Structural analysis of electrolyte solutions for rechargeable Mg batteries by stereoscopic means and DFT calculations

Article abstract of DOI:10.1021/ja1098512

We present a rigorous analysis of unique, wide electrochemical window solutions for rechargeable magnesium batteries, based on aromatic ligands containing organometallic complexes. These solutions are comprised of the transmetalation reaction products of Ph_xMgCl_2-x and Ph_yAlCl_3-y in different proportions, in THF. In principle, these reactions involve the exchange of ligands between the magnesium and the aluminum based compounds, forming ionic species and neutral molecules, such as Mg₂Cl₃⁺·6THF, MgCl₂· 4THF, and Ph_yAlCl_4-y^- (y = 0-4). The identification of the equilibrium species in the solutions is carried out by a combination of Raman spectroscopy, multinuclear NMR, and single-crystal XRD analyses. The association of the spectroscopic results with explicit identifiable species is supported by spectral analyses of specially synthesized reference compounds and DFT quantum-mechanical calculations. The correlation between the identified solution equilibrium species and the electrochemical anodic stability window is investigated. This study advances both development of new nonaqueous solution chemistry and possible development of high-energy density rechargeable Mg batteries.

Full text of DOI:10.1021/ja1098512

ARTICLE
pubs.acs.org/JACS
Structural Analysis of Electrolyte Solutions for Rechargeable Mg
Batteries by Stereoscopic Means and DFT Calculations
Nir Pour, Yossi Gofer, Dan T. Major, and Doron Aurbach
Department of Chemistry and the Lise Meitner-Minerva Center of Computational Quantum Chemistry, Bar-Ilan University,
Ramat-Gan 52900, Israel
S Supporting Information
b
ABSTRACT: We present a rigorous analysis of unique, wide electrochemical
window solutions for rechargeable magnesium batteries, based on aromatic
ligands containing organometallic complexes. These solutions are comprised of
the transmetalation reaction products of PhxMgCl2ꢀx and PhyAlCl3ꢀy in
different proportions, in THF. In principle, these reactions involve the exchange
of ligands between the magnesium and the aluminum based compounds,
þ
forming ionic species and neutral molecules, such as Mg2Cl3 6THF, MgCl2
3
3
ꢀ
4THF, and PhyAlCl4ꢀy (y = 0ꢀ4). The identification of the equilibrium
species in the solutions is carried out by a combination of Raman spectroscopy,
multinuclear NMR, and single-crystal XRD analyses. The association of the
spectroscopic results with explicit identifiable species is supported by spectral
analyses of specially synthesized reference compounds and DFT quantum-
mechanical calculations. The correlation between the identified solution
equilibrium species and the electrochemical anodic stability window is inves-
tigated. This study advances both development of new nonaqueous solution
chemistry and possible development of high-energy density rechargeable Mg
batteries.
’ INTRODUCTION
voltammograms (CV) with Pt working electrode, displayed electro-
chemical windows in excess of 3.3 V vs. Mg.5 Additionally, these
solutions possess specific conductivity of ∼2 ꢁ 10ꢀ3 Ωꢀ1 cmꢀ1 at
RT, and high Mg cycling efficiency. The widened electrochemical
stability window attained with the all phenyl complex (APC)-type
electrolytes is critical for future improvements of feasible recharge-
able magnesium batteries, particularly with respect to energy density.
In previous papers, we reported on the structural analyses of
the DCC family of electrolytes.1ꢀ4,6 The current study follows a
similar methodology for the complete identification of the whole
range of species existing in the APC solutions.
During the past decade we had been developing new classes of
nonaqueous electrolyte solutions, from which metallic magnesium
can be electrochemically deposited reversibly, and magnesium ions
may intercalate in appropriate inorganic hosts.1,2 The first electrolyte
solutions presented, the DCC (“Dichloro complex”) electrolyte
family, were based on the reaction products of the Lewis bases
RxMgCl2ꢀx with a variety of Lewis acids R0yAlCl3ꢀy (R,R0 = n-butyl
and/or ethyl, x = 0ꢀ2, y = 0ꢀ3), all dissolved in THF. The
equilibrium species in these solutions, which dominate its electro-
chemical characteristics, are the consequence of transmetalation
reactions, where the organic and inorganic ligands are rapidly
exchanged between the magnesium and aluminum cores.3 The
solvent also plays an important role in these reactions, both as a
strong donor and as a polar medium that yields high ionic
conductivity solutions. These electrochemical systems exhibit an
electrochemical window of ∼2.2 V, specific conductivity of few
10ꢀ3 Ωꢀ1 cmꢀ1, and as high as 100% Coulombic reversibility. The
main disadvantage of these electrolytes is their limited electroche-
mical anodic stability that limits its utilization in high-energy density
cells. The limited anodic stability of these electrolytes was proved to
be the due to the relatively weak aluminumꢀcarbon bond that
electrochemically breaks via β-H-elimination.2,4
In addition to the field of magnesium electrochemistry and
rechargeable batteries, these families of organometallic molecules
and their exact structures play central roles in high performance
polymerization schemes as specific function catalysts.
’ EXPERIMENTAL SECTION
All the chemical syntheses and electrochemical measurements were
carried out under pure argon atmosphere (99.999%) in a M. Braun, Inc.
glovebox (less than 1 ppm of water and oxygen).
Chemicals and Syntheses. All electrolyte solutions were synthe-
sized according to the same procedure: First, AlCl3 (Aldrich, anhydrous,
99.999%) in THF (Aldrich, 99.9%, less than 20 ppm water) solution was
Hence, we theorized that substituting the aliphatic ligands
for phenyl, which does not possess β hydrogen, will immunize
the complex salts to this oxidation pathway. Indeed, cyclic
Received: December 5, 2010
Published: April 01, 2011
r
2011 American Chemical Society
6270
dx.doi.org/10.1021/ja1098512 J. Am. Chem. Soc. 2011, 133, 6270–6278
|

Journal of the American Chemical Society
ARTICLE
prepared in the desired concentration by very slow addition to the
vigorously stirred solvent. Then, this solution was added dropwise to a
predetermined quantity of 2 M Phenylmagnesiumchloride (PhMgCl)
solution in THF (Aldrich). Both reactions are very exothermic. The
resulting solution was stirred for additional 16 h or more (at room
temperature).
