035209jes] All rights reserved.
Manuscript submitted February 23, 2012; revised manuscript received April 6, 2012. Published August 14, 2012.
The key properties of electrolytes are determined by the solvate
species present in solution, but only limited information is available
regarding the structure of liquid electrolytes. A preliminary study of
(glyme) n -LiX mixtures indicated the following order for increasing
ionic association with varying anions: 1
LiTFSI, LiAsF 6 < LiClO 4 , LiI < LiBF 4
< LiCF 3 SO 3 < LiNO 3 , LiBr < LiCF 3 CO 2
based upon the phase behavior and crystalline solvates that form in
relatively dilute mixtures. This resulted in the following classification:
LiN(SO 2 CF 3 ) 2 (LiTFSI) and LiAsF 6 are highly dissociated; LiClO 4 ,
LiI and LiBF 4 are intermediately associated; and LiCF 3 SO 3 , LiNO 3 ,
LiBr and LiCF 3 CO 2 are associated or highly associated. Upon dissolution
in aprotic solvents, Li + cations are coordinated by the solvent
molecules. The anions remain uncoordinated by the solvent as anion
solvation occurs principally through hydrogen bonding (which does
not occur in aprotic solvents such as AN). A competition therefore
exists between the solvent molecules and anions for coordination to
the Li + cations in solution. This results in the formation of a variety of
solvate species which may be designated as solvent-separated ion pair
(SSIP), contact ion pair (CIP) or aggregate (AGG) solvates in which
the anions are coordinated to zero, one and two or more Li + cations,
respectively. The solvates present in solution are a strong function
of salt concentration and the structure of the solvent and anions, and
to a lesser extent the temperature—with ionic association (desolvation)
tending to increase with increasing temperature. Thus, for dilute
solutions with dissociated (e.g., AsF − 6 ) and highly associated (e.g.,
CF 3 CO − 2 ) anions, SSIP and AGG solvates, respectively, are expected
to predominate.
This paper continues a previous exploration of electrolyte solution
structure of acetonitrile (AN) mixtures with intermediate (LiClO 4 and
LiBF 4 ) and associated (LiCF 3 SO 3 , LiNO 3 and LiCF 3 CO 2 ) salts. 2 AN
has been used in this study as this solvent has relatively simple interactions
with Li + cations due to the single electron lone-pair on
the nitrogen atom of the solvent molecule. Nitrile and dinitrile solvents
may also be of interest as practical lithium battery electrolyte
components. 3–5 The present manuscript reports the phase behavior
and solution structure of (AN) n -LiTFSI and -LiPF 6 mixtures (highly
dissociated salts). This work is part of a larger study linking electrolyte
∗ Electrochemical Society Student Member.
∗∗ Electrochemical Society Active Member.
z E-mail: whender@ncsu.edu
solution structure to the transport properties of aprotic solvent-lithium
salt mixtures.
Experimental and Computational Methods Section
Materials.— Anhydrous AN (Sigma Aldrich, 99.8%) was used
as-received. LiTFSI and LiPF 6 (battery grade) were purchased from
3M Company and Novolyte, respectively. LiTFSI was dried at 120 ◦ C
for 24 hr. Samples were prepared in a Vacuum Atmospheres inert
atmosphere (N 2 ) glove box (< 5 ppm H 2 O) by combining the appropriate
amounts of salt and solvent in hermetically sealed vials and
heating/stirring on a hot plate to form homogeneous solutions. The
water content in the samples was confirmed to be negligible using Karl
Fischer titration (Mettler Toledo DL39X Coulometer). The compositions
are described generally using three notations: (1-x) AN-(x) LiX,
(AN) n -LiX and (AN) n :LiX for discussions focusing on mole fraction,
ratio of AN/Li and specific crystalline solvates, respectively.
Thermal characterization.— DSC measurements were performed
with a TA Instruments Q2000 DSC with liquid N 2 cooling. The instrument
was calibrated with cyclohexane (solid-solid phase transition
at −87.06 ◦ C, melt transition (T m ) at 6.54 ◦ C) and indium (T m at
156.60 ◦ C). Hermetically sealed Al sample pans were prepared in the
glove box. Sample pans were cycled (5 ◦ Cmin −1 ) and annealed repeatedly
at subambient temperature to fully crystallize the samples when
possible. Once the samples were crystallized, the pans were cooled to
−150 ◦ C and then heated (5 ◦ Cmin −1 ) to fully melt the samples. Only
the final heating traces are reported. Peak temperatures from this data
were then used to construct the phase diagrams.
Raman measurements.— Raman vibrational spectra were collected
with a Horiba-Jobin Yvon LabRAM HR VIS high-resolution
confocal Raman microscope using a 632 nm −1 He-Ne laser as the
exciting source and a Linkam stage for temperature control with a
long distance 50X objective. Spectra were typically collected using a
10 s measurement time and five accumulations. Raman spectra were
processed using LabSpec software.
Quantum chemistry (QC) calculations.— QC calculations were
performed on (AN) n -Li-(anion) m clusters (where anion = PF − 6 or
TFSI − ;n= 0, 2 or 3; m = 1 or 2) using Gaussian 09 Revision B.01
software. 6 Geometries were optimized at the M05-2X/6-31+G(d)
level, while the binding energies were calculated at the M05-2X/6-
31+G(d), MP2/aug-cc-pvDz and MP2/aug-cc-pvTz levels. Basis set

A1490 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
Figure 1. DSC heating traces (5 ◦ Cmin −1 ) and the corresponding phase diagrams of (a) (1-x) AN-(x) LiTFSI and (b) (1-x) AN-(x) LiPF 6 mixtures.
superposition error (BSSE) was calculated using the counterpoise
correction methodology.
Molecular dynamics (MD) simulations.— MD simulations were
performed on AN doped with LiPF 6 or LiTFSI employing a recently
developed, many-body polarizable APPLE&P (version 1e19) force
field for AN and the PF 6 − and TFSI − anions. 7–9 Extensive details regarding
the MD simulation methodology are provided in the previous
related manuscript. 2
Results and Discussion
Solvent-lithium salt phase behavior.— Pure AN.—AN undergoes a
solid-solid phase transition at −56 ◦ C prior to the T m at −46 ◦ C
(Fig. 1). 10–17 The low temperature (β or II) and high temperature (α or
I) phases both consist of ordered AN molecules with the phase transition
consisting of a 90 ◦ rotation of double slabs of the AN molecules
packed in the crystal layers. 14
AN-LiTFSI.—DSC measurements and the corresponding phase diagram
for (AN) n -LiTFSI mixtures (Fig. 1) agree well with a previously
reported partial phase diagram. 18 Three different crystalline solvate
phases are found, consisting of 6/1, 4/1 and 1/1 AN/LiTFSI solvates,
respectively. The structures of the 6/1 and 4/1 solvates are not yet
known. For the 6/1 solvate, it is possible that the Li + cations are fully
solvated by six AN molecules (octahedral coordination), but several
studies have suggested that it is not energetically favorable (in the gas
phase) to coordinate more than four AN molecules to a Li + cation. 19–23

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
A1491
The spectroscopic data (see below) confirms that this is a SSIP phase
with uncoordinated TFSI − anions, as expected for a solvate containing
six AN molecules. The 4/1 solvate is also a SSIP phase which may
have coordination similar to that found in the structures of the SSIP
(AN) 4 :LiClO 4 24 and (AN) 4 :LiI 25 phases in which the Li + cations are
fully solvated by four AN molecules (tetrahedral coordination) and
the anions are uncoordinated. For compositions between the 4/1 and
1/1 phases, it was not possible to fully crystallize the samples despite
subjecting the samples to extensive heating-cooling cycles at
subambient temperature. Thus, a “crystallinity gap” occurs for these
compositions due to either the slow nucleation of crystalline solvates
(beyond the time frame of the crystallization procedure utilized) or
the inhibition of ordered crystalline solvate formation due to unfavorable
packing. Single crystals of the 1/1 phase, i.e., (AN) 1 :LiTFSI
solvate, were obtained and the structure was determined and reported
elsewhere (see Supporting Information). 26 In this phase, there are two
different Li + cations. One Li + cation has six-fold coordinated by six
oxygen atoms from four TFSI − anions. The second Li + cation has
four-fold coordination by two oxygen atoms from two TFSI − anions
and two nitrogen atoms from two AN solvent molecules. Each TFSI −
anion is coordinated to three Li + cations (AGG solvate).
