View Online / Journal Homepage / Table of Contents for this issue

Journal of Dynamic Article Links < C

Materials Chemistry
Cite this: J. Mater. Chem., 2011, 21, 11764
www.rsc.org/materials PAPER
Potential positive electrodes for high-voltage magnesium-ion batteries
Tetsu Ichitsubo,a Takero Adachi,a Shunsuke Yagia and Takayuki Doib
Received 23rd April 2011, Accepted 18th May 2011
DOI: 10.1039/c1jm11793a
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

Magnesium-ion batteries (MIBs) with a Mg-metal negative electrode are expected to combine high
Downloaded by Ryerson Polytechnic University on 15 November 2012

energy density and high electromotive force, owing to the divalent ion careers and its low redox
potential. However, it has been reported to date that the cell voltage of MIBs is not high enough
(
1.5 V), being far below that of lithium-ion batteries (LIBs) (4–5 V). In this work, we have investigated
the potentiality of Mg–Co–O and Mg–Ni–O complex oxides as the positive electrode for MIBs, which
are composed of these positive electrodes and a Mg negative electrode in acetonitrile with Mg(ClO4)2
salt as an electrolyte. These MIBs can exhibit a relatively high open circuit voltage, OCV, and can light
a blue diode after charge. However, as the combination of the acetonitrile electrolyte and metal Mg can
yield the passivation on the surface of the Mg electrode, we have also checked these materials for a well-
established Li ion battery system, and confirmed that the charged battery can show high OCV. In order
to attain such a high cell voltage, it would be significant to exploit the unstabilized ion after charge in
the host complex oxide.

1. Introduction ion that can accommodate up to 4 electrons.6 The mechanism of

the valence change of such a Mo6S8 cluster including metallic
Since an innovative magnesium-ion battery (MIB) was system- bonds is excellent to retain the local electro-neutrality rapidly
atically prototyped by Aurbach et al.,1 it has been expected as upon accepting Mg2+ ions, but its valence change presumably does
a potential rechargeable battery comparable to the lithium-ion not contribute to the enhancement of the cell voltage. Besides,
battery (LIB) from the viewpoints of safety of handling, its large both high electronic conductivity and high ionic conductivity are
abundance, and inexpensiveness of magnesium. This is mainly indispensable for a positive-electrode material, to facilitate redox
because, when magnesium metal is used as a negative-electrode reactions inside the active material. Thus, in order to find an
material in an MIB, it can provide a considerably higher energy appropriate positive-electrode material and to obtain a firm
density owing to the considerably higher capacity (2200 Ah kg1 guideline for high cell voltage, we dare to focus on classical but
or 3830 Ah dm3 for Mg, being comparable to 3860 Ah kg1 or unexplored Mg–TM complex oxide systems (TM: transition
2060 Ah dm3 for Li) and negatively large redox potential metal, such as Co or Ni) as the positive-electrode materials,
(e.g., 2.36 V for Mg2+/Mg, 3.05 V for Li+/Li versus SHE). because their high valence ions, Co(IV) or Ni(IV), are quite unstable
Nevertheless, there are not so many papers associated with MIBs (active), so that it is anticipated they will play a significant role in
so far. This is due to the fact that the MIBs reported to date increasing the cell voltage, as shown in Fig. 1(a). As shown in
deliver less voltage in comparison with the LIB cell voltage.1–3 In Fig. 1(b), in the proposed MIB system (e.g., for MgCo2O4), we
order to develop further MIB technology, in addition to the consider the following redox reactions during discharge:
research on the appropriate electrolyte for the Mg negative
electrode,4 we have to design an excellent positive-electrode Positive electrode (Cathode for discharge): Mg1xCo2O4
material that shows higher potential versus magnesium.5 + yMg2+ + 2ye / Mg1x+yCo2O4
Due to the divalence of Mg2+ ions, it is generally foreseen that
the intercalation of Mg2+ ions is somewhat difficul and complex. Negative electrode (Anode for discharge): yMg / yMg2+ + 2ye
From the viewpoint of fast intercalation of Mg2+ ions into the
host, the unusual structure of the Chevrel phase such as MgMo3S4 Here, x means the amount of prior charge, and y denotes the
seems to be surely suitable for the MIB positive electrode,1 where amount of the following discharge. In this battery, by the
the Mo6S8 blocks (Mo6-octahedral clusters surrounded by eight insertion of Mg2+ ions into the host positive electrode
chalcogen ions) are known to behave like a single transition-metal active material, the valence of the Co ions changes from Co(IV) to
Co(III). At this valence transition, it is expected that a large
a
potential change is yielded, resulting in the high cell voltage.
Department of Materials Science and Engineering, Kyoto University,
In this work, we have investigated the spinel structure and
Kyoto, 606-8501, Japan
b
Office of Society-Academia Collaboration for Innovation, Kyoto electrical conductivity for the single-phase MgCo2O4 samples
University, Kyoto, 615-8520, Japan synthesized by the inverse coprecipitation process, and examined

