Predicting the electrochemical behavior of lithium nitrite in acetonitrile with quantum chemical methods

Article abstract of DOI:10.1021/ja410766n

Electrolyte stability is an essential prerequisite for the successful development of a rechargeable organic electrolyte Li-O₂ battery. Lithium nitrate (LiNO₃) salt was employed in our previous work because it was capable of stabilizing a solid-electrolyte interphase on the Li anode. The byproduct of this process is lithium nitrite (LiNO₂), the fate of which in a Li-O₂ battery is unknown. In this work, we employ density functional theory and coupled-cluster calculations combined with an implicit solvation model for neutral molecules and a mixed cluster/continuum model for single ions to understand the chemical and electrochemical behavior of LiNO₂ in acetonitrile (AN). The redox potentials of oxygenated nitrogen compounds predicted in this study are in excellent agreement with the experimental results (the average accuracy is 0.10 V). Theoretical calculations suggest that the reaction between the nitrite ion and its first oxidation product, nitrogen dioxide (NO₂), in AN solution proceeds via the initial formation of a trans-ONO-NO₂ dimer that is subject to autoionization and the subsequent reaction of produced nitrosyl ion (NO ⁺) with NO₂^-. Good agreement between experimental and simulated cyclic voltammograms for electrochemical oxidation of LiNO₂ in AN provides support to the proposed mechanism of coupled electrochemical and chemical reactions. The results suggest a possible mechanism of regeneration of LiNO₃ in electrolyte in the presence of oxygen, which is uniquely possible under charging conditions in a Li-O₂ battery.

Full text of DOI:10.1021/ja410766n

Article
pubs.acs.org/JACS
Predicting the Electrochemical Behavior of Lithium Nitrite in
Acetonitrile with Quantum Chemical Methods
Vyacheslav S. Bryantsev,* Jasim Uddin, Vincent Giordani, Wesley Walker, Gregory V. Chase,
and Dan Addison
Liox Power, Inc., 129 N. Hill Ave., Suite 103, Pasadena, California 91106, United States
S
* Supporting Information
ABSTRACT: Electrolyte stability is an essential prerequisite
for the successful development of a rechargeable organic
electrolyte Li-O2 battery. Lithium nitrate (LiNO3) salt was
employed in our previous work because it was capable of
stabilizing a solid-electrolyte interphase on the Li anode. The
byproduct of this process is lithium nitrite (LiNO2), the fate of
which in a Li-O2 battery is unknown. In this work, we employ
density functional theory and coupled-cluster calculations
combined with an implicit solvation model for neutral
molecules and a mixed cluster/continuum model for single
ions to understand the chemical and electrochemical behavior of LiNO2 in acetonitrile (AN). The redox potentials of oxygenated
nitrogen compounds predicted in this study are in excellent agreement with the experimental results (the average accuracy is 0.10
V). Theoretical calculations suggest that the reaction between the nitrite ion and its first oxidation product, nitrogen dioxide
(NO2), in AN solution proceeds via the initial formation of a trans-ONO-NO2 dimer that is subject to autoionization and the
subsequent reaction of produced nitrosyl ion (NO+) with NO2 . Good agreement between experimental and simulated cyclic
−
voltammograms for electrochemical oxidation of LiNO2 in AN provides support to the proposed mechanism of coupled
electrochemical and chemical reactions. The results suggest a possible mechanism of regeneration of LiNO3 in electrolyte in the
presence of oxygen, which is uniquely possible under charging conditions in a Li-O2 battery.
2Li + LiNO3 → Li2O + LiNO2
1. INTRODUCTION
(1)
As hybrid and full electric vehicles continue to increase in sales
and popularity around the world, the race is on to develop
batteries that can deliver higher capacities at lower costs. Li-O2
batteries have been heavily investigated as a potential battery
system to replace Li-ion because of a high theoretical specific
energy, however research into Li-O2 cells has been hampered
by degradation issues with every major component of the
system upon cycling, and progress has been slow.1−5
Another product of this reaction, lithium nitrite (LiNO2), is
expected to be soluble and electroactive in the charging
potential range of a Li-O2 cell. Indeed, small oxidative processes
at 3.6−3.7 V observed in a Li-O2 cell with a LiNO3/
dimethylacetamide (DMA) electrolyte6 are consistent with
the presence of LiNO2 in an electrolyte solution. LiNO3 is used
as an electrolyte additive in rechargeable lithium-sulfur (Li-S)
batteries to promote the formation of a stable passivation film
on Li anode, which is known to significantly suppress the
reduction of polysulfide species in solution.9,10 However, the
progressive consumption of LiNO3 on the Li anode in Li-S
batteries limits the number of cycles over which LiNO3 is able
to prevent the redox shuttle of lithium polysulfides. We have
recently shown12 that LiNO3 can be regenerated from LiNO2
in the presence of dissolved O2 during the charging of a Li-O2
battery, which could be a contributing factor to the observed
interfacial stability and cycling of Li metal when both LiNO3
and O2 are present.11 Detailed knowledge of the chemical and
electrochemical behavior of LiNO2 in aprotic solvents is thus
essential to gain further insights into the mechanism of
Liquid electrolyte plays a critical role in determining the
nature of discharge products and cycling characteristics of a
rechargeable Li-O2 battery. One of the biggest challenges is to
develop an electrolyte composition that is sufficiently stable to
both the Li anode and O2 cathode environments upon long-
term cycling.1−5 We have recently reported6 on the Li-O2 cell
that enables longer duration cycling (>2000 h) with
significantly reduced decomposition of electrolyte materials
compared to Li-O2 cells previously reported in the field. This
performance is achieved by combining straight-chain alkyl
amides, which are significantly more stable to the reactions of
the O2 electrode than conventional electrolyte solvents,6,7 with
the lithium nitrate (LiNO3) salt, which is capable of stabilizing a
solid-electrolyte interphase (SEI) on the Li anode.8−11
−
regeneration of NO3 in rechargeable Li-O2 cells.
It was suggested8 that a stabilizing effect of LiNO3 on Li
metal cycling is due to the formation of a passivating layer of
Li2O on the Li electrode surface through the following reaction:
Received: October 21, 2013
© XXXX American Chemical Society
A
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 1. Comparison of the Calculated and Experimental Gibbs Free Energies of Gas-Phase Reactions involving Nitrogen−
a
Oxygen Compounds (kcal/mol)
B3LYP
M06
M06-L
M06-2X
c
reaction
2NO2 = N2O4
BS1
BS2
BS1
BS2
BS1
BS2
BS1
BS2
RCCSD(T)/CBS
expt
1.7
2.9
1.7
2.8
−5.8
−3.5
−13.0
−10.0
160.5
177.1
171.1
−28.1
−0.04
6.6
−5.8
−3.2
−12.2
−9.3
160.8
174.1
170.4
−28.6
−0.54
6.5
−5.4
−6.1
−8.2
−5.5
172.7
182.8
181.5
−6.1
−6.1
−7.9
−4.9
170.8
178.1
178.8
−33.9
2.3
−0.7
3.4
−1.4
2.5
−0.5
1.3
−1.1
0.4
NO2 + NO = N2O3
−
−
NO2 + NO2 = NO3 + NO
−12.1
−12.9
159.5
177.5
170.3
−18.6
11.0
−11.4
−12.3
159.9
175.9
170.1
−19.9
9.8
−19.3
−18.9
147.7
177.2
162.9
−8.9
4.4
−18.5
−17.8
148.2
175.3
162.8
−12.3
2.6
−13.9
−14.0
147.6
170.4
160.4
−16.7
3.0
−14.1
−13.5
149.4
−
−
NO2 + 1/2N2O4 = NO3 + NO
N2O4 = NO+ + NO3
−
+
−
N2O4 = NO2 + NO2
N2O3 = NO+ + NO2
161.9
−16.9
3.70
−
2NO + O2 = 2NO2
−33.2
3.6
2NO2 + 1/2O2 = N2O5
MUE, RCCSD(T)/CBS
b
5.1
5.1
11.4
10.5
10.6
3.5
2.6
0.8
b
MUE, expt
4.4
4.6
6.1
6.3
10.4
3.2
2.4
a
BS1 is the 6-311+G* basis set. BS2 is the aug-cc-pVTZ basis set. Geometries were optimized at each basis set and density functional. Zero point
b
energies and thermal corrections were calculated at the B3LYP/6-311+G* level. MUE versus benchmark calculations at the RCCSD(T)/CBS level
and experiment. Ref 46.