AN-LiPF 6 .—(AN) n -LiPF 6 mixtures form both 6/1 and 5/1 AN/LiPF 6
crystalline solvate phases. In dilute mixture (n ≥ 10), the 6/1 SSIP
solvate crystallizes with the peak at 18 ◦ C corresponding to the T m for
this phase (Fig. 1 and Supporting Information). It is difficult to characterize
the thermal behavior of this phase more definitively as more
concentrated mixtures (i.e., 10 > n ≥ 6) form both the 6/1 and 5/1
phases. For these samples, in each case, some of the sample crystallized
into the 5/1 phase upon cooling from the melt, even when a rapid
cooling procedure was used. The remainder of the sample then crystallized
at low temperature into the 6/1 phase resulting in the complicated
thermal behavior noted (Fig. 1). The crystal structure of the 6/1 phase,
i.e., (AN) 6 :LiPF 6 solvate, has recently been reported. 27 The structure
consists of Li + cations coordinated by four AN molecules with uncoordinated
PF 6 − anions and uncoordinated AN molecules (two per Li +
cation) located between the solvated cations (see Supporting Information).
The crystal structure of the 5/1 phase, i.e., (AN) 5 :LiPF 6 solvate,
has also been determined. 28 This resembles the 6/1 structure with Li +
cations coordinated by four AN molecules, uncoordinated PF 6 − anions
and uncoordinated AN molecules (one per Li + cation) located
between the solvated cations (see Supporting Information). The same
solvate structure is found for (AN) 5 :CuPF 6 . 29,30 The 5/1 solvate with
LiPF 6 has a T m at 67 ◦ C with a solid-solid phase transition at 38 ◦ C.
It is noteworthy that CuClO 4 and CuBF 4 form 4/1 phases 31–36 rather
than a 5/1 phase, as is also found for LiClO 4 and LiBF 4 2 given that
the Li + and Cu(I) + cations are nearly identical in size. 27,37 A eutectic
point is observed for more concentrated mixtures (n < 5) between the
5/1 solvate and a more aggregated solvate (composition unknown).
Despite the fact that the phase diagrams for both LiPF 6 and LiTFSI
suggests that these are highly dissociated salts (i.e., both are able to
form a 6/1 phase), significant differences are also evident. In particular,
LiPF 6 forms crystalline phases with a (relatively) high T m ,in
contrast with LiTFSI which forms phases which melt at low temperature.
Thus, (AN) n -LiTFSI mixtures remain liquid at −30 ◦ C over a
large concentration range, whereas all of the (AN) n -LiPF 6 mixtures
(except for the most dilute) crystallize readily at ambient temperature.
These differences are attributed to the differences in size, shape and
flexibility of the anions. The PF 6 − anions are essentially spherical and
easily pack together symmetrically with the solvated Li + cations. At
higher temperature, the uncoordinated PF 6 − anions may become disordered
by spinning about one axis or tumbling. The solid-solid phase
transition for the 5/1 solvate is a structural change in the crystal structure
which accomodates this disorder at elevated temperature. The
long-range order, however, is retained. The non-spherical shape of the
TFSI − anions, in contrast, likely results in a less symmetrical packing
of the solvated Li + cations amongst the uncoordinated anions. Further,
the anions are flexible 38,39 and may become conformationally disordered
at elevated temperature thus disrupting the long-range packing
of the solvate structure leading to melting instead of a disordered solid
crystalline (plastic crystalline) phase.
Raman characterization of AN-Li + cation solvation.— Upon coordination
with Li + cations, the vibrational bands associated with the
solvent C–C and C≡N stretching modes shift. Thus, these bands may
be utilized to determine the average solvation number (N) forAN
coordinated to the Li + cations. Uncoordinated AN has a ν 4 band at
918 cm −1 for the C–C stretching vibration and a ν 2 band at 2254 cm −1
for the C≡N stretching vibration (Fig. 2), with the 922 and 2251 cm −1
shoulders attributed to hot bands. 40,41 When the electron lone-pair
on the nitrogen is coordinated to a Li + cation, these bands shift to
930 and 2277 cm −1 , respectively (Fig. 2). 18,20,42,43 The following calculation
may be used to determine the fraction of uncoordinated and
coordinated AN molecules:
A AN-C
= N c LiX
[1]
A AN-C + A AN−UC c AN
where A AN-C and A AN-UC are the integrated area intensities of the bands
for the coordinated and uncoordinated AN, respectively, c LiX and c AN
are the concentrations of the salt and AN, respectively, and N is the
average solvation number. 44,45 From this information, values for N
may be determined by multiplying the fraction of coordinated solvent
by the total number of AN molecules present at a specified concentration
(Fig. 3). The analysis in Fig. 3 is based upon the assumption that
the relative activities of the bands associated with the uncoordinated
and coordinated AN have equivalent Raman activity (no scaling is required),
as was done in the previous manuscript. 2 Data were collected
for mixtures of (AN) n -LiTFSI and (AN) n -LiPF 6 at 60 ◦ C (note that
similar data has previously been reported for (AN) n -LiTFSI mixtures
at ambient temperature 18 ). For comparison, data for (AN) n -LiClO 4
mixtures is included in Fig. 3. 2 The data for the different salts may
then be compared for a specified composition. For example, for the
composition x = 0.20 (or four AN molecules per Li + cation), the average
solvation number is approximately 3.2 for LiPF 6 , 2.8 for LiTFSI,
and 2.7 for LiClO 4 (Fig. 3). This, along with a previous study for more
associated salts (for which the values were 2.1 for LiBF 4 and 1.0 for
LiCF 3 CO 2 for x = 0.20), 2 indicates the following order for the Li +
cation solvation number in liquid (AN) n -LiX mixtures:
LiPF 6 > LiTFSI ≥ LiClO 4 > LiBF 4 >> LiCF 3 CO 2
Based upon the suggested order for ionic association tendency indicated
previously, however, the results for LiTFSI are surprising as the
TFSI − anion was proposed to be a highly dissociated anion. 1 A comparison
of the crystalline phases that form for (AN) n -LiTFSI, -LiPF 6
and -LiClO 4 mixtures 2 does indeed suggest that, in the crystalline
phases, the TFSI − −
anions are highly dissociated (comparable to PF 6
anions). But in the liquid phase, as will be shown below, the TFSI −
anions may have an ionic association tendency closer to ClO − 4 than
PF − 6 anions. Note, however, that TFSI − also tends to form bidentate
coordination to a single Li + cation to a much greater extent than anions
such as PF − 6 ,ClO − 4 and BF − 4 . This is evident from the QC studies
of (AN) n -Li-(TFSI) m complexes and MD simulation results (see
below), as well as the known crystal structures of numerous LiTFSI
solvates. 18,26,46–49 Most of the solvated Li + cations have tetrahedral
coordination. Thus, if a given TFSI − anion contributes two rather
than one donor oxygens to the cation coordination, this may displace
an additional AN molecule. . . thereby lowering the value of N to some
extent from what would otherwise be predicted. This could therefore
be one possible explanation for the lower than expected N values noted
for the TFSI − anion.
Raman characterization of ionic association.— Fig. 4 shows the
Raman band vibration from the expansion and contraction of the entire
TFSI − anion 50,51 with changing concentration at −80 and 60 ◦ C. At
−80 ◦ C, all of the samples are crystalline solids, except for those in the
crystallinity gap which either remain liquid or are amorphous solids,
depending upon the sample T g (Fig. 1). Two bands are observed in the
dilute mixtures (n ≥ 3.6) which do not vary in position with varying

A1492 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
Figure 2. Raman spectra at 60 ◦ C of AN C–C stretching mode (920 cm −1 )andC≡N stretching mode (2250 cm −1 ) bands for (a) (AN) n -LiTFSI and (b) (AN) n -LiPF 6
mixtures (AN/LiX ratio indicated).
