Lithium electronic environments in recha

Lithium electronic environments in rechargeable battery electrodes - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Lithium electronic environments in rechargeable battery electrodes
Citation
Hightower, Adrian
(2001)
Lithium electronic environments in rechargeable battery electrodes.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/tb89-6g55.
Abstract
This work investigates the electronic environments of lithium in the electrodes of rechargeable batteries. The use of electron energy-loss spectroscopy (EELS) in conjunction with transmission electron microscopy (TEM) is a novel approach, which when coupled with conventional electrochemical experiments, yield a thorough picture of
the electrode interior.
Relatively few EELS experiments have been preformed on lithium compounds owing to their reactivity. Experimental techniques were established to minimize sample contamination and control electron beam damage to studied compounds. Lithium hydroxide was found to be the most common product of beam damaged lithium alloys. Under an intense electron beam, halogen atoms desorbed by radiolysis in lithium halides. EELS spectra from a number of standard lithium compounds were obtained in order to identify the variety of spectra encountered in lithium rechargeable battery electrodes. Lithium alloys all displayed characteristically broad Li K-edge spectra, consistent with
transitions to continuum states. Transitions to bound states were observed in the Li K and oxygen K-edge spectra of lithium oxides. Lithium halides were distinguished by
their systematic chemical shift proportional to the anion electronegativity. Good agreement was found with measured lithium halide spectra and electron structure calculations using a selfconsistant multiscattering code.
The specific electrode environments of LiC_6, LiCoO_2, and Li-SnO were investigated. Contrary to published XPS predictions, lithium in intercalated graphite was determined to be in more metallic than ionic. We present the first experimental evidence of charge compensation by oxygen ions in deintercalated LiCoO_2. Mossbauer studies on cycled Li-SnO reveal severely defective structures on an atomic scale.
Metal hydride systems are presented in the appendices of this thesis. The mechanical alloying of immiscible Fe and Mg powders resulted in single-phase bcc alloys of less than 20 at% Mg. Kinetic studies on LaNi_(5-x)Sn_x alloys proved that the mass transfer of hydrogen through these alloys was not hindered with increasing Sn substitutions for Ni. Collaborations with Energizer© found LanNi_(4.7)Sn_(0.3) alloys to possess limited utility in rechargeable nickel-metal-hydride sealed-cell batteries.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
Materials Science
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Materials Science
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Fultz, Brent T.
Thesis Committee:
Unknown, Unknown
Defense Date:
14 July 2000
Record Number:
CaltechTHESIS:11192010-080258624
Persistent URL:
DOI:
10.7907/tb89-6g55
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
6183
Collection:
CaltechTHESIS
Deposited By:
Rita Suarez
Deposited On:
22 Nov 2010 23:03
Last Modified:
19 Apr 2021 22:36
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
8MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
LITHIUM ELECTRONIC ENVIRONMENTS IN
RECHARGEABLE BATTERY ELECTRODES

Thesis by
Adrian Hightower
In Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy

California Institute of Technology
Pasadena, California
2001

(Defended July 14, 2000)

Adrian Hightower

iii

Acknowledgments
This work was made possible by the generous support of a great number of
people. I would like to thank the following people for making this journey a rewarding
one.
• Dr. Brent Fultz, my thesis advisor, who has provided steadfast support and
guidance during my graduate career;
• Dr. Ratnakumar Bugga, who taught me all I know on batteries, fuel cells and
accomplishing tasks at the Jet Propulsion Laboratory;
• Dr. Channing Ahn and Carol Garland, who patiently shared their art of Electron
Microscopy;
• Dr. Peter Rez, for his invaluable expertise in Electron Energy Loss Spectroscopy
and unfailing sense of humor;
• Dr. Charles Witham, Peter Bogdanoff, Stephen Glade, Claudine Chen, Sven
Bossuyt, Yun Ye, Jason Graetz and my fellow graduate students who
work as hard as they play;
• Pam Albertson, Elena Escot, and the entire staff of the Caltech Material Science
Department;
• Eddie Grado, Frank Vargas, Cheryl Hawthorne, Sue Borrego, Michelle Medley
and the Caltech Office of Minority Student Affairs;
• My family: Mom, Dad, Erica, Stephanie, Jeffery, Sheila, Natalie, Sarah,
Jonathan, and my beloved Michelle.
This work has been supported by the U. S. Dept. of Energy grant DE-FG03-94ERl4493.

Abstract
This work investigates the electronic environments of lithium in the electrodes of
rechargeable batteries. The use of electron energy-loss spectroscopy (EELS) in
conjunction with transmission electron microscopy (TEM) is a novel approach, which
when coupled with conventional electrochemical experiments, yield a thorough picture of
the electrode interior.
Relatively few EELS experiments have been preformed on lithium compounds
owing to their reactivity. Experimental techniques were established to minimize sample
contamination and control electron beam damage to studied compounds. Lithium
hydroxide was found to be the most common product of beam damaged lithium alloys.
Under an intense electron beam, halogen atoms desorbed by radiolysis in lithium halides.
EELS spectra from a number of standard lithium compounds were obtained in order to
identify the variety of spectra encountered in lithium rechargeable battery electrodes.
Lithium alloys all displayed characteristically broad Li K-edge spectra, consistent with
transitions to continuum states. Transitions to bound states were observed in the Li K
and oxygen K-edge spectra of lithium oxides. Lithium halides were distinguished by
their systematic chemical shift proportional to the anion electronegativity. Good
agreement was found with measured lithium halide spectra and electron structure
calculations using a selfconsistant multiscattering code.
The specific electrode environments of LiC6, LiCo0 2 , and Li-SnO were
investigated. Contrary to published XPS predictions, lithium in intercalated graphite was
determined to be in more metallic than ionic. We present the first experimental evidence
of charge compensation by oxygen ions in deintercalated LiCo02• Mossbauer studies on
cycled Li-SnO reveal severely defective structures on an atomic scale.
Metal hydride systems are presented in the appendices of this thesis. The
mechanical alloying of immiscible Fe and Mg powders resulted in single-phase bee
alloys of less than 20 at% Mg. Kinetic studies on LaNis_xSnx alloys proved that the mass
transfer of hydrogen through these alloys was not hindered with increasing Sn
substitutions for Ni. Collaborations with Energizer© found LanNi4.7Sno.3 alloys to
possess limited utility in rechargeable nickel-metal-hydride sealed-cell batteries.

Table of Contents
Title page
Copyright page

Acknowledgments

III

Abstract

Table of Contents

List of Tables and Figures

Introduction

1.1 Components of a Battery

1.2 References

Techniques and Instrumentation

2.1 Battery Electrochemistry

2.1.1

Thermodynamics

2.1.2

Kinetics

2.1.3

Reversibility

2.2 Mossbauer Spectrometry

2.2.1

Recoil Free Fraction

2.2.2

Hyperfine Parameters

2.2.2.1 Isomer Shift, IS

2.2.2.2 Electric Quadrupole Hyperfine Interactions, QS

2.2.2.3 Magnetic Dipole Hyperfine Interactions

2.3 Electron Energy Loss Spectroscopy

2.3.1

Kinematics

2.3.2

Data Acquisition and Analysis

2.3.3

Energy Resolution

2.3.4

Energy Loss Spectrum

2.3.4.1 Zero-Loss Peak

2.3.4.2 Low-Loss Spectra

2.3.4.2.1

Plasmons

2.3.4.2.2

Single-Electron Excitations

2.3.4.2.3

Excitons

2.3.4.3 Core-Loss Spectra

XPS and EELS

2.3.5

2.4 References

Lithium EELS Spectra

3.1 Near Edge Structure

3.2 Beam Damage

3.2.1

Electron Specimen Interactions

3.2.2

Radiation Damage of Metallic Lithium

3.2.3

Radiation Damage of Lithium Flouride

3.3 Metallic Lithium and Alloys

3.4 Lithium Oxides

3.5 Lithium Halides

3.6 Electron Structure Calculations

3.7 References

Specific Electrode Environments
4.1 Electron Energy Loss Spectrometry on Lithiated Graphite