11764 | J. Mater. Chem., 2011, 21, 11764–11772 This journal is ª The Royal Society of Chemistry 2011
 View Online

sodium carbonate solution, being kept at 70–80  C, continuously

keeping stirred, in which the coprecipitation proceeded at an
almost constant pH. The objective precipitates were filtered, then
washed with hot water until nitrate ions disappeared, and
subsequently dried at 80  C. The thus obtained precursor was
heated at 350  C for 2 h in air to fully oxidize. The crystal
structures of the MgCo2O4 and Mg0.67Ni1.33O2 samples were
confirmed by X-ray diffraction (XRD) with Mo-Ka radiation.
For the as-synthesized sample of MgCo2O4, we have evaluated
the site occupation by using the Rietveld analysis with the
powder diffractions data. The microstructures of the electrodes
were observed by a field-emission scanning electron microscope
(FE-SEM).
For electrical conductivity measurement, the pellet was
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

obtained by calcining at 500  C or 600  C for 24 h in air. For

Downloaded by Ryerson Polytechnic University on 15 November 2012

reference, the spinel-type Co3O4 sample was prepared by

calcining the commercial reagent at 800  C for 24 h in air
(actually, the Co3O4 sample for the impedance measurement also
should have been also prepared by the inverse precipitation
Fig. 1 (a) Concept of enhancement of OCV in the MIB. The illustration method, but the reagent was used here just for reference).
shows the qualitative magnitude relation of the redox potential. To MgCo2O4 and Co3O4 thus prepared form spinel structures and
exploit the unstable cations, e.g., Co(IV) or Ni(IV), in the positive-elec- Mg0.67Ni1.33O2 shows the rock salt structure. The electrical
trode active material is considered to be a key for enhancement of the cell
conductivities were analyzed by AC impedance measurement,
voltage. (b) Schematic illustration of the MIB system presented here:
performed in the frequency range from 10 Hz to 7 MHz. Before
positive electrode (in this figure, MgCo2O4); negative electrode, Mg
metal; electrolyte: Mg(ClO4)2 in acetonitrile.
the measurements, the top and bottom surfaces of each sample
were coated with gold paste. The conductivity at room temper-
the feasibility of the use of MgCo2O4 (spinel, ideal capacity: ature was measured in air, and the temperature dependence of
260 Ah kg1) and also Mg0.67Ni1.33O2 (rock salt, ideal capacity: the electrical conductivity of the samples was measured up to
283 Ah kg1) as a positive-electrode material, prototyping a cell. approximately 500  C in a mixed gas atmosphere with the frac-
The ideal capacity stated above is evaluated under the assump- tion of O2/Ar ¼ 1/4.
tion that only Mg2+ ions are intercalated/deintercalated into/ The MIB system was prototyped by using the powder of
from the host. Here we experimentally substantiate the MgCo2O4 or Mg0.67Ni1.33O2 as a positive electrode and Mg as
following: (1) MgCo2O4 is a semiconductor showing a very high a negative electrode, as shown in Fig. 1(b). Most of the cell
electrical conductivity comparable to LiCoO2. (2) The present constructions were done in a glove box filled with a high-purity
MIB with a MgCo2O4 positive electrode exhibits a high open Ar gas. The positive-electrode materials were a mixture of
circuit voltage (OCV) of about 3.5 V. Furthermore, we demon- MgCo2O4 or Mg0.67Ni1.33O2 powder (in a diameter less than 10–
strate for the first time that this new MIB successfully lighted 20 mm) of 75–80 wt%, polyvinylidene difluoride (PVDF) of 10 wt
a blue diode. The possibility of Co3O4 with the normal spinel %, and carbon black of 10–15 wt%, and it was pasted on the Al
structure was examined as a positive-electrode material of MIB plate, and the commercially available electrolyte (Kishida
so far,2 but as shown later, this is crucially different from Chemical Corporation), 1 mol L1 Mg(ClO4)2 dissolved in
MgCo2O4 in terms of realizing the unstable ions such a Co(IV). acetonitrile, was used in our MIB system. In most cases, the
(3) On the basis of the prospected guideline, Mg0.67Ni1.33O2 is reaction surface areas of each electrode were about 2 cm2. In all
also shown to be a potential positive-electrode material for MIB. of the cells, the powder samples annealed at 350  C for 2 h after
fabricated by the inverse coprecipitation method were used as the
positive-electrode active materials.
2. Experimental
The charge processes were done under cc–cv conditions. In this
The synthesis of magnesium cobalt oxide by the solid state mode, the cc (constant current) value is (mostly) set at
reaction is not applicable, because the treatment requires 50 mA/2 cm2, and after the current cannot be passed, the mode is
annealing or calcining at high temperatures but this leads to the changed to the cv (constant voltage) condition of 4 V. The
decomposition of cobaltite.7 Hence, MgCo2O4 samples were discharge processes other than use of the blue diode were made
synthesized by the inverse coprecipitation method, referring to under the condition of a lower limit of 1.5 V. Moreover, we show
the papers describing the synthesis of XCo2O4 (X ¼ Cu, Ni, Zn, a part of the results of the cyclic voltammetries (CVs) for the
Mg etc.).8–12 First, referring to the literature,13 by using Mg present positive electrodes. In the case of cyclic voltammetry,
(NO3)2$6H2O and Co(NO3)2$6H2O or Ni(NO3)2$6H2O used as MgCo2O4 or Mg0.67Ni1.33O2 was the working electrode (WE),
starting materials, we prepared the mixted nitrate solution with and two different pieces of Mg were used for the counter elec-
concentrations of 0.08 mol L1 Mg(NO3)2 and 0.16 mol L1 Co trode (CE) and reference electrode (RE). Prior to CVs, the cells
(NO3)2 or Ni(NO3)2 in deionized water. We used sodium were charged to some extent under the above cc–cv condition,
carbonate solution of 0.35 mol L1 in water as the precipitating where the cv condition of 4 V was set as the potential WE versus
agent, and poured the solution of mixed nitrates slowly into the CE. Subsequently, CVs have been performed for each cell.