Figure 3. Raman spectroscopic analysis at 60 ◦ C of solvent bands for uncoordinated AN and Li + cation coordinated AN in (AN) n -LiX mixtures with (a) LiPF 6 ,
(b) LiTFSI and (c) LiClO 4 (the latter shown for comparison). The calculated Li + cation average solvation number (N) is shown at the top. The dark solid line
corresponds to the average of the two sets of data from Fig. 2.

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
A1493
Figure 5. TFSI − anion band variation with temperature in (AN) n -LiTFSI
mixtures with (a) n = 6.0 (crystalline in the −80 to −45 ◦ C range) and (b)
n = 4.0 (crystalline in the −100 to −35 ◦ C range).
Figure 4. TFSI − anion band variation with concentration in (AN) n -LiTFSI
mixtures at (a) −80 ◦ Cand(b)60 ◦ C. The mixtures with n = 3.0, 2.5, 2.0 and
1.55 are in the crystallinity gap and remain either fully amorphous liquids or
glassy solids at −80 ◦ C.
concentration at −80 ◦ C. Although two different crystalline solvates
(6/1 and 4/1 phases) are present for this concentration range (n ≥ 4)
(Fig. 1), the anion coordination to the Li + cations does not change
because both phases appear to consist of SSIP solvates in which the
anions are uncoordinated. The TFSI − anion is known to have two
different low-energy conformational states: a cisoid form (C 1 ) with
the CF 3 groups on the same side of the S–N–S plane and a transoid
form (C 2 ) with the CF 3 groups on opposite sides of the plane. 51 This
difference in the band position was initially thought to originate from
different conformations of the anions in the solvate crystal structures.
The data in Fig. 4a thus suggested that the 6/1 phase consists of
uncoordinated TFSI − anions with the C 1 conformation, whereas the
4/1 phase consists of uncoordinated TFSI − anions with both the C 1 and
C 2 conformations. To confirm this, the 250–450 cm −1 region of the
spectra from (AN) n -LiTFSI mixtures with n = 6 and 4 was examined
as the anion vibrational bands in this region provide a fingerprint for
the uncoordinated anion conformations (Fig. 5a). 51 The data, however,
clearly indicate that both phases consist only of anions with the C 2
conformation. Note, however, that new bands are evident at 382 cm −1
for the n = 6 sample (Fig. 5a) and at 315 cm −1 for the n = 4 sample
(Fig. 5b) which do not correspond to the bands typically noted for
either the C 1 or C 2 conformations, so it may be that one or more
different conformations are present which would also account for the
band at 741 cm −1 . Perhaps these bands are related to a third lowenergy
conformation for the TFSI − anion which has been reported, 52
but unfortunately the Raman spectrum for this conformation is not
yet known. In the crystallinity gap for more concentrated mixtures,
the samples are liquid or amorphous solids at −80 ◦ C and broad bands
are evident in the Raman spectra. As the concentration of LiTFSI
increases, the Raman band variation indicates that the amount of SSIP
solvates decreases and more aggregated solvates, i.e., CIP and AGG
solvates, increase. As the composition approaches n = 1, a crystalline
phase is formed again (along with an amorphous phase if 1 < n
< 1.5). The Raman spectra show sharper bands for the crystalline
phases and broader bands for the amorphous phase. For the n = 1
mixture, the entire sample is crystalline with one sharp Raman band
at 751 cm −1 , corresponding to the AGG-IIb (AN) 1 :LiTFSI crystalline
solvate (with the TFSI − anion coordinated to three Li + cations through
four oxygens).
For the liquid mixtures at 60 ◦ C (Fig. 4b), the shift of the
Raman bands for the TFSI − anion varies with concentration in a similar
manner to that noted for the solid phase. However, unlike the solid

A1494 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
Figure 6. Solvate species distribution in (AN) n -LiTFSI mixtures at 60 ◦ C.
phase, the bands shift smoothly with changing concentration. Due to
the structural flexibility of the TFSI − anion, different conformations
are possible and each of these may be coordinated in varying ways
to one or more Li + cation(s). Therefore, numerous types of solvates
are formed in the liquid mixtures with changing concentration, giving
overlapping bands. This makes it challenging to deconvolute the
spectra conclusively to identify specific forms of anion coordination
to the Li + cations. This has been done, however, using preliminary
assignments from a study underway in which the Raman spectra of
crystalline LiTFSI solvates have been correlated with known crystal
structures. From this analysis, the peak positions at −80 ◦ C/60 ◦ Care
C 2 -SSIP (741/740 cm −1 – uncoordinated TFSI − anions), C 2 -CIP-II
(747/746 cm −1 – TFSI − anion coordinated to a single Li + cation
through two oxygen atoms), C 2 -AGG-Ib (749/748 cm −1 – TFSI − anion
coordinated to two Li + cations through three oxygen atoms) and
C 1 -AGG-IIb (750-752/749-750 cm −1 – TFSI − anion coordinated to
three Li + cations through four oxygen atoms). Note that in the liquid
phase, these bands may shift somewhat due to variability in the anion
conformations and coordination bond lengths to the Li + cations.
The close proximity of the bands for AGG coordination (along with
other forms of AGG solvates which have not yet been characterized)
suggests that it is fruitless to attempt the identification of different
forms of the AGG solvates in solution. Thus, Fig. 6 simply notes
the deconvolution of the peaks in Fig. 4 in terms of uncoordinated
TFSI − anions (SSIP), anions coordinated to a single Li + cation (CIP)
and those coordinated to more than one Li + cation (AGG). Such an
analysis suggests that dilute mixtures contain > 80% SSIP solvates,
with the remainder CIP solvates. With increasing salt concentration,
the fraction of both CIP and AGG solvates increases at the expense
of the SSIP solvates, but the latter persist even for very concentrated
mixtures.
To examine why the crystallinity gap occurs for the (AN) n -LiTFSI
mixtures, variable-temperature Raman spectra have been measured
for compositions within the crystallinity gap (Fig. 7 and Supporting
Information). For the n = 4.0 mixture, the Raman bands at 741
and 743 cm −1 (from uncoordinated TFSI − anions) increase as the
temperature decreases, and the band at 747–749 cm −1 (attributed
to CIP and AGG-I solvates) decreases, indicating that the amount
of SSIP solvates increases at lower temperature. This results in the
nucleation and growth of the SSIP 4/1 crystalline phase. The n = 3.0
sample, however, did not crystallize and the Raman spectra show that
even though the amount of SSIP solvates increased as the temperature
decreased, there remains a sizeable amount of coordinated anions
(corresponds to Raman band on 748 cm −1 ) thus hindering/preventing
the formation of the 4/1 crystalline phase. Similarly, although the n
Figure 7. Variable-temperature TFSI − anion band variation with concentration
in (AN) n -LiTFSI mixtures with (a) n = 4.0, (b) n = 3.0, (c) n = 2.0 and
(d) n = 1.0.
= 2.0 sample has a dominant peak at 750 cm −1 (corresponding to an
AGG-II solvate as found in the 1/1 phase), a significant fraction of
the anions persist as SSIP, CIP and perhaps AGG-I solvates which
hinder/prevent the nucleation and growth of the 1/1 crystalline phase.
In contrast, for the n = 1.0 sample, most of the anions have AGG-II
(or perhaps even higher) aggregation, thus facilitating the nucleation
of the 1/1 crystalline phase. Note that in the phase diagram for the
(AN) n -LiTFSI mixtures in Fig. 1, the“×” symbols indicate the T g

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
A1495
Figure 8. PF 6 − anion band variation with concentration in (AN) n -LiPF 6 mixtures
at (a) −80 ◦ Cand(b)60 ◦ C.
of the fully amorphous samples, whereas the “triangles” are the T g
for the amorphous phase which remains after some portion of a sample
has crystallized as the 1/1 phase. The composition of the latter
amorphous phase will thus be more dilute than the composition of
the sample itself. These samples have nearly the same T g value of
−82 ◦ C and this value can be used to estimate the composition of the
stabilized amorphous phase (i.e., x ∼ 0.25). This value corresponds
to an average of 3 AN molecules per Li + cation.