85
85

Vll

4.1.1

Introduction

4.1.2

Experimental

4.1.3

X-ray Diffraction Results

4.1.4

TEM Micrographs

4.1.5

Carbon K-edge

4.1.6

Lithium K-edge

4.1.7

Comparison to XPS Results

4.1.8

Discussion

4.1.9

Conclusion

4.1.10 References

4.2 EELS Analysis of LiCo02

4.2.1

Introduction

4.2.2

Experimental

100

4.2.3

Electrochemical Results

102

4.2.4

X-Ray Diffractometry

105

4.2.5

EELS Analysis

105

4.2.5.1 Li K-edge and Cobalt M 23 edge

107

4.2.5.2 Oxygen K-edge

107

4.2.5.3 Cobalt L23 Edges

112

Cycled Cathodes

112

4.2.6.1 TEM Micrographs of Cycled Cathodes

114

4.2.6.2 EELS Spectra of Cycled Cathodes

116

Discussion

121

4.2.6

4.2.7

V1l1

4.2.8

Conclusions

123

4.2.9

References

126

4.3 119Sn Mossbauer Spectrometry of the Li-SnO Anode

128

4.3.1

Introduction

128

4.3.2

Experimental

129

4.3.3

Electrochemical Results

133

4.3.4

X-Ray Diffractometry and TEM Results

133

4.3.5

Mossbauer Spectrometry Results

139

4.3.5.1 Mossbauer Spectra of Control Samples

139

4.3.5.2 Mossbauer Spectra of Anode Materials

143

4.3.5.3 Recoil-Free Fractions: Standards

145

4.3.5.4 Recoil-Free Fractions: Anode Materials

147

4.3.6

Oxidation in Ambient Air

150

4.3.7

Thermodynamics of Anode Reactions and Oxidation

153

4.3.8

Conclusions

158

4.3.9

Acknowledgments

160

4.3.10 References

Appendix A Mechanical Alloying of Fe and Mg

162

166

A.l Introduction

166

A.2 Experimental

167

A.3 Results

169

A.3.1 X-ray Diffractometry

169

A.3.2 Density

172

ix
A.3.3 Mossbauer Spectrometry

172

A.3A Rotating Magnetometry

177

A.4 Discussion

177

AA.l Chemical Distributions in bcc Alloys

177

AA.2 Limits of Solubility

183

A.5 Conclusion

185

A.6 References

187

Appendix B Kinetics of H diffusion in LaNis_xSnx Alloys

189

B.l Introduction

189

B.2 Experimental

191

B.3 Results

193

B.3.1

Steady State Measurements

193

B.3.!'1 Potentiodynamic Polarization

193

B.3.1.2 Discharge Characteristics

193

B.3.2 Transient Measurements

196

B.3.2.1 Chronoamperometry

196

B.3.2.2 Chronocoulometry

198

BA Temperature Studies

200

B.5 Conclusion

206

B. 7 References

208

Appendix C Performance of LaNi4,7SnO.3 Electrodes in Sealed Cells

210

C.I Introduction

210

C.2 Experimental

211

C.3 Results and Discussion

213

C.3.IIsothenns

213

C.3.2 Self Discharge

215

C.3.3 Rate Measurements

215

C.3.4 Pressure Measurements

217

C.3.5 Cycle Life

217

C.3.6 AC Impedance

223

C.4 Conclusions

223

C.5 References

225

List of Tables and Figures

Chapter 1
Figure 1.1

Diagram of a lithium ion cell.

Table 1.1

Anode materials

Table 1.2

Cathode materials

Table 1.3

Characteristics of organic solvents

Figure 2.1

Cathodic and anodic currents versus polarization, 11

Figure 2.2

Atom recoil, decay lineshape, lines separated by recoil

Figure 2.3

Elements that may be studied using Mossbauer spectroscopy

Figure 2.4

Mossbauer effect absorption and Doppler-derived energy

Figure 2.5

Mossbauer spectra with quadrupole and magnetic splitting

Figure 2.6

Nuclear quadrupole moment with nearby charges

Table 2.1

Relative Mossbauer peak intensities for 57 Fe

Figure 2.7

Kinematics of inelastically scattered electron

Figure 2.8

Bethe surface for K-shell ionization

Figure 2.9

Inner-shell ionization and associated nomenclature

Figure 2.10

Schematic of EELS spectrometer

Figure 2.11

Collection-angle geometry in image mode

Figure 2.12

Collection angle geometry in diffraction mode

Figure 2.13

Instrument response for EELS detector

Table 2.2

Characeristics of EELS processes

Chapter 2

xii
Figure 2.14

Aluminum volume plasmons

Figure 2.15

Electon energy levels and energy-loss spectrum

Figure 3.1

Knock-on displacement energy of atomic species

Figure 3.2

EELS spectra of beamed damaged metallic lithuim

Table 3.1

Thermodynamic constants for select Li compounds

Figure 3.3

Radiolytic displacement sequence for alkali halides

Figure 3.4

EELS spectra of beamed damaged lithium flouride

Figure 3.5

Micrograph of beamed damaged lithium flouride

Figure 3.6

Micrograph of beam damaged lithium tin

Figure 3.7

EELS low-loss spectra of lithium alloys

Figure 3.8

Lithium K-edges of lithium alloys

Figure 3.9

Aluminum L23 edge

Figure 3.10

EELS low-loss spectra of lithium oxides

Figure 3.11

Lithium K-edges of lithium oxides

Figure 3.12

Oxygen K-edges of lithium alloys

Figure 3.13

EELS low-loss spectra of lithium halides

Figure 3.14

Lithium K-edges of lithium halides

Figure 3.15

Chemical shift versus anion electronegativity

Figure 3.16

Simulated lithium K-edges for LiF

Figure 3.16

Simulated lithium K-edges for LiCl

Figure 3.17

Simulated lithium K-edges for LiBr

Chapter 3

xiii

Chapter 4
Figure 4.1

XRD pattern of KS44 graphite and lithiated graphite

Figure 4.2

Micrograph of lithiated graphite

Figure 4.3

Carbon K-edge for KS44 graphite and lithiated graphite

Figure 4.4

Lithium K-edge of metallic Li, LiF, and lithiated graphite

Figure 4.5

Initial charge-discharge curves of lithiated KS44 graphite

Figure 4.6

Layered ABCABC stacking of LiCo0 2

Figure 4.7

Initial cycles of LiCo0 2cell

103

Figure 4.8

Electrochemical discharge curves of LiCo02

104

Figure 4.9

XRD patterns of discharged LiCo0 2

106

Figure 4.10

Co M23 and Li K edge of cobalt oxide standards

108

Figure 4.11

Oxygen K-edge of cobalt oxide standards

109

Table 4.1

Occupation numbers of CoO

III

Table 4.2

Occupation numbers of LiCo02

III

Figure 4.12

Cobalt L23 edges of cobalt oxide standards

113

Figure 4.13

TEM micrographs of discharged cathodes

115

Figure 4.14

Li K-edge and cobalt M 23 edges of discharged cathodes

117

Figure 4.15

Oxygen K-edges of discharged cathodes

118

Figure 4.16

Cobalt L23 edges of discharged cathodes

119

Figure 4.17

Simulated oxygen K-edge in LiCo0 2

122

Table 4.3

Normalized areas of oxygen K-edge peaks

125

Table 4.4

Normalized areas of cobalt L-edge peaks

125

XIV

Figure 4.18

Electrochemical discharge curves for lithium-tin anode

134

Figure 4.19

Cycle life of lithium-tin anode

135

Figure 4.20

XRD pattern of lithium-tin alloy and anodes

137

Figure 4.21

TEM micrograph of cycled lithium-tin anode

138

Figure 4.22

Mossbauer spectra of tin standards and cycled lithium-tin anodes

140

Figure 4.23

XRD patterns of oxidized Liz.3SnO anode

151

Figure 4.24

XRD patterns of oxidized LisSnO anode

152

Figure 4.25

Deconvoluted Mossbauer spectra of oxidized Lh.3SnO anode

154

Figure 4.26

Deconvoluted Mossbauer spectra of oxidized LisSnO anode

155

Figure 4.27

Deconvoluted Mossbauer spectra of oxidized Li 22 Sns

156

Table 4.5

Relative spectral areas in LisSnO

161

Table 4.6

Relative spectral areas in Lh.3SnO

161

Figure Al

XRD patterns of Fe-Mg alloys

170

Figure A.2

XRD pattern of annealed Fe8sMglS

l71

Figure A.3

Lattice parameters of hcp and bcc phases

173

Figure A4

Fractions of hcp and bcc phases from XRD

174

Figure A5

Density measurements of as-milled alloys

175

Figure A6

Mossbauer spectra of as-milled Fe-Mg alloys

176

Figure A7

Hyperfine magnetic field distributions

178

Figure A8

Calculated nearest neighbor ratio

l79

Figure A.9

Saturation magnetization for Fe-Mg alloys

180

Appendix A

xv
Appendix B
Figure B.l

Potentiodynamic polarization curves for LaNis_xSnx alloys

194

Figure B-2

Diffusion limiting currents of LaNis_xSnx alloys

195

Figure B.3

Discharge curves of LaNis_xSnx alloys

197

Figure B.4

Analysis of chronoamperometric curves of LaNis_xSnx alloys

199

Figure B.5

Analysis of chronocouometric response of LaNis_xSn x alloys

201

Figure B.6

Variation of hydrogen duffusion coefficient

202

Figure B.7

Chronoamperometric response of LaNi4.8Sno.2 at various T

203

Figure B.S

Chronocoulmetric response of LaNi4 .8Sno.2 at various T

204

Figure B.9

Arrhenius plot of LaNi 4.8 Sno.2 diffusion coefficients

205

Table B.l

Determination of Diffusion coefficients

207

Figure C.l

Gas phase desorption isotherms of anode materials

214

Figure C.2

Self discharge curves at 23 DC and 45 DC

216

Figure C.3

Electrochemical capacities of AA cells

21S

Figure C.4

Fractions of H2 gas evolved during charging

219

Figure C.5

Cycle life of sealed cell measured at JPL

220

Figure C.6

Cycle life of sealed cell measured at Energizer

221

Figure C.7

Cell internal resistance vs. cycle number

222

Appendix C

Chapter 1

Introduction

1.1 Components of a Battery

The fundamental components of an electrochemical cell are the anode, cathode,
electrolyte, and external load to complete the circuit. The anode is the electrode where
molecules or ions from the electrolyte are oxidized upon spontaneous reaction
(discharge). The electrons from the oxidized species of the anode move as current
through the wires of the external load. Upon discharge, ions or molecules are reduced at
the cathode by reacting with electrons from the external load. The electrolyte provides
the media for ionic current with positive ions moving toward the cathode and negative
ions moving toward the anode upon discharge (Figure 1.1). For low currents, the cell
voltage is the difference in electrochemical potential between the cathode and the anode
[1]. For higher currents, a voltage drop across the cell may also be important [2].