The family of electrolytes containing phenyl ligands in various propor-
tions is referred to in this paper as “APC”. Electrolyte concentrations were
reported in terms of the aluminum-based species concentration. If not
indicated otherwise, the AlCl3 to PhMgCl ratio is 1:2. Thus, a 0.25 M APC
solution is synthesized by reacting equal quantities of 0.25 M AlCl3 and
0.5 M PhMgCl in THF. The standard electrolyte, to which others are
compared, is 0.25 M APC, as it exhibits favorable electrochemical properties.
In cases where solutions were prepared with AlCl3 and PhMgCl at ratios
other than 1:2, they are denoted as APC m:n solutions (m and n are the
relative amounts of AlCl3 and PhMgCl, respectively).
Synthesis of AlPh3 THF Reference Compound7. A 2 M THF solution
3
Figure 1. Raman spectra of (a) 0.5 M PhMgCl in THF in comparison
with reference spectra, all in THF (top to bottom): (b) toluene, (c)
benzene, and (d) THF. (Inset) (e) MgCl2 in THF in comparison
to THF.
of phenylmagnesium chloride (22.5 mmol) was slowly added to a 1.2 M
solution of AlCl3 (1.0 g, 7.5 mmol) in THF (6.25 mL) at 0 °C. The
solution was then stirred for 14 h at RT, after which, the solvent was
removed under reduced pressure. The solid residue (mainly MgCl2) was
separated and removed by toluene extraction and centrifuging. The
response, where the I vs E curves showed an appreciable increase in the
positive currents. Preparative, potentiostatic [3.00 V vs Mg] electrolysis
was carried out in a single compartment cell, passing 150 C (∼4 h)
(accounts to approximately 2 electrons per half the total amount of
anions). Magnesium foil served as the counter and reference electrode
[60 cm2], and stainless steel net [24.5 cm2] served as the WE.
Measurements were performed at 27 °C ( 3.
Single-Crystal XRD. Single-crystals (∼1ꢀ4 mm) were selected,
washed with dry THF, and transferred (slightly wet) in a sealed glass vial
under argon, to the X-ray facility. The crystals were coated with inert
crystallography oil and were flash-frozen with ∼100 K nitrogen. The
measurements were carried out using the Bruker “Kappa Apex II” system.
Data collection was performed at cryo-temperatures under nitrogen atmo-
sphere with Mo KR, 0.07103 nm, radiation @ φ and ω scans. Structure
calculation and refinement were performed with the SHELXS-97 program.
white powder left after toluene removal contained mainly Ph3Al THF
3
with less than 10% (molar) MgCl2. NMR indicated a Ph:THF ratio of
3:1 (see Figure S1, Supporting Information [SI]).
Other syntheses followed well-documented procedures, and are
similar to the one above.
Elemental compositions (Al, Mg, Cl): were determined by ICP and
SM-4500-Clꢀ D (for Cl) at Aminolab Inc. (analytical laboratories) Israel.
For these measurements, solution samples of 2 mL were withdrawn and
reacted with 40 mL of doubly distilled water, to which few drops of
concentrated nitric acid (Merck, 65%, extra pure) were added. Extra dry
THF and other solvents were further dehydrated with activated 4A
molecular sieves (Aldrich 4 Å, beads, 8ꢀ12 mesh). Trace water was
monitored by Karl Fischer titration with Metrohm 652 KF coulometer.
Spectral Studies. Raman spectra were measured with a JY Horiba
spectrometer using a HeꢀNe laser (632.817 nm) in screw-cap rectan-
gular spectrozile quartz cuvettes. All experiments were performed at RT
(25 ( 4 °C). We focusd on the 100ꢀ1150 cmꢀ1 spectral region, which
was found to be the most informative for the identification of the
organometallic compounds.
NMR spectra were measured with a Bruker DMX-600 spectrometer.
DFT Computations. All aluminum core geometry optimizations
were performed at the B3LYP/6-31þG* level of theory.8ꢀ10 Magne-
sium core geometry optimizations were done at the B3LYP/CEP-
31þG* level in the gas-phase and in solution. The solvent (THF, ε =
7.58) was taken into account via the IEFPCM model.11 We failed to
minimize the solvated PhAlCl3ꢀ molecule at the B3LYP/6-31þG* due
to a very flat potential energy surface; the optimization of this molecule
was done at the B3LYP/6-31G* with a single point calculation at
B3LYP/6-31þG*. Optimization at MP2/6-31þG** and B3LYP/6-
31þG** levels gave nearly identical results.
’ RESULTS AND DISCUSSION
Preliminary NMR and Raman spectroscopy studies with APC-
type solutions revealed that this chemical system is by no means
simple. Not only did the spectra exhibit numerous features, some
very weak and wide, there were also no spectral reference library
upon which the identification of the species could be based on.
Thus, as for the aliphatic system,6 in order to realize a complete
identification of the species we had to establish a comprehensive
reference spectra library containing the various possible compounds.
Only a few magnesium species exist in the complex solutions,6
the most important of which are MgCl2 MgClþ, Mg2Cl3
,
þ
Ph2Mg and PhMgCl, all coordinated by THF. Contrary to a
recent publication,15 we assume that in all the above cationic Mg
species, the magnesium atom is hexacoordinated by a combina-
tion of Clꢀ and THF ligands (for instance, see Figure 2). At
relatively low concentrations in THF, it is reasonable to assume
that these compounds exist predominantly as monomers.16
Figure 1 presents the Raman spectra for 0.5 M PhMgCl in THF,
and for comparison, pure THF, toluene and benzene in THF.
