Hydrate-Melt Electrolytes For High-Energy-Density Aqueous Batteries

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

ARTICLES

PUBLISHED: 26 AUGUST 2016 | ARTICLE NUMBER: 16129 | DOI: 10.1038/NENERGY.2016.129

Hydrate-melt electrolytes for high-energy-density


aqueous batteries
Yuki Yamada1,2, Kenji Usui1, Keitaro Sodeyama2,3,4, Seongjae Ko1, Yoshitaka Tateyama2,4
and Atsuo Yamada1,2*

Aqueous Li-ion batteries are attracting increasing attention because they are potentially low in cost, safe and environmentally
friendly. However, their low energy density (<100 Wh kg−1 based on total electrode weight), which results from the narrow
operating potential window of water and the limited selection of suitable negative electrodes, is problematic for their future
widespread application. Here, we explore optimized eutectic systems of several organic Li salts and show that a room-
temperature hydrate melt of Li salts can be used as a stable aqueous electrolyte in which all water molecules participate
in Li+ hydration shells while retaining fluidity. This hydrate-melt electrolyte enables a reversible reaction at a commercial
Li4 Ti5 O12 negative electrode with a low reaction potential (1.55 V versus Li+ /Li) and a high capacity (175 mAh g−1 ). The resultant
aqueous Li-ion batteries with high energy density (>130 Wh kg−1 ) and high voltage (∼2.3–3.1 V) represent significant progress
towards performance comparable to that of commercial non-aqueous batteries (with energy densities of ∼150–400 Wh kg−1
and voltages of ∼2.4–3.8 V).

T
here is a growing demand for rechargeable batteries that Herein, we report the use of a Li salt hydrate melt as an
are high energy density and retain a high level of safety1–3 . electrolyte to bridge the energy-density gap between aqueous and
Lithium-ion batteries have relied to date on non-aqueous non-aqueous Li-ion batteries. A hydrate melt is a metal salt mixed
electrolytes (with wide potential windows)4,5 in combination with an extremely small amount of water, in which all water
with low-potential and high-capacity negative electrodes, for molecules are coordinated with the metal cations while retaining
example, graphite (0.15 V versus Li+ /Li, 372 mAh g−1 ) and spinel their fluidity16 . Multivalent cation salts (for example, Mg(NO3 )2
Li4 Ti5 O12 (1.55 V versus Li+ /Li, 175 mAh g−1 ), to store the high and Mn(NO3 )2 ) are known to form hydrate melts at ambient
energies per weight (∼150–400 Wh kg−1 based on total electrode temperatures, and have been studied with a focus on their peculiar
weight) found in commercially available batteries1–3,6 . However, physicochemical properties, which are analogous to those of molten
the use of highly volatile, flammable, and toxic non-aqueous salts16,17 . By contrast, Li salt hydrate melts are usually obtainable
electrolytes compromises battery safety4,5 . Although various only at high temperatures (for example, >100 ◦ C; ref. 18) or in
measures have been taken to minimize the safety risk7 , accidental the form of an unstable supercooled droplet19 and have never
fires due to non-aqueous Li-ion batteries in notebook computers, previously been obtained at ambient temperatures. We discovered
automobiles, and aeroplanes have still occurred8,9 . In most cases, a room-temperature hydrate melt of Li salts by carefully selecting
these accidents have been aggravated by leakage of the highly Li salt anions and then exploring their optimized eutectic systems
flammable electrolyte induced by thermal runaway (that is, with the aim of expanding the liquidus range. Using this hydrate
out-of-control exothermic reactions)7,9 . melt as an electrolyte, we demonstrated the reversible reaction of
Aqueous Li-ion batteries, first reported by Li et al. in 199410 , are a commercial Li4 Ti5 O12 negative electrode in an aqueous system,
receiving considerable attention as a safer battery technology based thereby significantly increasing the energy density of aqueous
on non-flammable and low-toxicity aqueous electrolytes11–14 . Aque- batteries into the energy-density range of commercial non-aqueous
ous Li-ion batteries also have an advantage in terms of production Li-ion batteries.
cost because there is no need to build an ultradry manufacturing
facility, as is required for current non-aqueous batteries. However, a Room-temperature hydrate melts
critical issue that limits their application is their low energy density Our strategy for seeking a room-temperature hydrate melt of Li
resulting from their inherently narrow potential window (1.23 V for salts consisted of the careful selection of Li salt anions and the
pure water)10,13 . The limited selection of suitable negative electrodes utilization of a eutectic system thereof. We selected two organic
(with low reaction potential and high capacity) also makes it more imide anions, N(SO2 CF3 )−2 (TFSI) and N(SO2 C2 F5 )−2 (BETI). These
difficult to achieve a higher energy density13 . As a result, even with a bulky anions occupy a large volume in solution, thereby reducing
highly stable aqueous electrolyte14,15 , the energy density of aqueous the water concentration. Moreover, these anions are weakly Lewis
Li-ion batteries has remained limited to <100 Wh kg−1 , which is basic and thus interact only weakly with Li+20,21 , thereby promoting
much smaller than those of current commercial non-aqueous Li-ion the solvation of the Li+ ions by water molecules rather than the
batteries (∼150–400 Wh kg−1 ). abundant formation of ion pairs. Finally, these anions exert a

1 Department of Chemical System Engineering, The University of Tokyo, 7-3-1, Hongo, Bunkyo-ku, Tokyo 113-8656, Japan. 2 Elements Strategy Initiative for
Catalysts & Batteries (ESICB), Kyoto University, 1-30, Goryo-Ohara, Nishikyo-ku, Kyoto 615-8245, Japan. 3 PRESTO, Japan Science and Technology Agency
(JST), 4-1-8, Honcho, Kawaguchi, Saitama 333-0012, Japan. 4 Center for Green Research on Energy and Environmental Materials and Center for Materials
Research by Information Integration, National Institute for Materials Science (NIMS), 1-1, Namiki, Tsukuba, Ibaraki 305-0044, Japan.
*e-mail: yamada@chemsys.t.u-tokyo.ac.jp

NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy 1


© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ
ARTICLES NATURE ENERGY DOI: 10.1038/NENERGY.2016.129

a 4 b
Li(TFSI)x(BETI)1−x /(H2O)n

Liquid

Water content, n
LiTFSI LiBETI H2O
3
Eutectic
Salt
deposition x = 0.7
n = 2.0

0.0 0.2 0.4 0.6 0.8 1.0 Molar


ratio 0.7 0.3 2.0 Hydrate melt
Salt mixing ratio, x

Figure 1 | Preparation of a room-temperature hydrate melt. a, Liquidus line of Li(TFSI)x (BETI)1−x salt–water mixtures. The liquidus line was drawn by
connecting saturation points (filled circles) that were experimentally obtained through the gradual addition of water to Li(TFSI)x (BETI)1−x salts with the
given values of x. The light blue region above the liquidus line represents a stable liquid phase of fully miscible salts and water, whereas the shaded blue
region represents a partially miscible phase of salts and water containing deposited crystalline salt or hydrate. The Li(TFSI)0.7 (BETI)0.3 salt (x = 0.7)
corresponds to a eutectic mixture that enables the formation of a hydrate melt state, Li(TFSI)0.7 (BETI)0.3 · 2H2 O, with the lowest water content (n = 2.0).
b, Stoichiometric amounts of LiTFSI, LiBETI and water used to prepare a Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate melt. The red arrows indicate the liquid levels.