Anodes of rechargeable lithium battery systems are generally composites of a
lithium alloy, a conducting diluent (carbon black), and a polymer binder
(polyvinylidenefluoride). Anode materials are chosen for their high chemical potential
for lithium (low voltage) relative to the cathode environment. These materials, including
metallic lithium, are extremely reactive owing to their high chemical potential of lithium.
Lithiated anode materials react quickly with moisture from the air to form lithium
hydroxide and lithium oxides. Table 1 lists some common lithium anode materials.
For rechargeable batteries, critical characteristics of electrode materials are capacity,
chemical potential, cycle life, and kinetics. As a general rule, electrodes materials with
larger capacities have smaller cycle life and vise versa. The design of

SCHEMATIC DIAGRAM OF A Li-ION CELL
CARBON ANODE

OXIDE CATHODE

3.8 V (3.65 V under load) - - - - o M

Figure 1.1. Schematic diagram of a lithium-ion cell.

secondary lithium ion batteries has been the history of optimizing and compromising
these characteristics [3]. Good kinetics implies efficient charge transfer and mass
transfer reactions at and through the various interfaces in the cell and in the material.
Cycle life is the number of times an electrode can be charged and discharged before a
substantial loss of capacity occurs. Cycle life is generally dependent on the structural and
chemical stability of the electrode.
Table 1.1 Anode Materials
Material

Li metal
LiAI
Li-tin oxide amorphous
LiC 6 (doped coke or graphite)
Lio.sC6 (coke)
LiW0 2
LiMo0 2
LiTiS 2

Voltage range
vs. lithium,V
0.0
0.3
0.1-1.2
0.0-0.5
0.0-1.3
0.3-1.4
0.8-1.4
1.5-2.7

Theoretical
specific capacity,
Ah/g
3.86
0.8
0.65
0.370
0.185
0.12
0.199
0.226

Although metallic lithium has the best energy density and voltage, its practical
use as an anode material is limited by its dangerously fast kinetics, which lead to
heterogeneous plating of lithium on the anode during charging. This leads to the
formation of dendrites, which eventually short the battery, possibly leading to dangerous
run away reactions. Except for the smallest cells, the formation of dendrites inhibits the
cycle life of metallic lithium anodes to tens of cycles.
Lithiated graphite (LiC6) is currently the most popular anode material in lithium
ion systems. This work details how lithium ions intercalate between graphite planes
where they are reduced to neutral atoms. The preferential intercalation of lithium
between the graphite planes allows for rapid mass transfer through the graphite phase.

The chemical potential of lithium in graphite is very similar to that of lithium in its
metallic form owing to limited charge transfer from Li to the surrounding carbon. Strong
covalent bonding between the graphite layers provides a stable lattice to promote a cycle
life on the order of thousands of cycles.
The cathode of most lithium ion batteries is a composite of a lithium transition
metal oxide (Table 1.2), a conducting diluent, and a polymer binder. Lithium cobalt
oxide is the most widely used cathode material in lithium ion systems. LiCo02 has an
ordered rock-salt structure with lithium and cobalt planes alternating between closepacked oxygen layers. As with graphite, the layered structured of LiCo0 2 allows for
rapid diffusion in the lithium plane. A model of strict charge transfer views LiCo0 2 as
being composed of Li+, Co+3, and 0- 2 ions. Conventional wisdom views the lithium
deintercalation accompanied with a change in cobalt valence from Co+3 to Co+4 . Thus
oxygen atoms remain unchanged as 0 2- ions. This work refutes this model and suggests
significant charge compensation by oxygen atoms with lithium deintercalation. The high
cost of cobalt has stimulated interest in alternative cathode materials, most notably
manganese oxides.

Table 1.2 Cathode Materials
Material

LiCo02
LiMn20 4
Li 2Mn 2 0 4
LiV0 2
LiNi0 2

Voltage range
vs. lithium,V

3.73-4.0
3.9-4.1
2.6-2.7
2.9-3.0
3.5-3.7

Theoretical
specific capacity,
Ah/g
0.5
0.15
0.75
0.226
4.5

Electrolytes of rechargeable lithium systems are combinations of a lithium salt
and an organic solvent with less than 50 ppm of water. An electrolyte needs to be highly
conductive over a large temperature range as well as electrochemically stable over a wide
range in voltage (0 to 4.1 V vs LilLi+). Typical salts in lithium ion systems include
LiAsF6, LiBF6, LiCI0 4 and LiPF6, chosen for economic and environmental reasons. The
organic molecules used in electrolytes decompose on the surface of an electrode forming
a solid-electrolyte interphase (SEI) [4,5]. For lithium ion systems, the SEI is composed
of reduced electrolyte products and reduced electrode material. Common lithium
compounds of the SEI are lithium carbonate (LbC0 3), lithuim oxide (LbO), lithium
hydroxide (LiOH), lithium alkoxides, lithium flouride (LiF), and electrolyte salt
reduction products yet to be characterized [1]. The SEI can have the beneficial role of
protecting the electrode from decomposing and stable SEI's do not consume further
lithium, minimizing irreversible capacity.
Blends of carbonates are currently used as electrolyte solvents (Table 1.3). These
blends optimize the benefits and offset the deficiencies of each carbonate. For example,
the rapid kinetics of propylene carbonate (PC) are hampered by its destructive effects on
anode surfaces. Ethlylene forms a passivating SEI on the electrode, thus enhancing the
battery's cycle life. Blending EC and PC in a 1: 1 ratio has proved a very successful
electrolyte. Much work has been done to formulate electrolyte blends to lower the
operating temperature of lithium ion batteries to less than O°C [6].

Table 1.3 Characteristics of Organic Solvents *.

yBL

THF

CHz-CH z

CH z-CH 2

CH 2-O-CH,

CH z C=O

CH z CH z

CHz-O-CH,

Characteristic
Structural
formula

"- /

1,2-DME

DMC

DEC

DEE

Dioxolane

CH 2-O-CzH s

CHz-O-CzH s

"-0

CH 2-CH

"-CH,

Boiling
temperature, °C
Melting
temperature, °C
Density, g/cm'
Solution
conductivi ty,
S/cm
Viscosity at 25°C,
cP
Dielectric constant
at 20°C
Molecular weight
H 2 0 content, ppm
Electrolytic
conductivity at
20°C,
1M LiAsF~.
mS/cm

CH,

CH z-CH 2

"-0

CH,

CH 2

CH z

CH,

CHz-CH z

202-204

65-67

240

248

126

121

-43

-109

-58

-49

-39-40

4.6

-43

-74

-95

1.13
I.IxIO- H

0.887
2.1 X 10- 7

0.866
3.2 X lO- M

1.198
2.1XlO-~

1.322
<10- 7

1.071
<10- 7

0.98
<10- 7

0.842
<10- 7

1.060
<10- 7

1.75

0.48

0.455

2.5

0.59

0.75

0.65

0.58

7.75

7.20

64.4

3.12

2.82

5.1

6.79

86.09
<10
10.62

72.10
<10
12.87

90.12
<10
19.40

102.0
<10
5.28

1.86
(at 40°C)
89.6
(at 40°C)
88.1
<10
6.97

90.08
<10
11.00
(1.9 mol)

118.13
<10
5.00
(l.S mol)

118.18
<10
-10.00t

74.1
<10
-11.20t

• yBL = y·butyrolactone; THF = tetrahydrofuran; l,2.DME = l,2·dimethoxyethane; PC = propylene carbonate; EC = ethylene carbonate; DMC = dimethyl carbonate; DEC = diethyl
carbonate; DEE = diethoxycthane.
t Estimation based on Walden', rule.

1.2 References
[1] J.O.M Bockris and S. U. M. Khan, Suiface Electrochemistry: A Molecular Level

Approach, (Penum Press, New York, 1993).
[2] A. J. Bard and L. R. Faulkner, Electrochemical Methods: Fundamentals and

Applications, (John Wiley & Sons, New York 1980).
[3] D Linden, Handbook of Batteries, (McGraw-Hill. New York, 1976).
[4] D. T. Sawyer, A. Sobkowiak, J. L. Roberts Jr, Electrochemistryfor Chemists, (John
Wiley & Sons, New York, 1995).
[5] G. Prentice, Electrochemical Engineering Principles, Prentice-Hall, New Jersey,
1991).
[6] M. C. Smart, B. V. Ratnakumar* and S. Surampudi, Y. Wang, X. Zhang and S. G.
Greenbaum, A. Hightower, C. C. Ahn and B. Fultz, J. Electrochem. Soc. 146, 11 (1999).

Chapter 2

Techniques and Instrumentation

2.1 Battery Electrochemistry
Batteries, fuels cells, and corrosion process all convert the energy of chemical
reactions into electrical energy and thus adhere to the same laws of kinetics and
thermodynamics. The interconversion of chemical and electrical energy is accomplished
by an ionic current flow through an electrolyte. This electrolyte must lie between two
chemical potentials in electrical contact with each other.

2.1.1 Thermodynamics of a Battery
A battery can be thought of as an engine that derives work from a gradient in chemical
potential. From the First law of thermodynamics we can derive the relationship between
the work done by the battery and the Gibbs free energy of a reaction.
I::.U=q-w

(2.1)

Here L1U is the change in Internal Energy, q is the heat added to system, and W is the
work done by the system. We can further define the heat added as q= TL1S and the work
done by the battery as the sum of the mechanical work, wp = pI::.V and electrochemical
work We, W=Wp + We. Thus we have the internal energy defined as
1:1 U = TI:1S - pl:1V - We

(2.2)

Comparing this to the change in Gibbs Free energy,
!::.G =!::.U -TI::.S+ p!::.V

we find that the change in Gibbs free energy is equal to the negative of the
electrochemical work.

