Comprehensive Insights into the Reactivity of Electrolytes Based on Sodium Ions

Article abstract of DOI:10.1002/cssc.201501605

We report a systematic investigation of Na-based electrolytes that comprise various NaX [X=hexafluorophosphate (PF₆), perchlorate (ClO₄), bis(trifluoromethanesulfonyl)imide (TFSI), fluorosulfonyl-(trifluoromethanesulfonyl)imide (FTFSI), and bis(fluorosulfonyl)imide (FSI)] salts and solvent mixtures [ethylene carbonate (EC)/dimethyl carbonate (DMC), EC/diethyl carbonate (DEC), and EC/propylene carbonate (PC)] with respect to the Al current collector stability, formation of soluble degradation compounds, reactivity towards sodiated hard carbon (Na_x-HC), and solid-electrolyte interphase (SEI) layer formation. Cyclic voltammetry demonstrates that the stability of Al is highly influenced by the nature of the anions, solvents, and additives. GC-MS analysis reveals that the formation of SEI telltales depends on the nature of the linear alkyl carbonates and the battery chemistry (Li⁺ vs. Na⁺). FTIR spectroscopy shows that double alkyl carbonates are the main components of the SEI layer on Na_x-HC. In the presence of Na salts, EC/DMC and EC/DEC presented a higher reactivity towards Na_x-HC than EC/PC. For a fixed solvent mixture, the onset temperature follows the sequence NaClO₄646

Full text of DOI:10.1002/cssc.201501605

DOI: 10.1002/cssc.201501605
Full Papers
Very Important Paper
Comprehensive Insights into the Reactivity of Electrolytes
Based on Sodium Ions
Gebrekidan Gebresilassie Eshetu,[a, b] Sylvie Grugeon,[c, d] Huikyong Kim,[c, e] Sangsik Jeong,[a, b]
Liming Wu,[a, b] Gregory Gachot,[c, d] Stephane Laruelle,[c, d] Michel Armand,[f] and
Stefano Passerini*[a, b]
We report a systematic investigation of Na-based electrolytes
that comprise various NaX [X=hexafluorophosphate (PF6), per-
chlorate (ClO4), bis(trifluoromethanesulfonyl)imide (TFSI), fluo-
rosulfonyl-(trifluoromethanesulfonyl)imide (FTFSI), and bis-
(fluorosulfonyl)imide (FSI)] salts and solvent mixtures [ethylene
carbonate (EC)/dimethyl carbonate (DMC), EC/diethyl carbon-
ate (DEC), and EC/propylene carbonate (PC)] with respect to
the Al current collector stability, formation of soluble degrada-
tion compounds, reactivity towards sodiated hard carbon (Nax-
HC), and solid–electrolyte interphase (SEI) layer formation.
Cyclic voltammetry demonstrates that the stability of Al is
highly influenced by the nature of the anions, solvents, and ad-
ditives. GC–MS analysis reveals that the formation of SEI tell-
tales depends on the nature of the linear alkyl carbonates and
the battery chemistry (Li+ vs. Na+). FTIR spectroscopy shows
that double alkyl carbonates are the main components of the
SEI layer on Nax-HC. In the presence of Na salts, EC/DMC and
EC/DEC presented a higher reactivity towards Nax-HC than EC/
PC. For a fixed solvent mixture, the onset temperature follows
the sequence NaClO4 <NaFTFSI<NaPF6 <NaTFSI<NaFSI, and
the total heat generated increases in the order NaFSI<
NaTFSI<NaClO4 <NaPF6 <NaFTFSI.
Introduction
Amongst post-lithium energy storage devices, the sodium ion
battery (NIB) appears to be the most appealing and viable
technology.[1–6] Unlike Li, Na mineral deposits are unlimited, at-
tainable at low cost, and distributed evenly. If the practical de-
ployment of NIBs could be realized, there will be approximate-
ly a three-orders of magnitude relaxation in the constraints on
resources (Table 1), accompanied by sustainability, improved
environmental benignity, and cost reduction.[7] Besides these
salient features, NIBs also offer the advantage of using electro-
lyte systems with a lower decomposition potential because of
the higher half-reaction potential of Na than Li. The desolva-
tion energy of Na+ is generally ꢀ40–70 kJmolÀ1, lower than
that of Li+, which implies a lower activation barrier for Na (de)-
insertion and thereby faster charge/discharge characteristics.[8]
However, the projected gravimetric and volumetric energy
densities of NIBs are inevitably penalized by the inherent char-
acteristics of Na, such as larger ionic radii (>30% higher than
Li), atomic weight (>three times that of Li), and reduction po-
tential (ꢀ0.33 V vs. Li/Li+). Another anticipated challenge, even
more than that of LIBs, is long-term stability because of the
large size of Na, which causes a greater volume change during
Na (de)insertion into host materials. Hence, to compete and go
beyond the currently prevailing Li-ion technology, significant
progress in the understanding of the electrode materials (both
active and inactive components), electrolytes (salts, solvents,
additives), and interfacial dynamics is needed urgently.
[a] Dr. G. G. Eshetu, Dr. S. Jeong, Dr. L. Wu, Prof. Dr. S. Passerini
Helmholtz Institute of Ulm (HIU)
Helmholtz Str.11, 89081 Ulm (Germany)
E-mail: gebrekidan.eshetu@kit.edu
stefano.passerini@kit.edu
[b] Dr. G. G. Eshetu, Dr. S. Jeong, Dr. L. Wu, Prof. Dr. S. Passerini
Karlsruhe Institute of Technology (KIT)
P.O.Box 3640, 76021 Karlsruhe (Germany)
As in any electrochemical storage device, the scientific com-
munity has been active in the development of electrode mate-
rials for NIBs, however, studies that deal with Na+ electrolytes,
interphases, and their reactivity in the vicinity of the electrified
surface with (de)sodiated electrodes are few. Yet, the electro-
lyte should be considered equally as it is largely responsible
for the lifetime and the credibly possible performances, in
terms of practically accessible capacity, chemical/thermal stress
(safety), rate capability, and so on. Moreover, although the ex-
perience and knowhow accumulated from LIBs could help in
the development of electrolytes and the understanding of in-
terfaces for NIBs, it does not guarantee that electrolyte sol-
vents suitable for LIBs are also suitable for NIBs. Thus, one
[c] Dr. S. Grugeon, Dr. H. Kim, Dr. G. Gachot, Prof. Dr. S. Laruelle
Laboratoire de RØactivitØ et Chimie des Solide (LRCS)
CNRS UMR 7314 UniversitØ de Picardie Jules Verne
33 rue Saint Leu, 80039 Amiens (France)
[d] Dr. S. Grugeon, Dr. G. Gachot, Prof. Dr. S. Laruelle
RØseau sur le Stockage Electrochimique de l’Energie
CNRS RS2E FR3459 (France)
[e] Dr. H. Kim
Renault, DEA-IREB, Renault, Technocentre
1 avenue du Golf, 78084 Guyancourt (France)
[f] Prof. Dr. M. Armand
CIC Energigune, Parque Tecnologico de Alava
Albert Einstein 48, ED. CIC, 01510, Minano (Spain)
ORCID(s) from the author(s) for this article is/are available on the WWW
under http://dx.doi.org/10.1002/cssc.201501605.