‘plasticizing’ effect22,23 , inhibiting the crystallization of both the vibration modes of water molecules30–32 in various aqueous solutions
Li salts and the Li salt solvates (hydrates). These characteristics studied via Raman spectroscopy. In pure water, the O–H stretching
of TFSI and BETI anions result in an amorphous liquid state vibration gave rise to a broad Raman band consisting of several
at a minimal water concentration. In addition, these anions are components, which are attributed to various water molecules with
known to be highly resistant to hydrolysis24 . Next, to further different hydrogen-bonding environments in water clusters31,32 . This
expand the liquidus range, we explored a eutectic composition of broad band remained pronounced in saturated aqueous solutions
LiTFSI and LiBETI salts. Figure 1a shows the miscibility limits of of two conventional inorganic Li salts, LiNO3 and Li2 SO4 (Fig. 2a),
mixed LiTFSI–LiBETI salts in pure water at room temperature. We suggesting the presence of water clusters that were not participating
found that a eutectic composition of Li(TFSI)0.7 (BETI)0.3 showed in the Li+ hydration shells because of the high water content.
the highest miscibility with water, forming a stable, transparent, However, increasing the concentration of the organic LiTFSI salt led
colourless liquid with an extremely low water content (molar ratio: to a completely different situation (Fig. 2b); a new peak appeared at
n = H2 O/Li+ = 2.0), corresponding to a water molar concentration 3,565 cm−1 at the expense of the broad water-cluster band. Finally,
of 10.1 mol dm−3 (Fig. 1b). Because of this low water concentration, Li(TFSI)0.7 (BETI)0.3 · 2H2 O, with the lowest water content, showed
the water molecules can be isolated from each other by the bulky only a sharp peak at 3,565 cm−1 without the broad band, indicating
anions, thereby decreasing the opportunity for the formation of that the abundance of water clusters was significantly diminished
clusters of free water molecules (without coordination with Li+ ). in this extreme-state aqueous liquid. A sharp O–H vibration peak
Indeed, the discovered liquid material is regarded as a Li salt is characteristic of crystalline hydrates19,33 and has rarely been
hydrate melt, Li(TFSI)0.7 (BETI)0.3 · 2H2 O (dihydrate), with all water observed in aqueous solutions; thus, these findings suggest that the
molecules participating in Li+ hydration shells, as is evidenced both state of the water molecules in Li(TFSI)0.7 (BETI)0.3 · 2H2 O is rather
experimentally and theoretically (shown later). Very recently, we close to that in crystalline hydrates.
have found published works on water-in-salt electrolytes14,15 , but To achieve theoretical insight into the liquid structure, we
these other electrolytes are not hydrate melts and contain non- conducted first-principles density-functional-theory-based
negligible amounts of remaining free water molecules (more than molecular dynamics (DFT-MD) simulations. In a typical aqueous
10% of all water molecules) because of the use of much smaller and solution of 0.93 mol kg−1 LiTFSI/H2 O (n = 60), Li+ is solvated by
off-eutectic anions. four (or five) water molecules, and many free water molecules form
Information on the physicochemical properties of the hydrate a network with each other through hydrogen bonding (Fig. 2c).
melt is presented in Table 1. The hydrate melt has a vapour pressure For Li(TFSI)0.7 (BETI)0.3 · 2H2 O, after considering various initial
of 0.50 kPa at 30 ◦ C, which is quite low for an aqueous system structures at various temperatures, we found a stable state in which
(for example, 4.25 kPa for pure water25 at 30 ◦ C), suggesting that all water molecules were coordinated with Li+ by the Lewis-basic
the water activity is significantly decreased to 0.12 (=0.50/4.25) oxygen atoms and exhibited negligible hydrogen bonding with
because of a hydration effect in the hydrate melt. The viscosity other water molecules (Fig. 2d). The disappearance of free water
is high (203 mPa s at 30 ◦ C), but the ionic conductivity is molecules was further confirmed by tracking all water molecules’
3 mS cm−1 , which is comparable to that of Li-salt-based non- coordination states over 100,000 DFT-MD steps (Supplementary
aqueous solutions26 . Interestingly, on the basis of the Walden plot Fig. 2). This simulation result indicates the incorporation of all
(Supplementary Fig. 1), the hydrate-melt electrolyte can be classified water molecules into Li+ hydration shells and is fully consistent
as a ‘superionic solution’ because it lies above the ideal KCl line with the absence of the broad Raman band originating from water
(full dissociation)27,28 , which indicates that the ionic conductivity clusters, suggesting that Li(TFSI)0.7 (BETI)0.3 · 2H2 O can indeed be
is decoupled from the viscosity (fluidity); in other words, the ions classified as a hydrate melt.
move faster than would be expected based on the viscosity29 . The
‘superionicity’ of the hydrate melt is an interesting feature for its Li+ intercalation electrode reactions
use as a battery electrolyte and is worth scrutinizing in the future, Spinel Li4 Ti5 O12 (1.55 V versus Li+ /Li) is used for negative
although it is not the main focus of the present work. electrodes in commercial non-aqueous Li-ion batteries but has
never previously functioned reversibly in an aqueous electrolyte
Liquid structure because of the intensive hydrogen evolution that occurs instead
The hydrate melt state was characterized using both experimental of Li+ intercalation during charging13 . Figure 3 presents cyclic
and theoretical approaches. Figure 2a,b shows the O–H stretching voltammograms of a Li4 Ti5 O12 electrode (on an Al current collector)

2 NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy

© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ


NATURE ENERGY DOI: 10.1038/NENERGY.2016.129 ARTICLES
Table 1 | Physicochemical properties of the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate melt.

Material Salt molality Salt molar Density∗ Viscosity† Conductivity† Vapour pressure† Water activity‡
(mol kg−1 ) fraction (%) (g cm−3 ) (mPa s) (mS cm−1 ) (kPa) (-)
LiTFSI/H2 O solution 1.2 2.05 1.145 1.51 42.1 4.1 0.96
Hydrate melt 27.8 33.3 1.783 203 3.0 0.5 0.12
Data of a typical aqueous solution of 1.2 mol kg−1 LiTFSI/H2 O are listed for comparison. ∗ Density was measured at 25 ◦ C. † Viscosity, ionic conductivity and vapour pressure were measured at 30 ◦ C.

Water activity was evaluated from the vapour pressure divided by that of pure water (4.25 kPa) at 30 ◦ C. The vapour pressure of pure water was calculated from an equation presented in ref. 25.

a Hydration to Li+ b Hydration to Li+ d Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O

Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O

Sat. 9.5 mol kg−1


LiNO3/H2O LiTFSI/H2O
Intensity (a.u.)

Intensity (a.u.)

Sat. 1.2 mol kg−1 c Dilute LiTFSI/H2O


Li2SO4/H2O LiTFSI/H2O

Water clusters Water clusters


Pure Pure
water water

2,500 3,000 3,500 4,000 2,500 3,000 3,500 4,000


Raman shift (cm−1) Raman shift (cm−1)

Figure 2 | Structural characterization of the hydrate melt. a,b, Raman spectra of Li(TFSI)0.7 (BETI)0.3 · 2H2 O compared with those of inorganic Li salt
aqueous solutions, namely, saturated 12 mol kg−1 LiNO3 /H2 O and saturated 2.9 mol kg−1 Li2 SO4 /H2 O (a), and organic Li salt aqueous solutions, namely,
dilute 1.2 mol kg−1 LiTFSI/H2 O and concentrated 9.5 mol kg−1 LiTFSI/H2 O (b). The Raman bands observed in the range of 2,500–4,000 cm−1 correspond
to the O–H stretching modes of water molecules. The broad bands in the range of 2,900–3,700 cm−1 (between the blue dashed lines) are attributed to
clusters of water molecules with various hydrogen-bonding environments. The sharp peak at 3,565 cm−1 (orange dashed line) is attributed to Li+ -solvated
water molecules that are not clustered through hydrogen bonding. c,d, Snapshots of equilibrium trajectories obtained from first-principles DFT-MD
simulations of 1-LiTFSI/60-H2 O (corresponding to 0.93 mol kg−1 ) (c) and 7-LiTFSI/3-LiBETI/20-H2 O (corresponding to Li(TFSI)0.7 (BETI)0.3 · 2H2 O
hydrate melt, 28 mol kg−1 ) (d). Atom colours: Li, purple (presented with a larger size for emphasis); C, dark grey; H, light grey; O, red; N, blue; S, yellow;
F, light green. The cubic boxes represent the periodic boundaries of the supercells used in the DFT-MD simulations.