(2.3)

I1G =-We = -nFM

(2.4)

The work done by the battery upon spontaneous discharge acts to lower the Free
energy of the system by a value proportional to the number of electrons transferred, 11, the
Faraday constant, F, and the difference in electrochemical potential across cell, t1E.
Conversely, electrochemical work must be done on a battery to charge it. For practical
applications, this fundamental analysis must be expanded upon and applied to the various
reacting species within the battery. We now build upon these relations and expound on
the kinetics of electrochemical cells.

2.1.2 Kinetics
The kinetics of electrodes are best understood in the relationship between the flow
of current and driving electrochemical potential. We start with an electrode reaction
involving a number, 11, of electrons, e, an oxidized species, 0, and its reduced state, R.
kc

O+ne

(2.5)

For a galvanic cell, electrochemical convention defines the electrode at higher
electrochemical potential (lower free energy) as the cathode, while the one at lower
electrochemical potential (higher free energy) is the anode. Cathodic current results from
the reduction of electrode species. Anodic currents oxidize electrode species. Both
cathodic and anodic reactions occur simultaneously with electron-transfer rate constants
kc and ka defined by Arrhenius relations.
_ /',v c

k c =Ace

(2.6)

(2.7)

10
These rate constants have units of (secr! while the standard free energies of activation,
L1Ga and L1Gn are the sum of a charge-dependent and charge-independent components.

We assume Faraday's law governs the processes, i.e., the amount of
electrochemical reaction is proportional to the number of electrons transferred. The
charge dependent component of L1Ga and L1Ge , can be expanded in terms of a potential
difference across the electrode-solution interface, M, and Faradays constant, F. Only a
fraction of M is effective in accelerating the reaction rate and is represented by the
symmetry parameter a, 0< a < 1.
(2.8)

(2.9)
Free energies of activation can be expanded into their Arrhenius relations.
Building charge-independent terms and rate constants, A e, and A a, into keo and kao yields
the following relations.

k=ke
co

(2.10)

(I-a )l1FI1E

k a =kao e

(2.11 )

The rate of electron transfer at the electrode interface is a product of these electron
transfer rate constants and the concentration of the reacting species, Co and CR. The
amount of reacting species depends on mass transfer mechanisms at the electrodeelectrolyte interface. Diffusion, convection (stirring), and migration (movement of ions
in an electric field) are the primary mass transport mechanisms. Thus we can define the
net current, jlle(, at an electrode as a function of the above electron transfer rate constants,
Co, CR, F, and 11.

(2.12)
(2.13)

iller = nF( Co kco e

(l-a)IIF~E

CRkao

(2.14)

Of particular interest is the condition when the net current flow across the
electrode-solution interface is zero, CO=CR, ka=kc, and l1E =l1Eeq . Under these conditions
of equilibrium potential difference l1Eeq , the current densities represent the equilibrium
exchange current density, io. The activation overpotential, 1}, is the difference between
l1E and l1Eeq .
(2.15)
(mFM,

(l-a)I1FM,>

(2.16)

Finally, substitutingio and 1} into Equation 2.12 yields the Butler-Volmer equation.
_ anF?

J.net = J.0 (e

(i-a)nF1)

(2.17)

Note that Equation 2.17 assumes equal concentration of the oxidizing species and the
reduced species (Figure 2.1). More thorough forms of the Butler-Volmer equation allow
for a concentration dependent term [1].

2.1.3 Reversibility
From the Butler-Volmer relation we see that the reversibility of a reaction is only
dependent on the free energies of activation. The exchange current is a practical measure
of the reversibility of the process. Large exchange currents for a given reaction imply
facile activation energies in both directions. Primary batteries have other processes that
occur at lower activation energy than the reverse reaction. This is true for the zinc (Zn)

+11

-1}

Figure 2.1. Cathodic and anodic components of current as a function of electrode
polarization,11 [1]

13
anode of the most popular primary battery design. When the primary cell is discharged,
an anodic current flows across the Zn electrode.
'+

Zn~Zn-

+2e

(2.18)

If one attempts to recharge the anode by applying an opposing voltage larger than the cell

voltage, a cathodic current is observed in the production of hydrogen gas, H 2 •
(2.19)

Standard sealed primary cells are not designed to withstand the pressure from this gas
build up. Extreme cases result in cell case rupture and subsequent ignition of the evolved

2.2 Mossbauer Spectrometry
Mossbauer Spectroscopy involves the recoilless resonant absorbtion and emission of
gamma rays to determine the chemical state of atoms in a lattice. It is named for Rudolf L.
Mossbauer who was awarded the 1961 Nobel Prize in Physics for his Ph.D. thesis work.
Aspects of the Mossbauer effect are described in Figure 2.2. The energy of a gamma ray
photon, Eo, is the difference in energy between a nuclear excited state, Ee and the ground state,

Eg [3]. The atom can recoil with energy, ER during this emission process. Subsequently the
energy of the emitted gamma ray is Ey =Eo - ER (Figure 2.2a). The energy linewidth, r of this
process is proportional to bandwidth of excited states, ~ (Figure 2.2b).
The resonant process Mossbauer Spectroscopy implies that the emitted and absorbed

photon must have equivalent energies precise to r = 10-6 - 1O- eV. Such energy overlap only
occurs if individual atoms are not significantly displaced from their lattice sites by emission or
absorption of gamma rays. These recoilless processes are only possible for solids in which
the entire matrix recoils as a unit. Since the mass of the entire matrix exceeds that of atoms
by several orders of magnitude, the recoil energy, ER is much less than r. Figure 2.2c
describes the separation of Mossbauer lines by nucleus recoil. The stringent requir ement of
the resonant recoilless process limits the types of materials that can be studied. Figure 2.3
shows elements that may be studied with Mossbauer Spectroscopy [4].
Mossbauer spectra are typically generated by varying the energy of the gamma
ray source. The source is mounted onto a precision velocity transducer that induces an
appropriate Doppler shift. When the energy of source gamma rays equal nuclear resonant
energies, absoption of the gamma ray will result. The transducer typically scans a small

R - 2Mc 2

(recoil energy)

Nucleus of mass M
end mean energy Eo
(at rest before 'V - ray
emisSion)

Ey.Eo-ER
(" -ray energy)

(a)

I.r2 - - - - - -

(b)

(c)

Figure 2.2. Recoil of the nucleus during emission. (b) The Mossbauer line shape from
the decay of the nuclear excited state. (c) Separation of the Mossbauer lines by the recoil
of the nucleus. The Mossbauer effect is only possible in the absence of this recoil [3J.

Elements that Can Be Studied with Mossbauer Spectroscopy

Sc Ti

~ EASY TO STUDY. EXTENSIVE RESEARCH

'i% MORE DIFFICULT TO STUDY. SOME RESEARCH
::r{ VERY DIFFICULT OR LIMITED RESULTS

Figure 2.3. Elements that may be studied by Mossbauer spectroscopy [4].

Source 5

Absorber A

Detector

V :

Maximum
overlap
• E

Eo(S}
Eo(A}

JIm
DJ\

I(E}

V) 0:

Partial
overlap
• E

Eo(A} Eo(S}

Eo(S) Eo(A)

v (0:
Partial
overlap

• E
v»O

V" 0

No overlap
No resonance
Eo(S}

Eo(A}

Eo(S}
T (%)

c::

y ,",

~~--------,-~~--

R~ance

aDsorption line

"Moss bauer spectrum·

-y

Doppler velocity

Figure 2.4. Mossbauer effect absorption as a function of the Doppler-derived energy
(ER=O is assumed) [3].

18
velocity range on the order of millimeters per second. Figure 2.4 illustrates the
experiment.
The chemical, crystallographic, and magnetic environment of the absorber influence its
nuclear states and thus its Mossbauer spectra. The overall intensity (recoil-free fraction,J)
and the line energies (hyperfine parameters) are the two most common pieces of information
derived from the Mossbauer spectra.

2.2.1 Recoil Free Fraction,!
The percentage of emissions with no photon response is characterized by the
recoil free fraction, f This factor determines the overall intensity of the resonance
spectrum. The recoil free fraction can be predicted using the Debye model [3]

f =e

-2M

(2.20)

with

(2.21)

Here, T is temperature, kb is Boltzman's constant, 8D is the Debye temperature, and ER is the
recoil energy defined in Fig. 2.1. This temperature-dependent form off, also known as the
Lamb-Mossbauer factor, is comparable to the Debye-Waller factor in X-ray scattering. The
main difference being that lattice vibrations are short relative to an X-ray scattering process
but long compared to the lifetime of the Mossbauer excited state. The Lamb-Mossbauer
factor reduces to:
(2.22)

and

(2.23)

The recoil free fraction is influenced by same factors that determine the vibrational
response of a lattice. These include E R , T, and stiffness of the atom in the lattice, represented
by eD . The recoil free fraction is higher at lower temperatures and with less energetic
photons.

2.2.2 Hyperfine Parameters
Electronic effects on the nucleus determine the absorption peak energies of Mossbauer
spectra. These effects are isomer shifts (IS), electric quadrupole hyperfine interactions (QS),
and magnetic dipole hyperfine interactions (MD). A summary of these interactions is
displayed in Figure 2.S.