ChemSusChem 2016, 9, 462 – 471
462
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 1. Comparison of Li and Na.
Characteristics
Na
Li
ratio of reserves
price of carbonates [US $ per ton]
1000
150
1
5000
cathode current collectors
Al=1920[a]
23
4.44
1165
Cu=6755[a]
6.9
1.84
3829
atomic weight [gmolÀ1
ionic volume [Š]
]
theoretical capacity [mAhgÀ1
voltage [V] vs. SHE[b]
melting point [8C]
]
À2.714
À3.045
98
182
desolvation energy [kJmolÀ1] in
DEC
EC
PC
148
152
158
208
211
218
coordination preference
reactivity/flammable
distribution
octahedral and prismatic
highly flammable
everywhere
octahedral and tetrahedral
less reactive
70% in South America
[a] In US$ per ton. [b] SHE=standard hydrogen electrode.
should not assume a simple transition of knowledge from Li+
to Na+ chemistries. For instance, although ethylene carbonate
(EC)/dimethyl carbonate (DMC) and EC/diethyl carbonate (DEC)
are preferred solvent mixtures in LIBs, EC/propylene carbonate
(PC) has been described as the most stable and suitable for
NIBs. Hence, a systematic screening, formulation, and optimiza-
tion of electrolytes along with the evaluation of their interfacial
properties is needed.
(ARC). In this study, a NaPF6-based electrolyte was more reac-
tive than NaTFSI towards sodiated HC because of the high
thermal stability of NaPF6, which results in the absence of NaF
in the SEI layer. Moreover, in the presence of NaPF6, DMC and
DEC are more reactive than EC if in contact with Nax-HC be-
cause of the preferential solvation of Na+ by EC, which leaves
DMC and DEC in the outer sphere.
Ponrouch et al.[13] performed an exhaustive evaluation of the
thermal stability of a number of electrolytes, such as 1m
NaClO4 in EC/DEC, PC, and EC/PC and compared them with
1m NaPF6/EC/PC. NaPF6/EC/PC presented the highest onset
temperature and lowest heat release (ꢀ717 JgÀ1), comparable
to fully lithiated graphite (ꢀ700 JgÀ1, for 1m LiPF6 in EC/DMC).
However, Zhao et al.[14] made a comparative investigation on
the reactivity of 1m NaX (X=ClO4 and PF6) in EC/DMC, PC, and
their Li counterparts, which demonstrated that Na-based elec-
trolytes present a higher onset temperature and lower heat re-
lease to lead to the conclusion that NIBs could be safer battery
systems than LIBs.
An understanding of the reactivity of electrolytes in the vi-
cinity of the electrified interface is crucial as it plays a vital role
to improve the irreversible capacity, interfacial resistance,
cycle life, and safety of the battery system. Once again, the in-
triguing question here is whether the replacement of Li+ with
Na+ would result in a similar solid–electrolyte interphase (SEI)
layer, for example, with the same nature, composition, and re-
activity, or not? Although one would initially assume similar
trends, some of the peculiar characteristics, such as the larger
size of Na+, higher solubility of Na-based SEI components/
degradation products (e.g., Na2CO3 vs. Li2CO3), lower Lewis
acidity and higher reduction potential of Na+, could lead to
a quite different SEI layer and preferential reactivity of solvent
molecules.
If we consider these previous reports, to obtain reliable data
would require an in-depth and systematic investigation of Na+-
based electrolyte systems. However, to arrive at representative
conclusions, it requires data to be obtained from various elec-
trolyte formulations that comprise of different salts and solvent
blends.
Although their number is increasing at a fast pace, quite few
reports detail the reactivity of the electrolytes and the investi-
gation of the SEI layer in NIBs. Komaba et al.[9] conducted a de-
tailed investigation of surface films formed on Nax-hard carbon
(HC) by using X-ray photoelectron spectroscopy (XPS) and
time-of-flight secondary ion mass spectrometry (TOF-SIMS) and
obtained the same composition as for graphite in LIBs, that is,
mainly M2CO3, ROCO2M, CH2À, and ÀCOÀOÀ ester linkages
(M=Li or Na). It was also reported that the SEI on HC in NIBs
contains many inorganic species that are nonhomogeneous
and porous in nature. Xia et al.[10–12] studied the thermal re-
sponse of Nax-HC in the presence of various salts, such as
NaPF6 and sodium bis(trifluoromethanesulfonyl)imide (NaTFSI),
and solvent mixtures by using accelerated-rate calorimetry
In recognition of the existing gap and upcoming new op-
portunities with NIBs, a systematic and detailed comparative
investigation on diverse Na-based electrolytes that comprise
various NaX [X=PF6, ClO4, TFSI, fluorosulfonyl-(trifluorometha-
nesulfonyl)imide (FTFSI), and bis(fluorosulfonyl)imide (FSI)] salts
and solvent mixtures (EC/DMC, EC/DEC, EC/PC) was performed.
Moreover, ionic liquids (ILs) were utilized for anodic Al stability
comparison studies. A number of analytical tools were used,
such as cyclic voltammetry (CV), differential scanning calorime-
try (DSC), gas chromatography–mass spectrometry (GC–MS),
and Fourier transform infrared (FTIR) spectroscopy.