As can be seen in Figure 1, the spectrum of THF over this
region is characterized by six peaks: 289 (low and wide), 585 and
655 (very low and wide—seen for clarity in the inset) 914 (very
high), 1031 (medium), and 1072 cmꢀ1 (low).17 The spectrum of
toluene in THF exhibits the same set of peaks related to the
solvent, plus additional peaks, of which the most important one
Theoretical Raman spectra were calculated at the B3LYP/6-31þG*
level in the gas phase. A scaling factor of 0.965 was applied.12 Insights
regarding electron distributions were gained by natural bond analysis13-
(NBO), performed using B3LYP/6-31þG*. All NBO and energy
calculation were performed in solution using Gaussian 09 package.14
Electrochemistry. Electrochemical measurements were per-
formed with an EG&G PAR 273 potentiostat controlled by “Corrware”
software (Scribner, Inc.). Magnesium foils served both as the counter
and reference electrodes in three-electrode cells. Flame-cleaned plati-
num wire working electrodes were used for studying the Mg deposition-
dissolution processes and the anodic stability window of the solutions,
which was determined at the deflection point in their potentiodynamic
6271
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
Table 1. Summary of the Most Characteristic Peak Positions
in Raman Spectra for Some Reference Materials (M = Al
or Mg)
peak position, cmꢀ1
Figure 2. Structures of C4 and C2 solids (obtained from single-crystal
XRD) precipitated from THF solutions of 0.4 M AlCl3 þ 0.8 M
PhMgCl, and of 0.25 M AlCl3 þ 0.125 M PhMgCl, respectively.
species
type of vibration MꢀPh
type of vibration MꢀCl
PhꢀMg-Cl
MgCl2
635 (“E2g”)
993 ('B1u')
212
330
AlCl3 THF
ꢀ3
AlCl4 THF
180 (T2)
1031
348 (A1)
1072
3
THF
289
585ꢀ655
∼1000
phenyl group
624 ('E2g')
for this study is located at 624 cmꢀ1 (denoted as 'E2g', corre-
sponding to the notation for benzene, pertaining to phenyl ring
wagging, devoid of CꢀR stretching). As will be presented later,
this peak appears whenever phenyl group is present in the
system, regardless of its chemical environment, except for
benzene. For benzene, as shown in Figure 1, this peak appears
at 609 cmꢀ1. As will be seen later, the presence of a phenyl ligand
invariably results in the appearance of peaks around 1000 cmꢀ1
whose18 specific location depends both on the core metal
(magnesium or aluminum), and on the number of chlorine to
phenyl ligands ratio. The spectrum of 0.5 M PhMgCl/THF
displays three peaks, at 624 ('E2g'), 635 (“E2g”) (“ represent the
'E2g' phenyl ring wagging, combined with CꢀR stretching) and
993 cmꢀ1 ('B1u'). The peak at 635 cmꢀ1 is associated with a
PhꢀMg vibration. The expected feature around 1000 cmꢀ1
appears, in this case, as a single peak at 993 cmꢀ1. Unfortunately,
the reference Raman spectrum for Ph2Mg solution, synthesized
according to standard procedure19 showed only the THF peak at
914 cmꢀ1 due to strong fluorescence. In a previous Raman study6
we showed that MgCl2 in THF (inset) displays a single char-
acteristic peak, at 212 cmꢀ1. For further identification of the
Raman bands we were assisted by the results from the DFT
calculations. Figure S2 (SI) exemplifies the close correlation
between the DFT calculated and the experimental spectra.
In the course of establishing a reliable reference library of
spectra, we also measured the spectra of various aluminum
compounds containing only chlorine ligands (see Figure S3, SI).
Table 1 summarizes the relevant peaks for the above reference
materials.
Previous studies have shown that the transmetalation reac-
tions between organoꢀhalo aluminum and magnesium com-
pounds can be understood in terms of Lewis acidꢀLewis base
reactions.1ꢀ3 The reactions are frequently very exothermic and
result in new species, neutral and/or ionic, depending on the
reactants concentration, ratio, reaction temperature, and the
solvent used. In these reactions, the aluminum species are acidic,
and the magnesium ones are bases. Three important character-
istics were discovered regarding these reactions:
1. Aluminum is preferentially coordinated by organic ligands,
while the magnesium is preferentially coordinated by
chlorine and THF.
2. The final equilibrium species in the solution, whether ionic
or neutral, depend on the relative acidity (basicity) of the
products.
Figure 3. Raman spectra of (a) redissolved C4 in THF (saturated), (b)
∼1 M THF in water, (c) solid C4 precipitated from solution composed
of 0.4 M AlCl3 þ 0.8 M PhMgCl, and (d) solid C2 precipitated from
solution composed of 0.25 M AlCl3 þ 0.125 M PhMgCl.
When analyzing the electrochemical properties of such a
complex solutions, the exact equilibrium species present in the
solution are of paramount importance. Clear correlation was
established between the chemical and the electrochemical char-
acteristics of the solution, and its constituents for the DCC
case.1ꢀ3,5 Of the different solution-species, the one dominating
the anodic stability limit is the most oxidation-vulnerable moiety.
When we first started exploring the solution structure of the
phenyl ligand-containing complexes, we hypothesized that the same
principles found previously for DCC solutions would also govern the
reactions and the equilibrium products of this family of materials.5
Indeed, when PhMgCl solution reacts with AlCl3 solution, a very
exothermic reaction ensues, yielding a solution with high ionic
conductivity, which testifies that new materials are being generated.5
As demonstrated below, for the APC solutions, the systems indeed
undergo a complete trans-metalation of the organic ligands from
magnesium to aluminum. However, in this case we obtain a unique
mixture of organoꢀaluminum and chloroꢀaluminum species, of
seemingly very different Lewis acidities, coexisting in equilibrium.
This peculiar phenomenon is discussed further in the paper. The rich
and complex mixture of anionic and molecular species, and occa-
sionally precipitates, formed in APC solutions makes the analysis of
even the simplest solutions extremely challenging.
For instance, transparent, colorless crystals precipitated from
any solution made with 2:1 Al:Mg ratio, and in solutions with 1:2
Al:Mg, at 0.3 M or above, when synthesized at RT. Such crystals,
C2 and C4, respectively, analyzed by single-crystal X-ray diffrac-
tion (SCXRD) and solid-state NMR. Figure 2 displays the results
of the SCXRD. 27Al SSNMR of C4 fully supported the SCXRD
results (see Figure S5, SI).