c
The electrochemical oxidation of the oxygenated nitrogen
compounds has been studied previously in various media, but
the reported results are often inconsistent.12−22 For example,
mechanisms of electrochemical processes in organic media
and guiding the assignment of the cyclic voltammetry peaks.
+
2. COMPUTATIONAL METHODS
the redox potential of the NO2/NO2 couple was reported to
vary by more than 0.6 V in the same solvent.18,21 This
ambiguity was assigned to the problem of contamination by
traces of water that can react with nitrogen oxides to generate
new electroactive species.21 Due to lack of detailed mechanistic
understanding of nitrite oxidation, there are significant
contradictions and variations in the interpretation of the
successive oxidation waves of the nitrite anion.12,13,16,19,20,22
In this work, we employ theoretical calculations to predict
redox potentials of oxygenated nitrogen compounds in
acetonitrile (AN) solution. AN is chosen as a solvent in this
study because similarly to straight-chain alkyl amides, it shows
improved stability toward reduced O2 species7 but offers a
wider electrochemical window, thus extending the range of
accessible redox potentials. Ion solvation is modeled by explicit
inclusion of eight solvent molecules in the vicinity of the ion
and implicit treatment of the rest of the solvent with a dielectric
continuum model. This mixed cluster/continuum frame-
work23,24 avoids the ambiguities of assigning atomic O and N
radii for ionic compounds, because the results are not sensitive
to the choice of these parameters if a solute is completely
surrounded by solvent molecules. Excellent agreement with the
experimental redox potentials provides strong support for the
computational model, thereby lending credence to the
calculated solvation free energies of various ionic nitrogen−
oxygen species. Furthermore, we have investigated the
mechanism of the chemical reaction between the nitrite anion
and its first oxidation product, nitrogen dioxide (NO2) and its
dimer. We find that the interconversion between nitrogen
dioxide and its various dimers provides a low-energy path for
the reaction with the nitrite anion. Good agreement between
experimental and simulated cyclic voltammograms for electro-
chemical oxidation of LiNO2 in AN provides support to the
proposed mechanism of coupled electrochemical and chemical
reactions. The results have important practical consequences
for rechargeable Li-O2 batteries, suggesting a possible
mechanism of regeneration of LiNO3 in the electrolyte. In a
more general context, first principles calculations can now
provide valuable insight into solution-phase chemistry and open
new opportunities for the use of theory in elucidating
2.1. Gas-Phase Calculations. We used the restricted open-
shell second-order Moller−Plesset perturbation theory
̈
(MP2)25,26 in the complete basis set (CBS) limit and
coupled-cluster theory with singles, doubles, and perturbative
triples excitations for restricted reference wave functions
(RCCSD(T))26−29 in the aug-cc-pVnZ (n = D and T) basis
set30,31 to determine the benchmark reaction energies and
barriers involving oxygenated nitrogen compounds. Only the
valence electrons were correlated in the RMP2 and RCCSD(T)
calculations. Assuming that the difference between RCCSD(T)
and RMP2 energies exhibits only a small basis set dependence,
the RCCSD(T) energies at the CBS limit can be estimated as32
ΔE(RCCSD(T)/CBS)
= ΔE(RMP2/CBS) + δRCCSD(T)
(2)
δRCCSD(T) = ΔE(RCCSD(T)/aug‐cc‐pVnZ)
− ΔE(RMP2/aug‐cc‐pVnZ)
(3)
Test calculations for the NO2 dimer show that the
δRCCSD(T) corrections vary by 0.67 and 0.20 kcal/mol
upon extending the basis set aug-cc-PVnZ from n = D to T and
T to Q, respectively. Thus, the combination of RMP2/CBS
with RCCSD(T)/aug-cc-pVTZ corrections provides an ex-
cellent compromise between accuracy and computational cost.
The geometries of nitrogen−oxygen compounds in all post-
Hartree−Fock calculations were obtained at the RCCSD(T)/
aug-cc-pVDZ level. Minimization and transition-state optimi-
zation was performed using the Driver module of NWChem26
with default convergence criteria. The binding energy for the
NO2 dimer differs by only 0.05 kcal/mol when compared to
results obtained after full optimization at the RCCSD(T)/aug-
cc-pVTZ level. However, if the geometry is obtained at the
RMP2/aug-cc-pVDZ level, the binding energy for the NO2
dimer from single-point calculations at the RCCSD(T) level is
underestimated by ∼1 kcal/mol. For the basis set expansion in
our RMP2 calculations, we used a family of Dunning’s30,31
augmented correlation-consistent basis sets up to quadruple-
B
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
and quintuple-ζ quality with spherical harmonic functions. All
RMP2 and RCCSD(T) calculations were carried out using the
NWChem 5.1 program package.26
results gives reasonable values of Gibbs free energies for 2NO2
= N2O4 (−4.9 kcal/mol) and NO2 + NO = N2O3 (−3.7 kcal/
mol) in AN (the deviation from experiment20,40 is 1.3 and 1.9
kcal/mol, respectively).
It was shown by Liu and Goddard33 that the choice of the
reference wave function in coupled-cluster calculations has a
significant effect on the potential energy surface for the
dimerization of NO2. In agreement with this work33 we found
that the restricted wave function method, RCCSD(T), recovers
more correlation energy near the equilibrium bond distance
than does the unrestricted wave function method and gives a
more reliable estimate for the binding energy of the NO2 dimer.
We examined the ability of B3LYP34,35 and a family of M06
density functionals (M06, M06-L, and M06-2X)36,37 in the
unrestricted spin formalism to reproduce the benchmark
RCCSD(T) binding and reaction energies involving oxy-
genated nitrogen compounds. The proper open-shell initial
guess was necessary for several NO2 dimer geometries (open-
shell singlet state).33 Using a closed-shell initial guess can bias
the solution toward the closed-shell state with higher energy.
For each of these functionals, two basis sets were considered: 6-
311+G* (one set of polarization d-funtions; for systems
containing H atom the basis set was 6-311++G**) and aug-
cc-pVTZ. The ultrafine numerical integration grid was
employed in all DFT calculations. The B3LYP functional was
chosen to calculate the free energy barriers for the reaction of
2.3. Configurational Sampling of Solute−Solvent
Clusters. A search of the low-energy solute−solvent clusters
was performed using the Monte Carlo sampling algorithm
(called BLENDS)41,42 to generate 40 initial cluster config-
urations. Some initial configurations were also built upon visual
inspection and manual rearrangement of the optimized clusters.
