Cf
0
C~-
-.
0 C"Wo
C0
.....
Signature redacted.....................
1

Understanding and Designing High Power and High Energy Density
Cathode Materials for Lithium Ion Batteries by Experiments and
First Principles Computations by
Xiaohua Ma
Submitted to the Department of Materials Science and Engineering on October 19, 2011, in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Abstract
A new layered compound LiNi
2/ 3
Sb
1
/
3
0
2 was synthesized and tested electrochemically to understand the effect of the transition metals on the structural stability of the layered compound upon Li de-intercalation. The electrochemical results show that the structure of LiNi
2/ 3
Sb
1
/3
0 degrades upon cycling. XRD pattern refinement and TEM diffraction on the cycled
2
LiNi
2/ 3
Sbl/
3
0
2 indicate that the structure degradation is associated with the migration of Ni into the Li layer. First principles calculation also shows a very low barrier for the migration of a divalent Ni from the transition metal layer to the tetrahedral sites of the lithium layer in the partially delithiated Li
2/3
Ni
2/ 3
Sbl/
3
0
2
. The divalent Ni becomes highly mobile because of the strong electrostatic repulsion from the surrounding three Ni3+ and three Sb
5
The effect of the alkali ions on the structural stability in the layered AMO
2
(A = alkali ion; M transition metals) compounds is discussed by comparing layered LiMnO
2 and NaMnO
2
. The structure of layered LiMnO
2 transforms rapidly into a spinel-like structure upon delithation due to the formation of a Li/Mn dumbbell configuration. However, such kind of dumbbell does not form in NaxMnO
2
, indicating its better structural stability upon deintercalation. The electrochemical results of NaMnO
2 show a much better capacity retention than that of LiMnO
2
, confirming that
NaMnO
NaMnO
2 is more stable than LiMnO
2 upon deintercalation. The XRD results of the cycled
2 also show no significant structural change. The pronounced voltage steps and plateaus of NaMnO
2 upon cycling were also investigated.
First principles calculations show that the Li diffusivity in LiNio
5
Mn 504 is in the order of 10~' cm 2 /s, implying that LiNio.
5
Mn.
5
0
4 can be a high rate material even with a large particle size. The electrochemical tests of the micron-sized LiNi
LiNio
5
Mn.
5
O
4
0
.
5
Mn
1
,
5
0
4 show higher rate capability than nano
by Shaju, et al, indicating that the ionic and electronic transport may not be the rate limiting factors. It was also found that cell configurations, such as separators, mechanical pressure of the cell and the carbon content in the electrode, could dramatically affect the rate capability of the cell. When the cell is highly optimized in configuration, more than half of the theoretical capacity is obtained at a discharge rate of 167C (corresponding to 22 seconds) with a particle size in the range of 3-5 pm, which agrees with the high Li diffusivity by my calculation.
Thesis Supervisor: Gerbrand Ceder
Title: R.P. Simmons Professor of Materials Science and Engineering
2

I would like to thank my supervisor Prof. Gerbrand Ceder for his generous financial supports and great inspiration and motivations. He always encourages me to be a scientist rather than just a graduate student with his visions and insights.
I would like to thank my two committees, Prof. Tuller and Prof. Shao-Horn for their valuable time and great comments for my thesis.
I would like to thank former and present members of Ceder group. They shared their expertise with me and make my graduate life delightful: Maria Chan, Byungchan Han,
Shirley Meng, Kisuk Kang, Fei Zhou, Kristin Persson, Chris Fischer, Tim Muller, Rober
Doe, Denis Kramer, Gang Yang, Kevin Tibbetts, Yoyo Hinuma, Osman Burak Okan,
Byoungwoo Kang, Geoffroy Hautier, Charles Moore, Anubhav Jain, Jaechul Kim,
Shinyoung Kang, Rahul Malik, Shyue Ping Ong, Ruoshi Sun, Yabi Wu, Lusann Yang, Aziz
Abdellahi, Hailong Chen, Rickard Armiento, Vincent Chevrier and Kathryn Simons.
My gratitude goes to my parents for their unconditional love. This thesis would not be completed without their encourages and supports.
My deepest gratitude goes to my wife, Xiaoli Liu and our beloved daughter Run-Yi. Her devotion, motivation, and patience make this possible.
3

List of Figures.....................................................................7
List of Tables........................................................................13
Chapter 1. Motivation and outline.............................................14
Chapter 2. Introduction of the alkali ion batteries......................19
2.1 The configuration of the alkali ion battery......................................19
2.2 Charge and discharge processes in an alkali ion battery.......................25
2.3 Important properties for the electrode materials.............................26
2.4 The structural stability of the layered intercalation compounds.............28
2.5 Rate-limiting factors in an alkali ion battery.................................31
Chapter 3. First Principles Calculations...................................35
3.1 Density functional theory and first principles calculations..................35
3.2 Application of first principles calculations to battery electrode research.. .39
3.2.1 Intercalation/deintercalation voltage.........................................39
3.2.2 A lkali ionic diffusivity............................................................40
3.2.3 Kinetics of the structure transformation......................................41
Chapter 4. The effect of transition metals on the structural stability in layered compounds: study of the layered LiNi
2
/
3
Sb
1
30
2 .
.
. . . . . .
.. .
44
4.1 M otivation............................................................................ 44
4.2 Materials and methods............................................................ 48
4.2.1 Synthesis of NaNi
2
/
3
Sb302..................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.48
4

4.2.2 Ion exchange...................................................................
4.2.3 X-ray diffraction and TEM....................................................49
48
4.2.4 Electrochemical properties....................................................49
4.2.5 Calculation methodology....................................................49
4.3 Results and discussion..............................................................50
4.3.1 Structure of the precursor.........................................................50
4.3.2 Structure of LiNi
2
/
3
Sb/30
2 after ion exchange................................52
4.3.3 Electrochemical results...........................................................56
4.3.4 Structural degradation during electrochemical reaction.....................57
4.3.5 Relations between structure degradation and capacity decay..............63
4.4 Conclusion................................................................................63
2
. . . . . . . . . . . . . . . . . .
5.1 Introduction............................................................................66
5.1.1 Layered AMO
2
(A = Na, Li; M = 3d transition metal) system as electrode m aterials................................................................................. 66
5.1.2 Electrochemical properties of the layered LiMnO
2
............
. . . . . . . . . . . . . . . .
67
5.1.3 Structure and electrochemical properties of the layered NaMnO
2
.....
. .. ...
69
5.1.4 M otivation ........................................................................... 70
5.2 Experiment methods..................................................................70
5.3 Experimental results....................................................................72
5.4 Discussion...............................................................................86
5.5 Conclusion................................................................................91
5

5
1
5
4
. . . . . . ..
6.1 Introduction...............................................................................95
6.1.1 The structure of LiNio.
5
Mnj.
5
0
4
................
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
.95
6.1.2 The electrochemical properties of LiNio.
5
Mn.
5
0
4
........
. . . . . . . . . . . . . . .
.97
6.2 Materials and Methods...............................................................101
6.3 Results and Discussion................................................................104
6.3.1 Structure and Morphology of the as-prepared LiNio.
5
Mn .504 spinels ... 104
6.3.2 Electrochem ical tests .......................................................... 107
6.3.2.1 Capacity versus voltage by PITT in ordered LiNio.
5
Mn
1 5
0
4
....
. . . . .
107
6.3.2.2 Rate capability and cyclability............................................108
6.3.3 Effects of the cell configuration on the performance......................112
6.3.3.1 The effect of the separators...............................................113
6.3.3.2 The effect of external pressure on the coin cell........................114
6.3.3.3 The effect of the carbon content in the electrode......................118
6.3.4 First principles calculation of lithium diffusivity in ordered
L iN io
5
M n
5
0
4
...........................
. . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . .
119
6.4 Conclusions..............................................................................124
6

Figure 1-1 Comparison of different battery technologies in terms of volumetric and gravim etric energy density.........................................................................15
Figure 2-1 Schematic description of a rechargeable alkali ion battery. A+ is the alkali io n ..................................................................................................... 2 0
Figure 2-2 Cyclic voltammograms of Al metal (a) and Cu metal (b) respectively. 1 M
LiPF
6 in EC:DMC=1 :1 was used as the electrolyte and Li metal as the anode. Scanning started from open circuit potential (OCP) in the cathodic direction with a sweeping rate of
10 m V /s...............................................................................................2 5
Figure 2-3 The structure of the layered LiCoO2 with an 03 stacking sequence..... 29
Figure 4-1 Schematic electronic configuration of Ni at different valence state in the octahedral coordination.......................................................................... 45
Figure 4-2 XRD patterns of NaNi
2
/3
SbvI
3
0
2 synthesized at 800, 900 and 1000 C.........51
Figure 4-3 (a) X-ray diffraction pattern and refinement of LiNi
2
/3
SbI/30
2 obtained after
8h ion exchange from NaNi
2
/3SbI/ 3
0
2 synthesized at 900'C; (b) XRD pattern of
LiNi
2
/3SbI/ 3
0
2 calculated by first principles methods with VxV'-ahex ordering of Ni and
Sb in the T M layer..................................................................................53
Figure 4-4 (a) (b) TEM patterns for the pristine LiNi
2
/3SbI/
3
0
2
; (c) (d) TEM pattern for
LiNi
2
/3SbI/
3
0
2 after 10 cycles, in which the superstructure is enhanced.....................55
Figure 4-5 Charge and discharge capacity for first ten cycles measured at C/20 rate. The voltage w indow is 2.5-4.6 V ................................................................... 57
Figure 4-6 XRD results of LiNi
2
/3SbI/ 3
0
2 electrode after five cycles. The Rietveld
7

refinement indicates that cation mixing is around 10.4%.............................................58
Figure 4-7 Demonstration of the nickel migration and the formation of a Li/Ni dumbbell.
The process can be divided into three steps: (1) Li disorder in a partially delithiated structure creates a trivacancy around a tetrahedron in the Li layer. (2) A single Ni ion moves from the TM layer into the triangular face between the TM and Li layer, and ultimately into the tetrahedral site of the Li layer. (3) The latter step is usually accompanied by a lithium ion moving into the tetrahedral site that shares a face with the octahedral site that has been vacated by the Ni ion. This leads to a Litet-Nitet dumbbell around the TM vacancy and is believed to be a key defect in LiNi/
2
Mn
1 2
O
2 and an intermediate state in the migration of Mn in LiMnO
2
.............
. . . . . . . . .. . . . . . . . . . . . . . .
.60
Figure 4-8 (a) Energy for a Ni ion along the path from an octahedral site in the TM layer to a tetrahedral site in the Li/vacancy layer. (b) Integrated net spin for Ni cations along the migration path in LiNi
2
/
3
Sb302......................
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
Figure 5-1 (a) Cycling of the layered LiMnO
2 electrode between 3.4 V to 4.3 V vs.
Li/Lie; (b) Voltage profile and differential capacity vs. voltage for the cycling of the layered LiMnO
2 between 2.0 V to 4.5 V at the rate of C/15...............................68
Figure 5-2 XRD pattern and refinement of as-prepared NaMnO
2
. The red (black) line represents the experimental (calculated) data. The residual discrepancy is shown in blue.
The refinement results are preformed in the C2/m space group and give Rap = 10.2%, and
)
=
9 .9 8 .............................................................................................
7 3
Figure 5-3 Capacity vs voltage of Na deintercalation (upper) and intercalation (lower) for NaMnO
2 measured by PITT with 1 OmV steps. (a) is for the first cycle, (b) for the second cycle and (c) for the third cycle.......................................................75
8

Figure 5-4 Voltage profile of NaxMnO
2 upon Na deintercalation and intercalation measured by PITT. The cell is potentiostatically charged up to 3.8 V vs Na/Na+ and discharged to 2.0 V ................................................................................. 76
Figure 5-5 Voltage profile of NaMnO
2 after multiple cycles at C/10 (a) and C/30 (b) respectively. The cell is galvanostatically cycled between 2.0 V and 3.8 V.............78
Figure 5-6 Comparison of the voltage profiles of NaMnO
2 charged up to 3.8 V and 4.2 V respectively. The cell was galvanostatically cycled at C/10 rate..........................78
Figure 5-7 Voltage steps (black) and the corresponding current relaxation (red) during
PITT measurements in the first charge (a) and discharge (b) cycle..........................80
Figure 5-8 XRD of fresh NaMnO
2 electrode (black) and partially charged Nao
83
MnO
2
(red). New peaks corresponding to Na0.70MnO2 are indexed using a monoclinic lattice ............................................................................................ . . 82
Figure 5-9 Structure of monoclinic Nao.
93
MnO
2 projected in a-c plane. (purple Mn, yellow Na, red Oxygen) The solid line indicates the unit cell of monoclinic
Na
0
.
93
MnO
2
. The super cell indicated by the dashed line is more comparable to the unit cell of the N a
0
.
70
M nO
2 phase................................................................... 82
Figure 5-10 XRD patterns of fresh electrode (black) and cycled electrode (red)..........84
Figure 5-11 (a) Electrochemical cycling curve from monoclinic NaMnO
2
; (b) Open circuit voltage from monoclinic NaMnO
2
. 1 M NaClO
4 in PC was used as the electrolyte....................................................................................... . . 87
Figure 5-12 Cyclability of NaMnO
2 at lower cutoff voltage. The cell is cycled at C/10 within voltage window 2-3.4 V................................................................89
9

Figure 5-13 Demonstration of the Li/Mn dumbbell configuration in layered LiMnO
2
. Mn migrates to the tetrahedral site in Li layer leaving a Mn vacancy in Mn layer. Li trivacancy is necessary to avoid face sharing with Mn......................................91
Figure 6-1 Schematic structure of the ordered LiNio
5
Mn
1
.
5
0
4
. (Li green, Ni grey, Mn
- purple, O - red)....................................................................................97
Figure 6-2 The reversible potential vs. the Li composition in LixNio.
5
Mn
1 5
0
4
...
. . . . . ..
99
Figure 6-3 Cycling performance of nano LiNio.
5
Mn
1
.
5
0
4
(both ordered and disordered phases). The cell is charged at C/5 rate to 5.0 V and then discharged at various rates dow n to 3 .5 V ...................................................................................... 10 1
Figure 6-4 X-ray diffraction pattern of the sample after 900*C heat treatment for 12 hours in air. The arrows indicate the peaks that belong to the minor rock salt phase............104
Figure 6-5 X-ray diffraction pattern and refinement of as-prepared ordered
LiNio.
5
Mnj.
5
0
4 annealed at 700'C. The square root of the intensity is shown to highlight the minor peaks. The red (black) line is the experimental (calculated) data. The residual discrepancy is shown in blue. The refinement results agree with the ordered spinel model and give R, = 8.76%, R, =12.8%,
2
= 7.75 .....................................
106
Figure 6-6 SEM of the as-prepared ordered LiNio.
5
Mn .
5
0
4 shows an average particle size of 3-5 m icrom eter.................................................................................107
Figure 6-7 Capacity versus voltage measured upon potentiostatically charge and discharge with 1 Om V steps (PITT)..............................................................108
Figure 6-8 Discharge rate performance (a) and capacity retention (b) of the ordered and disordered LiNio.
5
Mni.
5
0
4
.
The electrode density is 3mg/cm 2 and the electrode contains
10

15wt% carbon. Two pieces of Celgard C480 separators were used. Charge rate is 0.5C for 20C, 40C discharge and 0.2C for 0.2C discharge. Voltage window is 3-5V..........110
Figure 6-9 Charge rate performance (a) and capacity retention (b) of LiNio.
5
Mn
1
.
5
0
4
. The electrode density is3mg/cm 2 and the electrode contains 15wt% carbon. Two pieces of
Celgard 2500 separators were used. Discharge rate is 5C. Voltage window is 3-
5 V ...................................................................................................
1 12
Figure 6-10 Performance comparison for three different Celgard separators. Two pieces of each type of separator were used. Charge rate is 0.5C for 20C, 40C discharge and 0.2C for 0.2C discharge. Voltage window is 3-5V..................................................114
Figure 6-11 The relaxation of external load applied on the coin cell with two sequential pressings is shown in (a). Three electric impedance spectra collected during the external load relaxation are compared in (b). Two pieces of Celgard 2325 separators were u sed ................................................................................................. 1 16
Figure 6-12 Rate (a) and cyclability (b) performance for different pressure loaded on the coin cell. Two pieces of 2325 separators were used. Cells were relaxed for one hour after application of pressure...........................................................................117
Figure 6-13 Rate (a) and cyclability (b) performance of LiNio
5
Mn
1
.504 with 65wt% carbon. Charge rate is 5C. Voltage window is 3-5V.........................................119
Figure 6-14 Schematic figure of two distinct paths for lithium migration in ordered
LiNio.
5
Mni.
5
0
4
. Green balls denote Li ions and red balls denote oxygen ions. Li diffusion path I is indicated by blue arrows and the blue ball is the 4a site in the middle of path I.
Dark scarlet arrows show the Li diffusion path II with dark scarlet balls at 12d sites in the m iddle of path II.................................................................................120
11

Figure 6-15 Calculated energy along the lithium migration paths in ordered
LiNio.
5
Mn.
5
0
4
. (a) path I defined by 3 Ni and 3 Mn around the active state; (b) path II defined by 1 Ni and 5 Mn around the active state.............................................123
12

Table 2-1 Comparison of three typical positive electrode materials for Li ion b atterie s...............................................................................................2 1
Table 4-1 The calculated activation barrier for Li motion for various transition metals near the activated state....................................................................................46
Table 4-2 Ionic radii of the related metals in the octahedral coordination...............47
Table 4-3 Refinement results for Na synthesized at three different temperatures..........51
Table 4-4 Structure refinement parameters for LiNi
2
/
3
Sb
1
/3
0
2 after ion exchange at
2 80
0
C ............................................................................................ . . 54
Table 5-1. Comparison of lattice parameters of Nao.
9 3
MnO
2 and Nao .oMnO
2
.....
. . . . . . . .
83
Table 5-2. Summary of ICP results of Na anode surface, fresh and cycled electrolyte in the w ater diluted solution........................................................................
Table 6-1. The related properties of three different Celgard separators....................114
86
13

Lithium ion rechargeable batteries have been widely used in portable electronic devices as their gravimetric and volumetric energy densities are higher than other battery technologies (Fig. 1-1). As these electronic devices become more compact but have more functionality, they demand higher energy density and longer cycle life at reduced cost.
The energy density of a battery is calculated as the product of its voltage and specific capacity, which are characteristic of the electrode materials in the cell. Researchers have spent significant effort to improve the reversible capacity of the electrode materials.
Layered compounds LiMO
2
(M=transition metals) remain most promising because of their theoretical capacity of about 270 mAh/g, which is much higher than that of their competitors such as spinel or olivine materials. However, for layered compounds with a single transition metal, only LiCoO
2 and LiNiO
2 can be reversibly cycled up to about half of their theoretical capacity. For other transition metals, the LiMO
2 compounds degrade structurally upon the first charge. Thus, all the layered compounds with mixed transition metals that can be reversibly cycled are derivatives of LiCoO
2 or LiNiO
2
. Doping of other metals into LiCoO
2 or LiNiO
2 is used to stabilize the structure or for other purposes.
Take LiNiO
2 as an example. The Mn-doped compound LiNio.
5
Mno.
5
0
2 shows higher capacity with much better cyclability than LiNiO
2 itself [1,2]. This indicates that replacing Co or Ni partially by other metals can improve the structural stability upon
14

higher delithiation. A better understanding of their mechanism in stabilization will be very helpful to improve the reversible capacity and cyclability.
3200
100
MH
0 50 100 150
Energy densky (W h kg
1
2bO 250
Figure 1-1 Comparison of different battery technologies in terms of volumetric and gravimetric energy density [3].
All layered LiMO
2 have their corresponding Na phases. As Li and Na are both alkali metals, they show many similarities in the electrochemical properties. However, the significant size difference between Na and Li may also result in some different characteristics. More specifically, in a layered compound, Na ions are less likely to be in the tetrahedral sites than Li. It has been found that the spinel structure with Li in tetrahedral sites is usually the competing phase to the delithiated layered LiMO
2
, which results in phase transformation and capacity fade. However, this type of phase transformation is less likely to happen in Na compounds. Therefore, the comparison of
15