2.2.2.1 Isomer Shift, IS
Isomer shifts are caused by interactions of nuclear charge with electron density inside
the nucleus. Electrons must have finite probability of being inside the nucleus (s-electrons) to
directly influence the isomer shift, though other electrons (p-, d-, and f-) indirectly influence
these shifts. This electronic interaction with the nucleus can be interpreted to give
unequivocal information about the valence of the absorbing atom. The isomer shift is:
(2.24)

with

=the radii of assumed spherical atomic nuclei,

/3i+ 2
3/Z

____
' _-3/2I=----=--~2
~~~~'<-t-3/
+1

"'::--

_ /2

1/2

Eledrostaric
interactions only

------~l ~
(a)

__I d ' k I

1/2

(b)

--1

Jk-

11
-liz

- 3/2

=:---- -!:r~ _liz

Quadrupole
splithng

--------

__.x_--xttx- + 1/2

Magnehc + quadrupole

(c)

Figure 2.5. Changes in nuclear energy caused by quadrupole and magnetic splitting with the resulting Mossbauer spectra [5].

1'11(0)1 2 = the probability density of the s-electron at the nucleus
~R

=Re-Rg

(e = excited, g= ground)

A and S subscripts -> absorber and source

2.2.2.2 Electric Quadrupole Hyperfine Interactions, QS

A nucleus with a spin greater than 112 will have an electric quadrupole moment that
will interact with an electric field gradient. An electric field will split the degenerate energy
levels of a nucleus with spin 3/2 level into two distinct degererate levels ±312 and ±112. This
allows for two absorption levels and thus a splitting of peaks in Mossbauer spectra. The
energy linewidth, QS, of this splitting is:

(2.25)

with

(2.26)
Here e is electron charge, Q is the nuclear quadrupole moment, Vii are the electric field
gradients along the given direction (i = x, y, z,), and c is an asymmetry parameter, which
describes the difference in electric field gradient in the x and y direction. Non-uniform
electron density within an atom can cause these field gradients. Thus quadrupole splitting is
greatly influenced by the bonding of the absorbing atom. An oversimplified model of this
effect is shown in Figure 2.6.

fz "

312

Ir ,,-312

1 LOW-ENERGY
CONFIGURATION

LIGAND AXIS
(r AXIS)

C~)

fz "

112

lr" -1/2

1 HIGH-ENERGY
CONFIGURATION

Figure 2.6. Coupling of the nuclear quadrupole moment with nearby charges. The
oversimplified drawing depicts the Iz=3/2 and Iz=1I2 sins exactly along and perpendicular
to the z-axis; the correct mechanical solution is more complex [4].

2.2.2.3 Magnetic Dipole Hyperfine Interactions
Magnetic dipole hyperfine interactions, also known as the nuclear Zeeman effect, arise
from interactions of the nuclear magnetic moment with a magnetic field at the nucleus. The
magnetic momentof a nucleus is related to its angular momentum. The angular momentum or
nuclear spin I, is the sum of the orbital, L and spin, S angular momentum, I

=L + S.The

magnetic field splits degenerate energy levels into 21+ 1 equally spaced, non-degenerate
energy levels. Splitting of the excited state, le=3/2, and ground state, Ig=1I2, creates eight
possible energy transitions. From these eight transitions, only six are allowed by selection
rules: ~1=1; Mn=O, ±l. The probabilities of these six transitions and thus their corresponding
peak intensities can be calculated by squaring the corresponding Clebsch-Gordon coefficients
(Table 2.1 [6]).
Table 2.1 Relative Mossbauer peak intensities for 57Fe assuming no thickness effects.
Normalized Intensities
Absorption Peaks

3 and 4
1 and 6
2 and 5

varies from to 4 depending on the moment
projection on the magnetic field; isotropic average
of2.

An absorbing atom influenced by the nuclear Zeeman effect will yield a distinct sextet
of peaks dependent on the magnetic field at its nucleus (hyperfine magnetic field, HMF).
When there are multiple magnetic environments for an absorbing element, the Mossbauer
spectra is a sum of superimposed sextet peaks. The values of the hyperfine magnetic fields,
HMG, can be extracted from spectra using the method of La Caer and Dubois [7].
The HMF is generated by 1) the orbital angular momentum of 4f and 5f electrons, 2)
densities of spin-up and spin-down electrons at the nucleus (polarization of local s-electrons),

3) externally applied fields, or 4) any combination of the above. In the case of iron, the
magnetic polarization model [8] explains the chemical origins of perturbations in the HMF.
The HMF perturbation, ~H, is the sum of perturbations caused by local environments, Id, and
those caused by nonlocal environments, H NL . With 57 Fe, HL arises from the polarization of selectrons by unpaired 3d-electrons at the same atom. HL is directly proportional to the change
of the local magnetic moment at the 57Fe atom. Neighboring atoms can affect the bonding
nature of the 57Fe 3d-electrons and thus influence lk.
The second component, H NL, arises from polarization of nonlocal 4s-electrons at the
57 Fe nucleus.

These perturbations are due to changes in the magnetic moments of neighboring

lattice sites. HN is dependent on whether neighboring sites are occupied by solute atoms or by
other 57Fe atoms whose moments may be perturbed by nearby solute atoms.

2.3 Electron Energy Loss Spectrometry
Electron Energy Loss Spectroscopy (EELS) measures the energy distribution of
electrons inelastically scattered by a specimen. A typical EELS experiment utilizes a
transmission electron microscope (TEM) to direct an electron beam through a magnetic
prism. The magnetic prism spatially separates electrons by energy. The energy
distribution of the scattered electrons gives valuable information on the valence, bonding,
and coordination of elements of the specimen. EELS is not affected by the fluorescence
yield limitation that restricts light elements X-ray analysis. The sensitivity of EELS to
light elements makes it a preferred technique for analysis of Li chemical environments.

2.3.1 Kinematics
The interaction of a fast moving electron with an atom is illustrated in Figure 2.7.
The difference between the initial and final wave vectors, ko and kl respectively, [9] is
given by
(2.27)
We assume the scattering is symmetric in the azimuthal angle to describe the scattering
process in terms of the scalar wave numbers. Conservation of energy reveals the
relationship between the energy loss of a scattering event and the scalar wave numbers of
the electron.
(2.28)
which simplifies to

...

- --

------..,,-

Figure 2.7. Kinematics of an electron inelastically scattered by an atom.

(2.29)
where y = (1_v2/c 2 rIl2 , mo is the rest mass of an electron, ao is the Bohr radius, and R=
13.6 eV is the Ryberg energy. Equation 2.29 is independent of the scattering angle e,
implying that all scattered wave vectors of the same magnitude correspond to inelastic
events of the same energy.
The law of cosines is applied to the vector triangle defined by the conservation of
momentum (Figure 2.7).
(2.30)
Substituting Equations 2.3.4 into 2.3.5 gives

(2.31)

We are now interested in the case for e =0, corresponding to q=qmill=ko-k j • From the
binomial expansion of the square root term, we find terms up to the second order in E
cancel, leaving
(2.32)
For nearly all collisions, y-3 E1Eo ~ 1 holds true and thus only the E2 terms are important.
We can then express qrnin as
(2.33)
8E=EI(2yEo) is the characteristic or most probable scattering semiangle. 8E can be

thought of as the classical scattering angle in "billiard-ball" collisions.

28
The excitations of core electrons show distinct profiles that give insight to the
coordination and valence state of the excited electrons. The intensity of inner-shell
transitions, ICE,~), can be expanded in tenns of N, the number of atoms per unit area in
the sample illuminated by the incident electron beam, It(~), the total number of electrons
collected by the spectrometer, and do/dE the energy differential cross section per atom.

dcr
I(E,[3) = NI,C[3)-(E, [3)
dE

(2.34)

The energy differential cross section is obtained by integrating the double
differential cross section dd/dEdq obtained from Fenni's Golden Rule of scattering
theory. Under the first Born approximation, which assumes only single scattering events
within each atom, da/dE is given by

(2.35)

where r is the position vector of the incident electron. The matrix element is evaluated
over the coordinates of all atomic electrons and summed over all degenerate energy
states. Lastly, 'VI is a continuum wave functions such that the square of the matrix has
units of inverse energy. The continuum wave function is nonnalized by the energy, EO. 25 ,
which gives an inverse square root dependence. This means. The energy differential
cross section can further be expressed in tenns of df/dE, the general oscillator strength
(GOS).

(2.36)
with

(2.37)
The GOS for the excitation of a Is electron (K-shell ionization) can be expressed
on a two-dimensional plot known as a Bethe surface (Figure 2.8). This plot clearly
demonstrates that maximum intensity occurs just above the inner-shell binding energy
EK, and at e=o, q=qmil1=ko eE . This regime corresponds to the dipole region of scattering

and, in terms of a particle model, represents "soft" collisions with relatively large impact
parameters. At a large energy loss, the intensity is concentrated in a region known as the
Bethe Ridge where q satisfies
(2.38)

The equivalent scattering angle, eC =(2eE )112, is known as the cut-off angle, above which
scattering intensity falls to zero. These "hard" collisions have small impact parameters
and primarily involve the electrostatic field of a single inner-shell.
Evaluation of Equation 2.36 yields an approximate behavior of
da
o c EdE

(2.39)

Here, s is constant over a limited energy range. The value for s tends to decrease as B, E
and the thickness of the sample increase. For inner-shell scattering, s is 4.5 for small