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
463
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
The first part of this paper deals with the study of the intrin-
sic Al stability in the aforesaid electrolyte solutions. The second
part addresses the trend in the generation of soluble degrada-
tion products that originate from the reduction of linear sol-
vents. The last section focuses mainly on the investigation of
the reactivity of sodiated HC with various Na+-based electro-
lytes by using DSC to extract information about the nature
and growth of the SEI layer, the origin of SEI layer disruption,
and the accompanying thermal stabilities of the various elec-
trolyte solutions.
Table 2. Maximum anodic current density [mAcmÀ2] evolution during cy-
cling recorded at 5.3 V vs. Na/Na+.
Electrolyte formulation
CV cycle
5th
1st
10th
NaPF6/EC/DEC
NaClO4/EC/DEC
NaFSI/EC/DEC
NaFSI/EC/DEC+5% NaPF6
NaFSI/Pyr14FSI
5.3”10À2
2.1”10À2
1.61”10À2
3.47
1.8”10À2
1.38”10À2
3.8”10À1
1.24
5.35”10À3
7.5”10À1
2.2”10À1
6.01”10À3
2.1
6.7”10À1
1.98”10À3
3.62”10À2
29.5
6.34
2.07”10À2
15.9
4.19”10À1
1.6”10À2
3.65
8.98”10À1
9.38”10À3
4.5
6.6”10À3
3.43
NaFTFSI/EC/DEC
NaFTFSI/EC/DEC+5% NaPF6
NaFTFSI/Pyr14FTFSI
NaTFSI/EC/DEC
NaTFSI/EC/DEC+5% NaPF6
NaTFSI/Pyr14TFSI
2.91”10À1
7.02”10À3
2.88
7.8”10À1
3.22”10À3
Results and Discussion
Comparative anodic stability of the Al current collector
One of the major benefits of NIBs is the possible use of Al foil
as anode and cathode current collectors; thus, it is required
that the electrolytes for NIBs have a good compatibility with it.
Accordingly, CVs were collected from Al/Na cells on diverse
electrolytes, namely, 1) 1m NaX (X=PF6, ClO4, TFSI, FTFSI, FSI)
in EC/DEC, 2) NaTFSI, NaFTFSI, and NaFSI in EC/DEC+5%
NaPF6, and 3) NaTFSI/Pyr14TFSI, NaFTFSI/Pyr14FTFSI, and NaFSI/
Pyr14FSI at 208C (Pyr14 =N-butyl-N-methylpyrrolidinium (Fig-
ure 1i–iv), to evaluate the anodic stability of Al. A comparison
of the different salt anions in EC/DEC showed that the trend in
the Al dissolution and/or anion decomposition increases in the
order NaPF6 <NaClO4 <NaTFSI<NaFTFSI<NaFSI (Figure 1i).
Imide-based electrolytes showed a hysteresis loop that was ini-
tiated at ꢀ3.25, 3.31, and 3.45 V for FSIÀ, FTFSIÀ, and TFSIÀ, re-
spectively, with a large irreversible oxidative current (e.g., up
to 30 mAcmÀ2 for FSIÀ) at 5.3 V vs. Na/Na+. However, a fast de-
crease in the current intensity is observed upon cycling for all
electrolytes, which may be associated with the formation of
a passivation layer during the first cycle, which in turn limits
the reactivity of the Al working electrode (Table 2).
The stability of Al in imide-based electrolytes could be im-
proved by the addition of a small amount of NaPF6 (5 wt%) to
the base electrolytes, which is because of the formation of flu-
oroaluminates (AlF3 or AlOxFy) on the Al surface (Figure 1ii–iv).
This observation is in line with that reported for LiTFSI/EC/
DMC,[15] NaTFSI/PC,[13] NaFSI/PC, and NaFSI/C3C1PyrFSI.[16]
To gain an insight into the effect of solvent on the anodic
stability of the Al current collector, CVs were measured in elec-
trolyte solutions based on IL/(NaFSI, NaTFSI, and NaFTFSI). Con-
trary to carbonate-based electrolytes, IL-based electrolytes sup-
press Al corrosion, which is attributed to the formation of
a protective passivation layer. Kühnel et al.[17] showed that the
solubility of Al(TFSI)3 in LiTFSI/PC is considerably higher than
that in LiTFSI/Pyr14TFSI, which implies that Al(TFSI)3 effectively
serves as a passivation layer in IL-based electrolytes to prevent
Al corrosion but not in organic carbonate-based electrolytes.
The improvement with other IL-based electrolytes is also be-
cause of the effective protection of the assumed formation of
Al(FSI)3 and Al(FTFSI)3, which depends on their stability and, es-
pecially, their solubility in the different electrolyte formulations.
Although there is no documented data on the solubility or ex-
istence of Al(FSI)3 and Al(FTFSI)3 in EC/DEC, Pyr14FSI, or
Pyr14FTFSI, on the basis of comparison with the behavior of
LiTFSI dissolved in carbonate solvents and ILs, it is reasonable
to assume that the layers will be more stable in the ILs than in
carbonate-based solvents.
In general, Al corrosion by NaA (A=FSI, TFSI, and FTFSI)
salts can be represented schematically by a set of simple equa-
tions [Eqs. (1)–(3)]:
NaA Ð NaþþAÀ
3 AÀþAl Ð AlðAÞ3þ3 eÀ
AlðAÞ3 Ð 3 AÀþAl3þ
ð1Þ
ð2Þ
ð3Þ
The equilibrium in Equation (3) determines the degree of
solubility of the layers and their effectiveness. Thus, on the
basis of the difference in the current density in the first CV
cycle (as indicated in Figure 1ii–iv and Table 2), it is reasonable
to assume that the solubility of the layers increases in the
order of Al(TFSI)3 <Al(FTFSI)3 <Al(FSI)3 both in carbonate- and
Figure 1. CVs measured on Al by using three-electrode Swagelok cells with
i) a) NaFSI/EC/DEC, b) NaFTFSI/EC/DEC, c) NaTFSI/EC/DEC, d) NaClO4/EC/DEC,
and e) NaPF6/EC/DEC; ii) NaTFSI in a) EC/DEC, b) EC/DEC+5% NaPF6, and
c) Pyr14TFSI; iii) NaFTFSI in a) EC/DEC, b) EC/DEC+5% NaPF6, and
c) Pyr14FTFSI; and iv) NaFSI in a) EC/DEC, b) EC/DEC+5% NaPF6, and
c) Pyr14FSI.