together with the linear sweep voltammograms of the Al current For the positive electrode side, layered LiCoO2 and spinel
collector. In a typical aqueous electrolyte of 1.2 mol kg−1 LiTFSI LiNi0.5 Mn1.5 O4 are established materials with high reaction
(1.0 mol dm−3 ), no reversible redox pair was observed; instead, a potentials of approximately 4.0 V and 4.7 V versus Li+ /Li,
cathodic current was observed below 2.0 V versus Li+ /Li for both the respectively3,13 . Figure 3 shows cyclic voltammograms of these
Li4 Ti5 O12 and Al electrodes (Fig. 3a), which was simply attributed active electrodes (on a Ti current collector) together with the
to hydrogen evolution by means of the reduction of water. The linear sweep voltammograms of the Ti current collector. The Ti
same behaviour was observed when using a Zn current collector current collector was found to be inactive up to high potentials
with a large overpotential for hydrogen evolution (Supplementary of above 5.5 V in the investigated aqueous systems by virtue of
Fig. 3) or aqueous electrolytes of conventional inorganic Li salts a stable native oxide layer on the surface. In a typical aqueous
(for example, LiNO3 or Li2 SO4 ) even at saturation (Supplementary electrolyte of 1.2 mol kg−1 LiTFSI, a reversible redox pair of Li+
Fig. 4). However, a reversible redox pair emerged on increasing the de-intercalation/intercalation was observed for LiCoO2 , with
concentration of organic Li salts, that is, LiTFSI and/or LiBETI, its reaction potential of 4.0 V, but not for LiNi0.5 Mn1.5 O4 , with its
at the expense of the hydrogen evolution (Supplementary Figs 5 higher reaction potential of 4.7 V, because of the high level of oxygen
and 6). In the extreme state of the Li(TFSI)0.7 (BETI)0.3 · 2H2 O evolution from the oxidation of water on the active LiNi0.5 Mn1.5 O4 .
hydrate melt, hydrogen evolution was suppressed on both Li4 Ti5 O12 By contrast, the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate-melt electrolyte
and Al (and Zn), and a fully reversible redox pair appeared at enabled the reversible Li+ de-intercalation/intercalation of both
1.85 V (Fig. 3b and Supplementary Fig. 3), which corresponds to the LiCoO2 and LiNi0.5 Mn1.5 O4 at 4.25 V and 5.00 V versus Li+ /Li,
Li+ intercalation/de-intercalation reaction of Li4+x Ti5 O12 . Notably, respectively. The reaction potentials were again approximately
the reaction potential in the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate 0.3 V higher than those in the conventional dilute organic
melt was found to be approximately 0.3 V higher than that in a electrolyte, as was the case for the Li4 Ti5 O12 electrode.
conventional organic electrolyte. This upward shift in the reaction
potential combined with the suppression of hydrogen evolution is High-energy-density aqueous Li-ion batteries
the key to achieving a reversible Li+ intercalation reaction, as shown Having confirmed the occurrence of reversible reactions on both
in the Discussion section. the negative and positive electrodes, we fabricated full Li-ion cells to

NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy 3


© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ
ARTICLES NATURE ENERGY DOI: 10.1038/NENERGY.2016.129

a 0.4 fast interfacial reactions. More importantly, such stable cycling was
1.2 mol kg−1 LiTFSI/H2O
observed exclusively for the hydrate-melt electrolyte. For the typical
aqueous electrolyte of 1.2 mol kg−1 LiTFSI, the Li4 Ti5 O12 negative
0.2
Current (mA)

electrode did not function at all, as is clear from Fig. 3a. Even a water-
in-salt electrolyte with high electrochemical stability (but also free
Ti water molecules)14 suffered severe capacity decay on cycling (only
0.0
Al LiNi0.5Mn1.5O4 12% capacity retention after 30 cycles at 0.2C with <90% Coulombic
Li4Ti5O12 LiCoO2 efficiency), as shown in Supplementary Fig. 10, suggesting that the
−0.2 hydrate melt state is important to achieve stable cycling.
Additionally, as an extreme example, the hydrate-melt electrolyte
1 2 3 4 5 enabled the partial use (that is, <0.5 Li+ extraction) of a high-
Potential (V) versus Li+/Li potential LiNi0.5 Mn1.5 O4 positive electrode. As shown in Fig. 4b and
b Supplementary Fig. 11, a LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 cell showed a
0.2 Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O voltage plateau at ∼3.0–3.3 V during both charging and discharging,
although the cell capacity was lower than that of the 2.4 V
0.1 LiCoO2 /Li4 Ti5 O12 cell because of the partial use of the positive
Current (mA)

electrode. Under high-rate cycling at 6.8C, the capacity retention


0.0 over 100 cycles was 63%, with the retention of nearly 100%
Coulombic efficiency (Fig. 4b). The reversible charge–discharge
−0.1
reaction was also confirmed under low-rate cycling at 0.5C and
2.4 V 1.0C, although the capacity faded and the Coulombic efficiency
−0.2 3.1 V declined during cycling (Supplementary Fig. 11), suggesting that
1 2 3 4 5 oxygen evolution occurred in parallel with the Li+ de-intercalation
Potential (V) versus Li+/Li at the LiNi0.5 Mn1.5 O4 during slow charging. These findings indicate
that the reaction potential of LiNi0.5 Mn1.5 O4 is positioned at
Figure 3 | Reversibility of Li+ intercalation/de-intercalation reactions. the edge of the anodic stability window of the hydrate melt,
a,b, Cyclic voltammograms (scan rate: 0.1 mV s−1 , electrode area: thereby hampering the full utilization of the high-potential positive
0.50 cm2 ) of Li4 Ti5 O12 , LiCoO2 and LiNi0.5 Mn1.5 O4 electrodes in a typical electrode. This is the voltage limitation of the present hydrate melt
aqueous solution of 1.2 mol kg−1 LiTFSI/H2 O (a) and in a hydrate-melt technology, although it implies the possibility of 3 V-class aqueous
electrolyte of Li(TFSI)0.7 (BETI)0.3 · 2H2 O (b). The three vertical lines batteries with further optimization of the hydrate-melt electrolyte.
indicate the redox potentials of the electrodes. The LiCoO2 and By virtue of the utilization of the low-potential and high-capacity
LiNi0.5 Mn1.5 O4 electrodes exhibited reaction potentials of 2.4 V and 3.1 V, Li4 Ti5 O12 negative electrode, the energy density of the aqueous cells
respectively, relative to that of the Li4 Ti5 O12 electrode. The current was considerably increased. Figure 4c shows the theoretical and
collectors were Al for the negative electrode (Li4 Ti5 O12 ) and Ti for the actual energy densities of various battery chemistries, including
positive electrodes (LiCoO2 and LiNi0.5 Mn1.5 O4 ). The dashed lines those investigated in this work, with respect to the total weight of the
represent the linear sweep voltammograms (scan rate: 0.1 mV s−1 , positive and negative active materials. Importantly, the actual energy
electrode area: 0.50 cm2 ) of the current collectors. density (as experimentally obtained) of the LiCoO2 /Li4 Ti5 O12 cell
was 130 Wh kg−1 (55.3 Ah kg−1 and 2.35 V on average), based on the
study their charge–discharge cyclability (Fig. 4 and Supplementary first discharge curve shown in Supplementary Fig. 12. This value can
Fig. 7). As shown in Fig. 4a, the fabricated LiCoO2 /Li4 Ti5 O12 be made to approach the theoretical value of 184 Wh kg−1 (Fig. 4c)
cell with the hydrate-melt electrolyte showed a voltage plateau by applying an engineering approach to optimize the cell design
at ∼2.3–2.5 V during both charging and discharging, suggesting to more effectively utilize the active materials, decrease the initial
the reversible operation of the battery. Under high-rate cycling irreversible capacity (Supplementary Fig. 7), and reduce the internal
at 10C, the capacity retention after 200 cycles was 75%, with the resistance of the cell. This energy density is comparable to that
retention of approximately 100% Coulombic efficiency. It should of a commercial non-aqueous LiMn2 O4 /Li4 Ti5 O12 Li-ion battery
be noted that irreversible capacity was observed in the initial (theoretical energy density: 160 Wh kg−1 ).
cycle (Supplementary Fig. 7) but was markedly decreased in the
following cycles (Fig. 4a), suggesting that a passivation reaction Discussion
occurred to suppress further electrolyte decomposition (as shown The hydrate melt state, in which all water molecules participate
in the Discussion section). It is well known that H+ -Li+ exchange in Li+ hydration shells, gives rise to a widening of the potential
reactions cause the gradual capacity decay of a LiCoO2 electrode window in both a thermodynamic and a kinetic manner (Fig. 5a)
on cycling34 . To circumvent this inessential electrode degradation and an upward shift in the Li+ intercalation potential (Fig. 5b),
by means of pH control11 , we used an alkaline hydrate melt, which synergistically contribute to enabling the reversible operation
Li1+x (TFSI)0.7 (BETI)0.3 (OH)x · (2 − x)H2 O (x = 0.035), for further of high-energy-density aqueous batteries.
prolonged cycling tests at low and high rates (Supplementary Fig. 8). Figure 5a shows the potential windows evaluated for various
Under low-rate cycling at 0.2C and 0.5C, the levels of capacity metal electrodes by means of linear sweep voltammetry using
retention after 100 cycles were 67% and 79%, respectively, with the a low scan rate (0.1 mV s−1 ). For the most catalytically active
Coulombic efficiency approaching 100% and with limited voltage Pt electrode, for which the kinetic factor (that is, overpotential)
fading during cycling (Supplementary Fig. 9). High-rate cycling can be minimized, the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate melt
at 10C resulted in a capacity retention of 84% after 500 cycles exhibited a wide potential window of 2.7 V (∼2.35–5.05 V versus
with the retention of approximately 100% Coulombic efficiency. Li+ /Li), whereas that of the typical 1.2 mol kg−1 LiTFSI aqueous
Notably, despite the high viscosity of the electrolyte, the voltage electrolyte was only 2.1 V (∼2.35–4.45 V versus Li+ /Li). Notably,
polarization was small even at this high rate, presumably because of this widening of the potential window primarily resulted from the
the decoupling of the ionic conductivity from the viscosity due to the extension of the anodic limit up to 5.05 V, suggesting that the
‘superionicity’ of the electrolyte (Supplementary Fig. 1) and/or the oxidation potential of the water molecules (or dissociated OH− ) was
small interfacial activation energy in aqueous systems35 that enables increased in the hydrate melt state, in which all water molecules