NaMO
2 and LiMO
2 will also give insights to understand the structural stability in the layered structure.
The Na ion battery itself is also promising for its potential cost advantage. Sodium is far more abundant than lithium, though it has not been conclusively demonstrated that lithium reserves would be an issue in the near future. The field of sodium-ion batteries can also offer possibility for novel intercalation structures, some of which may not exist in their Li equivalents.
Besides the pursuit of higher energy density, high power also becomes crucial for some large-scale battery applications such as in hybrid electric vehicles (HEV) or power tools.
In HEV, the rechargeable battery must be able to discharge sufficiently fast and supply enough power to the vehicle, especially during acceleration. It also needs to be able to quickly charge to take the advantage of regenerative braking. In power tools, the current available from the battery should be high enough to supply the desirable power. With the spread of these applications, demands for high power (ie. high rate capability) batteries with high energy density are expected to grow.
Traditionally, the Li diffusion in the active electrode material is viewed as the rate limiting process in a Li ion battery because the Li diffusivity in the solid electrode materials is much lower than in other components such as the liquid electrolyte. The diffusivity is mainly an intrinsic property of the material, which cannot be easily improved. A more practical way to facilitate the diffusion is shortening the diffusion
16

distance by making nano particles. Numerous efforts have been made to use nanomaterials to improve the rate capability, and nano-LiFePO
4 is one of the most successful materials among them. However, the low density of nanomaterials results in low volumetric energy density, and the large surface area of nano-materials results in more severe side reactions with the electrolyte, which deteriorates the cycle life of the battery.
Therefore, it is desirable to find high rate materials with large particle size. In 1D Li diffusion structures such as olivine structures, it is less likely to get high rate capability with large particle size because a small number of defects in the diffusion channel will block the entire channel. It has been shown by first principles calculation that the Li diffusion is much slower in bulk LiFePO
4 than in nano-LiFePO
4 even with a very low concentration of defects in the diffusion channel [4]. For 2D or 3D diffusion structures such as layered and spinel structures, this kind of blockage mechanism is no longer valid as each Li has more than one diffusion path. For example, in spinels such as LiMn
2
04 and LiNio.
5
Mn
1
.
5
0
4
, there are four diffusion paths for each Li site. Therefore, the particle size is not expected to affect the Li diffusion even if there is a low concentration of defects in the diffusion channels. With these considerations, spinels are potentially good candidates for high rate materials with large particle sizes.
The objective of the first part of my thesis is to understand the effects of transition metal and alkali cations on the structural stability in the layered intercalation compounds upon cycling. The objective of the second part is to investigate the rate capability of
LiNio.
5
Mni.
5
0
4 spinel with large particle sizes. Both experimental techniques and first principles calculations are used during my research. Chapter 2 gives a general
17

introduction of alkali ion batteries. Two specific topics are also covered in the introduction as they are closely related to the thesis research: structural stability in the layered intercalation compounds, and rate-limiting factors in an alkali ion battery.
Chapter 3 briefly introduces density functional theory, first principles calculations, and their applications in battery research. In chapter 4, I study the structural stability of layered LiNi
2
/
3
Sb/30
2 upon delithiation and discuss the mechanism of structural degradation. Chapter 5 discusses the effects of alkali metal on the structural stability in layered compounds. I specifically compare the properties of layered LiMnO
2 and
NaMnO
2 and try to explain their dramatic differences in the structural stability. In
chapter 6, I investigate the Li diffusivity in LiNio.
5
Mn
1
.
5
0
4 spinel by first principles calculations and find that it can be a high rate electrode material even with a large particle size. Experimental results confirm the prediction from my calculation. I also discuss the others components of the cell that can also affect the rate capability of the cell significantly. Chapter 7 summarizes the overall work with conclusions.
References
[1] Y. Makimura, T. Ohzuku, J. Power Sources 119-121 (2003) 156-160.
[2] M. Broussely, F. Perton, P. Biensan, J.M. Bodet, J. Labat, A. Lecerf, C. Delmas,
A. Rougier, J.P. Peres, J. Power Sources 54 (1995) 109-114.
[3] J.M. Tarascon, M. Armand, Nature 414 (2001) 359-67.
[4] R. Malik, D. Burch, M. Bazant, G. Ceder, Nano Lett. 10 (2010) 4123-7.
18

A battery is a device that converts chemical energy to electrical energy and vice versa. In practice, it usually consists of several electrochemical cells that are connected in parallel and/or in series to meet the voltage and current requirements. Each electrochemical cell consists of a positive electrode, a negative electrode, and a membrane separator between the two electrodes. Both electrodes and the separator are immersed in the electrolyte. The positive electrode and negative electrode consist of active materials that allow alkali ions to be inserted or extracted. The electrolyte and separator are conductive to ions but electronically insulating, allowing alkali ions but not electrons to pass between the two electrodes via the electrolyte. Both the positive and negative electrodes are in contact with current collectors, which provide paths for electrons to the external electric circuits.
Fig. 2-1 is a schematic description of a rechargeable alkali ion battery.
19

e-
V=-(PA nF
ApfCk
e-
Figure 2-1 Schematic description of a rechargeable alkali ion battery. A+ is the alkali ion.
The positive electrode (also known as the cathode) is a thin film consisting of an active material, a conductive agent, and a binder. The active materials are usually inorganic compounds where alkali ions are inserted or extracted through intercalation or conversion reactions. Within lithium ion batteries, layered LiMO
2
, spinel LiM
2
0
4 and olivine
LiMPO
4
(M=transition metals), are the most intensively investigated and many of them have been commercialized. Table 2-1 compares three typical positive electrode materials for Li ion batteries. For sodium ion batteries, layered NaMO
2 and NASICONs can be used as active materials. As most positive electrode materials are semiconductors or insulators, conductive agents such as carbon black or acetylene black are necessary to enhance the electronic conductivity. Polymer binders such as PTFE
20

(polytetrafluoroethylene) or PVDF (poly-vinylidene fluoride) are used to hold the active materials and conductive agent together with reasonable mechanical strength. Over cycling, the volume change of the active materials can be significant. This results in poor electric contacts between the active material, the conductive agent, and the current collector. A proper binder can improve the electric contact during the cycling, which in turn improves the cyclability.
Table 2-1 Comparison of three typical positive electrode materials for Li ion batteries.
Structure
Ionic conduction
Average Voltage (V)
Practical capacity (mAh/g)
LiCoO
2
LiMn
2
0
4
LiFePO
4 layered (R-3m) spinel (Fd3m) olivine (Pnma)
2D 1 D
4.0
3D
4.1 3.4
155 120 160
Gravimetric energy density
(Wh/kg)
620 492 544
Volumetric energy density
(Wh/L)
3100 2066 1959
The most widely used materials for the negative electrode (also known as the anode) are carbonaceous materials such as graphite or MCMB (Mesocarbon Microbead). Graphite's layered structure allows lithium ions to intercalate between the layers. The specific
21

capacity of graphite is 372 mAh/g and the average voltage about 0.1 0.2 V vs. Li/Lie.
Graphite as an anode shows very good cyclability with electrolytes containing ethylene carbonate because such electrolyte solvents decompose reductively on the carbonaceous anode, forming a protective film called the solid electrolyte interphase (SEI) [1,2]. Na ions can barely be intercalated into the layers in graphite. A hard or nanoporous carbon, however, contains pores between randomly stacked layers, where both sodium and lithium ions can be inserted [3,4]. Silicon, with a specific capacity as high as 4000 mAh/g, is also very promising as a negative electrode material. However, when silicon is fully lithiated, it forms the compound Li
4
.
4
Si, and can undergo a volumetric change as high as 300% [5]. During cycling, the large volume change causes the loss of mechanical integrity at the active material/electrode matrix interface. A better choice of binder, such as CMC (Carboxyl methyl cellulose), can improve the mechanical integrity over cycling and give better cyclability than common binders such as PVDF [6,7].
The separator is a porous membrane placed between the two electrodes, permeable to ionic flow but preventing electric contact of the electrodes. The separator should be chemically inert to both electrodes and the electrolyte. Commonly used separators include porous films of PE (Polyethylene), PP (Polypropylene), and glass fibers. Trilayer separators (PP/PE/PP) offer advantages by combining the lower melting temperature of
PE with the high-temperature strength of PP [8].
The electrolyte is a solution of alkali salts and solvent. For Li and Na ion batteries, the active nature of the strongly reducing negative electrode and the strongly oxidizing
22

positive electrode rules out the use of any aqueous electrolyte despite their many advantages. This is because the reduction of protons and the oxidation of anions such as
OH~ generally occur within 2.0 4.0 V vs. alkali metal [9], while the charged potentials of the positive and negative electrode average 0.0-0.2 V and 3.0-4.5 V respectively. On the other hand, non-aqueous solvents need polar groups such as carbonyl (C=0) or etherlinkage (-0-) to dissolve the salt sufficiently [1]. Among the non-aqueous solvents, carbonates such as ethylene carbonate (EC), propylene carbonate (PC), dimethyl carbonate (DMC) and diethyl caronbate (DEC) are most commonly used. The choices of salts with these solvents are rather limited because the solubility requirement in low dielectric nonaqueous solvent, together with the anodic stability, rules out most anions.
Although several salts such as LiClO
4
, LiAsF
6
, LiBF
4 have been intensively studied,
LiPF
6 wins out due to its well-balanced properties [10-12].
The current collectors are usually metallic foils, which serve as the substrates of the electrodes. In lithium ion batteries, aluminum is the choice for the positive electrode, due to its low cost, good electric conductivity, and anodic stability up to 5 V vs. Li/Li+. Fig.
2-2 (a) from reference [13] shows the cyclic voltammograms of Al metal in 1 M LiPF
6 in
EC:DMC = 1:1 as an electrolyte [13]. It is seen that the anodic current maintains a very low level from 1.5 V up to 5.0 V vs. Li/Li4, indicating a good anodic stability within the potential window of the positive electrode. Between 0 V to 1.0 V vs. Li/Lit, however, there is a strong redox reaction corresponding to the Li-Al alloying and de-alloying. This hinders the use of an Al foil as a current collector for the negative electrode. Copper, however, shows significantly high anodic and cathodic current between 1.5 V to 5.0 V vs.
23

Li/Lie but very low current between 0 V to 1.0 V (Fig. 2-2 (b)) [13]. The excellent cathodic stability makes Cu a good choice for the current collector of the negative electrode in lithium ion batteries. There are no systematic investigations on the current collectors for Na ion batteries at this point.
(a)
I
U aa(
I
I
4 .
I j
E
2
C
0 g0
-2
0
-4
-
0
1.4
2
OCP: 2.630 V
4 6 1 2 3
E vs. (LIILI*) I V
(b)
100
E 5o
0
.100
0
(OCP: 3.326 V)
I 2 3
E vs.(LI/ L*) I V
4 5
24

Figure 2-2 Cyclic voltammograms of Al metal (a) and Cu metal (b) respectively. 1 M
LiPF
6 in EC:DMC=1:1 was used as the electrolyte and Li metal as the anode. Scanning started from open circuit potential (OCP) in the cathodic direction with a sweeping rate of
10 mV/s [13].
The open circuit voltage across an alkali ion battery is established by the difference in the chemical potential of alkali ions on the two electrodes as shown in the equation in Fig. 2-
1. Therefore, the combination of the electrodes materials determines the open circuit voltage of the cell. The electrode with a lower chemical potential of alkali ions has a higher electric potential and is therefore called the positive electrode.
Charge and discharge occurs through the intercalation/deintercalation process at the electrode, which is driven by the reduction of the electrochemical potential of the alkali ion. As lithium ion and sodium ion batteries are similar, we can take a typical lithium ion battery as an example. During the charge process, the Li ions are deintercalated from the positive electrode, which is a layered LiMO
2 compound:
LiMO
2 txLi + xe~ + Li
1 xMO
2
The Li ions are driven by the electric force to the negative electrode through the electrolyte, and the electrons move through the external circuit performing work. At the negative electrode, the Li ions intercalate into the graphite: xLi + 6C + xe-t LixC
6
25

The above reactions at both electrodes are reversible. Upon discharge, the reactions at both electrodes occur in the reverse direction. Li ions are deintercalated from the negative electrode and intercalated into the positive electrode. Fig. 2-1 shows the ionic and electronic flow at the charge and discharge process.
The choice of electrode materials is important in the performance of a cell, since the electrochemical reaction in a cell is intimately tied to the electrode materials. The key parameters for electrode materials include voltage, gravimetric and volumetric energy density, power density, cycle life, safety, efficiency, cost, toxicity, etc.
High voltage is preferred to improve the energy density when there is no room to increase the specific capacity. However, the stability of today's electrolyte requires the voltage to be kept below 4.5 V to avoid significant reactions between the electrolyte and the electrode. The anodic stability of the electrolyte needs to be improved to make use of some promising high voltage electrode materials such as LiNio.
5
Mn.
5
0
4 and LiCoPO
4
[14,15].
The gravimetric energy density, defined as the energy per unit weight (Wh/kg), is the product of the voltage and the specific capacity. The specific capacity of an electrode is the charge that the electrode material can deliver per unit weight (mAh/g). If one mole of electrode material can supply x mole of electrons, then the specific capacity is (x -
26

F/M)x1OOO/3600[mAh/g]. F is the Faraday's constant defined as the amount of charge carried by one mole of electrons, and M is the molar mass of the electrode material.
Besides the specific energy density, volumetric energy density is defined as the energy per unit volume (Wh/L). As today's electronic devices require more energy within a limited size, volumetric energy density is becoming more important. Higher volumetric energy density can also cut down costs by reducing the use of separators, electrolytes and current collectors.
The power density of a cell is defined as the power per unit weight (W/kg), and is calculated as the product of the cell voltage and current. If the internal resistance is r and the load on the cell is R, the current is I = V,/(R + r) and the output power can be calculated by the following equation:
P = V - I - I2 - r= V, - RI(R + r)2, where Vc is the open circuit voltage. With higher current, more power will be distributed to the internal resistance and generates heat inside the cell. When the cell is discharging at high rate with low external load R, the output power is mainly restricted by the internal resistance.
The cycle life is defined as the number of charge/discharge cycles the cell can perform before the specific capacity falls below a certain percentage (such as 80%) of the initial capacity. The cycle life depends on many factors such as the structure stability of the electrode materials, the formation of the SEI layer and its stability, the stability of the
27

electrolyte, and the mechanical integrity. Besides the properties of the cell, working conditions such as the temperature and the discharge rate can also affect the cycle life.
Layered transition metal oxides with a formula AMO
2
(A=Li, Na; M=transition metals) are important compounds for the positive electrode in alkali ion batteries. LiMO
2 often possess a a-NaFeO
2 type structure. This is a distorted rock salt superstructure with an
FCC cubic close-packed (CCP) oxygen framework (ABCABC close-packed oxygen plane stacking sequence). The Li and transition metal cations are ordered in alternating
(111) planes at the octahedral interstitials of the oxygen framework. The rigid layered host possesses MO
2 layers composed of edge sharing MO
6 octahedra. The Li in between these layers can be deintercalated reversibly. This structure is called 03 because the MO
2 plane stacking sequence repeats every three layers. Fig. 2-3 shows the structure of an 03 type LiCoO
2 with a space group of R-3m [16]. In the Na equivalents, however, Na cations can sit at either octahedral or trigonal prismatic interstitials. The oxygen stacking sequence can also be different from ABCABC style. Various Na sites and oxygen stacking sequences result in polytypes such as 03 (ABCABC), 02 (ABAB), P3
(ABBCCA), and P2 (ABBA) [17,18].
28

-
-B
-A
C
-
-C
A
B
-A
03 host structure
0 octahedra surrounding Co
Edge sharing 0 octahedra in Li plane
Face sharing 0 octahedra in Li plane
Figure 2-3 The structure of the layered LiCoO2 with an 03 stacking sequence [19].
If the alkali ions can be fully deintercalated from the layered AMO
2
, the specific capacity can be as high as 270 mAh/g and 240 mAh/g for the lithium and sodium compound respectively. However, that high capacity cannot be achieved with good cyclability. For example, only half of the theoretical capacity can be achieved in LiCoO
2 with good cyclability because the removal of more than 0.5 Li will cause structural instability [20].
During the delithiation of the layered compounds, there are several possible structural changes that can affect the cyclability. Firstly, the c lattice parameter can change significantly in a highly delithiated layered structure. A dramatic change in lattice parameter will cause crystal faults especially in large particles. Secondly, the oxygen stacking sequence can change through gliding of the MO
2 layers. For example, after full delithiation of LiCoO
2
, the stacking sequence changes from 03 to Hl-3 and then to 01
[21,22]. Though this kind of structure change can be reversible, it is more likely to cause more defects and microfractures. Thirdly, transition metal ions can move to the vacant Li sites in the Li layer and the vacant octahedral sites in the transition metal layer also make
29

some face-sharing tetrahedral sites available to Li or transition metal cations [23]. Indeed, a spinel structure with Li in the tetrahedral sites was found after extensive cycling of
LiCoO
2 with deep charge and discharge [24]. The occupation of octahedral or tetrahedral sites in Li layers by transition metal ions will block the Li diffusion paths and make some
Li sites unavailable.
The structural stability in different layered compounds can vary. For example,
LiNio.
5
Mno.
5
0
2 can be cycled up to 200 mAh/g with excellent capacity retention [25], while LiCoO
2 shows obvious capacity fade when it is cycled with the capacity more than
150 mAh/g [20]. A more dramatic comparison is between LiCoO
2 and LiMnO
2
. Layered
LiMnO
2 can be obtained by from NaMnO
2 by ion exchange. During cycling of LiMnO
2 the capacity decays rapidly and the structure becomes spinel-like [26]. Layered LiCoO
2 however, can be cycled up to Lio.
5
CoO
2 with excellent capacity retention. In the delithiated LixMnO
2
, the transformation from layered to spinel requires the migration of
Mn ions from the transition metal layer into the Li layer. The first principles calculation shows that Mn can migrate from the octahedral sites of the transition metal layer into the tetrahedral sites of the Li layer with a very low energy barrier [27]. The Mn ions in the tetrahedral sites are stable in energy. The energy barrier for Co to migrate is, however, too high to be likely. Therefore, layered LiCoO
2 shows much better structure stability and cyclability than LiMnO
2 though the structures are identical. As the layered compounds can show very different structural stability upon the removal of alkali ions, a deeper understanding of the structural stability in layered compounds is helpful to improve the reversible capacity and the cyclability.
30

2.5
The rate capability is directly related to the resistance for both ionic and electronic transport. In alkali ion intercalation or deintercalation, the ionic transport can be viewed as a four-step process: ionic diffusion in the particle of the active materials, ionic transport at the electrode/electrolyte interface, diffusion in the electrolyte, and transport through the separator. With each step, there is some resistance that impedes the transport.
During intercalation and deintercalaction, alkali ions need to diffuse within a particle from certain interstitial sites to the surface of the particle. This step is traditionally regarded as the rate-limiting step due to the very low ionic diffusivity of the solid electrode material. However, with the discovery of some high diffusivity electrode materials such as LiFePO
4
, some other steps might be rate limiting. The reaction between the electrode and the electrolyte usually happens at the interface, forming a protective film that the alkali ions need to pass through. The characteristics of the film, mainly its resistance and thickness, determine the rate of this transport step. Though the liquid electrolyte usually has very high ionic diffusivity, its ionic transport within the composite electrode matrix can be slow. For an active particle that is far away from the electrode film/ electrolyte front, Li salts need to pass through the pores of the composite electrode matrix to reach the surface of the particle [28]. Therefore, the porosity and thickness of the electrode film play an important role in this transport step. Lastly, the porosity and permeability of the separator will affect the rate that alkali ions can pass through it.
31