3. The equilibrium aluminum species are a mixture of various
combinations of compounds, characterized by Cl/R
ratios.1ꢀ3,5
6272
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
Table 2. Calculated Raman Peak Positions for OrganoꢀHalo Aluminum Anions and Molecules in THF
anion
peak position, cmꢀ1
molecule
peak position, cmꢀ1
Ph4Alꢀ
Ph3AlClꢀ
186
193
614 (E2g)
614 ('E2g')
614 ('E2g')
614 ('E2g')
167 (T2)
638 (E2g)
644 (“E2g”)
652 (“E2g”)
662 (“E2g”)
326 (A1)
Ph3Al THFꢀ
196
272
275
614 ('E2g')
646 (“E2g”)
657 (“E2g”)
672 (“E2g”)
3
ꢀ
Ph2AlCl2
252
Ph2AlCl THF
614 ('E2g')
613 ('E2g')
3
ꢀ
PhAlCl3
276
PhAlCl2 THF
3
ꢀ
AlCl4
107 (E)
The Raman spectra of these crystals, containing tꢀhe unam-
biguously identified Mg2Cl3þ, Ph4Alꢀ, and AlCl4 species
(coordinated by THF), were also used for the Raman spectra
library. Figure 3 displays the Raman spectra of C4 (solid), C2
(solid), and redissolved C4 saturated solution in THF.
wagging ('E2g'), devoid of CꢀAl stretching, and the peaks in the
638ꢀ662 cmꢀ1 region relate, in all cases, to the ring stretching
(“E2g”) mode, combined with CꢀAl stretching. The appearance
of distinct and unique bands for each of the different ions allows
the identification of the aluminum core species in the solutions,
based solely on Raman spectroscopy.
C4 (c) and C2 (d) spectra are very unique, as they exhibit
unambiguously identified components, one for each metal core,
with the addition of crystal-bound THF. The crystal-bound THF
spectral characteristics, as expected, are somewhat different from
those of the free THF. In these spectra the main THF peak at
914 cmꢀ1 splits into two narrow peaks.17 In order to confirm that
these peaks indeed belong to THF, we dissolved ∼1 M THF in
water [see Figure 3b]. In this state the THF molecules are solvated
by water, and thus, presumably, lack THFꢀTHF interactions. This
assisted in identifying and excluding all the crystal-bound THF
peaks appearing in the C4 and C2 solids spectra. The C2 crystal
exhibits only four peaks, neglecting the THF ones: three at 115(E),
180(T2), and 348(A1) cmꢀ1, attributed to AlCl4ꢀ. The very weak
feature at 241 cmꢀ1, pertaining6 to Mg2Cl3þ, is attributed to a minor
residue. As did Sobota et al.17 who also investigated similar
complexes, we did not detect any features attributable to MgClþ.
Due to the low solubility of C2 in THF, no Raman spectrum of its
solution could be obtained.
Several interesting observations and trends may be deduced from
the calculated data and their relation to the measured values
(Table 2). First, as can be seen from the listed values, the calculated
spectra predict a clear trend for the first and third row peak positions
versus Ph:Cl ratio. The higher this ratio, the higher the peak position
in cmꢀ1, although the computed vibrations are consistently lower
than the experimental ones. Additionally, the peaks associated with
the middle column are practically constant, regardless of the Ph:Cl
ratio, both for the measured and the calculated spectra. This result is
expected as the 'E2g' vibration, at 624 cmꢀ1 involved only the
benzene-ring bonds.
Previous multinuclear NMR analysis3 of a homologue series of
compounds containing aliphatic ligands showed clearly that the 27Al
ꢀ
NMR spectra for the anions RyAlCl4ꢀy and the molecules
Ry AlCl3ꢀy THF (y = 1ꢀ4; y0 = 1ꢀ3) are identical whenever
0
3
y = y0; e.g. R2AlCl2ꢀ exhibits the same spectrum as R2AlCl THF.
3
Hence, from 27Al NMR analysis point of view, the Clꢀ ligand has the
same influence on the chemical shift as THF. Interestingly, we
observed a similar trend in the Raman spectra for the Al core anions
and the neutral molecules in the current study. For instance, the
most important features of the calculated and experimental Raman
þ
The C4 (c) spectrum displays clearly the expected Mg2Cl3
peak as a strong feature at 241 cmꢀ1. The other peaks, at 201,
624(”E2g'), 649(“E2g”), 985, 997, and 1152 cmꢀ1 are all attrib-
uted to the symmetrical tetraphenyl aluminate ion. The peak at
624 cmꢀ1 invariably appears whenever a phenyl group is present
in the system.
spectra for Ph3AlClꢀ are very similar to those for Ph3Al THF. The
3
calculated Raman spectra show the same trend for all the vibrations,
ꢀ
Interestingly, as can be seen in spectra (a) and (c) of Figure 3,
there is no difference in the peak positions for C4 as solid crystal or as
THF solution. This later allowed with great certainty the identifica-
tion of the existence of both Mg2Cl3þ and Ph4Alꢀ in solutions.
Spectrum (a) is characterized by five unique peaks, at 201, 649-
(“E2g”), 985, 997, and 1152 cmꢀ1. As gathered from the NMR,
elemental analysis, and the Raman spectra, the peaks in the
1000 cmꢀ1 region pertain solely to phenyl ligands around the
aluminum core. The exact peak positions change very little with the
nature of the aluminum-core molecule or ion, regardless of the Ph:Cl
ratio. The 201 and 649 cmꢀ1 peaks, on the other hand, are exclusive
to Ph4Alꢀ, and exist both for the solid C4 crystal, and its solution,
though with some peak-shape difference for the 201 cmꢀ1 peak.
We calculated the Raman spectra for the five possible
PhyAlCl4ꢀyꢀ ions (y = 0ꢀ4), and their neutral forms, focusing
only on the most significant (intense) peaks. Each molecule has a
single strong peak in the 250 cmꢀ1 region and two strong peaks
in the 650 cmꢀ1 region. Table 2 summarizes the relevant
calculated peaks for these materials:
except for the lowest vibration for Ph2AlCl2 and Ph2AlCl THF
3
(252 and 272 cmꢀ1, respectively). However, careful 13C NMR
analysis of the spectra of a wide range of anionic and neutral species
indicated small but significant and consistent differences that make a
distinction between these species. Based on these results (to be
published soon) we conclude that no detectable levels of aluminum-
based neutral compounds exist in the above solutions.
It is worth mentioning that the peak at 201 cmꢀ1 in the C4
spectrum is very close to the 212 cmꢀ1 peak of MgCl2, and to one
for the calculated Ph3AlClꢀ peak. This, in some cases, hampers
the unambiguous identification of all the components in some of
the more species-rich solutions.