4 NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy

© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ


NATURE ENERGY DOI: 10.1038/NENERGY.2016.129 ARTICLES
a 4.0 b 4.0
LiCoO2 /Li4Ti5O12 (10C) LiNi0.5Mn1.5O4/Li4Ti5O12 (6.8C)
3rd 2nd

3.0 5th−2nd 3.0


Voltage (V)

Voltage (V)
2.0 75 100 2.0 40 100
Capacity (Ah kg−1)

Capacity (Ah kg−1)


Efficiency (%)

Efficiency (%)
30
50 80 80
20
1.0 25 60 1.0 60
10
0 40 0 40
0 50 100 150 200 0 25 50 75 100
Cycle number Cycle number
0.0 0.0
0 20 40 60 0 10 20 30
Cell capacity (Ah kg−1) Cell capacity (Ah kg−1)

c 100 150 200 250 300 350 Wh kg−1


4.0
Non-aqueous LiMn2O4/C LiCoO2/C
LiNi0.5Mn1.5O4 /Li4Ti5O12

3.0 0.5 Li+


LiMn2O4/Li4Ti5O12
Cell voltage (V)

Hydrate melt
LiCoO2/Li4Ti5O12

2.0
LiMn2O4/Mo6S8 (ref. 14)

LiMn2O4/VO2 (ref. 10)


1.0
Aqueous LiFePO4 /LiTi2(PO4)3 (ref. 11)

20 40 60 80 100
Cell capacity (Ah kg−1)

Figure 4 | High-energy-density aqueous Li-ion batteries with Li4 Ti5 O12 negative electrodes. a,b, Charge–discharge voltage profiles of two Li-ion full cells,
2.4 V LiCoO2 /Li4 Ti5 O12 (a) and 3.1 V LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 (b), with Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate-melt electrolytes. Charging and discharging
were performed at 10C (1.37 A g−1 for LiCoO2 ) (a) and 6.8C (1.00 A g−1 for LiNi0.5 Mn1.5 O4 ) (b) at 25 ◦ C. The LiCoO2 /Li4 Ti5 O12 cell was composed of
1.4 mg cm−2 LiCoO2 , 1.4 mg cm−2 Li4 Ti5 O12 and a glass fibre separator. The LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 cell was composed of 3.0 mg cm−2 LiNi0.5 Mn1.5 O4 ,
1.0 mg cm−2 Li4 Ti5 O12 and a quartz crystal separator. The current collectors were Al and Ti for the negative and positive electrodes, respectively. The cell
capacity was calculated on the basis of the total weight of the positive and negative active materials. The insets show the discharge capacity (open red
circles) and Coulombic efficiency (open blue circles) of the cells on cycling. c, Theoretical (filled circles and stars) and actual (open circles and stars)
energy densities of various aqueous and non-aqueous Li-ion chemistries with respect to the total weight of the positive and negative electrodes (the
weights of the electrolyte and cell are not considered). The electrode capacities and voltages used to calculate the theoretical energy densities are shown
in Supplementary Table 1. The actual energy densities of the LiCoO2 /Li4 Ti5 O12 and LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 cells were experimentally obtained from the
first discharge curves shown in Supplementary Fig. 12. The hydrate-melt electrolyte, as represented by the red area, enables the fabrication of
high-energy-density and high-voltage Li-ion batteries that bridge the traditional gap between aqueous and non-aqueous Li-ion batteries, which are
represented by the blue and orange areas, respectively.

were coordinated with Li+ . When coordinated with Li+ (a Lewis that anions are predominantly reduced to create a peculiar anion-
acid), a water molecule (or dissociated OH− ) donates its electron based passivation film on a negative electrode because of the
from the lone pair of the oxygen atom to the Li+ , which lowers its downward shift of the anions’ lowest unoccupied molecular orbital
highest occupied molecular orbital (HOMO) level and thus raises its level37,38 . Subsequently, it was reported that a similar anion-based
oxidation potential36 . This thermodynamic extension of the anodic passivation film can also form in superconcentrated aqueous
limit enables not only the fully reversible reaction of LiCoO2 but also electrolytes14 . Motivated by these previous reports, we analysed
the partial use of the higher-potential LiNi0.5 Mn1.5 O4 (although with the surface of a cycled Li4 Ti5 O12 electrode via X-ray photoelectron
parasitic oxygen evolution at low rates). spectroscopy (XPS; Supplementary Fig. 13) and identified a sulfur-
By contrast, no remarkable extension of the cathodic limit based passivation film on the electrode. According to the S2p
was observed for a Pt electrode in the hydrate-melt electrolyte. spectra recorded following Ar+ etching, an SOx moiety (167 eV)
Interestingly, however, the reversible redox pair of Li4 Ti5 O12 existed on the electrolyte side and sulfide species (160–162 eV)
appeared outside the cathodic limit in the hydrate-melt electrolyte. existed on the electrode side (Supplementary Fig. 13). Neither
To understand this seemingly paradoxical phenomenon, one should of these is a post-mortem decomposition product of residual
also consider a kinetic factor (that is, passivation) suppressing salts due to X-ray radiation or Ar+ etching in the XPS chamber
hydrogen evolution, thus further extending the cathodic limit. In (Supplementary Fig. 13), and thus, they should instead have formed
our previous work on superconcentrated non-aqueous electrolytes electrochemically on the Li4 Ti5 O12 during charging as reductive
(that is, a non-aqueous analogue of hydrate melts), we reported decomposition products of the TFSI and/or BETI anions; this

NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy 5


© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ
ARTICLES NATURE ENERGY DOI: 10.1038/NENERGY.2016.129

a b 0.3
0.2 1.2 mol kg−1 LiTFSI/H2O V
Current (mA) Ag/AgCl Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O
0.1 in sat. KCl
Pt
SUS SUS Ti
0.0
Al Zn LiTFSI
Pt LixFePO4

Potential shift (V)


−0.1 0.2
Liquid
Passivation Lower HOMO

High γ Li
−0.2 junction
22 mol kg−1 (sat.)
Li(TFSI)0.7(BETI)0.3 ⋅ 2H2O SUS
0.2
Li4Ti5O12/Al
Current (mA)

0.1 0.1 Sat. LiNO3


γ Li = 1
Pt Ti
0.0
Al Zn SUS Pt

High mLi
−0.1 Sat. Li2SO4
LiNi0.5Mn1.5O4/Ti
−0.2 1.25 V 2.35 V 5.05 V
0.0
1.0 1.5 2.0 2.5 4.0 4.5 5.0 5.5 0 5 10 15 20 25 30
Potential (V) versus Li+/Li mLi (mol kg−1)

c
Pure water

Li4Ti5O12 LiCoO2 LiNi0.5Mn1.5O4

LiTFSI/H2O solution
with free H2O

Nernst shift

Hydrate melt

Passivation Lower HOMO

1.25 2.35 4.45 5.05


Potential (V) versus Li+/Li (std.)

Figure 5 | Thermodynamic and kinetic factors accounting for the battery operating mechanism. a, Potential window of Li(TFSI)0.7 (BETI)0.3 · 2H2 O
hydrate melt compared with that of a typical 1.2 mol kg−1 LiTFSI/H2 O solution, evaluated via linear sweep voltammetry (scan rate: 0.1 mV s−1 ) on various
electrodes (Pt, Al, Ti, Zn, and stainless steel (SUS)). The dashed curves represent the cyclic voltammograms (scan rate: 0.1 mV s−1 ) of Li4 Ti5 O12 (on an Al
current collector) and LiNi0.5 Mn1.5 O4 (on a Ti current collector). The reduction potential of the hydrate melt at Pt (red dashed line at 2.35 V) is the same as
that of the typical aqueous solution, but it is markedly extended at Al (red dashed line at 1.25 V) as a result of passivation. The oxidation potential of the
hydrate melt at Pt (red dashed line at 5.05 V) is extended compared with that of the conventional aqueous solution because of the lowered HOMO level of
the water (and OH− ) molecules due to Li+ hydration. b, Equilibrium potential shift of the Li+ intercalation reaction as a function of the Li+ molality (mLi ) in
the electrolyte. The potential of Lix FePO4 (x ∼ 0.5) was measured versus Ag/AgCl in the cell configuration shown in the inset and was compared with that
in a 1.2 mol kg−1 LiTFSI aqueous solution. The electrolyte solutions used were the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate melt (red star), 1.2–22 mol kg−1
LiTFSI/H2 O (blue circles), saturated 12 mol kg−1 LiNO3 /H2 O (orange circle), and saturated 2.9 mol kg−1 Li2 SO4 /H2 O (brown circle). The dashed violet
line represents the potential shift in an ideal solution based on the Nernst equation, in which the Li+ activity coefficient (γ Li ) is supposed to be 1. The
hydrate melt shows the highest potential shift of 0.25 V by virtue of its high γ Li and high mLi . c, Relation between the potential windows (light blue bands)
of pure water, the conventional LiTFSI/H2 O solution, and the hydrate melt and redox reaction potentials (vertical red bars) of Li4 Ti5 O12 , LiCoO2 and
LiNi0.5 Mn1.5 O4 . The LiTFSI/H2 O solution has a wider potential window (∼2.35–4.45 V at Pt) than that of pure water, but the reaction potentials of
Li4 Ti5 O12 and LiNi0.5 Mn1.5 O4 still lie outside this potential window. For the hydrate melt, the anodic limit is extended up to 5.05 V (at Pt) because of the
lowered HOMO level of the water molecules (violet arrow), and the cathodic limit is extended down to 1.25 V (at Al) as a result of passivation (blue arrow).
Furthermore, the redox reaction potentials of all electrodes are shifted upwards by approximately 0.25 V according to the Nernst equation (orange arrows).

mechanism is consistent with the irreversible capacity observed For a Pt electrode, the cathodic limit was found to remain nearly
in the initial cycle (Supplementary Fig. 7). Lithium fluoride (LiF) unchanged regardless of the electrolyte, suggesting that passivation
was also found on the cycled electrode; importantly, however, failed to occur because of its strong catalytic activity towards
this material can form as a post-mortem decomposition product hydrogen evolution. By contrast, other electrodes (Al, Zn, stainless
of LiTFSI and LiBETI salts due to X-ray radiation and/or Ar+ steel (SUS), and Li4 Ti5 O12 ) in combination with the hydrate-melt
etching (Supplementary Fig. 13), as reported previously39,40 . Hence, electrolyte resulted in markedly widened cathodic limits, and thus,
we cannot conclude that this LiF forms in an electrochemical anion-based passivation was successfully achieved. Notably, the
manner and is a major component of the passivation film. passivation effect was maximal on the Al electrode, yielding a
The sulfur-based passivation film, which predominantly forms wide cathodic limit extending down to 1.25 V versus Li+ /Li in the
through anion reduction, should play a key role in effectively hydrate-melt electrolyte, whereas for the SUS current collector,
suppressing parasitic hydrogen evolution, thus kinetically extending such as those previously used in aqueous systems14 , massive
the cathodic limit. hydrogen evolution was observed at a higher potential than the
Interestingly, the effectiveness of the passivation strongly redox pair of Li4 Ti5 O12 (1.85 V versus Li+ /Li). Hence, the use of
depends on the electrodes to be passivated, as well as the an Al current collector in combination with the hydrate-melt
electrolytes that participate in the passivation reaction (Fig. 5a). electrolyte is a key factor in enabling the reversible reaction of