angles and 2.5 for large angles.

d fK
dE

[eV-1x1Q-3 j

In (q a.)

~"-·-7--·---

-4

Figure 2.8. Bethe surface for K-shell ionization of carbon, calculated from a hydrogentic
model [9].

31
Ionization energies of electrons from the K, L, M, N, and 0 (quantum numbers
11=1-5 respectively) can be approximated by
-R(Z-E,f

(2.40)

Here R= 13.6 eV is the ionization energy of hydrogen and Z-En is the effective atomic
number decreased by screening [10]. The observed "core edges" are referred to by
nomenclature similar to that of X-ray spectroscopy. Core K, L, M, N, and 0 edges [11]
refer to excitations of K, L, M, N, and 0 shell electrons (Figure 2.9).
The above expressions assume that the incident electrons undergo one inelastic
scattering event. In practice EELS spectra are dominated by plural scattering (> 1
scattering event) and multiple scattering (>20 scattering events). Quantitative analysis of
EELS spectra usually requires data processing techniques to remove plural and multiple
scattering contributions. The Fourier-log method [12] deconvolutes the entire energy loss
spectrum and must be used to determine core edge losses at less than 100 eV. The
Fourier-ratio method [13] provides convenient deconvolution for high-energy core edges.

2.3.2 Data Acquisition and Analysis
EELS spectra were acquired at room temperature using a Gatan 666 parallel
detection magnetic prism spectrometer attached to a Philips EM 420 transmission
electron microscope. A schematic drawing of a parallel EELS spectrometer is shown in
Figure 2.10. Measurements were performed with 100 ke V electrons at a collection angle
of either 11 or 50 mrad. The TEM beam current was approximately 7 nA.

04,5
02,3
01

N6,7
e;~

«.,.0

N4,5

N2,3
Nl

e;<$'~

,,
,,
,,

M4,5

M2,3

L2,3
Ll

,,
,,

1 3

2 2

K shell

1 3

L shell

2 2
3p

3 5

2 2 2 2: 2
4s 4p
4d

M shell

N shell
Energy

5 7

l l ~J'

2222222
4f
5s 5p
5d

./ If 0 shell

Figure 2.9. Possible inner-shell ionization edges and their associated nomenclature [11].

entrance
aperture

ax OY
SX SY
ALIGN
TE cooler
90° prism

electrically
isolated
drift-tube

01
fiber-optic
window

Figure 2.10. Gatan model 666 parallel-recording spectrometer. Q and S represent
quadrupole and sextupole electron lenses respectively [18].

34
The spectrum signal to noise ratio and sample spatial resolution can be optimized
by operating the TEM in either image or diffraction mode. "Coupling" refers to the
object for the spectrometer at the projector crossover. We adopt the microscopists'
terminology referring to "image mode" as the case when the specimen image is projected
onto the view screen and the back focal plane of the objective lens contains the Fourier
transform of the image or diffraction pattern (diffraction coupling, Figure 2.11).
"Diffraction mode" refers to the case when the diffraction pattern is projected on the
screen (image coupling). In this case, the collection angle, B, is determined by the width
of the spectrometer aperture and the camera length (Figure 2.12). 120 mm is the shortest
camera length on the Philips EM 420 and the largest collection aperture is 5 mm. Thus
the maximum obtainable collection angle in diffraction mode is 21 mrad, while the
collection angles in excess of 100 mrad can easily be obtained in image mode. Spectra
are generally acquired at the same Bsince detailed intensity variations in the spectrum
depend on the range of electron scattering angles gathered by the spectrometer.
The magnetic prism bends the electron beam via the Lorentz Force, F = -e v x B,
with -e being the charge of an electron, v its velocity, and B the magnetic field. For v
perpendicular to B, the electron adopts a circular trajectory of radius R. Equating the
Lorentz force with the centripetal force, F = mv 2fR, we see R = mv/eB. Thus, electrons
of lower kinetic energies or lower velocities undergo larger deflection. If the detector is
positioned to accept electrons that traverse roughly one-fourth of the circular orbit, the
spectrometer is called a 90° -prism spectrometer. The electron beam is extremely
sensitive to external magnetic fields. These fields disrupt beam trajectory and influence
spectra resolution. The moving of metal chairs, fields from computer monitors,

Inc i dent Beam

-- - -

Collection Angle ~

Specimen
Ob j ect ive
Lens

Objective
Aperture

I nterm ed i ate
Lens

Viewing Screen

Spectrometer
- - Co 11 ect i on Aperture

Figure 2.11. Collection-angle geometry for energy loss spectra collected in a
transmission electron microscope in image mode [19].

Inc i dent Beam

........ Specimen

Collection
, Angle J3

Camera Length L

t---

Spectrometer
Co 11 ect i on Aperture

Figure 2.12. Collection-angle geometry for energy loss spectra acquired in diffraction
mode [19].

and the opening of metal doors can all cause measured spectra to drift or broaden. Care
was taken to ensure consistent experimental conditions.
The Gatan 666 parallel detector is a single crystal yttrium-aluminum garnet
(YAG) scintallator, fiber-optically coupled to a linear photodiode array of 1024 channels.
Each channel is made up of a 25 Ilm wide diode. A thermoelectric cooler attached to the
photodiode array regulates scintallator temperature, thus minimizing thermal noise
contributions.
Parallel detectors have high detective quantum efficiency but add noise to the raw
data and suffer non-uniform response across the array [14]. A number of steps were
taken to ensure accurate spectra were acquired from the detector. Thermal and electronic
noise components were measured by a detector dark count obtained from the array in the
absence of illumination. A detector dark count was measured for each spectrum under
identical acquisition times. Two techniques were employed to address the non-uniform
channel to channel gain of the detector array. First, an instrument response was recorded
by uniformly illuminating the detector. Subsequent spectra were normalized by this
instrument response (Figure 2.13). Secondly, multiple scans of a feature were taken by
shifting the spectra across the array. These separate scans were subsequently aligned and
summed.

2.3.3 Energy Resolution
The energy distribution in the incident electron beam is dictated by electrostatic
interactions between electrons at the filament crossover (Boersch effect). The resolution
of typical electron sources are 2.5 eV with Tungsten, 1.5 eV with LaB 6 , and 0.5 eV with

1.2
1.0
'CB
(.!)

0.8

"'0
Q)

0.6

ctS

100

0.4
0.2
0.0

200

400
600
Photodiode Channel

Figure 2.13. Instrument response of linear photodiode array.

800

1000

39
cold Field Emission Guns. Energy resolution decreases slightly as the energy loss
increases. This is never more than 1.5 times the zero-loss peak up to about 1000 eV
energy loss. Energy resolution also degrades as electron voltage increases, with the full
width half maximum tripling from 1OOkeV to 400keV. Lastly, resolution degrades with
larger spectrometer entrance apertures due to contributions from off-axis beams.
Optimum energy resolution is acheived with small projector crossover and a small (1 mm
or 2 mm) entrance aperture. The energy resolution of our Gatan 666 spectrometer is 1.2
eV with a dispersion of 0.2 eVlchannel and 2.5 eV at 0.5 eV/channel, as measured from
the shape of the zero-loss peak in the EELS spectra.

2.3.4 Energy Loss Spectrum
An EELS spectrum is a measure of the elastic and inelastic scattering processes of
the transmitted electrons. The important inelastic interactions, in order of energy loss, are
phonon excitations, inter- and intra-band transitions, plasmon excitations, and inner-shell
ionization. Values for the energy loss and scattering semiangle, 8E. are summarized in
Table 2.2. 8 E is the most probable scattering semi angle for a given energy loss.
Table 2.2 Characteristics of EELS Processes
Energy loss (e V)
Process
Phonons
Inter/intra-band transitions
Plasmons
Inner-shell ionization

8E (mrads)

-0.02

5 - 15

5 - 25

5 -10

-5 - 25
-10-1000

<- 0.1
1-5

2.3.4.1 Zero-Loss Peak
The zero-loss or elastic peak is made up of unscattered electrons, elastically
scattered electrons, and electrons which excite phonons. These electrons lose negligible

40
energy relative to EELS measurements upon passing through a sample. The full width
half maximum of the zero-loss peak is a few eV and is dependent on the energy
distribution of the incident electron beam. The intensity of the zero-loss peak can
damage the scintillator and saturate the photo diode array. Care must be taken to
minimize the area of the detector exposed to the zero-loss peak.

2.3.4.2 Low-Loss Spectra
The low-loss region of an EELS spectrum is typically defined as the region from

o e V-50 e V. Spectral features in this region are associated with interactions of the
incident electrons with valence electrons. The principle mechanisms that influence the
low-loss spectrum are plasmon interactions, single-electron excitations and excitons.
Although calculating the intensity of the low-loss region is technically feasible, the
potential results do not yet justify the labor to do so. Spectra from the low-loss region are
best identified by comparison to compiled EELS spectra. The EELS Atlas by Ahn and
Krivanek is an excellent resource [15].

2.3.4.2.1 Plasmons
Valence electrons can be excited to collectively oscillate with modes determined
by the sample's plasma frequency, wp' These oscillations of valence electrons, known as
plasmOllS, are longitudinal waves of charge density propagating through the sample. The

plasmon energy is proportional to the root of the free-electron density, 11.

(2.41 )

41
For h, planks constant, eo, the pennativinty of free space, COp, the plason frequency, e and
m are the electron charge and mass. Free electron densities for metals range from 1022 to