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
464
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
IL-based electrolytes. Beyond the first CV cycle, the combined
effect of the Al–imide anion complex and the electrolyte de-
composition layer could explain the attenuation of the Al dis-
solution with subsequent cycles.
Formation of SEI telltales and SEI investigation
Studies that deal with electrolyte degradation remain at the
frontline of the research endeavors in LiBs because of the fore-
seeable electrolyte degradation presumably at the negative
electrode. The formation of a robust SEI layer upon the initial
charging can, more or less, limit the continuous degradation of
electrolytes during the life of the battery system. Of the differ-
ent approaches used to monitor the evolution of the SEI layer,
the identification of soluble compounds formed concomitantly
could help us to understand the whole electrolyte degradation
process. The analysis of the ethylene biscarbonates (11 and 1’1)
as telltales of the SEI (Scheme 1) help us to monitor the extent
of linear carbonate solvent reduction and thereby the effec-
tiveness of the passivation layer as recently indicated by Kim
et al.[18] in the case of lithiated graphite. The question arose as
to whether or not the soluble compounds detected after the
first full charge (Nax-HC) of the HC composite were the same
in nature and content as in those in the case of Li salt-based
electrolytes if we consider the 0.33 V potential difference be-
tween Li and Na. Hence, 1m NaPF6 in EC/DMC and EC/DEC
electrolytes recovered from a charged cell that consists of
a HC composite powder negative electrode and a NaN3 plastic
film positive electrode were analyzed by using GC–MS. The
NaN3 sacrificial salt was used to replace the metallic Na elec-
trode known to contaminate the electrolyte with the same
soluble degradation compounds. The results reveal the pres-
ence of the respective ethylene biscarbonates 11 and 1’1 (Figur-
es 2a and b) as in the case of Li salt electrolyte, which is indi-
cative of the same electrochemical processes. The 11/EC and
1’1/EC peak area ratios (ꢀ12.5 and 4.1%, respectively) are
quite reproducible, and although the values are low, the trend
in the quantity of biscarbonates (11 >1’1) generated corrobo-
rates the data obtained by the extrapolation of results ob-
tained with a Li electrolyte at 0.33 V vs. Li/Li+ (Figures 2c and
d), that is, 20 and 5.1% for EC/DMC and EC/DEC, respectively.
Figure 2. Liquid GC–MS chromatograms of 1m NaPF6 in EC/DMC and EC/
DEC electrolytes measured after the first sodiation of HC in 1m NaPF6 in
a) EC/DMC and b) EC/DEC. Values provide the integrated intensity of each
area of 1(’)1, compared to the integrated intensity of EC. Liquid GC–MS anal-
ysis of the electrolytes recorded during the first lithiation of graphite in
c) 1m LiPF6 in EC/DMC and d) 1m LiPF6 in EC/DEC.[18]
The main difference comes from the potential at which the
soluble product is formed, that is, although the maximum pro-
duction of 1’1 in EC/DEC appears near the lithiation stage
(ꢁ0.2 V vs. Li/Li+), in NIB, the DEC reduction is limited because
of the higher half-reduction potential of Na than Li. For 11, the
highest quantity is generated in the 0.9–0.6 V vs. Li/Li+ poten-
tial range, and hence its generation is not limited by potential
in the case of Na salt-based electrolytes. We infer that the
nature and effectiveness of the SEI layer in both LIB and NIB
could be expected to be different, which impacts the per-
formance and safety of NIBs.
As in the case of LIBs, the evaluation of soluble compounds
can also help to screen effective electrolyte additives in NIBs as
Scheme 1. Elementary steps for biscarbonate generation in linear carbonate/EC-based electrolytes.
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
465
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
the inhibition of the generation of the former implies the ef-
fectiveness of the additives.
Lastly, a preliminary investigation on the composition of the
SEI layer formed on desodiated Nax-HC in contact with 1m
NaPF6 in 1) EC/DMC, 2) EC/DEC, and 3) EC/PC was performed
by using FTIR spectroscopy. The FTIR spectra (Figure 3) evi-
dence the presence of sodium double alkyl carbonate (NEDC),
a 1eÀ reduction product of EC, as the main component of the
SEI layer, from the bands at n˜ =1655, 1390, 1440, 1300, 1080,
and 820 cmÀ1. Na2CO3 with bands at n˜ =1440 and 870 cmÀ1 is
also part of this SEI but in a lower quantity in the case of linear
carbonate/EC blends.
Figure 4. TGA profiles of neat Na salts: NaPF6, NaClO4, NaTFSI, NaFTFSI, and
NaFSI. For comparison, the TGA curve of the state-of-the-art salt in LIBs LiPF6
is also presented.
at 2508C, whereas NaPF6 does not show any noticeable loss
up to 3008C. Thus, besides the gain in intrinsic safety, the
higher thermal stability of NaPF6 indicates that NIBs might per-
form much better than LIBs in terms of capacity retention
during cycling and storage at higher temperatures. Capacity
fading at elevated temperature is one of the most critical limi-
tations of LIBs, which is mainly attributed to the thermal insta-
bility of LiPF6-based electrolytes and leads to the generation of
Lewis acids that, in turn, catalyzes parasitic cascade reactions
inside the cell.
Figure 3. FTIR spectra of fully sodiated HC after one cycle using 1m NaPF6 in
a) EC/DMC, b) EC/DEC, and c) EC/PC.
The melting points of all of the Na salts measured in this
study are generally higher than those reported in literature,[7]
which demonstrates their high purity. Furthermore, the higher
melting point of Na salts makes them easier to dry than the Li
equivalents.
Reactivity of Na+-based electrolytes
Thermal stability of neat Na salts
A stability study of a variety of Na salts was performed 1) to es-
tablish a database on the thermal properties of the salts and
2) to investigate their effect on the thermal reactivity of Nax-
HC. Notably, the thermogravimetric analysis (TGA) traces
shown in Figure 4 indicated that the thermal stability of the
neat salts decreases in the order NaClO4 >NaTFSI>NaPF6 >
NaFTFSI>NaFSI. NaClO4 outperforms all of the other salts in
terms of thermal stability. However, although NaClO4 is the
most commonly used, it is restricted to academic interest be-
cause of its explosive nature in the dry state.