6 NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy

© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ


NATURE ENERGY DOI: 10.1038/NENERGY.2016.129 ARTICLES
Li4 Ti5 O12 with remarkably suppressed hydrogen evolution through the thermodynamic and kinetic extension of the potential window
kinetic hindrance. and the upward shift of the electrode potentials, both of which
In addition to the extension of the potential window in are consequences of the unusual aqueous liquid state unique to
both a thermodynamic and a kinetic manner, the equilibrium hydrate melts, in which all water molecules are coordinated with
potential shift of the Li+ intercalation reaction greatly contributes Li+ and free (un-coordinated) water molecules therefore disappear.
to enabling the reversible operation of the investigated aqueous This work will open up a new field of hydrate melt electrochemistry
cells. Figure 5b shows the equilibrium potential shift as evaluated and expand the potential avenues of research in the search for safe
on the basis of electromotive force measurements on Lix FePO4 aqueous batteries of high energy density.
(x ∼ 0.5) in various electrolyte solutions. With an increase in the
Li+ molality, the equilibrium potential shifted upwards compared Methods
with the 1.2 mol kg−1 LiTFSI/H2 O standard. In the extreme state Materials. Lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, LiN(SO2 CF3 )2 )
of the Li(TFSI)0.7 (BETI)0.3 · 2H2 O hydrate melt the equilibrium and lithium bis(pentafluoroethanesulfonyl)imide (LiBETI, LiN(SO2 C2 F5 )2 ) were
purchased from Kishida Chemical. Inorganic Li salts—LiNO3 , Li2 SO4 · H2 O and
potential shift reached 0.25 V. Such a large upward potential shift LiOH · H2 O—were purchased from Wako Pure Chemical Industries. The
was an exclusive characteristic of the Li(TFSI)0.7 (BETI)0.3 · 2H2 O electrolyte solutions were prepared by mixing the Li salts with ultrapure water
hydrate-melt electrolyte with the highest Li+ molality and without (Kanto Chemical). In particular, the hydrate-melt electrolyte,
free solvent molecules. These observations reasonably explain Li(TFSI)0.7 (BETI)0.3 · 2H2 O, was prepared by dissolving LiTFSI and LiBETI salts
the approximately 0.3 V higher reaction potentials of Li4 Ti5 O12 , in ultrapure water at a molar ratio of LiTFSI:LiBETI:H2 O = 7:3:20. To prepare an
LiCoO2 and LiNi0.5 Mn1.5 O4 observed in the cyclic voltammograms alkaline hydrate melt, a LiOH/H2 O solution (LiOH:H2 O = 1:57.6 by mol) was
prepared by first dissolving LiOH · H2 O in ultrapure water and then dissolving
compared with those in conventional dilute organic (and aqueous) LiTFSI and LiBETI salts in the LiOH/H2 O solution at a molar ratio of
electrolytes. The observed upward shift in potential by 0.25 V was LiTFSI:LiBETI:(LiOH/H2 O) = 7:3:20. The composition of the resultant alkaline
considerably larger than that expected in an ideal solution (that is, hydrate melt was Li1+x (TFSI)0.7 (BETI)0.3 (OH)x · (2 − x)H2 O (x = 0.035). To serve
with a Li+ activity coefficient of 1), which is shown as a dashed as active electrode materials, Li4 Ti5 O12 and LiCoO2 were provided by Ishihara
line in Fig. 5b, suggesting a large contribution from a high Li+ Sangyo and Nippon Chemical Industrial, respectively, and LiNi0.5 Mn1.5 O4 and
activity coefficient as well as the high Li+ molality. This effect can LiMn2 O4 were purchased from Hosen. LiFePO4 powder was synthesized via a
solid-state reaction from stoichiometric amounts of Li2 CO3 (Wako, 99.0%),
be attributed to the maximization of the hydration effect41–44 in FeC2 O4 · 2H2 O (JUNSEI, 99%) and (NH4 )2 HPO4 (Wako, 99.0%) as starting
the hydrate-melt electrolyte in the absence of free water molecules materials. During the synthesis, Ketjen black (ECP, Lion) was mixed with LiFePO4
(Supplementary Note 1). The large upward potential shift was also powder in a weight ratio of 90:10. The detailed synthesis conditions are reported
independently verified on the basis of the water activities (Table 1) elsewhere45 . To prepare electrode films (of Li4 Ti5 O12 , LiCoO2 , LiNi0.5 Mn1.5 O4 and
using the Gibbs–Duhem equation (Supplementary Note 2). LiMn2 O4 ), the active material powder was well mixed with acetylene black (AB,
On the basis of the above discussion, the high-energy-density HS-100, Denki Kagaku Kogyo) and polyvinylidene difluoride (PVdF, Kureha) in
N -methylpyrrolidone (NMP, Wako) solvent, with weight ratios of active
battery operating mechanism that is exclusively enabled by the material:AB:PVdF of 85:5:10, 85:9:6, 85:9:6 and 85:10:5, for Li4 Ti5 O12 , LiCoO2 ,
hydrate-melt electrolyte is summarized in Fig. 5c. In a conventional LiNi0.5 Mn1.5 O4 and LiMn2 O4 , respectively. For the preparation of a LiFePO4
aqueous electrolyte, the reaction potentials of Li4 Ti5 O12 and electrode film, the LiFePO4 /Ketjen black powder was well mixed with PVdF in a
LiNi0.5 Mn1.5 O4 are both completely outside the electrolyte’s weight ratio of 95:5 in NMP solvent. The obtained slurry was uniformly spread
potential window. By contrast, in the hydrate-melt electrolyte, onto a current collector using a doctor blade and dried at 60 ◦ C under vacuum
the anodic and cathodic potential limits are extended through a for 12 h. The current collectors were Al foil for the negative electrodes (Li4 Ti5 O12 )
and Ti foil for the positive electrodes (LiCoO2 , LiNi0.5 Mn1.5 O4 and LiMn2 O4 ),
thermodynamic mechanism (lowered HOMO level) and a kinetic unless otherwise noted. The dried LiFePO4 electrode was then electrochemically
mechanism (anion-based passivation), respectively, which results in oxidized in 1.1 mol kg−1 Li(TFSI)0.7 (LiBETI)0.3 aqueous electrolyte to obtain a
an extremely wide potential window with a maximum width of 3.8 V Lix FePO4 (x ∼ 0.5) electrode. The loading of the active material was controlled
(∼1.25–5.05 V versus Li+ /Li). In addition, the Li+ intercalation by the doctor blade depending on the usage, as described later.
potential shifts upwards by approximately 0.3 V, allowing the
reaction potential of Li4 Ti5 O12 to lie fully inside the potential Physicochemical properties. The density and viscosity of the electrolyte solutions
window. On the positive electrode side, the reaction potential of were evaluated using a DMA 35 density meter (Anton Paar) and a Lovis 2000 M
viscometer (Anton Paar), respectively. The ionic conductivity was measured via
LiCoO2 , although shifted upwards by approximately 0.3 V, is also the a.c. impedance method using a Solartron 147055BEC measuring system
fully encompassed by the remarkably widened potential window. (Solartron Analytical). A cell with a pair of platinum-plate electrodes was used,
Even for the higher-potential LiNi0.5 Mn1.5 O4 , the upward-shifted and the cell constant was defined with respect to a standard KCl aqueous
reaction potential remains at the edge of the potential window. solution. The vapour pressures of the electrolyte solutions were evaluated via a
The synergetic contributions of the thermodynamic and kinetic static method using a Baratron high-accuracy pressure manometer (MKS
effects, both maximized in the hydrate melt without free water Instruments) and a stainless-steel pressure vessel (Taiatsu Techno) in a circulation
bath (PolyScience).
molecules, enable the stable cycling of aqueous Li-ion batteries of
high energy density. Liquid structures. The coordination states in the electrolyte solutions were
In conclusion, we discovered a room-temperature Li salt studied via Raman spectroscopy using an NRS-5100 spectrometer (JASCO). A
hydrate melt by carefully selecting Li salt anions and exploring laser was excited at 532 nm. Each electrolyte solution was placed in a quartz cell,
their eutectics. The reversible Li+ intercalation reaction of the and the laser was directed through the quartz crystal window.
commercially used Li4 Ti5 O12 electrode material was demonstrated
in an aqueous system with the Li salt hydrate melt as the electrolyte. Electrochemical measurements. All electrochemical measurements except the
charge–discharge tests were performed using a VMP3 potentiostat (BioLogic).
The use of a low-potential and high-capacity Li4 Ti5 O12 negative The reaction reversibility of the Li4 Ti5 O12 , LiCoO2 and LiNi0.5 Mn1.5 O4 electrodes
electrode significantly increases the energy density of an aqueous was studied via cyclic voltammetry at a scan rate of 0.1 mV s−1 . The cyclic
battery, bridging the traditional energy-density gap between voltammetry measurements were performed in a three-electrode cell with a
aqueous (<100 Wh kg−1 ) and non-aqueous (>150 Wh kg−1 ) Li-ion LiMn2 O4 counter electrode (an excess amount compared with the working
batteries. As a proof of concept, charge–discharge cycling was electrode) and a Ag/AgCl (in saturated KCl aqueous solution) reference electrode
demonstrated in LiCoO2 /Li4 Ti5 O12 and LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 (3.239 V versus standard Li+ /Li, in which both the Li+ and H2 O activities are 1
and the temperature is 298.15 K). The surface area of the working electrode was
batteries with the hydrate-melt electrolyte, and the results pave the defined to be 0.50 cm2 using an O-ring. The active material loading was
way towards high-voltage (∼2.3–3.1 V) and high-energy-density approximately 2 mg cm−2 . The electrochemical stability windows of the electrolyte
(>130 Wh kg−1 ) aqueous Li-ion batteries. Reversible charge– solutions were evaluated via linear sweep voltammetry at a scan rate of
discharge cycling was achieved by means of the synergetic effect of 0.1 mV s−1 using a three-electrode cell. The cathodic and anodic limits were

NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy 7


© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ
ARTICLES NATURE ENERGY DOI: 10.1038/NENERGY.2016.129

measured using Al, Zn, stainless steel (SUS), Pt and Ti as working electrodes. The 8. Doughty, D. & Roth, E. P. A general discussion of Li ion battery safety. Interface
counter electrode was platinum mesh, and the reference electrode was Ag/AgCl 21, 37–44 (2012).
(in saturated KCl aqueous solution). To study the Li+ intercalation potential shift 9. Wang, Q. et al. Thermal runaway caused fire and explosion of lithium ion
as a function of the Li+ molality, the equilibrium potentials of the Lix FePO4 battery. J. Power Sources 208, 210–224 (2012).
(x ∼ 0.5) electrode in various electrolyte solutions were measured using a 10. Li, W., Dahn, J. R. & Wainwright, D. S. Rechargeable lithium batteries with
two-electrode cell with a Ag/AgCl (in saturated KCl aqueous solution) reference aqueous electrolytes. Science 264, 1115–1118 (1994).
electrode. The potentials versus Ag/AgCl were converted into potentials versus 11. Luo, J.-Y., Cui, W.-J., He, P. & Xia, Y.-Y. Raising the cycling stability of aqueous
standard Li+ /Li, under the assumption that the potential of the Ag/AgCl electrode lithium-ion batteries by eliminating oxygen in the electrolyte. Nat. Chem. 2,
was 3.239 V versus Li+ /Li. Charge–discharge tests were performed using a 760–765 (2010).
TOSCAT-3100 charge–discharge unit (Toyo System), using a 2032-type coin cell 12. Pasta, M., Wessells, C. D., Huggins, R. A. & Cui, Y. A high-rate and long cycle
with a Li4 Ti5 O12 negative electrode, a LiCoO2 or LiNi0.5 Mn1.5 O4 positive life aqueous electrolyte battery for grid-scale energy storage. Nat. Commun. 3,
electrode, and a quartz fibre separator of 380 µm in thickness (Advantec, QR100) 1149 (2012).
or a glass fibre separator of 420 µm in thickness (Whatman GF/F). The electrode 13. Kim, H. et al. Aqueous rechargeable Li and Na ion batteries. Chem. Rev. 114,
sheets were punched out to form disc electrodes. The active material loadings of 11788–11827 (2014).
the positive and negative electrodes are described in the corresponding figure 14. Suo, L. et al. ‘Water-in-salt’ electrolyte enables high-voltage aqueous
legends. For the LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 cells, the negative/positive capacity lithium-ion chemistries. Science 350, 938–943 (2015).
ratio was set to a small value (Li4 Ti5 O12 /LiNi0.5 Mn1.5 O4 = 0.4) with the intent of 15. Suo, L. et al. Advanced high-voltage aqueous lithium-ion battery
partial use of the LiNi0.5 Mn1.5 O4 capacity (that is, <0.5 Li extraction). To enabled by ‘water-in-bisalt’ electrolyte. Angew. Chem. Int. Ed. 55,
minimize side reactions between the electrolyte and the coin cell components, an 7136–7141 (2016).
18 mmφ Ti foil was placed between the positive electrode sheet and the positive 16. Angell, C. A. A new class of molten salt mixtures. The hydrated dipositive ion
casing bottom, another 18 mmφ Al foil was placed between the negative electrode as an independent cation species. J. Electrochem. Soc. 112, 1224–1227 (1965).
sheet and the spacer, and the spacer was wrapped with Al foil. Here, a rate of xC 17. MacFarlane, D. R. & Angell, C. A. Emulsion techniques for the study of glass
corresponded to an applied current density of 137 × x mAh g−1 or formation. 2. Low melting point salt hydrates. J. Phys. Chem. 88,
147 × x mAh g−1 for the LiCoO2 /Li4 Ti5 O12 or LiNi0.5 Mn1.5 O4 /Li4 Ti5 O12 cell, 4779–4781 (1984).
respectively, based on the active mass of the positive electrode. All 18. Tripp, B. Vapor pressures of aqueous melts: LiNO3 +KNO3 melts containing
electrochemical measurements were performed at 25 ◦ C. water or deuterium oxide. J. Chem. Thermodyn. 7, 263–269 (1975).
19. Zhang, Y. & Chan, C. K. Observations of water monomers in supersaturated
Electrode surface characterization. The surface chemistry of the Li4 Ti5 O12 NaClO4 , LiClO4 , and Mg(ClO4 )2 droplets using Raman spectroscopy. J. Phys.
electrode was analysed using an X-ray photoelectron spectrometer (PHI5000 Chem. A 107, 5956–5962 (2003).
VersaProbe II, ULVAC-PHI) with a monochromatic Al Kα X-ray source. A 20. Henderson, W. A. Glyme-lithium salt phase behavior. J. Phys. Chem. B 110,
LiCoO2 /Li4 Ti5 O12 cell was subjected to a one-cycle charge–discharge 13177–13183 (2006).
measurement at 10C and was then disassembled in an Ar-filled glove box; 21. Johansson, P. Electronic structure calculations on lithium battery electrolyte
subsequently, the cycled Li4 Ti5 O12 electrode was washed thrice with salts. Phys. Chem. Chem. Phys. 9, 1493–1498 (2007).
1,2-dimethoxyethane (DME) solvent to minimize the amount of residual Li salts. 22. Lascaud, S. et al. Phase-diagrams and conductivity behavior of
The washed electrode was dried and transferred into the XPS chamber without poly(ethylene oxide) molten-salt rubbery electrolytes. Macromolecules 27,
being exposed to air. For XPS measurements of the Li salts, LiTFSI and LiBETI
7469–7477 (1994).
powders were deposited on a carbon tape in an Ar-filled glove box and then
23. Henderson, W. A. et al. Glyme-lithium bis(trifluoromethanesulfonyl)imide and
transferred into the XPS chamber without being exposed to air. The depth profile
glyme-lithium bis(perfluoroethanesulfonyl)imide phase behavior and solvate
was obtained via Ar+ sputtering at 1 kV.
structures. Chem. Mater. 17, 2284–2289 (2005).
24. Lux, S. F. et al. LiTFSI stability in water and its possible use in aqueous
Computational details. The liquid structure was theoretically studied on the
lithium-ion batteries: pH dependency, electrochemical window and
basis of first-principles DFT-MD simulations. Car–Parrinello-type DFT-MD
temperature stability. J. Electrochem. Soc. 160, A1694–A1700 (2013).
simulations were conducted using CPMD code46,47 . Cubic supercells with linear
25. Saul, A. & Wagner, W. International equations for the saturation properties of
dimensions of 1.2337 nm and 1.4874 nm were used for the DFT-MD simulations
ordinary water substance. J. Phys. Chem. Ref. Data 16, 893–901 (1987).
of 1-LiTFSI/60-H2 O and 7-LiTFSI/3-LiBETI/20-H2 O, respectively. A fictitious
26. Ue, M. Mobility and ionic association of lithium salts in a propylene
electronic mass of 500 a.u. and a time step of 4 a.u. (0.10 fs) were chosen. The
carbonate-ethyl methyl carbonate mixed solvent. J. Electrochem. Soc. 142,
temperature was controlled using a Nosé thermostat48 with a target temperature
2577–2581 (1995).
of 298 K. To obtain the equilibrated liquid structure, we considered six, four and
27. Xu, W., Cooper, E. I. & Angell, C. A. Ionic liquids: ion mobilities, glass
four initial structures at 298 K, 350 K and 400 K, respectively. After 5 ps
equilibration steps, statistical averages were computed from trajectories of temperatures, and fragilities. J. Phys. Chem. B 107, 6170–6178 (2003).
at least 10 ps in length. The electronic wavefunction was quenched to the 28. Bressel, R. D. & Angell, C. A. Fluidity and conductance in aqueous electrolyte
Born–Oppenheimer surface approximately every 1 ps to maintain solutions. An approach from the glassy state and high-concentration limit. I.
adiabaticity. The energy cutoff of the plane wave basis was set to 90 Ry. Ca(NO3 )2 solutions. J. Phys. Chem. 1086, 3244–3253 (1979).
Goedecker–Teter–Hutter-type norm-conserving pseudopotentials49,50 for C, 29. McLin, M. & Angell, C. A. Contrasting conductance/viscosity relations in
H, O, N, S, F and Li were used. The detailed methods are presented in our liquid states of vitreous and polymer solid electrolytes. J. Phys. Chem. 92,
previous publication38 . 2083–2086 (1988).
30. Brooker, M. H., Hancock, G., Rice, B. C. & Shapter, J. Raman frequency and
intensity studies of liquid H2 O, H18 2 O and D2 O. J. Raman Spectrosc. 20,
Received 10 February 2016; accepted 29 July 2016; 683–694 (1989).
published 26 August 2016 31. Auer, B., Kumar, R., Schmidt, J. R. & Skinner, J. L. Hydrogen bonding and
Raman, IR, and 2D-IR spectroscopy of dilute HOD in liquid D2 O. Proc. Natl
Acad. Sci. USA 104, 14215–14220 (2007).
References 32. Auer, B. M. & Skinner, J. L. IR and Raman spectra of liquid water: theory and
1. Tarascon, J.-M. & Armand, M. Issues and challenges facing rechargeable interpretation. J. Chem. Phys. 128, 224511 (2008).
lithium batteries. Nature 414, 359–367 (2001). 33. Uriarte, L. M. et al. Reference Raman spectra of synthesized CaCl2 · nH2 O
2. Armand, M. & Tarascon, J.-M. Building better batteries. Nature 451, solids (n = 0,2,4,6). J. Raman Spectrosc. 46, 822–828 (2015).
652–657 (2008). 34. Wang, Y. & Xia, Y. Hybrid aqueous energy storage cells using activated carbon
3. Goodenough, J. B. & Kim, Y. Challenges for rechargeable Li batteries. Chem. and lithium-intercalated compounds. J. Electrochem. Soc. 153,
Mater. 22, 587–603 (2010). A1425–A1431 (2006).
4. Xu, K. Nonaqueous liquid electrolytes for lithium-based rechargeable batteries. 35. Nakayama, N. et al. Interfacial lithium-ion transfer at the LiMn2 O4 thin film
Chem. Rev. 104, 4303–4417 (2004). electrode/aqueous solution interface. J. Power Sources 174, 695–700 (2007).
5. Xu, K. Electrolytes and interphases in Li-ion batteries and beyond. Chem. Rev. 36. Yoshida, K. et al. Oxidative-stability enhancement and charge transport
114, 11503–11618 (2014). mechanism in glyme-lithium salt equimolar complexes. J. Am. Chem. Soc. 133,
6. Dunn, B., Kamath, H. & Tarascon, J.-M. Electrical energy storage for the grid: 13121–13129 (2011).
a battery of choices. Science 334, 928–935 (2011). 37. Yamada, Y. et al. Unusual stability of acetonitrile-based superconcentrated
7. Balakrishnan, P. G., Ramesh, R. & Prem Kumar, T. Safety mechanisms in electrolytes for fast-charging lithium-ion batteries. J. Am. Chem. Soc. 136,
lithium-ion batteries. J. Power Sources 155, 401–414 (2006). 5039–5046 (2014).