Similarly, electronic transport includes several steps: electronic conduction within the active material, electronic transport at the electrode/electrolyte interface, electron transport through conductive agents such as carbon, and transport from the electrode film to the current collector. The first two steps are very similar to those in the ionic transport.
For the electron transport through the conductive agent, carbon can be regarded as electronic wires connecting the active material and the current collector. For an active particle that is far from the current collector, electrons need to travel a long distance to reach the current collector, which results in a large resistance [29]. Therefore a thin electrode film is helpful in this electronic transport. In addition, a good electric contact between the electrode film and the current collector is important to facilitate the electronic transport.
Besides the intrinsic ionic and electronic conductivity of the electrode materials, other factors can also affect the rate capability of an electrochemical cell. These factors include the particle size, the characteristics of the surface film, the thickness and porosity of the electrode film, the ionic conductivity of the electrolyte, the porosity and permeability of the separator, the electric contact between the electrode and the current collector, etc. All these factors need to be considered to obtain an optimized rate capability. For a fast electrode material, the rate-limiting process of the cell is no longer the diffusion within the material. Removing all other rate-limiting factors becomes critical to correctly evaluate the intrinsic rate capability of the active material [30].
32

References
[1] K. Xu, Chem. Rev. 104 (2004) 4303-417.
[2] R. Fong, U. von Sacken, J.R. Dahn, J. Electrochem. Soc. 137 (1990) 2-6.
[3] D.A. Stevens, J.R. Dahn, J. Electrochem. Soc. 147 (2000) 1271.
[4] D.A. Stevens, J.R. Dahn, J. Electrochem. Soc. 148 (2001) A803.
[5] L.Y. Beaulieu, T.D. Hatchard, A. Bonakdarpour, M.D. Fleischauer, J.R. Dahn, J.
Electrochem. Soc. 150 (2003) A1457.
[6] J. Li, R.B. Lewis, J.R. Dahn, Electrochem. Solid-State Lett. 10 (2007) A17.
[7] W.-R. Liu, M.-H. Yang, H.-C. Wu, S.M. Chiao, N.-L. Wu, Electrochem. Solid-
State Lett. 8 (2005) A100.
[8] P. Arora, Z. Zhang, Chem. Rev. 104 (2004) 4419-62.
[9] A.J. Fry, Synthetic Organic Electrochemistry, 2nd ed., John Wiley, London, 1989.
[10] L.J. Krause, W. Lamanna, J. Summerfield, M. Engle, G. Korba, R. Loch, R.
Atanasoski, J. Power Sources 68 (1997) 320-325.
[11] M. Ue, S. Mori, J. Electrochem. Soc. 142 (1995) 2577.
[12] J.M. Tarascon, D. Guyomard, Solid State Ionics 69 (1994) 293-305.
[13] S.-T. Myung, Y. Hitoshi, Y.-K. Sun, J. Mater. Chem. (2011).
[14] D. Aurbach, B. Markovsky, Y. Talyossef, G. Salitra, H. Kim, S. Choi, J. Power
Sources 162 (2006) 780-789.
[15] Y. Talyosef, B. Markovsky, G. Salitra, D. Aurbach, S. Choi, J. Power Sources 146
(2005) 664-669.
[16] C. Delmas, C. Fouassier, P. Hagenmuller, Physica 99B (1980) 80.
[17] C. Delmas, J.J. Braconnier, C. Fouassier, P. Hagenmuller, Solid State Ionics 3
(1981) 165-169.
[18] A. Mendiboure, C. Delmas, P. Hagenmuller, J. Solid State Chem. 57 (1985) 323-
331.
[19] A. Van der Ven, PhD Thesis, MIT, Cambridge, MA, 2000.
33

[20] H. Xia, L. Lu, Y.S. Meng, G. Ceder, J. Electrochem. Soc. 154 (2007) A337.
[21] A. Van der Ven, M.K. Aydinol, G. Ceder, G. Kresse, J. Hafner, Phys. Rev. B 58
(1998) 2975-2987.
[22] Z. Chen, Z. Lu, J.R. Dahn, J. Electrochem. Soc. 149 (2002) A 1604.
[23] J. Reed, G. Ceder, Chem. Rev. 104 (2004) 4513-33.
[24] H. Wang, Y.-I. Jang, B. Huang, D.R. Sadoway, Y.-M. Chiang, J. Electrochem.
Soc. 146 (1999) 473.
[25] Y. Makimura, T. Ohzuku, J. Power Sources 119-121 (2003) 156-160.
[26] A.R. Armstrong, P.G. Bruce, Nature 381 (1996) 499-500.
[27] J. Reed, G. Ceder, A. Van der Ven, Electrochem. Solid-State Lett. 4 (2001) A78.
[28] P.A. Johns, M.R. Roberts, Y. Wakizaka, J.H. Sanders, J.R. Owen, Electrochem.
Commun. 11 (2009) 2089-2092.
[29] M. Gaberscek, M. Ktizma, J. Jamnik, Phys. Chem. Chem. Phys. 9 (2007) 1815-20.
[30] X. Ma, B. Kang, G. Ceder, J. Electrochem. Soc. 157 (2010) A925-A931.
34

First principles calculations, or ab initio calculations, are distinguished from empirical methods by requiring only nuclear charges and number of electrons as inputs, without any adjustable experimental parameters. Energy, structure, and other properties of a substance can be calculated from these inputs strictly through the application of quantum mechanics. First principles calculations have proven to be excellent complementary tools to laboratory experiments in research because they can calculate some characteristics of a modeled system that are very hard to obtain experimentally. First principles calculations also offer far greater ability to control and manipulate a system, providing the modeled system reflects the real system accurately. In this chapter, we will discuss first principles calculations and their applications in the research of positive electrode materials.
The objective of a first principles calculation is to obtain the wave function that describes a given system by solving the Schr6dinger equation. In the investigation of the basic chemistry of the materials, only time-independent interactions between electrons and nuclei need to be considered. Thus the wave function is the product of a time-dependent exponential phase factor and a time-independent wave function qi that is determined by the time-independent Schrddinger equation:
35

(3.1)
In equation 3.1, H is the Hamiltonian operator, y is the time-independent wave function, and E is the total energy of the system. For systems with many particles like the oxides crystals, it is necessary to introduce a number of approximations to arrive at a solution for qf. The first is the Born-Oppenheimer approximation, in which the electrons are assumed to move so much faster than the nuclei that the nuclei can be treated as stationary. This approximation allows a separate wave function that contains only information about electrons to be determined by equation 3.1. The electronic wave functions that satisfy the
Schr6dinger equation must be anti-symmetric to obey the Pauli exclusion principle. The coordinates of nuclei Rn act simply as parameters in the Hamiltonian that can be written as
(3.2)
Rm-Rn where T is the kinetic energy operator of electron, Vee is the Coulomb interactions between the electrons and v(';) is the Coulomb interactions between the electrons and the nuclei. The last term arises from the Coulomb interactions between nuclei having charge Zm and Zn, which can be dropped in solving the electronic wave function since it is simply an additive term.
A variational method is often used to solve this intractable many-body Schr6dinger equation. This method is based on the fact that given an arbitrary square-integrable function p, an upper bound for the ground state energy EO of the system governed by the
Hamiltonian H is provided by [1]
36

((qHj)
=E[q] > E
0 (3.3)
To find the upper bound that lies closely above the true ground state energy, it is critical to use a trial function p that closely resembles the true ground state wavefunction. The functional E[p] is variationally minimized to optimize the trial function and hence minimize the upper bound. One type of trial function is a Slater determinate of single electron orbitals; such a method is called the Hartree-Fock method [2].
A different approach to solving equation 3.1 relies on Density Functional Theory (DFT).
The fundamental principle of DFT is that the ground state of a substance is uniquely determined by the electron density p(r'),
(?)
=
(P1;
(3.4) and the ground state energy is a functional of the electron density [3]
E[p(i')] = F[p(i)] + f p()v(ir)dr. (3.5)
In equation 3.5, the only inputs into the functional E[p(-)] that are unique to a given system are v(7), the electrostatic potential created by the nuclei, and N, the total number of the electrons as a constraint on p(r'). F[p(r')] T[p(r')] + Vee [p(r')] is a universal functional independent of the nuclear arrangement and charge. If the universal functional
F[p(r')] is known, the ground state energy can be obtained by variationally minimizing the functional E[p(r?)] with respect to the electron density p(r').
Kohn and Sham introduce a separation of the unknown F[p(r')] as follows: [4]
F[p] = T [p] +J[p] + Ex,[p], (3.6)
37

where T[p] is the kinetic energy of a system of non-interacting electrons, and J[p] is the classical Coulomb energy. The last term Ex, [p] is called the exchange-correlation energy, including the difference in kinetic energies between non-interacting electrons and actual interacting electrons, as well as a correction to J [p] arising from the correlations between electrons. The non-interacting electrons kinetic energy functional T[p] can be calculated exactly with a Slater determinant of independent electron orbitals. By replacing F[p] by equation 3.6, the minimization of E[p] in equation 3.5 leads to the Kohn-Sham equations in the form of [4]
H(() =
[-}V 2
+ veff (r)]pj(r) = Ej pj(r)
(3.7)
Veff
(1) = v(1) + 8%p + SExc[p]
The ground state can be obtained by solving the Kohn-Sham equations self-consistently.
The exchange-correction term is unknown and approximations are necessary. A simple approximation to Exc is the local density approximation (LDA) proposed by Kohn and
Sham [4]. In LDA, exchange correlation energy per electron is set equal to the exchange correlation energy per electron of a homogeneous electron gas with the same density.
Therefore, LDA is a good approximation for a system with a slowly varying electron density. For systems where electrons are highly localized, however, LDA fails to capture the strong correlation between the localized electrons.
Purdue and Yue developed the generalized gradient approximation (GGA) in order to address the major source of error in LDA [5,6]. GGA incorporates polynomial terms with
38

the modulus of the electron density gradient to correct some of the deviation of the exchange-correlation energy from the uniform electron gas results.
Among many numerical programs that solves the Kohn-Sham equations iteratively, we have chosen "Vienna Ab inito Simulation Package" (VASP) in our research [7,8]. GGA is used as it typically does better reproducing spin polarization on transition metal ions in oxides [9]. Besides GGA, we also include the self-consistent U parameter to cancel the electron self-interaction errors [10].
By solving the Kohn-Sham equations, we can obtain important properties of materials for electrochemical applications. We will briefly discuss the basic ideas on how properties, such as the intercalation voltage, alkali ionic diffusivity, and structural transformation rate are obtained from first principles calculations.
3.2.1 Intercalation/deintercalation voltage
The equilibrium intercalation voltage is determined by the chemical potential difference of alkali ions in the anode and cathode. The open circuit voltage of a cathode with the alkali ion composition of x, is obtained by
V(x) = _AA cathode _ anode ze I!A ,(3.9)
39

and the voltage is only dependent on the change in the chemical potential on the cathode athode
(x). In most cases, the average voltage is of main interest, though it is possible to obtain the voltage as a function of alkali ion composition with a higher computation cost
[11,12]. The average voltage between the composition x
1 and x
2 is
V = -AGz,
(x2-x1)ze'
(3.10) where AG is the Gibbs free energy for the reaction between the alkali ion composition of x
1 and x
2 in the cathode. The reaction free energy can be considered in three parts by the equation AG = AE + PAV TAS. The internal energy change, AE, can be obtained from first principles calculations. The PAV term can be neglected in a solid-state reaction where the volume changes are usually small. In fact PAV is in the order of 10- eV, whereas AE is in the order of 3 to 4 eV per formula unit. TAS can also be neglected as it is only the order of thermal energy, which is about 0.025 eV. Therefore, we can replace
A G by AE in equation 3.10 to calculate the voltage with small errors [13,14].
3.2.2 Alkali ionic diffusivity
Diffusion is a non-equilibrium phenomenon that refers to the transport of atoms over a chemical potential gradient. Alkali ionic diffusion in an intercalation reaction is not too far from equilibrium but rather evolves between local equilibrium states. During the alkali ionic diffusion, alkali ions spend most of their time at the equilibrium sites of the crystal, and only a small fraction of time is spent on the paths connecting the adjacent equilibrium sites. Therefore, the alkali ionic motion can be regarded as a succession of
40

discrete hops. To obtain the ionic diffusivity, it is necessary to calculate the frequency with which alkali ions hop between adjacent sites. A good estimation of the hop frequency is given by the transition state theory, in which the hop frequency is written as
[15] r = v * exp
(
,Eb (3.11) where v is a vibrational prefactor, kB is the Boltzmann constant and T is the temperature.
AEb is the activation energy barrier.
The elastic band method can be used to calculate the activation energy barrier [16]. In the calculation, the initial and end states of a hop are calculated first. Then some intermediate states are created by interpolation of the initial and end states. As these intermediate states are metastable, they are bound to one another with a so-called "elastic band" during the calculation so that they do not relax back to their stable initial or end points. After relaxing all the intermediate states, the activation energy barrier can be estimated. All the possible diffusion paths need to be calculated and the paths with the lowest energy barrier will be used to evaluate the diffusivity.
3.2.3 Kinetics of the structure transformation
During the deintercalation/intercalation reaction, structural transformations may happen in the electrode compound. These transformations usually involve the migration of alkali and/or transition metal ions [17]. Therefore, the rate of the structural transformation is sometimes also determined by the rate of the ionic migration. If the position of the
41

activated state is known, an alternative approach to the elastic band method called the selective dynamics method can be used to calculate the activation energy barrier [18]. In the selective dynamics scheme, the migrating ion at the unstable activation state is fixed relative to distant ions in the supercell. Thus the migrating ion is held at the unstable position without being relaxed to the stable initial or end position during the relaxation.
On the other hand, the ions surrounding the migration ion are allowed to freely relax.
This approach relies on the assumption that atoms are not distorted significantly from their crystal positions by the presence of a far away defect. In this thesis, the selective dynamic method is used to calculate the rate of the transition metal migration during the structural transformation.
References
[1] B.H. Bransden, C.J. Joacin, Introduction To Quantum Mechanics, Longman Group
UK Limited, 1989.
[2] R.G. Parr, W. Yang, Density Functional Theory Of Atoms and Molecules, Oxford
England, 1989.
[3] P. Hohenberg, W. Kohn, Phys. Rev. 136 (1964) 864.
[4] W. Kohn, L.J. Sham, Phys. Rev. 140 (1965) Al 133-Al 138.
[5] J.P. Perdew, W. Yue, Phys. Rev. B 33 (1986) 8800-8802.
[6] J.P. Perdew, K. Burke, M. Ernzerhof, Phys. Rev. Lett. 77 (1996) 3865-3868.
[7] G. Kresse, J. Furthmtller, Phys. Rev. B 54 (1996) 11169-11186.
[8] G. Kresse, J. Furthmuller, Comp. Mater. Sci. 6 (1996) 15-50.
[9] S.K. Mishra, G. Ceder, Phys. Rev. B 59 (1999) 6120-6130.
42

[10] F. Zhou, M. Cococcioni, C. Marianetti, D. Morgan, G. Ceder, Phys. Rev. B 70
(2004) 1-8.
[11] A. Van der Ven, G. Ceder, Electrochem. Comm. 6 (2004) 1045-1050.
[12] A. Van der Ven, M.K. Aydinol, G. Ceder, G. Kresse, J. Hafner, Phys. Rev. B 58
(1998) 2975-2987.
[13] M. Aydinol, A. Kohan, G. Ceder, K. Cho, J. Joannopoulos, Phys. Rev. B 56 (1997)
1354-1365.
[14] M.K. Aydinol, G. Ceder, J. Electrochem. Soc. 144 (1997) 3832.
[15] G. Vineyard, J. Phys. Chem. Solids 3 (1957) 121-127.
[16] G. Mills, H. Jonsson, G.K. Schenter, Surf. Sci. 324 (1995) 305-337.
[17] J. Reed, G. Ceder, A. Van der Ven, Electrochem. Solid-State Lett. 4 (2001) A78.
[18] J.S. Reed, Ph.D. Thesis, pp 253-257, MIT, USA, 2003.
43

2
A transition metal redox couple is necessary to produce charge compensation for the intercalation of ions in a crystal structure. Among 3d transition metals, the Ni
2
+/Ni
4 redox couple contributes two electrons per atom. More importantly, the voltages of
Ni 2
+/Ni and Ni
3
+/Ni
4
* are very similar and both in the range of practical interest. The interesting Ni
2
+/Ni
4
+ redox couple is due to its unique electronic configuration in the octahedral coordination. Fig. 4-1 shows the electronic configuration of Ni 2
+, Ni
3
+ and
Ni4* in an octahedral coordination, where five d orbitals split into eg orbitals with two fold degeneracy, and t2g levels with three fold degeneracy. While Ni
4
+ has a relatively stable configuration with filled t2g orbitals, Ni 3
+ becomes unstable because of its unpaired electron in an eg orbital. The unpaired electron can be stabilized by either removing it
(Ni
4
*) or making it paired (Ni 2
+). Therefore, the transition between Ni
2
+ and Ni
4
* is feasible during an intercalation reaction [1].
44

2
+
t2g
3
+
4
Figure 4-1 Schematic electronic configuration of Ni at different valence state in the octahedral coordination.
As Ni
2
+/Ni
4 + contributes two electrons per atom, only half the amount of Ni is needed in a layered LiTMO
2
(TM = transition metals) compound, offering the feasibility to incorporate half of some other metals to stabilize the structure or for other purposes.
Several Li intercalation electrodes with the composition LiNiXTM(VX)0
2
(TM = Ti, Mn) have been synthesized and evaluated. In these materials, only Ni participates in the electrochemical reaction while TM acts only as a structure stabilizer [2-6]. A material such as LiNil/
2
Mnl/20
2 is perfectly balanced in extractable Li and electron content. Even though Mn remains unchanged in valence, Ni2+ can be oxidized to Ni4+ thereby supplying all the electrons needed to extract Li+. While this is ideal from a capacity perspective, it is likely that the rate capability could be increased by adding more Ni 2
+ to the system. Li extraction from LiNi/
2
Mn/20
2 creates Ni
4
*, which at the high state of charge leads to a reduction in the electronic conductivity [7]. Oxidation of Ni 2
+ to Ni
4 also leads to a significant increase in the activation barrier for Li hopping [8,9]. The
45

electrostatic repulsion between Li+ in its activated state and the transition metal ion has been shown to have a significant influence on the Li migration barrier [8-10]. In fact, a recent computational study has investigated the Li migration barriers of a series of common transition metal cations shown in Table 4-1. Among them, Ni2+ was found to have the lowest energy barrier [9]. Hence, one would expect that increasing the Ni
2 content in layered LiNixTM(X)O
2 materials could lead to materials that have improved rate capability at the high state of charge. Since the average valence of the transition metals TM in layered LiTMO
2 materials needs to be +3, more Ni 2
+ can only be accommodated by incorporation of high valent cations in the structure. Using a +5 cation such as Sb 5 increases the Ni content to 66% as in LiNi
2
/
3
Sb/
3
O
2
. Even when all Li is removed from this material there should still be 1/6 of residual Ni 2
+ present, and at a typical charge limit corresponding to removal of 2/3 Li [4,11,12] one third of all Ni would still be Ni
2
+ assuming the other Ni ions have been oxidized to +4.
Table 4-1 The calculated activation barrier for Li motion for various transition metals near the activated state. [8]
Transition metal Co4+ Ni Mn 4 Ni 3 + Co-+ Cu
2
+ Ni
2
Activation barrier (meV) 490 490 340 310 310 270 210
On the other hand, the effect of incorporating higher valent cations on the structural stability needs to be considered. By replacing half Ni with Mn into LiNiO
2
LiNio.
5
Mno.50
2 can be cycled up to 200 mAh/g with excellent capacity retention,
46

implying a better structural stability than LiNiO
2
[12,13]. This suggests that incorporation of 4+ cations into LiNixTM
1
.,0
2 can enhance the stability. It would also be interesting to investigate the structural stability when a higher valent cation such as Sb 5 is incorporated.
When synthesized through a solid state route LiNi
2
/3
Sb
1
/30
2 forms an orthorhombic structure with Fddd space group [14]. To create a layered structure we synthesized
LiNi
2
/
3
Sb/30
2 through an ion exchange route from NaNi
2
/
3
Sb/30
2
. Since the layered R-
3m structure is stabilized by a large size difference between the alkali and the TM cations, it is more likely to form a layered structure with a larger alkali ion such as Na+
(Table 4-2) [15,16]. Indeed, NaNi
2
/
3
Sb
1
/30
2 has been synthesized with a solid state process by V. Nalbandyan, et al. and the structure was confirmed to be layered R-3m
[17]. Ion exchange is a soft chemical approach performed at relatively low temperature so that only Na+ is replaced by Li+ with the rest of the structure intact. It is reasonable to expect a well-layered LiNi
2
/
3
Sb
1
30
2 with very little cation mixing through such a synthesis route.
Table 4-2 Ionic radii of the related metals in the octahedral coordination. [18]
Metals Na+ Li+ Ni
2
+
Ni 4
+
Sb 5
Ionic radii (A) 1.02 0.76 0.69 0.48 0.6
47