1023 cm-3 , giving plasmon energies - lOeV [16].
For 100 keY incident electrons, the typical mean-free paths for plasmon
excitations is lp, ;:::0 100 nm and scattering angles are typically < 0.1 mrad. Therefore
incident electrons interacting with plasmons are strongly forward scattered. Multiple
plasmon collision events may occur depending on sample thickness. Figure 2.14 displays
plasmons of a thick Al sample. It is speculated that plasmons can be carried by one atom
at a time explaining their occurrence in insulating materials [9]. Unfortunately, all metals
and alloys have similar plasmon energies, although with careful observation it is possible
to observe changes in composition from small shifts in the plasmon peaks.
Above a critical wave vector, qc = w/v F' a plasmon can transfer all its energy to
a single electron, which can then lose energy through an interband transition. This
transfer in energy takes place when the plasmon phase velocity is comparable to the
Fenni velocity, Vp. The transfer of energy to single-electron transitions is the dominant
mechanism of plasmon damping.

2.3.4.2.2 Single -Electron Excitations
In addition to plasmon excitations (a collective electron response), incident
electrons can transfer energy to a single atomic electron within the specimen. If
sufficient energy is transferred, the atomic electron can undergo inter or intra-band
transitions. These single-electron interactions (i.e., the creation of electron-hole pairs)

---

- 15.4 eV

AI volume plasmon
15.8 eV (theory)
15.03 eV (exp.)

100

Enerav Loss. eV

Figure 2.14. Aluminum volume plasmons.

150

43
can add fine structure to the energy-loss spectrum and broaden (possibly shift) plasmon
peaks [17].

2.3.4.2.1 Excitons
In semiconductors and insulators, electrons can be excited from the valence band
to a series of states lying below the bottom of the conduction band. Such an excitation
creates a bound conduction-band electron and a valence-band hole. This bound electronhole pair is called an exciton. The resulting energy loss, Ex, is dependent on the band gap
energy, Eg , and the exciton binding energy Eb (n is an integer).
(2.42)
An exciton can move through the crystal and transport energy, though remains
electrically neutral.
The ability to detect excitons by EELS depends greatly on binding energy. In
weakly bound excitons, the radius of the "orbiting" electron is larger than the interatomic
spacing. These are know as Mott-Wannier excitons and have binding energies, EI> < I
eV. These are commonly found in high-permittivity semiconductors (e.g. Cup, CdS)
and generally not observed in EELS. Frenkel excitons are tightly bound and the hole is
usually on the same atom as the electron. These excitons are common in molecular
crystals and alkali halides (EI> • several electron volts). In alkali halides, energy-loss
peaks due to exciton transitions are observed below the "plasmon" resonance peak
(Section 3.5).

2.3.4.3 Core-Loss Spectra

44
Spectral features above 50 eV are associated with ionization of atoms by
excitation of their core electrons into higher unoccupied energy states (Figure 2.15). The
relaxation of the atom to its initial state may produce a characteristic X-ray or Auger
electron.
Relative to plasmon excitations, the ionization cross sections of core transitions
and their mean-free paths are small. As follows, core edges intensities are small
compared to plasmon peaks and become smaller with increasing energy loss (Equation
2.39). The characterisitc scattering angles, 8E , of core transitions is much larger than
those of plasmons (Table 2.2) and thus, for thick samples, it is common to observe
plasmon signatures superimposed on the near edge structure of a core loss.
The shapes of ionization edges contain information about the valence and
coordination of the excited atom. The shape of the region within 50 eV from the onset
energy, termed the energy loss near edge structure (ELNES), is greatly influenced by the
nature of the lowest unoccupied states (Section 2.3.1). There are no simple relationships
between specific edge types and shapes, although near edge structures can be classified
into various categories: saw-tooth, delayed edge, whitelines, plasmon-like, and sharp
maximum / delayed continuum [10]. Small intensity oscillations can sometimes be
detected at energies extending beyond 50 eV above the onset energy. These oscillations
arise from the constructive and destructive interference of backscattered electrons. These
oscillations are called extended energy-loss fine structure (EXELFS) and are analogous
to extended X-ray absorption fine structure (EXAFS) in X-ray Spectra.

Conduction/valence bands

Neighboring

~c=~================~~~~~============~~~~ruoms
~5~

Counts
70 k

NiOEELS
spectrum

x100

60 k
50k
40k
30k
20k
10k

-0

200

400

600

800
Energy-Loss (eV)

Figure 2.15. The correspondence between the energy levels of electrons surrounding Ni
and 0 atoms and the energy loss spectrum. The zero-loss peak is above the Fermi energy
EF , the plasmon peak is at the energy level f the conduction/valence bands, and the
critical ionization energy required to eject specific K-, L-, and M- shell electrons is show.

46
Ionization-loss electrons, like plasmon-loss electrons, are strongly forward
scattering (Table 2.2). The angular distribution of ionization-loss electrons is generally
confined to < 15 mrad, which is well within the spectrometer collection angle.

2.3.5 XPS and EELS
X-ray photoelectron spectroscopy (XPS) data provide information on corebinding energies by measuring the energy spectrum of ejected photo-electrons.
Differences in the EELS and XPS spectra are a consequence of the different electrons
these methods detect. XPS utilizes photons (AI Ka = 1486.6 eV) to excite bound
electrons into the vacuum level. The energy of an ejected electron, E, is given by
E =tiro - E" -

binding energy and

matrix (Equation 2.37), the final state of the excited electron,

"'I, is within the vacuum

level. In contrast, EELS measures the difference between energy of initial and final
states where the final states,

"'I. include the lowest unoccupied states of the excited

electron. The mean free path of photo-electrons is on the order of 1 nm. Thus XPS
measurements provide analysis of the top few atomic layers. EELS measures the bulk
properties by measuring the energy lost by an electron transmitted through 50 - 100 nm
of material.

47
2.4 References
[1] D. T. Sawyer, A. Sobkowiak, J. L. Roberts Jr, Electrochemistry for Chemists, (John
Wiley & Sons, New York, 1995).
[2] D Linden, Handbook of Batteries, (McGraw-Hill. New York, 1976).
[3] P. Glitlich, R. Link, and A. Trautwein, Mossbauer Spectroscopy and Transition Metal

Chemistry (Springer-Verlag, New York, 1978).
[4] R. L. Cohen, Applications of Mossbauer Spectroscopy (Academic Press, New York,
1976) Volume I.
[5] P. E. J. Flewitt and R. K. Wild, Physical Methods for Materials Characterization,
(Institute of Physics Publishing, London, 1994).
[6] T. Stephens, Chemical Environment Selecitivity in Mossbauer Diffraction, Thesis,
(California Institute of Technology, Pasadena 1996).
[7] G. Le Caer and J. M. Dubois, J. Phys. E : Sci. Instrum. 12, 1083 (1979).
[8] B. T. Fultz, Mossbauer Spectroscopy Applied to Magnetism and Materials Science,
editted G. J. Long and F. Grandjean (Plenum Press, New York, 1993), Chap. 1.
[9] R.F. Egerton, Electron Energy-Loss Spectroscopy in the Electron Microscope,
Plenum Press, New York, 1986.
[10] L. Reimer, Transmission Electron Microscopy, Springer-Verlag, New York. (1997).
[11] D. B. Williams and C. B. Carter. Transmission electron Microscopy; Spectrometry

IV, (Plenum Press, New York, 1996).
[12] R. E. Burge and D. L. Misell, Phil. Mag. 18,261 (1968).
[13] R. F. Egerton and M. J. Whelan, Phil. Mag. 30, 739 (1974).

48
[14] R.F. Egerton, Transmission Electron Energy Loss Spectrometry in Materials
Science, the Minerals, Metals, and Materials Society, Pennsylvania, 1992.
[15] C. C. Ahn and O. L. Krivanek" EELS Atlas, Gatan Inc, Warrendale, PA (1983).
[16] Raether, Heinz, Exiton of Plasmons Interband Transsitions by Electrons, SpringerVerlag, New York, 1980.
(17) W. Y. Liang and S. L. Cundy, Phil. Mag. 19,1031 (1969).
[18] M. M. Disko, C. C. Ahn, and B. Fultz, Transmision Electron Energy Loss

Spectroscopy in Materials Science, (TMS, Warrendale, 1992).
[19] D. H. Pearson, Measurements of White Lines in Transition Metals and Alloys using

Electron Energy Loss Spectroscopy, Thesis, (California Institute of Technology,
Pasadena 1991).

Chapter 3

Lithium EELS Spectra

To interpret local chemical environment effects in EELS spectra of Li materials, it
is important to investigate a systematic series of Li systems to establish standards and
trends. In this section we discuss the near edge structure of EELS spectrum before
exploring various Li systems.

3.1 Near Edge Structure
The electron energy loss near-edge structure (ELNES) of a core excitation can
give insight to the structural and electronic properties of the excited atom. ELNES
essentially probes the local symmetry -projected unoccupied density of states (DOS) at a
particular atomic site. EELS fingerprints arise when the local DOS available to the
excited electron is dominated by the interaction of the central atom with neighboring
atoms. Thus ELNES is sensitive to the valence and coordination of the excited species.
After studying a number of reference compounds with known structural and chemical
properties, it is possible to identify unknown phases in complex microstructures.
The valence of the central atoms greatly affects the edge onset energy of an
excited electron. Shifts in core binding energies rom the effective charge of a central
atom have been well characterized by XPS. The final state energies of the excited