The initial structures were subjected to full geometry
optimization at the B3LYP/6-31G** level, followed by single
point energy calculations at either M06-L/6-311++G** or
M06-2X/6-311++G** level (see Section 4.3). These were
combined with the implicit solvation free energies computed at
the B3LYP/6-311++G** level to yield the final ranking of the
clusters. Pure AN clusters were taken from our previous
study.24
3. EXPERIMENTAL SECTION
3.1. General. Anhydrous AN (99.8%) was purchased from Sigma-
Aldrich and used as received. Lithium nitrite (LiNO2) was synthesized
in our lab according to the procedure described previously.12 LiNO3
(Sigma Aldrich, ≥99.5%) and LiTFSI (Novolyte) were dried at 200 °C
for 3 days under vacuum prior to use. Nitronium tetrafluoroborate
(NOBF4, >95%) was purchased from Sigma Aldrich. All electrolytes
had a final water content <10 ppm as determined by Mettler Toledo
Karl Fischer titration. Li rod for reference electrode was purchased
from Alfar Aesar and washed with hexane before used. Ultra-high-
purity argon and O2 gases (Airgas) were used for purging the
electrolyte. NO2 gas (Sigma Aldrich, ≥99.5%) was used as received.
The UV−vis spectrometry measurements were performed using a
Perkin-Elmer Lambda 25 Spectrometer with quartz SUPRASIL cells.
Mass spectrometry measurements were made with a Stanford
Research Systems RGA100 (1−100 amu) residual gas analyzer
attached to high-vacuum system (1 × 10−8 Torr) from Pfeiffer
Vacuum.
3.2. Cyclic Voltammetry. The cyclic voltammetry (CV) experi-
ments were performed with a VMP3 (BioLogic Science Instruments)
potentiostat. The electrochemical cell (Pine Instrument Company) is
constructed in a dry glovebox. The cell consisted of a traditional three
electrode system utilizing a Li-rod reference electrode salt-bridged in
an electrolyte containing 5 M LiTFSI in AN. The lithium reference
electrode potential was converted to the Li+/Li potential scale using
the Nernst equation. A platinum wire was used as the counter
electrode in a fritted compartment, and a platinum disk (0.196 cm2)
was used as working electrode, which was polished with 0.05 mm
alumina paste prior to each experiment. Cyclic voltammetric data
indicate that the 0.5 M LiTFSI in AN has high anodic stability (>5.0 V
versus Li+/Li). The accessible redox potentials are within the range
exhibited by oxygenated nitrogen compounds. Digital simulations were
conducted with DigiElch,43 version 6.F, a software package for the
digital simulation of common electrochemical experiments.
−
NO2 with NO2 and N2O4, based on adequate description of
thermodynamic data for these reactions (Table 1). Additionally,
B3LYP calculations showed reasonable agreement with
RCCSD(T)/CBS potential energy surfaces for interconversion
between NO2 and its dimers.33 A transition-state search with
the B3LYP method was performed using a standard quasi-
Newton method in Jaguar,38 starting from the partially
optimized geometry at the same level of theory along the
chosen reaction coordinate and the precalculated Hessian. For
transition states, one single imaginary frequency at the B3LYP/
6-311+G* level was obtained in all cases, which upon
visualization was found to correspond to the expected motion
along the reaction coordinate. Additionally, intrinsic reaction
coordinate calculations (IRC) were performed at the B3LYP/6-
311+G* level (as implemented in Jaguar)38 to ensure that
transition-state structures connect their respective reactants and
products. Structural features of the transition states connecting
NO2 and its dimers are in agreement with previous work.33 The
standard Gibbs free energy of each species in the gas phase (T
= 298 K, P = 1 atm) was calculated using the rigid rotor-
harmonic oscillator approximation without scaling, except that
vibrational frequencies lower than 50 cm−1 were raised to 50
cm−1. This procedure is similar to that proposed by Truhlar and
co-workers39 as a way to correct for the well-known breakdown
of the harmonic oscillator model for the free energies of low-
frequency vibrational modes.
2.2. Implicit Solvation Calculations. Solvation calcula-
tions were carried out using the Poisson−Boltzmann
continuum model in Jaguar38 for the AN solvent, characterized
by a dielectric constant of 35.97 and the probe radius of 2.180.
The electronic energies in the solvent reaction filed were
obtained at the B3LYP/6-311+G* levels. Using the default
solute van der Waals radii in Jaguar leads to a systematic
overestimation of the solvation free energy for polar solutes in
AN.24 To obtain the optimal performance for polar solvents if
only the electrostatic contribution to the solvation free energy
is included, the default values of solute atomic radii were scaled
by 1.20. This description combined with RCCSD(T)/CBS
3.3. Reaction of NO2/N2O4 with LiNO2. The reaction between
NO2/N2O4 and LiNO2 in AN was carried out in the dry glovebox
under argon. NO2 gas was condensed to N2O4 at −78 °C in a sealed
tube, and then AN was added to make a NO2/N2O4 solution. A
portion of solution was placed in a sealed flask under argon, and 21
mM LiNO2 was slowly added into it. A colorless gas evolved
immediately upon addition of LiNO2. The gas from the reaction of
NO2/N2O4 with LiNO2 was identified as NO by a residual gas
−
analyzer (RGA). The formation of NO3 was confirmed by UV−vis
analysis (further experimental details are given in the Supporting
Information).
C
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
4. RESULTS AND DISCUSSION
Article
effects could be computationally very demanding. The
RCCSD(T)/CBS method offers a good compromise between
accuracy and computational cost, because it incorporates the
dynamic correlation quite well and provides accurate results for
bound molecules and transition-state geometries that are far
from the dissociation limit. Based on extended benchmark
studies of coupled-cluster theory,44,45 we expect the database of
RCCSD(T)/CBS reaction energies to have an accuracy of 1.0
kcal/mol or better and thus serve as a reliable benchmark for
the use of theory in describing the electrochemical processes
and evaluating the reliability of more approximate methods.
4.2. The Accuracy of DFT Calculations in the Gas
Phase. The RCCSD(T)/CBS reaction energies collected in
the previous section enable us to evaluate the reliability of
several density functionals for oxygenated nitrogen compounds.
Table 1 compares the performance of the B3LYP, M06, M06-L,
and M06-2X density functionals with the 6-311+G* and aug-
cc-pVTZ basis sets for predicting the Gibbs free energies of
reactions involving oxygenated nitrogen compounds (ΔGro).