Thermal reactivity of Na-based electrolytes in contact with
Nax-HC
The first Na-insertion profiles of HC in various Na salts dis-
solved in an archetypal solvent mixture, EC/DEC, are shown in
Figure 5. The HC electrodes showed similar electrochemical
characteristics, with sharp slopes that start at 1.2 V, followed
by well-defined small plateaus at lower potentials (<ꢀ0.1 V)
attributed to Na insertion. As reported for LIB,[19,20] the heat re-
leased is a function of the degree of insertion (state of charge,
SOC), particularly for lower values of x in LixC6, which may also
be expected for NIBs. Thus, all the sodiated HCs are kept at
nearly the same capacity to ensure a realistic comparison of
the inherent reactivity of the electrolytes.
The thermal properties of the Na salts, which include the
melting point (Tm) and decomposition temperature (Tdecom) ob-
tained from DSC and dynamic TGA measurements, respective-
ly, are summarized in Table 3. Values provided in brackets refer
to the corresponding Li salts. A simple comparison with the lit-
erature values indicates that Na salts are much more stable
than their homologous Li salts, which is explained by their
higher Magdelung energy, a parameter linked to the electro-
static energy in ionic crystals. For instance, the large difference
in the thermal stability of LiPF6 and NaPF6 is evidenced by their
decomposition, which starts at ꢀ125 and 3258C, respectively.
TGA measurements showed that LiPF6 loses ꢀ83% of its mass
Although neat Na salts are intrinsically stable, their fate is
capricious if they exist as electrolyte solutes in contact with so-
diated HC electrodes. As a figure of merit to compare the reac-
tivity of each solvent mixture and salt, it is imperative to con-
sider three parameters, namely, the exothermic onset tempera-
ture (related to SEI layer cracking), the temperature at which
the maximum heat is released, and the total heat generated
DH [JgÀ1] by the electrode/electrolyte reactions. Although the
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
466
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Table 3. Melting and decomposition temperatures of Na salts for NIB electrolytes from DSC and TGA measurements, respectively (heating rate: 108CminÀ1
except for NaPF6, which was at 18CminÀ1). The melting points indicated in brackets refer to the corresponding Li salts.
Na salt
Anion chemical structure
Tm [8C]
TGA [8C] (% mass loss)
(Li salt)
NaPF6
302 (200)
474 (236)
400 (8.14%), 535 (74.56%)
500 (0.09%), 555 (2.2%)
NaClO4
NaTFSI
263 (234)
400 (3.21%), 520 (85.50%)
NaFTFSI
NaFSI
160 (94.5)
122 (130)
300 (2.75%), 420 (64.22%)
300 (16.15%), 400 (57.96%)
investigated and is reported to be in the order of EC (1.4m)>
PC (1.0m)>DEC (0.8m)>DMC (0.6m),[10] which implies that
NaPF6 is coordinated more strongly by EC followed by PC in
the electrolytes.
Shakourian-Fard et al.,[21] studied the trends in the Na+ solva-
tion of a variety of single solvents (EC, DMC, DEC, and EMC)
and mixtures (EC/PC, EC/DMC, EC/DEC, and EC/EMC) by using
classical molecular dynamics simulation and hybrid DFT calcu-
lations. A comparison of the Na+(EC), Na+(PC), Na+(DEC), and
Na+(DMC) complexes evidenced the highest number of EC
molecules in the first solvation sphere followed by PC and
then the linear carbonates, the number of which decreases
with increasing size. In EC/PC, three EC and two PC molecules
were found in the primary solvation sphere. With EC/DMC,
only EC molecules were found in the primary solvation sphere,
whereas DMC was found in the vicinity of the highly reducing
interface because of its weaker interaction with Na+ ions.
The improved thermal stability of EC/PC could be attributed
to the intrinsic nature of the single solvents (e.g., their high
polarity), in which both EC and PC are coordinated preferen-
tially by Na+, which is less able to move freely and react with
Nax-HC. If NaPF6-containing electrolytes are made with EC/DMC
or EC/DEC, Na+ coordinates predominantly to EC to leave the
more reactive DMC and DEC available to diffuse through the
porous SEI layer and thereby react with the highly reducing so-
diated HC at lower temperatures.
Figure 5. Potential–composition profiles recorded upon the first sodiation of
HC in 1m NaX (X=PF6, ClO4, TFSI, FTFSI, and FSI) in EC/DEC (1:1 weight
ratio).
onset exothermic temperature could be regarded as the first
possible step for thermal runaway, the amount of heat re-
leased dictates the magnitude of the electrolyte decomposi-
tion in contact with Nax-HC.
The rate of exothermic heat release of NaPF6-, NaClO4-, and
NaTFSI-based electrolytes in contact with Nax-HC decreases de-
pending on the solvent mixture in the order of EC/PC>EC/
DEC>EC/DMC (Figure 6a–c). Hence, regardless of the nature
of the salt, lower rates of exothermic heat release are recorded
for the electrolytes that contain EC/PC, which shows that this
mixture is suitable for safer solvent and NIB electrolyte formu-
lations. EC/DMC and EC/DEC showed nearly the same behavior
but much higher activities, which is linked to the higher reac-
tivity of the linear carbonates (DMC, DEC) towards sodiated
HC. The solubility of NaPF6 in various single solvents has been
In the case of NaFSI-based electrolytes, the exothermic peak
occurs at the same temperature for all solvent mixtures (Fig-
ure 6e), which shows that the thermal reactivity is mainly dic-
tated by FSIÀ reduction and has been reported for LiFSI-based
electrolytes.[20] Similar to LiFSI, NaFSI undergoes a two-step,
3À
four-electron reduction to produce a trication salt (N(SO2)2
)
and NaF (Scheme 2). Hence, the relatively strong and energetic
peak in the range of 197–2078C is caused by the reduction of
the FSIÀ, ascribed to the presence of the weak SÀF bond.
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
467
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 6. DSC profiles of Nax-HC and various electrolyte formulations (which enlist the effect of the nature of the solvents): a) 1m NaPF6, b) NaClO4, c) NaTFSI,
d) NaFTFSI, and e) NaFSI in EC/DMC, EC/DEC, and EC/PC.