In this chapter, the layered LiNi
2
/
3
Sbl/30
2 was prepared by an ion exchange method. We also study the structure and electrochemical properties, and discuss the structural stability during intercalation by both experimental and first principles techniques.
4.2.1 Synthesis of NaNi
2
/3
Sbl/
3
0
2
LiNi
2
/
3
Sb
1
/30
2 was prepared by solid state reaction from Na
2
CO
3
(99.5%, Aldrich), Sb
2
O
3
(99.9%, Alfa Aesar) and Ni(OH)
2
(61% Ni, Alfa Aesar). Stoichiometric amount of these starting materials were ball-milled for 12 h. After drying, the mixture was ground and pressed into a pellet. The pellet was heated in air for 24h at three different temperatures
(800, 900, and 1000'C). The resulting products are compared and the optimized sodium precursor was chosen for ion exchange.
4.2.2 Ion exchange
The obtained powder was mixed with 10 times excess amount of the eutectic composition of LiNO
3
(99.9+%, ACROS) and LiC (99.9+%, Mallinckrodt Baker). The mixture was heated at 2800C for 8h in air. After ion exchange, the mixture was rinsed several times with distilled water and filtered to recover the powder. The resulting powder was dried overnight in the air.
48

4.2.3 X-ray diffraction and TEM
XRD patterns were collected using a Rigaku diffractometer equipped with Cu KAradiation by step scanning in the 20 range of 10-80'. Rietveld refinement and profile matching of the powder diffraction data were performed with Fullprof [19]. In the refinement, the oxygen occupancy is fixed to the default value of 2 for the layered structure with space group R-3m. For the lithium or sodium and transition layers, we restrict both the Ni, Sb and Li/Na occupancy on the 3a sites, and the Ni, Li/Na occupancy on 3b sites, to sum to 1.
Electron diffraction patterns were collected under an accelerating voltage of 200 keV on a JEOL 2010 microscope. The powders were suspended on a copper grid with lacey carbon.
4.2.4 Electrochemical properties
Electrochemical cells were configured in the following way: Li/i M LiPF
6 in EC:DMC =
1:1/ LiNi
2
/3
Sb
1
30
2 with carbon black (15 wt%) used as a conductive agent and polyethylenetetrafluoride (PTFE) (5 wt%) as a binder. Cells were assembled in an argon filled glovebox and cycled at room temperature using a Maccor 2200 operating in galvanostatic mode.
4.2.5 Calculation methodology
49

Energies are derived from first principles calculations based on density functional theory
(DFT). The spin polarized generalized gradient approximation, Perdue-Wang exchange correlation function, and the projector augmented-wave method were used as implemented in Vienna ab initio simulation package (VASP) [20]. A plane-wave basis with a kinetic energy cutoff of 370 eV was used. A reciprocal-space k-point grid of
5 x 5 x 3 or 3 x
x 5 was used depending on the size of the supercell considered. Structures were fully relaxed. The +U correction term in the Dudarev scheme was used with U
5.96 for Ni only [21].
4.3.1 Structure of the precursor
XRD patterns for NaNi
2
/
3
SbI/30
2 precursors synthesized at three different temperatures
800, 900 and 1000
0
C are shown in Fig. 4-2 and the results of the refinement are listed in
Table 4-3. All three patterns can be well refined using the a-NaFeO
2
(R-3m) structure with sodium in 3b(O, 0, 0.5) sites, transition metals (Ni, Sb) in 3a(0, 0, 0) sites and oxygen in 6c(0, 0, z) sites. Both the Na occupancy and Na/Ni exchange are allowed to vary in the refinement. For NaNi
2
/
3
SbI/
3
0
2 synthesized at 800'C, the Na occupancy is
0.938 and the Na/Ni exchange is determined to be 0.9%. For the compound synthesized at 900'C, the Na occupancy and Na/Ni exchange are 0.987% and 1%, respectively. For the Na phase precursor synthesized at 1000*C, Na occupancy is 0.954 and the Na/Ni
50

cation mixing is 4.5%. We choose the 900'C material for ion exchange as it has high Na occupancy with low Na/Ni exchange. The lattice parameters a=3.06 A, c=16.05
agree very well with the reported ones in reference [17], where a=3.06
c=16.02
9000C
Ij I A.
10 20 30 40
Figure 4-2 XRD patterns of NaNi
2
/3
SbI/30
2 synthesized at 800, 900 and 1000
0
C.
Table 4-3 Refinement results for Na synthesized at three different temperatures
Synthesis temperature (*C) Na occupancy
Na/Ni inter-layer mixing
800 0.9
93.8
98.4 900
1000 95.4
1.0
4.5
51

4.3.2 Structure of LiNi
2
/
3
Sb
1
30
2 after ion exchange
The XRD pattern for LiNi
2
/
3
SbI/30
2 after ion exchange is shown in Fig. 4-3(a). The difference between the observed and calculated XRD patterns for the R-3m structure is also indicated. The results of the Rietveld refinement in Table 4-4 gives a Li occupancy
0.973 which is a little bit lower than the Na amount in precursor. Also, the Li/Ni exchange is only 0.2%, which can essentially be neglected. The lattice parameters are a=2.99
c=14.56
Our XRD data indicates that the starting sodium precursor phase is not present anymore after the ion exchange although a small amount of residual Na ions may still be present in the final material.
52

(a)
TrIw~I...................................
(b)
C
4-
C I-
Ia
10
20 30 40 50 60 70 80
C
4-0
C
10 20
40 50 60 70
20
Figure 4-3 (a) X-ray diffraction pattern and refinement of LiNi
2
/
3
Sbl/30
2 obtained after
8h ion exchange from NaNi
2
/3
Sbl/30
2 synthesized at 900'C; (b) XRD pattern of
53

LiNi
2
/
3
SbI/
3
O
2 calculated by first principles methods with -VxV4ahex ordering of Ni and
Sb in the TM layer.
Table 4-4 Structure refinement parameters for LiNi
2
/3Sbl/
3
0
2 after ion exchange at 280'C atom site x/a y/b z/c occupancy
0.973
Li(1)
Ni(2)
Ni(1)
Li(2)
Sb(1)
0(1)
3b 0 0
3b 0 0
3a 0 0
3a 0 0
3a 0 0
6c 0 0
0.5
0.5
0
0
0
0.25678
0.002
0.664
0.002
0.333
2.000
a = 2.99 A, c = 14.56 A, Rp = 15.1%, Rp = 16.71%,/ = 1.81%
Fig. 4-4(a) and (b) shows two representative electron diffraction patterns from zone axes
[111]hex and [512 1 hex of the pristine LiNi
2
/
3
SbvI
3
0
2
, respectively. The fundamental reflections and the zone axes are indexed to the parent hexagonal cell with rhombohedral symmetry and space group R-3m. The (11l)(l=3n) type fundamental reflections are clearly tripled with additional superstructure intensities, indicating that a v3x
N3ahex superstructure is present in LiNi
2
/
3
SbyI
3
0
2
. Ordering in a NF3xFdahex supercell is common on a triangular lattice at composition 1/3 [22] and has also been speculated to occur in LiNil/
2
Mn
1
20
2 with 10-12% Li/Ni interlayer mixing [23]. The ordering pattern is a honeycomb of Ni
2
+ with Sb 5 in the center of the honeycomb thereby maximizing the
Ni 2
+/Sb
5
+ nearest neighbor contacts.
54

(a)
2T3
112 e
.
.101
[1T]
(C) 2 213
4
-113
(d)
[512] iO 4 aft1e2 102
015
113
[1821.
213 [251]
Figure 4-4 (a) (b) TEM patterns for the pristine LiNi2/3Sb1/302;
(c) (d) TEM pattern for
LiNi2/3Sb1/302 after 10 cycles, in which the superstructure is enhanced.
We also investigated the structure of LiNi
2
/
3
Sb
1 i
3
O
2 with first principles calculations. In our calculation, the transition metal layer is ordered in the V3xV'3ahex superstructure and no inter-layer cation mixing is considered. Computational details are the same as in previous work on related materials [6]. The calculated lattice parameters are a=3.015
c=14.709
which are within 1% deviation from the experimental values. The simulated
55

XRD pattern based on these parameters is shown in Fig. 4-3(b), and agrees very well with the experimental results. The small peaks around 20' in both experimental and calculated pattern are the result of the superstructure in the transition layer.
In LiNi
2
/
3
Sb
1
/30
2 synthesized by solid-state method [14], the structure is a rock salt superstructure in the orthorhombic space group Fddd, where the isolated SbO
6 octahedra share edges with 12 adjacent octahedra randomly occupied by Li or Ni. This arrangement maximizes the Sb-Sb distances.
4.3.3 Electrochemical results
The theoretical capacity of LiNi
2
/
3
Sbl/30
2 is 225.4 mAh/g assuming that all the lithium ions can be extracted. As shown in Fig. 4-5, we find a first discharge capacity around 92 mAh/g at C/20 rate, which is less than half of the theoretical capacity. After 10 charge/discharge cycles, the capacity drops to 38 mAh/g. Fig. 4-5 shows that the capacity decays rapidly in the first 10 cycles.
56

5.0
4.5-
0
4.0
2.5-
20 40 60 80
Capacity (mAh/g)
100 120
Figure 4-5 Charge and discharge capacity for first ten cycles measured at C/20 rate. The voltage window is 2.5-4.6 V.
4.3.4 Structural degradation during electrochemical reaction
To investigate possible structural changes as the origin for the capacity degradation, the
XRD pattern for an electrode was collected after it had been cycled five times. The XRD spectrum and refinement are shown in Fig. 4-6. The refinement results show that the
Li/Ni inter-layer mixing grows from 0.2% to around 10.4%. This dramatic change in structure was also confirmed from TEM patterns. Fig.4-4(c) and (d) shows the [182]hex and [251]hex zone axes patterns collected from the sample in the discharged state after 10 charge/discharge cycles. The FXV'3ahex superstructure is still present with somewhat enhanced intensities. Such phenomenon is observed in all ten crystals examined by TEM.
57

Single crystal electron diffraction is known to be more sensitive to superstructure reflections than powder X-ray diffraction [24]. The superstructure peaks in XRD are completely diminished due to the amorphous carbonaceous additives and polymer binders in the electrode. Both XRD and TEM show that, although the pristine material has no cation mixing at all, once it is electrochemically cycled nickel moves into lithium layer.
cj)
C
Mod
10 20
W11 ri i1h IN WOO i
K L
30 40 50
20
60 70 80
Figure 4-6 XRD results of LiNi
2
/
3
Sbi
1 3
O
2 electrode after five cycles. The Rietveld refinement indicates that cation mixing is around 10.4%.
To investigate whether Ni can migrate into Li vacancies, we calculated the activation barrier for Ni migration when the material is partially delithiated (Li concentration is
58

2/3). There are two possible pathways for the nickel migration: one is the most direct path travelling straight through the edge shared by neighboring octahedra; the other is a longer pathway where nickel goes through a nearest neighboring tetrahedral site via the faces that shares with the neighboring octahedra. Our previous work has consistently shown the tetrahedral path Oh -Td -Oh to be substantially lower in energy than the direct octahedral hop for any ion migration in layered structures [2,10,25]. Fig. 4-7 demonstrates the migration pathway. One nickel migrates into the tetrahedral site through the triangular face between the TM layer and Li layer. A Li trivacancy around this tetrahedral defect is necessary to prevent face sharing of Li and the Ni tetrahedral defect.
Meanwhile, one Li will move to the tetrahedral site that face-shares with the octahedral
TM site vacated by Ni to lower the energy and form a Li/Ni dumbbell structure. We calculated the energy along this migration pathway in a Li8Ni
8
Sb
4
O
24 unit cell with a 3 x
3 x 5 k-point mesh. Fig. 4-8(a) shows that the energy barrier is around 0.63 eV. It is likely that this barrier is too low to keep the material stable. The Boltzmann success rate, exp(-0.63 eV/kBT) 1.87 x 10-11 at room temperature. Assuming a vibrational pre-factor
of 1012 to 101 S-1, the Ni hopping rate would be around 20-200 s-
1
. One may also compare this barrier to the Mn migration barrier in half delithiated LiMnO
2
(-0.4 eV) and the Co migration barrier in Lit/
2
CoO
2
(-1.6 eV) [25]. Experimental evidence indicates that half delithiated LiMnO
2 will essentially transform into a spinel-like structure, while the Lil/
2
CoO
2 layered structure remains stable due to the much higher migration barrier [26,27]. The Ni migration barrier in LiNi
2
/
3
Sb1/
3
O
2 is between that of
Mn and Co but substantially closer to that of Mn in Li
1
/
2
MnO
2
. This seems further indication that some Ni will migrate into Li/vacancy layer resulting in a layered structure
59

with high Li/Ni inter-layer mixing consistent with our XRD results. The Ni migration also explains the observation in electron diffraction (Fig. 4-4) that the V'3xV3ahe, superstructure appears with somewhat enhanced intensities after cycling. When the material is cycled some Ni migrate into the lithium layer leaving the Ni sites vacated.
Due to the larger contrast between Sb and a vacancy than between Sb and Ni, the superstructure in electron diffraction is enhanced.
1 -(I)
2
LI
~Lithium
3
Lithium vacancy
o *Oxygen
Figure 4-7 Demonstration of the nickel migration and the formation of a Li/Ni dumbbell.
The process can be divided into three steps: (1) Li disorder in a partially delithiated structure creates a trivacancy around a tetrahedron in the Li layer. (2) A single Ni ion moves from the TM layer into the triangular face between the TM and Li layer, and ultimately into the tetrahedral site of the Li layer. (3) The latter step is usually accompanied by a lithium ion moving into the tetrahedral site that shares a face with the octahedral site that has been vacated by the Ni ion. This leads to a Litet-Nitet dumbbell
60

around the TM vacancy and is believed to be a key defect in LiNij/
2
Mn/
2
O
2
[11] and an intermediate state in the migration of Mn in LiMnO
2
[25].
In our calculations, the valence state of the migrating Ni can be determined by integrating the electron spin density in a sphere about the Ni cation centers [2,25]. Fig. 4-8(b) shows the net electron spin (in units of
2 ptB) as a function of the integration radius. The net spin rises fast initially as the 3d orbitals of Ni are integrated over. When the integration reaches the nonpolarized oxygen, however, the net spin levels off. The total spin (around
2 in units of 1/29B) at the plateau in Fig. 4-8(b) indicates that the migrating Ni corresponds to a high spin Ni2+ ion. Our work on LiNil/
2
Mn/
2
O
2 has indicated that both the location of the Ni in the transition metal layer and the redox state of the Ni are critical factors that appear to control Ni motion [28]. In our layered LiNi
2
/
3
Sb
1
/30
2
, each Ni is surrounded by three Sb ions and three Ni ions. A Nitet site may be occupied or a Li/Ni dumbbell may form only when the Ni redox state is 2+, since both Ni
3
+ and Ni
4
+ have a strong preference for the octahedral coordination [29]. If Ni is oxidized to 3+/4+ before migrating to a tetrahedral site, it becomes immobile. This may explain why not all Ni are prompted to migrate during charge and discharge. The comparatively lower migration barrier for Ni2+ in LiNi
2
/
3
Sb
1
/30
2 than in LiNi/
2
Mn
1
20
2 can be explained by the fact that the divalent Ni ion is surrounded by three Sb 5 and three Ni 3
+ when LiNi
2
/
3
Sb
1
30
2 is partially charged, and such strong electrostatic repulsion provides the driving force for the Ni 2
+ to migrate. It is important to realize the driving force may vary for Ni ions with different local environments and redox states. In fact, in previous first principles calculations on LiNil/
2
Mn/20
2 we have shown that Ni 2
+ in the transition metal layer
61

surrounded by six Mn" is unstable when it is oxidized to a higher redox state [28].
(a)
0.6
0.4
>
< 0.0
-0.2
(b) 2.4-.
2.0
1.6-
-a 1.2 --
(I
0.8
octa
Q-
-+-tetra
0.0-
0.0 0.2 0.4 0.6 0.8 1.0 1.2
Figure 4-8 (a) Energy for a Ni ion along the path from an octahedral site in the TM layer to a tetrahedral site in the Li/vacancy layer. (b) Integrated net spin for Ni cations along
62

the migration path in LiNi
2
/3
Sbl/30
2
Using similar calculations, the possibility of Sb 5 migration in LiNi
2
/
3
Sb
1
/
3
0
2 is also investigated. The calculated results show that the energy required to create a Li/Sb dumbbell is about 1.5 eV making Sb 5 migration essentially impossible.
4.3.5 Relations between structure degradation and capacity decay
The Li/Ni dumbbell structure in partially delithiated electrodes can influence the electrochemical properties in two ways. Firstly, occupation of a tetrahedral site in the lithium layer by nickel requires that the nearest three octahedral sites are always vacant thereby reducing the insertion capacity. Additionally, the most likely pathway for lithium diffusion is identified as octahedral-tetrahedral-octahedral [10,30] so that stable Li/Ni dumbbells will block some possible lithium diffusion pathways. Hence, both the capacity and rate capability may be reduced by the formation of Li/Ni dumbbell.
4.4 Conclusion
Layered LiNi
2
/
3
SbI/
3
O
2 was successfully synthesized by ion exchange from
NaNi
2
/
3
Sb/
3
O
2 and tested in Li half-cells. Both experimental and computational results imply that nickel migrates from the transition metal layer to the Li layer when the material is partially delithiated, leading to rapid capacity fade.
63