Two important reference solutions provided important spec-
tral information: Ph3Al THF in THF and an APC solution
3
containing PhAlCl3ꢀ as the major anion. The later solution, APC
1:1 (cs), was synthesized by reacting equimolar quantities of
PhMgCl with AlCl3 in THF at 0ꢀ°C. NMR analysis indicated that
this solution contains PhAlCl3 as the major anionic product
plus low concentration of Ph2AlCl2ꢀ (see Figure S6, SI).
Not only these bands are the most intense in the specified
region, but also each column represents the same vibrational
mode, as judged from visualization of the normal vibrational
modes. For instance, the 614 cmꢀ1 band pertains to phenyl ring
Figure 4 displays the Raman spectra of Ph3Al THF/THF and
APC 1:1 (cs) solutions.
3
As can be seen in Figure 4, both Ph3Al THF and APC 1:1 (cs)
3
solutions exhibit the expected peaks at 624, 985, 997, and
6273
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
Table 3. Summary of Peak Positions for the Measured Raman
Spectra of OrganoꢀHalo Aluminum and Magnesium Species
in THF
species
peak position, cmꢀ1
PhMgCl nTHF
PhAlCl3ꢀ/PhAlCl2 THF
Ph2AlCl2ꢀ/Ph2AlCl THF
Ph3AlClꢀ/Ph3Al THF
624a 635
993
3
297
270
624 680 985 997 1027 1152
624 667 985 997 1027 1152
624 658 985 997 1027 1152
624 649 985 997 1027 1152
3
3
209
3
Ph4Alꢀ
201
MgCl2 4THF
212
3þ
Mg2Cl3 6THF
241
3
MgClþ 5THF
unknown
330
3
AlCl3 THF
860
ꢀ3
AlCl4
180 (T2) 348 (A1)
a The 600 cmꢀ1 region vibrations pertain to E2g.
Figure 4. Raman spectra of (a) Ph3Al in THF (b) solution made by
reacting 0.25 M AlCl3 þ 0.25 M PhMgCl (APC 1:1 (cs)) in THF at 0
°C. (Inset) Comparison between (a) Ph3Al in THF and (c) MgCl2 in
THF. [(cs) = cold temperature synthesis].
This spectrum reflects the coexistence of a wide variety of
aluminum species, with Mg2Cl3þ as the major magnesium species.
All the peaks discernible in Figure 5, apart from the two at 270 and
667 cmꢀ1, have already been identified and presented in
Figures 2ꢀ4 and S2ꢀS3 (SI). On the basis of the various
possibilities for Al-containing solution species and the NMR data,
ꢀ
we hypothesized that these two peaks pertain to Ph2AlCl2
.
Moreover, this pair of peaks and their approximate positions were
predicted by the theoretical calculations. It is important to recognize,
as can be inferred from the multitude of features in the 650 cmꢀ1
and the 250 cmꢀ1 regions, that this solution contains almost all
possible organoꢀhalo aluminum species. Furthermore, the spec-
trum contains no features attributed to raw materials, proving that
the trans-metalation reaction goes to completion. NMR spectros-
copy supports all of these conclusions, as is further reported.
Table 3 summarizes the dominant experimental Raman
vibrational bands for the most important magnesium and alumi-
num core species.
From Table 3 and Figure 5 it is deduced that the equilibrium
species present in the standard APC solution are Ph4Alꢀ,
Ph2AlCl2ꢀ, PhAlCl3ꢀ, AlCl4ꢀ and Mg2Cl3þ. This list takes into
account only anionic aluminum species, as inferred from the high
ionic conductivity, of 1.8 ꢁ 10ꢀ3 Ωꢀ1 cmꢀ1, and the 13C NMR
analyses. On the basis of these data, the following reaction
formula (APC electrolyte formation scheme and solution equi-
librium species) is suggested for the formation of the standard
APC solution:
Figure 5. Raman spectrum of a standard 0.25 M APC solution (0.25 M
AlCl3 þ 0.5 M PhMgCl in THF).
1152 cmꢀ1, the common features for phenyl containing aluminum
species. The Ph3Al THF spectrum contains additional peaks, at 209
3
and 658 cmꢀ1, predicted also from the calculated spectra (see
Table 2, data for Ph3AlClꢀ). Despite the proximity of MgCl2 and
Ph3Al peaks, the two features are exclusive, and can be used to
identify the existence of each of them with certainty. This is
exemplified in the inset in Figure 4. The spectrum of this solution
shows a prominent peak at 680 cmꢀ1 as the strongest feature in the
region of 650 cmꢀ1. It also possesses a peak at 297 cmꢀ1, as the
strongest band in this spectral region. The existence of the two
ꢀ
9PhMgCl þ 4:5AlCl3 f Ph4Alꢀ þ 2Ph2AlCl2
þ PhAlCl3ꢀ þ 0:5AlCl4ꢀ þ 4:5Mg2Cl3
þ
As noted at the beginning of this paper, the explicit electro-
chemical characteristics of the complex solutions is dominated by
the various equilibrium species existing in the solution. This, in
turn, is very sensitive to a variety of factors, such as the total
concentration, solvent, components ratio (R:Mg:Al:Cl), and
even the preparation procedure. In Figures S7 and S8 (SI) the
influence of the synthesis temperature and the total complex
concentration is analyzed, as well as LiCl addition.
The results presented in Figures S7 and S8 (SI) prove that LiCl
acts as a Clꢀ donor in the APC system. Nonetheless, one striking
difference between the DCC and the APC electrolytes solutions is
reflected in these results. In contrast to the case of DCC solutions,
APC solutions form rather puzzling solution equilibria. As studied
species, Ph2AlCl2 and Ph3AlClꢀ, at approximately 1:5 ratio can
ꢀ
clearly be identified from the Raman spectrum, with solid support
from the calculated data and the NMR spectrum (see Figure S6, SI).
The spectrum also shows the expected MgCl2 and Mg2Cl3þ peaks
at 212 and 241 cmꢀ1, respectively.
After identifying the main species that might exist in the complex
electrolytes and establishing a reference spectra library, we elabo-
rated on the identification of the equilibrium species in the
“standard” APC solution (i.e., AlCl3 and PhMgCl at 1:2 ratio).
Figure 5 displays the Raman spectrum of a standard 0.25 M
APC solution, at room temperature.