Surprisingly, the heat released by 1m NaFSI in EC/DMC in
contact with sodiated HC is only approximately half of that of
1m LiFSI in EC/DMC in contact with lithiated graphite. This
might be because 1) there might not be free Na+ (eÀ) to be
utilized for the reduction of FSIÀ as Na+ is coordinated strongly
by EC and/or 2) the low degree of sodiation in HC may limit
the capacity.
which indicates that the FTFSI and TFSI anions are not able to
form a good SEI layer upon sodiation, although, according to
the procedure described by Eshetu et al.,[20] the FTFSI anion
can be reduced by the anthracene radical anion (ꢀ1.0 V vs. Li/
Li+). However, the lower thermal stability of FTFSI than TFSI
(Figure 4 and Table 3) results in a high heat release at lower
temperatures. The response of the sample in NaFSI is rather
different because this anion is able to form a robust SEI layer
that protects the electrolyte salt from reduction up to much
higher temperatures.
In the case of NaFTFSI (Figure 6d), the exothermic peaks
occur at a much lower temperature than that with NaFSI inde-
pendent of the solvent (or solvent mixture) used. This can be
explained if we consider that the thermal behavior of the FTFSI
anion (Figure 6d) resembles that of the TFSI anion (Figure 6c),
For a fixed solvent blend (e.g., EC/PC), the reactivity of Nax-
HC could be ranked in the order NaFSI>NaTFSI>NaPF6 >
Scheme 2. Proposed four-electron reduction of NaFSI.
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
468
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
NaFTFSI>NaClO4 with respect to the decreasing onset temper-
ature of the exothermal process or NaFTFSI>NaPF6 >NaClO4 >
NaTFSI>NaFSI in terms of the total heat generated (Fig-
ure 7a–c). Overall, NaFSI, NaTFSI, and NaPF6 seem to offer the
best performance if we consider the safety aspects of electro-
lytes for NIBs. However, NaFSI and NaTFSI are susceptible to Al
corrosion and hence may not be used as single salts in electro-
lyte formulations.
Thermal reactivity of model SEI compounds with Na salts/
electrolytes
The cascading thermal reactions of lithiated graphite are initi-
ated by SEI decomposition at ꢀ908C.[20,22,23] The origin of this
decomposition is reported to be either caused by an acid–
base reaction of the SEI compounds with Lewis acid species
(PF5, HF, POF3, etc.) [Eqs. (4)–(6)] and/or the transformation of
metastable SEI compounds into more stable ones [Eqs. (7) and
(8)].
Detailed thermal data such as the onset temperature of the
first predominant exothermic peak and the total normalized
heat generated for all the electrolyte formulations in contact
with fully sodiated HC are summarized in Table 4.
PF5þLi2CO3 ! 2 LiFþPOF3þCO2
ð4Þ
ð5Þ
ð6Þ
ð7Þ
ð8Þ
In general, two parameters are considered if we compare
the reactivity of electrolytes with Nax-HC and the analogous
LixC6: onset temperature of the exothermic reaction and total
heat generation. In terms of the former, LiBs are safer than
NIBs because the SEI layer of LIBs more robust. In terms of
total heat generation, however, NIBs appear to be slightly
better than LIBs, but the difference in the degree of lithiation
and sodiation prevents us from concluding whether Na- or Li-
based carbonaceous electrodes are safer in terms of the power
they release.
PF5þtrace H2O ! 2 HFþPOF3
2 HFþLi2CO3 ! 2 LiFþH2OþCO2
ðCH2OCO2LiÞ2 ! Li2CO3þC2H4þCO2þ1=2 O2
2 ROCO2Li ! Li2CO3þCO2þRÀRþ1=2 O2
If we consider the high thermal stability of NaPF6, however,
it would be irrational to assume that the generation of PF5 at
such a low temperature and the initiation of the SEI layer
Figure 7. DSC profiles of Nax-HC and various electrolyte formulations (which enlist the effect of the nature of the anions): 1m NaX (X=PF6, ClO4, TFSI, FTFSI,
and FSI) in a) EC/DMC, b) EC/DEC, and c) EC/PC.
Table 4. Summary of the thermal parameters extracted from the DSC measurements for all the above-mentioned electrolyte formulations in the presence
of fully sodiated HC.
Electrolyte system
salt
Onset temperature of first
exothermic peak [8C]
Tmax to first
peak [8C]
Total heat generated
[JgÀ1
]
solvent mixture
NaPF6
EC/DMC
EC/DEC
EC/PC
EC/DMC
EC/DEC
EC/PC
EC/DMC
EC/DEC
EC/PC
EC/DMC
EC/DEC
EC/PC
75
80
135
60
70
105
82
105
115
145
152
159
82
105
123
187
85
257
494
316
733
388
232
512
240
140
486
238
50
NaClO4
NaTFSI
NaFSI
90
170
144
169
201
197
202
207
144
169
201
NaFTFSI
EC/DMC
EC/DEC
EC/PC
715
538
414
105
115
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
469
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
cracking are caused by the aforementioned acid–base reac-
tions. To verify this hypothesis and thereby to gain more in-
sight, a DSC investigation of Na2CO3 mixed with NaPF6, NaClO4,
NaTFSI, NaFTFSI, and NaFSI was performed (Figure 8a–e). The
phase (SEI) telltales, and reactivity with sodiated hard carbon
(HC) enabled the screening of electrolytes for safe and per-
formance-inspired sodium ion battery (NIB) applications.
The results can be summarized as:
1) The stability of the Al current collector in various Na-
based electrolytes depends on both the salt anion and the sol-
À
vents. In particular, the PF6 anion in organic carbonate sol-
vents and TFSI-based IL solvents were the best electrolyte
choices.
2) The generation of soluble SEI telltales depends heavily on
the linear carbonates (DMC vs. DEC) and battery chemistry (Li+
vs. Na+). Unlike DMC, the reduction of DEC in NIBs was limited
by the higher half-reduction potential compared to that of lith-
ium ion batteries (LIBs).
3) The comparative reactivity study of the various Na-based
electrolytes in contact with Nax-HC shows that solvent blends
that contain linear carbonates present lower onset tempera-
tures for the exothermic process. Thus, for a fixed Na salt, the
reactivity of solvent mixtures decreases in the order of EC/
DMC>EC/DEC>EC/PC. The salt also plays a major role, and
the reactivity decreases in the order NaClO4 > NaFTFSI>
NaPF6 >NaTFSI>NaFSI for a fixed solvent mixture.