This journal is ª The Royal Society of Chemistry 2011 J. Mater. Chem., 2011, 21, 11764–11772 | 11765
 View Online

3. Results conductivity. Thus, it is noteworthy that MgCo2O4 has sufficient

potential as a positive-electrode material from the viewpoint of
First, we show the structure and electrical properties of the electrical conductance.
sample synthesized by the inverse coprecipitation method. Fig. 2 Fig. 3(a) and 3(b) show OCVs of the two test cells for Co3O4
(a) shows the XRD profile of the sample annealed at 500  C for and MgCo2O4 used as a positive-electrode material. It is seen
24 h after being synthesized by the inverse coprecipitation that the OCV for the Co3O4 positive electrode is about 1.5 V, but
method, which indicates that the sample has a single spinel phase. the initial OCV value for MgCo2O4 is approximately 3.5 V,
Thus, it is found that the present method can produce the spinel which exceeds two-fold of that for Co3O4. However, the OCV
MgCo2O4 sample successfully. The Rietveld analysis revealed value for MgCo2O4 gradually decreases with the rest time
that the sample is of the configurational disordered spinel (defined as the elapsed time without any charging or discharg-
structure (Mg0.6Co0.4)(Mg0.2Co0.8)2O4, on the assumption of the ing). This is due to the structural relaxation caused by the delayed
ideal composition of MgCo2O4. Such a disordered structure may ionic diffusion process in Mg1xCo2O4, Mg2+ ions near the
be preferred in terms of the diffusion of Mg2+ ions inside the surface of which are extracted by charging. Namely, Mg2+ ions
structure. Fig. 2(b) shows temperature dependence of the elec- on the inner side of each MgCo2O4 particle diffuse to spread out
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

trical conductivities of the two samples of MgCo2O4 and Co3O4. to the surface region where the concentration of Mg2+ ions is low
Downloaded by Ryerson Polytechnic University on 15 November 2012

The conductivity of each sample was elevated with increasing by the charge (this is discussed later in Fig. 5). Most importantly,
temperature, which indicates a typical semiconductor. The as is shown in Fig. 3(c), we have successfully demonstrated that
MgCo2O4 sample shows appreciable electrical conductance at the present MIB has sufficiently high voltage to light the blue
room temperature; 2.07  101 S cm1, being comparable to that diode of about 2.6 V band gap. To the authors’ knowledge, to
of LiCoO2, whose conductivity is in order of 102 S cm1.18 From show the MIB indeed working is a first in MIB technology. We
the slopes of Arrhenius plots, the activation energies (the band have also examined the rechargeability of the present MIB
gap in this case) are estimated to be 3.43  102 eV for MgCo2O4
and 1.23  101 eV for Co3O4. The lowering of band gap by
substituting Mg into Co3O4 may contribute to the increase of the

Fig. 3 The OCV values of the test cells after charge: positive electrode,
(a) Co3O4 and (b) MgCo2O4; negative electrode, Mg; electrolyte, Mg
Fig. 2 (a) The structural data determined by the Rietveld analysis shows (ClO4)2 in acetonitrile. For Co3O4, the remaining Co(II) ions can be
the disordered spinel (Mg0.6Co0.4)(Mg0.2Co0.8)2O4. (b) The temperature changed to Co(III), when the other Co(II) ions are extracted into the
dependence of the AC electrical property of MgCo2O4. The MgCo2O4 electrolyte by charge, but Co(III) inevitably changes to Co(IV) for
pellet dimensions for the electrical conductivity measurements are MgCo2O4, because of the absence of the Co(II) ions. (c) The potential of
0.958 cm in diameter, 0.0661 cm in thickness, and the mass density is (+)MgCo2O4|Mg(ClO4)2 in acetonitrile|Mg() revealed by lighting the
2.94 g cm3, indicating the relative density of 0.593. Those of the Co3O4 blue diode having a band gap of about 2.6 V. For reference, it is also
pellet calcined at 800  C are 0.731 cm in diameter, 0.199 cm in thickness, shown that a commercial button cell of 1.5 V cannot light the blue diode,
and the mass density is 3.06 g cm3, meaning the relative density of 0.504. but two cells can do it.