The accuracy of each method is characterized by the mean
unsigned error (MUE) versus RCCSD(T)/CBS and experi-
ment.46
The RCCSD(T)/CBS calculations yield results that fall
within the experimental uncertainty46 of ΔGro. We note that
none of the employed DFT methods provides a completely
satisfactory description of reaction energies of oxygenated
nitrogen compounds. Among them, the M06-2X and B3LYP
functionals significantly outperform the RMP2 method and
provide the best overall performance. M06-2X is the most
accurate functional for describing the autoionization energies of
N2O4 and N2O3, while B3LYP does a good job in reproducing
−
The electrochemistry of NO2 within the charging potential
range of the O2 electrode in Li-O2 batteries was extensively
studied by means of CV in several aprotic organic solvents
under inert atmosphere.16,19,20 In AN16 and sulfolane19,20 the
voltammetric curves show three consecutive oxidation and two
reduction waves. The unambiguous interpretation of the anodic
and cathodic waves, however, remains difficult because of the
nature of the electrochemical mechanism comprising complex
sequences of electrochemical and chemical steps13,16,19−22 and
complications resulting from the presence of water at trace
levels.19,21 In particular, there is no consensus on the origin of
the third anodic wave16,19 and the mechanism of a fast chemical
reaction between nitrite and its first oxidation product, NO2/
16,19,21
N2O4
−
−
NO2 + 1/2N2O /NO2 → NO + NO3
(4)
4
Furthermore, there are serious discrepancies among several
measurements of relative redox potentials of oxygenated
nitrogen compounds in aprotic solvents.18,21 In the following
discussion we will show that with a good choice of
computational methods in the gas phase and in solution, first
principles computations can predict quantitatively correct redox
potentials, provide useful mechanistic insights into reaction
mechanisms, and with the use of digital simulations43 of CVs
can resolve existing controversies in the interpretation of the
−
electrochemical behavior of NO2 . The results suggest a
possible mechanism of regeneration of LiNO3 in electrolyte in
the presence of oxygen, which may explain the positive
synergistic effect of LiNO3 and O2 in enabling long-term
cycling of Li metal.11
−
the thermodynamics for the reaction of NO2 with NO2 and
4.1. Database of Accurate Reaction Energies. The
ability of first-principles computational methods to provide
sufficiently reliable results to the level at which they can
complement and guide the experimental measurements is
contingent upon accurate thermodynamic description of
reaction processes in the gas phase and in solution. As a
benchmark for determining the accuracy of different DFT
functionals in the gas phase, we employed reaction and binding
energies obtained at the CBS limit of the RCCSD(T) theory
(Table S1). The compiled database consists of 9 reaction and 5
complexation energies involving oxygenated nitrogen com-
pounds and 8 reaction/activation energies for interconversion
between NO2 and its dimers.
We find that the RMP2 reaction energies are well converged
with respect to the basis set size at the aug-cc-pV5Z level (see
Table S1). The inclusion of a higher order of electron
correlation at the RCCSD(T) level has resulted in some
dramatic changes of the calculated reaction energies and
barriers compared to the RMP2 results. The magnitude of the
δRCCSD(T) correction is >7 kcal/mol in about one-third of
reactions shown in Table S1 and >10 kcal/mol in the most
difficult cases involving NO2 radical. Reactions involving
interaction of multicenter free radicals (such as NO2) require
great care because of their high sensitivity to both static and
dynamic correlation effects.33 MP2 method is the first step in
going beyond Hartree−Fock that often performs quite well for
closed-shell systems. However, for reactions involving the
transition between closed- and open-shells, neither RMP2 nor
UMP2 is reliable.33 While, in principle, the use of the
multireference wave functions naturally accounts for the static
correlation, satisfactory description of dynamic correlation
N2O4. Based on this performance, selective DFT methods were
employed where RCCSD(T)/CBS calculations were not
feasible.
4.3. Calculations of Solvation Free Energies of NO+,
+
−
−
NO2 , NO2 , and NO3 in AN. The capability of theoretical
models to accurately predict redox potentials without the use of
empirical fitting parameters relies upon their ability to
accurately describe the solvation free energies of ions. Pure
dielectric continuum models are not sufficiently reliable for
predicting single-ion solvation free energies. In this work we
employ a mixed cluster/continuum approach23,24 for calculating
absolute solvation free energies of single ions, where the most
important solute−solvent interactions are included explicitly in
the quantum chemical description, while the solvent effects
beyond the first solvation shell are described using a dielectric
continuum model. From the thermodynamic cycle shown in
Figure 1, the solvation free energy of Xm , ΔG*solv(Xm ), can
be expressed as the algebraic sum of the gas-phase complex-
ation free energy (ΔGog,bind), the difference in the solvation free
energy for [X(solvent)n]m and (solvent)n clusters calculated
Figure 1. Thermodynamic cycle for the calculation of ΔG*solv(X ).
D
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
using a dielectric continuum model, and the standard state/
concentration correction terms23,47 as described below:
energies (ΔG*solv) are shown in Figure 2. The protocol for the
configurational sampling of solute−solvent clusters is described
ΔG*solv (Xm ) = ΔGo
+ ΔG*solv ([X(solvent)n]m
)
g,bind
− ΔG*solv (solvent)n − ΔGo→* − RTln([solvent]/n)
(5)
For the correct construction of the thermodynamic cycle, each
reactant and product in the upper and lower legs of the cycle
must be in the same standard state.23 The conversion from an
ideal gas standard state of 1 atm (24.46 L mol−1) to an ideal
solution standard state of 1 M (1 mol L−1) is defined by47
ΔGo→* = RTln(24.46) = 1.89 kcal/mol (T = 298.15K)
(6)
Similarly, RT ln([solvent]/n) is defined as the free energy
change of 1 mol of (solvent)n ideal gas from [solvent]/n M
liquid state to 1 M,23 where [solvent] is molarity of solvent, and
n is the number of solvent molecules in a cluster.
The two most important considerations with mixed cluster/
continuum models are the size of the explicit solvation shell and
the choice of the DFT method for describing the solute−
solvent interactions. Explicit modeling of the first solvation shell
for neutral and singly charged solutes was found to be sufficient
to give a satisfactory description of their thermodynamic
properties in solution.23,24 Following our previous work on
+
−
Figure 2. Structures of the X(AN)8 clusters (X = NO+, NO2 , NO2 ,
NO3 ) optimized at the B3LYP/6-31G** level.
−
in Section 2.3. The solvation free energies computed using
mixed cluster/continuum and pure dielectric continuum model
are listed in Table 3. The results of implicit solvation
−
solvation of O2 and neutral lithium−oxygen compounds in
AN,24 the inclusion of eight solvent molecules was deemed
sufficient to provide a reasonable saturation of the first
solvation shell around small inorganic ions. As for the choice
of the accurate density functional for noncovalent interaction,
we examined the ability of several DFT methods to reproduce
the CCSD(T)/CBS binding energies of small ions with the AN
dimer. The results, Table 2, indicate that the M06-L density
Table 3. Solvation Free Energies of NO+, NO2 , NO2 , and
NO3 Ions in AN Calculated Using the Mixed Cluster/
Continuum and Pure Dielectric Continuum Solvent Models
(kcal/mol)
+
−
−
dielectric continuum model
a
b
c
ΔG*solv
NO+
cluster/continuum model
default
scaled
Table 2. The Accuracy of Several DFT Methods for
Predicting the Binding Energies between the AN Dimer and
−80.8
−77.5
−61.4
−59.7
−91.5
−88.3
−71.8
−68.4
−77.0
−73.7
−65.4
−61.1
NO+, NO2 , NO2 , and NO3 Ions in the Gas Phase (kcal/
mol)
NO2
+
−
−
+
−
NO2
−
NO3
DFT/6-311++G**/
/B3LYP/6-311++G**
a
Using the thermodynamic cycle shown in Figure 1. The structures of
the ion-(AN)8 clusters employed in these calculations are depicted in
M06- CCSD(T)/
b
Figure 2. Default solute van der Waals radii in Jaguar (rO = rN = 1.60
reaction
B3LYP
M06
M06-L
2X
CBS
c
Å). van der Waals radii for O and N in Jaguar scaled by a factor 1.2
2AN = (AN)2
−4.15
−6.14
−5.94
−6.33
−6.40
(rO = rN = 1.92 Å).
NO+ + (AN)2 =
−52.38 −51.93 −56.19 −46.68
−44.09 −44.62 −44.01 −45.89
−22.58 −23.73 −24.44 −23.10
−20.53 −21.65 −22.04 −21.64
−43.48
NO+(AN)2
+
NO2 + (AN)2 =
−42.79
−22.25
−21.24
calculations are highly dependent on the choice of atomic
radii that define the solute−solvent boundary, which are
typically optimized for neutral organic molecules in water.