Figure 8. DSC profiles of powder and electrolyte mixtures (1:1 molar ratio)
of a) Na2CO3+NaFSI, b) Na2CO3+NaFTFSI, c) Na2CO3+NaClO4,
d) Na2CO3+NaTFSI, e) Na2CO3+NaPF6, and f) Na2CO3+NaPF6 in EC/DEC.
4) Differential scanning calorimetry analysis of Na2CO3 (a
model SEI compound) mixed with NaPF6 and NaPF6/EC/DEC re-
vealed that the origin of the SEI layer cracking on HC differs
from that of graphite.
results suggest that there is no discernible reactivity between
the Na salts and Na2CO3 within the temperature range tested.
Unlike mixtures of Li2CO3 with LiPF6 and 1m LiPF6 in EC/DMC[20]
that show noticeable exothermic peaks around 908C, the
Na2CO3/NaPF6 mixture did not show any exothermic event,
which evidences that the SEI layer cracking in NIBs could not
be triggered by a simple acid–base reaction and reaffirms the
absence of PF5 generation (from NaPF6) at such a low tempera-
ture. However, one can object that the solid (e.g., Na2CO3)–gas
(e.g., PF5) reaction may not reflect the actual reactivity of PF5
towards Na2CO3 realistically in NIB cells. To validate this, the
DSC evaluation of Na2CO3 mixed with NaPF6 in EC/DEC was
performed (Figure 8 f). As a result of the strong solvation
strength of EC for Na+, one may assume that the release of PF5
is facilitated. However, once again, the results oppose this hy-
pothesis. Thus, the thermal breakdown of the SEI layer in NIBs
could be caused by the simple transformation of metastable
SEI compounds (e.g., NEDC) into more stable ones [Eq. (9)].
In summary, besides the screening of safety and extrapolat-
ed performance for NIB electrolytes, our study could help to
clarify the confusing remarks that exist in the literature with
regard to the comparative reactivity of NIB and LIB electrolytes.
Moreover, although practical validation by the assembly of
large-scale NIB cells is needed, we can conclude at this stage
that the SEI layer in NIBs is less robust, which leads to a lower
onset temperature of the exothermic reaction and probably
early thermal runaway, compared to that of LIBs.
Experimental Section
Electrolyte preparation
All tested electrolytes consist of 1m solution of salts, namely,
NaPF6, NaClO4, NaFSI, NaFTFSI, and NaTFSI dissolved in EC/DMC,
EC/DEC, and EC/PC, all prepared in a 1:1 weight ratio. DMC, DEC,
and EC from UBE Europe GmbH, Germany, and PC from BASF USA
were used as received to prepare the solvent blends. The IL-based
electrolytes used in this study, namely, NaFSI/Pyr14FSI, NaFTFSI/
Pyr14FTFSI, and NaTFSI/Pyr14TFSI, were prepared in a 1:9 molar
ratio. The ILs and salts were first dried by using a turbo pump at
ꢀ10À8 mbar for ꢀ72 h. Binary mixtures of solvents were prepared
carefully inside an Ar-filled drybox (O2 and H2O<0.1 ppm). All salts
were vacuum dried before use.
ðCH2OCO2NaÞ2 ! Na2CO3þC2H4þCO2þ1=2 O2
ð9Þ
Conclusions
A comparative investigation of diverse Na-based electrolytes
that comprise various NaX [X=PF6, ClO4, bis(trifluoromethane-
sulfonyl)imide (TFSI), fluorosulfonyl-(trifluoromethanesulfonyl)i-
mide (FTFSI), and bis(fluorosulfonyl)imide (FSI)] salts and alkyl
carbonate blends [ethylene carbonate (EC)/dimethyl carbonate
(DMC), EC/diethyl carbonate (DEC), and EC/propylene carbon-
ate (PC)] and ionic liquids (Pyr14FSI, Pyr14FTFSI, and Pyr14TFSI;
Pyr14 =N-butyl-N-methylpyrrolidinium) as solvents was per-
formed by using various analytical tools. A detailed study of
the Al corrosion, formation of soluble solid–electrolyte inter-
Electrochemistry
Cell assembly and cycling
For thermal analysis, Swagelok-type half-cells composed of 90 wt%
HC and 10 wt% Super P carbon (no binder) as the working compo-
site electrode (ꢀ10 mg), an electrolyte-impregnated Whatman GF/
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
470
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
D borosilicate glass fiber as the separator, and a sodium metal foil
as counter electrode were used and assembled inside an Ar-filled
glovebox (O2 and H2O<0.1 ppm). Once assembled, the cells were
subjected to C/20 galvanostatic discharge/charge in the potential
window of the open-circuit voltage (OCV) to 0.005 V vs. Na/Na+.
heat flux differential calorimeter at a heating rate of 10 KminÀ1 in
the temperature range of 25–5008C under a constant Ar flow of
200 mLminÀ1. To ensure reproducibility, two DSC measurements
were conducted on each sample.
TGA tests were performed by using a Netzch instrument to
ꢀ5808C at 108CminÀ1 under a N2 atmosphere.
For GC–MS analysis, Swagelok cells were composed of a NaN3 sac-
rificial salt that contained Bellcore-type 2.5 mAh film (NaN3, Super
P
carbon, and polyvinylidene difluoride-hexafluoropropylene
(PVDF-HFP) in a weight ratio of 69:12:19) as the positive electrode
and 9 mg of 90 wt% HC and 10 wt% Super P carbon composite
powder as the negative electrode. The two Whatman GF/D borosi-
licate glass fiber separators were impregnated with a fixed amount
of 160 mL electrolyte. The cells were subjected to a C/50 galvano-
static charge in the potential window of OCV to 3.8 V vs. Na/Na+.
Acknowledgements
The authors are thankful to Dr. Gurpreet Singh (CIC Energigune,
Spain) for helpful discussions.
The HC used in this work for the study of the reactivity of the elec-
trolytes was prepared following the procedure described in
Ref. [24] by the pyrolysis of biomass waste at 11008C for 6 h under
Ar flow. The annealed powder was ground manually and ball
milled for 3 h in a zirconium oxide ball mill jar (ball to powder
weight ratio: 10:1) at the main disk rotation speed of 400 rpm
(main rotating disc) and a reversed planetary rotation speed of
800 rpm (rotating planets). For GC–MS analysis, a low-surface-area
HC (2.8 m2 gÀ1) was used to allow a fair comparison with the
graphite used for LIBs.