11766 | J. Mater. Chem., 2011, 21, 11764–11772 This journal is ª The Royal Society of Chemistry 2011
 View Online
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A
Downloaded by Ryerson Polytechnic University on 15 November 2012

Fig. 4 Rechargeability of the present MIB with MgCo2O4. Two cycles

were performed only for checking the rechargeability. The charge was
performed under the condition of cc–cv mode (cc: constant current ¼ 50
mA/2 cm2; cv: constant voltage ¼ 4 V). The current spike in the second
charge is experimental noise.

system by using the blue diode; a series of processes is shown in

Fig. 4. After charging MIB for several hours under the cc–cv
condition, and then we made MIB light the blue diode until the
cell voltage decreases down to 2.6 V for about several minutes,
then we recharged it for several hours, and subsequently dis-
charged with a blue diode. Thus, we have succeeded to observe
the rechargeability of the present MIB system.
We can see experimentally the slowness of diffusion of Mg
inside the MgCo2O4 particles. As seen in Fig. 5(a), charging to
some extent can be done with the several times of the charge Fig. 5 (a) Intermittent charge procedure and corresponding OCV value
processes; that is, only one time process cannot charge the cell with the rest time for MIB with MgCo2O4, indicating the slowness of
sufficiently. This is a typical characteristic relating to solid-state diffusion of Mg ions inside the MgCo2O4 particles. (b) Experimental
diffusion. Besides, Fig. 5(b) strongly indicates the slow diffusion results showing the time-dependence of structural relaxation of
of Mg ions in the solid state. Just after the charge, to some extent a Mg1xCo2O4 particle due to the diffusion of Mg2+ ions toward the
(at position A in the figure), the OCV of the MIB is shown to be surface and its corresponding OCV, where the dark red color indicates
the high concentration of Mg2+ ions, the color diluted to orange or light
3.5 V, then the discharge at 1.5 V was made until the current
red means a lesser concentration of Mg2+ ions. A: Mg-poor region near
became almost zero, which took about 1 h (at position B);
surface after charge; B: Mg-rich region after discharge for short duration;
subsequently, when the mode changed to the rest mode, the rapid C: structural relaxation by diffusion of Mg ions from surface;
recovery of OCV was observed up to 2.4 V, and gradually the D: successive relaxation by Mg-ion diffusion from core; E: the next
OCV value decreased to 2.3 V with the rest time. Its total elapsed charge after full relaxation. The current spike to the positive side is
time from A to E was about 100 h, and the fact that such a long experimental noise, but that to the negative side means the done-by-
duration is required to attain the steady state strongly indicates intention to check the dischargeability at the stage.
that the this phenomenon is controlled by the solid-state diffu-
sion of Mg2+ ions inside the active-material particles.
Similarly, to utilize unstable ions such as a higher valence Ni reach 4%-charge in the case of MgCo2O4, it took about 300 h (see
ion, e.g., Ni(IV), is expected to be also effective in obtaining an Fig. 5), whereas it took only for about 150 h to charge about 50%
excellent cell voltage. Fig. 6(a) shows the XRD profile obtained for Mg0.67Ni1.33O2. This is because MIB with Mg0.67Ni1.33O2 is
for Mg0.67Ni1.33O2, which indicates that the sample forms the rock capable of passing relatively higher currents than MgCo2O4,
salt structure (this composition originates from our initial inten- which indicates that the mobility of Mg in Mg0.67Ni1.33O2 is higher
tion of synthesizing the spinel MgNi2O4, but it forms the rock salt than that in MgCo2O4. Of further importance, the OCV value
structure). As seen in Fig. 6(b), the OCV value of the MIB with with a Mg0.67Ni1.33O2 positive electrode is 3.5–3.8 V at the initial
Mg0.67Ni1.33O2 approximates about 3.5 V in the initial stage just stage just after charge, which indicates that Mg0.67Ni1.33O2 seems
after charge; of course, we have confirmed that this MIB can also to be more excellent than MgCo2O4. Fig. 6(c) compares the rates
light the blue diode with rechargeability (not presented here). of the structural relaxation after charge. Since the rate of struc-
Interestingly, for the Mg0.67Ni1.33O2 positive electrode, the tural relaxation depends on the diffusion of Mg inside the host
duration required for some amount of charge is much less than material, the Mg diffusion in Mg0.67Ni1.33O2 is also suggested to
that for MgCo2O4 by several intermittent charge processes; to be faster than that in MgCo2O4.

This journal is ª The Royal Society of Chemistry 2011 J. Mater. Chem., 2011, 21, 11764–11772 | 11767
 View Online
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A
Downloaded by Ryerson Polytechnic University on 15 November 2012