Using the default solute radii in Jaguar leads to a significant
overestimation (on average by ∼10 kcal/mol) of ΔG*solv for all
ions. In this work, the default solute radii are scaled by 1.20 to
improve the accuracy of solvation calculations for small neutral
molecules in AN. Using a single scaling factor also helps to
improve the calculated ΔG*solv for ions, which are now in much
better agreement with the values obtained using the mixed
cluster/continuum model. However, even with the optimized
scaling parameter, the implicit solvation calculations over-
estimate ΔG*solv for anions and underestimate ΔG*solv for
cations (Table 3).
+
NO2 (AN)2
−
NO2 + (AN)2 =
NO2−(AN)2
−
NO3 + (AN)2 =
NO3−(AN)2
functional employed in our previous work24 to model the
complexation of Li+ and O2 with AN does a good job in
−
+
−
reproducing the interaction energies for NO2 , NO2 , and
NO3 but strongly overbinds for the NO+ ion. M06-2X is the
−
only functional that gives a reasonable estimation of the binding
energy between NO+ and (AN)2. Thus, based on the evaluation
of several DFT methods, M06-2X was utilized to model the
solvation of the NO+ ion, while M06-L was employed to model
+
−
−
the solvation of NO2 , NO2 , and NO3 ions.
4.4. Calculations of Redox Potentials of Nitrogen
Oxides in AN. Redox potentials versus Li+/Li scale are
calculated using the thermodynamic cycle shown in Figure 3. In
The structures of the most stable X(AN)8 clusters (X = NO+,
NO2 , NO2 , NO3 ) used to calculate ion solvation free
+
−
−
E
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Table 4. Comparison of Calculated and Experimental Redox
Potentials for Oxygenated Nitrogen Compounds in AN (vs
Li+/Li, V)
a
half-reaction
calcd
expt
b
−
NO2 + e ⇌ NO2
3.44
4.30
4.84
5.07
3.50
4.17
c
NO+ + e ⇌ NO
d
e
+
NO2 + e ⇌ NO2
∼4.77
∼4.95
−
Figure 3. Thermodynamic cycle for the calculation of redox potentials
NO3 + e ⇌ NO3
versus Li+/Li scale.
a
Using the thermodynamic cycle shown in Figure 3. Electron
attachment/detachment energies are obtained at the RCCSD(T)
this cycle, ΔfGo[Li+(g)] is the standard gas-phase free energies
of formation of Li+, ΔGao[X+/o(g)] is the electron attachement
free energy of X+/o(g), F is Faraday’s constant, E* is the
standard reduction potential vs Li+/Li, ΔG*solv is the solvation
level. Zero point energies and thermal corrections are obtained at the
b
B3LYP/6-311+G* level. Obtained from the simulated CVs for the
oxidation of 10 mM LiNO2 under Ar and corrected for the ∼30 mM/
c
bar solubility18 of NO2 in AN. Determined from the CV experiments
for 10 mM NO+BF4− (Figure S4) and corrected for the ∼10 mM/bar
free energy of Li+, X+/o, and Xo/− (Xo/− = NO2 , NO, NO2,
−
d
solubility17 of NO in AN. Estimated from the reported21 difference of
NO3 ), and ΔGo→* is defined by eq 6. Through the use of the
−
+
0.60
0.10 V in standard potentials between the NO2 /NO2 and
thermodynamic cycle in Figure 3, we can write
e
NO+/NO couples in several solvents. Estimated from the anodic peak
current potential in the CV experiments for 10 mM LiNO3 at a scan
rate of 100 mV/s.
−FE*(vs Li+/Li) = Δf Go[Li+(g)] + ΔGao[X+/o(g)] + ΔGo→*
+ ΔG*solv (Li+) + ΔG*solv (Xo/−) − ΔG*solv(X+/o
)
(7)
+
place in the same oxidation potential range as that of NO2 ,
The superscript asterisk is used to underline that the electrode
potential refers to a 1 M ideal solution. Substituting the
tabulated46 value for ΔfGo[Li+(g)] (154.59 kcal/mol) and
ΔGo→* (1.89 kcal/mol) into eq 7, together with the value of
ΔG*solv(Li+) (−121.7 kcal/mol) calculated using the explicitly
solvated Li+(AN)4 cluster, we obtain
while the study by Boughriet et al.21 found that the oxidizing
+
character of NO2 ions is much higher (by 0.6 0.1 V) than
that of NO+ ions. Our calculated difference in the redox
potentials of NO+/NO and NO2 /NO2 couples (0.54 V) agrees
+
closely with the results of Boughriet et al.21 We therefore
concur with their conclusion that the disparate results found by
Lee et al.18 are due to problems of solvent contamination by
E*(vs Li+/Li) = −1.51(V) − (ΔGao[X+/o(g)] + ΔG*solv(Xo/−
− ΔG*solv (X+/o))/F
)
+
traces of water, which promote the conversion of NO2 to
(8)
NO+.
The calculations of E* reported here are based on accurate
values of ΔGao[X+/o(g)] obtained from RCCSD(T)/CBS
calculations and our best estimates of ΔG*solv(X ) derived
from mixed cluster/continuum calculations (the second column
of Table 3). CV was used to approximate the redox potential of
4.5. Mechanistic Insights into the Reaction Between
−
NO2 and NO2/N2O4 in AN. There is direct experimental
−
evidence that the reaction between NO2 and NO2/N2O4 in
AN to form NO and NO3− is very fast on the time scale of the
CV experiments (reaction 4). However, there is a lack of
fundamental understanding of the underlying mechanism that
drives this reaction. Quantum chemical calculations were
carried out to estimate various reaction pathways for the
overall reaction 4.
−
NO2/NO2 and NO+/NO couples. If a redox process involves
oxidation or reduction of a gas, a voltage correction for gas
solubility in the units of M/bar was applied in order to make a
proper comparison of the measured redox potentials derived
from CV with computed ones based on quantum chemical
calculations (see ref 48 for a more thorough discussion). At low
LiNO2 and LiNO3 concentrations (10 mM), it is reasonable to
assume that the two ions are well separated and can be very
well modeled as “free” ions.
In many reactions with protic substrates, it is not the NO2
monomer itself, but its asymmetric dimer, ONO-NO2, that is
considered to be the active species.33 The explanation for the
enhanced reactivity of ONO-NO2 lies in its ability to
autodissociate and to generate strongly oxidizing nitrosyl
(NO+) ions. Assuming a very fast reaction between oppositely
charged NO+ and NO2− ions, we might similarly expect ONO-
Table 4 shows a comparison of the calculated and
experimental redox potentials of oxygenated nitrogen com-
pound in AN. Overall, our computational protocol is capable of
accurately predicting the redox potentials of oxygenated
nitrogen compounds in AN solution, with the maximum
error and MUE of only 0.13 and 0.10 V, respectively. Among
the DFT methods (Table S2), the M06 density functional with
the MUE of 0.17 V shows the best overall performance and
thus can be recommended to calculate electron-transfer
energies in cases where the RCCSD(T) method is not feasible.
It is important to note that no adjustable parameters were
employed in these calculations. In contrast, a close agreement
between computed and experimentally determined redox
potentials would not be possible with pure dielectric continuum
solvent models without the use of empirical fitting parameters.
Previous results showed some ambiguity regarding the
−
NO2 to directly participate in the reaction with NO2 . To test
this hypothesis, we studied the interconversion between NO2
and its dimers at the RCCSD(T)/CBS level, with solvation
effects included through the polarizable continuum mode
(Figure 4). We note that the mechanism presented here is
qualitatively similar to the gas-phase mechanism suggested by
Liu and Goddard.33 All stationary points obtained previously at
the B3LYP level33 were refined at the RCCSD(T)/aug-cc-
pVDZ level of theory.