8 NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy

© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ


NATURE ENERGY DOI: 10.1038/NENERGY.2016.129 ARTICLES
38. Sodeyama, K., Yamada, Y., Aikawa, K., Yamada, A. & Tateyama, Y. Sacrificial 49. Goedecker, S., Teter, M. & Hutter, J. Separable dual-space Gaussian
anion reduction mechanism for electrochemical stability improvement pseudopotentials. Phys. Rev. B 54, 1703–1710 (1996).
in highly concentrated Li-salt electrolyte. J. Phys. Chem. C 118, 50. Krack, M. Pseudopotentials for H to Kr optimized for gradient-corrected
14091–14097 (2014). exchange-correlation functionals. Theor. Chem. Acc. 114, 145–152 (2005).
39. Ota, H., Sakata, Y., Wang, X., Sasahara, J. & Yasukawa, E. Characterization of
lithium electrode in lithium imides/ethylene carbonate and cyclic ether Acknowledgements
electrolytes. J. Electrochem. Soc. 151, A437 (2004). This work was supported by a JSPS Grant-in-Aid for Specially Promoted Research
40. McCloskey, B. D. et al. Twin problems of interfacial carbonate formation in (No. 15H05701). The calculations in this work were performed on the K computer at the
nonaqueous Li–O2 batteries. J. Phys. Chem. Lett. 3, 997–1001 (2012). RIKEN AICS and on the Oakleaf-FX at the University of Tokyo with the support of the
41. Pearce, J. N. & Nelson, A. F. The vapor pressures of aqueous solutions of HPCI Systems Research Projects (Proposal Numbers hp150275 and hp150068) and on
lithium nitrate and the activity coefficients of some alkali salts in solutions of the supercomputers at NIMS, ISSP and ITC at the University of Tokyo and
high concentration at 25◦ . J. Am. Chem. Soc. 54, 3544–3555 (1932). Kyushu University.
42. Stokes, R. H. A thermodynamic study of bivalent metal halides in aqueous
solution. Part XVII-Revision of data for all 2:1 and 1:2 electrolytes at 25◦ , and Author contributions
discussion of results. Trans. Faraday Soc. 44, 295–307 (1947). Y.Y. and A.Y. proposed the concept. Y.Y., K.U. and S.K. designed the experiments.
43. Stokes, R. H. & Robinson, R. A. Ionic hydration and activity in electrolyte K.U. developed the hydrate melt system and analysed the structures and basic
solutions. J. Am. Chem. Soc. 70, 1870–1878 (1948). physicochemical/electrochemical properties. S.K. designed the electrochemical cell and
44. Glueckauf, B. Y. E. The influence of ionic hydration on activity performed the cycling tests and the post-mortem electrode analyses. K.S. and Y.T.
coefficients in concentrated electrolyte solutions. Trans. Faraday Soc. 51, designed and performed the theoretical calculations. All authors contributed to the
discussion. Y.Y. and A.Y. wrote the manuscript. A.Y. supervised the overall project.
1235–1244 (1955).
45. Kobayashi, G. et al. Isolation of solid solution phases in size-controlled
Lix FePO4 at room temperature. Adv. Funct. Mater. 19, 395–403 (2009). Additional information
46. Car, R. & Parrinello, M. Unified approach for molecular dynamics and Supplementary information is available online. Reprints and permissions information is
density-functional theory. Phys. Rev. Lett. 55, 2471–2474 (1985). available online at www.nature.com/reprints. Correspondence and requests for materials
47. CPMD (IBM Corp. 1990–2008, MPI für Festkörperforschung Stuttgart should be addressed to A.Y.
1997–2001); http://www.cpmd.org/
48. Nosé, S. A unified formulation of the constant temperature molecular Competing interests
dynamics methods. J. Chem. Phys. 81, 511–519 (1984). The authors declare no competing financial interests.

NATURE ENERGY | VOL 1 | OCTOBER 2016 | www.nature.com/natureenergy 9


© ƐƎƏƖɥMacmillan Publishers LimitedƦɥ/13ɥ.$ɥ/1(-%#1ɥ341#. All rights reservedƥ

You might also like