Fig. 8a shows the band variation of the PF 6 − anion with varying
concentration at −80 and 60 ◦ C. Based upon O h symmetry, the band
assignments have been determined. 53–58 The most intense Raman band
for the PF 6 − anion is observed at 740–750 cm −1 .At−80 ◦ C, all of the
samples are crystalline solids. For dilute mixtures (n ≥ 5), only a single
band is evident at 744 cm −1 . This band corresponds to SSIP solvates
in which the PF 6 − anion is uncoordinated. There appears to be a slight
shift to higher wavenumber between the 6/1 and 5/1 samples (perhaps
corresponding to differences in structure (i.e., lattice packing) between
the (AN) 6 :LiPF 6 and (AN) 5 :LiPF 6 crystalline solvates). A new band
appears at 748 cm −1 as the concentration increases (n ≤ 5). This
band corresponds to an as yet undetermined more aggregated phase
(perhaps a 1/1 phase).
For the liquid mixtures at 60 ◦ C(Fig.8b), the Raman band for the
PF 6 − anion shifts to lower wavenumber and broadens. In dilute mixtures,
one Raman band is observed at 741 cm −1 . This band corresponds
to SSIP solvates. As the concentration increases, an asymmetric shoulder
at higher wavenumbers grows with increasing concentration. This
Figure 9. Geometries of the (a) (AN) 3 -Li-TFSI (A: C 1 -CIP-I), (b) (AN) 3 -
Li-TFSI (B: C 2 -CIP-I), (c) (AN) 3 -Li-TFSI (C: C 1 -CIP-II), (d) (AN) 3 -Li-TFSI
(D: C 1 -CIP-II), (e) (AN) 2 -Li-(TFSI) 2 (A: C 1 -CIP-I, C 2 -CIP-II), (f) (AN) 2 -Li-
(TFSI) 2 (B: C 2 -CIP-II, C 2 -CIP-II), (g) Li-(TFSI) 2 (C 2 -CIP-II, C 2 -CIP-II), (h)
(AN) 3 -Li-PF 6 and (i) (AN) 2 -Li-(PF 6 ) 2 clusters optimized using the M05-2X
functional with the 6-31+G(d) basis set (optimized in the gas phase) (Li-purple,
N-blue, O-red, S-yellow, F-light green).
is attributed to one or more bands due to CIP and AGG solvates. It is
clear that SSIP solvates dominate the mixtures for dilute concentrations
and such solvates are found even in the concentrated mixtures.
This confirms that LiPF 6 is a highly dissociated salt.
QC studies and force field validation.— QC studies of (AN) n -Li-
(anion) m complexes focused on the calculation of the binding energies,
which were then compared with the corresponding energies obtained
from molecular mechanics (MM) geometry optimizations employing
a polarizable force field (FF). QC studies were performed on the
clusters shown in Fig. 9. Binding energies relative to the isolated
species in the gas-phase are shown in Table I together with the previously
reported binding energies for the analogous complexes with
BF 4 − and ClO 4 − anions. 2 The binding energies, after BSSE correction
from the (computationally) inexpensive DFT M05-2X/6-31+G*
method, are found to be generally in fair agreement (< 3 kcal mol −1 )
with the significantly more expensive results from MP2/aug-cc-pvTz
calculations. The MP2/aug-cc-pvDz binding energies, on the other
hand, showed larger deviations from the MP2/aug-cc-pvTz energies
(Table I). Thus, the inexpensive M05-2X/6-31+G* binding energies
were deemed acceptable for the solvent-Li-anion force field validation,
instead of the MP2/aug-cc-pvDz energies, for the cases when
the MP2/aug-cc-pvTz energies are too expensive to calculate, such as
for the (AN) n -Li-(TFSI) m clusters shown in Fig. 9. An examination of

A1496 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
Table I. (AN) n -Li-(anion) m complex binding energies (in kcal mol −1 ) from QC calculations of the geometry using the M05-2X/6-31+G(d) level
and molecular mechanics MM (FF).
M052x/6-31+G(d) MP2/Dz MP2/Tz MM (FF) MM(FF)-MP2/Tz MM(FF)-M052x/6-31+G*
(AN) 3 -Li-BF 4 −191.4 −185.9 −189.1 −188.9 0.2 2.5
(AN) 3 -Li-ClO 4 −188.5 −184.3 −188.0 −188.0 0.1 0.5
(AN) 3 -Li-TFSI(A) (C 1 -CIP-I) −184.0 −180.4 −183.6 0.5
(AN) 3 -Li-TFSI(B) (C 2 -CIP-I) −180.8 −177.3 −182.0 −1.2
(AN) 3 -Li-TFSI(C) (C 1 -CIP-II) −181.5 −178.2 −181.4 0.1
(AN) 3 -Li-TFSI(D) (C 1 -CIP-II) −180.9 −177.4 −180.8 0.0
(AN) 3 -Li-PF 6 −184.5 −179.0 −182.8 −183.4 −0.6 1.1
(AN) 2 -Li-(BF 4 ) 2 −205.8 −199.6 −202.9 −203.8 −0.9 2.0
(AN) 2 -Li-(ClO 4 ) 2 −201.1 −196.4 −200.6 −201.6 −1.0 −0.5
(AN) 2 -Li-(TFSI) 2 (A) −199.7 −198.8 0.9
(AN) 2 -Li-(TFSI) 2 (B) −198.5 −194.5 −197.8 0.7
(AN) 2 -Li-(PF 6 ) 2 −193.6 −188.9 −192.1 −191.8 0.3 1.8
Li-(TFSI) 2 −185.0 −178.8 −183.3 1.6
the solvate structures in Fig. 9 suggests that these do not necessarily
represent the lowest energy structures in solution as the AN molecules
‘wrap’ around the anions to optimize the solvate energetics as there
is no surrounding shell of solvent molecules and ions with which to
interact. Nevertheless, this is not expected to impact the results significantly
and a comparison of the calculated solvate binding energies
and geometries is very informative.
The DFT binding energies shown in Table I indicate that the magnitude
of the complex stability follows the order PF − 6 ∼ TFSI −
< ClO − 4 < BF − 4 with the BF − 4 anion being the most stable. This
order is different from the order of the Li + . . . anion binding energies
without any solvent present (i.e., TFSI − < ClO − 4 < PF − 6 < BF − 4 )
reported by Johansson from MP2/6-31G(d) level calculations. 59 Thus,
inclusion of the solvent in the Li + . . . anion complex calculations
changes the order of the binding energy stabilization. Fig. 9a, 9b
and 9h also shows that the Li + cation binds to one TFSI − oxygen
or one PF − 6 fluorine atom in the most stable configurations for the
(AN) 3 -Li-anion clusters (Table I), which is different from the binding
arrangement in the most stable Li-PF 6 and Li-TFSI complexes (without
AN), where the Li + cation is bound to three fluorine atoms of
the PF − 6 anion (CIP-III) or to two oxygen atoms of the TFSI − anion
(CIP-II). 60–64 It was found, however, that a stable configuration of the
(AN) 3 -Li-TFSI complex with two TFSI − oxygen atoms bound to a
Li + cation (geometry C - Fig. 9c) is only 2-3 kcal mol −1 less stable
than the configuration with one TFSI − anion oxygen bound to a Li +
cation (geometry A - Fig. 9a). The (AN) 3 -Li-TFSI complexes C and D
have the TFSI − anion in the same conformation with φ C-S..S-C = 105 ◦
and bidentate binding to a Li + cation, but different arrangements of the
AN molecules (Fig. 9c and 9d) resulting in a slight (< 1 kcal mol −1 )
energy difference between the solvates. Amongst the (AN) 3 -Li-TFSI
complexes A and B (Fig. 9a and 9b) with monodentate binding of a
TFSI − anion to a Li + cation, complex A with the C 1 TFSI − anion
conformer is more stable than complex B with the C 2 TFSI − anion
conformer.