Fig. 6 (a) XRD profile for Mg0.67Ni1.33O2 after annealing (at 350  C, 2
h) the sample synthesized by the inverse coprecipitation method. It is
clearly seen that Mg0.67Ni1.33O2 forms the rock salt structure. (b) The
intermittent charge procedure (with cc–cv mode) and corresponding
OCV value with the rest time (cc: constant current ¼ 50 mA/0.53 cm2; cv:
constant voltage ¼ 4 V). The MIB with Mg0.67Ni1.33O2 can be charged for
a relatively shorter duration than the MIB with MgCo2O4. The sudden
drop of OCV in the first rest time seems to be experimental noise. (c) The
OCV versus rest time curve for three samples. MCO means MgCo2O4,
and MNO denotes Mg0.67Ni1.33O2, and ‘‘in GB’’ means that the cell was
constructed in a glove box filled with Ar gas, ‘‘in air’’ means that the cell
was made under ambient conditions. The structural relaxation in the case
of Mg0.67Ni1.33O2 is found to be faster than that in the case of MgCo2O4. Fig. 7 FE-SEM micrographs of (a) MgCo2O4 (MCO) and (b)
Mg0.67Ni1.33O2 (MNO) of as-synthesized samples and after-charged
samples. XRD profiles for (c) MgCo2O4 (MCO) and (d) Mg0.67Ni1.33O2
(MNO) after annealing (at 350  C for 2 h) the sample synthesized by the
Next we show the FE-SEM micrographs of the initial powder inverse coprecipitation method and the samples after charge to some
of (a) MgCo2O4 (MCO) and (b) Mg0.67Ni1.33O2 (MNO), and the amounts. The charge amount described here is for reference, because the
XRD profiles (c) MgCo2O4 and (d) Mg0.67Ni1.33O2 before and host material is not so stable after deep charge.
after charge tests in Fig. 7; the powder samples before tests are
the as-synthesized ones, and those after the charge tests are of the
active materials on the Al collector. As seen in the FE-SEM that we could easily grind into finer particles. The right micro-
micrographs (left photos), the initial MNO powder appears to be graphs show the morphologies of the active materials (where the
more homogeneous and slightly finer than the initial MCO powder was mechanically milled and pasted on the Al collector)
powder. The characteristic features are that the MCO powder is on the electrode after charging to some extent. From these
relatively ductile (comparatively difficult to get fine particles by photographs, it is found that the MNO particles on the electrode
grinding), but in contrast the MNO powder is hard and brittle, so are finer than the MCO ones on the electrode. These

11768 | J. Mater. Chem., 2011, 21, 11764–11772 This journal is ª The Royal Society of Chemistry 2011
 View Online

microstructural characteristics of MNO can be a reason why the

current is more likely to pass when using the MNO powder as
a positive electrode. Incidentally, before and after the charge
tests, we confirmed that the morphologies of the active materials
on the electrodes were not so significantly changed in both cases.
In the XRD profiles for MCO, after 10% change, the peak
intensity of the strongest peak in the spinel phase tends to
decrease relatively in comparison to the intensity of the Al
collector, but the lattice parameter seems to be unchanged (this is
probably due to a small amount of charge). Also for MNO, even
after charge to some amount, the crystal structure remains still
unchanged, but the whole peaks appear to shift slightly toward
the higher angle side, which may be due to the lattice shrinking
by the extraction of the cations from the host.
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

Finally, Fig. 8 shows the CV voltammograms obtained for

Downloaded by Ryerson Polytechnic University on 15 November 2012

3-electrode (3EL) cells using MgCo2O4 (MCO) or

Mg0.67Ni1.33O2 (MNO) as a positive electrode, in which WE
potential versus RE was controlled in the CV tests and the scan
range of CVs was between 1 and 3.5 V. This low upper limit
(3.5 V) is to avoid the decomposition of the electrolyte for many
scans in CV. The two cells were charged in advance using WE
and CE under the cc–cv condition described in Experimental
section. The initial WR-OCV (the OCV value between WE and
RE) was, in the present CVs, 2.0 V for MCO and 3.2 V for MNO.
Although the WR-OCV for MCO is seemingly lower than the
above value (about 3.0 V) in the charge–discharge tests, this is
due to the structural relaxation and depends on the initial charge
amount, and furthermore in the present cells WR-OCV tends to
be lower by about 0.5 V than the OCV value between WE and CE
in the static state (WC-OCV), which corresponds to the OCV
value in the 2-electrode cell discussed above. Thus, we here
intend to see the reversibility and the characteristic features of the
whole cyclic voltammograms. It is found that the cathodic
currents tend to be larger than the anodic ones and markedly Fig. 8 The cyclic voltammograms for the 3-electrode (3EL) cells with (a)
increase below 1 V (this feature is especially prominent for MgCo2O4 (MCO) and (b) Mg0.67Ni1.33O2 (MNO) positive electrodes.
MNO). This is because the discharge is possible to occur without The two insets in (a) and (b) show, respectively, the wire connection in
any lower limit of voltage, whereas the cell cannot be fully cyclic voltammetry (CV) for the 3EL cell [working electrode (WE):
charged due to the upper limit of 3.5 V. Furthermore, since the Mg0.67Ni1.33O2; counter electrode (CE): Mg; reference electrode (RE):
Mg] and the cyclic voltammogram obtained for a 3EL cell consisting of
cell is not well charged due to this upper limit in each cycle, the
Al-WE, Mg-CE, and Mg-RE.
cathodic currents reasonably decrease gradually with increase in
the cycle number. Incidentally the inset in Fig. 8(b) shows the CV
curve measured for a 3electrode cell consisting of the Al-WE (i.e., be utilized at the valence transition change in the positive-elec-
only a collector), Mg-CE, and Mg-RE. Except the condition trode active materials. In the case of Co3O4, when Co(II) ions are
whether MCO or MNO is present or absent, the other conditions extracted from Co3O4, remaining Co(II) ions could be oxidized to
are almost the same as in the present CV for the MIBs. There- Co(III) ions, whereas Co(III) ions are inevitably oxidized to Co(IV)
fore, in this case, we may consider that the significantly large ions when extracting Mg2+ ions from MgCo2O4. This is essen-
cathodic currents are mainly due to the insertion of Mg2+ ions tially different from the case of the Co3O4 active material,
into the host, although we always have to take notice that there is because mainly the valence change from Co(II) to Co(III) would
a possibility that Mg metal can be electrodeposited on the be realized even when the divalent ions are extracted from
positive electrode (MCO or MNO) and the electrolyte can be Co3O4. As for the MIB with Mg0.67Ni1.33O2, the high-OCV
decomposed. mechanism is uncertain but can be deduced as follows. In view of
the fact that the MIB can deliver such a high voltage, Ni(IV) ions
would be probably formed in the host material after charge. This
4. Discussion
can be realized if DE( / ) < 2DE( / ) when extracting one Mg2+
II IV II III