References
[1] K. Kang, Ph.D. Thesis, pp 45-48, MIT, Cambridge, USA, 2006.
[2] J. Reed, G. Ceder, Electrochem. Solid-State Lett. 5 (2002) A145.
[3] Z. Lu, D.D. Macneil, J.R. Dahn, Electrochem. Solid-State Lett. 4 (2001) A191.
[4] T. Ohzuku, Y. Makimura, Chem. Lett. (2001) 744.
[5] B. Ammundsen, J. Paulsen, Adv. Mater. 13 (2001) 943-956.
[6] K. Kang, D. Carlier, J. Reed, E.M. Arroyo, G. Ceder, L. Croguennec, C. Delmas,
Chem. Mater. 15 (2003) 4503-4507.
[7] S. Kumar, H.L. Li, J. Breger, C.P. Grey, Y. Shao-Horn, International Meeting on
Lithium Batteries (IMLB) 13 (2006).
[8] K. Kang, Y.S. Meng, J. Breger, C.P. Grey, G. Ceder, Science 311 (2006) 977-980.
[9] K. Kang, G. Ceder, Phys. Rev. B 74 (2006) 094105.
[10] A. Van der Ven, G. Ceder, Electrochem. Solid-State Lett. 3 (2000) 301.
[11] J. Brdger, Y.S. Meng, Y. Hinuma, S. Kumar, K. Kang, Y. Shao-Horn, G. Ceder,
C.P. Grey, Chem. Mater. 18 (2006) 4768-4781.
[12] Y. Makimura, T. Ohzuku, J. Power Sources 119-121 (2003) 156-160.
[13] T. Ohzuku, A. Ueda, M. Nagayama, J. Electrochem. Soc. 140 (1993) 1862.
[14] G.C. Mather, R.I. Smith, J.M.S. Skakle, J.G. Fletcher, M.A. Castellannos, M.P.
Gutierrez, A.R. West, J. Mater. Chem. 5 (1995) 1177-1182.
[15] E.J. Wu, P.D. Tepesch, G. Ceder, Philos. Mag. B 77 (1998) 1039-1047.
[16] T.A. Hewston, B.L. Chamberland, J. Phys. Chem. Solids 48 (1987) 97-108.
[17] V. Nalbandyan, I. Shukaev, 0. Smirnova, ICDD PDF#00-0530344, 2002.
[18] R.D. Shannon, Acta Cryst. A32 (1976) 751.
64

[19] Fullprof available at the Following URL: http://www.llb.cea.fr/fullweb/%OAfp2k
/fp2k.htm.
[20] G. Kresse, J. Furthmuller, Comp. Mater. Sci. 6 (1996) 15-50.
[21] F. Zhou, M. Cococcioni, C. Marianetti, D. Morgan, G. Ceder, Phys. Rev. B 70
(2004) 1-8.
[22] G. Ceder, A. Van der Ven, Electrochim. Acta 45 (1999) 131-150.
[23] Y.S. Meng, G. Ceder, C.P. Grey, W.S. Yoon, M. Jiang, J. Breger, Y. Shao-Horn,
Chem. Mater 17 (2005) 2386-2394.
[24] B. Fultz, J.M. Howe, Transmission Electron Microscopy and Diffractometry of
Materials, 2nd ed., Springer.
[25] J. Reed, G. Ceder, Chem. Rev. 104 (2004) 4513-33.
[26] H. Wang, Y.-I. Jang, B. Huang, D.R. Sadoway, Y.-M. Chiang, J. Electrochem.
Soc. 146 (1999) 473.
[27] H. Gabrisch, R. Yazami, B. Fultz, J. Power Sources 119-121 (2003) 674-679.
[28] Y. Hinuma, Y.S. Meng, K. Kang, G. Ceder, Chem. Mater 19 (2007) 1790-1800.
[29] B.N. Figgis, M.A. Hitchman, Ligand Theory
2000.
and Its Applications, wiley-VCH,
[30] A. Van der Ven, G. Ceder, J. Power Sources 97-98 (2001) 529-531.
65

2
5.1.1 Layered AMO
2
(A = Na, Li; M = 3d transition metal) system as electrode materials
Layered LixMO
2 compounds have been intensively studied as positive electrode materials for lithium ion batteries, and LiCoO
2 and LiNiO
2
-based materials operate well as Liintercalation materials. For several other 3d transition metals LiMO
2 either does not easily form a layered structure (M = Ti, Fe, Mn), or does not reversibly deintercalate lithium (M = V, Cr, Fe, Mn) [1-3]. Indeed, layered LiFeO
2 and LiMnO
2 can only be obtained from their Na phase by Li/Na ion exchange [3-5] but LiFeO
2 does not deintercalate lithium well [6,7], and LiMnO
2 transforms to spinel in the first charge or discharge [8]. This situation is in sharp contrast to the NaMO
2 compounds. NaxMO
2 systems form layered structures more easily due to the larger ionic size difference between alkaline and transition metals [9,10]. The deintercalation and intercalation of sodium in layered NaMO
2 has been reported for M = Ti, V, Cr, Mn, Fe, Co, Ni [2,11
18]. Among these 3d transition metals, Mn is interesting because of its low cost and
66

proper redox potential. In the following, we will give a brief review of the layered
LiMnO
2 and NaMnO
2
5.1.2 Electrochemical properties of the layered LiMnO
2
Layered 03-LiMnO
2 is not thermodynamically stable, but can be synthesized by ion exchange from layered 03-NaMnO
2
[3]. When the layered LiMnO
2 is cycled as a positive electrode, the capacity decays rapidly though the capacity of first charge can be as high as 200 mAh/g as is shown in Fig. 5-1(a) [3]. A closer investigation shows that the voltage profile changes after the first cycle. After the first charge, the discharge shows two plateaus around 4.0 V and 3.0 V in Fig. 5-1(b) [19]. This voltage profile agrees with the intercalation of LiMn
2
0
4 spinel, where the 4.0 V plateau corresponds to the reaction of LiMn
2
0
4
<-4 Li+ + e- + Mn
2
0
4 and the 3.0 V plateau to the reaction of LiMn
2
0
4
Li + e- --- Li
2
Mn
2
O
4
[19]. Therefore, the analysis of the voltage profile indicates that the layered structure has been transformed into a spinel related structure after first charge.
Structural analysis techniques such as TEM also confirmed the layered to spinel transformation [20]. First principles calculations have proposed a model of Mn2+ migration to explain the mechanism of the structural transformation [21,22].
67

(a)
5.0
4.5
4.0
Ir
3.5
-
2.5
0
200 m /g
5 10
15
--
20
(b)
-F/
V v J. A I .0
4
0.
F V: i~~
0
2 p * p-
-0.~
()
25
IE/
3.0
V VS 1A
3.5 4.0 4.5
Figure 5-1 (a) Cycling of the layered LiMnO
2 electrode between 3.4 V to 4.3 V vs.
Li/Lie (Figure from reference [3]); (b) Voltage profile and differential capacity vs.
voltage for the cycling of the layered LiMnO
2 between 2.0 V to 4.5 V at the rate of C/15
(Figures from reference [19]).
68

5.1.3 Structure and electrochemical properties of the layered NaMnO
2
The known phases of NaxMn0
2
(x =0.2, 0.40, 0.44, 0.70, 1) have been summarized by
Parant, et al [23]. There are two phases for NaMnO
2
. Low temperature a- NaMnO
2 has an 03 layered structure with a monoclinic structural distortion due to the Jahn-Teller distortion of the Mn 3 ion. At high temperature, the orthorhombic 0- NaMnO
2 forms in a different layered structure containing MnO
2 sheets consisting of a double stack of edgesharing MnO
6 octahedra. Na occupies the octahedral sites between two neighboring sheets [15]. First principles computations indicate that the monoclinic NaMnO
2 is energetically more stable than other competing phases [24]. This is in contrast to
LiMnO
2
, which prefers an orthorhombic structure [24,25].
Both a- and 0- NaMnO
2 have been electrochemically tested as a positive electrode material by Mendiboure, et al. in 1985. Their results showed that only 0.22 and 0.15 Na could be reversibly extracted and re-intercalated in a- and 0- NaMnO
2 respectively.
Besides NaMnO
2
, NaOAMnO
2
, Nao.
6
MnO
2
, and Nao
7
MnO2.2
5 have also been studied as a positive electrode in sodium ion batteries [15,26]. Among them, P2-Nao.
6
MnO
2 shows the highest capacity of about 150 mAh/g at the first cycle. The capacity decays to about 70 mAh/g over 10 cycles [26].
69

5.1.4 Motivation
As LiMnO
2 and NaMnO
2 share the same 03 layered structure, the difference in structural stability can only arise from the effects of the alkali ions. A comparison of their electrochemical properties together with some structural analysis techniques can provide information about the role the alkali ions play on the structural stability, which will give us insights to improve the structural stability of LiMnO
2 during deintercalation.
In this chapter, we will report the synthesis and electrochemical testing of monoclinic a-
NaMnO
2
. Our results show that 0.85 Na can be deintercalated and 0.8 Na intercalated back reversibly, corresponding to 210 mAh/g charge capacity and 197 mAh/g discharge capacity. The reasonable capacity retention indicates that the layered NaMnO
2 is structurally stable upon deintercalation in contrast with layered LiMnO
2
. We also explain why the different alkali ions will result in distinct structural stability.
NaMnO
2 was synthesized by solid-state reaction. Stoichiometric amounts of NaCO
3
(100%, Baker) and Mn
2
O
3
(98%, Alfa Aesar) were mixed and ball milled in acetone for 6 hours at 300 rpm rate. The mixture was dried into a powder. About 0.5g of powder was pressed into a pellet. The pellet was fired at 700
0
C in air for 10 hours before it was quenched to room temperature and moved to a glove box filled with argon.
70

X-ray diffraction (XRD) patterns were collected on a Rigaku Rotaflex or PANalytical
X'pert pro diffractometer equipped with Cu Ka radiation in the 20 range of 10-80'. All the samples were sealed with Kapton film to avoid air exposure. Rietveld refinement and profile matching of the powder diffraction data of the as-prepared NaMnO
2 were performed with Fullprof using space group C2/m.
The atomic ratio of Na and Mn in as-prepared NaMnO
2 powder was determined by direct current plasma (DCP) emission spectroscopy (Luvak Inc., Boylston, MA 01505). After cycling, the electrolyte and the Na metal anode were collected and analyzed for Mn content by ICP technique (HORIBA Jobin Yvon ACTIVA ICP-AES Spectrometer) to investigate potential Mn dissolution.
Electrochemical cells were configured in the following way: Na/ 1 M NaPF
6 in EC:DMC
= 1:1/ NaMnO
2 with carbon black (15 wt %) as conductive agent and polyethylenetetrafluoride (PTFE) (5 wt %) as binder. The 1 M NaPF
6 in EC:DMC electrolyte was prepared by dissolving anhydrous NaPF
6
(98%, Sigma Aldrich) into
EC:DMC (anhydrous, Sigma Aldrich, 1:1 in volume ratio). Two pieces of glass fiber served as separators and stainless steel as current collectors in Swagelok cells, assembled in an argon-filled glove box. The cells were cycled at room temperature using a Maccor
4000 operating in the galvanostatic mode.
The Na content vs. voltage was measured by potentiostatic intermittent titration (PITT) on a Solartron 1287 electrochemical potentiostat. Steps of 10 mV were taken to fully
71

charge and discharge the cell. The capacity was measured at each voltage step until the current was below C/50.
5.3
The XRD pattern of as-prepared monoclinic NaMnO
2 is shown in Fig. 5-2. The background and three broad peaks between 10-30* are from the Kapton film. Rietveld refinement gives the lattice parameters a = 5.672
b = 2.856
c = 5.807
P =
113.20. These values are fairly close to the results of Mendiboure, et al. (a = 5.63
b =
2.86
c = 5.77
P = 112.90) [4,15,23,27]. The Mn-O bond lengths in the MnO
6 octahedron are 2.39
(2x) and 1.94
(4x) respectively, confirming that Mn
3 is Jahn-
Teller active [24,27].
C
SI 11 11 I l iii I I II| | | IuIII l Illhl lil Il l
10 20 30 40 50
Two Theta
60 70 80
72

Figure 5-2 XRD pattern and refinement of as-prepared NaMnO
2
. The red (black) line represents the experimental (calculated) data. The residual discrepancy is shown in blue. The refinement results are preformed in the C2/m space group and give R.p=
10.2%, and - = 9.98.
The DCP data shows that the as-prepared NaMnO
2 consists of 20.5 % Na and 49.1 %
Mn in weight, giving an atomic ratio of Na and Mn to be 1:1. Assuming that the asprepared NaMnO
2 is 100% phase-pure and that the remainder 30.4 % of the total weight is oxygen, the chemical formula can be determined as NaMnO
2
.
06
Fig. 5-3 shows the capacity vs. voltage measured by PITT on the as-prepared
NaMnO
2
. Fig. 5-3(a), (b) and (c) are for the 1't,
2 nd and
3 rd PITT cycle respectively.
The results were plotted in a manner resembling cyclic voltammograms except that the current is replaced by the capacity. The results show a series of peaks in both charge and discharge, indicating a multitude of possible phase transitions as the system is de-sodiated. In the
1 st
PITT cycle, the charge process consists of eight oxidation peaks: 2.59V, 2.63V, 2.73V, 2.80V, 2.97V, 3.14V, 3.48V and 3.59V, while the discharge process consists of only five considerably less pronounced reduction peaks: 2.47V, 2.62V, 2.87V, 3.08V, 3.45V. Similar differences between the charge and discharge profile were found in the sequential two PITT cycles with the same cell. In the
2 nd and
3 rd cycles, most peaks are at the same positions, except that two oxidation peaks at 2.59V, 2.63V are merged into one peak at 2.56V and one new peak appears at 2.91V (shown in Fig. 5-3(b) and (c)). The distinctly different charge and
73

discharge peaks indicate that the charge and discharge process probably go through different reaction paths with different intermediate phases. However, this hysteresis seems to be reversible, given that the charge and discharge profile are largely preserved in the
2
"d and
3 rd cycle.
(a)
0.
4g0.
3
0.
2 u0.
1
(b)
-0.
1-
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8
Voltage (V)
0.
4
0.
3
0.
2-
I 0.
1
0
-0.
1
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8
Vdtage (V)
74

(c)
0.2-
0 .1
0
-0.1-
2 2.2 2.4 2.6 2.6 3 3.2 3.4 3.6 3.8
Voltage (V)
Figure 5-3 Capacity vs voltage of Na deintercalation (upper) and intercalation
(lower) for NaMnO
2 measured by PITT with 1 OmV steps. (a) is for the first cycle, (b) for the second cycle and (c) for the third cycle.
The data in Fig. 5-3 can also be represented in Fig. 5-4 where Na content is plotted vs. voltage. The Na content is calculated from adding the capacity at each voltage step in the PITT measurement. The initial Na content is 1 as determined by DCP.
The series of plateaus in Fig. 5-4 correspond to the peaks in Fig. 5-3. The difference between charge and discharge can also be observed in the different charge/discharge voltage profiles. About 0.85 Na, corresponding to a charge capacity of 210mAh/g, can be deintercalated from NaMnO
2 when the cell is charged up to 3.8 V vs. Na/Nat
75

in PITT. Upon discharge to 2 V, 0.8 Na (corresponding to discharge capacity
197mAh/g) can be reversibly intercalated back.
3.8
3.6
3.4
53.2
.3
2.8
0
> 2.6
2.4
2.2
2
-
I
0.2
I
0.4
I
0.6
x of Na p
0.8
1
Figure 5-4 Voltage profile of NaMnO
2 upon Na deintercalation and intercalation measured by PITT. The cell is potentiostatically charged up to 3.8 V vs Na/Nat and discharged to 2.0 V.
Fig. 5-5 shows the galvanostatic charge and discharge of NaMnO
2 at C/10 and C/30 at various cycles (IC = 240mA/g). At C/10, the charge/discharge capacities in the first and tenth cycle are 214/185 mAh/g and 165/149 mAh/g respectively. At C/30,
209/194 mAh/g is achieved in the first cycle and 207/144 mAh/g in the tenth cycle.
The coulombic efficiencies are low at both C/10 and C/30 rate (C/10: 86.4% for first cycle, 90.3% for tenth cycle; C/30: 92.8% for first cycle, 70% for tenth cycle). The low coulombic efficiency implies that side reactions, possibly with the electrolyte,
76

take place during charge, especially at higher voltage. When testing with a C/30 rate, the cell experiences high voltage for longer time, causing more parasitic reaction and/or degradation, hence the lower Coulombic efficiency. It is not clear from our data whether the poor Coulombic efficiency is the result of materials degradation, or due to a self-discharge shuttle in the electrolyte. Fig. 5-6 compares the voltage profiles of NaMnO
2 charged up to 3.8 V and 4.2 V respectively. When the cell is charged up to 4.2 V, it shows an additional charge capacity of about 90 mAh/g between 3.8 V and 4.2 V. However, this additional capacity is not shown during discharge, indicating the charge capacity between 3.8 V and 4.2 V is mainly from the decomposition of the electrolyte. On the other hand, the electrode material NaMnO
2 does not seem to be significantly affected by the overcharge as the discharges show similar profiles after the cell is charged up to 3.8 V and 4.2 V.
(a)
10
3.5
.
3
2
0 50 100 150
Capacity (mAhg)
200
77

(b)
3.6
3.4
3.2
S3 i2.8
7
2.6
2.4
2.2
2
0 50 100 150
Capacity (mAh/g)
Figure 5-5 Voltage profile of NaMnO
2 after multiple cycles at C/10 (a) and C/30 (b) respectively. The cell is galvanostatically cycled between 2.0 V and 3.8 V.
41
-
a3.5
~35
-
2.5
2
0
-
50 100 150
----
200 250
Figure 5-6 Comparison of the voltage profiles of NaMnO 2 charged up to 3.8 V and
4.2 V respectively. The cell was galvanostatically cycled at C/10 rate.
78

We have investigated the long voltage plateau at 2.63 V, starting from ~Nao.
93
MnO
2 and ending at ~Nao.
7 oMnO
2
. In Fig. 5-7(a) we show the result of current relaxations after 1 OmV voltage steps in the first charge. In the voltage plateau region, the current is initially low after a small potential step, but then slowly ramps up to its maximum after about 2.5 hours. This current behavior is typical of a two-phase reaction with a significant phase transformation barrier (i.e. a first order phase transition) [28].
Initially, the current is low as nucleation needs to occur, but once nucleation starts the current increases significantly as the second phase grows. The first charge in Fig. 5-5 also shows the typical voltage overshoot at both C/10 and C/30, typically associated with the onset of a first order phase transformation.
(a)
3.8-
3.6-
3.4
&3.2
03
-
2.8
2.6
2.4
1
0.06
0.05
6
11
Tim (h)
16 21
-0.03
0.02
26
0.01
79

(b)
3.8
3.6
3.4
s3.2
CD
S3
2.8
2.6
2.4
2.2
2
1 6
I
('((((a fin-'
-0.05
I
11
Time (h)
16
' '
'
21
' '_''_L
26
-0.06
I
-0.01
-0.02
-0.03 1
E
C
-0.04 )
Figure 5-7 Voltage steps (black) and the corresponding current relaxation (red) during
PITT measurements in the first charge (a) and discharge (b) cycle.
To confirm the two-phase reaction, ex situ XRD analysis was performed on an electrode that was removed from a cell, which was charged to the middle of the plateau (nominal Na content about 0.83). Fig. 5-8 compares the XRD pattern of the charged electrode Na
0
.
8 3
MnO
2 to that of uncharged NaMnO
2
. The results confirm the two-phase coexistence: One phase in the XRD of the partially charged cathode material matches with pristine NaMnO
2 with very little peak shift. From Fig. 5-4, we estimate the composition of this phase to be ~Nao.
93
MnO
2
. The lattice parameters of the Nao.
93
MnO
2 phase are a = 5.671 A, b = 2.817
c = 5.806
6 = 113.2*, as obtained by fitting the diffraction pattern. All the values are almost identical to those
80

of the pristine NaMnO
2 except that the b-lattice parameter has contracted by 1.4%.
From the charge/discharge curve in Fig. 5-4, we estimate that the composition of the second phase in the XRD pattern of the partially charged material, is ~Nao.
7
OMnO
2
The Nao70MnO
2 phase is indexed in Fig. 5-8 on a monoclinic lattice. The lattice parameters of this NaO.
7 oMnO
2 phase are shown in Table 5-1. In layered NaxMnO
2
Na and Mn are ordered in alternative (111) planes of an FCC cubic close-packed oxygen framework. The monoclinic cell used for the Nao.
7
0MnO
2 phase, contains six metal layers, contrary to two layers in the Nao
93
MnO
2 phase. This cell tripling can be due to the Na/vacancy ordering (both in-plane and inter-plane) that lowers the symmetry or to oxygen layer gliding to modify the 03-stacking [29]. To compare the lattice parameters of the coexisting two phases, we use the tripled unit cell for
Nao.
93
MnO
2 as indicated by the dashed lines in Fig. 5-9.
10-
8[
C -
100
2
10 20 30 40
Two Theta
50 60 70 K
81