6274
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
Table 4. Summary of the HOMO and LUMO Energy Levels
and Their Energy Difference in eV for the HaloꢀOrgano
Aluminum and Magnesium Species in THF
energy level (eV)
molecule
Ph4Alꢀ
Ph3AlClꢀ
HOMO
LUMO
Δ
ꢀ5.33
ꢀ5.76
ꢀ5.98
ꢀ6.20
ꢀ6.55
ꢀ6.71
ꢀ6.99
ꢀ7.46
ꢀ7.51
ꢀ7.53
0.64
0.38
5.97
6.15
6.32
6.43
5.93
6.00
6.22
6.67
6.82
8.54
ꢀ
Ph2AlCl2
0.34
ꢀ
PhAlCl3
0.23
Ph3Al THF
ꢀ0.62
ꢀ0.71
ꢀ0.77
ꢀ0.79
ꢀ0.69
1.01
3
Ph2AlCl THF
3
PhAlCl2 THF
3
MgClþ THF5
þ3
Mg2Cl3 THF6
3
ꢀ
AlCl4
Figure 6. Steady-state cyclic voltammograms of four APC electrolyte
solutions, measured with Pt wire working electrodes, Mg strips as
counter, and reference electrodes at 25 mV/s.
from the spectrum in Figure S8 (SI), the equilibrium species exhibit
coexistence of a wide variety of aluminum compounds, among which
are AlCl4ꢀ and Ph4Alꢀ. This is a very unusual situation, as these two
compounds (as well as all the other aluminum core species in this
solution) are expected to exhibit two extremes of Lewis acidity. This
is in striking contrast to the DCC electrolytes family, which usually
contain either a single aluminum-based species in the solution or the
smallest possible variety of species, forming a uniform mixture of Clꢀ
(or THF) and organic ligands around the aluminum core. The origin
of this phenomenon is not yet understood.
The comprehensive spectral analysis presented above led to
the identification of the various species coexisting in equilibrium
in APC-type solutions. One of the important characteristics of
the APC electrolytic solution, in contrast to DCC, is its apparent
wide electrochemical stability window. The anodic stability limit
for any electrochemical system is dominated by its most oxida-
tion susceptible component. For the APC family of solutions, it is
expected that the most oxidation vulnerable component would
be a molecule or an ion containing the weakest metalꢀcarbon
bond. Since none of these solutions include organoꢀmagnesium
compounds, the weakest link is predicted to contain an alumi-
numꢀcarbon bond. It was shown in studies with the DCC family
of electrolytes that, as a rule of thumb, the greater the Cl:Al ratio,
the higher its electrochemical oxidation potential. In other words,
the more inorganic character the aluminum species possesses,
the greater its oxidation stability.
The relative stability toward oxidation may be evaluated,
experimentally, using electrochemistry. In the simplest case,
cyclic voltammetry (CV) with inert working electrode can be
utilized for this purpose. Keeping all the experimental setup and
conditions constant, the anodic stability window can be deter-
mined as the deflection point in the positive scan where the
current increases appreciably. It is also possible to estimate the
relative sensitivity toward oxidation from the calculated molec-
ular orbital HOMO energies, as the oxidation reaction involves
removal of an electron from the compound’s HOMO level.20
Table 4 summarizes the calculated HOMO and LUMO
energy levels for a variety of haloꢀorganic aluminum and
magnesium species, in THF, and the pure solvent. The values
listed were calculated for the structure-optimized species, as
explained in the Experimental Section, including the THF
solvent influence. Considering the organometallic species only,
a clear trend is obtained from this table. As predicted, the greater
the Cl:Al ratio, the lower its HOMO energy level, and therefore
the greater its expected stability toward oxidation.
In relation to the relevant organoꢀhalo species only, these results
predict Ph4Alꢀ to be the most oxidation vulnerable species, while
ꢀ
AlCl4 to show the greatest stability. Thus, electrochemically, we
can conclude that any noticeable concentration of Ph4Alꢀ will have
detrimental influence on the solution anodic stability limit. The
same argument applies to all other APC-type solutions, in such a
way that the presence of the most oxidation-prone species will
govern the anodic stability window for the whole system. As such,
the weakest link in the chain is the species containing the highest Ph:
Cl ratio. The picture described in this section ignores the anodic
stability of pure THF, and THF in the presence of the solution
species. Based on the literature-reported oxidation stability window
of ∼4.5 V vs Li for THF in a variety of systems21 and the rule of
thumb that puts the magnesium depositionꢀdissolution at roughly
1 V above lithium, the anodic stability limit for THF in the
magnesium scale is estimated to lie around 3.5 V, far above the
anodic stability windows measured for most APC-type solutions.
Support for this hypothesis was found in a preparative electrolysis
experiment that we have carried out with an APC solution, in which
the main product was found to be benzene, presumably due to the
electrochemical breakdown of phenylꢀaluminum bonds. No THF
oxidation products were detected (see Figure S9, SI).
In order to investigate whether the theoretical calculations
predict the experimental results adequately, we performed cyclic
voltammetry with a variety of APC electrolyte solutions.
Figure 6 displays the anodic side of CVs for several APC
solutions with Pt working electrodes. The inset in this figure
shows the full CVs (including the reversible Mg deposition
domain at the left-hand side of the voltammogram) with the
potential domain of interest marked.
The four APC solutions’ CVs presented in Figure 6 contain,
according to the spectral analysis described above, the following
oxidizable species: APC 1:1 (cs) contains PhAlCl3ꢀ and Ph2AlCl2ꢀ
;
0.4 M APC (cs) contains all the various aluminum core species,
including Ph4Alꢀ, and so does the 0.25 M APC solution (see Figure
S7, SI); 0.4 M APC (cs) solution with the addition of 0.8 M LiCl
contains chiefly Ph4Alꢀ and AlCl4ꢀ. Thus, considering the argu-
ments presented above, we expected that all other CV’s will show
6275
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
lower anodic stability window than that for APC 1:1 (cs), which
contains no Ph4Alꢀ. Unexpectedly, Figure 6 reveals, among other
things, that this is not the case. The CVs show that regardless of the
identity of the solution species, discernible oxidation currents
commence at around 2.7ꢀ2.9 V. For instance, APC 1:1 (cs)
solution, which was expected to show the highest anodic stability,
starts to oxidize at as low as ∼2.7 V, while 0.4 M APC (cs) containing
0.8 M LiCl, which was predicted to possess the lowest anodic
stability, does not show appreciable oxidation currents at up to 3.1 V.
It is important to note that this CV does exhibit a deflection point in
the positive currents at ∼2.9 V. However, in contrast to the other
cases, after the slight increase in the currents, it stabilizes and even
shows decrease in the oxidation currents. In the usual cases, following
the deflection point, the currents continue to increase steadily,
rapidly reaching substantial values (as the potentials are higher).
Whatever the reason for the observed exceptional anodic stability
of the APC electrolytes is, recent important measurements revealed
unequivocally that these electrolyte solutions are indeed very
promising candidates for practical high energy density, high voltage,
magnesium batteries. In these experiments we reversibly interca-
lated/deintercalated lithium ions into two cathode materials, namely
V2O5 and LiFe0.2Mn0.8PO4 in APC solutions containing 0.5 M LiCl
within the potential window of ꢀ0.8 to 3 V vs Mg (a detailed
description of these studies is beyond the scope of this paper). The
reversible insertion of Li ions into these systems testifies that the
APC solution is anodically stable at least up to 2.8 V vs Mg.
In almost all CVs carried out with APC-type solutions with Pt as
WE, as seen very clearly in Figure 6 for 0.25 APC solution, the
anodic side of the CV is characterized by random, positive current
spikes. We hypothesize that this phenomenon relates to kind of
quasi-passivation of the Pt electrode by solution species. This
behavior is unique to APC type solutions, and was not observed
for the DCC-type solutions. Hence, we theorize that the aromatic
phenyl groups are responsible for this unique behavior. We suspect
that specific adsorption of the aromatic AlꢀPh-containing mol-
ecules onto the platinum surface plays an important role in this
phenomenon, leading to the unexpected anodic stability of these
solutions. The spiky CVs observed for many APC solutions in the
anodic domain are hypothesized to be a manifestation of a break-
and-repair-like mechanism. Such mechanism may explain the un-
predictably high anodic stability for 0.4 M APC (cs) containing 0.8
M LiCl that contains the largest Ph4Alꢀ concentration. This also
may explain the insensitivity of the anodic stability window mea-
sured for many APC-like solutions with platinum. Support for this
assumption can be obtained from a closer examination of the CVs in
Figure 6, particularly in the 2.8ꢀ2.9 V region. As can be seen, even
the CV for 0.4 M APC (cs) solution containing 0.8 M LiCl, which
does not attain substantial oxidation currents at up to 3.0 V, clearly
shows a positive deflection point in the current. However, soon after
this rise in current, the current density evens out, and even decreases
a little, suggesting that the electrochemical activity is inhibited. The
exploration of this phenomenon deserves comprehensive electro-
chemical and surface analysis studies.
Figure 7. Visualization of the HOMO NBO for Ph3AlClꢀ in THF.
from the aromatic system of the phenyl ligands, probably yielding
a phenyl radical. According to the electrolysis results reported
above, indeed the AlꢀC bond between the phenyl ligand and the
aluminum core breaks down. At this stage we have no further
information as to the entire set of final reaction products. We
have no data regarding the Al-based compounds, nor do we know
from which species the phenyl radical captures hydrogen to yield
the main product, benzene, although it is very probable that it
takes it from THF.
The NBO analysis for the entire range of the aluminum-based
species in this study reveals an interesting phenomenon. For all the
chlorophenyl aluminum species, the DFT calculations yield
aromatic phenyl rings in the expected alternating order of σ- and
π-bonds. However, for PhAlCl3ꢀ, which contains the largest number
of electron-withdrawing chlorine ligands, the NBO calculated
molecule shows clear breakdown in the aromaticity of the phenyl
ring. As seen in Figure 8, the NBO’s HOMO level is suggestive of a
quinone-like structure, with two double and four single bonds.
This phenomenon, which is also obtained with higher com-
putation levels, e.g. B3LYP/6-31þG(d,p), MP2/6-31(d,p), can
be studied using the NBO occupancy analysis (see Table. S10). It
can be attributed to the strong electron-withdrawing capacity of
the triply chlorinated aluminum core, imparting excess positive
charge on the organic ligand. The same phenomenon is found
also for PhAlCl2 THF. Additional support for this phenomenon
3
is presented in Figure S11, SI.
Similar quinoid structures were reported for substituted benzene
rings, and for quinone-methide.22ꢀ24 Analysis of the NBO models
shows that despite the quinoid character of the ring in this case, all the
CꢀC and CꢀH bond lengths remain practically identical, as in the
aromatic systems.25,26 However, analysis of the CꢀCꢀCbondangles
within the rings of PhAlCl3ꢀ, as presented in Table 5, does show that
unusual distortion occurs in the ipso position, relative to the other
anions. Only slight differences are calculated for the CꢀCꢀC angles
in the other positions, namely ortho, meta, and para.
Unfortunately, experimental validation of this phenomenon
could not be obtained from the Raman and NMR measurements.
As reported above, the changes in the spectral data versus the Cl:
Ph ratio show only gradual changes.
In order to shed some light on the mechanism through which
the organoꢀhalo aluminum species undergo electrooxidation,
we visualized the natural bond orbitals (NBO) for selected
phenylaluminum chloride molecules and anions. This is exem-
plified below for Ph3AlClꢀ. The picture obtained indicates that
the highest lying HOMO orbital is located at the aromatic π-
systems of the phenyl ligands, as seen in Figure 7.
These observations lead us to suggest that the first stage in the
electrooxidation process commences by electron abstraction
6276
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
Figure 8. Visualization of the HOMO NBO for PhAlCl3ꢀ in THF.
decade ago. A further followup challenge is research and devel-
opment of Mg insertion cathodes with redox potentials as high as
2.5 V in the Mg scale.
Table 5. Summary of CꢀCꢀC Bond Angles (deg) for the
Phenyl Rings in PhyAlCl4ꢀyꢀ Ions (i = ipso, o = ortho,
m = meta, p = para carbon)
anion
θi
θo
θm
θp
’ ASSOCIATED CONTENT
Ph4Alꢀ
Ph3AlClꢀ
115.37
115.77
116.23
116.53
122.78
122.51
122.22
122.03
119.95
119.96
119.96
120
119.145
119.27
119.39
119.4
S
Supporting Information. Supplementary figures and
b
ꢀ
complete ref 14. This material is available free of charge via the
Internet at http://pubs.acs.org.
Ph2AlCl2
ꢀ
PhAlCl3
’ AUTHOR INFORMATION
Additional interesting information is revealed from the DFT-
calculated structures pertaining to the ligandꢀaluminum bonds.
In all cases the RꢀAl bond is established through a lone-pair on
the ipso carbon, indicating polarꢀcovalent/coordination bonds
(see Figure S12, SI). The same type of bonding is also found for
the ClꢀAluminum bonds. Interestingly, the bond lengths for all
the ligands show a clear trend: the higher the Cl:Ph ratio, the
shorter the Alꢀligand bond length. This constitutes an addi-
tional point of view for the expected trend in bonds strengths, as
found from the HOMO energy levels.
Corresponding Author
aurbach@mail.biu.ac.il
’ REFERENCES
(1) Aurbach, D.; Lu, Z.; Schechter, A.; Gofer, Y.; Gizbar, H.;
Turgeman, R.; Cohen, Y.; Moshkovich, M.; Levi, E. Nature 2000,
407, 724.
(2) Aurbach, D.; Gizbar, H.; Schechter, A.; Chusid, O.; Gottlieb,
H. E.; Gofer, Y.; Goldberg, I. J. Electrochem. Soc. 2002, 149, A115.
(3) Gizbar, H.; Vestfrid, Y.; Chusid, O.; Gofer, Y.; Gottlieb, H. E.;
Marks, V.; Aurbach, D. Organometallics 2004, 23, 3826.
(4) Gofer, Y.; Chusid, O.; Gizbar, H.; Vestfrid, Y.; Gottlieb, H. E.;
Marks, V.; Aurbach, D. Electrochem, Solid-State Lett. 2006, 9, A257.
(5) Mizrahi, O.; Amir, N.; Pollak, E.; Chusid, O.; Marks, V.;
Gottlieb, H.; Larush, L.; Zinigrad, E. J. Electrochem. Soc. 2008, 155, A103.
(6) Vestfries, Y.; Chusid, O.; Goffer, Y.; Aped, P.; Aurbach, D.
Organometallics 2007, 26, 3130.
(7) Chen, C. R.; Gau, H. M. Acta Crystallogr. 2008, E64, m1381.
(8) Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652.
(9) Lee, C.; Yang, W.; Parr, R. G. Phys. Rev. B 1988, 37, 785–789.
(10) Hehre, W. J.; Radom, L.; Schleyer, P. v. R.; Pople, J. A. Ab Initio
Molecular Orbital Theory; John Wiley & Sons: New York, 1986.
(11) (a) Cance, M. T.; Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997,
107, 3032. (b) Mennucci, B.; Tomasi, J. J. Chem. Phys. 1997, 106, 5151.
(c) Mennucci, E.; Cances, M. T.; Tomasi, J. J. Phys .Chem. B 1997,
101, 10506. (d) Tomasi, J.; Mennucci, B.; Cances, E. J. Mol. Struct.
(THEOCHEM) 1999, 464, 211.
’ CONCLUSIONS
We have identified the major equilibrium species constituting
the electrolyte solutions formed by reacting PhMgCl and AlCl3
in THF. Raman spectroscopy, together with DFT calculations,
NMR, and single-crystal XRD provided most of the evidence
necessary for the elucidation of this complicated system. Some
ambiguous and less clear issues remain, such as the existence of
the elusive MgClþ ion, which has only been identified in
SCXRD, and the close spectral similarity between the anionic
and neutral aluminum species. The various solution species in the
APC family depend to a large extent, apart from the raw materials
proportions, on the synthesis temperature. The oxidation stabi-
lity trend among the computed aluminum-based compounds
does not follow the experimental stability window. We assume
that the anodic electrochemical behavior of these species is
governed by an inhibiting mechanism, such as specific adsorp-
tion, that leads to quasi-passivation.
(12) Merrick, J. P.; Moran, D.; Radom, L. J. Phys. Chem. A 2007,
111, 11683.
(13) Foster, J. P.; Weinhold, F. J. Am. Chem. Soc. 1980,
102, 7211–7218.
(14) Frisch, M. J.; Trucks, G. W.; et al. Gaussian 09, revision A.02;
Gaussian, Inc.: Wallingford, CT, 2009.
(15) Nakayama., Y; Kudo, Y.; Oki, H.; Yamamoto, K.; Kitajima, Y.;
Noda, K. J. Electrochem. Soc. 2008, 155 (10), A754.
On the basis of this work, it appears that adequate selection of
the raw materials, their stoichiometric ratios, and the synthesis
conditions can yield solutions for rechargeable magnesium
batteries with operating voltages in excess of 2.5 V (vs Mg).
This presents an extension of more than 1 V compared to the first
generation of the rechargeable Mg battery system presented a
6277
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278

Journal of the American Chemical Society
ARTICLE
(16) (a) Ashby, E. C.; Parris, G. E. J. Am. Chem. Soc. 1971, 93 (5),
1206. (b) Walker, F. W.; Ashby, E. C. J. Am. Chem. Soc. 1969, 91 (4),
3845.
(17) Shamir, J.; Sobota, P. J. Raman Spectrosc. 1991, 22, 535.
(18) Jon, D. M.; Harker, Y. D.; David, F. E. J. Chem. Phys. 1969,
50, 5420.
(19) Wotiz, J. H.; Hollongsworth, C. A.; Dessy, R. E. J. Am. Chem.
Soc. 1956, 78, 1221.
(20) Murakami, M.; Ue, A.; Nakamura, S. J. Electrochem. Soc. 2002,
149, A1572.
(21) Nonaqueous Electrochemistry, Dekker, Inc.: New York, 1999.
(22) Lee, J. Y.; Mhin, B. J.; Kim, K. S. J. Phys. Chem. A 2003, 107 (19),
3578.
(23) Bader., R. F. W.; Chang, C. J. Phys. Chem. 1989, 93, 2946.
(24) Vigalok, A.; Milstein, D. Acc. Chem. Res. 2001, 34, 798.
(25) Shaik, S.; Shurki, A.; Danovich, D.; Hiberty, P. C. Chem. Rev.
2001, 101, 1501.
(26) Itzhaki, L.; Rozental, E.; Altus, E.; Basch, H.; Hoz, S. J. Phys.
Chem. A 2008, 112 (50), 12812.
6278
dx.doi.org/10.1021/ja1098512 |J. Am. Chem. Soc. 2011, 133, 6270–6278