Based on the present experimental results, we describe a funda- ion from the host, where DE(X/Y) denotes the energy increase due
mental guideline to obtain the excellent cell voltage in the MIB to the oxidization of a Ni ion from X to Y. This will have to be
system. In order to enhance the cell voltage, as shown in Fig. 1 elucidated in near future.
(a), it can be crucial to exploit Co(IV) or Ni(IV) ions that are The problem to be overcome in the present MIB is that the
unstable in the host matrix, namely, the unstable cations should current density is appreciable but low, due to the divalent

This journal is ª The Royal Society of Chemistry 2011 J. Mater. Chem., 2011, 21, 11764–11772 | 11769
 View Online

characteristics as predicted, so that we could not charge and the difficulty of redistribution of the ions’ charge is considered to
discharge the test cells quickly. That is, one of the big drawbacks be a key for the diffusion of Mg2+ ions.6 It should be noted here
of these positive-electrode materials is the high polarization, that MgCo2O4 forms a configurational disordered spinel struc-
which is not good for efficient energy storage. This is mainly due ture (and probably Mg0.67Ni1.33O2 also disordered in terms of the
to the slowness of the solid-state diffusion in the host materials, charge rate and composition), which appears to be advantageous
as discussed below. As seen in Fig. 4, upon charging, the current in terms of the diffusion for intercalation, because the redistri-
density of, at most, about 25 mA cm2 can be realized for short bution of the ions’ charge should be easier in such a disordered
duration (i.e., cc condition), but fleetingly turns to the cv structure, where the configuration of the valences is intrinsically
condition, in which the current decreases gradually with time. As disordered. Thus, to attain a more disordered structure appears
briefly mentioned earlier, this is mostly due to the slowness of to facilitate the diffusion of Mg2+ ions inside the host active
diffusion of Mg2+ ions inside the active-material particles. material.
Therefore, in order to charge to a large extent, we have to do it Finally, we here address the problem that the Mg negative elec-
for a much longer time under a very low current condition. trode is passivated in the acetonitrile–Mg(ClO4)2 electrolyte.
Otherwise, the extraction of Mg2+ ions can be done intermittently Actually, many investigations on the proper electrolyte for MIBs
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

(unless the current is not quite low), as seen in Fig. 5(a) and Fig. 6 have been made intensively to date; the acetonitrile electrolyte with
Downloaded by Ryerson Polytechnic University on 15 November 2012