A trans form of ONO-NO2 is expected to be more prone to
autodissociation, because it is reported to provide a low-energy
pathway for exchange of NO+ between oxygen atoms in NO3 .
−
A direct transformation 2NO2 → trans-ONO-NO2 is associated
with an activation energy of 21.8 kcal/mol. Following the work
of Liu and Goddard,33 we have also found a low energy reaction
path from the two NO2 monomers to the trans-ONO-NO2
electrochemical behavior of NO+ and NO2 ions. The work
+
by Lee at al.18 suggested that the reduction of NO+ is taking
F
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
Figure 4. Reaction free energy profile for interconversion between NO2 and its dimers. Energies are estimated at the CCSD(T)/CBS level with
solvation effects included through the polarizable continuum model. CCSD(T)/aug-cc-pVTZ-optimized bond distances are shown in Å.
dimer, which consists of the following steps: 2NO2 → C2v-
N2O4 → cis-ONO-NO2 → trans-ONO-NO2. The transition
state for the rate-determining step (TS2), C2v-N2O4 → cis-
ONO-NO2, lies only 10.6 kcal/mol above the two NO2
monomers or 15.5 kcal/mol above the most stable symmetric
N2O4 dimer. Furthermore, the autodissociation of trans-ONO-
−
NO2 in the presence of NO2 is found to be barrierless,
suggesting that the reaction barrier for trans-ONO-NO2 →
NO+ + NO3 is essentially determined by the thermodynamic
−
stability of reaction products. The low-energy pathway for
interconversion between NO2 monomers and trans-ONO-NO2
is in accordance with a relatively high rate observed for reaction
4. Conversely, the reaction pathway involving the direct oxygen
transfer from either NO2 or symmetric N2O4 dimer (D2h) to
−
NO2 requires much higher activation energy (>25 kcal/mol),
and therefore it is not supported by our calculations (Figures
S1 and S2).
4.6. Electrochemical Oxidation of Lithium Nitrite
(LiNO2) in AN: A Comparison of Experimental and
Simulated Cyclic Voltammograms. Based on the computa-
tional results presented in the preceding sections, we can
provide a mechanistic analysis of electrochemical oxidation of
−
NO2 . To verify the validity of a given mechanism of coupled
electrochemical and chemical reactions in AN, the one-
dimensional simulation program DigiElch 6.F43 was used to
model the CVs of 10 mM LiNO2 on a Pt macroelectrode under
inert atmosphere at scan rates of 50, 100, and 200 mV/s. To
evaluate how the formation of various NO2 dimers influences
−
the electrochemistry of NO2 in AN, two reactions scheme of
different degrees of complexity were considered.
The basic reaction scheme comprises the following electro-
chemical and chemical steps (model 1, reactions 9−12):
−
Figure 5. Comparison of experimental and simulated the first scan
cyclic voltammograms for the oxidation of 10 mM LiNO2 in 0.5 M
LiTFSI/AN electrolyte on a Pt macroelectrode (d = 5 mm) at scan
rates of 50, 100, and 200 mV/s. Cyclic voltammograms are shown after
subtracting the background current and full IR compensation. Digital
simulation of the CV curves based on (a) reactions 9−12 (model 1)
and (b) reactions 9, 11, 12, and 14−19 (model 2).
NO2 + e → NO2
(9)
NO2 + NO2 ⇌ [N2O ] ⇌ NO+ + NO3
−
(10)
4
NO2 + NO+ → NO2 + NO
−
(11)
(12)
NO+ + e → NO
The voltammograms of LiNO2 in AN under argon show
three distinct anodic and two cathodic waves (Figure 5). The
anodic peak at ∼3.5 V vs Li+/Li is assigned to electrode
reaction 9, followed by a series of rapid chemical reactions 10
G
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
and 11. In this mechanism, the nitrite oxidation product, NO2,
undergoes dimerization and subsequent autoionization to form
cathodic current at 3.8−4.0 V drops and rises more quickly
than indicated by experiment. Theoretical calculations estab-
lished (Section 4.5) that the autodissociation process is more
complex than that described by reaction 10. To determine if the
additional reactions predicted by the theory improve the
agreement between simulations and the experimental voltam-
mograms, a more comprehensive reaction scheme was
employed in the simulations (model 2), which in addition to
reactions 9, 11, and 12 includes the following steps:
NO+ and NO3 ions. The ionic recombination ultimately
−
−
enables the conversion of NO2 into equimolar mixture of
−
NO3 and NO during the first oxidation wave (the sum of
reactions 9−11)
−
−
NO3 + NO + e → 2NO2
(13)
The first oxidation step appears irreversible at scan rates 50−
200 mV/s, because there is no associated cathodic peak when
the potential scan is reversed after the first oxidation wave
(Figure S5).
NO2 + NO2 ⇌ N2O
(14)
(15)
4
NO2 + NO2 ⇌ trans‐ONO‐NO2
The second anodic peak at ∼4.00 V and the third anodic
peak at ∼4.16 V vs Li+/Li were both assigned to NO oxidation
according to reaction 12. CV of NOBF4 in AN solution showed
that the NO/NO+ couple is quasi-reversible, with a midpoint
potential of 4.05 V (Figure S4). However, when LiNO3 was
incrementally added to solution of NOBF4, a peak splitting
similar to the one shown in Figure 5 was observed (see
Supporting Information and Figure S6 for more details). We
offer the following interpretation of this phenomenon. First we
note that the first anodic wave for NO oxidation in the
presence of NO2− and NO3− is slightly shifted to a less positive
potential (by ∼0.05 V). This observation can be rationalized by
considering a fast transformation of electrogenerated NO+
according to reactions 10 and 11. The in situ generated NO2
−
trans‐ONO‐NO2 ⇌ NO+ + NO3
(16)
(17)
(18)
(19)
NO2 + NO ⇌ N2O3
−
trans‐ONO‐NO2 + e → NO + NO3
−
N2O + e → NO + NO3
4
As evident from Figure 5b, the inclusion of a more complete set
of reactions provides significant improvements to the fits. The
equilibrium constants for reactions 11 and 14−17 used in the
simulation (Table S3) are in good agreement with the
calculated reaction free energies in AN. The high rates of
reactions 15 and 16 used in the simulation are consistent with
the relatively low barriers of these reactions estimated from
theoretical calculations (Figure 4). Furthermore, the ratio of
diffusion coefficients, DNO2/DNO2 and DNO/DNO < 1, found
by simulation (Table S2) is in line with the increase of the
effective hydrodynamic radius for the stronger solvated ionic
species compared to the neutral counterpart.49 It should be
emphasized that all of the above reactions provide incremental
improvements in the quality of the fit. The inclusion of N2O3
(reaction 17) was motivated by the increased stability of this
oxide in aprotic organic solvents, such as AN.50 The addition of
electrode reactions 18 and 19 was needed in order to
adequately reproduce the cathodic part of the CVs. Thus, the
results demonstrate that quantum chemical calculations provide
a useful framework to identify the intermediate products
responsible for voltammetric behavior of nitrite ions under Ar
in AN solution.
−
−
dimer can subsequently react with NO2 to form NO3 and
NO again. As a result, NO is being continuously generated
during the second oxidation wave at ∼4.00 V, giving rise to the
third anodic wave at ∼4.16 V vs Li+/Li. Fast kinetics for the
−
+
−
reaction of NO3 with the excess of NO+ near the electrode
−
surface can account for the absence of NO3 oxidation peak at
∼4.95 V. The two small successive cathodic waves at ∼4.05 and
∼3.46 V were attributed to the reduction of NO+ and NO2/
N2O4, respectively.