Keywords: carbon · electrochemistry · lithium · ionic liquids ·
sodium
[1] D. Larcher, J.-M. Tarascon, Nat. Chem. 2014, 7, 19–29.
[2] S. Y. Hong, Y. Kim, Y. Park, A. Choi, N.-S. Choi, K. T. Lee, Energy Environ.
Sci. 2013, 6, 2067.
[3] H. Pan, Y.-S. Hu, L. Chen, Energy Environ. Sci. 2013, 6, 2338.
[4] L. P. Wang, L. Yu, X. Wang, M. Srinivasan, Z. J. Xu, J. Mater. Chem. A
2015, 3, 9353–9378.
[5] B. L. Ellis, L. F. Nazar, Curr. Opin. Solid State Mater. Sci. 2012, 16, 168–
177.
[6] S. W. Kim, D. H. Seo, X. Ma, G. Ceder, K. Kang, Adv. Energy Mater. 2012,
2, 710–721.
Anodic behavior of Al
[7] A. Ponrouch, D. Monti, A. Boschin, B. Steen, P. Johansson, M. R. Palacín,
J. Mater. Chem. A 2015, 3, 22–42.
[8] M. Okoshi, Y. Yamada, A. Yamada, H. Nakai, J. Electrochem. Soc. 2013,
160, A2160–A2165.
[9] S. Komaba, W. Murata, T. Ishikawa, N. Yabuuchi, T. Ozeki, T. Nakayama,
A. Ogata, K. Gotoh, K. Fujiwara, Adv. Funct. Mater. 2011, 21, 3859–3867.
[10] X. Xia, M. N. Obrovac, J. R. Dahn, Electrochem. Solid-State Lett. 2011, 14,
A130.
[11] X. Xia, J. R. Dahn, J. Electrochem. Soc. 2012, 159, A515.
[12] X. Xia, W. M. Lamanna, J. R. Dahn, J. Electrochem. Soc. 2013, 160, A607–
A609.
The anodic behavior of Al in all the aforementioned Na salts dis-
solved in EC/DEC (chosen as a representative solvent mixture) and
ILs was determined by using CV. The experimental setup consisted
of a three-electrode Swagelok cell in which an Al sheet rinsed with
ethanol and vacuum dried served as the working electrode and
metallic sodium served as both the counter and reference electro-
des. Typically, the cell was left at OCV for at least 3 h before cycling
at a sweep rate of 10 mVsÀ1 between 0.1 and 5.3 V vs. Na/Na+.
[13] A. Ponrouch, E. Marchante, M. Courty, J.-M. Tarascon, M. R. Palacín,
Energy Environ. Sci. 2012, 5, 8572.
GC–MS and FTIR spectroscopy
The HC-based electrode and the two separators were recovered
after disassembling the cycled Swagelok cells inside an Ar-filled
drybox. The recovered samples were treated with 1 mL of dry ace-
tonitrile (H2O<0.001%), filtered, and diluted 100 times before in-
jection into the GC–MS. GC, which was interfaced with an ISQ
mass spectrometer, was performed by using a trace 1300 series GC
ultra-gas chromatograph. The chromatographic separation was
performed by using a “BPX70” cyanopropylpolysilene-siloxane-
based capillary column (30 m”0.25 mm i.d., 0.25 mm) from SGE.
[14] J. Zhao, L. Zhao, K. Chihara, S. Okada, J. Yamaki, S. Matsumoto, S. Kuze,
K. Nakane, J. Power Sources 2013, 244, 752–757.
[15] M. Dahbi, F. Ghamouss, F. Tran-Van, D. Lemordant, M. Anouti, J. Power
Sources 2011, 196, 9743–9750.
[16] L. Otaeguin, E. Goikolea, F. Aguesse, M. Armand, T. Rojo, G. Singh, J.
Power Sources 2015, 297, 168–173.
[17] R.-S. Kühnel, M. Lübke, M. Winter, S. Passerini, A. Balducci, J. Power Sour-
ces 2012, 214, 178–184.
[18] H. Kim, S. Grugeon, G. Gachot, M. Armand, L. Sannier, S. Laruelle, Elec-
trochim. Acta 2014, 136, 157–165.
[19] H. Yang, H. Bang, K. Amine, J. Prakash, J. Electrochem. Soc. 2005, 152,
A73.
[20] G. G. Eshetu, S. Grugeon, G. Gachot, D. Mathiron, M. Armand, S. Lar-
uelle, Electrochim. Acta 2013, 102, 133–141.
[21] M. Shakourian-Fard, G. Kamath, K. Smith, H. Xiong, S. K. R. S. Sankaranar-
ayanan, J. Phys. Chem. C 2015, 119, 22747–22759.
[22] J. Kalhoff, G. G. Eshetu, D. Bresser, S. Passerini, ChemSusChem 2015, 8,
2154–2175.
For the FTIR measurements, once the HC-based cells were in the
desodiated state, they were opened inside an Ar-filled drybox, and
the recovered powders were rinsed three times with DMC to elimi-
nate most of the solvents and salts before they were dried in the
antechamber. FTIR spectra were recorded on KBr pellets by using
a Nicolet Avatar 370DTGS spectrometer.
[23] Z. Chen, Y. Qin, Y. Ren, W. Lu, C. Orendorff, E. P. Roth, K. Amine, Energy
Environ. Sci. 2011, 4, 4023–4030.
[24] K. Hong, L. Qie, R. Zeng, Z. Yi, W. Zhang, D. Wang, W. Yin, C. Wu, Q. Fan,
W. Zhang, Y. Huang, J. Mater. Chem. A 2014, 2, 12733–12738.
TGA and DSC
Once the cells were in the sodiated state, they were dismantled
carefully inside a drybox and the samples recovered were trans-
ferred into Al pans, and used then directly for DSC analysis. All the
DSC experiments were conducted by using a Netzsch DSC 204F1
Received: December 2, 2015
Published online on February 2, 2016
ChemSusChem 2016, 9, 462 – 471
www.chemsuschem.org
471
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim