Synthesis and Characterization of Novel Fluoride
and Oxide Cathodes for Rechargeable Batteries
by
Nancy Twu
B.S., Rutgers, The State University of New Jersey (2008)
Submitted to the Department of Materials Science and Engineering
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Materials Science and Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2015
c Massachusetts Institute of Technology 2015. All rights reserved.
○
Author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Department of Materials Science and Engineering
April 3, 2015
Certified by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Gerbrand Ceder
R. P. Simmons Professor of Materials Science and Engineering
Thesis Supervisor
Accepted by . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Donald R. Sadoway
Chair, Department Committee on Graduate Students
2
Synthesis and Characterization of Novel Fluoride and Oxide
Cathodes for Rechargeable Batteries
by
Nancy Twu
Submitted to the Department of Materials Science and Engineering
on April 3, 2015, in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy in Materials Science and Engineering
Abstract
Developing new cathode materials is key to improving the energy density of rechargeable batteries and enabling new applications of energy storage.
In this thesis, two
families of materials were explored as candidate cathode materials: the dirutile and
rutile polymorphs of LiMnF4 , and layered lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 .
Dirutile LiMnF4 was identified from high-throughput computation as a promising conversion cathode. The dirutile polymorph was synthesized through a new lowtemperature route, and the rutile polymorph was discovered upon mechanical milling.
With simple synthesis and electrode preparation methods, both dirutile and rutile
polymorphs of LiMnF4 showed electrochemical activity.
Electron diffraction con-
firmed both polymorphs to convert upon lithiation along different reaction paths. As
with other fluorides, specific capacity was strongly linked with processing conditions.
The layered lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds were designed from
recent understanding of diffusion channels in lithium-excess materials.
Increasing
lithium content was found to improve both discharge capacity and capacity retention. Structural studies revealed a complex nanostructure pattern of Li-Sb and Ni-Sb
ordering where the interface between these domains formed the correct local configuration for good lithium mobility. The < 1nm Li-Sb stripe domains enable percolation
of the low barrier lithium diffusion channels at lower lithium excess levels.
The redox mechanisms of the lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 materials were
then studied as a function of lithium content and rate. Li1.15 Ni0.47 Sb0.38 O2 surprisingly exhibited higher discharge capacities at faster rates, and traversed distinct voltage curves at different rates.
Characterization of redox processes confirmed nickel
redox and oxygen loss, with oxygen redox proposed to account for the balance of the
capacity. Finally, irreversible nickel migration is suggested as an explanation for the
rate-dependent voltage curve features.
Thesis Supervisor: Gerbrand Ceder
Title: R. P. Simmons Professor of Materials Science and Engineering
3
4
Acknowledgments
First, I would like to thank my thesis advisor, Professor Gerbrand Ceder, for his
insightful comments, for holding a high standard to papers and presentations, and
for his enthusiasm for research. I am also grateful for the opportunity to pursue ideas
in any direction I found interesting.
I would like to thank my thesis committee members, Professor Jeff Grossman and
Professor Yang Shao-Horn, for their suggestions and complementary perspectives to
my research.
Also, a special thanks to Professor Shao-Horn for connecting me to
Professor Hubert Gasteiger’s group at Technische Universität München.
I am grateful for financial support from the National Science Foundation’s Graduate Research Fellowship. I also thank the sponsors of the lithium subgroup, Robert
Bosch Corporation and Umicore, for their support of fundamental science, patience
through 200 slide presentations, and encouragement.
I owe many thanks to my collaborators in the Ceder group (Dr. Xin Li, Charles
Moore, Dr. Alex Urban, Jinhyuk Lee, and Dr. Lei Liu), at Argonne National Laboratory (Dr. Mali Balasubramanian), and at Technische Universität München (Michael
Metzger and Dr.
Cyril Marino) for their time, interest in my projects, and many
fruitful discussions. I thank all my collaborators for helping to build a complete story
in my research. I also thank Dr. Scott Speakman for many helpful discussions around
x-ray diffraction.
I would like to thank the entire Ceder group for building a supportive and intellectually stimulating environment. In particular, I thank Dr. Xiaohua Ma, Dr. Hailong
Chen, and Dr.
Jae Chul Kim for mentoring me when I first joined the group, the
lithium excess subgroup for lively discussions and comradery, the experimentalists of
the Ceder group for bonding time in the lab, my officemates past and present for
the friendly and positive environment, and Wenhao Sun and Jinhyuk Lee for all the
coffee and milkshakes. Also, a huge thanks to Kathy Simons for always beng helpful,
welcoming, and making the group run smoothly.
Graduate school would not be complete without great friends.
5
In particular, I
thank my classmates Sema Ermez, Tim Milakovich, Kunal Mukherjee, and Wenhao
Sun, who have been there for me through the best and worst of times, and always
inspire me with their enthusiasm and interest in life. I have also been exceptionally
lucky to have many lovely roommates throughout graduate school. Mariah Mandt,
Sema Ermez, Gözde Rainville, Ece Alpaslan, and Angie King – thank you for making
home feel like home.
To my mom, dad, and brother – I would not be where I am today without your
love and support.
Finally, to my fiancé, Vincent Lee – thank you for always believing in me.
6
To my Mom and Dad
7
8
Contents
1
2
Introduction
25
1.1
Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
1.2
Overview of rechargeable batteries . . . . . . . . . . . . . . . . . . . .
26
1.2.1
Configuration and operation of a battery . . . . . . . . . . . .
26
1.2.2
Important properties of rechargeable batteries
. . . . . . . . .
30
1.2.3
Conversion cathodes
. . . . . . . . . . . . . . . . . . . . . . .
32
1.2.4
Intercalation cathodes
. . . . . . . . . . . . . . . . . . . . . .
33
1.3
Diffusion in layered oxides
. . . . . . . . . . . . . . . . . . . . . . . .
35
1.4
Lithium-excess and 0-transition metal (0-TM) diffusion channels . . .
37
1.5
Overview of the thesis
40
. . . . . . . . . . . . . . . . . . . . . . . . . .
Dirutile and Rutile LiMnF4 : Two New Conversion Cathodes
43
2.1
. . . . . .
43
2.1.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
43
2.1.2
Advantages
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
44
2.1.3
Synthesis challenges . . . . . . . . . . . . . . . . . . . . . . . .
46
2.1.4
Other challenges
. . . . . . . . . . . . . . . . . . . . . . . . .
47
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
48
2.2.1
Experimental characterization . . . . . . . . . . . . . . . . . .
48
2.2.2
Electrochemistry
. . . . . . . . . . . . . . . . . . . . . . . . .
49
2.2.3
Computations . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
Materials preparation . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
2.3.1
50
2.2
2.3
Fluorides and oxyfluorides as conversion cathode materials
Synthesis of dirutile LiMnF4 . . . . . . . . . . . . . . . . . . .
9
2.3.2
2.4
3
Synthesis of rutile LiMnF4 . . . . . . . . . . . . . . . . . . . .
Structure characterization
51
. . . . . . . . . . . . . . . . . . . . . . . .
52
2.4.1
Rietveld refinement . . . . . . . . . . . . . . . . . . . . . . . .
52
2.4.2
Mechanically-driven phase transformations . . . . . . . . . . .
54
2.4.3
Computed polymorph energies . . . . . . . . . . . . . . . . . .
56
2.5
Electrochemical performance . . . . . . . . . . . . . . . . . . . . . . .
57
2.6
Lithiation mechanisms
. . . . . . . . . . . . . . . . . . . . . . . . . .
60
2.6.1
Observed conversion reactions . . . . . . . . . . . . . . . . . .
60
2.6.2
Calculated intercalation properties
61
2.6.3
Comparison of conversion and intercalation reactions
. . . . . . . . . . . . . . .
. . . . .
65
2.7
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
2.8
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
Layered Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 : Designing lithium-excess cathode materials from percolation theory
71
3.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
71
3.2
Materials design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
72
3.3
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
3.3.1
Synthesis
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
75
3.3.2
Experimental characterization . . . . . . . . . . . . . . . . . .
75
3.3.3
Electrochemistry
. . . . . . . . . . . . . . . . . . . . . . . . .
76
3.3.4
Computations . . . . . . . . . . . . . . . . . . . . . . . . . . .
76
3.4
Validating materials design by electrochemical performance . . . . . .
77
3.5
Structure characterization
78
. . . . . . . . . . . . . . . . . . . . . . . .
3.5.1
Transmission electron microscopy
. . . . . . . . . . . . . . . .
78
3.5.2
Synchrotron x-ray diffraction
. . . . . . . . . . . . . . . . . .
82
3.5.3
Density functional theory
. . . . . . . . . . . . . . . . . . . .
84
3.5.4
Extended x-ray absorption fine structure . . . . . . . . . . . .
85
3.6
Nanohighways in Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2
. . . . . . . . . . . . . . . . . .
88
3.7
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
91
10
4
Layered Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 : Understanding the origins of higher capacities at faster rates
93
4.1
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
93
4.2
Experimental methods
. . . . . . . . . . . . . . . . . . . . . . . . . .
94
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
Synthesis
4.2.2
Ex situ
experimental characterization . . . . . . . . . . . . . .
94
4.2.3
In situ
X-ray diffraction
95
4.2.4
In situ
on-line electrochemical mass spectroscopy
. . . . . . . . . . . . . . . . . . . . .
. . . . . . .
96
Electrochemical properties . . . . . . . . . . . . . . . . . . . . . . . .
96
4.3.1
Performance as a function of lithium content . . . . . . . . . .
96
4.3.2
Performance as a function of rate
. . . . . . . . . . . . . . . .
99
4.4
Characteristics of the 4.45V plateau . . . . . . . . . . . . . . . . . . .
101
4.5
Characterization of nickel redox activity
. . . . . . . . . . . . . . . .
103
4.6
Characterization of oxygen loss
. . . . . . . . . . . . . . . . . . . . .
105
4.7
Assessment of rate-dependent redox activity
. . . . . . . . . . . . . .
109
4.8
Quantification of capacity contributions . . . . . . . . . . . . . . . . .
112
4.9
Proposed effect of nickel migration on reversible capacity . . . . . . .
114
4.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
116
4.3
5
4.2.1
Conclusion
119
11
12
List of Figures
1-1
Comparison of different battery technologies in terms of volumetric
and gravimetric energy density. [1]
1-2
. . . . . . . . . . . . . . . . . . .
Schematic of a rechargeable lithium battery.
e
−
Upon charge, Li
+
26
and
move from the cathode to the anode through the electrolyte and
external circuit, respectively. Upon discharge, the reverse process occurs. 27
1-3
Schematic open circuit diagram for a thermodynamically stable battery. 29
1-4
Representative crystal structures of cathode materials for lithium-ion
batteries: (a) layered
𝛼-LiCoO2 ;
(b) cubic LiMn2 O4 spinel; (c) olivine-
structured LiFePO4 . Li ions are shown as light green spheres, CoO6 octahedra in blue, MnO6 octahedra in mauve, Fe–O polyhedra in brown,
and PO4 tetrahedra in purple. [2] . . . . . . . . . . . . . . . . . . . .
1-5
Schematic illustration of Li
fects in 1D channels. [3]
1-6
+
diffusion impeded by immobile point de-
. . . . . . . . . . . . . . . . . . . . . . . . .
32
Delithiation of a layered LiMO2 cathode during charge occurs by extraction of lithium from the lithium layers. . . . . . . . . . . . . . . .
1-7
31
Schematics of diffusion from octahedral lithium site
i
35
j
to site . a) Top
view in rocksalt (111)-direction. Large empty circles denote octahedral
sites. The arrow indicates one of two lithium diffusion channels that
connets sites
i
and
j.
Only one gate site,
k,
is depicted. b) 3D view
of the same diffusion channel along with both gate sites,
k
l
and . The
di-vacancy lithium diffusion mechanism requires one of the gate sites
to be vacant. [4] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13
36
1-8
Diffusion in layered materials is largely affected by (a) lithium slab
spacing and (b) transition metal valence. . . . . . . . . . . . . . . . .
1-9
36
Comparison of diffusion pathways in stoichiometric LiMO2 and lithiumexcess Li1+𝑥 M1−𝑥 O2 .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
1-10 Possible environments for an
o-t-o
37
hop in rocksalt-like Li-TM oxides.
o-t-o )
(A) octahedral–tetrahedral–octahedral (
diffusion: Two tetrahe-
dral paths connect each pair of neighboring octahedral sites.
(B to
D) The activated state can share faces with no octahedral transition
metals (0-TM channel) (B), one transition metal (1-TM channel) (C),
or two transition metals (2-TM channel) (D). [5] . . . . . . . . . . . .
38
1-11 0-TM accessible lithium atoms per formula unit as a function of the
overall lithium content and the degree of cation mixing in the layered
structure.
The percolation threshold is indicated by the thick black
contour line. Compositions falling into the region left of the contour
line are not 0-TM percolating. The thin line indicates the composition
at which one lithium atom per formula unit becomes 0-TM accessible.
[4]
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
39
1-12 0-TM percolation requires individual 0-TM channels to connect across
a particle. Segregation of 0-TM channels into high aspect ratio domains
lowers the number of 0-TM channels necessary for percolation. . . . .
2-1
40
Comparison of relative energies of bonding and antibonding molecular
orbitals in transition metal oxides (left) and transition metal fluorides
(right)
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
14
45
2-2
XRD reveals the formation of rutile LiMnF4 by mechanochemical reaction, and transformation of rutile LiMnF4 to dirutile LiMnF4 by solid
state reaction. XRD patterns are shown for (a) Ball milled LiF and
MnF3 . M = MnF3 and L = LiF still unreacted after milling. (b) Af-
∘
ter heating the ball milled precursors to 200 C for 10 hr under argon.
∘
(c) After heating the ball milled precursors to 250 C for 10 hr under
argon.
∘
(d) After heating the ball milled precursors to 300 C for 10
hr under argon.
phase, Li2 MnF5 .
Black circles denote the main peaks of an impurity
(e) Theoretical peak positions of dirutile LiMnF4 ,
space group P21 /c.
2-3
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
XRD patterns for (a) Rutile LiMnF4 obtained by ball milling precursors LiF and MnF3 . L and M denote the main peaks of unreacted LiF
and MnF3 precursor, respectively. (b) Dirutile LiMnF4 obtained from
∘
heating sample A for 10 hours at 300 C under argon.
Black circles
denote the impurity Li2 MnF5 phase. (c) Rutile LiMnF4 obtained by
ball milling sample B for 12 hours at 500 rpm. (d) Dirutile LiMnF4
∘
obtained from heating sample C for 10 hours at 300 C under argon.
Black circles again denote the impurity Li2 MnF5 phase. . . . . . . . .
2-4
53
(a) Dirutile LiMnF4 has [100] planes of corner-sharing octahedral of
a single cation species.
(b) The projection of dirutile LiMnF4 along
[101] shows empty channels formed between the edge-sharing octahedral chains.
2-5
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
54
Rietveld refinement of dirutile LiMnF4 on the P21 /c space group shows
good agreement between the experimental (red) and refined (black)
diffraction patterns with R𝑤𝑝 = 12.3% and GOF = 3.336. The peak
positions for dirutile LiMnF4 are shown by the green vertical lines, and
the difference plot is shown by the bottom blue line. . . . . . . . . . .
15
55
2-6
Rietveld refinement of the new rutile LiMnF4 polymorph on the P42 /mnm
space group shows good agreement between the experimental (red) and
refined (black) diffraction patterns with R𝑤𝑝 = 9.6% and GOF = 1.825.
The peak positions for rutile LiMnF4 are shown by the green vertical
lines, and the difference plot is shown by the bottom blue line. . . . .
2-7
56
(a) First and second galvanostatic charge-discharge curves for dirutile
LiMnF4 milled for 2 hours with carbon at C/20 and C/50. (b) First and
second galvanostatic charge-discharge curves for rutile LiMnF4 milled
for 2 hours with carbon at C/20 and C/50. . . . . . . . . . . . . . . .
2-8
58
Bright field TEM images show the order of magnitude difference in
particle size between dirutile and rutile LiMnF4 .
(a) TEM of 100-
200nm particles of dirutile LiMnF4 . (b) TEM of <30nm particles of
rutile LiMnF4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2-9
59
First and second galvanostatic charge-discharge curves for a two-phase
sample of dirutile and rutile LiMnF4 milled for 12 hours with carbon
at C/20, C/50, and C/100. . . . . . . . . . . . . . . . . . . . . . . . .
60
2-10 SAED patterns showing conversion products of (a) rutile LiMnF4 discharged to 3V (b) rutile LiMnF4 discharged to 2V (c) rutile LiMnF4
discharged to 1V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
2-11 SAED patterns showing conversion products of (a) dirutile LiMnF4
discharged to 3V (b) dirutile LiMnF4 discharged to 2V and 1V, which
are indexed to the same phases
. . . . . . . . . . . . . . . . . . . . .
63
2-12 Lithium insertion into dirutile LiMnF4 may induce a lithium site shift
from the octahedral site (outlined in orange) to the two neighboring
tetrahedral sites (outlined in blue). Partially lithiated Li(1+𝑥) MnF4 is
shown here with MnF6 octahedra represented in purple in the background and LiF6 octahedra represented in green in the foreground. . .
16
64
2-13 The Li-Mn-F phase diagram at 0K is derived from first principles.
Thermodynamically stable compounds are denoted by solid red dots.
The lithiation path of LiMnF4 is marked by the dotted line. Insertion
into LiMnF4 produces metastable Li2 MnF4 , which is denoted by the
open red dot.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
2-14 Energies of compositions between LiMnF4 and Li2 MnF4 are plotted
relative to that of dirutile LiMnF4 . Rutile polymorphs and insertion
products (Li1+𝑥 MnF4 ) are represented by blue points, and competing
conversion products at specific lithiation levels are represented by red
points. Lines connecting points are labeled with the calculated equilibrium voltages between compounds. The equilibrium voltage is linearly
related to the slope between points. . . . . . . . . . . . . . . . . . . .
66
2-15 Starting with LiMnF4 , possible reactions of fluoride compounds with
lithium are labeled with their expected onset voltages.
The lithium
fraction and specific capacity listed to the left of each box are calculated
relative to LiMnF4 .
Less plausible reactions are denoted by dashed
arrows, while more plausible reactions are denoted by solid arrows.
The gray box containing Li2 MnF4 assumes insertion and lithium site
shift.
All other boxes represent conversion products from reactions
with lithium.
3-1
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
Maximal capacity based on various assumptions: The electron-limited
capacities of the LNSO compounds are calculated from the Ni
2+/4+
redox couple, and peak at 9% lithium excess. The 0-TM capacities in
the disordered rocksalt and layered rocksalt structures are calculated as
the amount of Li in the 0-TM percolating pathway. The dotted vertical
lines mark the 0-TM percolation thresholds in these two structures,
which are at
x
= 1.09 and 1.14, respectively. The LNSO compounds
studied are marked by black squares.
17
. . . . . . . . . . . . . . . . . .
73
3-2
SEM shows the LNSO particles to be 50-100nm after ball milling with
carbon at 300rpm for variable times. (a) LNSO-0 for 6 hours (b) LNSO5 for 3 hours (c) LNSO-10 for 2 hours (d) LNSO-15 for 1 hour 40 minutes. 78
3-3
Galvanostatic cycling of LNSO-0 (no Li excess) and LNSO-15 (15% Li
excess) at 1C between 2.5–4.6V. . . . . . . . . . . . . . . . . . . . . .
3-4
79
(a) Discharge capacity of the LNSO compounds over 50 cycles at 1C
between 2.5–4.6V. (b) Fraction of theoretical capacity achieved on discharge for the LNSO compounds over 50 cycles at 1C between 2.5–4.6V. 80
√
3-5
HRTEM of pristine LNSO-15 showing the coexistence of
√
and
3×1
Li-Sb domains within the transition metal
ab
√
3× 3 Ni-Sb
layer. In (b)-
d
(d), the black crosshatch arrows label the 1/3 110 streaks corresponding
√
to the
√
3× 3
ordering, while the red striped arrows label the set of
d
1/2 110 spots. (a) HRTEM along the [1-10] zone axis. The inset shows
the intensity line profile along the direction of the white arrow for one
projected
ab
layer. The HRTEM image is enlarged in the bottom of
the inset, where
3d
and
2d
label the tripling or doubling of the
d 110
spacing. (b) Fourier transform of the region inside the dashed square
in (a).
(c) Fourier transform of the region inside the solid square in
(a). (d) Electron diffraction pattern of the particle. (e) Enlarged view
√
of
3-6
3×
√
3
ordered regions showing random interlayer stacking. . . .
81
Two types of orderings coexist in the transition metal layer of lithium-
√
√
√
√ √
3× 3 and 3×1. (a) 3× 3 honeycomb
√
5+
and Sb . (b)
3×1 stripe ordering between
excess LNSO compounds,
ordering between Ni
Li
+
and Sb
5+
2+
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
18
82
3-7
Synchrotron x-ray diffraction patterns of pristine (a) LNSO-0, (b)
LNSO-5, (c) LNSO-10, and (d) LNSO-15. All LNSO compounds form
in the layered rocksalt structure. The inserts in each pattern show the
superstructure diffraction peaks deriving from ordering in the transition metal layer. Unlabeled superstructure peaks are indexed to the
√
√
3× 3 honeycomb ordering; superstructure peaks marked by vertical
√
arrows are attributed to
3×1 stripe ordering. . . . . . . . . . . . . .
3-8
83
(a) Convex hull of formation energies for LNSO compositions with
different lithium excess contents. The stable endpoints LiNi2/3 Sb1/3 O2
(LNSO-0) and Li(Li1/2 Sb1/2 )O2 (LNSO-50) are indicated with filled
black circles. Red crosses mark metastable and unstable configurations.
(b) Sketch of the lowest-energy cation orderings in the LNSO
[110]
plane for the four intermediate compostions with unit cells outlined in
black.
3-9
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
LNSO-20 cation orderings with one, two, and three atom wide Li(Li1/2 Sb1/2 )O2
domains (shaded red).
The computed (DFT) formation energy per
Li(Li1/5 Ni2/5 Sb2/5 )O2 formula unit (f.u.)
increases with growing do-
main width. Energies are shown relative to the most stable ordering.
Unit cells are outlined in black.
. . . . . . . . . . . . . . . . . . . . .
86
3-10 Distribution of Sb-Sb nearest neighbor distances extracted from all
DFT computed LNSO-12.5 structures. A distance of
as input for fitting EXAFS data.
∼ 3.1 Å was used
. . . . . . . . . . . . . . . . . . . .
87
3-11 Raw EXAFS at the LNSO-15 Ni and Sb K-edges showing the high
quality of the data to
∼
18 Å
−1
.
. . . . . . . . . . . . . . . . . . . .
87
3-12 Fits to the LNSO-15 Sb K-edge show improvement with inclusion of the
DFT-computed Sb-Sb correlation at 3.1Å, as highlighted by the black
box. The dotted lines denote the fit range covered. (a) Fit without SbSb correlation included (𝜒𝛾
2
= 514, R-factor = 0.0048). (b) Improved
fit with Sb-Sb correlation included (𝜒𝛾
19
2
= 339, R-factor = 0.0022) . .
89
3-13 Schematic illustration showing coexistance of the two types of ordering
and 0-TM diffusion channels at the domain interfaces. For clarity, the
projected position of Li𝐿𝑖 is only shown in the interface of regions of
both domains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
in situ
4-1
Configuration of a custom
cell for x-ray diffraction . . . . . . .
4-2
Schematic of a on-line electrochemical mass spectrometer (OEMS) with
90
96
a battery cell connected directly to a mass spectrometer. All gas products evolved are continuously sampled. [6]
4-3
. . . . . . . . . . . . . . .
97
(a) From top: Galvanostatic cycling of LNSO-0, LNSO-5, LNSO-10,
and LNSO-15 at C/20 between 2.5–4.6V. (b) Fraction of theoretical
capacity achieved on discharge over ten cycles at C/20 between 2.5–
4.6V. Capacity retention improves with increasing lithium content, and
all lithium excess samples access a higher fraction of their theoretical
capacities than LNSO-0.
4-4
. . . . . . . . . . . . . . . . . . . . . . . . .
98
(a) First cycle voltage curves of LNSO-15 obtained from galvanostatic
cycling between C/20 and 5C. At faster C-rates, the 4.45V charge
plateau disappears, and higher discharge capacities are achieved. (b)
Discharge capacity over ten cycles for LNSO-15 at rates between C/20
and 5C. Capacity increases increases from C/20 to 1C, then decreases
between 1C and 5C.
4-5
. . . . . . . . . . . . . . . . . . . . . . . . . . .
100
GITT measurements on LNSO-15 show a low overpotential along the
first charge plateau, and a larger overpotential along the second charge
plateau.
The 4.45V overpotential of the second plateau agrees well
with voltage curves obained by galvanostatic cycling. The larger overpotential along the second plateau implies a slower process at 4.45V.
4-6
102
Galvanostatic cycling with the first charge conducted at 1C, and first
discharge at C/20.
All subsequent cycles are at C/20.
The 4.45V
plateau appears on the second charge. . . . . . . . . . . . . . . . . . .
20
102
4-7
(a) Galvanostatic cycling of LNSO-15 at 1C for the first twenty cycles.
On the 21st cycle, the charge current is reduced to C/20, and a second
plateau is observed at 4.6V. The second plateau is observed only on
the first slow charge, regardless of previous time spent at high voltages
during cycling. (b) Charge and discharge capacity versus cycle number. 103
4-8
Voltage curves and
ex situ
XANES spectra of the nickel K-edge for
pristine and fully-charged LNSO-15 samples charged at 1C and C/20.
2+
3+
Reference edge positions for Ni
and Ni
are obtained from NiO
and NaNiO2 .
The edge positions of the charged LNSO-15 samples
show that for both rates, nickel is oxidized to a similar valence state
between Ni
4-9
2+
and Ni
3+
.
. . . . . . . . . . . . . . . . . . . . . . . . .
Voltage curves and L3/L2 ratio of the nickel K-edge obtained from
situ
104
ex-
EELS measurements for pristine, half-charged, and fully-charged
LNSO-15 samples charged at 1C and C/20. Reference L3/L2 ratios for
Ni
2+
and Ni
3+
are represented by horizontal bars. At the end of charge
for both rates, nickel is oxidized to a similar valence state between Ni
and Ni
4-10
3+
In situ
.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
XRD of LNSO-15 at C/20.
a -lattice
2+
parameter evolution, and
106
(a) First charge voltage curve,
c -lattice
parameter evolution. (b)
(003) and (104) peak evolution over the first charge. . . . . . . . . . .
4-11 Voltage curves and gas evolution measurements from
in situ
107
OEMS of
(a) LNSO-0 and (b) LNSO-15, both charged at C/20. Only LNSO-15
shows oxygen evolution starting at 4.4V. The O2 signal is multiplied
by 10x to be plotted on the same scale as the CO2 and CO signals.
.
109
4-12 HRTEM comparing bulk and surface structures of LNSO-15 after cycling at 1C and C/20 . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
110
22
List of Tables
3.1
Lattice parameters, lithium slab spacing, and cation mixing levels of
pristine Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds determined by Rietveld refinement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2
74
EXAFS structural parameters (coordination number, CN; correlation
distance, R; and mean-squared relative displacement,
𝜎 2 ) for LNSO-15
extracted by fitting the first two peaks of the Fourier transforms of the
Ni K-edge and Sb K-edge. The many body amplitude reduction factor
2
(S0 ) was fixed to be 0.83(4) for Ni and 1.03(3) for Sb; these values were
determined by fits to stoichiometric LNSO-0.
Possible Ni-Li antisite
disorder was not explicitly included. The number(s) in the parenthesis
is the uncertainty in the final digit(s).
The Ni and Sb EXAFS data
were simultaneously fit. . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1
88
Theoretical capacities of Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds calculated from
the Ni
2+/4+
redox couple. . . . . . . . . . . . . . . . . . . . . . . . . .
23
97
24
Chapter 1
Introduction
1.1 Motivation
As the world’s demand for energy continues to grow, it is clear that the limited supply of non-renewable fossil fuels will not satisfy our long-term needs. Replacing our
dependence on fossil fuels with renewable energy resources requires significant advances in energy storage, as energy converters such as wind turbines and photovoltaic
cells rely on intermittent resources [7]. Future applications will require batteries with
even higher energy densities,
e.g.
more energy per volume or mass.
Lithium-ion
has emerged as the frontrunner battery technology due to superior volumetric and
gravimetric energy density [1], as shown in Figure 1-1.
However, despite commer-
cial success in electronics and power tools, the deployment of lithium-ion batteries
to broader markets is delayed by several factors.
Current cathode chemistries are
unable to satisfy all desired characteristics for target applications, including energy
density, rate performance, safety, cycle life, cost, and toxicity. It is clear that new
materials are necessary to enable future applications in grid storage and long-range
electric vehicles [8, 9].
25
Figure 1-1: Comparison of different battery technologies in terms of volumetric and
gravimetric energy density. [1]
1.2 Overview of rechargeable batteries
1.2.1 Configuration and operation of a battery
Figure 1-2 shows a schematic of a rechargable lithium battery, which has three main
components – anode, cathode, and electrolyte. The anode and cathode are the negative and positive electrodes, respectively, where oxidation and reduction processes
occur upon charge and discharge. The electrolyte is an ionic conductor but electronic
insulator, allowing ions to shuttle between the anode and the cathode. Connected to
the outside of the electrodes are current collectors, which are electronic conductors but
ionic insulators. The current collectors are connected by an external circuit through
which electrons can travel. During discharge, the spontaneous reaction, lithium ions
and electrons separate at the anode and travel separately through electrolyte and
external circuit to the cathode, reducing the oxidation state of the cathode.
26
The
opposite reaction, charge, uses electrical work to drive the lithium ions and electrons
from cathode to anode. The separation of pathways for the ions and electrons enables
us to convert chemical energy to electric energy.
Discharge
e-
e-
V
Charge
Li
Li+
Li+
Li+
Li+
Li+
Li+
Anode
Current collector
Current collector
Discharge
Charge
Electrolyte
Cathode
Figure 1-2: Schematic of a rechargeable lithium battery. Upon charge, Li
+
−
and e
move from the cathode to the anode through the electrolyte and external circuit,
respectively. Upon discharge, the reverse process occurs.
In lithium-ion batteries, graphite typically serves as the anode, intercalating lithium
ions between its sheets to form LiC6 [10]. The cathode contains not only the active
material, but also carbon to enhance conductivity and a polymeric binder such as
polytetrafluorethylene (PTFE) or polyvinylidene fluoride (PVDF) to hold the film
together. Few active materials – layered oxides, spinels, and olivines – have passed
the minimum requirements for voltage and capacity to achieve commercialization [11].
Layered oxides of the LiMO2 formula are the most ubiquitous cathodes. LiCoO2 is
attractive for its high voltage, but is limited by the cost of cobalt and that only 0.5
lithium per formula unit can be reversibly removed from the layered structure during
cycling. Other common layered materials derive from LiCoO2 ; substitution of cobalt
with other transition metals such as nickel, manganese, or aluminum stabilizes the
layered structure and allows for higher capacities to be achieved. Spinel LiMn2 O4 is
27
interesting for its high-rate performance, but has a large step in its voltage profile and
has lower capacity.
Olivine LiFePO4 is attractive for its low cost, environmentally
friendly elements, and safety, but has a relatively lower voltage and only moderate
capacity.
Commercial electrolytes consist of a lithium salt and organic solvent. Electrolytes
must have high ionic conductivity, high chemical stability versus the electrodes, electrochemical stability in a large voltage window, low melting point, high boiling point,
and be non-toxic and low cost [12]. The most common salts are LiPF6 , LiBF4 , and
LiClO4 [13], which have good solubility and charge separation in solution, due to the
bulky nature of the anion groups [14]. For the solvent, alkyl carbonates are known to
be the most suitable for lithium-ion batteries [15]. Binary solvent mixtures combining ethylene carbonate (EC) with dimethyl carbonate (DMC), ethyl methyl carbonate
(EMC), or diethyl carbonate (DEC) are popular. The combination of two solvents
with different properties results in an electrolytes with large electrochemical windows
and optimized ionic conductivity deriving from low viscosity, high polarity, and high
relative permitivity [10].
Gravimetric energy density is the primary metric for new battery chemistries, and
is calculated as the product of cell voltage (V) and specific capacity (mAh/g). A high
voltage indicates a larger driving force for reaction. The open circuit voltage of a cell
is calculated by
𝑉 𝑜𝑙𝑡𝑎𝑔𝑒 =
where
𝜇𝑐𝐿𝑖 −𝜇𝑎𝐿𝑖
and anode,
F
z
−(𝜇𝑐𝐿𝑖 − 𝜇𝑎𝐿𝑖 )
𝑧𝐹
is the difference in lithium chemical potential between the cathode
is the charge associated with lithium displacement from the anode, and
is Faraday’s constant (96485 C/mol). Thus, to obtain a high voltage to increase
energy density, the difference in lithium chemical potentials should be large.
However, we cannot choose a cathode material of infinitely high voltage.
As
seen in Figure 1-3, the open circuit voltage is restricted by the electrochemical window of the electrolyte, defined as the energy difference between the highest occupied
28
Electron Energy
LUMO
Anode
μ
Electrolyte
HOMO
μ e-
Cathode
Figure 1-3: Schematic open circuit diagram for a thermodynamically stable battery.
molecular orbitals (HOMO) and lowest unoccupied molecular orbitals (LUMO). A
chemical potential of either electrode lying beyond this electrochemical window results in either oxidation or reduction of the electrolyte to lower the energy at the
electrode/electrolyte interface [16].
The second component of the energy density product is capacity, which is determined by how much lithium can be reversibly intercalated between the anode and
cathode. To date, the overall energy density of lithium-ion batteries remains limited
by the capacity of the cathode. The specific capacity, measured in mAh/g, can be
calculated by
𝑆𝑝𝑒𝑐𝑖𝑓 𝑖𝑐 𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦 =
where
z
is the charge associated with lithium displacement from the anode,
Faraday’s constant (96485 C/mol), and
in kg/mol.
𝑧𝐹
3600𝑀
M
F
is
in the molecular weight of the compound
Depending on the cathode of the rechargeable battery, high capacities
can be achieved through conversion (Section 1.2.3) or intercalation (Section 1.2.4)
reaction mechanisms.
29
1.2.2 Important properties of rechargeable batteries
While high energy density is often cited as the key commercialization driver for new
cathode materials, there are many other factors to consider, namely rate capability,
safety, cycle life, cost, and toxicity.
The rate capability of a material refers to the time required for a battery to charge
and discharge. This metric is especially critical for applications such as electric vehicles which require high power during acceleration or regenerative braking.
Rate
capability is determined by both the electronic conductivity and lithium diffusivity of
the cathode. Common strategies to enhance electronic conductivity through processing include finely mixing the cathode active material with a conductive agent such as
carbon, or carbon coating individual particles [17]. The intrinsic electronic conductivity of a material can also be improved by doping. In the case of LiFePO4 , a material
known to have low conductivity, doping on the lithium site was shown to improve
8
electronic conductivity by a factor of 10 by inducing p-type conductivity [18].
Lithium diffusivity is more commonly the limiting factor on rate capability. For
intercalation to proceed, lithium ions must be able to diffuse from the inside of a
particle to the surface. Thus, smaller particles typically show better rate capability
than larger particles, as lithium needs to diffuse over shorter distances. At an intrinsic
level, practical lithium diffusivity is determined by diffusion barriers, dimensionality
of diffusion, and defects.
Diffusion barriers are the activation barrier for lithium to move from one site to
another, and are largely determined by the spacing of the diffusion channel and the
electrostatic repulsion forces exerted by neighboring ions on the diffusing lithium ion.
Diffusion barriers for the layered structure will be described in more detail in Section
1.3.
The dimensionality of diffusion depends on the crystal structure.
Shown in
Figure 1-4 are three common crystal structures of cathode materials: layered, spinel,
and olivine.
Layered materials show 2D lithium diffusion within the lithium layer,
spinel materials exhibit 3D diffusion in all three directions of the crystal structure,
and olivine materials are limited to 1D diffusion through lithium channels. Lithium
30
diffusion in all three crystal structures has been shown to be facile, barring defects.
Defects are particularly problematic in materials limited to 1D diffusion, where a
single point defect can block an entire diffusion channel (Figure 1-5) [3].
Figure 1-4:
batteries:
Representative crystal structures of cathode materials for lithium-ion
(a) layered
𝛼-LiCoO2 ;
(b) cubic LiMn2 O4 spinel; (c) olivine-structured
LiFePO4 . Li ions are shown as light green spheres, CoO6 octahedra in blue, MnO6
octahedra in mauve, Fe–O polyhedra in brown, and PO4 tetrahedra in purple. [2]
Safety concerns can refer to a number of problems encountered in rechargeable
batteries. On the anode side, safety concerns preclude lithium metal from being used
as an anode, as continuous cycling leads to dendritic growth on the metal which can
short-circuit a battery. On the cathode side, safety typically refers to the propensity
of a material to thermally degrade upon charge. Often, this decomposition reaction
evolves oxygen gas, which can react with the electrolyte and lead to thermal runaway
[19, 20].
Cycle life is defined as the number of charge-discharge cycles a battery can sustain
before specific capacity falls below some percentage of the initial capacity. While the
cycle life of current chemistries can satisfy the 3–4 year lifetimes of portable electronics, electric vehicles and stationary grid storage will require service over 10+ year
lifetimes and likely undergo more than 10,000 cycles [21].
The cycle life metric is
closely tied to coulombic efficiency, which measures the capacity loss between a single
charge-discharge cycle.
Undesired side reactions such as formation of a solid elec-
trolyte interface (SEI) and electrolyte decomposition are known to lower coulombic
31
+
Figure 1-5: Schematic illustration of Li diffusion impeded by immobile point defects
in 1D channels. [3]
efficiency.
Other permanent structure changes such as phase transformations, and
oxygen evolution can also lower overall cycle life.
Finally, cost and toxicity go hand in hand with materials choice. Ideally, cathode
materials should be low cost and non-toxic. The search for alternative cathodes to
LiCoO2 , a high performance material, stems from cobalt’s high cost and moderate
toxicity. Thus, first row transition metals such as nickel, iron, and manganese are attractive due to their lower cost, higher abundance, and more environmentally-friendly
nature.
1.2.3 Conversion cathodes
In conversion reactions, cathode materials are synthesized in the charged state. Upon
discharge, multiple lithium atoms react with the cathode to utilize all the redox states
of the transition metal and fully reduce the metal. The host structure is not conserved;
instead, it converts to a multitude of other phases upon reaction with lithium. The
overall reactions are as follows
MY𝑥 +
x Li(s) → x LiY + M (discharge )
x LiY + M → MY𝑥
+
x Li(s) (charge )
32
Of the conversion materials explored to date, fluoride materials have shown the
most promise in lithium batteries, reaching 600-700 mAh/g [22, 23] on discharge. The
status of fluoride cathodes will be discussed in more detail in Section 2.1.
Other
batteries that operate by conversion reactions are lithium-sulfur and lithium-oxygen
batteries, which have theoretical capacities of 1672 mAh/g and 3840 mAh/g, respectively. These conversion theoretical capacities far exceed the highest demonstrated
intercalation capacities of 250-300 mAh/g.
However, conversion materials have many significant challenges to overcome before
commercialization, including reversibility of reactions, kinetics, and voltage hystresis
[24].
Because conversion requires new phases to form upon charge and discharge,
the stability of intermediate and final phases may preclude full reversibility. Kinetics
of conversion reactions are also much slower than intercalation, requiring transition
metal migration and nucleation of metal particles.
Finally, because overpotentials
may be necessary to drive reactions, or because different reaction pathways may be
accessed on charge and discharge, significant hystresis is often observed in the voltage
curves of conversion materials, leading to round-trip energy losses.
1.2.4 Intercalation cathodes
Intercalation is defined as the reversible insertion and extraction of an ion into a
crystal structure. In the case of lithium-ion batteries, lithium ions are intercalated
between the anode and cathode, whose structures remain largely unchanged.
The
cathode contains a transition metal that is oxidized upon lithium extraction and
reduced upon lithium insertion. Materials containing lightweight elements and open
crystal structures that allow for easy diffusion of lithium ions are desired for high
capacity cathodes. As discussed in Section 1.2.1, only a few materials have achieved
commercial success as cathodes.
One of the most established classes of cathode materials is the layered lithium
transition metal oxides, LiMO2 , where M is a transition metal with a 3
state.
The layered structure has a hexagonal unit cell with
R -3m
+
valence
symmetry, and
is defined by a close-packed oxygen sublattice in which layers of lithium ions and
33
layers of transition metal ions alternate in stacking. Layered cathode materials are
synthesized in the discharged state. The reactions upon charge and discharge are as
follows
LiMO2
→ x Li
x Li + x e−
+
x e−
charge )
+ Li1−𝑥 MO2 (
+ Li1−𝑥 MO2
→
discharge )
LiMO2 (
Intercalation cathodes tend to suffer from structural or chemical instability if too
much lithium is extracted upon cycling. For example, less than 0.5Li can be extracted
from LiCoO2 before it becomes prone to oxygen release and irreversible structure
changes which severely degrade safety and cycle life of the battery. One strategy taken
by researchers to increase stability and reversible capacity is transition metal substitution. This pathway has led to the development of several commercial chemistries,
including LiNi1/3 Co1/3 Al1/3 O2 [25], LiNi0.5 Mn0.5 O2 [26], LiNi1/3 Mn1/3 Co1/3 O2 [27].
Layered compounds that have some of their transition metal substituted with
lithium are known as lithium-rich, or lithium-excess materials (Li1+𝑥 M1−𝑥 O2 ).
To
date, the highest reversible capacities in lithium-ion chemistries have been achieved in
lithium-excess transition metal oxides [5, 28–31]. The most promising of the lithium-
2+
3+
4+
excess materials commonly contain Ni , Co , and Mn , and form as layered single
phase solid solutions [32, 33] or two-phase materials containing LiMO2 and Li2 MnO3
domains [29, 34–36].
While these materials have long been of interest given their
high cycling capacities, they still require improvement in rate capability [34, 37–40],
as well as elucidation of the first charge activation processes and their consequences
[30, 41–43]. Upon first charge, after oxidation of nickel and cobalt to 4
+
valence states
[44], lithium-excess materials often exhibit oxygen loss accompanied by structure
reorganization [42, 45–48].
Upon discharge, nickel, cobalt,
and
manganese are all
reduced and contribute to capacity [30, 49, 50]. After a decade of research exploring
manganese-containing lithium-excess layered materials, the mechanisms responsible
for their promising characteristics remain unclear [51].
34
1.3 Diffusion in layered oxides
Designing and engineering cathode materials to achieve maximum reversible capacity
requires clear understanding of lithium diffusion mechanisms. In the layered LiMO2
structure, lithium is extracted from the lithium layer during charge, and reinserted
into the lithium layer during discharge (Figure 1-6). Within the lithium layer, lithium
diffuses by a di-vacancy mechanism [52, 53] from one octahedral site to an empty,
neighboring octahedral site through a tetrahedral site.
lithium diffusion channel.
These three sites define a
As shown in Figure 1-7, the tetrahedral site face-shares
with four octahedral sites: the initial and final lithium sites (
k
sites (
and
l ) [4].
i
and
j ),
and two gate
The lithium sites are in the lithium layer, while the gate sites are
part of the transition metal layer.
Figure 1-6: Delithiation of a layered LiMO2 cathode during charge occurs by extraction of lithium from the lithium layers.
The intermediate tetrahedral site is considered an activated state, as it defines the
energy barrier for lithium to move through the diffusion channel. The height of the
energy barrier is primarily affected by two factors: lithium slab spacing and transition
metal valence [54]. With larger slab spacing, such as in Figure 1-8a, there is more
space for lithium to diffuse within and consequently lower lithium diffusion barriers.
Transition metal valence is also significant as it determines the electrostatic repulsion exerted by the gates sites on the tetrahedral site. In layered LiMO2 compounds
containing a single M
3+
transition metal, all lithium diffusion barriers are the same.
If layered LiMO2 compounds contain multiple transition metals, charge balance can
35
Figure 1-7: Schematics of diffusion from octahedral lithium site
view in rocksalt (111)-direction.
i
to site
j.
a) Top
Large empty circles denote octahedral sites.
The
i and j. Only
k, is depicted. b) 3D view of the same diffusion channel along with
both gate sites, k and l. The di-vacancy lithium diffusion mechanism requires one of
arrow indicates one of two lithium diffusion channels that connets sites
one gate site,
the gate sites to be vacant. [4]
Figure 1-8: Diffusion in layered materials is largely affected by (a) lithium slab spacing
and (b) transition metal valence.
also be achieved with metals of different valence states as long as the average valence
+
2+
4+
state is 3 . For example, in Figure 1-8b, both M
and M
ions occupy the transition metal layer. Here, diffusion channels containing gate sites with lower valence
2+
transition metals (M ) will have lower diffusion barriers than gate sites filled by high
36
4+
4+
valence transition metals (M ), as the M
ions exert a stronger repulsive force on
lithium.
1.4 Lithium-excess and 0-transition metal (0-TM)
diffusion channels
Figure 1-9: Comparison of diffusion pathways in stoichiometric LiMO2 and lithiumexcess Li1+𝑥 M1−𝑥 O2 .
The lithium diffusion mechanism is the same in layered LiMO2 and layered lithiumexcess Li1+𝑥 M1−𝑥 O2 compounds.
occupancy of the gate sites.
The difference in lithium-excess materials is the
As shown in Figure 1-9, some sites in the transition
metal layer are now occupied by lithium, meaning some of the gate sites neighboring
the activated tetrahedral state are occupied by lithium instead of transition metal.
+
The presence of Li in the gate site, compared to higher valence transition metals,
results in a much lower diffusion barrier.
Classification of diffusion channels based on gate site occupancy was established
by Lee et al. [5], and is summarized in Figure 1-10. As shown in Figure 1-10a, there
are two diffusion channels available for a lithium to traverse from one octahedral site
to another, both passing through an activated tetrahedral state. Figures 1-10b-d show
the possible configurations of the activated tetrahedral state, which are named based
37
on the occupancy of the gate sites. In the
𝛾 -LiFeO2
structure, both of the gate sites
are occupied by transition metals, forming a two transition metal (2-TM) channel
(Figure 1-10d). Layered LiMO2 compounds contain only one transition metal in a
gate site, forming one transition metal (1-TM) channels (Figure 1-10c). With layered
lithium-excess LiM1+𝑥 M1−𝑥 O2 compounds, the presence of lithium in the transition
metal layer makes the zero transition metal (0-TM) configuration possible, as shown
in Figure 1-10b. In 0-TM channels, diffusion barriers remain low even if lithium slab
spacing is reduced [5].
Figure 1-10: Possible environments for an
o-t-o hop in rocksalt-like Li-TM oxides.
(A)
o-t-o )
diffusion: Two tetrahedral paths connect
each pair of neighboring octahedral sites.
(B to D) The activated state can share
octahedral–tetrahedral–octahedral (
faces with no octahedral transition metals (0-TM channel) (B), one transition metal
(1-TM channel) (C), or two transition metals (2-TM channel) (D). [5]
38
Figure 1-11: 0-TM accessible lithium atoms per formula unit as a function of the
overall lithium content and the degree of cation mixing in the layered structure.
The percolation threshold is indicated by the thick black contour line. Compositions
falling into the region left of the contour line are not 0-TM percolating. The thin line
indicates the composition at which one lithium atom per formula unit becomes 0-TM
accessible. [4]
The significance of 0-TM channels is further established in the context of percolation theory. While a single 0-TM channel may have a low diffusion barrier, it will not
have an appreciable effect on macroscopic diffusion unless there is a critical number
of 0-TM channels percolating across a particle. This critical number is defined as the
percolation threshold. In a perfectly layered structure, the percolation threshold is
calculated to occur at 14% lithium excess level (Figure 1-11). At higher lithium excess levels, more lithium become part of the percolating 0-TM network,
e.g.
become
0-TM accessible. These values have been calculated for various cation orderings, including the layered, spinel-like, and
𝛾 -LiFeO2
structures [4]. Figure 1-11 shows the
percolation map for the layered structure as an example. The color mapping given
by the right scalebar indicates the amount of lithium which are 0-TM accessible, the
39
thick black contour line indicates the percolation threshold, and the thin black line
indicates the composition at which one lithium per formula unit becomes 0-TM accessible.
x
These values change as a function of overall lithium content ( -axis) and
y
degree of cation mixing ( -axis).
The percolation threshold marked in Figure 1-11 assumes individual 0-TM channels to distribute homogenously across a particle. While high lithium excess levels
are necessary to achieve percolation of the low diffusion barrier 0-TM channels, the
tradeoff of high lithium content is reduced transition metal content, which dictates
the capacity of the cathode. However, if 0-TM channels can be segregated into high
aspect ratio domains, such as stripes, percolation can be achieved at lower lithium
content. Percolation of stripe domains is shown schematically in Figure 1-12.
Figure 1-12: 0-TM percolation requires individual 0-TM channels to connect across
a particle. Segregation of 0-TM channels into high aspect ratio domains lowers the
number of 0-TM channels necessary for percolation.
1.5 Overview of the thesis
This thesis explores the potential of novel fluoride and oxide chemistries as cathode
materials for rechargeable batteries, and characterizes structural changes and redox
processes during their operation.
The first part of the thesis covers synthesis of
new fluoride conversion cathodes and characterization of their lithiation mechanisms.
The second part of the thesis discusses the design and structure of new lithium-excess
40
nickel antimony oxides from the concepts of 0-TM diffusion channels and percolation
theory, and proposes an explanation for the anomalous rate behavior from characterization of redox and structure.
Chapter 2 covers synthesis, electrochemical performance, and characterization of
conversion processes of two fluoride polymorphs, dirutile and rutile LiMnF4 . Chap-
ter 3 establishes the concept of 0-TM percolation in the design of lithium-excess
Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 , and confirms a uniquely patterned microstructure containing 0TM nanohighways. Chapter 4 discusses the effect of lithium content on electrochemical performance in lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 , and characterizes the active
redox processes at different rates. Chapter 5 concludes the thesis.
41
42
Chapter 2
Dirutile and Rutile LiMnF4: Two
New Conversion Cathodes
2.1 Fluorides and oxyfluorides as conversion cathode
materials
2.1.1 Introduction
Of the conversion cathode materials considered to date, fluorides are the most attractive due to their high voltages. The electrochemical activity of the binary metal
fluorides FeF3 , TiF3 , VF3 , and MnF3 was first reported by Arai et al. [55].
De-
spite having high theoretical specific capacities of greater than 200 mAh/g with one
lithium, only 80 mAh/g could be reversibly cycled in the former three compounds,
and no discharge capacity was observed for MnF3 . This illustrates one of the main
challenges in fluoride systems: while the electronegativity of fluorine gives rise to high
voltages, fluorides also tend to be insulators.
Badway et al.
later demonstrated a means of compensating for the inherently
low conductivity of fluoride cathodes. By aggressively mixing FeF3 with carbon to
obtain sub-30nm particles and form carbon metal fluoride nanocomposites, much
higher capacities could be realized [22, 56]. FeF3 reached a specific capacity of 200
43
3+
2+
mAh/g at a 2V cutoff by accessing the Fe /Fe
redox couple, and greater than 600
2+
0
mAh/g at a 1.5V cutoff when further reducing Fe
to Fe . The high specific capacity
achieved in the FeF3 system prompted a series of studies on other metal fluoride
and oxyfluoride systems, including FeF2 [57–59], FeOF [60, 61], CuF2 [58, 62], BiOF
[63], and BiF3 [64]. Research efforts on conversion materials include optimization of
particle morphology for improved performance [65], as well as elucidation of reaction
paths through computation [66,67] and
in situ TEM studies [58,59].
Despite successes
in these efforts, there continues to be a need for new, high energy density materials.
High-throughput computations have emerged as an attractive way to identify
promising new materials for cathodes [68]. The Materials Project runs calculations
across compounds known in nature and that are computationally designed, notably
looking at material stability, voltage, and capacity. From these screening criteria, a
number of candidate fluoride and oxyfluoride materials were identified as high energy density materials. In Sections 2.2–2.8, one of these materials, LiMnF4 , will be
discussed in detail.
2.1.2 Advantages
The fluoride and oxyfluoride materials represent a relatively untapped opportunity in
battery materials [69]. Few fluoride compounds have been reported in the literature
as intercalation materials [69–73].
More common is the use of binary fluorides as
conversion cathode materials [22, 24, 57, 58, 62, 63, 74, 75]. Unlike intercalation compounds, which maintain their crystal structure as lithium is inserted and extracted,
conversion materials react with lithium in the reaction
MF𝑥 + xLi(s)
→
xLiF + M
While much higher capacities are achieved by utilizing all available redox states of
the transition metal, the conversion reaction occurs much more slowly than intercalation and is less reversible. Some work has also been done with the fluoro-polyanion
systems [76–79], and small amounts of fluorine-doping [80–82], where fluorine has been
seen to stabilize the structure, raise voltage, or enhance electrochemical performance.
The high voltage of fluoride and oxyfluoride materials is generally attributed to
44
the high ionicity in the fluorine bonds.
As discussed in Section 1.2.1, voltage is
calculated as the difference in the chemical potential of lithium between the anode
and the cathode. This chemical potential term can be broken down into
chemical potential of lithium ions, and
the effect of
𝜇𝐿𝑖+
𝜇𝐿𝑖+ ,
𝜇𝑒− , the chemical potential of electrons.
the
While
on voltage is not well understood, the higher voltage of fluorides
can be explained by the electronic term.
deltaE (oxide)
deltaE (fluoride)
Energy
antibonding
molecular orbital
Metal
Metal
antibonding
molecular orbital
Oxygen
Fluorine
bonding
molecular orbital
bonding
molecular orbital
Figure 2-1:
Comparison of relative energies of bonding and antibonding molecular
orbitals in transition metal oxides (left) and transition metal fluorides (right)
Figure 2-1 illustrates the difference in energy levels between fluorides and less
electronegative elements, such as oxides.
Given a transition metal and ligand, the
higher antibonding level is primarily of transition metal character, while the lower
bonding level is primary of the ligand character. Comparing fluorides to oxides, there
is more overlap in the density of states between the transition metal d-bands and
oxygen p-bands, resulting in a more covalent bond and greater separation between
the bonding and antibonding energy levels. In fluorides, there typically is no overlap
between transition metal d-bands and fluoride p-bands, resulting in a very ionic bond
and less separation between the bonding and antibonding energy levels. Thus, because
of the lower position of the antibonding energy level in fluorides, more energy is
required to remove an electron, resulting in higher voltage compounds.
45
2.1.3 Synthesis challenges
Pure fluorides and oxyfluorides are known to be difficult to prepare. Historically, synthesis of fluorides was achieved through relatively complicated or dangerous routes
utilizing flowing F2 or HF gas, or sealed gold or platinum ampoules [83–86]. Both
of these methods require specialized setups and expensive equipment.
Recent ad-
vances in synthetic chemistry have shown success with much simpler methods such
as precipitation, hydrothermal, and mechanochemical synthesis.
Precipitation at room temperature or low temperature conditions are common,
avoiding the energy-intensive firing step of solid state reactions, and no special equipment is necessary. Synthesis of fluoride compounds via precipitation reactions typically requires an acidic solution and fluorine source. The oldest and most commonly
used solvent is hydrofluoric acid (HF) [84]. Some fluoride compounds successfully prepared via precipitation include CaF2 nanoparticles [87], Li3 FeF6 [72, 73], FeOF [57],
and VO2 F [88].
Hydrothermal reactions, which employ high pressure and high temperature in a
closed system, have been particularly successful as a means to explore new chemistries
and synthesize new fluoride compounds. Of particular interest is the work on vanadium and molybdenum fluorides by Aldous et al. [89–91], vanadium and molybdenum framework materials by Kotsapa [92], vanadium fluorides and oxyfluorides by
DeBurgomaster et al. [93], and fluoride cathode materials by Gocheva et al [71].
The main advantages of hydrothermal reactions over solid state processing are the increased mobility of the reagents in a liquid phase and the greater likelihood of forming
metastable products with special structures or properties. This makes hydrothermal
a good choice as a route to explore new chemistries.
Variations on hydrothermal reactions are solvothermal and ionothermal reactions,
which employ non-aqueous solutions and ionic liquids as solvents, respectively. Other
than the difference in solvent, these reactions operate on the same principles as hydrothermal reactions. Non-aqueous solvents are useful when products are moisture
sensitive. Ionic liquids have the advantage that the nature of the solvent can be var-
46
ied greatly with the choice of cation and anion. The ionic liquid takes a dual role of
being a structure directing agent, making ionothermal synthesis applicable to many
materials, including fluorides [94, 95].
In addition, ionothermal synthesis has been
seen to stabilize higher oxidation states that are hard to access with other methods.
Solid state synthesis typically involves a ball-milling step to reduce precursor particle size, followed by a high-temperature heat treatment to activate the reaction.
Mechanochemical synthesis differs from solid state in that chemical reaction proceeds
during ball milling due to the input of significant mechanical energy.
High-energy
milling causes a) reduction in particle size and increased surface area, b) intimate
precursor mixing, c) introduction of defects and eventual amorphization of the material, and d) frictional heating from impact of the milling media [96]. Sintering after
mechanochemical synthesis is optional depending on the degree of crystallinity desired.
This technique has been successful in the synthesis of NaMF3 (M=Fe, Mn,
Ni) [97], Cu2 PO4 F [98], and AgMo(O3 F3 ) [99]. Additionally, this method is advantageous as it produces particles on the 100nm scale. In insulating materials, small
particle size is critical to shortening the paths for ion diffusion and electron conduction
to enable electrochemical activity.
2.1.4 Other challenges
Fluorides tend to be moisture sensitive and air sensitive, requiring manipulation in a
dry environment. This requires much more care in their preparation and handling,
but given the necessity of similar conditions for the electrolyte and lithium metal, this
does not rule out fluorides as a viable cathode material. However, some fluorines and
volatile fluorides are corrosive and highly toxic, and require extra safety precautions
in handling. All experiments should be carried out under perfectly ventilated hoods.
In the context of cathode materials, a major challenge for fluorides is their insulating character. Fluorine’s high electronegativity results in strongly ionic bonds and
large bandgaps. This would appear to rule out the viability of fluorides as cathode
materials, as both high ionic conductivity and electronic conductivity are necessary to
operate a battery. However, clever strategies have been employed by other researchers
47
to overcome this same limitation in other materials. These include reducing the particles to the nanoscale, thereby reducing diffusion lengths and conduction paths, carbon
coating the particles to provide a continuous conductive matrix, and doping materials to intrinsically improve the electronic conductivity.
All of these methods can
be employed in fluoride cathode materials and are likely necessary to achieve good
electrochemical performance.
2.2 Methods
2.2.1 Experimental characterization
The X-ray diffraction (XRD) patterns in Figure 2-2 were collected on a Rigaku RU300
∘
Cr-source diffractometer with step scans between 10-120 2𝜃 , and the XRD patterns
in Figure 2-5 and Figure 2-6 were collected on a Bruker D8 Advance Da Vinci Mo-
∘
source diffractometer with step scans between 5-30 2𝜃 . These patterns are all converted to the 2𝜃 range corresponding to a Cu XRD source. All samples were sealed
with Kapton film to avoid air exposure. Powder diffraction data in Figure 2-3 was
collected in capillaries at Brookhaven National Laboratory NSLS, beamline x14A (𝜆
= 7.75756Å). Rietveld refinement and profile matching of the powder diffraction data
were performed using PANalytical High Score Plus.
Transmission electron microscopy (TEM) samples were prepared in an argon filled
glovebox with O2 and H2 O levels lower than 0.1ppm. TEM samples of pristine rutile
and dirutile LiMnF4 powder were prepared by the dry method to minimize artifacts.
The TEM samples for the pristine and discharged cathode films were prepared by a
gentle sonication process in anhydrous dimethyl carbonate (DMC) with 20ppm H2 O.
TEM samples were sealed in an airtight bottle in the glovebox and then transferred
into the TEM vacuum column immediately after the sample bottle was opened. Approximately 30 Selected Area Electron Diffraction (SAED) patterns for each type of
sample were obtained from a JEOL 2010F Transmission Electron Microscope and
indexed with PolyCrystalline Electron Diffraction Pattern (PCED2.0).
48
2.2.2 Electrochemistry
To maintain single phase samples of dirutile or rutile LiMnF4 , the active compound
was milled with carbon black (super P) in a Retsch PM200 planetary ball mill for 2
hours at 500rpm. XRD confirmed no dirutile-to-rutile phase transitions to occur after
the two hour milling cycle with carbon.
A two-phase sample of dirutile and rutile
LiMnF4 was obtained by milling dirutile LiMnF4 for 12 hours at 500rpm with carbon.
All milling was performed under inert atmostphere by assembling and reopening ball
mill jars in an argon-filled glovebox.
The mixture of active material and carbon was then hand-mixed with polyethylenetetrafluoride (PTFE) binder in a ratio of 45:45:10 by weight, and rolled into a film.
Swagelok cells were assembled in an argon-filled glovebox using lithium metal, Celgard C480 separator, and 1M LiPF6 in 1:1 ethylene carbonate:dimethyl carbonate
(EC:DMC) solution. Cells were also tested using 1M LiClO4 in propylene carbonate
(PC) as an alternative electrolyte, given the observations of fluoride instability in the
presence of trace amounts of HF from LiPF6 dissolution [63]. No noticeable differences in electrochemical performance were observed between the LiMnF4 cells tested
with different electrolytes.
All cells were tested on a Maccor 2200 or Maccor 4000
operating at room temperature.
2.2.3 Computations
All computations were performed using density functional theory (DFT) using the
generalized gradient approximation with a Hubbard-like U correction (GGA+U)
[100, 101].
The U value for the
d
states of Mn was set to 3.9 as determined by
fitting experimental oxidation energies [102]. The Vienna ab initio Simulation Package (VASP) and the included projector-augmented pseudopotentials were used for
all calculations. All calculations were performed with ferromagnetic spin alignment.
The calculated Li-Mn-F phase diagram, the expected conversion products, and their
associated energies and voltages were obtained from the Materials Project [102].
49
2.3 Materials preparation
2.3.1 Synthesis of dirutile LiMnF4
Stoichiometric amounts of LiF and MnF3 were milled together in a Retsch PM200
planetary ball mill for 6 hours at 500rpm, with up to 10% excess MnF3 . This milling
process induces a mechanochemical reaction between the precursors, and a subsequent
thermal treatment completes the reaction to dirutile LiMnF4 . Figure 2-2a shows the
XRD pattern collected after milling stoichiometric amounts of LiF and MnF3 precursor for 6 hours at 500rpm. While the main peaks of LiF and MnF3 can still be seen,
as marked by L and M, respectively, the primary peaks in this diffraction pattern
belong to a new mechanochemically-formed rutile phase, which will be discussed in
more detail in Section 2.3.2. Some dirutile LiMnF4 has also formed mechanochemi-
∘
∘
∘
cally. The largest peaks of dirutile LiMnF4 are found at 17.6 , 25.5 , and 27 Cu 2𝜃 ,
and can already be seen in Figure 2-2a after the first milling process.
Figure 2-2b-d shows the evolution of the reaction with increased firing tempera-
∘
tures under argon. The rutile phase crystallizes at temperatures up to 250 C. Comparing the pattern from the milled precursors (Figure 2-2a) to the sample heated at
∘
200 C (Figure 2-2b), we see that the intensity of the precursor peaks have decreased
and the rutile peaks have sharpened, indicating the progression of the solid state reaction. The dirutile peaks grow as the sample is heated to higher temperatures. At
∘
250 C (2-2c), the precursors have completely reacted, and the XRD pattern shows
a partial transformation from the rutile phase to the dirutile LiMnF4 phase.
The
∘
∘
∘
dirutile peaks at 17.6 , 25.5 , and 27 Cu 2𝜃 are significantly larger in Figure 2-2c
∘
than Figure 2-2b. Finally, at 300 C (Figure 2-2d), the rutile phase has completely
∘
disappeared, indicated by the missing rutile peak at 25.8 Cu 2𝜃 . The main peaks
of Figure 2-2d match those of dirutile LiMnF4 , and the two small impurity peaks
denoted by black circles belong to Li2 MnF5 .
The impurity phase, Li2 MnF5 , pre-
ferrentially forms over dirutile LiMnF4 when firing in air instead of argon. Further
∘
heating to temperatures between 300-600 C produces no changes in the obtained
∘
XRD patterns. Above 600 C, LiMnF4 decomposes.
50
Figure 2-2: XRD reveals the formation of rutile LiMnF4 by mechanochemical reaction,
and transformation of rutile LiMnF4 to dirutile LiMnF4 by solid state reaction. XRD
patterns are shown for (a) Ball milled LiF and MnF3 . M = MnF3 and L = LiF still
∘
unreacted after milling. (b) After heating the ball milled precursors to 200 C for 10
∘
hr under argon. (c) After heating the ball milled precursors to 250 C for 10 hr under
∘
argon. (d) After heating the ball milled precursors to 300 C for 10 hr under argon.
Black circles denote the main peaks of an impurity phase, Li2 MnF5 . (e) Theoretical
peak positions of dirutile LiMnF4 , space group P21 /c.
2.3.2 Synthesis of rutile LiMnF4
Rutile LiMnF4 appears as the dominant phase after mechanochemical reaction of LiF
and MnF3 .
The same rutile phase also appears when ball milling dirutile LiMnF4
by itself. In Figure 2-3, the reversibility of the phase transition between dirutile and
51
rutile phases with ball milling and heating cycles is shown by distinct changes in the
diffraction patterns. Figure 2-3a shows the diffraction pattern obtained after milling
the precursors, where the main peaks in the pattern belong to a rutile phase and “L”
and “M” denote the peaks of unreacted LiF and MnF3 precursors.
This sample is
∘
then heated under argon gas flow for 10 hours at 300 C, producing the crystalline
dirutile phase in Figure 2-3b.
Excluding the peaks marked by black circles, which
belong to a Li2 MnF5 impurity, all other peaks belong to dirutile LiMnF4 . Milling the
dirutile reverts it to rutile (Figure 2-3c), and reheating this rutile transforms it back
to dirutile (Figure 2-3d).
Phase pure rutile LiMnF4 is obtained after milling dirutile LiMnF4 by itself for
twelve hours. We could not match the peaks to a known phase of LiMnF4 , but noticed
a similarity of the experimental diffraction pattern to that of rutile MnF2 . At just
∘
300 C, the rutile transforms completely to the dirutile phase; lower temperatures also
cause partial transformations.
2.4 Structure characterization
2.4.1 Rietveld refinement
Dirutile LiMnF4 is monoclinic with space group P21 /c and unit cell parameters of a
= 5.41Å, b = 4.63Å, c = 5.69Å, and
𝛽
∘
= 113.24 . As described by Lacorre et al. [103],
the dirutile structure has twice the periodicity of a true rutile due to ordering of planes
+
3+
of Li and Mn
ions in the [100] direction (Figure 2-4a). Along the [101] direction,
edge-sharing LiF6 and MnF6 octahedra alternate to form a rutile-like structure with
empty channels between edge-sharing octahedral chains (Figure 2-4b).
Refinement on the P21 /c space group for dirutile LiMnF4 gives unit cell parameters
of a = 5.43 Å, b = 4.64 Å, c = 5.70 Å, and
𝛽
= 113.18
∘
(Figure 2-5). These values
∘
are within 0.01 Å and 0.1 of those reported by Wandner et al [104].
As mentioned in Section 2.3.2, the rutile phase of LiMnF4 is not known polymorph,
but structurally similar to rutile MnF2 . We performed Rietveld refinement using the
52
Figure 2-3: XRD patterns for (a) Rutile LiMnF4 obtained by ball milling precursors
LiF and MnF3 . L and M denote the main peaks of unreacted LiF and MnF3 precursor,
respectively. (b) Dirutile LiMnF4 obtained from heating sample A for 10 hours at
∘
300 C under argon. Black circles denote the impurity Li2 MnF5 phase. (c) Rutile
LiMnF4 obtained by ball milling sample B for 12 hours at 500 rpm. (d) Dirutile
∘
LiMnF4 obtained from heating sample C for 10 hours at 300 C under argon. Black
circles again denote the impurity Li2 MnF5 phase.
P42 /mnm space group, assuming exactly half occupancy of the cation site by Li and
Mn (Figure 2-6). The unit cell parameters of the new rutile phase are determined
to be a = b = 4.786Å and c = 2.985Å. We were unable to obtain a more crystalline
sample of the new rutile phase for characterization.
53
Figure 2-4: (a) Dirutile LiMnF4 has [100] planes of corner-sharing octahedral of a
single cation species. (b) The projection of dirutile LiMnF4 along [101] shows empty
channels formed between the edge-sharing octahedral chains.
2.4.2 Mechanically-driven phase transformations
Mechanochemistry has been widely employed in recent years, but the precise reaction
mechanisms are still speculative, as they are hard to evaluate either experimentally or
computationally [96]. The cause of mechanically-induced phase transitions in certain
materials is also not well understood; they may occur as a combined result of shear,
high defect concentration, and high temperatures [105–108].
The dirutile-to-rutile
phase transformation observed in our work follows similar trends seen in other oxides
54
Figure 2-5: Rietveld refinement of dirutile LiMnF4 on the P21 /c space group shows
good agreement between the experimental (red) and refined (black) diffraction patterns with R𝑤𝑝 = 12.3% and GOF = 3.336. The peak positions for dirutile LiMnF4
are shown by the green vertical lines, and the difference plot is shown by the bottom
blue line.
and fluorides. Short milling times can produce polymorphs that are metastable at ambient temperature and pressure. These metastable phases tend to be the equilibrium
phases at high pressure and/or high temperature conditions, and have more compact
structures [105]. The mechanically-formed rutile LiMnF4 follows this behavior with
a smaller volume than its dirutile counterpart, 68 Å
2
vs.
71 Å
2
per formula unit.
It is unclear if the new rutile polymorph of LiMnF4 forms because of the extreme
processing conditions seen in the ball mill or because of stabilizing surface energy
effects at the nanoscale.
55
Figure 2-6: Rietveld refinement of the new rutile LiMnF4 polymorph on the P42 /mnm
space group shows good agreement between the experimental (red) and refined (black)
diffraction patterns with R𝑤𝑝 = 9.6% and GOF = 1.825. The peak positions for rutile
LiMnF4 are shown by the green vertical lines, and the difference plot is shown by the
bottom blue line.
2.4.3 Computed polymorph energies
While mechanically-induced phase transformations are difficult to model, computations can still be used to investigate the stability of the final rutile LiMnF4 polymorph.
Rietveld refinement suggests that rutile LiMnF4 is disordered on the cation site with
equal probability of occupation of the site by Li or Mn. To estimate the energy scale
of cation disorder in the rutile framework, thirteen symetrically-distinct structures
with alternate Li-Mn orderings were calculated with a unit cell of less than or equal
to 18 atoms. Six of the thirteen structures were within 31 meV/atom of the ground
state dirutile LiMnF4 , and four were within 5 meV/atom.
While these alternative
orderings are not identical to the disordered structure used in Rietveld refinement,
the small energetic difference between Li-Mn orderings in LiMnF4 is consistent with a
56
phase which would show Li-Mn disorder. This agrees with experimental observations
where simple mechanical milling causes a phase transition from the ordered dirutile to
disordered rutile LiMnF4 , and the reverse reaction is activated by thermal treatment.
2.5 Electrochemical performance
Three different LiMnF4 samples were tested for electrochemical performance. Dirutile
LiMnF4 , obtained by solid state reaction, was ball milled with carbon for two hours
before being made into electrodes. Rutile LiMnF4 , obtained by ball milling dirutile
LiMnF4 by itself for twelve hours, was then ball milled with carbon for two hours.
The last sample was obtained by ball milling dirutile LiMnF4 for twelve hours with
carbon. Dirutile LiMnF4 partially transforms to rutile LiMnF4 during the twelve hour
milling cycle with carbon; these samples are referred to as the two-phase samples.
Figure 2-7a and b show the first cycle galvanostatic charge-discharge data of the
dirutile and rutile samples, respectively. Cells were cycled at rates of C/20, C/50, and
C/100. C/100 data is not shown in Figure 2-7 as the slower C/100 test rate shows
no improvement in specific capacity over the C/50 test rate in either sample. In the
second discharge, a specific capacity of 45 mAh/g was attained in the dirutile samples,
and 55 mAh/g in the rutile samples. We hypothesize that the specific capacity of both
these single-phase samples is low due to inadequate conductivity. The particle size
distribution seen in TEM after milling with carbon is
100-200nm for the dirutile phase (Figure 2-8).
<30nm for the rutile phase, and
Only short milling times were used
to mix the active cathode material with carbon in order to maintain single-phase
samples. Given the small particle sizes after milling and the high carbon content in
the electrodes, a fraction of the capacity observed in samples may be due to capacitive
storage in carbon.
Galvanostatic charge-discharge on the two-phase sample of dirutile and rutile
LiMnF4 was also performed at C/20, C/50, and C/100 rates. As shown in Figure 2-9,
the C/100 sample reaches 170 mAh/g on first discharge, which is three times the capacities observed in single phase dirutile or rutile samples, and 92% of the theoretical
57
Figure 2-7: (a) First and second galvanostatic charge-discharge curves for dirutile
LiMnF4 milled for 2 hours with carbon at C/20 and C/50.
(b) First and second
galvanostatic charge-discharge curves for rutile LiMnF4 milled for 2 hours with carbon
at C/20 and C/50.
capacity expected for a 1 Li reaction. For the two-phase samples tested at C/50 and
C/20, the capacity of first discharge is 103 mAh/g and 84 mAh/g, respectively. The
higher capacities of the two-phase sample are likely due to the longer milling time
58
Figure 2-8: Bright field TEM images show the order of magnitude difference in particle
size between dirutile and rutile LiMnF4 . (a) TEM of 100-200nm particles of dirutile
LiMnF4 . (b) TEM of <30nm particles of rutile LiMnF4 .
of active material with carbon, which reduces particle size and increases intermixing
with carbon. As with other fluoride materials, we believe the key to realizing higher
capacities is reducing particle sizes and achieving good intermixing with a conductive
matrix.
59
Figure 2-9: First and second galvanostatic charge-discharge curves for a two-phase
sample of dirutile and rutile LiMnF4 milled for 12 hours with carbon at C/20, C/50,
and C/100.
2.6 Lithiation mechanisms
2.6.1 Observed conversion reactions
To distinguish between insertion and conversion of the LiMnF4 phases upon lithiation,
the rutile and dirutile samples were discharged at a C/100 rate to 3V, 2V, and 1V
and disassembled for
ex situ
TEM electron diffraction.
Additionally,
ex situ
XRD
patterns were collected on rutile and dirutile films discharged to 1V before preparing
the discharged films for TEM. The XRD showed LiMnF4 peaks to be reduced in
intensity but still present, and lattice parameters to be unchanged.
Neither of the
intercalated Li2 MnF4 phases nor conversion products were detectable from XRD.
In contrast, distinct conversion products are visible in TEM for each of the three
discharge levels of rutile LiMnF4 .
Figure 2-10 shows the SAED patterns obtained
from rutile LiMnF4 discharged to 3V, 2V, and 1V. At the most shallow discharge level
of 3V, rutile LiMnF4 phase is found most frequently in the indexed ring diffraction
60
patterns (Figure 2-10a), but some of the rutile LiMnF4 has been converted into LiF
and MnF2 . At the 2V discharge level, most diffraction patterns show a superposition
of the LiF and MnF2 phases (Figure 2-10b). The LiF diffraction rings are much more
diffuse than the MnF2 ones at 2V, which indicate LiF is poorly crystallized. Some
rutile LiMnF4 is still present at 2V. At the 1V discharge level, the indexed phases
are
𝛼-Mn,
MnF2 , and LiF (Figure 2-10c). The intensity of the MnF2 peaks is lower
than those found in the 2V sample, indicating that some MnF2 is further reduced to
𝛼-Mn.
For both the 2V and 1V samples, while there are still areas with unreacted
rutile LiMnF4 , we only show the SAED patterns from areas undergoing the obvious
conversion reaction.
In contrast to the ring diffraction patterns observed in the rutile LiMnF4 samples, dirutile LiMnF4 spot diffraction patterns are observed at all levels of discharge
(Figure 2-11a and b).
The amorphous ring diffraction patterns of the conversion
products, which are indexed to
spot diffraction patterns.
by 1V, the
𝛼-Mn
𝛼-Mn
At 3V, the
metal, superimpose with the dirutile LiMnF4
𝛼-Mn
rings are observed occasionally, while
rings are seen in all images. It was necessary to enhance the con-
trast or overexpose the dirutile pattern in order to observe the faint Mn amorphous
rings. The difference in the diffraction patterns between rutile samples (ring pattern)
and dirutile samples (spot pattern) is due to the different particle sizes from sample
preparation: the particle sizes of dirutile LiMnF4 prepared by solid state reaction
were determined from TEM to be 100-200nm, while rutile LiMnF4 particles formed
mechanochemically were <30nm.
2.6.2 Calculated intercalation properties
Two compounds were computationally investigated as candidates for Li2 MnF4 , the intercalation product of dirutile LiMnF4 . The first structure, Li2 MnF4 -oct, was formed
by inserting additional Li into the distorted octahedral voids in the lithium layer of
dirutile LiMnF4 , resulting in columns of face-sharing Li octahedra. The second structure, Li2 MnF4 -tet, avoids Li octahedral face-sharing by splitting each octahedral Li
site into two neighboring Li tetrahedral sites. The resultant Li2 MnF4 -tet structure
61
Figure 2-10: SAED patterns showing conversion products of (a) rutile LiMnF4 discharged to 3V (b) rutile LiMnF4 discharged to 2V (c) rutile LiMnF4 discharged to
1V.
62
Figure 2-11:
SAED patterns showing conversion products of (a) dirutile LiMnF4
discharged to 3V (b) dirutile LiMnF4 discharged to 2V and 1V, which are indexed to
the same phases
is topologically identical to La2 CuO4 [109].
All calculations were initialized with
monoclinic symmetry.
We determined lithium’s preference for the tetrahedral sites by comparing energetics of the two possible Li2 MnF4 intercalation products. Li2 MnF4 -tet is preferred
by a sizeable 97meV/atom to Li2 MnF4 -oct, corresponding to 339.5meV/lithium. This
is because lithium tetrahedra in Li2 MnF4 -tet only edge-share with neighboring MnF6
octahedra, while the lithium octahedra in Li2 MnF4 -oct face-share with other lithium
octahedra. The XRD patterns of Li2 MnF4 -oct and Li2 MnF4 -tet are clearly distinct
from the dirutile LiMnF4 or rutile LiMnF4 patterns.
Additionally, two structures at intermediate lithiation levels between LiMnF4
and Li2 MnF4 were calculated.
To gain insight into the dilute limit of insertion,
63
Li(1+𝛿) MnF4 , a single octahedral Li was replaced by two tetrahedral Li in a 2x2x2
supercell of LiMnF4 , resulting in the composition Li17/16 MnF4 .
Similarly, for the
saturated limit, Li(2−𝛿) MnF4 , all but one octahedral lithium were replaced by two
tetrahedral Li in the same 2x2x2 supercell, resulting in composition Li31/16 MnF4 .
Figure 2-12: Lithium insertion into dirutile LiMnF4 may induce a lithium site shift
from the octahedral site (outlined in orange) to the two neighboring tetrahedral sites
(outlined in blue). Partially lithiated Li(1+𝑥) MnF4 is shown here with MnF6 octahedra
represented in purple in the background and LiF6 octahedra represented in green in
the foreground.
Figure 2-12 shows a possible structure of partially lithiated Li(1+𝑥) MnF4 , assuming some lithium intercalation into dirutile LiMnF4 . MnF6 octahedra are represented
by the purple octahedra in the background, LiF6 octahedra are represented by green
octahedra in the foreground, and the original lithium octahedral site is outlined in
orange.
When an extra lithium ion enters the lithium layer, the original lithium
shifts into one of the neighboring tetrahedral sites, and the extra lithium ion occupies the other tetrahedral site. The new lithium tetrahedral sites after intercalation
are outlined in blue. These tetrahedral sites are face-sharing only with the original
octahedral Li site and vertex-sharing with all other “non-split” Li octahedral sites.
64
2.6.3 Comparison of conversion and intercalation reactions
Figure 2-13 shows the calculated Li-Mn-F equilibrium phase diagram at 0K. Ground
state compounds are denoted by the solid red dots.
The two proposed structures
for Li2 MnF4 , the intercalation product of dirutile LiMnF4 , are both metastable and
marked by the open red dot on the phase diagram.
From the phase diagram, the
following equilibrium conversion reactions and voltages are expected with increasing
lithium content:
Li + 3LiMnF4
→
2Li2 MnF5 + MnF2 (4.13V)
Li + Li2 MnF5
→
MnF2 + 3LiF (3.99V)
2Li + MnF2
→
2LiF + Mn (1.89V)
Figure 2-13: The Li-Mn-F phase diagram at 0K is derived from first principles. Thermodynamically stable compounds are denoted by solid red dots. The lithiation path
of LiMnF4 is marked by the dotted line. Insertion into LiMnF4 produces metastable
Li2 MnF4 , which is denoted by the open red dot.
The equilibrium conversion reactions are compared to possible non-equilibrium
reaction paths in Figure 2-14. All compound energies are plotted relative to dirutile
LiMnF4 which sits at 0 meV per Li𝑥 MnF4 on the y-axis. At composition LiMnF4 , the
points in the range of 3-100meV on the y-axis represent structures with alternative
65
Li-Mn orderings. The other blue points at the top of Figure 2-14 show the energies of
two intermediate insertion compounds (Li17/16 MnF4 and Li31/16 MnF4 ) and two fully
intercalated compounds (Li2 MnF4 -oct and Li2 MnF4 -tet). The computed energies for
both intermediate compounds are higher in energy than the tie line between LiMnF4
and Li2 MnF4 -tet, suggesting that if intercalation were to occur, it would take place
as a two-phase insertion reaction directly from LiMnF4 to Li2 MnF4 -tet at 3.87V.
Figure 2-14:
Energies of compositions between LiMnF4 and Li2 MnF4 are plot-
ted relative to that of dirutile LiMnF4 .
Rutile polymorphs and insertion products
(Li1+𝑥 MnF4 ) are represented by blue points, and competing conversion products at
specific lithiation levels are represented by red points. Lines connecting points are labeled with the calculated equilibrium voltages between compounds. The equilibrium
voltage is linearly related to the slope between points.
The red points at the bottom of Figure 2-14 represent the equilibrium conversion
products given by the phase diagram. While the equilbrium lithiation path traverses
the Li2 MnF5 + MnF2 point before reaching LiF +MnF2 , converting dirutile LiMnF4
directly to LiF + MnF2 is not much higher in energy, and hence can occur with even
66
modest underpotentials upon discharge.
2.7 Discussion
As shown by galvanostatic charge-discharge tests and
ex situ
TEM electron diffrac-
tion, rutile and dirutile LiMnF4 are both electrochemically active but convert upon
lithiation along different reaction paths.
The observed reactions do not match the
equilibrium conversion reactions given by the phase diagram.
an artifact of TEM performed
ex situ
rather than
in situ
This may either be
and not all intermediate
reactions being captured, or because rutile and dirutile LiMnF4 simply convert via
nonequilibrium reaction paths upon lithiation. All reactions are summarized graphically in Figure 2-15.
Figure 2-15: Starting with LiMnF4 , possible reactions of fluoride compounds with
lithium are labeled with their expected onset voltages.
The lithium fraction and
specific capacity listed to the left of each box are calculated relative to LiMnF4 . Less
plausible reactions are denoted by dashed arrows, while more plausible reactions are
denoted by solid arrows.
The gray box containing Li2 MnF4 assumes insertion and
lithium site shift. All other boxes represent conversion products from reactions with
lithium.
In Figure 2-15, boxes are placed top to bottom in order of increasing lithium chemical potential (lower voltage). Based on the experimental data, we assume lithium
67
insertion occurs at most to a small extent (e.g. exclude Li2 MnF4 -like phases). We also
assume the metastable rutile phase is close in energy to dirutile LiMnF4 , so that we
have a single LiMnF4 starting point at the top of the flow chart. Solid arrows connecting boxes denote likely reaction paths, while dashed arrows represent less plausible
reaction paths which are shown only for the sake of completeness. The numbers next
to each arrow give the equilibrium reaction voltage between compounds in connected
boxes.
In the equilibrium reaction, the first expected reaction upon lithiation is the conversion of LiMnF4 to Li2 MnF5 and MnF2 at 4.13V, followed by the conversion of
Li2 MnF5 to MnF2 and LiF at 3.99V. However, in the SAED patterns of discharged
rutile samples, neither Li2 MnF5 nor superpositioned Li2 MnF5 and MnF2 patterns
are observed. The first observed conversion products are MnF2 and LiF, which first
appear alongside rutile LiMnF4 at the 3V discharge level. These observations suggest
that either the Li2 MnF5 formed at 4.13V quickly decomposes into MnF2 and LiF at
3.99V, or rutile LiMnF4 converts directly to LiF and MnF2 at 4.04V.
If we assume that rutile LiMnF4 follows the center 4.04V path in Figure 2-15
(LiMnF4
→
LiF + MnF2 ), the subsequent conversion to Mn and LiF is not expected
to begin until 1.89V. Our experimental results agree well with theory: at 2V, MnF2
and LiF are the dominant phases. Some rutile LiMnF4 is also still observed, but Mn
metal is not. The presence of some rutile LiMnF4 in the 2V sample suggests that the
direct conversion of rutile LiMnF4 to Mn metal is not preferred. As seen in Figure 215, the direct reaction of LiMnF4
→
Mn + LiF is predicted to begin at 2.60V. If this
2.60V reaction path were active, we would expect to see some Mn metal in the 2V
sample.
The rutile sample further discharged to 1V is indexed to
𝛼-Mn,
LiF, and some
remnant MnF2 . Rutile LiMnF4 is no longer observed at 1V. Between 2V and 1V, rutile
LiMnF4 may either convert directly to LiF + Mn, or form MnF2 before converting
to LiF + Mn. The MnF2 peaks in the 1V sample are weaker than those in the 2V
sample, suggesting there is less MnF2 present in the sample at 1V because of the
further conversion to
𝛼-Mn.
The diffused ring patterns of the converted LiF and Mn
68
metal phases show that they are much smaller in particle size than the starting phase
of rutile LiMnF4 , which is typical of conversion products in battery cathodes. It is
clear that rutile LiMnF4 is converting, but we cannot entirely rule out the possibility
of lithium insertion at the start of discharge, where the discharge specific capacity
may have contributions from both insertion and conversion reactions. A small amount
of lithium insertion may be possible if conversion of LiMnF4 is kinetically limited.
In contrast to the ring patterns observed in the rutile samples, dirutile LiMnF4
samples discharged between 3V and 1V all show a component of single crystal spot
diffraction patterns (Figure 2-11a and b) which belong to 100-200nm dirutile LiMnF4
particles. The superimposed amorphous ring patterns are from the conversion product,
𝛼-Mn,
formed during discharge.
This is the only conversion product indexed
in the discharged dirutile LiMnF4 samples.
𝛼-Mn’s
unexpected appearance at the
3V discharge level suggests that lithiation of dirutile LiMnF4 is dominated by direct
conversion to LiF + Mn.
It is not yet determined why
the conversion of LiMnF4
→
𝛼-Mn
is seen at 3V when
Mn + LiF is predicted to begin at 2.60V. LiF is not
observable in the SAED patterns because it is a weaker scatterer than Mn.
Like
rutile LiMnF4 , dirutile LiMnF4 is clearly converting to LiF and Mn, but may also be
inserting at the start of discharge.
Although the possibility of electron beam damage cannot be excluded as an explanation for
𝛼-Mn’s
presence in the 3V discharged sample,
𝛼-Mn’s
amorphous ring
pattern is only observed in several SAED patterns out of the 30 measured patterns
taken from different areas of the TEM sample. In contrast, in the 2V and 1V discharged samples,
𝛼-Mn
is observed in most SAED patterns. Considering the mea-
surement condition for SAED is kept constant for all measurements with the broadest
electron beam, this difference observed by SAED is most likely not an artifact. Using
scanning TEM techniques including STEM, we found that STEM as a focused beam
technique caused obvious artifacts and beam damage to these fluorides compared with
the conventional TEM technique of SAED.
Finally, we consider the remaining reaction paths shown in Figure 2-15.
The
reaction at 2.41V converting Li2 MnF5 and MnF2 directly to LiF and Mn is unlikely;
69
if this reaction path were active, Li2 MnF5 should have been observed in a 3V sample.
Thus, the 2.41V reaction is labeled with a dashed arrow to denote a less plausible
reaction. Insertion to Li2 MnF4 at 3.89V is also unlikely - no evidence of insertion is
seen in XRD or TEM, and thermodynamically, conversion is preferred over insertion
(Figure 2-14).
2.8 Conclusion
Dirutile LiMnF4 was computationally identified to be a promising new conversion
cathode material for lithium-ion batteries. In this work, two polymorphs of LiMnF4
were synthesized, tested as lithium-ion battery cathodes, and studied using electron
diffraction and computation to understand their lithiation mechanisms.
Synthesis
of dirutile LiMnF4 , a known compound, was achieved by a new low-temperature
mechanochemical/solid state reaction route, and the new rutile polymorph of LiMnF4
was synthesized by mechanical milling. Both phases were shown to be electrochemically active, but took different non-equilibrium reaction paths upon lithiation. Dirutile LiMnF4 converts directly to LiF and Mn, while rutile samples forms MnF2 before
converting to LiF and Mn metal. Computations suggest that the two nonequilbrium
paths taken by the samples are close in energy to the ground state path.
Due to lack of stability under long ball milling, phase-pure samples of dirutile
and rutile cannot be well mixed with carbon or reduced in particle size, leading
to low specific capacities. Two-phase samples of dirutile and rutile LiMnF4 , which
were derived from longer milling times, showed substantially larger capacity. Hence,
these results identify an opportunity for improving capacity if alternative sample and
electrode preparation methods are found to obtain smaller particle size and better
carbon-mixed electrodes.
70
Chapter 3
Layered Li𝑥Ni2−4𝑥/3Sb𝑥/3O2:
Designing lithium-excess cathode
materials from percolation theory
3.1 Introduction
The layered transition metal oxides (LiMO2 ) are well established cathode materials
for lithium ion batteries [11, 110, 111], achieving high reversible capacities through extraction of lithium from the lithium layer [52]. Historically, the chemical design space
of layered cathode materials has been limited to transition metal cations resistant
to migration into the lithium layer. Anti-site disorder between transition metal and
lithium cations, the result of transition metal migration, degrades lithium mobility
due to reduced oxygen spacing around the lithium layer [54]. This layer spacing, also
known as the lithium slab spacing, has been shown to be the most controlling factor
on lithium diffusivity and hence on practical lithium extraction capacity [52–54, 112].
Anti-site disorder can be present in the synthesized material or created upon severe
delithiation [113–116]. Thus, significant efforts have been made to design well-layered
materials resistant to structural instability [26, 27, 117–120].
Several recent observations have cast doubt upon this ordering paradigm by dis-
71
playing very high lithium cycling capacity, notwithstanding being almost fully cation
disordered [5, 28, 121, 122]. Insight into the remarkable performance of materials with
high disorder was given by Lee et al. [5] and Urban et al. [4], who demonstrated
that high capacity can be achieved even in disordered rocksalt materials if they have
high enough lithium content to form percolating 0-transition metal (0-TM) diffusion
channels.
As described in Section 1.3, the diffusion channel in close-packed oxides
connects one octahedral site to another octahedral site through a tetrahedral activated state [52]. The 0-TM channels have no transition metals around the activated
state, making them considerably less sensitive to the lithium slab spacing distance,
and as a result, largely insensitive to cation disorder. This finding enlarges the chemical design space for cathodes by including transition metal ions which, due to their
size or electronic structure [123], are prone to migration upon lithium cycling. Consequently, new opportunities emerge to discover lithium ion battery materials with
very high capacity.
To validate our understanding of compounds with 0-TM diffusion channels, we revisited a material prone to cation mixing to see if increasing lithium content improves
electrochemical performance. Layered LiNi0.67 Sb0.33 O2 has a high discharge voltage
of 3.9V and a theoretical capacity of 226 mAh/g, but reaches only 90 mAh/g on
its first discharge and shows rapid capacity fade on subsequent cycles [124].
Ex situ
x-ray diffraction suggests 10% Li/Ni mixing to occur over ten cycles, which reduces
the lithium slab spacing and explains the poor performance of the material.
Both
the high theoretical energy density of LiNi0.67 Sb0.33 O2 and its cation mixing tendencies make LiNi0.67 Sb0.33 O2 an ideal candidate material to explore the effect of lithium
excess and the formation of 0-TM diffusion channels.
3.2 Materials design
To assess the impact of increasing lithium excess on observed capacity, four layered
Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 (LNSO) compounds containing 0–15% lithium excess (
1.05, 1.10, and 1.15) were synthesized.
x
= 1.00,
In each compound, lithium excess is acco-
72
modated by adjusting the Sb/Ni ratio to keep the ions as 5
+
+
and 2 , respectively.
Hereafter, we abbreviate each compound by the amount of excess lithium,
e.g.
LNSO-
0 is LiNi0.67 Sb0.33 O2 and LNSO-15 is Li1.15 Ni0.47 Sb0.38 O2 .
Figure 3-1: Maximal capacity based on various assumptions: The electron-limited
2+/4+
capacities of the LNSO compounds are calculated from the Ni
redox couple,
and peak at 9% lithium excess. The 0-TM capacities in the disordered rocksalt and
layered rocksalt structures are calculated as the amount of Li in the 0-TM percolating
pathway.
The dotted vertical lines mark the 0-TM percolation thresholds in these
two structures, which are at
x
= 1.09 and 1.14, respectively. The LNSO compounds
studied are marked by black squares.
The theoretical capacity of the LNSO compounds is calculated based on the
Ni
2+/4+
redox couple, and varies as a function of lithium excess.
The solid black
line in Figure 3-1 represents the theoretical electron-limited specific capacity of LNSO
compounds, and the black squares denote the LNSO compounds studied in this work.
5+
With increasing lithium content, the fraction of high valence Sb
increases, and the
2+
amount of Ni
decreases. Although nickel is the only redox-active transition metal,
2+/3+
3+/4+
+
its two redox couples, Ni
and Ni
, allow for two Li
to be extracted per
nickel ion.
Therefore, despite decreasing nickel content, the LNSO chemistries can
incorporate excess lithium at a smaller expense to theoretical capacity. The theoret-
73
Table 3.1: Lattice parameters, lithium slab spacing, and cation mixing levels of pristine Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds determined by Rietveld refinement.
x
Composition
a
1.00
Li1.00 Ni0.67 Sb0.33 O2
2.984
14.549
1.05
Li1.05 Ni0.60 Sb0.35 O2
2.981
14.552
1.10
Li1.10 Ni0.53 Sb0.37 O2
2.981
1.15
Li1.15 Ni0.47 Sb0.38 O2
2.983
(Å)
c
(Å)
Oxygen
z
Li slab (Å)
Ni𝐿𝑖 (occ)
0.2522
2.492
0.030
0.2540
2.543
0.048
14.564
0.2575
2.646
0.054
14.568
0.2575
2.646
0.060
ical capacity of the LNSO compounds initially increases between 0% and 9% lithium
excess as a larger amount of lithium becomes available to extract and oxidize all Ni
4+
to Ni .
2+
Above 9% lithium excess, where the Li/Ni ratio equals 2, the theoretical
capacity decreases as there is insufficient Ni redox to extract all lithium by transition
metal oxidation.
The dashed lines in Figure 3-1 denote the calculated 0-TM theoretical capacities
in an ideal disordered rocksalt structure (red line) and in an ideal layered rocksalt
structure (blue line).
We define 0-TM capacity as the lithium that are part of a
percolating network of 0-TM diffusion channels, which should be easy to extract.
0-TM capacity increases as a function of lithium excess as an increasing number of
0-TM diffusion channels are formed. The slope of the 0-TM capacity is determined
by the topology of the reference crystal structure (disordered or layered) [4].
The 0-TM percolation thresholds of the disordered rocksalt structure and the
layered rocksalt structure are at 9% and 14% lithium excess, respectively, assuming
excess lithium is distributed randomly throughout the transition metal layer.
For
lithium contents below these values, the corresponding 0-TM capacity is exactly zero
in a macroscopic structure.
Due to finite-size effects in the numerical percolation
simulations, the onset of the 0-TM curves is slightly smoothed out and begins before
the percolation threshold [125]. Rietveld refinement shows that all LNSO compounds
are synthesized in the layered rocksalt structure with slight cation mixing (Table 3.1).
Accordingly, we expect the 0-TM percolation threshold for the LNSO compounds to
be close to 14%, the threshold for a perfectly layered rocksalt, and only the LNSO-15
sample to contain percolating 0-TM channels.
74
3.3 Methods
3.3.1 Synthesis
Li2 CO3 , Sb2 O3 , and Ni(OH)2 were ball milled in acetone for 24 hours at 300 rpm.
In the precursor mixture, 5% excess of Li2 CO3 was incorporated over the desired
stoichiometry to compensate for lithium losses during synthesis. After ball milling,
the precursors were dried and pressed into pellets. The LNSO-0 composition was fired
∘
under oxygen for ten hours at 900 C, while all lithium-excess LNSO compositions were
∘
fired at 800 C.
3.3.2 Experimental characterization
XRD patterns were collected in capillaries at the Brookhaven National Laboratory
National Synchrotron Light Source, beamline X14A (𝜆 = 7.792947 Å), and then
converted to a 2𝜃 range corresponding to a Cu source.
Rietveld refinement was
performed on the XRD patterns using PANalytical High Score Plus. Scanning electron images were collected on a Zeiss Merlin SEM. Selected area electron diffraction
(SAED) patterns and high resolution transmission electron microscopy (HRTEM)
images were obtained on a JEOL 2010F. Ni and Sb K-edge extended x-ray absorption fine structure (EXAFS) data were collected in transmission mode at the Argonne National Laboratory Advanced Photon Source, beamline 20-BM-B. The x-ray
beam was monochromatized using a fixed exit Si (111) double crystal monochromator, and energy calibration was performed using Ni/Sb foils. Harmonic contamination
was minimized using a harmonic rejection mirror in conjunction with crystal detuning.
EXAFS data reduction and analysis was performed using IFEEFIT suite of
software [126].
The fits to determine structural parameters were carried out using
amplitude and phase functions generated from
75
ab intio
FEFF6 calculations [127].
3.3.3 Electrochemistry
The LNSO active material and carbon black (Timcal, super P) were first mixed
in a Retsch PM200 planetary ball mill at 300 rpm between 100 minutes and six
hours to reduce particles to 50-100nm. Electrodes were then made by mixing LNSO,
carbon, and polytetrafluoroethylene (PTFE) binder in a ratio of 80:15:5 by weight
using a morter and pestle, and rolling the electrode into thin films. Swagelok cells
were assembled in an argon-filled glovebox using lithium metal as the anode, Celgard
C2320 as separator, and 1M LiPF6 in 1:1 ethylene carbonate:dimethyl carbonate
(EC:DMC) as electrolyte. Cells were cycled in galvanostatic mode on a Maccor 2200
or Maccor 4000 operating at room temperature.
3.3.4 Computations
The 0-TM capacities shown in Figure 1a were computed using a Monte-Carlo (MC)
approach [4]. The simulations were carried out on a face-centered cubic lattice using
a
16 × 16 × 16 supercell of the conventional 4-atom unit cell (16, 384 sites).
simulation comprised
1, 000
Each MC
MC steps.
All reported first-principles results are based on Kohn–Sham density-functional
theory (DFT) [128, 129] using the PBE exchange–correlation functional [100, 130] as
implemented in the Vienna
ab-initio simulation package (VASP) [131,132].
a Hubbard-𝑈 correction [101, 133, 134] of
reference 102.
𝑈 = 6.0 eV
For nickel,
was employed, as suggested in
A plane-wave energy cutoff of 520 eV was used, and
𝑘 -point
grids
were chosen to converge the total energy within 1 meV per Li(Li,Ni,Sb)O𝑧 formula
unit. LNSO orderings for specific compositions were generated using the approach
of Hart and Forcade [135–137] as implemented in the Python Materials Genomics
library [138].
76
3.4 Validating materials design by electrochemical
performance
To compare intrinsic electrochemical performance of all compounds as much as possible, all LNSO samples were ball milled to achieve similar final particle sizes of 50-100
nm, as confirmed by SEM (Figure 3-2). All cells were then galvanostatically cycled at
the 1C currents corresponding to their respective Ni-based theoretical capacities. The
charge-discharge curves for LNSO-0 compared to LNSO-15 are shown in Figure 3-3;
the other two samples show similar charge plateaus at 4V and discharge plateaus at
3.9V. The slight change in voltage profile between the first charge and all subsequent
charges may be caused by some minor structural rearrangements. Of the four LNSO
compounds, LNSO-15 exhibits the best performance, achieving 131 mAh/g on first
discharge and retaining 78% of its capacity over 50 cycles with negligible voltage fade.
To quantitatively compare the cyclability between the four LNSO compounds, we
plot discharge capacity versus cycle number in two ways. The discharge capacities are
shown as specific capacities in Figure 3-4a, and as fractions of theoretical capacity in
Figure 3-4b, which is the discharge capacity normalized by each LNSO compound’s
theoretical capacity.
Figures 3-4b and c clearly show that with increasing lithium
content, we achieve both higher specific capacities and higher fractions of theoretical
capacity on discharge.
The greatly improved performance of LNSO-15 is expected, as 15% lithium excess exceeds the calculated 0-TM percolation threshold of 14%. Surprisingly, however,
LNSO-10 also shows significant improvement over LNSO-0 although its lithium excess
level is below the calculated 0-TM percolation threshold. These unexpected results
suggest that further understanding of 0-TM percolation is necessary. To elucidate the
origins of the enhanced electrochemical performance of the lithium-excess LNSO compounds, detailed structure characterization was completed using transmission electron
microscopy (TEM), x-ray diffraction (XRD), density functional theory (DFT), and
extended x-ray absorption fine structure (EXAFS). We focus the structural discussion
in this work on LNSO-15, as it shows the best electrochemical performance among
77
Figure 3-2: SEM shows the LNSO particles to be 50-100nm after ball milling with
carbon at 300rpm for variable times. (a) LNSO-0 for 6 hours (b) LNSO-5 for 3 hours
(c) LNSO-10 for 2 hours (d) LNSO-15 for 1 hour 40 minutes.
the LNSO compounds.
3.5 Structure characterization
3.5.1 Transmission electron microscopy
Figure 3-5a shows the high resolution transmission electron microscopy (HRTEM)
image of pristine LNSO-15 along the [1-10] zone axis. Within a single particle, two
domains with different superstructures are present. To distinguish the two domains,
we look at the Fourier transform (FT) of two different regions marked by the dashed
square and solid square in Figure
3-5a. The FT of regions inside the dashed square
78
Figure 3-3: Galvanostatic cycling of LNSO-0 (no Li excess) and LNSO-15 (15% Li
excess) at 1C between 2.5–4.6V.
is shown in Figure 3-5b.
d
The black crosshatch arrows identify the 1/3 110 streaks
corresponding to a single domain of
√
√
3× 3
ordering.
The streaks are due to
different types of stacking between transition metal layers within the domain, as
√
shown in Figure 3-5e, an enlargement of the HRTEM image.
3×
√
3
ordering, also
known as honeycomb ordering, occurs in LNSO-0 as well as other mixed-transition
metal compounds [124, 139–142]. The local charge is naturally balanced in LNSO-0
by filling two of the sublattices in the
with Sb
5+
√
√
3× 3
2+
superstructure with Ni
and one
.
Having identified LNSO-0 to be one domain within a LNSO-15 particle, we consider the ordering and composition of the second domain. The simplest way for the
remaining Li
+
5+
and Sb
cations to balance the local charge is to order in a
√
3×1
Li-Sb stripe. Schematics of the two types of ordering are shown in Figure 3-6. The
√
superposition of the
3×
√
3
√
and
3×1
orderings is present only in some regions
of the LNSO-15 particle, such as inside the solid square in Figure 3-5a. The FT of
regions inside the solid square is shown in Figure 3-5c. Again, black crosshatch arrows
79
Figure 3-4:
(a) Discharge capacity of the LNSO compounds over 50 cycles at 1C
between 2.5–4.6V. (b) Fraction of theoretical capacity achieved on discharge for the
LNSO compounds over 50 cycles at 1C between 2.5–4.6V.
√
d
point to 1/3 110 streaks corresponding to
d
striped arrows mark 1/2 110 spots, spaced
ordering.
3×
∼3Å
√
3
Ni-Sb ordering. Additionally, red
√
apart, corresponding to
3×1
Li-Sb
The electron diffraction pattern of the whole particle (Figure 3-5d) also
shows the coexistance of the two domains.
√
The
3×1 Li-Sb stripe ordering previously has not been observed as a bulk phase,
80
√
Figure 3-5: HRTEM of pristine LNSO-15 showing the coexistence of
and
√
3×1
Li-Sb domains within the transition metal
d
ab
√
3
Ni-Sb
layer. In (b)-(d), the black
crosshatch arrows label the 1/3 110 streaks corresponding to the
d
3×
√
√
3× 3
ordering,
while the red striped arrows label the set of 1/2 110 spots. (a) HRTEM along the
[1-10] zone axis. The inset shows the intensity line profile along the direction of the
white arrow for one projected
of the inset, where
3d
and
2d
ab
layer. The HRTEM image is enlarged in the bottom
label the tripling or doubling of the
d 110
spacing. (b)
Fourier transform of the region inside the dashed square in (a). (c) Fourier transform
of the region inside the solid square in (a).
particle. (e) Enlarged view of
√
√
3× 3
(d) Electron diffraction pattern of the
ordered regions showing random interlayer
stacking.
81
Figure 3-6: Two types of orderings coexist in the transition metal layer of lithium-
√
√
√
√
√
3 × 3 and 3×1. (a) 3 × 3
√
3×1 stripe ordering between Li+
excess LNSO compounds,
2+
5+
between Ni
and Sb . (b)
honeycomb ordering
5+
and Sb .
though it is equivalent to the Li-vacancy pattern in Li0.5 CoO2 [142, 143]. In LNSO-15,
the
√
3×1 Li-Sb stripe is observed to be mostly one or two periods wide, corresponding
to widths of approximately 3Å and 6Å, respectively. The widest observed stripe is
three periods wide, forming a < 1nm interface. The insert in Figure 3-5a contains
the intensity line scan taken along the white arrow in Figure 3-5a, and shows the
√
3×1 ordering (marked as 2d ) embedded in a region with
√
√
3 × 3 ordering (marked
d
as 3 ). Note that close to the boundary between the two types of orderings, some
lattice spacings are neither 3
d
nor 2
d
(unlabeled spacing in the inset), which may be
a buffer zone needed to connect the two ordered domains.
3.5.2 Synchrotron x-ray diffraction
Superstructure peaks in synchrotron XRD further confirms the presence of both domains in the high lithium-excess LNSO compounds. The strong XRD peaks in all
four pristine LNSO compounds can be indexed to the layered structure with
R 3m
∘
symmetry (Figure S2). The weak superstructure peaks between 19-34 2𝜃 (Figure S2
inserts) derive from ordering in the transition metal layer. In LNSO-0, the five strong
∘
superstructure peaks between 19-34 2𝜃 all arise from
82
√
√
3× 3
ordering, which is
the only ordering present in LNSO-0 [144]. In LNSO-5, these five peaks are also all
present, but are much weaker in intensity. For LNSO-10 and LNSO-15, in addition
to the five weak
√
√
3× 3
∘
peaks, there are two additional peaks at 21.3 and 30.4 2𝜃
that are marked by black arrows in the inserts of Figure S2c and d. These peaks have
d -spacings
of 4.2Å and 2.94Å, respectively, which correspond to doubled spacings of
√
the (104) and (-108) planes in the
on a conventional
R 3m
3×1
domain. The
hkl
planes are defined based
∘
unit cell. The peak at 30.4 2𝜃 corresponding to a
d -spacing of
√
2.94Å agrees with TEM observations of ∼3Å d 110 spacing in the
3×1 stripe domain.
Figure 3-7: Synchrotron x-ray diffraction patterns of pristine (a) LNSO-0, (b) LNSO5, (c) LNSO-10, and (d) LNSO-15. All LNSO compounds form in the layered rocksalt
structure.
The inserts in each pattern show the superstructure diffraction peaks
deriving from ordering in the transition metal layer. Unlabeled superstructure peaks
√
are indexed to the
3×
√
3
honeycomb ordering; superstructure peaks marked by
vertical arrows are attributed to
√
3×1
stripe ordering.
83
3.5.3 Density functional theory
√
The observed structure with
3×
√
3
√
domains separated by
3×1
stripes is further
supported by DFT calculations. Without input from XRD and TEM, we computed
the energies of 108 different orderings for four lithium-excess LNSO compositions.
The lithium excess levels of 9%, 12.5%, 20%, and 28.6% were selected to minimize
the size of the unit cell used in the calculations. All symmetrically distinct orderings
with up to 11 formula units (44 atoms) were considered.
Figure 3-8a shows the convex hull of formation energies on the composition line
between LiNi2/3 Sb1/3 O2 (LNSO-0) and Li(Li1/2 Sb1/2 )O2 (LNSO-50). Each computed
ordering is represented by a red cross.
Although the 0-K formation energy of all
intermediate compositions is positive, the lowest energy orderings are just 3–10 meV
per formula unit above the convex hull. Hence, the driving force for phase separation is small, and several factors which favor intermediate ordering, such as entropy
and lattice parameter constraints from a common matrix, are likely to stabilize the
intermediate orderings.
Figure 3-8b shows the cation ordering within the transition metal layer for each
of the lowest energy computed structures, with the unit cell for each composition
outlined in black.
+
2+
Cations are denoted by blue (Li ), white (Ni ), and orange
5+
(Sb ) circles. LNSO-9, LNSO-12.5, and LNSO-20 contain domains of
Sb honeycomb ordering separated by a single period of
√
3×1
√
3×
√
3
Ni-
Li-Sb stripe ordering.
With increasing lithium content, the Li-Sb stripe occurs more frequently.
At the
highest computed lithium-excess composition of LNSO-28.6, the ordering becomes
inverted,
the
e.g.
√
3×1
Li-Sb orders in the
√
√
3× 3
honeycomb pattern and Ni-Sb orders in
stripe pattern.
Of the computed structures, LNSO-12.5 is closest in composition to LNSO-15.
The computed structure confirms the structure model observed by TEM: regions of
√
3×
√
3
ordering of Ni-Sb are separated by stripes of Li-Sb.
supports the TEM observations of narrow < 1nm Li-Sb stripes.
Furthermore, DFT
For the LNSO-20
composition, we calculated energies of three unit cells assuming different widths of
84
Figure 3-8: (a) Convex hull of formation energies for LNSO compositions with different lithium excess contents. The stable endpoints LiNi2/3 Sb1/3 O2 (LNSO-0) and
Li(Li1/2 Sb1/2 )O2 (LNSO-50) are indicated with filled black circles. Red crosses mark
metastable and unstable configurations. (b) Sketch of the lowest-energy cation orderings in the LNSO
[110]
plane for the four intermediate compostions with unit cells
outlined in black.
the
√
3×1
Li-Sb stripe. As shown in Figure 3-9, increasing the width of the Li-Sb
stripe from one period to two or three periods was found to increase to the energy
of the structure by 20 and 26 meV per formula unit, respectively.
The increase in
energy with increasing stripe width implies that narrower stripe domains of Li-Sb are
preferred and may be stabilized by interfacial energy.
3.5.4 Extended x-ray absorption fine structure
Sb and Ni K-edge EXAFS data were collected for LNSO-15, specifically to look for
√
3×1 stripe domain. A key local feature of this stripe do√
√
the
3 × 3 domain, is the presence of a short-ranged Sb-Sb
evidence of the Li-Sb
main, in contrast to
85
Ni
Li
Sb
0 meV/f.u.
Figure
3-9:
LNSO-20
20 meV/f.u.
cation
orderings
with
26 meV/f.u.
one,
two,
and
three
atom
wide
Li(Li1/2 Sb1/2 )O2 domains (shaded red). The computed (DFT) formation energy per
Li(Li1/5 Ni2/5 Sb2/5 )O2 formula unit (f.u.) increases with growing domain width. Energies are shown relative to the most stable ordering. Unit cells are outlined in black.
correlation. TEM and XRD studies suggest that such a correlation might be present
around
∼3Å.
Initial fits of EXAFS data to identify a Sb-Sb correlation at
√
unsuccessful. However, inspection of the structure of the
∼3Å
were
3×1 stripe generated from
the DFT computation suggested a distribution of distances with a larger, averaged
Sb-Sb distance closer to
∼3.1Å
(Figure 3-10). Using this DFT-computed structure
as input and explicitly including a Sb-Sb correlation at the larger
∼3.1Å
provided
much improved fits. In particular, there is a clear improvement in the visual quality
of the fits, as well as improvements in the R-factor by a factor of
2
reduced-chi-square (𝜒𝛾 ) by
∼30% [145].
∼2
and in the
Structural parameters are reported in Table
3.2, raw EXAFS data showing excellent signal quality is shown in Figure 3-11, and
the fits are plotted in Figure 3-12.
Allowing for the presence of local distortions, the EXAFS results are consistent
with the coexistance of two ordered domains in the lithium-excess LNSO compounds.
TEM, XRD, and DFT all show a
The
∼3.1Å
∼3Å
√
interplanar spacing for the
3×1
domain.
interatomic distance between Sb ions, deduced from EXAFS data, can
86
Frequency of occurrence
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
3.00
3.05
3.10
3.15
Sb−Sb distance (Å)
3.20
Figure 3-10: Distribution of Sb-Sb nearest neighbor distances extracted from all DFT
computed LNSO-12.5 structures. A distance of
∼
3.1 Å was used as input for fitting
EXAFS data.
Figure 3-11:
Raw EXAFS at the LNSO-15 Ni and Sb K-edges showing the high
−1
quality of the data to ∼ 18 Å .
87
Table 3.2: EXAFS structural parameters (coordination number, CN; correlation dis2
tance, R; and mean-squared relative displacement, 𝜎 ) for LNSO-15 extracted by
fitting the first two peaks of the Fourier transforms of the Ni K-edge and Sb K2
edge. The many body amplitude reduction factor (S0 ) was fixed to be 0.83(4) for Ni
and 1.03(3) for Sb; these values were determined by fits to stoichiometric LNSO-0.
Possible Ni-Li antisite disorder was not explicitly included.
The number(s) in the
parenthesis is the uncertainty in the final digit(s). The Ni and Sb EXAFS data were
simultaneously fit.
−4
2
(10
Å )
Shell
CN
R (Å)
𝜎2
Ni-O
6
2.076(4)
47(2)
Ni-Ni
2.4(8)
2.996(9)
40(30)
Ni-Sb
2.8(3)
2.988(9)
41(5)
Sb-O
6
1.990(3)
36(3)
Sb-Ni
3.2(4)
1,4
0.75
2.988(9)
1
2
1
Sb-Sb
3
2
3.109(20)
3
41(5)
4
43(20)
1 Fixed
at provided values in the final fits
bond distance constrained to be the same from both Ni and Sb viewpoints
3 Disorder parameter constrained to be the same from both Ni and Sb viewpoints
4 Sb-Sb CN (0.75) and R (∼3.10Å) were included using the DFT computed structure as a guide
2 Ni-Sb
be accomodated well if the Sb ions are offset from the ideal site positions inside two
∼3Å interplanar spacing.
√
of the
3×1 domain.
neighboring Sb stripes with
the nano-interface nature
The site offsets may be due to
3.6 Nanohighways in Li𝑥Ni2−4𝑥/3Sb𝑥/3O2
Having built a structure model for the lithium-excess LNSO compounds, we now
describe how 0-TM diffusion channels percolate in these materials.
As mentioned
above, two coexisting nanoscale domains were observed in the 10% and 15% lithiumexcess LNSO compounds. Figure 3-13 shows a schematic of the microstructure with
√
3×
√
3
√
domains divided by
3×1
domains.
The interface between the two do-
mains is highlighted in green. The significance of this two-domain microstructure is
many-fold. 0-TM diffusion channels form at the interface of these two domains on
the nanoscale and provide low diffusion barrier pathways for the lithium, effectively
serving as nanohighways for lithium diffusion. Lithium now only needs to diffuse from
88
Figure 3-12: Fits to the LNSO-15 Sb K-edge show improvement with inclusion of
the DFT-computed Sb-Sb correlation at 3.1Å, as highlighted by the black box. The
dotted lines denote the fit range covered. (a) Fit without Sb-Sb correlation included
2
(𝜒𝛾 = 514, R-factor = 0.0048). (b) Improved fit with Sb-Sb correlation included
2
(𝜒𝛾 = 339, R-factor = 0.0022)
89
√
inside the
3×
√
3
√
domains to the interface with the
3×1
domain, where it can
join the percolating 0-TM path.
Figure 3-13: Schematic illustration showing coexistance of the two types of ordering
and 0-TM diffusion channels at the domain interfaces.
For clarity, the projected
position of Li𝐿𝑖 is only shown in the interface of regions of both domains.
The segregated microstructure also lowers the threshold for 0-TM percolation.
The two-domain microstructure allows for percolation to be achieved through interconnected nanohighways rather than homogeneous percolation through the bulk.
Because TEM and DFT suggest that the Li-Sb stripe domain only forms as < 1nm
stripes, increasing lithium is not incorporated as wider
√
3×1
Li-Sb domains, but as
a larger number of < 1nm stripes. Beyond a certain lithium excess level, these stripes
form a percolating network of 0-TM diffusion channels.
In general, percolation is
achieved at lower volume fractions when the aspect ratio of the percolating object
becomes larger [146, 147]. Hence, such nanohighway microstructuring is an effective
90
way to achieve percolation of 0-TM diffusion channels, even at low lithium excess
concentrations.
3.7 Conclusion
In conclusion, a new family of cathode materials, the lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2
compounds, were designed from percolation theory, synthesized, and tested for electrochemical activity for the first time.
We hypothesized that excess lithium would
improve electrochemical performance through increased percolation of 0-TM diffusion
channels. In agreement with theory, we found increasing lithium content to improve
both capacity and cyclability, with the best performing compound, LNSO-15, achieving 132 mAh/g at 1C at a discharge voltage of 3.9V. While further work is needed to
understand phenomena that limit capacity and further engineer these materials, the
improvement of electrochemical performance with increasing lithium content validates
our materials design strategy.
Through a combination of TEM, XRD, DFT, and EXAFS, we solved the structure
of the new lithium excess LNSO compounds. Additional lithium is incorporated in
√
lithium excess LNSO compounds as < 1nm
3×1 Li-Sb stripes subdividing
√
√
3× 3
Ni-Sb domains. 0-TM diffusion channels form at the interface of the two domains, acting as nanohighways for lithium diffusion. The lithium excess LNSO compounds show
that 0-TM percolation is possible at lower lithium excess levels through patterned,
low-dimensional, lithium-rich domains.
Additionally, because 0-TM percolation is
achieved without disordering materials, we maintain the advantage of high and fairly
flat voltage in layered materials. This work extends upon our previous understanding
of 0-TM diffusion channels in disordered materials, demonstrating that 0-TM percolation can be achieved in ordered materials, and laying the foundation for future
cathode materials design.
91
92
Chapter 4
Layered Li𝑥Ni2−4𝑥/3Sb𝑥/3O2:
Understanding the origins of higher
capacities at faster rates
4.1 Introduction
In Chapter 3, we introduced the layered lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 (LNSO)
materials, a new family of cathode materials for lithium-ion materials. The design
strategy for these materials proposed by Lee and Urban [4, 5] was to increase lithium
content and enhance lithium diffusion through percolation of low barrier lithium diffusion channel. These low barrier lithium diffusion channels, termed 0-TM channels,
are described in Section 1.4.
The lithium-excess LNSO materials show greatly improved discharge capacity and
cyclability over their stoichiometric counterpart. However, despite being an exciting
proof-of-concept, the experimental capacities of the LNSO materials are far below
their theoretical capacities, and require further understanding on which phenomena
limit capacity. Current demonstrated capacities in the LNSO materials cannot compete with capacities of other established lithium-excess transition metal oxides.
In this chapter, the redox mechanisms of four lithium-excess LNSO materials
93
with 0-15% lithium excess are studied as a function of lithium content and rate.
n
Hereafter, we refer to these materials as LNSO- , where
of excess lithium.
n
represents the amount
We show that an oxygen loss plateau emerges with increasing
lithium content, similar to other lithium-excess layered oxides. More interestingly, at
the highest lithium excess level, LNSO-15, the oxygen loss plateau disappears with
increased charge rate, and we observe
higher discharge capacities at faster rates.
From
careful characterization of nickel, antimony, and oxygen redox activity, we determine
which redox processes are active at slow and fast rates, and propose nickel migration
as an explanation for the anomalous rate behavior.
4.2 Experimental methods
4.2.1 Synthesis
The LNSO materials were synthesized by solid state reaction, as described in Section 3.3.1. Electrode films were made with LNSO, carbon, and polytetrafluoroethylene (PTFE) binder in a ratio of 80:15:5 by weight. Swagelok cells for galvanostatic
cycling and galvanostatic intermittent titration technique (GITT) measurements were
assembled in an argon-filled glovebox using LNSO as the cathode, Li metal as the
anode, Celgard 2320 as separator, and 1M LiPF6 in EC:DMC as electrolyte. Cells
were cycled on a Maccor 2200 or Maccor 4000 operating at room temperature. For
the GITT measurement, a C/20 current pulse was applied for 30 minutes followed by
an OCV step of 10 hours.
4.2.2 Ex situ experimental characterization
Cells for
ex situ
measurements were charged or cycled in Swagelok cells using 5/16”
diameter films, then diassembled in a glovebox to minimize air exposure.
Samples
for electron energy loss spectroscopy (EELS) and high resolution transmission electron microscopy (HRTEM) were lightly sonicated in DMC in a glovebox to disperse
the cathode film into small particles; EELS spectra and HRTEM images were then
94
collected on a JEOL 2010F. Cathode films for
ex situ
x-ray absorption near edge
structure (XANES) were cut in half and taped between two pieces of Kapton tape
in a glovebox. XANES measurements were taken quickly after extraction from inert
atmosphere.
XANES spectra were collected in transmission mode at the Argonne
National Laboratory Advanced Photon Source, beamline 20-BM-B.
4.2.3 In situ X-ray diffraction
In situ
x-ray diffraction (XRD) patterns were obtained in 40 minute intervals from
a Bruker D8 Advance Da Vinci Mo-source diffractometer scanned between 8–30
∘
2𝜃 ,
and then converted to a 2𝜃 range corresponding to a Cu source. The cell was cycled
in galvanostatic mode on a Solatron SI12287 in the voltage window of 2.5–4.6V at a
current rate of C/20. Rietveld refinement on the XRD patterns was completed using
PANalytical High Score Plus.
Figure 4-1 shows the configuration of the custom
in situ
cell. The top stainless
steel plate has a cutout for x-ray penetration, and is tapped at the four corners for
screws.
On the inside of the top plate, a shallow ledge supports a rubber O-ring,
followed by the beryllium window. The beryllium window and bottom stainless steel
plate provide pressure to the cell stack, which is contained between a second O-ring.
The cell stack consists of current collectors, a cathode film, two pieces of glass fiber
separator, lithium foil, and a rubber spacer. The interior of both stainless steel plates
are taped with Kapton tape to prevent a short-circuit of the cell, and current collectors
with leads are taped to the plates.
The x-ray beam enters through the beryllium window, and reflects to the detector.
To maximize intensity of diffraction peaks from the cathode, the cathode is situated
immediately behind the beryllium window of the top stainless steel plate, and a higher
loading (thicker film) is used for the measurement. Additionally, the cell components
exposed to the beam are minimized to the extent possible. For example, thin metal
foils are used as current collector leads, and center cutouts are made in the current
collector to minimize copper, aluminum, or stainless steel signal.
95
Figure 4-1: Configuration of a custom
in situ
cell for x-ray diffraction
4.2.4 In situ on-line electrochemical mass spectroscopy
In situ
gas analysis by on-line electrochemical mass spectroscopy (OEMS) was per-
formed for LNSO-15/Li metal and LNSO-0/Li metal half-cells. The OEMS test cells
were assembled in custom cell hardware [6, 148] using a Li metal counter electrode,
two glass fiber separators, and 1M LiPF6 in EC:DMC as electrolyte. The OEMS setup
is previously described in detail [6]. The gas evolution during charge was measured
∘
at 25 C, with cells first held at OCV for two hours, then galvanostatically charged
at a rate of C/20 to a cutoff potential of 4.6V. Conversion of the mass spectrometer
currents to concentrations was done for oxygen, carbon dioxide, and carbon monoxide
using a calibration gas (H2 , O2 , CO, and CO2 at 2000 ppm each in argon, Westfalen
AG). A schematic of the OEMS system [6] is shown in Figure 4-2.
4.3 Electrochemical properties
4.3.1 Performance as a function of lithium content
The voltage versus specific capacity plots for the four LNSO compounds cycled at
C/20 are shown in Figure 4-3a, in order of increasing lithium content from top to
96
Figure 4-2: Schematic of a on-line electrochemical mass spectrometer (OEMS) with
a battery cell connected directly to a mass spectrometer. All gas products evolved
are continuously sampled. [6]
bottom.
Although the LNSO compounds have different theoretical capacities, the
plots are stacked vertically on the same
x -axis
to visually highlight trends in first
charge, capacity retention, and discharge voltage profiles. Table 4.1 lists the theoret-
2+/4+
ical capacities of the LNSO compounds calculated from the Ni
redox couple.
Table 4.1: Theoretical capacities of Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds calculated from
2+/4+
the Ni
redox couple.
Sample
Formula
Theoretical Capacity
LNSO-0
Li1.00 Ni0.67 Sb0.33 O2
226 mAh/g
LNSO-5
Li1.05 Ni0.60 Sb0.35 O2
240 mAh/g
LNSO-10
Li1.10 Ni0.53 Sb0.37 O2
247 mAh/g
LNSO-15
Li1.15 Ni0.47 Sb0.38 O2
219 mAh/g
The first charge of LNSO-0 shows a single plateau at 4V, while the first charge of
the lithium-excess LNSO compounds shows a plateau at 4V and one at 4.45V. The
2+/4+
4V plateau accounts for approximately half the theoretical Ni
redox capacity of
each compound, and the length of the 4.45V plateau increases with increasing lithium
content. Charge capacity obtained along the 4.45V plateau is absent in discharge.
97
Figure 4-3: (a) From top: Galvanostatic cycling of LNSO-0, LNSO-5, LNSO-10, and
LNSO-15 at C/20 between 2.5–4.6V. (b) Fraction of theoretical capacity achieved on
discharge over ten cycles at C/20 between 2.5–4.6V. Capacity retention improves with
increasing lithium content, and all lithium excess samples access a higher fraction of
their theoretical capacities than LNSO-0.
98
Figure 4-3b shows the C/20 discharge capacities plotted as a function of cycle
number.
Because each LNSO compound has a different theoretical capacity, the
discharge capacity is normalized by dividing the experimental discharge capacity by
the theoretical capacity. Compared to LNSO-0, the lithium-excess LNSO compounds
show two noticable improvements. First, the fraction of theoretical capacity accessed
on discharge by all the lithium-excess LNSO compounds is much higher.
Second,
increasing lithium content reduces capacity fade, leading to flatter slopes in Figure 43b.
One final feature in Figure 4-3a to note is the small voltage step at 2.6V on
discharge.
The LNSO-0, LNSO-5, and LNSO-10 compounds show this 2.6V step
to evolve over ten cycles, becoming most pronounced in the tenth cycle. The 2.6V
feature is less pronounced with increasing lithium content, and is not observed at all
in the LNSO-15 voltage curves.
At present, we do not know what phenomena are
responsible for this voltage step.
4.3.2 Performance as a function of rate
At faster rates, the LNSO compounds containing 0-10% lithium excess exhibit typical
rate behavior: higher C-rates lead to lower capacities and increased polarization. In
contrast, LNSO-15 shows the opposite, highly unexpected behavior between C/20
and 1C. Upon charge, all cells between C/20 and 1C reach approximately the same
final charge capacity, but with increasing rate leading to a deferral of the 4.45V
plateau. Upon discharge, as the rate increases from C/20 to 1C, the discharge capacity
increases.
This increase in discharge capacity increases the first-cycle coulombic
efficiency between C/20 and 1C. Only when the rate is increased to 2C and 5C do
both charge and discharge capacity decrease, which is likely due to traditional polarization effects. With the 2C and 5C cells, we observe that the charge voltage is
increased and the discharge voltage is depressed. The first cycle voltage curves for
LNSO-15 obtained at C/20, C/10, C/5, 1C, 2C and 5C are overlayed in Figure 44a. The C-rate dependence of the discharge capacity is summarized in Figure 4-4b.
Pristine cells are used for each rate test.
99
Figure 4-4: (a) First cycle voltage curves of LNSO-15 obtained from galvanostatic
cycling between C/20 and 5C. At faster C-rates, the 4.45V charge plateau disappears,
and higher discharge capacities are achieved. (b) Discharge capacity over ten cycles
for LNSO-15 at rates between C/20 and 5C. Capacity increases increases from C/20
to 1C, then decreases between 1C and 5C.
The anomalous rate behavior of LNSO-15 motivates the remainder of this work.
First, we conducted additional electrochemical tests to characterize the 4.45V plateau
and understand its relationship to reversible capacity. Second, using a range of char-
100
acterization techniques, we determine which redox processes are active at fast and
slow rates.
4.4 Characteristics of the 4.45V plateau
Rate tests on LNSO-15 suggest that the processes occurring along the 4.45V plateau
are slow, as this plateau disappears upon higher C-rates. This observation is further
supported by galvanostatic intermittent titration technique (GITT) measurements,
shown in Figure 4-5. Small overpotentials are observed along the first 4V plateau,
indicating a faster process, and larger overpotentials are observed along the second
plateau, indicating a slower process. The overpotentials from the GITT voltage curve
agree with the C/20 galvanostatic voltage curve.
The last few steps of the GITT charge yield very little capacity. Ultimately, the
2+/4+
charge capacity does not exceed the theoretical capacity calculated from the Ni
redox couple although there is plenty of lithium remaining in LNSO-15 to extract.
This is in contrast to other lithium-excess materials which continue to charge along
a 4.5V plateau beyond their theoretical capacities [28, 149].
The rate tests and GITT measurement suggest that the 4.45V plateau is a kineticallylimited process, and that higher capacities can be obtained if we bypass this process.
Indeed, the 4.45V plateau can be bypassed until
the first slow charge
of the cell. Fig-
ure 4-6 shows the voltage curves where the first charge is at 1C, and all subsequent
discharges and charges are at C/20. The 4.45V plateau appears on the second charge.
The 4.45V plateau can be delayed over extended cycles, suggesting that prior time
spent at high voltage does not affect the appearance of the 4.45V plateau. Figure 47a shows the voltage curves of a cell where the first twenty cycles are at 1C, but
the 21st charge has been slowed to C/20. While the first twenty cycles bypass the
slow process, the characteristic 4.45V plateau appears on the 21st charge, although
its voltage is slightly pushed up.
Thus, prior time spent at high voltage does not
affect the appearance of the 4.45V plateau. A cell charged at C/20 in its first cycle
spends equivalent time at high voltages as a cell cycled 20 times at 1C. The plateau
101
Figure 4-5:
GITT measurements on LNSO-15 show a low overpotential along the
first charge plateau, and a larger overpotential along the second charge plateau. The
4.45V overpotential of the second plateau agrees well with voltage curves obained by
galvanostatic cycling.
The larger overpotential along the second plateau implies a
slower process at 4.45V.
Figure 4-6: Galvanostatic cycling with the first charge conducted at 1C, and first
discharge at C/20. All subsequent cycles are at C/20. The 4.45V plateau appears on
the second charge.
102
Figure 4-7: (a) Galvanostatic cycling of LNSO-15 at 1C for the first twenty cycles.
On the 21st cycle, the charge current is reduced to C/20, and a second plateau
is observed at 4.6V. The second plateau is observed only on the first slow charge,
regardless of previous time spent at high voltages during cycling.
(b) Charge and
discharge capacity versus cycle number.
still appears on the first slow charge of the cell. This phenomena can be seen more
clearly in Figure 4-7b, where the specific charge and discharge capacities are plotted
as a function of cycle number, and we observe a spike in charge capacity on the slow,
21st charge.
4.5 Characterization of nickel redox activity
The rate-dependent voltage curves of LNSO-15 motivate further characterization on
the redox processes active at fast and slow rates. We first investigate the redox activity
of nickel with x-ray absorption near edge structure (XANES), and look at the nickel
103
K-edge position to determine its oxidation state.
Ex situ
XANES measurements were
made for three LNSO-15 samples: pristine, charged to 4.6V at 1C, and charged to
4.6V at C/20. Reference measurements for the edge positions of Ni
made on LNSO-0 and NaNiO2 , respectively.
of the
ex situ
2+
3+
and Ni
were
Figure 4-8a shows the voltage curves
charged LNSO-15 samples, and Figure 4-8b shows the nickel K-edge
positions from XANES of the reference samples and LNSO-15 samples.
Figure 4-8:
Voltage curves and
ex situ
XANES spectra of the nickel K-edge for
pristine and fully-charged LNSO-15 samples charged at 1C and C/20. Reference edge
2+
3+
positions for Ni
and Ni
are obtained from NiO and NaNiO2 . The edge positions
of the charged LNSO-15 samples show that for both rates, nickel is oxidized to a
2+
3+
similar valence state between Ni
and Ni .
Ex situ
XANES confirms that pristine LNSO-15 contains Ni
2+
; the edge position
aligns well with the LNSO-0 reference. Upon charging, we expect nickel to oxidize.
104
The 1C sample reaches 117 mAh/g at end of charge, and the C/20 sample reaches
185 mAh/g. Both charge capacities fall between the one-electron and two-electron
theoretical capacities of 109.5 mAh/g and 219 mAhg, calculated from Ni
Ni
2+/4+
2+/3+
and
, respectively. Surprisingly, despite a difference of 70 mAh/g and both exper-
imental capacities exceeding the one-electron theoretical capacity, the nickel K-edge
positions are nearly identical and lie between the Ni
2+
and Ni
3+
standards. We cannot
3+
say what the exact valence of Ni is, but it clearly does not exceed Ni .
To support the XANES measurements, we also characterize the nickel valence
state using electron energy loss spectroscopy (EELS). In EELS measurements, the
L3/L2 ratio of the transition metal edges is used as a measure of the oxidation state
of the metal [150].
Ex situ
EELS measurements on the nickel K-edge were made for
pristine LNSO-15, LNSO-15 charged to 69 mAh/g at 1C and C/20 through the end
of the first plateau, and LNSO-15 fully charged to 4.6V at 1C (138 mAh/g) and C/20
(158 mAh/g).
The voltage curves for the half-charged and fully charged LNSO-15
samples are shown in Figure 4-9a.
Figure 4-9b plots the L3/L2 ratio of the Ni K-edge for the LNSO-15 samples
versus reference samples.
Ni
2+
From XANES, we know that pristine LNSO-15 contains
; hence, a thin horizontal bar is drawn next to pristine LNSO-15 to represent
the Ni
2+
3+
reference. The wider horizontal bar represent the Ni
reference, obtained
from EELS measurements on NaNiO2 .
The L3/L2 ratios obtained for the charged
LNSO-15 samples are given with error bars.
The EELS measurements agree with
the findings from XANES: Ni oxidizes to similar states at half charge and full charge
3+
regardless of C-rate, but does not oxidize past Ni .
4.6 Characterization of oxygen loss
4.5V plateaus in many lithium-excess materials have been extensively characterized,
and attributed to a range of phenomena including spinel formation and oxygen loss
[42, 151–156]. Based on these prior studies, we hypothesized that the 4.45V plateau
in lithium-excess LNSO materials may be due to oxygen loss, although the plateau
105
Figure 4-9: Voltage curves and L3/L2 ratio of the nickel K-edge obtained from
ex-situ
EELS measurements for pristine, half-charged, and fully-charged LNSO-15 samples
2+
3+
charged at 1C and C/20. Reference L3/L2 ratios for Ni
and Ni
are represented
by horizontal bars. At the end of charge for both rates, nickel is oxidized to a similar
2+
3+
valence state between Ni
and Ni .
length is much shorter than in other materials. Lu and Dahn [157] were one of the
first to correlate changes in structure with oxygen loss phenomena, studying lithiumexcess Ni-Mn oxides with
in situ
x-ray diffraction (XRD). From diffraction patterns
taken along an electrochemical curve, changes in lattice parameters can be correlated
to voltage curve features. Some evidence for transition metal undercoordination was
also seen in prior extended x-ray absorption fine structure (EXAFS) studies [158].
The top graph in Figure 4-10a shows the first charge voltage curve of LNSO-15
obtained during
in situ
XRD at C/20.
Black dots along the voltage curve mark
the points where XRD patterns were obtained.
106
The middle and bottom graphs in
Figure 4-10:
In situ
XRD of LNSO-15 at C/20. (a) First charge voltage curve,
lattice parameter evolution, and
c -lattice
a-
parameter evolution. (b) (003) and (104)
peak evolution over the first charge.
107
Figure 4-10a show
a -lattice
and
c -lattice
parameter evolution. Figure 4-10b shows
the evolution of the (003) and (104) peaks during first charge.
The shifting (003) and (104) peaks, characteristic peaks of the layered structure, show that LNSO-15 evolves as a single phase. There is no peak splitting observed, in contrast to observations by Koga et al. [159, 160] made on lithium-excess
Li1.20 Mn0.54 Co0.13 Ni0.13 O2 . The
c -lattice parameter of LNSO-15 initially expands due
to increased repulsion between oxygen atoms surrounding the vacated lithium layer,
and then contracts at higher levels of delithiation. The
a -lattice parameter decreases
linearly in the first part of charge as a result of the oxidation of the transition metal,
which reduces its radius [157, 161]. At the end of charge along the 4.45V plateau, the
a -lattice parameter is constant.
The changes in the
c
and
a
lattice parameters result
in a 4% volume decrease over the first charge.
OEMS measurements are conducted for two LNSO materials:
LNSO-0, which
does not contain lithium excess and does not show a 4.45V plateau, and LNSO-15,
the 15% lithium excess material that shows the 4.45V plateau. Working under the
hypothesis that the 4.45V plateau is correlated to oxygen loss, the LNSO-0 sample
acts as a control sample that should not show oxygen loss.
Figure 4-11 shows the first charge voltage curves of LNSO-0 and LNSO-15 obtained
with
in situ
OEMS, and the corresponding concentration of gas species in the cell in
thousands of ppm. The ppm quantity can be converted to moles using the volume
of the OEMS cell.
In both LNSO-0 and LNSO-15, CO2 evolution begins at 4.2V,
but only in LNSO-15 do we also see an O2 signal at the start of the 4.45V plateau.
The O2 signal is multiplied by 10x to be plotted on the same scale as the CO2 signal.
While small, the O2 signal is significant.
Finally, HRTEM images were taken to highlight differences in surface evolution
after cycling. Figure 4-12 shows HRTEM images of LNSO-15 cycled at C/20 and 1C.
The white dashed lines mark the boundary of surface regions that have undergone
structural modification from the layered structure to more three-dimensional like
structures.
108
Figure 4-11: Voltage curves and gas evolution measurements from
in situ
OEMS of
(a) LNSO-0 and (b) LNSO-15, both charged at C/20. Only LNSO-15 shows oxygen
evolution starting at 4.4V. The O2 signal is multiplied by 10x to be plotted on the
same scale as the CO2 and CO signals.
4.7 Assessment of rate-dependent redox activity
The electrochemical performance of LNSO-15 inspires two questions. First, how can
a material show higher discharge capacities at higher rates?
Second, how can a
sample cycled at high rate, presumably deviating more from equilibrium, show lower
voltages at any point of charge than a sample cycled at low rate? The different voltage
curves accessed at slow and fast rates can only be explained by the LNSO-15 material
traversing different pathways at different rates. In particular, the electrochemical data
suggests that the improved rate performance of LNSO-15 at higher rates is linked to
the bypass of the 4.45V plateau.
From electrochemical tests, we know that the 4.45V plateau appears with increasing lithium content in the LNSO chemistry. Similar analogies in the literature
can be found in other chemistries where a lithium-excess compound shows a plateau
at
∼4.5V,
but its stoichiometric equivalent does not [149, 157]. We also know that
109
Figure 4-12: HRTEM comparing bulk and surface structures of LNSO-15 after cycling
at 1C and C/20
LNSO-15’s 4.45V plateau is a kinetically-limited process that leads to irreversible
capacity loss on discharge. Higher capacities can be obtained if we bypass this 4.45V
process, and the process can indeed be bypassed until the first slow charge of the
cell. To determine what occurs along the 4.45V plateau and what redox processes are
active at different rates, we step through evidence for all possible redox reactions in
LNSO-15: nickel redox, oxygen loss, and oxygen redox.
Based on nickel valence measurements, nickel redox activity cannot explain the
rate-dependent voltage curves of LNSO-15.
Ex situ
XANES and EELS measurements
on nickel taken at half charge and full charge at 1C and C/20 are in agreement.
Although the 1C and C/20 samples achieve different capacities at the end of charge
110
and show different voltage curve features, nickel is oxidized to a similar valence state
2+
3+
between Ni
and Ni .
3+/4+
The limited oxidation of nickel implies that the Ni
redox couple may not be accessible, contrary to our assumption in our initial material
design. We also confirm that nickel is the only redox-active metal in the LNSO system.
XANES on the antimony K-edge shows that no changes to antimony are observed
either in charged or cycled samples, confirming that Sb
5+
is electrochemically inactive.
Having determined there is no antimony activity and limited nickel oxidation,
we consider evidence for oxygen loss. Characterization on the C/20 charge with
situ
XRD indirectly suggests oxygen loss to occur along the 4.45V plateau.
XRD shows the
a -lattice
in
In situ
parameter to decrease linearly at the beginning of charge,
corresponding to transition metal oxidation and shrinking ionic radii. Along the 4.45V
plateau, the
a -lattice
parameter is constant, implying there is no longer transition
metal oxidation and that capacity instead comes from oxygen loss. Since the initial
study by Lu and Dahn correlating constant
a -lattice parameter along a 4.5V plateau
with oxygen loss [157], similar observations have been made in other lithium-excess
materials [162].
Direct detection of oxygen gas by electrochemical mass spectrometry is another
proven method to characterize oxygen loss [45–47, 163–166].
OEMS shows oxygen
evolution in LNSO-15 to correlate with the second plateau starting at 4.4V. However,
unlike previous reports, the O2 levels are secondary to the CO2 levels by an order of
magnitude. While O2 evolution cannot account for all the capacity observed along
the 4.45V plateau, the amount is significant enough that it cannot be ignored as an
artifact. When considered in context of oxygen loss in lithium-excess materials, it is
also not surprising that oxygen loss occurs along the 4.45V plateau in LNSO-15.
The majority of the CO2 evolution likely results from direct decomposition of the
electrolyte catalyzed by nickel, although some contributions to CO2 and CO may
be attributed to evolved oxygen reacting with the electrolyte.
This hypothesis is
supported by our observation of two distinct voltage thresholds for CO2 (4.2V) and
O2 (4.4V) evolution, as well as significant CO2 released in LNSO-0, which does not
show a 4.45V plateau.
The OEMS measurements on LNSO-0 illustrates that even
111
in a sample that does not release O2 , high levels of CO2 are still released starting at
4.2V. While this voltage is lower than expected, different chemistries are known to
show a range of electrolyte decomposition voltages [167]. Notably, LiNiO2 has a very
low decomposition voltage of 4.2V [168].
Unfortunately, due to experimental restrictions, neither
in situ
XRD nor
in situ
OEMS measurements could be conducted at high C-rates for LNSO-15. XRD patterns
require 40 minutes to collect, and thus could not take an accurate snapshot of any
structure along a 1C voltage curve charging in one hour.
With OEMS, the high
loading of the cathode films required for the OEMS experiment precluded high rate
tests from being meaningful in analysis, as significant polarization on the materials
resulted in very low capacities. However, surface changes in cycled samples suggest
that oxygen loss may indeed occur at both slow and fast rates. The HRTEM images
in Figure 4-12 show that the surface structure has transformed in both the C/20
and 1C cycled samples from the original layered features into more 3-dimensional like
structures, possibly by transition metal migration [154]. Structure reorganization is
known to accompany oxygen loss, which we presume to occur only from the surface
[151, 159, 160, 165, 169].
4.8 Quantification of capacity contributions
In the LNSO system, only nickel and oxygen oxidize with delithiation. Nickel shows
limited oxidation to somewhere between Ni
LNSO-15 calculated from the Ni
2+/3+
2+
and Ni
3+
. The theoretical capacity of
redox couple is 109.5 mAh/g. Therefore, based
on the observed oxidation state of nickel in LNSO-15, nickel can only account for less
than 109.5 mAh/g of reversible capacity.
From
ex situ
XANES, we conclude that antimony has no contributions to observ-
5+
able capacity. It is an electrochemically-inactive spectator, remaining Sb
throughout cycling. Antimony’s inactivity explains why we observe irreversible capacity loss
after oxygen loss along the 4.45V plateau. Because antimony cannot be reduced, we
cannot activate a second transition metal upon discharge after oxygen loss. This is
112
in contrast to lithium-excess manganese-containing materials, which show manganese
reduction on discharge after oxygen evolves on the first charge [30, 35].
We now consider oxygen’s contributions to capacity. At end of charge, LNSO-15
releases 150 ppm of O2 , which is equivalent to 6E-8 moles. The detected 6E-8 moles
of O2 is far less than the theoretical maximum of 0.0815 moles of evolved O2 . The
theoretical maximum is calculated assuming all 38 mAh/g from 4.4V, the onset of O2
evolution, is due to oxygen loss. The 38 mAh/g of capacity along the 4.45V plateau
corresponds to 0.163 Li being deintercalated and a maximum of 0.0815 moles of O2
evolved. Because the detected O2 is so low, we believe oxygen loss can only account
for a small part of capacity, even with the assumption that some O2 has reacted with
electrolyte.
Having quantified nickel, antimony, and oxygen loss capacity contributions, significant capacity remains unaccounted for. To summarize the known redox activities
in LNSO-15: Nickel shows limited oxidation between Ni
2+
3+
and Ni
when charging
to 4.6V at both slow and fast rates. Antimony is inactive and has no contributions
to capacity.
Oxygen loss occurs on the first slow charge starting at 4.4V. We now
consider the only remaining redox process that could contribute to capacity: oxygen
redox from O
2−
−
to O .
While experiments remain to be completed to confirm oxygen redox, and a reliable
means of calculating the oxygen redox voltage has yet to be established, we believe
2−
−
that oxidation of O
to O can explain the remaining capacity observed in LNSO-15.
Given the observation of oxygen loss at 4.4V, it is likely that oxygen redox occurs
at a similar voltage and is within the electrochemical cycling window. Additionally,
reversible oxygen redox has been suggested in other lithium excess materials as a
mechanism with good coulombic efficiency [47, 160, 170, 171]. Thus, we propose the
following redox sequence for charge: nickel is oxidized first from Ni
2+
3+
to Ni , followed
by oxygen oxidation, which occurs as both oxygen loss on the surface and oxygen redox
in the bulk. Given limitations of
in situ
techniques, oxygen loss is only confirmed at
slow rates, but likely occurs to a lesser extent at fast rates.
113
4.9 Proposed effect of nickel migration on reversible
capacity
Careful characterization on LNSO-15 sheds light on the active redox processes, but
the questions posed at the beginning of the discussion remain unanswered. Because
similar oxidation processes are observed at slow and fast rate, the rate-dependent
voltage curves cannot simply be explained by competition between different redox
processes. Redox processes cannot explain why LNSO-15’s voltage curves diverge at
fast and slow rates, resulting in the bypass of the 4.45V plateau and higher discharge
capacities at faster rates. At slow rate, a critical change in the material occurs by
mid-charge after the 4V plateau. It is important to note here that oxygen loss, while
observed at slow rates, occurs
after
this critical point, and thus cannot be the cause
of the divergent rate behavior.
To reconcile the observed electrochemical behavior, we propose that
gration
nickel mi-
is the rate-dependent critical process and dictates the reversible capacity of
LNSO-15 in the following manner:
1. In the initial charge along the 4V plateau, delithiation is accompanied by oxi-
2+
3+
dation of nickel ions in the transition metal layer from Ni
to Ni
(a) With slow charge, an increasing number of lithium site vacancies allow
2+
some Ni
ions to migrate from the transition metal layer to stablize the
2+
delithiated structure. Ni
migrates either to tetrahedral sites or to octahedral sites in the lithium layer to lower its site energy.
(b) With fast charge, few Ni
2+
ions migrate from the transition metal layer,
and the material remains well-layered.
2. In the second half of charge from
Ni
3+
∼4.1V–4.6V,
2+
nickel oxidation from Ni
to
occurs concurrent with oxygen oxidation in the form of oxygen loss at the
surface (O
2−
→
2−
−
O2 (g)) and oxygen redox in the bulk (O → O )
(a) With slow charge, the 4.45V plateau accompanies oxidation of the
114
migrated
Ni
2+
.
3+
Once oxidized to Ni , nickel will prefer octahedral sites in the
lithium layer, and further migration becomes energetically unfavorable
(b) With fast charge, because little Ni
2+
migration has occurred, delithiation
continues with oxidation of nickel ions in the transition metal layer, hence
a longer 4V plateau
3+
2+
3. On discharge, lithiation is accompanied by reduction of Ni
to Ni
and reduction of oxygen from O
−
to O
2−
3+
(a) Following a slow charge, a significant number Ni
ions becomes “stuck” in
3+
the migrated sites. Any migrated Ni
cannot move back to the transition
2+
3+
metal layer until reduced to Ni . At the same time, not all migrated Ni
can be reduced, as the migrated ions block all neighboring vacant lithium
sites for intercalation
(b) Following a fast charge, because little nickel migration has occurred, more
lithium can be reintercalated as all sites are available, yielding higher discharge capacities
Although further experiments are needed to confirm nickel migration in LNSO15 [114, 172], nickel-containing cathode materials, including LiNi2/3 Sb1/3 O2 [124], are
known to be prone to nickel migration and cation mixing [26, 114, 157, 172–175]. Having thoroughly characterized redox activity at slow and fast rate, we believe that nickel
migration can explain the anomalous rate behavior of LNSO-15 and is substantiated
by the literature.
2+
We first consider nickel’s migration tendencies and site preferences. Ni
has low
migration barriers due to its low valence state, and can occupy both tetrahedral and
octahedral sites in close-packed crystal structures [124,172,176]. In contrast, Ni
Ni
4+
3+
and
have much higher migration barriers [54], and only prefer to occupy octahedral
sites [114, 177].
2+
Pristine LNSO-15 contains only Ni . Until appreciable delithiation occurs, few
vacancies on the lithium sites may be available to facilitate nickel migration. Thus,
115
the critical point where fast and slow voltage curves diverge occurs
has traversed the 4V plateau and
∼0.3Li
after
LNSO-15
has been extracted. At this critical point
2+
halfway through charge, LNSO-15 contains both lithium vacancies and Ni
ions.
2+
At high rate, nickel migration occurs at a slower rate than delithiation; thus, Ni
3+
largely remains in octahedral sites in the transition metal layer and is oxidized to Ni
before it has a chance to migrate. In contrast, gradual delithiation at slow rate allows
sufficient time for Ni
2+
migration either to tetrahedral sites or vacant octahedral
sites in the lithium layer.
If Ni
2+
goes to tetrahedral sites, it will later move to
3+
2+
octahedral sites in the lithium layer upon oxidation to Ni . If Ni
migrates directly
to octahedral sites in the lithium layer, it will stay in these sites upon oxidation to
Ni
3+
.
At the end of charge, Ni
3+
ions are stable in octahedral sites in the lithium layer.
3+
If these migrated Ni
do not move back to their original positions in the transition
metal layer, then the irreversible capacity loss observed after first charge can be explained as a combined effect of oxygen loss and site availability [112, 153, 154, 178,179].
When LNSO-15 evolves oxygen upon charge, lithium ions extracted with oxygen evolution cannot be reinserted upon discharge, as Sb
5+
is not redox active. Additionally,
fewer sites for lithium are available for lithium reinsertion on discharge if Ni
3+
mains in the lithium layer, as it blocks all neighboring vacant lithium sites.
re-
The
detrimental effects of nickel migration are pronounced at slow rates, especially in the
first slow charge when the majority of the nickel migration occurs. The reduction of
site availability from nickel migration also explains why cycling at slow rates (more
nickel migration) leads to larger capacity fade, while cycling at fast rates (less nickel
migration) has better capacity retention.
4.10 Conclusion
The lithium-excess LNSO materials containing 0-15% lithium excess were studied
as a function of lithium content and rate.
At slow C-rates, we observe improved
cyclability with increasing lithium content, but also irreversible capacity loss after
116
the first cycle correlating with a plateau at 4.45V. The length of the 4.45V plateau
scales with increasing lithium content. For the LNSO-15 sample, faster C-rates lead
to the disappearance of the 4.45V plateau, as well as unexpectedly higher discharge
capacities.
Through a number of characterization techniques, we set out to understand what
phenomena are responsible for different voltage curve features at slow and fast rates,
and how can higher discharge capacities be achieved at faster C-rates. Using
ex situ
XANES and EELS, we confirm that nickel is oxidized to a similar valence state at
2+
3+
end of charge, but only to somewhere between Ni
and Ni . From
in situ
XRD and
OEMS, we determine that nickel oxidation occurs along the first 4V plateau, followed
by oxygen loss along the second 4.45V plateau. The case for oxygen loss is further
supported by surface structure evolution observed in HRTEM.
The experimental capacity cannot be fully explained by the limited nickel oxidation and oxygen loss. Thus, we propose that the final redox process that is active in
2−
−
the LNSO-15 system may be reversible oxygen redox from O
to O . Additionally,
we suggest that the degree of nickel migration in LNSO-15 dictates the reversible
capacity in the material.
When significant nickel migration occurs to the lithium
layer during slow charge, the surrounding vacant lithium sites become unavailable on
discharge, thus reducing the discharge capacity. In contrast, because nickel migration
is minimized in fast charge, higher discharge capacities can be attained at higher
C-rates.
117
118
Chapter 5
Conclusion
The importance of energy storage is undeniable, from improving performance and
extending lifetimes of current devices to enabling widespread adoption of electric
vehicles and renewable resources. To date, few chemistries have achieved commercial
success as cathode materials in rechargeable lithium batteries.
The challenge for
cathode materials is that a single chemistry must satisfy all requirements across energy
density, cost, safety, rate performance, and lifetime metrics. While known chemistries
have incrementally improved with engineering, significant breakthroughs likely will
rely on new materials.
This has motivated our group’s work to find new cathode
chemistries and accelerate materials development.
In this thesis, two new chemistries were targeted for consideration as cathodes for
rechargeable lithium batteries, the polymorphs of LiMnF4 (Chapter 2) and the layered
lithium-excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 compounds (Chapters 3 and 4). Through careful
characterization of microstructure, structure evolution, and redox mechanisms, we
sought to understand the phenomena that govern stability, energy density, and rate
performance in these materials.
High-throughput computation is one method of identifying new cathodes, screening materials based on calculated stability, voltage, and capacity. While some candidate materials are known but have not been tested for electrochemical properties,
others are predicted and have never been synthesized. Dirutile LiMnF4 is an example
of the former, and was targeted for its high conversion voltage and large theoreti-
119
cal capacity.
To address the need for safe and scalable synthesis methods, dirutile
LiMnF4 was synthesized via a new low temperature solid state reaction. Upon ball
milling of the dirutile phase, the new rutile polymorph of LiMnF4 appeared. Both
polymorphs of LiMnF4 were found to be electrochemically active. Electron diffraction on discharged samples confirmed both dirutile and rutile LiMnF4 to convert upon
lithiation, although not along the equilibrium path.
As in the case of other fluoride cathodes, electrochemical performance was found
to be strongly linked to processing conditions. The ionic nature of fluorides benefits
voltage, but worsens capacity and rate performance as ionic compounds tend to be
insulators. To compensate the insulating character, fluoride cathodes often require
nanosizing and intimate mixing with conductive media. In addition to conductivity
issues, challenging synthesis may explain why few fluorides have been reported in the
literature as cathode materials. Prior to synthesis of dirutile LiMnF4 , the synthesis
of several predicted fluoride compounds was attempted with no success.
Lithium-excess layered oxides are a more fruitful research direction than fluorides.
The lithium-excess Ni-Sb chemistry was designed from theory developed on the percolation of 0-TM diffusion channels in lithium-excess compounds. We validated the
positive effect of lithium excess on electrochemical performance with a new series
of Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 layered materials containing 0-15% lithium excess. Increasing
lithium content in this chemistry was found to improve both discharge capacity and
cyclability at 1C. Characterization on the lithium excess structures revealed a two
domain microstructure in the transition metal layer, where the interface of the two
domains forms the desired low barrier lithium diffusion channels.
The significance of the first part of our study on Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 is many-fold.
First, we confirm the design principle of incorporating lithium excess to improve
lithium diffusion in a new chemical system. Second, the distinct microstructure enables percolation of the low barrier lithium diffusion channels to be achieved at lower
lithium excess levels, maximizing the redox-active transition metal content. Finally,
we demonstrate that percolation of these channels can be achieved in ordered materials, thus maintaining the high voltage advantage of ordered materials to achieve
120
higher energy densities. These results point to an exciting strategy for future design
of new lithium excess cathode materials.
In the second part of our study on Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2 , we investigated the highly
unusual rate behavior of 15% lithium-excess Li1.15 Ni0.47 Sb0.38 O2 . Cathode materials
typical achieve their highest capacities at slow rates, as lithium diffusion is often the
rate-limiting process. In contrast, when increasing cycling rates by a factor of twenty
from C/20 to 1C, Li1.15 Ni0.47 Sb0.38 O2 achieved 25% higher discharge capacity and
improved capacity retention.
Through a range of techniques, we characterized the nickel, antimony, and oxygen
redox activity in Li1.15 Ni0.47 Sb0.38 O2 at slow and fast rates.
Surprisingly, similar
redox behavior was observed of nickel and oxygen at slow and fast rate.
Because
nickel redox and oxygen loss could not account for all observed capacity, the balance
of the capacity was hypothesized to derive from oxygen redox.
This is in contrast
to traditional cathode materials that are designed around a redox-active transition
metal, not the anion.
From careful characterization of the active redox processes, it was concluded that
redox activity could not explain the origins of higher capacities at faster rates. Instead, nickel migration was proposed as an explanation for the rate-dependent voltage
curves. At slow rates, significant nickel migration can occur during charge, blocking
vacant lithium sites and leading to irreversible capacity loss. At fast rates, nickel migration occurs to a lesser extent, allowing higher discharge capacity to be achieved at
faster rates. Understanding around the unusual redox behavior and structural changes
in Li1.15 Ni0.47 Sb0.38 O2 can shed light on mysteries in other high capacity lithium-excess
materials, contributing to future design and engineering of new materials.
While the polymorphs of LiMnF4 and the layered lithium excess Li𝑥 Ni2−4𝑥/3 Sb𝑥/3 O2
compounds may never be commercially viable cathodes, our insights into their properties are still valuable.
Generalities across battery chemistries are few, and the
mechanisms in each chemistry are complex. Only by understanding materials at a
fundamental level can we design better battery materials for the future.
121
122
Bibliography
[1] J.-M. Tarascon and M. Armand.
lithium batteries.
Issues and challenges facing rechargeable
Nature, 414:359–367, 2001.
[2] M. Saiful Islam and Craig A. J. Fisher. Lithium and sodium battery cathode
materials:
properties.
computational insights into voltage, diffusion and nanostructural
Chemical Society Reviews, 43:185–204, 2014.
[3] Rahul Malik, Damian Burch, Martin Bazant, and Gerbrand Ceder.
Size Dependence of the Ionic Diffusivity.
Particle
Nano Letters, 10:4123–4127, 2010.
[4] Alexander Urban, Jinhyuk Lee, and Gerbrand Ceder.
The Configurational
Space of Rocksalt-Type Oxides for High-Capacity Lithium Battery Electrodes.
Adv. Energy Mater., 4(13):1400478, 2014.
[5] Jinhyuk Lee, Alexander Urban, Xin Li, Dong Su, Geoffroy Hautier, and
Gerbrand Ceder.
Unlocking the Potential of Cation-Disordered Oxides for
Rechargeable Lithium Batteries.
Science, 343:519–522, 2014.
[6] N. Tsiouvaras, S. Meini, I. Buchberger, and H. A. Gasteiger. A Novel On-Line
Mass Spectrometer Design for the Study of Multiple Charging Cycles of a Li-O2
Battery.
J. Electrochem. Soc., 160(3):A471–A477, 2013.
[7] Paul Denholm, Erik Ela, Brendan Kirby, and Michael Milligan.
The Role of
energy storage with renewable electricity generation. Technical report, National
Renewable Energy Laboratory, 2010.
[8] Bruce Dunn, Haresh Kamath, and Jean-Marie Tarascon.
Storage for the Grid: A Battery of Choices.
Electrical Energy
Science, 334:928–935, 2011.
[9] M. Armand and J.-M. Tarascon. Building better batteries.
Nature, 451:652–657,
2008.
[10] Vinodkumar Etacheri, Rotem Marom, Ran Elazari, Gregory Salitra, and Doron
Aurbach. Challenges in the development of advanced Li-ion batteries: a review.
Energy & Environmental Science, 4:3243–3262, 2011.
[11] M. Stanley Whittingham. Lithium Batteries and Cathode Materials.
Reviews, 104:4271–4301, 2004.
123
Chemical
Lithium Batteries: Sci-
[12] Gholam-Abbas Nazri and Gianfranco Pistoia, editors.
ence and Technology.
Kluwer Academic Publishers, 2004.
[13] Vanchiappan Aravindan, Joe Gnanaraj, Srinivasan Madhavi, and Hua-Kun Liu.
Chemistry -
Lithium-Ion Conducting Electrolyte Salts for Lithium Batteries.
A European Journal, 17(51):14326–14346, 2011.
[14] Doron Aurbach.
Nonaqueous Electrochemistry.
Marcel Dekker, Inc., 1999.
[15] D.
and
Li
Guyomard
J.
M.
Tarascon.
Metal-Free
Rechargeable
LiMn2O4/Carbon Cells: Their Understanding and Optimization.
the Electrochemical Society, 139(4):937–948, 1992.
Journal of
[16] John B. Goodenough and Youngsik Kim. Challenges for Rechargeable Li Batteries.
Chemistry of Materials, 22:587–603, 2010.
[17] Jianlin Li, Claus Daniel, and David Wood. Materials processing for lithium-ion
batteries.
Journal of Power Sources, 196(5):2452–2460, 2011.
[18] Sung-Yoon Chung, Jason T. Bloking, and Yet-Ming Chiang.
conductive phospho-olivines as lithium storage electrodes.
Electronically
Nature Materials,
1:123–128, 2002.
[19] L. Wang, T. Maxisch, and G. Ceder. A First-Principles Approach to Studying
the Thermal Stability of Oxide Cathode Materials.
Chemistry of Materials,
19(3):543–552, 2007.
[20] Shyue Ping Ong, Anubhav Jain, Geoffroy Hautier, Byoungwoo Kang, and Gerbrand Ceder. Thermal stabilities of delithiated olivine MPO4 (M = Fe, Mn)
cathodes investigated using first principles calculations.
munications, 12:427–430, 2010.
Electrochemistry Com-
[21] A. J. Smith, J. C. Burns, S. Trussler, and J. R. Dahn. Precision Measurements
of the Coulombic Efficiency of Lithium-Ion Batteries and of Electrode Materials
for Lithium-Ion Batteries.
Journal of The Electrochemical Society, 157(2):A196–
A202, 2010.
[22] F. Badway, F. Cosandey, N. Pereira, and G. G. Amatucci.
Carbon Metal
Fluoride Nanocomposites: High capacity Reversible Metal Fluoride Conversion
Materials as Rechargable Positive Electrodes for Li Batteries.
Electrochemical Society, 150(10):A1318–A1327, 2003.
Journal of The
[23] I. Plitz, F. Badway, J. Al-Sharab, A. DuPasquier, F. Cosandey, and G. G. Amatucci.
Structure and Electrochemistry of Carbon-Metal Fluoride Nanocom-
posites Fabricated by Solid-State Redox Conversion Reaction.
Electrochemical Society, 152(2):A307–A315, 2005.
124
Journal of The
[24] Jordi Cabana, Laure Monconduit, Dominique Larcher, and M. Rosa Palacín.
Beyond Intercalation-Based Li-Ion Batteries: The State of the Art and Challenges of Electrode Materials Reacting Through Conversion Reactions.
vanced Energy Materials, 22:E170–E192, 2010.
Ad-
[25] G. Ceder, Y.-M. Chiang, D. R. Sadoway, M. K. Aydinol, Y.-I. Jang, and
B. Huang. Identification of cathode materials for lithium batteries guided by
first-principles calculations.
Nature, 392:694–696, 1998.
[26] Tsutomu Ohzuku and Yoshinari Makimura. Layered Lithium Insertion Material
of LiNi1/2Mn1/2O2: A Possible Alternative to LiCoO2 for Advanced LithiumIon Batteries.
Chem. Lett., 8:744–745, 2001.
[27] Tsutomu Ohzuku and Yoshinari Makimura. Layered Lithium Insertion Material
of LiCo1/3Ni1/3Mn1/3O2 for Lithium-Ion Batteries.
Chem. Lett.,
30(7):642–
643, 2001.
[28] N. Yabuuchi, M. Takeuchi, D. Endo, T. Ozaki, T. Inamasu, J.-Y. Son, Y.T. Cui, H. Oji, and S. Komaba.
Synthesis, Electrochemical Properties, and
Reaction Mechanisms of New Lithium-Excess Transition Metal Oxides with
Cation Disordered Rock-Salt-Type Structure.
Meeting, 2013.
In
Abstract #874, 224th ECS
[29] C. S. Johnson, J-S. Kim, C. Lefief, N. Li, J. T. Vaughey, and M. M. Thackeray.
The significance of the Li2MnO3 component in ‘composite’ xLi2MnO3
(1-x)LiMn0.5Ni0.5O2 electrodes.
Electrochem. Commun., 6:1085–1091, 2004.
[30] Naoaki Yabuuchi, Kazuhiro Yoshii, Seung-Taek Myung, Izumi Nakai, and
Shinichi Komaba.
Detailed studies of a high-capacity electrode material for
rechargeable batteries, Li2MnO3-LiCo(1/3)Ni(1/3)Mn(1/3)O2.
Soc., 133:4404–4419, 2011.
J. Am. Chem.
[31] Michael M. Thackeray, Christopher S. Johnson, John T. Vaughey, N. Li, and
Stephen A. Hackney. Advances in manganese-oxide ‘composite’ electrodes for
lithium-ion batteries.
J. Mater. Chem., 15:2257–2267, 2005.
[32] E. McCalla, A. W. Rowe, J. Camardese, and J. R. Dahn. The Role of Metal
Site Vacancies in Promoting Li-Mn-Ni-O Layered Solid Solutions.
Materials, 25:2716–2721, 2013.
[33] Jing Li, John Camardese, Stephen Glazier, and J. R. Dahn.
Chemistry of
Structural and
Electrochemical Study of the Li-Mn-Ni Oxide System within the Layered Single
Phase Region.
Chemistry of Materials, 26:7059–7066, 2014.
[34] Michael M. Thackeray,
Sun-Ho Kang,
Christopher S. Johnson,
Vaughey, Roy Benedek, and S. A. Hackney.
= Mn, Ni, Co) electrodes for lithium-ion batteries.
3125, 2007.
125
John T.
Li2MnO3-stabilized LiMO2 (M
J. Mater. Chem.,
17:3112–
[35] Haijun Yu, Hyunjeong Kim, Yarong Wang, Ping He, Daisuke Asakura, Yumiko
Nakamura, and Haoshen Zhou.
High-energy ‘composite’ layered manganese-
rich cathode materials via controlling Li2MnO3 phase activation for lithium-ion
batteries.
Phys. Chem. Chem. Phys., 14:6584–6595, 2012.
[36] E. McCalla, J. Li, A. W. Rowe, and J. R. Dahn.
The Negative Impact of
Layered-Layered Composites on the Electrochemistry of Li-Mn-Ni-O Positive
Electrodes for Lithium-Ion Batteries.
Journal of the Electrochemical Society,
161(4):A606–A613, 2014.
[37] Haijun
and
Yu,
Yarong
Haoshen
Wang,
Zhou.
Daisuke
Asakura,
Electrochemical
Eiji
kinetics
Hosono,
Tao
of
0.5Li2MnO3
the
Zhang,
0.5LiMn0.42Ni0.42Co0.16O2 ‘composite’ layered cathode material for lithiumion batteries.
RSC Adv., 2:8797–8807, 2012.
[38] Andrew van Bommel and J. R. Dahn.
Kinetics Study of the High Potential
Range of Lithium-Rich Transition-Metal Oxides for Lithium-Ion Batteries by
Electrochemical Methods.
Electrochem. Solid-State Lett., 13(5):A62–A64, 2010.
[39] Sun-Ho Kang and Michael M. Thackeray.
Enhancing the rate capability of
high capacity xLi2MnO3 (1-x)LiMO2 (M=Mn, Ni, Co) electrodes by Li-NiPO4 treatment.
Electrochem. Commun., 11:748–751, 2009.
[40] Xiqian Yu, Yingchun Lyu, Lin Gu, Huiming Wu, Seong-Min Bak, Yongning
Zhou, Khalil Amine, Steven N. Ehrlich, Hong Li, Kyung-Wan Nam, and XiaoQing Yang. Understanding the Rate Capability of High-Energy-Density Li-Rich
Layered Li1.2Ni0.15Co0.1Mn0.55O2 Cathode Materials.
Adv. Energy Mater.,
4:1300950, 2013.
[41] Shijian Zheng, Rong Huang, Yoshinari Makimura, Yoshio Ukyo, Craig A. J.
Fisher, Tsukasa Hirayama, and Yuichi Ikuhara.
Microstructural Changes in
LiNi0.8Co0.15Al0.05O2 Positive Electrode Material during the First Cycle.
Electrochem. Soc., 158(4):A357–A362, 2011.
J.
[42] Bo Xu, Christopher R. Fell, Miaofang Chi, and Ying Shirley Meng. Identifying surface structural changes in layered Li-excess nickel manganese oxides in
high voltage lithium ion batteries: A joint experimental and theoretical study.
Energy Env. Sci., 4:2223–2233, 2011.
[43] Adrien Boulineau, Loïc Simonin, Jean-François Colin, Carole Bourbon, and
Sébastien Patoux. First evidence of manganese-nickel segregation and densification upon cycling in Li-rich layered oxides for lithium batteries.
Nano Lett.,
13:3857–3863, 2013.
[44] Jung-Min Kim and Hoon-Taek Chung.
The first cycle characteristics of
Li[Ni1/3Co1/3Mn1/3]O2 charged up to 4.7 V.
944, 2004.
126
Electrochimica Acta,
49:937–
[45] A. Robert Armstrong, Michael Holzapfel, Petr Novák, Christopher S. Johnson,
Sun-Ho Kang, Michael M. Thackeray, and Peter G. Bruce. Demonstrating oxygen loss and associated structural reorganization in the lithium battery cathode
Li[Ni0.2Li0.2Mn0.6]O2.
J. Am. Chem. Soc., 128(26):8694–8698, 2006.
[46] Fabio La Mantia, Fabio Rosciano, Nicolas Tran, and Petr Novák.
Quantifi-
cation of Oxygen Loss from Li1+x(Ni1/3Mn1/3Co1/3)1-xO2 at High Potentials by Differential Electrochemical Mass Spectrometry.
J. Electrochem. Soc.,
156(11):A823–A827, 2009.
[47] M. Sathiya, G. Rousse, K. Ramesha, C. P. Laisa, H. Vezin, M. T. Sougrati, ML. Doublet, D. Foix, D. Gonbeau, W. Walker, A. S. Prakash, M. Ben Hassine,
L. Dupont, and J-M. Tarascon.
Reversible anionic redox chemistry in high-
capacity layered-oxide electrodes.
Nat. Mater., 12:827–835, 2013.
[48] Sunny Hy, Felix Felix, John Rick, Wei-nien Su, and Bing Joe Hwang. Direct In
situ Observation of Li2O Evolution on Li Rich High Capacity Cathode Material,
Li[NixLi(1-2x)/3Mn(2-x)/3]O2 (0<x<0.5).
Society, 136(3):999–1007, 2013.
Journal of the American Ceramic
[49] Jason R. Croy, Joong Sun Park, Fulya Dogan, Christopher S. Johnson, Baris
Key, and Mahalingam Balasubramanian. First-Cycle Evolution of Local Structure in Electrochemically Activated Li2MnO3.
Chemistry of Materials, 26:7091–
7098, 2014.
[50] Hideyuki Koga, Laurence Croguennec, Michel Ménétrier, Philippe Mannessiez,
François Weill, Claude Delmas, and Stéphanie Belin.
Operando X-ray Ab-
sorption Study of the Redox Processes Involved upon Cycling of the Li-Rich
Layered Oxide Li1.20Mn0.54Co0.13Ni0.13O2 in Li-Ion Batteries.
of Physical Chemistry C, 118(11):5700–5709, 2014.
[51] Haijun Yu and Haoshen Zhou.
The Journal
High-Energy Cathode Materials (Li2MnO3-
LiMO2 ) for Lithium-Ion Batteries.
J. Phys. Chem. Lett., 4:1268–1280, 2013.
[52] A. Van der Ven and G. Ceder. Lithium diffusion mechanisms in layered intercalation compounds.
Journal of Power Sources, 97-98:529–531, 2001.
[53] Anton Van der Ven, Jishnu Bhattacharya, and Anna A. Belak. Understanding
Li Diffusion in Li-Intercalation Compounds.
Accounts of Chemical Research,
46(5):1216–1225, 2013.
[54] Kisuk Kang and Gerbrand Ceder.
lithium transition metal oxides.
Factors that affect Li mobility in layered
Phys. Rev. B, 74:094105, 2006.
[55] Hajime Arai, Shigeto Okada, Yoji Sakurai, and Jun-ichi Yamaki. Cathode performance and voltage estimation of metal trihalides.
68:716–719, 1997.
127
Journal of Power Sources,
[56] F. Badway, N. Pereira, F. Cosandey, and G. G. Amatucci. Carbon-Metal Fluoride Nanocomposites: Structure and Electrochemistry of FeF3:C.
The Electrochemical Society, 150(9):A1209–A1218, 2003.
Journal of
[57] N. Pereira, F. Badway, M. Wartelsky, S. Gunn, and G. G. Amatucci.
Oxyfluorides as High Capacity Cathode Materials for Lithium Batteries.
nal of The Electrochemical Society, 156(6):A407–A416, 2009.
[58] Feng Wang,
Rosa Robert,
Natasha A. Chernova,
Nathalie Pereira,
Iron
JourFred
Omenya, Fadwa Badway, Xiao Hua, Michael Ruotolo, Ruigang Zhang, Lijun
Wu, Vyacheslav Volkov, Dong Su, Baris Key, M. Stanley Whittingham, Clare P.
Grey, Glenn G. Amatucci, Yimei Zhu, and Jason Graetz. Conversion Reaction
Mechanisms in Lithium Ion Batteries:
Electrodes.
Study of the Binary Metal Fluoride
Journal of the American Chemical Society, 133:18828–18836, 2011.
[59] Feng Wang, Hui-Chia Yu, Min-Hua Chen, Lijun Wu, Nathalie Pereira, Katsuyo Thornton, Anton Van der Ven, Yimei Zhu, Glenn G. Amatucci, and Jason
Graetz. Tracking lithium transport and electrochemical reactions in nanoparticles.
Nature Communications, 3:1201, 2012.
[60] Vincent L. Chevrier, Geoffroy Hautier, Shyue Ping Ong, Robert E. Doe, and
Gerbrand Ceder.
FeOF.
First-principles study of iron oxyfluorides and lithiation of
Physical Review B, 87:094118, 2013.
[61] Kamila M. Wiaderek, Olaf J. Borkiewicz, Elizabeth Castillo-Martínez, Rosa
Robert, Nathalie Pereira, Glenn G. Amatucci, Clare P. Grey, Peter J. Chupas,
and Karena W. Chapman.
Comprehensive Insights into the Structural and
Chemical Changes in Mixed-Anion FeOF Electrodes by Using Operando PDF
and NMR Spectroscopy.
Journal of the American Chemical Society, 135:4070–
4078, 2013.
[62] F. Badway, A. N. Mansour, N. Pereira, J. F. Al-Sharab, F. Cosandey, I. Plitz,
and G. G. Amatucci.
Structure and Electrochemistry of Copper Fluoride
Nanocomposites Utilizing Mixed Conducting Matrices.
Chemistry of Materials,
19:4129–4141, 2007.
[63] M. Bervas, L. C. Klein, and G. G. Amatucci.
Reversible Conversion Reac-
tions with Lithium in Bismuth Oxyfluoride Nanocomposites.
Electrochemical Society, 153(1):A159–A170, 2006.
Journal of the
[64] Robert E. Doe, Kristin A. Persson, Geoffroy Hautier, and Gerbrand Ceder.
First Principles Study of the Li-Bi-F Phase Diagram and Bismuth Fluoride
Conversion Reactions with Lithium.
Electrochemical and Solid-State Letters,
12(7):A125–A128, 2009.
[65] Sung-Wook Kim, Dong-Hwa Seo, Hyeokjo Gwon, Jongsoon Kim, and Kisuk
Kang. Fabrication of FeF3 Nanoflowers on CNT branches and their application
128
to high power lithium rechargeable batteries.
Beach, Fla.), 22(46):5260–4, December 2010.
Advanced materials (Deerfield
[66] A.-L. Dalverny, J.-S. Filhol, and M.-L. Doublet.
in conversion materials for Li-ion batteries.
Interface electrochemistry
Journal of Materials Chemistry,
21:10134–10142, 2011.
[67] Robert E. Doe, Kristin A. Persson, Y. Shirley Meng, and Gerbrand Ceder. FirstPrinciples Investigation of the Li-Fe-F Phase Diagram and Equilibrium and
Nonequilibrium Conversion Reactions of Iron Fluorides with Lithium.
istry of Materials, 20:5274–5283, 2008.
[68] Gerbrand Ceder.
Chem-
Opportunities and challenges for first-principles materials
MRS Bulletin,
design and applications to Li battery materials.
35:693–702,
2010.
[69] Glenn G. Amatucci and Nathalie Pereira. Fluoride based electrode materials for
advanced energy storage devices.
Journal of Fluorine Chemistry,
128(4):243–
262, 2007.
[70] Anna Basa, Elena Gonzalo, Alois Kuhn, and Flaviano García-Alvarado. Facile
synthesis of
rial.
𝛽 -Li3VF6:
A new electrochemically active lithium insertion mate-
Journal of Power Sources, 207:160–165, 2012.
[71] Irina Gocheva, Kuniko Chihara, Shigeto Okada, and Jun-ichi Yamaki. Synthesis
and properties of Li3MF6 ( M = V , Cr , Fe ) as positive electrode material for
rechargeable batteries. In
LiBD-5, number 15, 2011.
[72] Irina D. Gocheva, Yuichi Kamimura, Takayuki Doi, Shigeto Okada, Jun-ichi
Yamaki, and Tetsuaki Nishida. Direct synthesis of Cryolite type Li3FeF6 and
its characterization as positive electrode in Li cell.
Kyushu University, 31(1):7–11, 2009.
Engineering Sciences Report,
[73] A. Kuhn, A. Basa, E. C. Gonzalo, F. Garcia-Alvarado, I. Sobrados, and J. Sanz.
Structure and electrochemical Li insertion properties of cryolite-type Li3FeF6
prepared by precipitation from aqueous-based solutions. In
LiBD-5, 2011.
[74] Dae Hoe Lee, Kyler J. Carroll, Scott Calvin, Sungho Jin, and Ying Shirley
Meng. Conversion Mechanism of Nickel Fluoride and NiO-doped Nickel Fluoride
in Li Ion Batteries.
Electrochimica Acta, 59:213–221, 2012.
[75] Hong Li, Gunther Richter, and Joachim Maier. Reversible Formation and Decomposition of LiF Clusters Using Transition Metal Fluorides as Precursors and
Their Application in Rechargeable Li Batteries.
Advanced Materials, 15(9):736–
739, 2003.
[76] P. Barpanda, M. Ati, B. C. Melot, G. Rousse, J-N. Chotard, M-L. Doublet,
M. T. Sougrati, S. A. Corr, J-C. Jumas, and J-M. Tarascon. A 3.90 V iron-based
129
fluorosulphate material for lithium-ion batteries crystallizing in the triplite
Nature Materials, 10:772–779, 2011.
structure.
[77] J. Barker, R. K. B. Gover, P. Burns, A. Bryan, M. Y. Saidi, and J. L.
Swoyer.
Performance Evaluation of Lithium Vanadium Fluorophosphate in
Lithium Metal and Lithium-Ion Cells.
Journal of The Electrochemical Society,
152(9):A1776–A1779, 2005.
[78] Masatoshi Nagahama, Norifumi Hasegawa, and Shigeto Okada. High Voltage
Performances of Li2NiPO4F Cathode with Dinitrile-Based Electrolytes.
of The Electrochemical Society, 157(6):A748–A752, 2010.
Journal
[79] N. Recham, J-N. Chotard, L. Dupont, C. Delacourt, W. Walker, M. Armand,
and J-M. Tarascon.
A 3.6 V lithium-based fluorosulphate insertion positive
electrode for lithium-ion batteries.
Nature Materials, 9:68–74, 2009.
[80] Oliver Clemens, Melanie Kuhn, and Robert Haberkorn.
Synthesis and char-
acterization of the La1-xSrxFeO3-𝛿 system and the fluorinated phases La1xSrxFeO3-xFx.
Journal of Solid State Chemistry, 184:2870–2876, 2011.
[81] Xiao-Zhen Liao, Yu-Shi He, Zi-Feng Ma, Xiao-Ming Zhang, and Liang Wang.
Effects of fluorine-substitution on the electrochemical behavior of LiFePO4/C
cathode materials.
Journal of Power Sources, 174:720–725, 2007.
[82] A. R. Naghash and Jim Y. Lee. Lithium nickel oxyfluoride (Li1-zNi1+zFyO2-y)
and lithium for lithium rechargeable batteries. Part I . Synthesis and characterization of bulk phases.
Electrochimica Acta, 46:941– 951, 2001.
[83] M. Leblanc, G. Ferey, and R. De Pape. Hydrothermal synthesis of transition
metal fluorides in binary, ternary and quaternary systems.
Mat. Res. Bull.,
19:1581–1590, 1984.
[84] Werner Massa. Fluorides: Solid-state Chemistry, 2010.
[85] C. F. Weaver and H. A. Friedman. A Literature Survey of the Fluorides and
Oxyfluorides of Molybdenum. Technical report, 1976.
[86] Von Rudolf Hoppe, Werner Liebe, and Wolfgang Dahne. Uber Fluoromanganate
der Alkalimetalle.
Zeitschrift fur anorganische und allgemeine Chemie, 307:276–
289, 1961.
[87] Nakhaei Omolfajr, Shahtahmassebi Nasser, Rezaeeroknabadi Mahmood, and
Ahmad Kompany.
Synthesis and Characterization of CaF2 NPs with Co-
precipitation and Hydrothermal Method.
otechnology, 2(5):10–13, 2011.
Journal of Nanomedicine and Nan-
[88] George A. Kolta, David W. A. Sharp, and John M. Winfield.
and redox reactions involving vanadium oxotrifluorine.
Chemistry, 14:153–162, 1979.
130
Substitution
Journal of Fluorine
[89] David W. Aldous, Nicholas F. Stephens, and Philip Lightfoot.
The role of
temperature in the solvothermal synthesis of hybrid vanadium oxyfluorides.
Dalton Transactions, pages 4207–4213, 2007.
[90] David W. Aldous.
Solvothermal Chemistry of Early Transition Metal Fluorides.
PhD thesis, University of St. Andrews, 2008.
[91] David W. Aldous and Philip Lightfoot.
Crystallisation of some mixed Na/V
and K/V fluorides by solvothermal methods.
Solid State Sciences, 11:315–319,
2009.
Synthesis and Characterization of Molybdenum and Vanadium
Framework Materials. PhD thesis, University of Southampton, 2010.
[92] Eleni Kotsapa.
[93] Paul DeBurgomaster, Wayne Ouellette, Hongxue Liu, Charles J. O’Connor,
Gordon T. Yee, and Jon Zubieta.
Solvatothermal chemistry of organically-
templated vanadium fluorides and oxyfluorides.
Inorganica Chimica Acta,
363:1102–1113, 2010.
[94] Farida H. Aidoudi, David W. Aldous, Richard J. Goff, Alexandra M. Z. Slawin,
J. Paul Attfield, Russell E. Morris, and Philip Lightfoot.
prepared S = 1/2 vanadium oxyfluoride kagome lattice.
An ionothermally
Nature Chemistry,
3:801–806, 2011.
[95] Farida H. Aidoudi, Peter J. Byrne, Pheobe K. Allan, Simon J. Teat, Philip
Lightfoot, and Russell E. Morris. Ionic liquids and deep eutectic mixtures as
new solvents for the synthesis of vanadium fluorides and oxyfluorides.
Transactions, 40:4324–4331, 2011.
Dalton
[96] Stuart L. James, Christopher J. Adams, Carsten Bolm, Dario Braga, Paul Collier, Tomislav Friščić, Fabrizia Grepioni, Kenneth D. M. Harris, Geoff Hyett,
William Jones, Anke Krebs, James Mack, Lucia Maini, A. Guy Orpen, Ivan P.
Parkin, William C. Shearouse, Jonathan W. Steed, and Daniel C. Waddell.
Mechanochemistry: opportunities for new and cleaner synthesis.
ciety Reviews, 41:413–447, 2012.
Chemical So-
[97] Irina D. Gocheva, Manabu Nishijima, Takayuki Doi, Shigeto Okada, Jun-ichi
Yamaki, and Tetsuaki Nishida. Mechanochemical synthesis of NaMF3 (M=Fe,
Mn, Ni) and their electrochemical properties as positive electrode materials for
sodium batteries.
Journal of Power Sources, 187:247–252, 2009.
[98] A. Skrzypczak, F. Badway, and G. G. Amatucci. Electrochemical Properties of
a Copper Fluorophosphate Electrode Material Fabricated By Mechanochemical
Synthesis.
Electrochemical and Solid-State Letters, 12(4):A66–A68, 2009.
[99] Wei Tong and Glenn G. Amatucci. Mechanochemical synthesis of metallic Ag2F
and application to Ag3MoO3F3 based nanocomposites for lithium batteries.
Journal of Power Sources, 196:1449–1454, 2011.
131
[100] John P. Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized Gradient
Approximation Made Simple.
Physical Review Letters, 77(18):3865–3868, 1996.
[101] Vladimir I. Anisimov, Jan Zaanen, and Ole K. Andersen. Band theory and Mott
insulators: Hubbard U instead of Stoner I.
Physical Review B,
44(3):943–954,
1991.
[102] Anubhav Jain, Geoffroy Hautier, Charles J. Moore, Shyue Ping Ong, Christopher C. Fischer, Tim Mueller, Kristin A. Persson, and Gerbrand Ceder.
high-throughput infrastructure for density functional theory calculations.
putational Materials Science, 50:2295–2310, 2011.
A
Com-
[103] P. Lacorre, J. Pannetier, F. Averdunk, R. Hoppe, and G. Ferrey. Crystal and
Magnetic Structures with a Dirutile Structure of LiCoF4 : The First Compound
with a Dirutile Structure.
Journal of Solid State Chemistry, 79:1–11, 1989.
[104] K.-H. Wandner and R. Hoppe. Der erste Di-Rutil-Vertreter: LiMnF4 (Mit einer
Bemerkung uber LiCoF4).
Zeitschrift fur anorganische und allgemeine Chemie,
546:113–121, 1987.
[105] Vladimir Sepelák, Sylvie Bégin-Colin, and Gérard Le Caër.
in oxides induced by high-energy ball-milling.
Transformations
Dalton Transactions,
41:11927–
11948, 2012.
[106] M. Bervas, F. Badway, L. C. Klein, and G. G. Amatucci.
Bismuth Fluoride
Nanocomposite as a Positive Electrode Material for Rechargeable Lithium Batteries.
Electrochemical and Solid-State Letters, 8(4):A179, 2005.
[107] Yiquan Wu, Yufeng Zhang, Giuseppe Pezzotti, and Jingkun Guo.
Influence
of AlF3 and ZnF2 on the phase transformation of gamma to alpha alumina.
Materials Letters, 52:366–369, 2002.
[108] Georges Denes, Delphine Le Roux, and M. Cecilia Madamba. Phases driven far
from equilibrium by applying mechanical energy: phase transformations to yPbSnF4 upon ball milling.
Materials Research Society Symposium Proceedings,
481:667–672, 1998.
[109] C. Chaillout and M. Marezio. Structural and Physical Properties of Superconducting La2CuO4+delta. In E. Kaldis, editor,
Materials and Crystallographic
Aspects of HTc-Superconductivity, pages 129–144. Kluwer Academic Publishers,
1994.
[110] John B. Goodenough. Cathode materials: A personal perspective.
Power Sources, 174:996–1000, 2007.
Journal of
[111] K. Mizushima, P. C. Jones, P. J. Wiseman, and J. B. Goodenough. LixCoO2
(0<x<1): A New Cathode Material for Batteries of High Energy Density.
Res. Bull., 15(6):783–789, 1980.
132
Mat.
[112] J. P. Peres, C. Delmas, A. Rougier, M. Broussely, F. Perton, P. Biensan, and
P. Willmann. The relationship between the composition of lithium nickel oxide
and the loss of reversibility during the first cycle.
J. Phys. Chem Solids,
57(6-
8):1057–1060, 1996.
[113] Haijun Yu, Yumin Qian, Minoru Otani, Daiming Tang, Shaohua Guo, Yanbei
Zhu, and Haoshen Zhou.
Study of the lithium/nickel ions exchange in the
layered LiNi0.42Mn0.42Co0.16O2 cathode material for lithium ion batteries:
experimental and first-principles calculations.
Energy & Environmental Science,
7:1068–1078, 2014.
[114] Julien Bréger, Ying S. Meng, Yoyo Hinuma, Sundeep Kumar, Kisuk Kang, Yang
Shao-Horn, Gerbrand Ceder, and Clare P. Grey. Effect of High Voltage on the
Structure and Electrochemistry of LiNi0.5Mn0.5O2 : A Joint Experimental and
Theoretical Study.
Chem. Mater., 18(20):4768–4781, 2006.
[115] A. Rougier, P. Gravereau, and C. Delmas.
Optimization of the Composition
of the Li1-zNi1+zO2 Electrode Materials: Structural, Magnetic, and Electrochemical Studies.
Journal of The Electrochemical Society,
143(4):1168–1175,
1996.
[116] Christopher R. Fell, Kyler J. Carroll, Miaofang Chi, and Ying Shirley Meng.
Synthesis-Structure-Property Relations in Layered, "Li-excess" Oxides Electrode Materials Li[Li1/3-2x/3NixMn2/3-x/3]O2 (x=1/3, 1/4, and 1/5).
of The Electrochemical Society, 157(11):A1202–A1211, 2010.
Journal
[117] Kisuk Kang, Ying Shirley Meng, Julien Breger, Clare P. Grey, and Gerbrand Ceder. Electrodes with High Power and High Capacity for Rechargeable
Lithium Batteries.
Science, 311:977–980, 2006.
[118] Xiaoyu Zhang, W. J. Jiang, A. Mauger, Qilu, F. Gendron, and C. M. Julien.
Minimization of the cation mixing in Li1+x(NMC)1-xO2 as cathode material.
Journal of Power Sources, 195:1292–1301, 2010.
[119] Jie Xiao, Natasha A. Chernova, and M. Stanley Whittingham. Layered Mixed
Transition Metal Oxide Cathodes with Reduced Cobalt Content for Lithium
Ion Batteries.
Chemistry of Materials, 20:7454–7464, 2008.
[120] Kenji Shizuka, Takako Kobayashi, Kenji Okahara, Kaoru Okamoto, Shou Kanzaki, and Ryoji Kanno. Characterization of Li1+yNixCo1-2xMnxO2 positive
active materials for lithium ion batteries.
Journal of Power Sources,
146:589–
593, 2005.
[121] Valerie Pralong, Venkatesh Gopal, Vincent Caignaert, Victor Duffort, and
Bernard Raveau.
Lithium-Rich Rock-Salt-Type Vanadate as Energy Storage
Cathode: Li2-xVO3.
Chemistry of Materials, 24:12–14, 2012.
133
[122] Valerie Pralong, Gopal Venkatesh, Sylvie Malo, Vincent Caignaert, Radu Baies,
and Bernard Raveau. Electrochemical Synthesis of a Lithium-Rich Rock-SaltType Oxide Li5W2O7 with Reversible Deintercalation Properties.
Chemistry, 53:522–527, 2014.
Inorganic
[123] John Reed and Gerbrand Ceder. Role of Electronic Structure in the Susceptibility of Metastable Transition-Metal Oxide Structures to Transformation.
Chemical Reviews, 104:4513–4534, 2004.
[124] Xiaohua Ma, Kisuk Kang, Gerbrand Ceder, and Ying Shirley Meng. Synthesis
and electrochemical properties of layered LiNi2/3Sb1/3O2.
J. Power Sources,
173:550–555, 2007.
[125] K. Binder and D. Heerman.
Introduction.
Monte Carlo Simulation in Statistical Physics: An
Springer-Verlag, 2010.
[126] B. Ravel and M. Newville. ATHENA, ARTEMIS, HEPHAESTUS: data analysis for X-ray absorption spectroscopy using IFEFFIT.
Radiation, 12:537–541, 2005.
Journal of Synchrotron
[127] S. I. Zabinsky, J. J. Rehr, A. Ankudinov, R. C. Albers, and M. J. Eller.
Multiple-scattering calculations of x-ray-absorption spectra.
Physical Review
[128] P. Hohenberg and W. Kohn. Inhomogeneous Electron Gas.
Physical Review,
B, 52(4):2995–3009, 1995.
136(3B):B864–B871, 1964.
[129] W. Kohn and L. J. Sham. Self-Consistent Equations Including Exchange and
Correlation Effects.
Physical Review, 140(4A):A1133–A1138, 1965.
[130] John P. Perdew, Kieron Burke, and Matthias Ernzerhof. ERRATA: Generalized
Gradient Approximation Made Simple.
Physical Review Letters,
78(7):1396–,
1997.
[131] G. Kresse and J. Furthmuller.
Efficient iterative schemes for ab initio
total-energy calculations using a plane-wave basis set.
Physical Review B,
54(16):11169–11186, 1996.
[132] G. Kresse and J. Furthmuller. Efficiency of ab-initio total energy calculations
for metals and semiconductors using a plane-wave basis set.
Materials Science, 6:15–50, 1996.
Computational
[133] Vladimir I. Anisimov, F. Aryasetiawan, and A. I. Lichtenstein. First-principles
calculations of the electronic structure and spectra of strongly correlated systems: the LDA + U method.
J. Phys.: Condens. Matter, 9:767–808, 1997.
[134] A. Rohrbach, J. Hafner, and G. Kresse.
transition-metal sulfides.
Electronic correlation effects in
J. Phys.: Condens. Matter, 15:979–996, 2003.
134
[135] Gus L. W. Hart and Rodney W. Forcade. Algorithm for generating derivative
structures.
Physical Review B, 77:224115, 2008.
[136] Gus L. W. Hart and Rodney W. Forcade.
Generating derivative structures
from multilattices: Algorithm and application to hcp alloys.
B, 80:014120, 2009.
Physical Review
[137] Gus L. W. Hart, Lance J. Nelson, and Rodney W. Forcade. Generating deriva-
Computational Materials Science,
tive structures at a fixed concentration.
59:101–107, 2012.
[138] Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffroy Hautier,
Michael Kocher, Shreyas Cholia, Dan Gunter, Vincent L. Chevrier, Kristin A.
Persson, and Gerbrand Ceder.
Python Materials Genomics (pymatgen):
robust, open-source python library for materials analysis.
terials Science, 68:314–319, 2013.
A
Computational Ma-
[139] Julien Bréger, Nicolas Dupré, Peter J. Chupas, Peter L. Lee, Thomas Proffen,
John B. Parise, and Clare P. Grey. Short- and Long-Range Order in the Positive
Electrode Material, Li(NiMn)0.5O2: A Joint X-ray and Neutron Diffraction,
Pair Distribution Function Analysis and NMR Study.
Chemical Society, 127:7529–7537, 2005.
Journal of the American
[140] Dingding Yuan, Xinmiao Liang, Lin Wu, Yuliang Cao, Xinping Ai, Jiwen Feng,
and Hanxi Yang. A Honeycomb-Layered Na3Ni2SbO6 : A High-Rate and CycleStable Cathode for Sodium-Ion Batteries.
Advanced Materials,
26:6301–6306,
2014.
[141] A. Boulineau, L. Croguennec, C. Delmas, and F. Weill.
Reinvestigation of
Li2MnO3 Structure: Electron Diffraction and High Resolution TEM.
of Materials, 21:4216–4222, 2009.
Chemistry
[142] Gerbrand Ceder and Anton Van der Ven. Phase diagrams of lithium transition
metal oxides: investigations from first principles.
Electrochimica Acta,
45:131–
150, 1999.
[143] Jan N. Reimers and J. R. Dahn. Electrochemical and In Situ X-Ray Diffraction
Studies of Lithium Intercalation in LixCoO2.
Society, 139(8):2091–2097, 1992.
Journal of The Electrochemical
[144] Elena A. Zvereva, Maria A. Evstigneeva, Vladimir B. Nalbandyan, Olga A.
Savelieva, Sergey A. Ibragimov, Olga S Volkova, Larisa I. Medvedeva, Alexander
N.
Vasiliev,
Ruediger
Klingeler,
and
Bernd
Buechner.
Monoclinic
honeycomb-layered compound Li3Ni2SbO6: preparation, crystal structure and
magnetic properties.
Dalton transactions, 41:572–580, January 2012.
[145] E. A. Stern, M. Newville, B. Ravel, Y. Yacoby, and D. Haskel. The UWXAFS
analysis package: philosophy and details.
135
Physica B, 208&209:117–120, 1995.
[146] E. J. Garboczi, K. A. Snyder, J. F. Douglas, and M. F. Thorpe. Geometrical
Physical Review E,
percolation threshold of overlapping ellipsoids.
52(1):819–
828, 1995.
[147] Jing Li, Peng Cheng Ma, Wing Sze Chow, Chi Kai To, Ben Zhong Tang,
and Jang-Kyo Kim.
Correlations between Percolation Threshold, Dispersion
State, and Aspect Ratio of Carbon Nanotubes.
Advanced Functional Materials,
17:3207–3215, 2007.
[148] Stefano Meini, Michele Piana, Nikolaos Tsiouvaras, Arnd Garsuch, and Hubert A. Gasteiger. The Effect of Water on the Discharge Capacity of a NonCatalyzed Carbon Cathode for Li-O2 Batteries.
Electrochem. Solid-State Lett.,
15(4):A45–A48, 2012.
[149] Tsutomu Ohzuku, Masatoshi Nagayama, Kyoji Tsuji, and Kingo Ariyoshi. Highcapacity lithium insertion materials of lithium nickel manganese oxides for
advanced lithium-ion batteries: toward rechargeable capacity more than 300
mAhg-1.
J. Mater. Chem., 21:10179–10188, 2011.
[150] Jason Graetz, Channing C. Ahn, Hao Ouyang, Peter Rez, and Brent Fultz.
White lines and d-band occupancy for the 3d transition-metal oxides and
lithium transition-metal oxides.
Phys. Rev. B, 69:235103, 2004.
[151] N. Tran, L. Croguennec, M. Ménétrier, F. Weill, C. Jordy, and C. Delmas.
Mechanisms Associated with the "Plateau" Observed at High Voltage for the
Overlithiated Li1.12(Ni0.425Mn0.425Co0.15)0.88O2 System.
Chem. Mater.,
20:4815–4825, 2008.
[152] Atsushi Ito, Kaoru Shoda, Yuichi Sato, Masaharu Hatano, Hideaki Horie, and
Yasuhiko Ohsawa. Direct observation of the partial formation of a framework
structure for Li-rich layered cathode material Li[Ni0.17Li0.2Co0.07Mn0.56]O2
upon the first charge and discharge.
J. Power Sources, 196:4785–4790, 2011.
[153] Andrew van Bommel, L. J. Krause, and J. R. Dahn. Investigation of the Irreversible Capacity Loss in the Lithium-Rich Oxide Li[Li1/5Ni1/5Mn3/5]O2.
Electrochem. Soc., 158(6):A731–A735, 2011.
J.
[154] Meng Jiang, Baris Key, Ying S. Meng, and Clare P. Grey. Electrochemical and
Structural Study of the Layered, "Li-Excess" Lithium-Ion Battery Electrode
Material Li[Li1/9Ni1/3Mn5/9]O2.
Chem. Mater., 21:2733–2745, 2009.
[155] Jatinkumar Rana, Marian Stan, Richard Kloepsch, Jie Li, Gerhard Schumacher,
Edmund Welter, Ivo Zizak, John Banhart, and Martin Winter.
Changes in Li2MnO3 Cathode Material for Li-Ion Batteries.
Mater., 4:1300998, 2014.
Structural
Adv. Energy
[156] M. N. Ates, Q. Jia, A. Shah, A. Busnaina, S. Mukerjee, and K. M. Abraham.
Mitigation of Layered to Spinel Conversion of a Li-Rich Layered Metal Oxide
136
Cathode Material for Li-Ion Batteries.
J. Electrochem. Soc., 161(3):A290–A301,
2014.
[157] Zhonghua Lu and J. R. Dahn.
Understanding the Anomalous Capacity of
Li/Li[NixLi(1/3-x/3)Mn(2/3-x/3)]O2 Cells Using In Situ X-Ray Diffraction and
Electrochemical Studies.
[158] M. Balasubramanian,
J. Electrochem. Soc., 149(7):A815–A822, 2002.
J. McBreen,
I. J. Davidson,
P. S. Whitfield,
and
I. Kargina. In Situ X-Ray Absorption Study of a Layered Manganese-Chromium
Oxide-Based Cathode Material.
J. Electrochem. Soc., 149(2):A176–A184, 2002.
[159] Hideyuki Koga, Laurence Croguennec, Michel Ménétrier, Philippe Mannessiez,
François Weill, and Claude Delmas.
Different oxygen redox participation for
bulk and surface: A possible global explanation for the cycling mechanism of
Li1.20Mn0.54Co0.13Ni0.13O2.
J. Power Sources, 236:250–258, 2013.
[160] H. Koga, L. Croguennec, M. Ménétrier, K. Douhil, S. Belin, L. Bourgeois,
E. Suard, F. Weill, and C. Delmas.
Reversible Oxygen Participation to the
Redox Processes Revealed for Li1.20Mn0.54Co0.13Ni0.13O2.
Soc., 160(6):A786–A792, 2013.
[161] Zhonghua Lu and J. R. Dahn.
J. Electrochem.
In Situ and Ex Situ XRD Investigation of
Li[Cr[sub x]Li[sub 1/3-x/3]Mn[sub 2/3-2x/3]]O[sub 2] (x=1/3) Cathode Material.
J. Electrochem. Soc., 150(8):A1044, 2003.
[162] Debasish Mohanty, Sergiy Kalnaus, Roberta A. Meisner, Kevin J. Rhodes,
Jianlin Li, E. Andrew Payzant, David L. Wood, and Claus Daniel.
Struc-
tural transformation of a lithium-rich Li1.2Co0.1Mn0.55Ni0.15O2 cathode during high voltage cycling resolved by in situ X-ray diffraction.
J. Power Sources,
229:239–248, 2013.
[163] Fabio La Mantia, Fabio Rosciano, Nicolas Tran, and Petr Novák. Direct evidence of oxygen evolution from Li1+x (Ni1/3Mn1/3Co1/3)1-xO2 at high potentials.
J. Appl. Electrochem., 38:893–896, 2008.
[164] Rosa Robert, Christa Bunzli, Erik J. Berg, and Petr Novák. Activation Mechanism of LiNi0.80Co0.15Al0.05O2: Surface and Bulk Operando Electrochemical,
DEMS and XRD Analyses.
Chemistry of Materials, 27:526–536, 2015.
[165] Jihyun Hong, Hee-Dae Lim, Minah Lee, Sung-Wook Kim, Haegyeom Kim,
Song-Taek Oh, Geun-Chang Chung, and Kisuk Kang.
Critical Role of Oxy-
gen Evolved from Layered Li-Excess Metal Oxides in Lithium Rechargeable
Batteries.
Chem. Mater., 24:2692–2697, 2012.
[166] Elias Castel, Erik J. Berg, Mario El Kazzi, Petr Novák, and Claire Villevieille. Differential Electrochemical Mass Spectrometry Study of the Interface
of xLi2MnO3 (1-x)LiMO2 (M= Ni, Co, and Mn) Material as Positive Electrode
in Li-ion Batteries.
Chem. Mater., 26:5051–5057, 2014.
137
[167] Hongsen Wang, Eric D Rus, Takahito Sakuraba, Jun Kikuchi, Yasuyuki Kiya,
and Héctor D Abruña. CO2 and O2 Evolution at High Voltage Cathode Materials of Li-ion Batteries: A Differential Electrochemical Spectrometry (DEMS)
Study.
Anal. Chem., 86:6197–6201, 2014.
[168] Roman Imhof and Petr Novák. Oxidative Electrolyte Solvent Degradation in
Lithium-Ion Batteries: An In Situ Differential Electrochemical Mass Spectrometry Investigation.
J. Electrochem. Soc., 146(5):1702–1706, 1999.
[169] Cecile Genevois, Hideyuki Koga, Laurence Croguennec, Michel Ménétrier,
Claude Delmas, and François Weill.
Insight in the Atomic Structure of Cy-
cled Lithium-rich Layered Oxide Li1.20Mn0.54Co0.13Ni0.13O2 using HAADF
STEM and Electron Nano Diffraction.
J. Phys. Chem. C, 119:75–83, 2015.
[170] M. Sathiya, K. Ramesha, G. Rousse, D. Foix, D. Gonbeau, A. S. Prakash, M. L.
Doublet, K. Hemalatha, and J.-M. Tarascon. High Performance Li2RuMnO3
(0.2<y<0.8) Cathode Materials for Rechargeable Lithium-Ion Batteries: Their
Understanding.
[171] Eric McCalla,
Chem. Mater., 25:1121–1131, 2013.
Moulay Tahar Sougrati,
Gwenaelle Rousse,
Erik Jamstorp
Berg, Artem M. Abakumov, Nadir Recham, Kannadka Ramesha, Mariyappan
Sathiya, Robert Dominko, Gustaaf Van Tendeloo, Petr Novák, and Jean-Marie
Tarascon. Understanding the roles of anionic redox and oxygen release during
electrochemical cycling of lithium-rich layered Li4FeSbO6.
J. Am. Chem. Soc.,
137:4804–4814, 2015.
[172] Naoaki Yabuuchi, Sundeep Kumar, Hayley H. Li, Yong-Tae Kim, and Yang
Shao-Horn. Changes in the Crystal Structure and Electrochemical Properties
of Li[sub x]Ni[sub 0.5]Mn[sub 0.5]O[sub 2] during Electrochemical Cycling to
High Voltages.
J. Electrochem. Soc., 154(6):A566–A578, 2007.
[173] J R Dahn and C A Michal. Structure and electrochemistry of Li1+-yNiOz and
a new Li2Ni02 phase with the Ni (OH) 2 structure.
Solid State Ionics, 44:87–97,
1990.
[174] Zhonghua Lu, D. D. MacNeil, and J. R. Dahn.
Layered Li[Ni[sub x]Co[sub
1-2x]Mn[sub x]]O[sub 2] Cathode Materials for Lithium-Ion Batteries.
trochem. Solid-State Lett., 4(12):A200–A203, 2001.
Elec-
[175] Jason R. Croy, Mahalingam Balasubramanian, Donghan Kim, Sun-Ho Kang,
and Michael M. Thackeray.
Designing High-Capacity, Lithium-Ion Cathodes
Using X-ray Absorption Spectroscopy.
Chem. Mater., 23:5415–5424, December
2011.
[176] Y Arachi, H Kobayashi, S Emura, Y Nakata, M Tanaka, T Asai, H Sakaebe,
K Tatsumi, and H Kageyama.
Li de-intercalation mechanism in LiNiMnO
cathode material for Li-ion batteries.
Solid State Ionics, 176:895–903, 2005.
138
[177] BN Figgis and MA Hitchman.
Ligand Field Theory and its Applications. Wiley-
VCH, New York, 2000.
[178] T. A. Arunkumar, Y. Wu, and A. Manthiram.
Factors Influencing the Irre-
versible Oxygen Loss and Reversible Capacity in Layered Li[Li1/3Mn2/3]O2Li[M]O2 (M = Mn 0.5-yNi0.5-yCo2y and Ni1-yCoy) Solid Solutions.
Mater., 19:3067–3073, 2007.
[179] Z. Q. Deng and A. Manthiram.
Chem.
Influence of Cationic Substitutions on the
Oxygen Loss and Reversible Capacity of Lithium-Rich Layered Oxide Cathodes.
J. Phys. Chem. C, 115:7097–7103, April 2011.
139