In order to further investigate the tendency of the TFSI − anion
to contribute one or two oxygen atoms to the anion. . . Li + cation
coordination, two (AN) 2 -Li-(TFSI) 2 complexes were examined: in
complex A, one TFSI − anion is bound to a Li + cation with one oxygen,
while the other TFSI − anion is bound to the Li + cation with two
oxygen atoms (Fig. 9e); in complex B, both TFSI − anions contribute
two oxygen atoms each to the anion. . . Li + cation coordination (Fig.
9f). The DFT geometry optimization of these complexes started from
a configuration with two AN molecules bound to a Li + cation, but
the optimized geometry instead converging to a structure with only
one of the AN molecules bound to the Li + cation while the other
AN is located in close proximity, but directed away from the Li +
cation (Fig. 9e and 9f). Complex A was found to be more stable
by about 1 kcal mol −1 than complex B (Table I) indicating that, for
the (AN) 2 -Li-(TFSI) 2 complexes, the configuration with two TFSI −
anions contributing both oxygen atoms to the anion. . . Li + cation
coordination is not the most energetically favorable (although the
difference is not large). In contrast to the (AN) 2 -Li-(TFSI) 2 complexes,
which do not have two AN molecules bound to the Li + cationinthe
most stable configurations, the most stable (AN) 2 -Li-(PF 6 ) 2 complex
does have two AN molecules, in addition to the two PF 6 − anions,
bound to the Li + cation (Fig. 9i).
Next, the ability of the many-body polarizable force field to predict
the binding energies and geometries (relative to the QC calculations)
was examined for the (AN) n -Li-(anion) m clusters. The MM
optimization using the developed FF predicted cluster geometries is in
acceptable agreement with the geometries from the DFT calculations.
For example, for the (AN) 3 -Li-PF 6 cluster, the distances are R(Li-N)
= 2.0–2.1 Å from both DFT and MM(FF) and R(Li-F) = 1.8 Å from
MM(FF) and 1.9 Å from DFT. For the (AN) 3 -Li-(PF 6 ) 2 complex,
R(Li-N) = 2.1 Å from both DFT and MM(FF), while R(Li-F) was in
the range of 1.7–1.9 Å from MM(FF) and 1.9–2.1 Å from the DFT calculations.
Excellent agreement (less than 0.05 Å deviations) between
R(Li-O) and Li-N(AN) distances from the DFT and MM(FF) calculations
were also observed for both the A and B (AN) 3 -Li-TFSI geometries.
The complex binding energies from the MM(FF) calculations are
compared with the QC results in Table I. Excellent agreement is observed
between the DFT M05-2X/6-31+G* and MM(FF) predictions
with deviations of 1.1 kcal mol −1 or less for the (AN) 3 -Li-anion complexes
and 1.8 kcal mol −1 or less for the (AN) 2 -Li-(anion) 2 complexes
with the PF 6 − and TFSI − anions. Similar quality MM(FF) predictions
were observed when compared to a more accurate MP2/aug-cc-pvTz
level. Note that MM(FF) correctly predicts that the A geometries
of (AN) 3 -Li-TFSI and (AN) 2 -Li-(TFSI) 2 are more stable than the B
geometries.
DFT calculations were also employed to examine the shift of the
TFSI − anion Raman bands and intensities upon Li + cation complexation.
While previous DFT calculations 62–64 have investigated the influence
of Li + cation coordination on the TFSI − anion Raman and
IR band vibrations, no solvent (i.e., AN) was included in the studies.
The solvent is included, however, explicitly and implicitly through a
PCM model in the present work. For this, an additional set of DFT
calculations was performed on the (AN) n -Li-(TFSI) m solvates at a
M05-2X/6-31+G* level using a self-consistent reaction field with the
polarized continuum PCM model with AN parameters from the Gaussian
g09 package. Initial geometries were taken from the gas-phase
DFT calculations without PCM (Fig. 9). The resulting optimized geometries
with PCM(AN) are shown in the Supporting Information.
The band frequencies and Raman activities are given in Table II for
the spectral range around 740–750 cm −1 that is used for the experimental
characterization of solvates. In this region, similar frequencies
are noted for the C 1 and C 2 TFSI − anion conformations. Bidentate

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
A1497
Table II. Raman frequency (ν) and activity (I) for (AN) n -Li-(TFSI) m solvates from M05-2X/6-31+G* calculations with PCM(AN) (solvates shown
in Supporting Information Fig. S10).
complex ν (cm −1 ) scaled ν a (cm −1 ) ν complex −ν SSIP I complex /I SSIP φ C-S...S-C ( ◦ )
TFSI (C 1 -SSIP) 749.7 742.2 −39
TFSI (C 2 -SSIP) 751.4 743.9 −174
(AN) 3 -Li-TFSI(A) (C 1 -CIP-I) 753.9 746.3 3.3 0.93 −45
(AN) 3 -Li-TFSI(B) (C 2 -CIP-I) 750.5 743.0 0.0 0.95 177
(AN) 3 -Li-TFSI(C) (C 1 -CIP-II) 755.5 747.9 5.0 0.98 −127
(AN) 3 -Li-TFSI(D) (C 1 -CIP-II) 756.6 749.0 6.1 0.98 −112
(AN) 2 -Li-TFSI 758.3 750.7 7.7 1.01 −169
(AN) 2 -Li-(TFSI) 2 758.9 751.3 8.4 1.09 179,179
Li-(TFSI) 2 756.4 748.9 5.9 1.01 178,178
a M05-2X/6-31+G* frequency multiplied by scaling factor of 0.99.
binding of a Li + cation to the TFSI − anion (i.e., CIP-II) results in a
frequency shift of 5–8 cm −1 (from the uncoordinated anion), while the
monodentate binding of the TFSI − anion to a Li + cation (i.e., CIP-I)
in (AN) 3 -Li-TFSI clusters results in a much smaller frequency shifts
of 0 and 3 cm −1 , respectively, for C 2 and C 1 TFSI − anion conformers.
The observed frequency shift upon bidentate binding of 5–8 cm −1
is in good agreement with previous DFT calculations 64 and experimental
data for solvates with known crystal structures. The calculated
much smaller shift of 0-3 cm −1 for monodentate coordination is also
qualitatively in accord with previously reported calculated results for
Li + (TFSI(C 2 )) 4 complexes. 64 The small shift for TFSI − anion CIP-I
coordination (relative to the uncoordinated anion) is an important
point to note as this is potentially a source of error, perhaps a significant
one, in the analysis performed in Figs. 4 and 6. No experimental
data is available at present, however, to validate the DFT prediction
that Li + cation complexation to a single oxygen atom of the TFSI −
anion results in a frequency shift of less than < 3cm −1 for the TFSI −
anion vibrational band located near 740 cm −1 . Finally, the total energy
of the (AN) 3 -Li-TFSI complexes A-D was calculated to estimate
their stability with the inclusion of the implicit solvent PCM(AN).
It was found that the energy of the (AN) 3 -Li-TFSI (A-D) complexes
varied by less than 1 kcal mol −1 with the relative binding energies (to
A) following the order: A (0 kcal mol −1 ) > B (0.4 kcal mol −1 ) > D
(0.7 kcal mol −1 ) > C (1.0 kcal mol −1 ) (in contrast to the results reportedinTableI
which did not included the polarized continuum PCM
model for the surrounding solvent).
MD simulations of (AN) n -LiTFSI and -LiPF 6 mixtures.— MD simulations
have been used to examine the calculated solution solvate
structures and to explore the ionic association tendence of the salts in
AN. Three-dimensional snapshots of the simulations for the (AN) n -
LiTFSI and -LiPF 6 mixtures with n = 30, 20, 10, 5 and 2 may be
viewed using the .xyz files provided in the Supporting Information and
a freeware structure viewing program such as Mercury. The snapshots
were created by modifying the simulation box results as follows—
unwrapping all of the molecules/ions so that they were not divided,
applying periodic boundary conditions so as not to divide the solvate
structures across the boundaries of the simulation box (thereby retaining
entire solvates) and removing solvent molecules if they were >
2.90 Å from the Li + cations (uncoordinated and not in close proximity)
to facilitate the viewing of the solvates.
Table III shows the fraction of the uncomplexed Li + cations and
anions (TFSI − or PF 6 − ) and the coordination of the anions as determined
by the MD simulations. Li + cation coordination to the anions
was defined as Li + cations within 4.74 or 3.70 Å, respectively, of
the TFSI − or PF 6 − anions (N or P atoms, respectively). These values
were chosen because the radial distribution function (RDF) g Li–N (r)
for Li–N(anion) and g Li–P (r) for Li–P equals one after the first peak
for these distances (Fig. 10). In the previous manuscript associated
with this study of solution structure, 2 it was found to be advantageous
to also scrutinize the coordination using a distance defined by g Li–X (r)
= max(1st peak)/2, where for the present study X = P or N(anion)
(i.e., 4.56 or 3.50 Å, respectively for TFSI − or PF 6 − (Table III)),
which may be interpreted as representing the strongly bound Li +
cations, whereas the former values (4.74 or 3.70 Å) include both the
strongly and loosely bound Li + cations.
A comparison of the MD simulation results (Table III) with the
experimental data (Fig. 8) for the (AN) n -LiPF 6 mixtures suggests that
there is reasonable agreement. The dilute mixtures are dominated
by uncoordinated anions and fully solvated Li + cations, i.e., SSIP
solvates (see Supporting Information). Examples of the solvate complexes
found in the simulation dilute mixtures are shown in Fig. 11.
Note that the bidentate coordination of the PF 6 − aniontoaLi + cation
(i.e., Fig. 11f) is only rarely observed, as reflected by the similarity
of the values for the number of fluorine and phosphorus atoms in
close proximity (within 2.40 and 3.70 Å, respectively) to a Li + cation
(Table III). The average solvation numbers (for AN coordinated to Li +
cations) from the MD simulations (Table III) are also in reasonable
accord with the experimental values (Fig. 3) for the most concentrated
mixtures, but the values deviate for the n = 10 composition
Figure 10. Radial distribution function (RDF) from the MD simulations for
the (a) (AN) n -LiTFSI and (b) AN-LiPF 6 (n = 10) mixtures.

A1498 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
Table III. Composition of the ion coordination shell from MD simulations of (AN) n -LiTFSI and (AN) n -LiPF 6 mixtures at 60 ◦ C (Note: some of the
percentages do not sum to 100% due to the rounding off of the values).
LiTFSI
Property (AN:Li ratio): 30 20 10 5 2
No. solvent in MD box 480 640 640 640 512
No. LiTFSI in MD box 16 32 64 128 256
Concentration (M) 0.56 0.81 1.46 2.43 3.98
Molality (mol kg −1 ) 0.72 1.03 1.78 2.80 4.27
Simulation run length a (ns) 17.3 (6) 12.2 (4) 7.0 (9) 15.3 (16) 15.0 (16)
Simulation box length (Å) 36.30 40.35 41.74 44.38 43.11
MD density (g cm −3 ) 0.844 0.896 1.019 1.197 1.469
Expt density (g cm −3 ) 0.861 0.887 1.005 1.176 (1.385) b
Fraction of free Li (r Li–N > 4.74 Å) (SSIP) 0.37 0.31 0.20 0.10 0.02
Fraction of free N* (r Li–N > 4.74 Å) (SSIP) 0.36 0.27 0.16 0.06 0.01
Li + coordination numbers
# N** (within 2.40 Å of Li + ) 3.00 2.88 2.63 2.23 1.45
# O (within 2.40 Å of Li + ) 0.83 0.97 1.24 1.67 2.49
# N* (within 4.74 Å of Li + ) 0.75 0.89 1.14 1.57 2.42
Probability of finding the following number of Li + cations within the given distance from the N of TFSI −
0Li + within 4.56/4.74 Å of N (SSIP) 0.41/0.36 0.32/0.27 0.25/0.16 0.10/0.06 0.02/0.01
1Li + within 4.56/4.74 Å of N (CIP) 0.52/0.55 0.57/0.58 0.50/0.56 0.47/0.41 0.20/0.12
2Li + within 4.56/4.74 Å of N (AGG-I) 0.07/0.10 0.11/0.14 0.22/0.25 0.36/0.42 0.45/0.41
3Li + within 4.56/4.74 Å of N (AGG-II) 0.00/0.00 0.01/0.01 0.03/0.03 0.07/0.10 0.28/0.37
4Li + within 4.56/4.74 Å of N (AGG-III) 0.00/0.00 0.00/0.00 0.00/0.00 0.00/0.01 0.05/0.09
* N from TFSI − , ** N from AN
LiPF 6
Property (AN:Li ratio): 30 20 10 5 2
No. solvent in MD box 480 640 640 640 512
No. LiPF 6 in MD box 16 32 64 128 256
Concentration (M) 0.58 0.87 1.67 3.06 5.89
Molality (mol kg −1 ) 0.72 1.03 1.78 2.80 4.27
Simulation run length a (ns) 12.0 (4) 11.0 (4) 7.7 (2) 18.0 (4) 14.0 (4)
Simulation box length (Å) 35.78 39.42 39.71 40.51 40.02
MD density (g cm −3 ) 0.802 0.844 0.938 1.091 1.379
Expt density (g cm −3 ) 0.814 0.842 0.936
Fraction of free Li (r Li–P > 3.70 Å) (SSIP) 0.66 0.61 0.47 0.30 0.06
Fraction of free P (r Li–P > 3.70 Å) (SSIP) 0.66 0.60 0.44 0.24 0.02
Li + coordination numbers
# N (within 2.40 Å of Li + ) 3.42 3.37 3.15 2.84 1.81
# F (within 2.40 Å of Li + ) 0.38 0.44 0.68 1.02 2.13
# P (within 3.70 Å of Li + ) 0.36 0.42 0.62 0.96 2.00
Probability of finding the following number of Li + cations within the given distance from the P of PF − 6
0Li + within 3.50/3.70 Å of P (SSIP) 0.70/0.66 0.65/0.60 0.50/0.44 0.31/0.24 0.11/0.02
1Li + within 3.50/3.70 Å of P (CIP) 0.29/0.33 0.34/0.38 0.46/0.50 0.56/0.58 0.34/0.24
2Li + within 3.50/3.70 Å of P (AGG-I) 0.01/0.01 0.01/0.02 0.04/0.06 0.12/0.18 0.36/0.50
3Li + within 3.50/3.70 Å of P (AGG-II) 0.00/0.00 0.00/0.00 0.00/0.00 0.00/0.01 0.16/0.22
4Li + within 3.50/3.70 Å of P (AGG-III) 0.00/0.00 0.00/0.00 0.00/0.00 0.00/0.00 0.02/0.02
a equilibration run lengths are given in parentheses,
b experimental density value in parentheses was extrapolated from experimental data.
(N = 3.15 from the MD simulation, but ∼4 from the experimental
results). Note that for the n = 10 composition at 60 ◦ C, there
is no significant evidence of the presence of CIP or AGG solvates
(Fig. 8b), but CIP and AGG solvates are found in the MD simulations
for the most dilute mixtures studied with n = 10, 20 and 30
(see Supporting Information). Thus, the simulations (i.e., force fields)
appear to overpredict the degree of ionic association of the electrolyte
mixtures. This was also true for the mixtures with LiBF 4 (and perhaps
LiClO 4 ). 2
For the highly concentrated (AN) n -LiTFSI mixtures, reasonable
agreement is noted between the MD simulations and experimental
data. For example, for the n = 5 concentration, simulations predict
that the Li + cations have (on average) 0.6 less AN in the first coordination
shell for the (AN) n -LiTFSI mixture as compared to the analogous
(AN) n -LiPF 6 mixture, in reasonable agreement with the experimental
observation for x = 0.20 (i.e., n = 4) where coordination of the Li +
cations in the (AN) n -LiTFSI mixture contains 0.4 less AN (on average)
than in the (AN) n -LiPF 6 mixture (Fig. 3). For the dilute mixtures,
however, poor agreement is found between the MD simulations
(Table III) and experimental data (Fig. 6) for the SSIP, CIP and AGG
distributionin(AN) n -LiTFSI mixtures. The experimental spectroscopic
data for the anion (Figs. 4 and 6) indicates that the LiTFSI salt
is highly dissociated, although to a lesser extent than LiPF 6 .Thisis
also evident in the phase behavior (Fig. 1), as both LiPF 6 and LiTFSI
form a SSIP 6/1 crystalline phase which is not found for the LiClO 4
and LiBF 4 (intermediately associated) salts. 2 The results from the MD
simulations, however, suggest that, for the dilute mixtures, the majority
of the TFSI − anions are coordinated to the Li + cations as CIP

Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
A1499
Figure 11. Representative Li + cation solvate species (i.e., coordination shells)
extracted from the MD simulations for the (AN) n -LiPF 6 mixtures (n = 30, 20
and 10) at 60 ◦ C: (a) SSIP, (b) CIP-I, (c) AGG-I, (d) CIP-I (×2), (e) CIP-I,
AGG-I and (f) CIP-I, CIP-II (Li-purple, N-blue, P-orange, F-light green).
solvates. Further, from the MD simulations, the solvation number for
LiTFSI (Table III) is lower than for LiClO 4 2 which also conflicts with
expectations from the experimental data—note that most of the Li +
cations have 4-fold coordination. . . thus, a lower solvation number
equates to an increase in anion coordination. There is no experimental
evidence which supports the MD simulation conclusion that LiTFSI
is more associated than LiClO 4 . One potential source of error in the
estimated experimental solvate distribution (Fig. 6) is the use of a
band at 747 cm −1 to estimate the fraction of CIP solvates during the
band deconvolution. The QC calculation results (Table II), however,
suggest that this may only account for the CIP-II (bidentate) coordinated
TFSI − anions, with the CIP-I (monodentate) coordinated anions
included, in part,with the SSIP fraction. In reality, a new band would
need to be introduced during the deconvolution (for CIP-I solvates)
which would change both the SSIP and CIP (i.e., CIP-II) band areas.
Thus, the fraction of SSIP and CIP solvates should perhaps be
smaller and larger, respectively. There is no experimental proof yet
available regarding the band position for the monodentate anion coordination,
but this is discussed to bring this possible source of error
in the analysis to the reader’s attention. Even so, it is still evident that
the MD simulations predict for the dilute mixtures increased ionic
association (fraction of CIP and AGG solvates) relative to the experimental
data—this is true for all of the salts studied thus far (i.e., LiPF 6 ,
LiTFSI, LiClO 4 and LiBF 4 ).
Despite the discrepancies between the simulation and experimental
results, it is quite informative to examine the manner in which the
TFSI − anions coordinate the Li + cations in the solvates found in the
MD simulations (Fig. 12). In contrast with semi-spherical anions such
as PF 6 − ,ClO 4 − and BF 4 − , bidentate coordination of the TFSI − anion
Figure 12. Representative Li + cation solvate species (i.e., coordination shells) extracted from the MD simulations for the (AN) n -LiTFSI mixtures (n = 30, 20 and
10) at 60 ◦ C: (a) C 1 -SSIP, (b) C 1 -CIP-I, (c) C 2 -AGG-I, (d) C 2 -CIP-I, C 1 -CIP-II (e) C 2 -CIP-I (×2), C 1 -AGG-I, (f) C 2 -CIP-II (×2), (g) C 1 -CIP-II, (h) C 1 -AGG-I
(×2), (i) C 1 -AGG-I, C 2 -AGG-I and (j) C 1 -CIP-I (×2), C 1 -AGG-I, C 2 -AGG-I, C 2 -AGG-II (Li-purple, N-blue, O, red, S-yellow, F-light green).

A1500 Journal of The Electrochemical Society, 159 (9) A1489-A1500 (2012)
to a given Li + cation in the MD simulations is not uncommon
(Fig. 12d, 12f and 12g). For example, for the n = 10 concentration,
12% of the TFSI − anions have two oxygen atoms from the same anion
bound to the same Li + cation, which is similar to the observations from
previous simulations of (EC) n -LiTFSI 65 and (ionic liquid) n -LiTFSI 66
mixtures. As noted above, this, in part, may explain the lower than
expected average solvent coordination numbers noted experimentally
for LiTFSI (as comparable to LiClO 4 ) despite the additional experimental
evidence (i.e., phase diagram and Raman spectroscopy of the
TFSI − anion coordination) which suggest that LiTFSI is more dissociated
than LiClO 4 (but less dissociated than LiPF 6 ) in AN. Thus,
the designation of the TFSI − anion as “highly dissociated” may be
questionable and it may be more appropriate to categorize this anion
instead as “dissociated” with the PF − 6 anion retaining the “highly
dissociated” appellation.
Conclusions
The solid-state and solution structure of electrolyte mixtures consisting
of AN and either LiTFSI or LiPF 6 have been examined in
detail using phase diagrams and Raman spectroscopy. Both salts are
able to form a 6/1 AN/LiX crystalline phase which is not found for
more associated salts. LiPF 6 forms crystalline phases with a high T m ,
in contrast with LiTFSI which forms phases which melt at low temperature.
Thus, (AN) n -LiTFSI mixtures remain liquid at −30 ◦ C over
a large concentration range, whereas the (AN) n -LiPF 6 mixtures (except
for the most dilute) crystallize readily at ambient temperature.
LiPF 6 and LiTFSI are found to be highly dissociated and dissociated,
respectively, in dilute mixtures. When CIP or AGG solvates do form,
bidentate coordination of Li + cations is more prominent for TFSI −
than for PF − 6 anions. QC calculations and MD simulations have been
used to extensively complement this study. Taken together, the thermal
phase behavior and spectroscopic analysis, when paired with the
structural information provided by the crystalline solvates and simulations,
give tremendous insight into the solution structure of these
electrolytes.
Supporting Information
Supporting Information files include expanded views of the DSC
data for the (AN) n -LiTFSI and LiPF 6 mixtures. In addition, Li +
cation coordination and ion packing diagrams are provided for the
(AN) 1 :LiTFSI, (AN) 6 :LiPF 6 and (AN) 5 :LiPF 6 solvate crystal structures.
Files are also provided for the MD simulations in .xyz format
(text files). These files are linked to the electronic version of this
manuscript. 67
Acknowledgments
The authors wish to express their gratitude to the U.S. Department
of Energy, Office of Basic Energy Sciences, Division of Materials
Sciences and Engineering which fully supported the experimental research
under Award DE-SC0002169. The computational work was
partially supported by an Interagency Agreement between the U.S.
Department of Energy and the U. S. Army Research Laboratory under
DE-IA01-11EE003413 for the Office of Vehicle Technologies
Programs including the Batteries for Advanced Transportation Technologies
(BATT) Program.
References
1. W. A. Henderson, J. Phys. Chem. B, 110, 13177 (2006).
2. D. M. Seo, O. Borodin, S.-D. Han, Q. Ly, P. D. Boyle, and W. A. Henderson, J.
Electrochem. Soc., 159, A553 (2012).
3. Y. Abu-Lebdeh and I. Davidson, J. Electrochem. Soc., 156, A60 (2009).
4. M. Nagahama, N. Hasegawa, and S. Okada, J. Electrochem. Soc., 157, A748 (2010).
5. Y. Abu-Lebdeh and I. Davidson, J. Power Sources, 189, 576 (2009).
6. M. J. Frisch et al., Gaussian, Inc., Wallingford CT, 2010.
7. J. B. Hooper and O. Borodin, Phys. Chem. Chem. Phys., 12, 4635 (2010).
8. O. Borodin, J. Phys. Chem. B, 113, 11463 (2009).
9. O. Borodin and G. D. Smith, J. Phys. Chem. B, 113, 1763 (2009).
10. M. J. Barrow, Acta Crystallogr. B, 37, 2239 (1981).
11. O. K. Antson, K. J. Tilli, and N. H. Andersen, Acta Crystallogr. B, 43, 296
(1987).
12. B. H. Torrie and B. M. Powell, Mol. Phys., 75, 613 (1992).
13. H. Abramczyk and K. Paradowska-Moszkowska, Chem. Phys., 265, 177 (2001).
14. R. Enjalbert and J. Galy, Acta Crystallogr. B, 58, 1005 (2002).
15. Y. Suzuki, M. Sato, K. Takanohashi, T. Ida, and M. J. Mizuno, J. Phys. Chem. A, 112,
13481 (2008).
16. A. Olejniczak and A. Katrusiak, J. Phys. Chem. B, 112, 7183 (2008).
17. S. Hore, R. Dinnebier, W. Wen, J. Hanson, and J. Maier, Z. Anorg. Allg. Chem., 635,
88 (2009).
18. D. Brouillette, D. E. Irish, N. J. Taylor, G. Perron, M. Odziemkowski, and
J. E. Desnoyers, Phys. Chem. Chem. Phys., 4, 6063 (2002).
19. W. Kunz, J. Barthel, L. Klein, T. Cartailler, P. Turq, and B. Reindl, J. Soln. Chem.,
20, 875 (1991).
20. E. M. Cabaleiro-Lago and M. A. Ríos, Chem. Phys., 254, 11 (2000).
21. X. Xuan, H. Zhang, J. Wang, and H. Wang, J. Phys. Chem. A, 108, 7513 (2004).
22. D. Spångberg and K. Hermansson, Chem. Phys., 300, 165 (2004).
23. E. Pasgreta, R. Puchta, A. Zahl, and R. van Eldik, Eur. J. Inorg. Chem., 1815 (2007).
24. Y. Yokota, V. G. Young Jr., and J. G. Verkade, Acta Crystallogr. C, 55, 196 (1999).
25. C. L. Raston, C. R. Whitaker, and A. H. White, Aust. J. Chem., 42, 201 (1989).
26. D. M. Seo, P. D. Boyle, and W. A. Henderson, Acta Crystallogr. E, 67, m534 (2011).
27. D. M. Seo, P. D. Boyle, O. Borodin, and W. A. Henderson, RSC Adv., in-press (2012).
28. D. M. Seo, P. D. Boyle, and W. A. Henderson, Acta Crystallogr. E, 67, m1148 (2011).
29. J. R. Black, W. Levason, and M. Webster. Acta Crystallogr. C, 51, 623 (1995).
30. L. A. Dakin, P. C. Ong, J. S. Panek, R. J. Staples, and P. Stavrospoulos,
Organometallics, 19, 2896 (2000).
31. I. Csöregh, P. Kierkegaard, and K. Norrestam, Acta Crystallogr. B, 31, 314 (1975).
32. G. A. Bowmaker, D. S. Gill, B. W. Skelton, N. Somers, and A. H. White, Z. Naturforsch.,
B59, 1307 (2004).
33. E. K. Beloglazkina, A. V. Shimorsky, A. G. Mazhuga, O. V. Shilova, V. A. Tafeenko,
and N. V. Zyk, Russ. J. Gen. Chem., 79, 1504 (2009).
34. H.-G. Hao, X.-D. Zheng, and T.-B. Lu, Angew. Chem. Int. Ed., 49, 8148 (2010).
35. P. G. Jones and O. Crespo, Acta Crystallogr. C, 54, 18 (1998).
36. J. W. Bats, T. Kretz, and H.-W. Lerner, Acta Crystallogr. C, 65, m94 (2009).
37. R. D. Shannon, Acta Crystallogr. A, 32, 751 (1976).
38. P. Johansson, S. P. Gejji, J. Tegenfeldt, and J. Lindgren, Electrochim. Acta, 43, 1375
(1998).
39. M. Herstedt, W. A. Henderson, M. Smirnov, L. Ducasse, L. Servant, D. Talaga, and
J.-C. Lassègues, J. Mol. Struct., 783, 145 (2006).
40. P. Neelakantan, P. Indian Acad. Sci. A, 60, 422 (1964).
[http://www.springerlink.com/content/t428037518568053/
41. G. Fini and P. Mirone, Spectrochim. Acta A, 32, 439 (1976).
42. J. Barthel, R. Buchner, and E. Wismeth, J. Soln. Chem., 29, 937 (2004).
43. J. M. Alia, H. G. M. Edwards, and J. Moore, Spectrochim. Acta A, 51, 2039 (1995).
44. J.-S. Seo, B.-S. Cheong, and H.-G. Cho, Spectrochim. Acta A, 58, 1747 (2002).
45. B. G. Oliver and G. J. Jantz, J. Phys. Chem., 74, 3819 (1970).
46. M. G. Davidson, P. R. Raithby, A. L. Johnson, and P. D. Bolton, Eur. J. Inorg. Chem.,
18, 3445 (2003).
47. Q. Zhou, P. D. Boyle, L. Malpezzi, A. Mele, J.-H. Shin, S. Passerini, and
W. A. Henderson, Chem. Mater., 23, 4331 (2011).
48. Q. Zhou, K. Fitzgerald, P. D. Boyle, and W. A. Henderson, Chem. Mater., 22, 1203
(2010).
49. W. A. Henderson, F. McKenna, M. A. Khan, N. R. Brooks, V. G. Young Jr., and
R. Frech, Chem. Mater., 17, 2284 (2005).
50. I. Rey, P. Johansson, J. C. Lassègues, J. Grondin, and L. Servant, J. Phys. Chem. A,
102, 3249 (1998).
51. M. Herstedt, M. Smirnov, P. Johansson, M. Chami, and J. Grondin, J. Raman Spectrosc.,
36, 762 (2005).
52. L. Xue, C. W. Padgett, D. D. DesMarteau, and W. T. Pennington, Solid State Sciences,
4, 1535 (2002).
53. A. M. Heyns and D. de Waal, J. Chem. Phys., 97, 8086 (1992).
54. X. Xuan, J. Wang, and H. Wang, Electrochim. Acta, 50, 4196 (2005).
55. R. Aroca, M. Nazri, G. A. Nazri, A. J. Camargo, and M. Trsic, J. Soln. Struct., 29,
1047 (2000).
56. A. M. Heyns, Spectrochim. Acta A, 33, 315 (1977).
57. C. M. Burba and R. Frech, J. Phys. Chem. B, 109, 15161 (2005).
58. J. Grondin, L. Ducasse, J.-L. Bruneel, L. Servant, and J.-C. Lassègues, Solid State
Ionics, 166, 441 (2004).
59. P. Johansson, Phys. Chem. Chem. Phys., 9, 1493 (2007).
60. O. Borodin, G. D. Smith, and R. L. Jaffe, J. Comput. Chem., 22, 641 (2001).
61. S. E. Popov, A. E. Nikiforov, O. V. Bushkova, and V. M. Zhukovskii, Russ. J. Electrochem.,
41, 546 (2005).
62. O. Borodin and G. D. Smith, J. Phys. Chem. B, 110, 6293 (2006).
63. S. P. Gejji, C. H. Suresh, K. Babu, and S. R. Gadre, J. Phys. Chem. A, 103, 7474
(1999).
64. J. C. Lassegues, J. Grondin, C. Aupetit, and P. Johansson, J. Phys. Chem. A, 113, 305
(2009).
65. O. Borodin and G. D. Smith, J. Phys. Chem. B, 110, 4971 (2006).
66. O. Borodin, G. D. Smith, and W. Henderson, J. Phys. Chem. B, 110, 16879 (2006).
67. See Supporting Information at http://dx.doi.org/10.1149/2.035209jes.html.