Figure 5-8 XRD of fresh NaMnO
2 electrode (black) and partially charged Na
0
.
8 3
MnO
2
(red). New peaks corresponding to Na
0
.
70
MnO
2 are indexed using a monoclinic lattice.
00,-000
* 1.
CI
19 0
1
a
Figure 5-9 Structure of monoclinic Nao.
93
MnO
2 projected in a-c plane. (purple -
Mn, yellow
-
Na, red
-
Oxygen) The solid line indicates the unit cell of monoclinic
Na
0
.
93
MnO
2
. The super cell indicated by the dashed line is more comparable to the unit cell of the Na,0
7
0MnO
2 phase.
Table 5-1 compares the lattice parameters of Na
0
.oMnO
2 and Na.
93
MnO
2 in the same supercell setting. The changes in lattice constants upon Na removal are significant, with a reduced by 9.6% and b and c increasing. The angle 6 also changes by 2.6%.
The significant lattice misfit implies considerable strain between the two phases, which could be responsible for the significant phase transformation barrier observed in the current relaxation data in Fig. 5-7(a).
82

Table 5-1. Comparison of lattice parameters of Nao.
9
3
MnO
2 and Nao.
7 oMnO
2
Samples Misfit (%)
a (A)
b (A) c (A)
NaO.
9 3
MnO
2
5.671
2.817
16.054
Nao.
7
0MnO
2
5.126
2.921
16.737
9.6
3.7
4.3
(0)
Volume (A 3 )
94.25
255.76
91.82
250.50
2.6
2.1
It has been reported that chemically synthesized Nao.70MnO
2 usually shows oxygen over-stoichiometry [30], so the compound is usually written as Nao.
7
0MnO
2
+y
(0 <y <
0.25). There are two known structures reported for Nao.7oMnO
2
+y depending on the amount of excess oxygen [23,30]. One is orthorhombic with space group Cmcm (for
0 < y < 0.05) and the other hexagonal with space group P63/mmc (for 0.05 < y <
0.25). The XRD pattern of the Nao.70MnO
2 phase we obtained by electrochemical deintercalation, however, does not match either of the above two phases. Its diffraction pattern is actually quite similar to that of Na
0
.
7
4
CrO
2 whose structure is unsolved as of yet [31].
Some capacity decline with cycling is evident from Fig. 5-5. The relative decay of discharge capacity is more severe at C/30 rate than that at C/10. To investigate the reason for the capacity decay, a cell cycled for 20 cycles was dismantled in the discharged state, and the cathode electrode washed in anhydrous DMC in an argon
83

filled glove box for XRD. Fig. 5-10 compares the XRD patterns of the cycled and fresh electrodes. The two patterns are very similar, but the peaks after cycling become broadened. Rietveld refinement results from the cycled electrode gives lattice parameters: a = 5.654
b = 2.861
c = 5.808
= 112.970, less than 0.3% different from the refinement results of pristine NaMnO
2
. In the peak shape refinement, the Gaussian composition of the peak broadening does not change after cycling but the Lorentzian composition increases. The related parameters are x =
0.247, y = 0.073 in the fresh electrode and x = 0.159, y = 0.257 in the cycled electrode respectively. The Gaussian broadening x is typically related to crystal strain, while y is related to crystal size and faulting [32]. In the cycled electrode, x is decreased implying less crystal strain. The increase of y parameter indicates that either the crystal size decreased, or more likely, that more faulting occurred.
10
.
6
After 20 cycles
4
2
0 i Fres Eectrods
10 20 30 40
Two Theoa
50 60 70 s0
Figure 5-10 XRD patterns of fresh electrode (black) and cycled electrode (red).
84

With Mn3 containing cathode materials, possible dissolution of Mn due to its disproportionation to Mn2 is always a cause of concern [33]. To find out how much
Mn dissolved into the electrolyte or deposited on the Na anode surface after cycling, the electrolyte from the cycled cell and the Na anode surface were collected for analysis. The electrolyte was diluted into water, and the surface of the Na anode was also scraped and dissolved into water to form a dilute solution. The diluted solutions were analyzed by ICP to measure the concentration of Na and Mn. In the data analysis of the electrolyte, the Na concentration was used as a reference since the Na concentration in the undiluted electrolyte is known as 1 mol/l. For the surface of the
Na anode, the total weight of the dilute solution was measured to estimate the amount of the possible Mn deposition. Table 5-2 shows the Na and Mn concentration in the diluted solutions. The Mn concentration in the fresh electrolyte and on the fresh Na anode surface is negligible. In the diluted solution of the cycled electrolyte, the concentration of Mn is less than 0.2 ppm. As the Na concentration in undiluted electrolyte is known as 1 mol// (~ 23 x 103 ppm), we can estimate the concentration of
Mn in undiluted cycled electrolyte to be less than 32 ppm. In our cell the electrolyte is less than 1 ml, therefore the amount of Mn dissolved in the electrolyte is less than 32 pg. The typical cathode weight in our cell is about 2-3 mg. Therefore, only a few percent of Mn is dissolved, which is unlikely to cause the significant capacity decay we observed. It should however not be excluded that dissolved Mn can be deposited on other cell components, such as in the separator.
85

Table 5-2. Summary of ICP results of Na anode surface, fresh and cycled electrolyte in the water diluted solution.
Samples
Fresh electrolyte
Cycled electrolyte
Na anode surface
Na (ppm)
361
145
701
Mn (ppm)
-0
<0.2
-0
We achieve a very high charge and discharge capacity from monoclinic NaMnO
2 with the 03 structure. Our results are significantly better than those obtained by
Mendiboure, et al, who evaluated monoclinic NaMnO
2 with 1 M NaClO
4 in PC as electrolyte [15]. Their open circuit voltage curve in Fig. 5-11(a) shows a voltage plateau around 2.7 V ranging from Na
0
.
93
MnO
2 to Nao.
73
MnO
2
, which is close to the plateau we observe at 2.63 V. However, only about 0.22 Na could be deintercalated/intercalated from NaMnO
2 in their work, and no obvious voltage plateau appeared in their study (Fig. 5-11(b)). In our results, about 0.7 Na can be deintercalated up to a charge cutoff of 3.5V. The reasons for this difference are not clear though they may be related to the difference in electrolyte used.
86

.
(VJ
4.07
B
70pA/cm 2
0'3
3.0
2.5-
2.0
1.5
phas
.75
A
I
1- A-
I I X
.8 .85 .9 .95 1.0
Figure 5-11 (a) Electrochemical cycling curve from monoclinic NaMnO
2
; (b) Open circuit voltage from monoclinic NaMnO
2
[15]. 1 M NaClO
4 in PC was used as the electrolyte.
The voltage versus capacity shows very pronounced features with strong voltage steps and plateaus indicative of phase transitions upon desodiation. Such strong features have also been observed for intercalation of Na in NaCoO
2
[34], but are less common in Li-intercalation systems unless structural or electronic transitions [35] are involved. The phase changes in NaMnO2 can be due to Na-vacancy ordering or transitions that involve the gliding of oxygen planes. The latter is a possibility in Nadeintercalation reactions as Na prefers both octahedral and trigonal prismatic
87

environments and the trigonal prismatic coordination can only be achieved in an 03 stacking by sliding some of the oxygen layers [11].
Careful inspection of Fig. 5-4 and 5-5 indicates that the sequence of phase transformation on charge and discharge is not the same. This is consistent with the strong first order character of the phase transformations. When the two phases on each side of the transition are very distinct, nucleation is often not symmetric. This is similar to the asymmetry between the melting and solidification transition of solids.
At this point we do not understand why transitions in NaxMn0
2 are so pronounced. It is possible that Na-vacancy ordering interactions are stronger than for Li-vacancy due to the larger size of Na, which would add large elastic contributions to the screened electrostatic interactions between the alkali and the vacancy [36]. Another likely possibility is the occurrence of structural transitions, which affect the oxygen stacking. Such oxygen layer gliding would modify the Na coordination and allow it to be optimized at each Na concentrations. In each case, the strong phase transitions in
Na-intercalation oxides may create more hysteresis and reduce the energy efficiency, similar to the problems with conversion reactions in lithium systems [37]. Besides the voltage profile, the current responses also show some asymmetry in Fig. 5-7(a) and
(b). Specifically, the current responses around 2.5 V during discharge become more like a second order phase transition in contrast with the first order transition at 2.63 V during charge.
88

Several other NaMO
2
(M = V, Cr, Fe) oxides have been evaluated as intercalation cathodes. While NaMO
2
(M
=
V, Cr, Fe) can be reversibly cycled up to Nao.
5
MO
2 only a very small amount of Na can be intercalated back when these materials are further charged [17,18,38]. However, NaMnO
2 can be deintercalated up to
Nao.
15
MnO
2 with 0.8 Na being reversibly intercalated back. Secondly, for NaMO
2
(M
= V, Cr, Fe) the reversibility is improved with lower charge cutoff voltage [17,18,38], but for NaMnO
2 the cyclability becomes worse when the charge voltage is lowered from 3.8 V to 3.4 V, as is shown in Fig. 5-12.
1 3 2
3.4
3.2
3-
E
0 2.8-
62.6
2.4
2.2
0 20 40 60 s0 100
Capacity (mAh/g)
120 140 160
Figure 5-12 Cyclability of NaMnO
2 at lower cutoff voltage. The cell is cycled at C/10 within voltage window 2-3.4 V.
It is instructive to compare the remarkable difference in the electrochemical behavior of NaMnO
2 with that of LiMnO
2
. Layered LiMnO
2 can be obtained from NaMnO
2 by
89

ion exchange. During cycling of LiMnO
2
, the capacity decays rapidly and the structure becomes spinel-like [8]. The transformation from layered to spinel requires the migration of Mn ions from the transition metal layer into the alkali layer. It is clear from comparing the electrochemical behavior of 03-LiMnO
2 to 03-NaMnO
2 that the alkali plays a critical role in this migration. Removing the alkali from both compounds leads to the same structures, but the LiMnO
2 system transforms to spinel while NaMnO
2 does not. This difference in behavior is consistent with the model proposed by Reed, et al. using first principles computation [21]. Reed, et al. argued that in LiMnO
2
, Mn can migrate into tetrahedral sites when trivacancies are formed in the Li layer. Meanwhile, a Li ion can move to a tetrahedral site in the adjacent Li layer to form the Li/Mn dumbbell configuration as is shown in Fig. 5-13. This type of dumbbell configuration can stabilize the Mn in tetrahedral sites and serves as nucleus for the layered to spinel phase transformation. It was pointed out that Mn in a tetrahedral site without an accompanying Li in a tetrahedral site was actually unstable energetically [39]. In NaMnO
2
, however, Na is unlikely to move to a tetrahedral site and can therefore not stabilize Mn in a tetrahedral site. Hence, with Na, the intermediate state in the transformation of layered to spinel is not stable giving the layered NaxMnO
2 good metastability, thereby resulting in good cyclability of the material.
90

~}
*%
L!
Lithium
Maganese
Li Vacancy
Mn Vacancy
Oxygen Gee*
Figure 5-13 Demonstration of the Li/Mn dumbbell configuration in layered LiMnO
2
Mn migrates to the tetrahedral site in Li layer leaving a Mn vacancy in Mn layer. Li trivacancy is necessary to avoid face sharing with Mn.
Monoclinic NaMnO
2 is synthesized and tested electrochemically as a positive electrode material for a sodium ion battery. About 0.85 Na can be deintercalated out of NaMnO
2 and 0.8 Na can be reversibly intercalated back. The charge and discharge profile show a different series of plateaus indicating different reaction paths with different intermediate phases. Current relaxation in PITT indicates a phase transformation barrier at the 2.63 V plateau, which was confirmed to be a two-phase reaction by XRD. While some capacity fade was observed after 20 cycles, no significant structural change was found after cycling. From our results it is clear that cyclability of corresponding Na and Li compounds can be very different.
91

[1] L. Zhang, K. Takada, N. Ohta, M. Osada, T. Sasaki, J. Power Sources 174 (2007)
1007-1011.
[2] S. Komaba, C. Takei, T. Nakayama, A. Ogata, N. Yabuuchi, Electrochem. Comm.
12 (2010) 355-358.
[3] A.R. Armstrong, P.G. Bruce, Nature 381 (1996) 499-500.
[4] B. Fuchs, S. Kemmler-Sack, Solid State Ionics 68 (1994) 279-285.
[5] T. Shirane, R. Kanno, Y. Kawamoto, Y. Takeda, M. Takano, T. Kamiyama, F.
Izumi, Solid State Ionics 79 (1995) 227-233.
[6] M. Tabuchi, C. Masquelier, T. Takeuchi, K. Ado, I. Matsubara, T. Shirane, R.
Kanno, S. Tsutsui, S. Nasu, H. Sakaebe, 0. Nakamura, Solid State Ionics 90
(1996) 129-132.
[7] J. Li, J. Luo, L. Wang, X. He, Int. J. Electrochem. Sci. 6 (2011) 1550 1561.
[8] Y. Shao-Horn, S.A. Hackney, A.R. Amstrong, P.G. Bruce, R. Gitzendanner, C.S.
Johnson, M.M. Thackeray, J. Electrochem. Soc. 146 (1999) 2404.
[9] E.J. Wu, P.D. Tepesch, G. Ceder, Philos. Mag. B 77 (1998) 1039-1047.
[10] T.A. Hewston, B.L. Chamberland, J. Phys. Chem. Solids 48 (1987) 97-108.
[11] C. Delmas, J.J. Braconnier, C. Fouassier, P. Hagenmuller, Solid State Ionics 3
(1981) 165-169.
[12] J.J. Braconnier, C. Delmas, P. Hagenmuller, Mat. Res. Bull. 17 (1982) 993-1000.
[13] A. Maazaz, C. Delmas, P. Hagenmuller, J. Inclusion Phenom. 1 (1983) 45-5 1.
[14] S. Kikkawa, S. Miyazaki, M. Koizumi, J. Power Sources 14 (1985) 231-234.
[15] A. Mendiboure, C. Delmas, P. Hagenmuller, J. Solid State Chem. 57 (1985) 323-
331.
[16] L.W. Shacklette, T.R. Jow, L. Townsend, J. Electrochem. Soc. 135 (1988) 2669.
[17] Y. Takeda, K. Nakahara, M. Nishijima, N. Imanishi, 0. Yamamoto, M. Takano, R.
Kanno, Mat. Res. Bull. 29 (1994) 659-666.
92

[18] C. Didier, M. Guignard, C. Denage, 0. Szajwaj, S. Ito, I. Saadoune, J. Darriet, C.
Delmas, Electrochem. Solid-State Lett. 14 (2011) A75.
[19] G. Vitins, K. West, J. Electrochem. Soc. 144 (1997) 2587.
[20] Y. Shao-Horn, S.A. Hackney, A.R. Armstrong, P.G. Bruce, R. Gitzendanner, C.S.
Johnson, M.M. Thackeray, J. Electrochem. Soc. 146 (1999) 2404.
[21] J. Reed, G. Ceder, Chem. Rev. 104 (2004) 4513-33.
[22] J. Reed, G. Ceder, A. Van der Ven, Electrochem. Solid-State Lett. 4 (2001) A78.
[23] J.P. Parant, R. Olazcuaga, M. Devalette, C. Fouassier, E.T.P. Hagenmuller, J. Solid
State Chem. 3 (1971) 1-11.
[24] 0.1. Velikokhatnyi, C.-C. Chang, P.N. Kumta, J. Electrochem. Soc. 150 (2003)
A1262.
[25] R. Hoppe, G. Brachtel, M. Jansen, Z. Anorg. Allg. Chem. 417 (1975) 1.
[26] A. Caballero, L. Hernan, J. Morales, L. Sanchez, J. Santos Pefia, M.A.G. Aranda,
J. Mater. Chem. 12 (2002) 1142-1147.
[27] M. Jansen, R. Hoppe, Z. Anorg. Allg. Chem. 399 (1973) 163.
[28] N. Meethong, H.-Y.S. Huang, S.A. Speakman, W.C. Carter, Y.-M. Chiang, Adv.
Funct. Mater. 17 (2007) 1115-1123.
[29] Y. Hinuma, Y. Meng, G. Ceder, Phys. Rev. B 77 (2008) 1-16.
[30] R. Stoyanova, D. Carlier, M. Sendova-Vassileva, M. Yoncheva, E. Zhecheva, D.
Nihtianova, C. Delmas, J. Solid State Chem. 183 (2010) 1372-1379.
[31] S. Miyazaki, S. Kikkawa, M. Koizumi, Rev. Chim. Miner. 19 (1982) 301.
[32] R.J. Hill, Advances in X-Ray Analysis 35 (1992) 25.
[33] Y. Terada, Y. Nishiwaki, I. Nakai, F. Nishikawa, J. Power Sources 97-98 (2001)
420-422.
[34] R. Berthelot, D. Carlier, C. Delmas, Nat. Mat. 33 (2010).
[35] M. Mendtrier, I. Saadoune, S. Levasseur, C. Delmas, J. Mater. Chem. 9 (1999)
1135-1140.
93

[36] A. Van der Ven, M.K. Aydinol, G. Ceder, G. Kresse, J. Hafner, Phys. Rev. B 58
(1998) 2975-2987.
[37] R.E. Doe, K. Persson, Y.S. Meng, G. Ceder, Chem. Mater. 20 (2008) 5274-5283.
[38] S. Komaba, T. Nakayama, A. Ogata, T. Shimizu, C. Takei, S. Takada, A. Hokura,
I. Nakai, ECS Trans. 16 (2009) 43-55.
[39] J.S. Reed, Ph.D. Thesis "Ab-Initio Study of Cathode Materials for Lithium
Batteries", pp 74-76, MIT, USA, 2003.
94

1
Developing positive electrode materials with high energy density is one of the key challenges for lithium ion batteries. High energy density can be obtained either by high voltage or high capacity. Several compounds have been investigated for their high voltage [1-3]. LiNio.
5
Mni.
5
0
4 is one of these that can work with a conventional carbonate based electrolyte though some side reaction occurs [4-7]. With average voltage around
4.7V and comparable capacity (around 140mAh/g) to LiCoO
2
(~620Wh/kg) and LiFePO
4
(~591Wh/kg), LiNio.
5
Mn.
5
0
4
(~658Wh/kg) gives higher specific energy than many commercialized compounds.
6.1.1 The structure of LiNio.
5
Mn
1
.
5
O
4
Spinel LiNio.
5
Mn.504 has a cubic structure where Ni and Mn occupy the octahedral sites of a close packed oxygen framework while Li sits in the tetrahedral site. In a stoichiometric LiNio.
5
Mni.50
4
, Ni and Mn are ordered in the octahedral sites, and the space group of the structure is P4
3
32. Fig. 6-1 shows the structure of the ordered
LiNio.
5
Mn
1
.
5
0
4 spinel. When the material is heated above ~ 650-700'C in air, oxygen deficiency often occurs accompanied by some Mn + being reduced to Mn 3
+.
This also
95

leads to disordering of Ni and Mn, restoring the symmetry to Fd-3m. The lattice parameter also increases, as Mn
3
* is larger than Mn . With further oxygen loss at a temperature above -750-800'C, a secondary rock salt phase with the space group Fm-3m appears. The composition of the secondary phase was proposed to be LixNi..O by comparison of the XRD pattern [8]. However, because of the low scattering factor of Li and the similarity in those of Ni and Mn, other possible stoichiometries such as
(Lio.
3
33
Nio.
167
Mno.
5
),O also agree with the XRD pattern [9]. On the other hand, the composition of the secondary phase also varies with temperatures. Above ~95
0
'C, the rock salt phase becomes the major phase and the metal:oxygen ratio in the rock salt phase increases with higher oxygen deficiency. When the temperature increases to ~ 100'C with the oxygen deficiency above 0.8, the material is close to stoichiometric single-phase rock salt, perhaps with small amount of the secondary spinel phase remaining. The changes of the structures and phases with temperatures are summarized as follows: [9]
LiNio
5
: 0 < 6 _ 0.05, spinel single phase, P4
3
32
LiNio.
5
MnI.
5
O
4
-
6
: 0.05 < 6 < 0.18, spinel single phase, Fd-3m
Two phases: 0.05 < 6 < 0.18, spinel phase, Fd-3m, and rock salt phase, Fm-3m
Two phases: 0.05 < 6 < 0.18, rock salt main phase, Fm-3m and secondary spinel phase,
Fd-3m.
96

These changes represent thermodynamically reversible behavior. Upon slow cooling, oxygen deficiency can be restored, and the structures and phases change reversibly. The oxygen deficiency is also affected by the oxygen partial pressure during synthesis under the same temperature [10].
Figure 6-1 Schematic structure of the ordered LiNio.
5
Mn
1
.
5
0
4
. (Li green, Ni grey, Mn
purple, 0 - red)
6.1.2 The electrochemical properties of LiNio.
5
Mn
1
.
5
0
4
Fig. 6-2 from reference [11] shows the reversible potential vs. the composition of Li in the ordered LiNio.
5
Mni.
5
0
4
. At the voltage of about 4.7 V vs. Li/Li+, one Li can be deintercalated from LiNio.
5
Mn
1
.
5
0
4 to form Nio.
5
Mnj.
5
0
4 giving a capacity of 148 mAh/g.
Li can also intercalates into LiNio.
5
Mni.
5
0
4 to form Li
2
Nio.
5
Mn
1
.
5
0
4 at the voltage of about 2.8 V with another capacity of about 148 mAh/g. The reaction can be written as
97

Li++e-+Lite'NiMn0
4 <-+
Li""Ni
0
Mn.
5
O
4
. It is necessary to point out that the intercalated Li ions are in the octahedral sites. Meanwhile, the intercalated Li pushes the pre-existed Li in the tetrahedral sites to octahedral sites. This reaction involves a cubic to tetragonal phase transformation with one lattice parameter expanded and the other two shrunk. Between 1.4 V and 1.9 V, the half Mn 4 in Li
2
Nio.
5
Mn
1
.
5
O
4 can be further reduced to Mn
3
+ with an additional capacity of about 74 mAh/g. The reaction can be formally written as Li2[Ni2'Mn3+Mn4l
4+5Li
+O.5e-
<
Li.
[Ni20.Mn
14. As the octahedral sites in Li
2
Nio.
5
Mn
1
.
5
0
4 has already been fully occupied, the additional half Li can be in the tetrahedral sites but not yet confirmed. Overall, the total reversible capacity of LiNio.
5
Mn
1
.
5
0
4 a spinel can be as high as 370 mAh/g, consisting of three regions: 4.7
V (148 mAh/g), 2.8 V (148 mAh/g) and 1.4-1.9 V (74 mAh/g). However, only the capacity at 4.7 V is practically used as the reaction at 2.8 V was found to be very slow
[11]. The cubic to tetragonal phase transformation at -2.8 V may also cause some structural degradation and deteriorate the cyclability.
98

6.0 -
I
5.0 -
I I
1
I
I
I
> 4.0
3.0
2.0 -
1.0 -
0.0
0.0 0.5 1.0 1.5 2.0 x in Li,[Ni
1
,2Mn,]O
4
2.5 3.0
Figure 6-2 The reversible potential vs. the Li composition in LixNio.
5
Mn
1
.
5
0
4
[11].
For the disordered LiNio.
5
Mn.
5
0
4
.5, the voltage profile around 4.7 V becomes less flat and the average voltage is slightly lower than that of the ordered LiNio.
5
Mn.
5
0
4
. Most investigators have shown disordered LiNio.
5
Mni.
5
04.s to have better rate capability and cyclability than ordered LiNio.
5
Mn.
5
0
4
, though the difference in performance between the two phases varies [8,12,13]. One explanation put forward is that the disordered phase shows higher electronic conductivity than the ordered one due to the existence of Mn
3
* in disordered phase [8,13]. However, Mn
3
+ is oxidized to 4+ in the beginning of charge and hence would not contribute much to electronic conductivity in most of the charge/discharge regime. Furthermore, this explanation is based on the assumption that electronic conductivity is rate limiting instead of lithium conductivity, which has not been directly confirmed experimentally or computationally.
99

The rate capability of LiNio.
5
Mni.
5
0
4 also varies significantly with synthesis conditions.
For micron-sized material synthesized by a molten salt method, the ordered phase showed poor rate performance, while the disordered performed better [12]. Creating nano particles, a typical strategy to achieve high rate performance, has given mixed results in
LiNio.
5
Mn
50
4
. While Shaju, et al. showed high rate capability and cyclability with
50nm-sized LiNio.5Mn.
5
0
4
(shown in Fig. 6-3) [14], some other groups did not achieve comparable rate performance on nano-materials with similar size [15,16]. This indicates that particle size may not be the critical factor affecting the electrochemical performance.
On the other hand, nano LiNio.
5
Mnl.
5
0
4 may cause some disadvantages when it is used in the lithium ion battery. Firstly, the reaction between the electrode and the electrolyte at the interface becomes more severe as the surface area increases in nano materials, which will eventually deteriorate the cycle life of the battery. This reactivity issue will become even more problematic for LiNio.
5
Mn .
5
0
4 because of its high voltage, which promotes electrolyte decomposition. Secondly, the lower density of nano LiNi
0
.
5
Mn
1 5
0
4 results in a lower volumetric energy density. Therefore, it is desirable to obtain high rate
LiNio.
5
Mn .
5
0
4 with a large particle size.
100

M140
g'
20C
'80
40
...
0
50 100 150 200 250 300
Figure 6-3 Cycling performance of nano LiNio
5
Mnl.
5
0
4
(both ordered and disordered phases). The cell is charged at C/5 rate to 5.0 V and then discharged at various rates down to 3.5 V [14].
In this chapter, we show results on micron-sized ordered LiNio.
5
Mn
1
.
5
0
4 with very high rate performance and good cyclability. First principle calculations of the lithium diffusion barrier confirm that the intrinsic rate capability of this compound is very high even with micron-sized particles, indicating that LiNio.
5
Mni.
5
0
4 can be a high rate material with high tap density, unlike most nanosystem. The effects of the cell configuration on the rate testing are also investigated and discussed.
6.2 Materials and Methods
101

Synthesis of ordered LiNio.
5
Mn
1
.
5
0
4 spinels. Ordered LiNio.
5
Mnl.
5
0
4 spinel was prepared by solid state reaction from Li
2
CO
3
(99.995%, Alfa Aesar), NiCO
3
(99%, Alfa
Aesar) and MnO
2
(99.9%, Alfa Aesar). A stoichiometric amount of these starting materials was ball milled for 6 hours at 300 rpm with 5mm diameter Yittrium-stabilized
Zirconia balls. After drying, the mixture was ground and pressed into a pellet. The pellet was calcined in air at 900 C for 12 hours and slowly cooled down. The product was ground and re-pelletized. The pellet was annealed at 700
0
C in air for 48 hours and slowly cooled down.
X-ray diffraction and SEM. XRD patterns were collected using a Rigaku diffractometer equipped with Cu Ka radiation by step scanning in the 28 range of 10 -800.
Rietveld refinement and profile matching of the powder diffraction data were performed with Fullprof with space group P4
3
32 for the ordered LiNio.
5
Mni.504.
SEM images were collected under an accelerating voltage of 5KV on a JEOL 6320 microscope. The powders were coated with Carbon by evaporation.
Electrochemical Properties. Electrochemical cells were configured in the following way: Li/lM LiPF
6 in EC:DMC = 1:1/ LiNio.
5
Mni.504 with carbon black (15 wt%) as conductive agent and polyethylenetetrafluoride (PTFE) (5 wt%) as binder. The electrode density is 3 4mg/cm 2 . Two pieces of Celgard 2500, 2325, or C480 separators were used in 2016 coin cells, assembled in an argon-filled glove box, and cycled at room temperature using a Maccor 4000 operating in galvanostatic mode. To control the
102

external pressure on the coin cell the cell was fixed on a C-shape clamp with two stainless steel leads. The load applied on the cell can be controlled with the C-clamp and is measured by inserting a load cell (Omega Engineering Inc., LCM307-1KN) between the coin cell and the C-clamp. The relaxation of the load on the coin cell was recorded with a Keithley 2701 digital multimeter.
The capacity versus voltage was measured by potentiostatic intermittent titration (PITT) on a Solartron 1287 Electrochemical Interface. Steps of IOmV were taken to fully charge and discharge the cell. The capacity was measured at each voltage step until the current was below C/100. Electric impedance spectra were collected on a Solartron 1260
Impedance/gain-phase Analyzer coupled to Solartron 1287 Electrochemical Interface.
The amplitude of the AC signal was kept at 1 OmV and the frequency ranged from 1 MHz to 10 mHz.
Computational Methodology. Energies were obtained with first principles calculations based on density functional theory (DFT). The spin-polarized generalized gradient approximation with Perdue-Wang exchange correlation, and the projector augmentedwave (PAW) method were used as implemented in Vienna ab initio simulation package
(VASP) [17]. A plane-wave basis with a kinetic energy cutoff of 370eV and a reciprocalspace k-point grid of 5 x 5 x 5 were used. The +U correction term in the Dudarev scheme was used with U = 5.96 for Ni and U = 5.0 for Mn respectively [18]. Structures were fully relaxed and the nudged elastic band method was used for determining activation barriers.
103

6.3.1 Structure and Morphology of the as-prepared LiNio.
5
Mn
1
.
5
0
4 spinels
Fig. 6-4 shows the XRD pattern of the sample after the first heat treatment at 900'C. The pattern confirms a disordered spinel with the space group of Fd-3m, and a minor amount of secondary phase with the space group of Fm-3m. The composition of the secondary phase has been suggested to be NiO, LixNil.,O or (Lio.
3 3 3
Nio.
16
7
Mno.
5
)xO [8,9]. Because of the low scattering factor of Li and the similarity in those of Ni and Mn, the exact cation ratio can not be determined from the XRD results.
40
30
-. .
.
.III
10
0
1 0 20 30 40 50
60 70 60
Figure 6-4 X-ray diffraction pattern of the sample after 900'C heat treatment for 12 hours in air. The arrows indicate the peaks that belong to the minor rock salt phase.
104

The XRD pattern of the LiNio.
5
Mn
1
504 annealed at 700'C is shown in Fig. 6-5 with the square root of the intensity plotted to more clearly show weak peaks due to Ni/Mn ordering. After annealing, the peaks from the secondary rock salt phase disappear, and several minor peaks due to Ni/Mn ordering appear. Rietveld refinement gives a lattice parameter a =8.160A in good agreement with the value reported previously for ordered
LiNio.
5
Mni.50
4 synthesized in air [10,11], but different from the value of 8.1176(2)A reported by Pasero, D. et al. Their sample, prepared in high oxygen pressure (I20bar) [9], showed no oxygen deficiency and Ni/Mn disordering as low as 2%. As the lattice parameter increases with the presence of Mn 3 resulting from oxygen deficiency, it is reasonable to conclude that the ordered LiNio.
5
Mn
1
.
5
0
4 synthesized in air, as in this paper and refs [10,11], has a certain amount of oxygen deficiency while retaining Ni/Mn ordering [9].
105

240-.
210
180
S150
120
90
~60-
30
-30
9 17 25 33 41 49 57 65 73 81
Two Theta
Figure 6-5 X-ray diffraction pattern and refinement of as-prepared ordered
LiNio.
5
Mn
1
.
5
0
4 annealed at 700'C. The square root of the intensity is shown to highlight the minor peaks. The red (black) line is the experimental (calculated) data. The residual discrepancy is shown in blue. The refinement results agree with the ordered spinel model and give R, = 8.76%, R, = 12.8% ,
2 =7.75.
The SEM of the as prepared ordered LiNio.
5
Mn
1
504 in Fig. 6-6 shows that the material is well crystallized with a particle size in the range of 3-5 pm .
106

Figure 6-6 SEM of the as-prepared ordered LiNio.5Mni.504 shows an average particle size of 3-5 micrometer.
6.3.2 Electrochemical tests
6.3.2.1 Capacity versus voltage by PITT in ordered LiNio.
5
Mn
1
.
5
0
4
Fig.6-7 shows capacity versus voltage measured by PITT in the as-prepared
LiNio.
5
Mn
1
.
5
0
4
. The results were plotted in a manner resembling cyclic voltammograms
(CV) except that the current is replaced by the capacity [19]. The major oxidation consists of two peaks (-4.73V and ~4.75V) and the corresponding reduction peaks are
-4.71V and -4.73V. The 20mV splitting agrees with the result obtained by Ariyoshi, K.
et al in ordered LiNio.
5
Mnl.504 (~20mV) [11], but is much smaller than that of disordered
LiNio.
5
Mn
1
.
5
0
4
(~60mV) [12,20]. The inset figure shows the additional redox couples.
107

Below 4.7V, there are three redox couples -3.94V/3.93V, ~4.08V/4.07V, ~4.40V/4.33V
as are labeled in Fig. 6-7. Above 4.7V, one additional redox couple was observed at
~4.90V/4.88V. While two of these redox couples (~4.08V/4.07V, ~4.40V/4.33V) have been reported previously in ordered LiNio.sMn .
5
0
4
[10], the other two (-3.94V/3.93V,
~4.90V/4.88V) are found for the first time in this work. The mechanisms of the additional redox couples and their possible effect on the performance of our materials are not clear at present.
0.3
.O
U)
E
0
0
0
0.2
0.1
0.0
-o
0.000
-0.002
3.0
%.
(U
-0.1
-0.2
-
-0.3
3.5
3.5 4.0 4.5 l
04
4.0
Voltage (V)
4.5
5
Figure 6-7 Capacity versus voltage measured upon potentiostatically charge and discharge with IOmV steps (PITT).
6.3.2.2 Rate capability and cyclability
108

The discharge rate capability is shown in Fig. 6-8. The C-rates are calculated based on IC
= 150mA/g, and the charge rate was 0.5C for the high rate discharge tests and 0.2C for the 0.2C discharge. For the ordered LiNiO.5Mnl.504, the discharge capacities are around
147mAh/g, 134mAh/g and 11 OmAh/g at 0.2C, 20C and 40C respectively. The discharge capacity at 20C decays slowly from 137mAh/g at the 2nd cycle to 134mAh/g at the 50th cycle (Fig. 6-8(b)), while at 40C, no obvious capacity decay was seen over 50 cycles.
Coulombic efficiency was measured to be 93.1% for the first cycle upon 0.2C charge and discharge. After 50 cycles, the Coulombic efficiency reaches 98.8%. The disordered
LiNio.
5
Mni.
5
0
4 shows slightly less capacity than the ordered one at a slow rate of C/5 due to the existence of impurities. At the high rates of 20C and 40C, the disordered
LiNio.
5
Mni.
5
0
4 shows similar rate capability as the ordered one.
(a)
5 -
4.5
4. -
0
-Ordered
20 40 80 80 100 120 140
109

(b)
:120
0 u'80-
-
0 20 40 60
Cycle No.
80 10i
Figure 6-8 Discharge rate performance (a) and capacity retention (b) of the ordered and disordered LiNio.
5
Mn
1
.
5
0
4
. The electrode density is 3mg/cm
2 and the electrode contains
15wt% carbon. Two pieces of Celgard C480 separators were used. Charge rate is 0.5C for 20C, 40C discharge and 0.2C for 0.2C discharge. Voltage window is 3-5V.
Rate capability we obtained is better than previously reported results with various synthesis methods [8,12-16,21-28]. It is particularly interesting to compare with the rate capability of small particles. Shaju, K. M. et al synthesized ordered and disordered
LiNio.
5
Mni.
5
0
4 with particle size around 50nm and showed high rate capability
(~l20mAh/g at 20C, ~100mAh/g at 40C for disordered LiNio.
5
Mn
1
.
5
0
4
; ~80mAh/g at
20C, ~50mAh/g at 40C for ordered LiNio.
5
Mn
1
.
5
0
4
) [14]. The LiNio.
5
Mni.504 we synthesized shows average particle size around 3-5 prn but better rate capability than both
110

ordered and disordered nano-LiNio.
5
Mni.
5
0
4
. The fact that around 100-times larger particle shows higher rate capability strongly indicates that ion or electron transport in the particle is not the rate limiting process at least for 40C.
The results of the rate capability in charge are shown in Fig. 6-9(a). The discharge rate was 5C for all the cycles and lower discharge rates did not significantly modify the subsequent charge capacity. Charge capacities for 0.2C, 1C, 5C rates are respectively
149mAh/g, 132mAh/g and 89mAh/g. No significant decay was observed in 20 cycles as is shown in Fig. 6-9(b). One major reason that the charge rate capability can not be as high as the discharge rate capability lies in the fact that the charge voltage is cut at 5V which is only about 0.3V above the equilibrium voltage. Hence, polarization at the higher rate leads to premature charge cutoff.
(a)
5.0]
4.8
4.6
CM
4.4
4.2
4.0-
3.8
0
-- 1C
SC
20 40 60 80 100 120 140
Capacity (mAhlg)
111

(b)
150-
140-
130- "......mm..uum.u
0120 -C
(
E
110-
100
0.2C
80
70
60
50
0 10 20
Cycle No.
.
30
SC
I
40
Figure 6-9 Charge rate performance (a) and capacity retention (b) of LiNio.5Mnl.
5
0
4
. The electrode density is 3mg/cm 2 and the electrode contains 15wt% carbon. Two pieces of
Celgard 2500 separators were used. Discharge rate is 5C. Voltage window is 3-5V.
6.3.3 Effects of the cell configuration on the performance
When testing electrode materials at very high rate it is important to ensure that no other components of the cell and electrode become rate limiting. In addition, homogeneity is required in the electrode as any inhomogeneity in charging or discharging leads to higher effective rates for some fraction of the electrode and hence reduced capacities.
Several aspects of the cell configuration were found to influence the rate performance.
112

6.3.3.1 The effect of the separators
Three different separators were used for comparison. All other conditions were the same when assembling the coin cell. The rate capability and capacity retention with three different separators are compared in Fig. 6-10. The relevant properties of the three separators (Celgard C480, 2500, 2325) are listed in Table 6-1. The C480 separator gives the best performance both in rate capability and cyclability. While C480 is the thinnest separator, the thickness difference does not seem large enough to explain the performance variation based solely on diffusion length through the separator. This is supported by the comparison of performances between Celgard 2325 and 2500 which have the same thickness, but different porosity and pore size.
150~
140
130 so-
50
0
20C
$I120
S110
*.**-****
20C
20C
100 -
90
0.
o 80
70
10
0.2C C480
+
2500
2325
~ i6iu
U
40C
40C
20
Cycle No.
30 40
113

Figure 6-10 Performance comparison for three different Celgard separators. Two pieces of each type of separator were used. Charge rate is 0.5C for 20C, 40C discharge and 0.2C for 0.2C discharge. Voltage window is 3-5V.
Table 6-1. The related properties of three different Celgard separators
Type
C480
2500
2325
Layers
PP/PE/PP
PP
PP/PE/PP
Celgard Separators Comparison
Thickness (pm) Porosity
21.5
25
25
50%
55%
39%
Pore Size (pm)
0.038
0.064
0.028
6.3.3.2 The effect of external pressure on the coin cell
When external pressure was applied to the coin cell by squeezing the C-clamp, the load initially decayed but stabilized after some time. The relaxation is likely due to the creep of the lithium foil, separator and binder in the positive electrode, etc. [29]. Fig. 6-11 (a) shows a typical load relaxation profile when external pressure was applied. The same coin cell was pressed two times. After each pressing, the pressure during relaxation was recorded and the electric impedance measured. Fig. 6-11(b) shows that pressing significantly decreases the resistance that results in the semi-circle in the impedance plot.
114

(a)
(b)
140.
120-
1st pressing
0.
100
S80-
S60-
2nd pressing
EIS III
40-
20-
EIS I
0
EIS 1I
5000
Time
10000
(second)
15000
-8000-
--EIS I
--EIS 11
EIS III
-6000-
E
O -4000-
-200
-100
-2000-
0 i
0
0
0 100 200-
2000 4000 6000 8000
Z' (Ohm)
115

Figure 6-11 The relaxation of external load applied on the coin cell with two sequential pressings is shown in (a). Three electric impedance spectra collected during the external load relaxation are compared in (b). Two pieces of Celgard 2325 separators were used.
The rate capability under two external pressures is compared in Fig. 6-12. The cell, which was pressed up to about 80psi and relaxed for 1 hour, shows low rate capability
(lOOmAh/g at 20C, 60mAh/g at 40C). Another cell, which was pressed up to about
240psi and relaxed for 1 hour, shows much higher rate capability (l30mAh/g at 20C,
1OOmAh/g at 40C). The discharge overpotential also decreased significantly with higher pressure, which agrees with the decrease of impedance.
(a)
5.0
-480p
*-~240 si psi
4.5
C
C
.2C
4.0
3.5
20
OIC
3.0-
I*I* I~I*
0 20 40 60 80 100
Capacity (mAhlg)
10 140 160
116

(b)
.2C
120-
%.100
980- o
60
40-
0
100 -
20C
10
AA
A A
20
Cycle No.
30
+
A
~240psi
A A AA A A
40
Figure 6-12 Rate (a) and cyclability (b) performance for different pressure loaded on the coin cell. Two pieces of 2325 separators were used. Cells were relaxed for one hour after application of pressure.
Gaberscek, M. et al have pointed out that the high frequency impedance semi-circle is due to the contact impedance between the current collector and the electrode [30]. The external mechanical load improves this interface contact, and improves the rate capability of the cell. It is possible that pressure on the cell also improves the Li ion transport in electrolyte. Johns, P. A. et al have shown that Li ion transport in electrolyte especially through the composite electrode can be rate limiting at high rate [31]. It is possible that pressure increases the penetration and wetting of the electrolyte in the composite cathode.
117

6.3.3.3 The effect of the carbon content in the electrode
To investigate the effect the amount of carbon in the electrode has on the rate performance, electrodes with 30wt% active material, 65wt% carbon and 5wt% binder were prepared. The discharge capability of this cell, charged at 5C, and discharged at various rates is shown in Fig. 6-13. After 151 cycles, the discharge capacity reaches
107mAh/g at 83C. After 202 cycles, the discharge capacity reaches 78mAh/g at 167C.
These increases in rate capability observed when diluting the active mass with more carbon are similar to what has been observed in LiFePO
4
[32] and point to the electrode morphology being rate limiting for very fast electrode materials.
(a)
4.4
4.2
0)
0
0
4.0
3.8
3.6
5.0
4.8
4.6
3.2
3.0
0
-
\,67C
5C
83C
I
20
I
I
40 60 80
Capacity (mAh/g)
100 120
118

(b)
180-
160
140a 120-
100 --
80
0.17C
83C
0.17C
-Z 60 -1
M M e 40-
20-
0-
167C
167C
-20
0 20 40 60 80 100 120 140 160 180 200
Cycle No.
Figure 6-13 Rate (a) and cyclability (b) performance of LiNio.
5
Mn .504 with 65wt% carbon. Charge rate is 5C. Voltage window is 3-5V.
6.3.4 First principles calculation of lithium diffusivity in ordered LiNio.
5
Mni.
5
0
4
The diffusion path of Li in spinels is believed to be a three dimensional network [33,34].
Lithium moves from one tetrahedral site to the next by migration through a vacant octahedral site [20]. While in a spinel such as LiMn
2
0
4 all vacant octahedral sites and the lithium migration paths are equivalent, ordering in LiNio.
5
Mn
1
.
5
0
4 creates a distinct Ni site in 4b(5/8,5/8,5/8) and a Mn site in the 12d(1/8, 3/8, -1/8) position of the P4 3
32 space group [11]. Similarly the vacant octahedral sites are divided into 4a(1/8, 1/8, 1/8) and
12d(1/8, 5/8, -3/8) sites. The 4a site is surrouned by 3 Ni and 3 Mn atoms, and the
119

12d(1/8, 3/8, -1/8) site is surrounded by 1 Ni and 5 Mn atoms. Migration of lithium between tetrahedral sites occurs through intermediate vacant octahedral sites. Since each tetrahedral site is surrounded by one 4a and three 12d(1/8, 3/8, -1/8) sites two distinct lithium diffusion paths exist as shown in Fig. 6-14.
Path I through 4a site
Path 11 through 12d site
Li
Oxygen
Figure 6-14 Schematic figure of two distinct paths for lithium migration in ordered
LiNio.
5
MnI.
5
0
4
. Green balls denote Li ions and red balls denote oxygen ions. Li diffusion path I is indicated by blue arrows and the blue ball is the 4a site in the middle of path I.
Dark scarlet arrows show the Li diffusion path II with dark scarlet balls at 12d sites in the middle of path II.
The energy along both paths was calculated in supercells containing eight formula units
(LiNio.
5
Mni.
5
0
4 as a formula unit). The periodic images of the migrating lithium are approximately 8.32
apart. Activation barriers are determined in the dilute Li limit (
120

x "
0 ) by including one Li in an otherwise delithiated cell with eight formula units, and in the dilute vacancy limit (x ~w1) with seven Li and one vacancy site in the cell. All lattice parameters are fixed at x = 0 or x = 1 values for the activation barrier calculations and all internal degrees of freedom are relaxed [35].
Fig. 6-15 shows the energy along the distinct migration paths in lithium rich and vacancy rich phases. Only the energy from the tetrahedral site to the octahedral site is plotted as the other half of the path from octahedral site to tetrahedral site is symmetrically equivalent. In path I the octahedral site surrounded by three Ni and three Mn atoms is effectively the activated state with a barrier of about 416 meV in the lithium rich phase and 536 meV in the vacancy rich phase. In path II the activated state is either the octahedral site (in the Li rich phase) or the crossing through the joint oxygen triangle between the tetrahedral and octahedral site (in the vacancy rich phase). For this path, the activation barriers are lower (about 351 meV in the lithium rich phase and 305 meV in the vacancy rich phase). It is interesting to note that the activation barrier is higher in the lithium rich phase than in the vacancy rich phase for path I, while the opposite is the case for path II. The electrostatic potential has been shown to play a critical role in determining the activation barrier [36]. In the vacancy rich phase less electrostatic repulsion from other Li+ is expected to lower the barrier. Meanwhile, the change of valence state of the nearest transition metal(s) will also influence the electrostatic repulsion [37]. In path II, one Ni atom changes from 2+ to 4+ upon delithitation, which will increase repulsion with the diffusing Li+. The results in Fig. 6-15(b) indicate that the effect of valence increase from Ni is smaller than that from Li+ interaction, and the total
121

barrier becomes lower upon delithiation. This is similar to what happens in LiFePO
4 upon delithiation [35] but opposite to what happens in LixCoO
2
[38]. In path I, however, three Ni atoms change from 2+ to 4+, which increases the electrostatic repulsion more than the effect of the Li vacancies. Hence, the total barrier along this path increases upon delithiation.
(a)
0.6-
0.5-
.
,.7
T I . . .
0.4-
C
.2
0.3-
.0
0.2-
,
0.1-
Path 1: 3N113 Mn.
Li rich phase
-. - Vacancy rich phase
0.0-
0. 0 0.2
I I I
.
I I ' I ' I ' I
0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
diffusion distance (A)
122

(b)
0.40
0.35-
0.30-
0.25-
.:
M
0.
C
0.20-
0.15-
0.10-
Path Ii: 1NII5Mn
0O
0.05-
0.00-
- -
- -
Li rich phase
Vacancy rich phase
-0.05
0.0 0.2
0.4 0.6 0.8 1.0 1.2 1.4
diffusion distance (A)
1.6 1.8
Figure 6-15 Calculated energy along the lithium migration paths in ordered
LiNio.
5
Mn
1
.0
4
. (a) path I defined by 3 Ni and 3 Mn around the active state; (b) path II defined by 1 Ni and 5 Mn around the active state.
The activation barriers in path II are significantly lower than those in path I both in the lithium rich phase and vacancy rich phase. Since path II percolates the structure, lithium does not need to migrate through the high barrier path I. Consequently, we use the activation energy along path II to evaluate the lithium diffusivity of this material. With a simple dilute diffusion model
D = a
2 v exp(-E,, / kBT), (1) where a is the lithium hopping distance, v is the attempt frequency and E,,, is the activation barrier. With hopping distance a = 3.46
in diffusion path II and v =10 12 Hz
(typical range of phonon frequencies), the diffusivities are calculated to be
123

1.41xI0-
9 cm 2 /s in the lithium rich phase and 8.25x10-9cm 2 /s in the vacancy rich phase
[35]. Using the simple estimate L 2 = Dt for the diffusion length, Li can migrate 3pm in
10 seconds. Our experimental results show that more than half of the theoretical capacity was obtained at discharge rate of 167C (corresponding to 22 seconds) with a particle size in the range of 3-5 pm . Hence, both experiment and calculation are in good agreement and confirm the high rate capability of this material.
Our results with high loading density and 15wt% carbon also indicate again that the intrinsic rate capability of active materials can only be measured with a highly optimized cell, as for fast materials, tested under high rate, the Li transport in the electrolyte and composite electrode is typically rate limiting [31]. Only when the active mass is sufficiently diluted and other kinetic limitations are resolved can one observe the intrinsic rate limits of the material.
Ordered LiNio.
5
Mn
1
504 spinel was prepared by solid state reaction. Though the particle size is several microns, this material shows high rate capability and excellent cyclability.
The high rate capability observed agrees with our first principles calculations, which show a low migration barrier for Li. Both experiment and calculation suggest that micron-sized ordered LiNio.
5
Mni.
5
0
4 can be a high rate electrode material with excellent density.
124

Reference:
11] J. Wolfenstine, J. Allen, J. Power Sources 142 (2005) 389-390.
[21 S. Sawa, S. Okada, A. Yoshino, J. Power Sources 97-98 (2001) 430-432.
[31 R. Ruffo, R.A. Huggins, C.M. Mari, M. Piana, W. Weppner, Ionics 11 (2005) 213-
219.
[41
Zhong, A. Bonakdarpour, M. Zhang, Y. Gao, J.R. Dahn, J. Electrochem. Soc
144 (1995) 205-213.
[51 Y. Talyosef, B. Markovsky, G. Salitra, D. Aurbach, S. Choi, J. Power Sources 146
(2005) 664-669.
[61 Y. Talyosef, B. Markovsky, R. Lavi, G. Salitra, D. Aurbach, D. Kovacheva, M.
Gorova, E. Zhecheva, R. Stoyanova, J. Electrochem. Soc. 154 (2007) A682.
[71 D. Aurbach, B. Markovsky, Y. Talyossef, G. Salitra, H. Kim, S. Choi, J. Power
Sources 162 (2006) 780-789.
[8] M. Kunduraci, G.G. Amatucci, J. Electrochem. Soc. 153 (2006) A1345.
[91 D. Pasero, N. Reeves, V. Pralong, A.R. West, J. Electrochem. Soc. 155 (2008)
A282.
[101 M. Kunduraci, G.G. Amatucci, J. Power Sources 165 (2007) 359-367.
[111 K. Ariyoshi, Y. Iwakoshi, N. Nakayama, T. Ohzuku, J. Electrochem. Soc. 151
(2004) A296.
[121 J.H. Kim, S.T. Myung, C.S. Yoon, S.G. Kang, Y.K. Sun, Chem. Mater. 16 (2004)
906.
[131 M. Kunduraci, J. Al-Sharab, G. Amatucci, Chem. Mater. 18 (2006) 3585-3592.
[141 K. Shaju, P. Bruce, Dalton Trans. (2008) 5471-5475.
[15] H. Fang, L. Li, G. Li, J. Power Sources 167 (2007) 223-227.
[16] J.C. Arrebola, A. Caballero, J. Morales, Electrochem. Solid-State Lett. 8 (2005)
A641.
125

[17] G. Kresse, J. Furthmuller, Comp. Mater. Sci. 6 (1996) 15-50.
[181 M. Cococcioni, C.A. Marianetti, F. Zhou, D. Morgan, G. Ceder, Phys. Rev. B 70
(2004) 1-8.
[19] N. Meethong, H.-Y.S. Huang, S.A. Speakman, W.C. Carter, Y.-M. Chiang, Adv.
Funct. Mater. 17 (2007) 1115-1123.
[201 H. Xia, Y.S. Meng, L. Lu, G. Ceder, J. Electrochem. Soc. 154 (2007) A737.
[211 S. Patoux, L. Daniel, C. Bourbon, H. Lignier, C. Pagano, F. Le Cras, S. Jouanneau,
S. Martinet, J. Power Sources 189 (2009) 344-352.
[221 T. Yi, X. Hu, J. Power Sources 167 (2007) 185-191.
[23] J. Liu, A. Manthiram, J. Electrochem. Soc. 156 (2009) A833.
1241 J. Liu, A. Manthiram, J. Electrochem. Soc. 156 (2009) A66.
1251 H.M. Wu, I. Belharouak, H. Deng, A. Abouimrane, Y.-K. Sun, K. Amine, J.
Electrochem. Soc. 156 (2009) A 1047.
[26] H.-B. Kang, S.-T. Myung, K. Amine, S.-M. Lee, Y.-K. Sun, J. Power Sources 195
(2010) 2023-2028.
[271 X. Zhang, J. Liu, H. Yu, G. Yang, J. Wang, Z. Yu, H. Xie, R. Wang, Electrochim.
Acta 55 (2010) 2414-2417.
[281 K. Ariyoshi, S. Yamamoto, T. Ohzuku, J. Power Sources 119-121 (2003) 959-963.
1291 T.E. Chin, U. Rhyner, Y. Koyama, S.R. Hall, Y.-M. Chiang, Electrochem. Solid-
State Lett. 9 (2006) A 134.
1301 M. Gaberscek, J. Moskon, B. Erjavec, R. Dominko, J. Jamnik, Electrochem. Solid-
State Lett. 11 (2008) A 170.
[311 P.A. Johns, M.R. Roberts, Y. Wakizaka, J.H. Sanders, J.R. Owen, Electrochem.
Commun. 11 (2009) 2089-2092.
[32] B. Kang, G. Ceder, Nature 458 (2009) 190-3.
[33] B. Ammundsen, J. Roziere, M.S. Islam, J. Phys. Chem. B 101 (1997) 8156-8163.
1341 M. Wakihara, Electrochemistry 73 (2005) 328-335.
126

D. Morgan, A. Van der Ven, G. Ceder, Electrochem. Solid-State Lett. 7 (2004)
A30.
[361 K. Kang, G. Ceder, Phys. Rev. B 74 (2006) 094105.
K. Kang, Y.S. Meng, J. Brdger, C.P. Grey, G. Ceder, Science 311 (2006) 977-80.
1381 A. Van der Ven, G. Ceder, J. Power Sources 97-98 (2001) 529-531.
127

The focus of this research is to understand and design positive electrode materials for lithium ion batteries by combining first principles calculations directly with experimental efforts. To understand the effect of the transition metals on the structural stability of the layered compounds upon Li deintercalation, I synthesized a new layered compound
LiNi
2
/
3
SbI/30
2
by means of ion exchange from NaNi
2
/
3
SbI/30
2
. The incorporation of a high valent Sb 5 increases the content of Ni
2
+in the compound, which improves the Li mobility and electronic conductivity especially when LiNi
2
/
3
SbI/30
2 is in a highly charged state. However, the structure of LiNi
2
/
3
SbI/30
2 degrades upon cycling. The structure degradation is found to be associated with the migration of Ni into the Li layer, which was determined by XRD pattern refinement and TEM diffraction on the cycled
LiNi
2
/
3
Sb/30
2 material. First principles calculation also shows a very low barrier for the migration of a divalent Ni from the transition metal layer to the tetrahedral sites of the lithium layer in the partially delithiated Li
2
/
3
Ni
2
/
3
Sbl/30
2
. The divalent Ni becomes highly mobile because of the strong electrostatic repulsion from the surrounding three Ni
3
+, and three Sb".
I also investigated the effect of the alkali ions on the structural stability in the layered
AMO
2 compounds by comparing layered LiMnO
2 and NaMnO
2
. Layered 03-LiMnO
2 can be obtained by means of ion exchange from 03-NaMnO
2
. However, the structure of layered LiMnO
2 transforms rapidly into a spinel-like structure upon delithation. While in
128

a Li-compound divalent Mn can easily migrate into the tetrahedral sites of the Li layer, as
Mn 2
in the tetrahedral sites is stabilized by a Li in the tetrahedral sites forming a Li/Ni dumbbell configuration, such a stabilizing dumbbell does not form in NaxMn0
2
Therefore, NaMnO
2 is expected to show a better structural stability upon deintercalation.
The electrochemical results of NaMnO
2 show a much better capacity retention than that of LiMnO
2
, confirming that NaMnO
2 is more stable than LiMnO
2 upon deintercalation.
The XRD results of the cycled NaMnO
2 also show no significant structural change. The pronounced voltage steps and plateaus of NaMnO
2 upon cycling were also investigated. I used ICP technique to study the possible Mn dissolution and found no significant Mn dissolution upon cycling.
To obtain high power and high volumetric energy density, materials with large particle but fast ionic and electronic conductivity are desired. I calculated the Li diffusivity in the
LiNio.
5
Mn
5
O
4 spinel by first principles calculations, and found that the Li diffusivity is in the order of 10~9 cm 2 /s, implying that LiNio.
5
Mn
.504 can be a high rate material even with a large particle size. To confirm the results of my calculations, ordered
LiNio.
5
Mn.
5
0
4 with particle size of 3-5 ptm was synthesized. The electrochemical tests of this micron-sized LiNio.
5
Mn
1 5
0
4 show higher rate capability than nano LiNio.5Mn
1 5
0
4 by
Shaju, et al, indicating that the ionic and electronic transport may not be the rate limiting factors. It was also found that cell configurations, such as separators, mechanical pressure of the cell and the carbon content in the electrode, could dramatically affect the rate capability of the cell. When the cell is highly optimized in configuration, more than half of the theoretical capacity is obtained at a discharge rate of 167C (corresponding to 22
129

seconds) with a particle size in the range of 3-5pm, which agrees with the high Li diffusivity by my calculation.
Integrating first principles methods with experimental efforts has been both effective and efficient in my research. Taking advantage of the merits of each tool greatly helped me understand the physics behind new phenomena, and identify ways to improve the properties of the materials. The iterative practice of computational and experimental tools presented ways to solve problems whenever either of the tools alone was insufficient.
Therefore, this kind of approach, integrating computational and experimental techniques, holds promise in developing advanced materials.
130