The nonlinear fitting procedure implemented in DigiElch43
was used to establish the set of simulation parameters that
provides the best average agreement between simulation and
experiment at various scan rates (Table S3). The points in
Figure 5a represent the best-fit simulation voltammograms
based on reactions 9−12. We note that it was not possible to
obtain a satisfactory fit to the experimental CVs by using the
overall reaction 4 directly (as opposed to using reactions 10 and
11). This is consistent with quantum chemical calculations
demonstrating a complex multistep mechanism for reaction 4 in
AN solvent.
The proposed mechanism for the oxidation of LiNO2 under
Ar can be extended to describe the electrochemical behavior of
LiNO2 in the presence of oxygen by considering a fast
reaction51 between NO and O2
It is important to note that our calculations reproduce the
anodic peak splitting with a single electrochemical step
(reaction 12), which lends further credit to our mechanistic
interpretation of NO oxidation in CV experiments. Sensitivity
analysis of simulation parameters further reveals that the NO/
NO+ peak splitting can only be observed if the diffusion
coefficients of NO and NO2 and the rate constant for reaction
10 are above a certain threshold, while the diffusion coefficient
NO + 1/2O2 → NO2
(20)
NO generated at the first oxidation wave, according to the
overall reaction 13, is quickly converted51 to NO2, which in
turn rapidly reacts with NO2− (the overall reaction 4) to create
a catalytic loop where more NO is produced to react again with
O2. Taking the reactions steps 4, 13, and 20 together, the
overall process can be summarized as follows
−
of NO3 is below a certain limit. This could explain why the
peak splitting is observed in some solvents16,19 but not in
−
−
nNO2 + (n − 1)/2O2 → (n − 1)NO3 + NO2 + e
others.12,22
(21)
Although a reasonable fit between simulated and exper-
imental CVs was obtained in Figure 5a, some notable
discrepancies are evident. The simulation significantly under-
estimates the height of the cathodic peak at ∼3.46 V and does
not reproduce the relative heights of the second and third
anodic waves as the scan rate increases. Additionally, the
where n depends on the relative rates of the chemical and
electrochemical steps. The changes in first-scan voltammo-
grams of 10 mM LiNO2 in AN observed upon switching from
Ar to O2 atmosphere are fully consistent with this mechanistic
picture (Figure 6).
H
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
successive oxidation waves of the nitrite anion. In this work we
apply first-principles computational methods to understand the
chemical and electrochemical behavior of LiNO2 in AN and
identify the intermediate products responsible for voltammetric
behavior of nitrite ion in AN solution. Our computational
protocol based on coupled-cluster calculations at the complete
basis set limit (RCCSD(T)/CBS) in the gas phase and mixed
cluster/continuum calculations is solution is capable of
accurately predicting the redox potentials of oxygenated
nitrogen compounds in AN. A detailed analysis of reaction
between NO2− and NO2/N2O4 suggests a mechanism in which
the interconversion between nitrogen dioxide and its various
dimers provides a low-energy path for the reaction with the
nitrite anion. A set of coupled electrochemical and chemical
steps predicted by the theory was used to build the reaction
network for the digital simulation of cyclic voltammograms of
LiNO2 in AN under argon. Good agreement between
simulations and the experimental voltammograms is only
achieved with inclusion of a more complete set of reactions
suggested by the theory. The results establish a possible
mechanism of regeneration of LiNO3 in electrolyte in the
presence of oxygen, which is uniquely possible under charging
conditions in a Li-O2 battery.
Figure 6. IR-corrected first scan cyclic voltammograms for the
oxidation of 10 mM LiNO2 under argon (black curve) and oxygen
(red curve) and oxidation of 10 mM LiNO3 under Ar (blue curve) in
0.5 M LiTFSI/AN electrolyte on a Pt macroelectrode (surface area =
0.196 cm2) at a scan rate of 100 mV/s.
In a more broad content, this study provides a theoretical
framework for computational electrochemistry in solution. We
find that with a choice of an accurate computational method in
the gas phase, implicit solvation model optimized for neutral
molecules, and explicit consideration of the most important
solute−solvent interactions to describe the solvation of ions,
first principles methods can provide valuable insights into
mechanisms of electrochemical processes in organic media and
guide the assignment of the cyclic voltammetry peaks.
Indeed, the first anodic peak current is considerably lower
−
under O2, because the consumption of NO2 via anodic
oxidation competes with consumption resulting from the
chemical reaction in the presence of O2. In addition, two
new peaks at ∼4.85 and ∼4.95 V vs Li+/Li are generated in the
voltammogram under O2. The first peak at ∼4.85 V appears in
+
the same potential range where the oxidation of NO2 to NO2
occurs (as evidenced both by measurements21 and calculations
ASSOCIATED CONTENT
* Supporting Information
given in Table 4). The second peak at ∼4.95 V is unequivocally
■
−
attributed to the oxidation of NO3 , as shown by the direct
S
comparison with the anodic peak for the oxidation of 10 mM
LiNO3 in AN (Figure 6). Finally, we note that the first anodic
peak at ∼3.5 V completely disappears under steady-state
Complete ref 26, RMP2 and RCCSD(T) reaction/activation
energies as a function of basis set size (Table S1), performance
of several DFT methods in reproducing the redox potentials of
oxygenated nitrogen compounds (Table S2), parameters
−
conditions, indicating a full conversion of NO2 according to
−
process (reaction 21) near the electrode surface. This provides
employed in the CV simulation of NO2 oxidation under Ar
−
a mechanism by which NO3 can at least be partially
(Table S3), additional computational (Figures S1, S2) and
experimental details (Table S4, Figures S3−S6), Cartesian
coordinates, electronic states, and absolute energies (in
Hartrees) for tested molecular structures obtained with the
B3LYP functional. This material is available free of charge via
the Internet at http://pubs.acs.org.
−
regenerated from NO2 under conditions that are uniquely
possible in Li-O2 cells during the process of cell charging.
Preventing or mitigating the loss of LiNO3 served to stabilize
the SEI on Li metal (reaction 1) could be a contributing factor
to the observed synergistic effect of O2 and LiNO3 on the
interfacial stability and cycling of Li metal, as reported11 by our
group previously.
AUTHOR INFORMATION
Corresponding Author
slava@liox.com
■
5. CONCLUSIONS
Degradation of electrolyte components under conditions that
exist in rechargeable Li-O2 batteries is one of the most difficult
challenges to overcome. Recently we have presented a new
strategy for stabilizing the Li/electrolyte interface in solvents
that do not form a stable SEI on their own.6,11 LiNO3 was
found to serve the dual purpose of an electrolyte salt and an SEI
film stabilizing agent. The reduction of LiNO3 on the Li metal
results in the formation of insoluble Li2O that contributes
toward stabilizing the SEI film, while the other more soluble
product, LiNO2, is electroactive within the operating potential
window of a Li-O2 cell.6 Due to lack of detailed mechanistic
understanding of nitrite oxidation, there are significant
contradictions and variations in the interpretation of the
Notes
The authors declare no competing financial interest.
REFERENCES
■
(1) Girishkumar, G.; McCloskey, B.; Luntz, A. C.; Swanson, S.;
Wilcke, W. J. J. Phys. Chem. Lett. 2010, 1, 2193−2203.
(2) Bruce, P. G.; Freunberger, S. A.; Hardwick, L. J.; Tarascon, J.-M.
Nat. Mater. 2012, 12, 19−29.
(3) Christensen, J.; Albertus, P.; Sanchez-Carrera, R. S.; Lohmann,
T.; Kozinsky, B.; Liedtke, R.; Ahmed, J.; Kojic, A. J. Electrochem. Soc.
2012, 159, R1−R30.
(4) Shao, Y.; Ding, F.; Xiao, J.; Zhang, J.; Xu, W.; Park, S.; Zhang, J.-
G.; Wang, Y.; Liu, J. Adv. Funct. Mater. 2013, 23, 987−1004.
I
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(5) Cheng, F.; Chen, J. Chem. Soc. Rev. 2012, 41, 2172−2192.
(6) Walker, W.; Giordani, V.; Uddin, J.; Bryantsev, V. S.; Chase, G.
V.; Addison, D. A. J. Am. Chem. Soc. 2013, 135, 2076−2079.
(7) Bryantsev, V. S.; Uddin, J.; Giordani, V.; Walker, W.; Addison, D.;
Chase, G. V. J. Electrochem. Soc. 2013, 160, A160−A171.
(8) Poris, J.; Raistrick, I.; Huggins, R. Proc. Electrochem. Soc. 1984,
84−2, 313−321.
(42) Fan, C. F.; Olafson, B. D.; Blanco, M.; Hsu, S. L. Macromolecules
1992, 25, 3667−3676.
(43) Rudolph, M. J. Comput. Chem. 2005, 26, 1193−1204; see also
www.ElchSoft.com.
(44) Feller, D.; Dixon, D. A. J. Chem. Phys. 2001, 115, 3484−1−13.
̌
(45) Rezac
2155.
́ ̌
, J.; Hobza, P. J. Chem. Theory Comput. 2013, 9, 2151−
(46) NIST Chemistry Webbook; NIST: Gaithersburg, MD, 2005,
http://webbook.nist.gov/chemistry/
(9) Aurbach, D.; Pollak, E.; Elazari, R.; Salitra, G.; Kelley, C.; Affinito,
J. J. Electrochem. Soc. 2009, 156, A694−A702.
(47) Kelly, C. P.; Cramer, C. J.; Truhlar, D. G. J. Phys. Chem. B 2007,
111, 408−422.
(10) Zhang, S. S. Electrochim. Acta 2012, 70, 344−348.
(11) Giordani, V.; Walker, W.; Bryantsev, V. S.; Uddin, J.; Chase, G.
V.; Addison, D. A. J. Electrochem. Soc. 2013, 160, A1544−A1550.
(12) Uddin, J.; Bryantsev, V. S.; Giordani, V.; Walker, W.; Chase, G.
V.; Addison, D. A. J. Phys. Chem. Lett. 2013, 4, 3760−3765.
(13) Martins, M. E.; Calandra, A. J.; Arvía, A. J. Electrochim. Acta
1970, 15, 111−126.
(14) Castellano, C. E.; Wargon, J. A.; Arvía, A. J. Electroanal. Chem.
Interfacial Electrochem. 1973, 47, 371−372.
(15) Bontempelli, G.; Mazzocchin, G.-A.; Magno, F. Electroanal.
Chem. Interfacial Electrochem. 1974, 55, 91−100.
(48) Bryantsev, V. S.; Addison, D.; Chase, G. V. J. Electrochem. Soc.
2013, 160, H818−H819.
(49) Macías-Ruvalcaba, N. A.; Evans, D. H. J. Phys. Chem. B 2010,
114, 1285−1292.
(50) Boughriet, A.; Wartel, M.; Fischer, J. C.; Auger, Y. J. Electroanal.
Chem. 1985, 186, 201−209.
(51) Gary, R. M. Synthetic Nitrogen Products: A Practical Guide to the
Products and Processes; Kluwer Academic/Plenum Publishers: New
York, 2004.
(16) Bontempelli, G.; Mazzocchin, G.-A.; Magno, F.; Seeber, R.
Electroanal. Chem. Interfacial Electrochem. 1974, 55, 101−107.
(17) Lee, K. Y.; Kuchynka, D. J.; Kochi, J. K. Inorg. Chem. 1990, 29,
4196−4204.
(18) Lee, K. Y.; Amatore, C.; Kochi, J. K. J. Phys. Chem. 1991, 95,
1285−1294.
(19) Boughriet, A.; Fischer, J. C.; Leman, G.; Wartel, M. Bull. Soc.
Chim. Fr. 1985, 1, 8−13.
(20) Boughriet, A.; Coumare, A.; Wartel, M. Electrochim. Acta 1990,
35, 389−397.
(21) Boughriet, A.; Wartel, M. J. Electroanal. Chem. 1993, 362, 167−
176.
(22) Broder, T. L.; Silvester, D. S.; Aldous, L.; Hardacre, C.;
Compton, R. G. J. Phys. Chem. B 2007, 111, 7778−7785.
(23) Bryantsev, V. S.; Diallo, M. S.; Goddard, W. A., III J. Phys. Chem.
B 2008, 112, 9709−9719.
(24) Bryantsev, V. S. Theor. Chem. Acc . 2012, 131, 1250−1−7.
(25) Moller, C.; Plesset, M. S. Phys. Rev. 1934, 46, 618−622.
̈
̀
(26) Straatsma, T. P.; Apra, E.; Windus, T. L.; Bylaska, E. J.; de Jong,
W.; Hirata, S.; Valiev, M.; Hackler, M.; Pollack, L.; Harrison, R. et al.
NWChem, A Computational Chemistry Package for Parallel Computers,
version 5.1; Pacific Northwest National Laboratory, Richland, WA,
2007.
(27) Purvis, G. D., III; Bartlett, R. J. J. Chem. Phys. 1982, 76, 1910−
1918.
(28) Raghavachari, K.; Trucks, G. W.; Pople, J. A.; Head-Gordon, M.
Chem. Phys. Lett. 1989, 157, 479−483.
(29) Watts, J. D.; Gauss, J.; Bartlett, R. J. J. Chem. Phys. 1993, 98,
8718−8733.
(30) Dunning, T. H., Jr. J. Chem. Phys. 1989, 90, 1007−1023.
(31) Kendall, R. A.; Dunning, T. H., Jr.; Harrison, R. J. J. Chem. Phys.
1992, 96, 6796−6806.
(32) Koch, H.; Fernandez, B.; Christiansen, O. J. Chem. Phys. 1998,
108, 2784−1−7.
(33) Liu, W.-G.; Goddard, W. A., III. J. Am. Chem. Soc. 2012, 134,
12970−12978.
(34) Becke, A. D. Phys. Rev. A 1988, 38, 3098−3100.
(35) Lee, C. T.; Yang, W. T.; Parr, R. G. Phys. Rev. B 1988, 37, 785−
789.
(36) Zhao, Y.; Truhlar, D. G. A J. Chem. Phys. 2006, 125, 194101−1−
17.
(37) Zhao, Y.; Truhlar, D. G. Theor. Chem. Acc . 2008, 120, 215−241.
(38) Jaguar, version 7.7, Schrodinger, LLC, New York, 2010.
̈
(39) Ribeiro, R. F.; Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. J.
Phys. Chem. B 2011, 115, 14556−14562.
(40) Boughriet, A.; Wartel, M.; Fischer, J. C. J. Electroanal. Chem.
1985, 190, 103−115.
(41) Blanco, M. J. Comput. Chem. 1991, 12, 237−247.
J
dx.doi.org/10.1021/ja410766n | J. Am. Chem. Soc. XXXX, XXX, XXX−XXX