(b). Namely, after extracting of Mg2+ ions from near the surfaces Mg(ClO4)2 used here has been proposed,14–16 but the passivation
of the active-material particles, we have to wait until the Mg2+ problem has been recently raised in the case of the Mg negative
ions around the cores of the particles diffuse nearer towards their electrode.17 In such a case, it may be thought that the present
surfaces. This typical feature of the solid-state diffusion can be methodology is inadequate to examine positive electrodes for
easily understood from Fig. 5(b). Since the diffusion/mobility of MIBs. However, in the present demonstration, we have tested the
Mg2+ ions inside MgCo2O4 or Mg0.67Ni1.33O2 particles is low, MIBs at the very small current density of less than 50–100 mA cm12.
when MIB is discharged, the cell voltage decreases rapidly Thus, unless Mg is ‘‘completely’’ passivated, the redox reactions
because Mg2+ ions are saturated only near the surface. However, supposed in the present MIBs are possible to occur. Actually, we see
when the mode is changed to the ‘‘rest’’ condition, its OCV value from Fig. 5(a) and (b) and Fig. 6(b) that the rate of the reaction is
recovers to increase gradually, because the inserted Mg2+ ions considered to be controlled by the solid-state diffusion inside the
move into the inner side of the particles and consequently the respective active materials. That is, this should be one of the
concentration of the Mg2+ ions in the vicinity of the surface evidences that the cell reactions are closely related with the positive-
becomes diluted. For further elapsed time, the OCV decreases electrode active material. Taking account of the fact that the present
due to the Mg diffusion from the core to the surrounding. When MIBs work as a battery, it is reasonable to consider that the rate of
the concentration of Mg2+ ions comes to be high to some extent, diffusion inside the active material is much lower than the disso-
we can again extract Mg2+ ions from the host, that is, we can lution rate of the Mg electrode passivated in the electrolyte. In
again charge the cell. Thus, it should be stressed here that the addition, in the light of non-aqueous Mg electrochemistry of today,
intercalation/deintercalation of Mg2+ ions into/from these active probably there are no stable electrolyte solutions with wide enough
materials is surely possible despite the solid-state diffusion of electrochemical windows in which Mg electrodes are truly reversible
divalent ions. for many cycles. In the present demonstration, since at most several
The diffusion mechanism in the spinel structure is briefly dis- charge/discharge tests were performed, it is conceivable that the
cussed based on the two papers.6,19 According to Ishizawa and electrolyte remains stable during the charge/discharge tests from the
Tateishi,19 the diffusion of Li+ ions located at 8a sites in the kinetics point of view, not from the view of thermodynamics.
LiMn2O4 normal spinel occurs via the octahedral 16c interstices In order to obtain more evidence for the abilities of these active
(the largest spatial void in the spinel structure). The radius of the materials, we performed some additional experiments by using
 whereas the radius of the Mg2+ ion is 0.66 A
Li+ ion is 0.68 A,  and the Cu negative electrode and the frequently used electrolyte for
almost the same as that of the Li+ ion. Thus, the diffusion Li ion batteries (1 M LiPF6 in a solution of 33% ethylene
mechanism may be discussed in terms of hopping activation carbonate and 67% dimethyl carbonate). This system is very
energy controlled by the ion charges and atomic configuration. stable, in that only a small amount of Mg is electrochemically
This geometric situation can be also applied to the present cases deposited on the Cu negative electrode while the battery is
of MgCo2O4 regardless of the normal or disordered spinel, and charged; Li metal is considered to be mainly deposited on the Cu
probably the octahedral 16c interstices play a significant role in negative electrode and Li is not passivated in the electrolyte.
the diffusion of Mg2+ ions with the similar ionic radius. In Here, we have tested four types of the materials as the positive
general, it is believed that the divalence is disadvantageous for electrode: MgCo2O4, Mg0.67Ni1.33O2, MgNiO2, and carbon
the ionic conductor, because the activation energy for diffusion black (CB) only. First, each battery was charged under the cc–cv
or ion hopping is roughly predicted to be large in proportion to z2 condition, and then we observed the OCVs after 30 min.
(z: the valence), but this is just an intuitive understanding. Subsequently, the electrolyte was renewed to remove the possi-
According to the literature,6 since the repulsion exceeds the bility that the test cell is a sort of electrochemical capacitor, and
attraction if the interatomic distance decreases, the strong then the OCV was again monitored with the rest time. As is
attractions between inserted Mg2+ ions and the host anions clearly seen in Fig. 9, the OCV values for MgCo2O4
should rather decrease the potential barriers by decreasing the Mg0.67Ni1.33O2 and MgNiO2 maintain high values of about 2.5–
repulsion upon squeezing, thus promoting the diffusion of Mg2+ 3.5 V, even though the electrolyte is renewed. Furthermore, we
ions. The most important thing is to achieve the local electro- have also confirmed by SEM-EDX that Mg and also Co, Ni are
neutrality when accepting the inserted Mg2+ ions, i.e., to reduce deposited on the Cu electrode (not presented here). Thus, the

11770 | J. Mater. Chem., 2011, 21, 11764–11772 This journal is ª The Royal Society of Chemistry 2011
 View Online
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A
Downloaded by Ryerson Polytechnic University on 15 November 2012

Fig. 9 The OCV re-examination by using the Li-ion battery system. The negative electrode is Cu metal, the electrolyte is an ethylene carbonate and
dimethyl carbonate with LiPF6 salt, which are commonly used for the LIB systems. (a) Schematic illustration of constructing the test cell and the way to
measure the OCV. After charging of the cell, we have checked with FE-SEM that Mg and also Ni are deposited on the Cu surface, indicating that
divalent Mg and Co, Ni ions are deintercalated from the host active materials. (b) The time-dependence of the cell voltage (OCV) of the test cell with each
positive-electrode material. ‘‘Electrolyte renewed’’ means that the electrolyte was exchanged to a new one on the way of the measurement to remove any
effects of the capacitor.

values of charge amount in Fig. 6 etc. are not accurate in this a high cell voltage, it would be significant to exploit the unsta-
viewpoint. Of course, these high OCV values are obtained versus bilized ion after charge, such as Co(IV) or Ni(IV) ions in the host
Li metal, not Mg metal; therefore, when it is versus Mg, the OCV matrix after extracting divalent ions by charge.
may be reduced. Namely, this means that the validity of these (3) We have also observed in Fig. 5 the diffusion-like behavior
active materials was checked in the Li ion battery system. inside the solid active materials, which suggested that the cell
However, these experiments strongly indicate that the Mg ions reaction is strongly related to the solid active materials.
can be deintercalated from the host matrix and the transition (4) Since the Mg negative electrode tends to be passivated in
metals can then be oxidized to higher values. the acetonitrile electrolyte, we have also confirmed the potenti-
ality of MgCo2O4, Mg0.67Ni1.33O2 and MgNiO2 active materials
by utilizing the commonly used Li-ion battery system. In
5. Conclusions
particular, the MgNiO2 positive electrode can yield a high OCV
In conclusion, considering that MIB will be one of candidates of (3.5–3.9 V) versus Li/Li+.
post-LIB in the future, we have investigated the potentiality of Thus, judging from overall experimental results, the present
the magnesium complex oxides (MgCo2O4, Mg0.67Ni1.33O2, active materials are considered to possess high potential to
MgNiO2, etc.) for the MIBs, which are composed of a Mg negative deliver high voltage as the positive electrodes in MIB systems.
electrode and an acetonitrile electrolyte with 1 M Mg(ClO4)2. However, there are many remaining problems to be settled in the
(1) MgCo2O4 was found to show an appreciably high elec- future for practical use, such as seeking a more appropriate wide-
trical conductance at room temperature: 2.07  101 S cm1, potential window electrolyte to circumvent the passivation
being comparable to that of LiCoO2. problem of the Mg negative electrode and dissolution in the
(2) We have presented that the MIBs with the positive elec- electrolyte, enhancing the current density, etc. Although the MIB
trodes (using MgCo2O4, Mg0.67Ni1.33O2, etc.) exhibit high OCVs study tends to be undermined due to its low cell voltage, we
exceeding 3 V, and have demonstrated that these batteries can believe that the present study to gain an excellent cell voltage will
light a blue diode with a 2.6 V bandgap. In order to attain such trigger off much more extensive works on the MIB technology.

This journal is ª The Royal Society of Chemistry 2011 J. Mater. Chem., 2011, 21, 11764–11772 | 11771
 View Online

Acknowledgements 5 J. O. Besenhard and M. Winter, ChemPhysChem, 2002, 3, 155–159.

6 E. Levi, M. D. Levi, O. Chasid and D. Aurbach, J. Electroceram.,
We are very grateful to Professor Eiichiro Matsubara for his 2009, 22, 13–19.
7 N. Kamioka, T. Ichitsubo, T. Uda, S. Imashuku, Y. Taninouchi and
great experimental help and various support, and also thank E. Matsubara, Mater. Trans., 2008, 49, 824–829.
Professor Tetsuya Uda and Mrs Natsumi Kamioka for great 8 D. Klissurski and E. Uzunova, J. Mater. Sci., 1994, 29, 285–293.
experimental help and fruitful discussion. This work was also 9 D. G. Klissurski and E. L. Uzunova, J. Mater. Sci. Lett., 1990, 9,
partly supported by Grant-in-Aid for ‘‘Wakate Kenkyu A 1255–1258.
10 D. G. Klissurski and E. L. Uzunova, Chem. Mater., 1991, 3, 1060–
(22686064)’’ from JSPS, Japan. One of the authors, T. I., is very 1063.
grateful to his mother, Nobuko Ichitsubo, for various supports 11 U. Chellam, Z. P. Xu and H. C. Zeng, Chem. Mater., 2000, 12, 650–
to date, and dedicates this article to her who passed away on 658.
12 P. Peshev, A. Toshev and G. Gyurov, Mater. Res. Bull., 1989, 24, 33–
January 26th, 2011, and also to his father, Takeshi Ichitsubo,
40.
who was devoted to looking after her. 13 V. S. Holgersson and A. Karlsson, Z. Anorg. Allg. Chem., 1929, 183,
384–394.
References 14 P. Novak and J. Desilvestro, J. Electrochem. Soc., 1993, 140, 140–144.
Published on 01 July 2011 on http://pubs.rsc.org | doi:10.1039/C1JM11793A

15 P. Novak, W. Scheifele and O. Haas, J. Power Sources, 1995, 54, 479–

1 D. Aurbach, Z. Lu, A. Schechter, Y. Gofer, H. Gizbar, R. Turgeman, 482.
Downloaded by Ryerson Polytechnic University on 15 November 2012

Y. Cohen, M. Moshkovich and E. Levi, Nature, 2000, 407, 724–727. 16 P. Novak, W. Scheifele, F. Joho and O. Haas, J. Electrochem. Soc.,
2 T. D. Gregory, R. J. Hoffman and R. Winterton, J. Electrochem. Soc., 1995, 142, 2544–2550.
1990, 137, 775–780. 17 Z. Lu, A. Schechter, M. Moshkovich and D. Aurbach, J. Electroanal.
3 M. Morita, N. Yoshimoto, S. Yakushiji and M. Ishikawa, Chem., 1999, 466, 203–217.
Electrochem. Solid-State Lett., 2001, 4, A177–179. 18 M. Menetrier, I. Saadoune, S. Levasseur and C. Delmas, J. Mater.
4 P. Novak, R. Imhof and O. Haas, Electrochim. Acta, 1999, 45, 351– Chem., 1999, 9, 1135–1140.
367. 19 N. Ishizawa and K. Tateishi, J. Ceram. Soc. Jpn., 2009, 117, 6–14.

11772 | J. Mater. Chem., 2011, 21, 11764–11772 This journal is ª The Royal Society of Chemistry 2011