electron also shift with the effective charge of the central atom. The chemical shift of an
edge energy onset, measured by EELS, is the combination of these effects. The chemical
shifts of lithium compounds are especially prominent due to the large electronegativity of
lithium leading to substantial band gap energies in insulating samples.

50
Relaxation effects play an import role in the ELNES of lithium compounds.
When a core hole is created by an inner-shell excitation, the surrounding electron orbitals
are pulled inward. The screening by the surrounding orbitals reduces the magnitude of
the measured binding energy by an amount equal to the "relaxation energy". Leapman et
al. [1] found that a difference in relaxation energy between a metal and its compounds
has a significant effect on the measured chemical shift. Lithium, with its small atomic
number (Z=3), should be particularly susceptible to relaxation effects.

3.2 Beam Damage
3.2.1 Electron Specimen Interactions
The reactivity of reduced Li and its susceptibility to radiation damage under an
electron beam present a formidable challenge to EELS analysis of Li samples. The nature
of the radiation damage depends greatly on the microstructure of the sample. The
dominant effect of radiation damage on crystalline microstructures is atomic
displacement (knock-on damage) resulting in vacancy / interstitial pairs know as Frenkel
defects.
Thin samples of metallic Li and LiF were prepared by thermal evaporation onto
amorphous holey carbon TEM grids. A large glove bag was placed over the evaporator
to minimize atmospheric exposure during transfers. The evaporating chamber was backfilled with Ar where the samples were immersed in Flourinert® FC-43 gettered with Li
chips.
The calculations of threshold energy for knock-on damage suggest a significant
number of inelastic collisions result in atomic Li displacement. In an elastic collision of a

51
stationary particle of mass M and an incident particle of mass m<<.M and velocity v,
conservation of energy and momentum dictate the maximum energy transfered to the
stationary particle, T'ma;., is
(3.1)
Here c is the speed of light. The energy required to displace an atom, Td , is related to the
details of inter-atomic bonding. Average atomic displacement energies estimated from
sublimation energies were found to be 6.1 eV for Li and 31.2 eV for carbon in graphite
[2]. Rearranging Equation 3.1, one can calculate the threshold incident electron energy,

Uth in MeV, which transfers energy, T'max = Tdin eV, to an atom of atomic weight A.
Figure 3.1 displays the values of Uth plotted versus Td [3] .
U th =

.J(l4.4 + 0.186 ·A·Td )
-0.511
20

(3.2)

Assuming elastic collisions, Uth for incident electrons are calculated to be 19 ke V
for Li and 150 keY for carbon in graphite. The dependence of Uth on the atomic weight,
A, indicates we need primarily be concerned with knock-on damage to Li atoms as

opposed to other elements present in our compounds (the exception being Li halides
described below).
To ensure beam damage was not resulting from sample heating, we calculated the
rise in temperature of the sample. The rise in specimen temperature can be derived from
the radial form of the radial form of the differential equation for heat conduction
(3.3)

>Cl
a::
z 50

:E
of
...J

11.

II)

100

200

300

DISPLACEMENT

400

THRESHOLD

500

600

U th • keY

Figure 3.1. Knock-on displacement energy of atomic species [3].

100

800

53
for the appropriate boundary conditions. Here dUldZ is the average rate of energy loss, K
is the specimen thermal conductivity,j is the beam current, and e is the charge of an
electron. The average rate of energy loss to electronic processes in a solid of density N
given determined from the Thomson-Widdington law [3]
(3.4)

where a is the Bohr radius, R is the Ryberg constant and Z is the atomic number. T min is
the minimum energy transferable to an atom constrained in a solid (phonon - 10.2 eV).
For a specimen of uniform thickness, thermally anchored to a good conducting medium
at the periphery r =s, the maximum temperature rise at the center of an irradiated area of
radius b is

(3.4)

Assuming a range in sample thickness of 400 nm to 1000 nm, b =400 nm, S = 20 !lm, j =
7 nA, a rise in sample temperature of 0.5°K to 1.00K is calculated from irradiation of 100
keVelectrons.
Metallic Li samples appeared to "boil" during TEM observations at 200 ke V.
This was determined to be a consequence of knock -on damage since the calculated
temperature rise was negligible. Thus, the majority of radiation damage is a result of the
direct transfer of momentum from incident electrons to the Li nucleus. 100 keV was
chosen as a reasonable compromise between knock-on damage and instrument
performance.

3.2.2 Radiation Damage of Metallic Lithium
Beam damage of metallic Li was studied at 100 keV by concentrating the electron
beam onto a thin area of the Li sample. Figure 3.2 presents a time sequence of EELS
spectra for this exposed region. Over a few minutes, the area of the low-loss region was
significantly reduced, resulting in the suppression ofLi plasmons at 7.5 eV and 15eV.
This is consistent with sample thinning as Li atoms are liberated from the sample under
the electron beam. Under TEM image mode, the specimen was observed to undergo
physical shrinkage, especially under a focused incident beam. Over slightly longer time
periods Li plasmon peaks evolved into distinct profiles commonly associated with
oxidized Li, primarily LiOH [4]. Under a concentrated electron beam, the broad profile
of the Li K-edge evolves to two sharp peaks at 59.5 eV and 64.5 eV, similar to peaks
observed in samples exposed to atmosphere [4]. Exposure to the intense electron beam
can remove metallic Li, exposing preexisting LiOH or stimulate Li reactions with
moisture and oxygen present in the microscope, leading to the formation of LiOH [5].
The possibility of excited lithium reacting with the amorphous carbon substrate has been
ruled out in light of beam damage experiments preformed on Si02 substrates. Radiation
damage of metallic Li on Si0 2 substrate show identical results as those preformed on
amorphous carbon substrate. The Li K-edge in LiC 6 and other reactive Li alloys (LiAl,
Li22Sns, Lh2Sis) all show the same peaks at 59.5 eV and 64.5 eV when exposed to the
atmosphere.
The majority of the observed LiOH is inherent to our specimen and not formed in
situ during exposure to the focused electron beam. The region displayed in Figure 3.2
loses 26% of its low-loss intensity during exposure to the electron beam.

>......
·00

Q)
......

4 minutes

Energy Loss, eV

Figure 3.2. EELS spectra of metallic lithium exposed to focused 100 keY electron beam.

56
This reduction in spectral intensity is consistent with thinning of the specimen. If
LiOH were growing in situ, we would expect the low-loss intensity to increase as the
region's thickness increased relative to the plasmon mean free path.
There is some degree of moisture present in the TEM chamber. When metallic Li
is exposed to an H 20 partial pressure in the TEM vacuum of 10-8 Torr, one finds a
prevalence of Li oxides rather than Li hydroxides [6]. However, it is possible that rate of
LiOH growth is enhanced by the focused electron beam.
Table 3.1 Thermodynamic constants for select lithium compounds [7].
Material
H 2O
LhO
LiOH
LiH)
Lh0 2
LhC02
Li 2C2
LiCI

~Go

~so

(kcaUmol)
-54.6
-134.3
-105.1
-16.4
-136.5
-37.1
-13.4
-91.79

(kcal/mol)
-57.8
-143.1
-116.6
-21.7
-151.2
-47.5
-14.2
-97.58

(kcal/mol)
45.1
9.06
10.23
4.79
13.5
14
14.17

Although from Table 3.1 we find Li 20 is more thermodynamically stable,
hydroxylation of the oxide dominates in the presence of H 20. Zavadil and Armstrong [6]
propose the following reaction pathways for metallic Li exposure to H 20.
(3.5)
The lack of spectroscopic evidence of LiH suggest further hydrolysis and reaction with
metallic Li:
(3.6)
(3.7)
(3.8)

3.2.3 Radiation Damage of Lithium Fluoride
Electron irradiation of alkali halides simulates desorption of halogen atoms
through the mechanism of radiolysis. This leads to the aggregation of the alkali metal and
the formation of color centers [8,9]. Figure 3.3 displays the energy -to-momentum
conversion sequence for the radiolysis of alkali halides. In the case of LiF, this results in
an F 2- molecular ion, called an H center, on a single anion site and the exciton electron at
the anion vacancy, known as an F center. The H-F centres constitute Frenkel defects.
Numerous low-energy electron energy loss spectroscopy (LEELS) studies on LiF
correlate peaks at 2.1eV with point defect (F centers) excitations [10]. These defects are
produced along with a large flux of secondary electrons which induce F centers to
coalesce into F2 centers. EELS spectra of beam damaged LiF show a strong temperature
dependence, most likely attributed to surface relaxation and restoration processes [11].
Correspondingly, metallic Li clusters are smaller at lower temperatures as agglomeration
of metallic islands is supressed.
Figure 3.4 displays EELS spectra of LiF under a concentrated electron beam. The
low-loss spectrum of Figure 3.4a clearly shows a decrease in the characteristic LiF 22 eV
bulk plasmon and the evolution of a 7 eV bulk plasmon of metallic Li. The intensity of
the 7eV plasmon decreases as Li atoms are liberated from the sample with further
irradiation. The area under the LiF spectra decreases with exposure consistent with
thinning of the LiF sample. Figure 3.4b expands the region around the Li K-edge. The
distinct peaks at 62 eV and 70 eV evolve into a broad Li K -edge consistent with metallic
Li. We expect the bold spectra of Figure 3.4 to represent the specimen before much of

( bl

( 01

(c 1

Figure 3.3. Radiolytic displacement sequence for rock-salt structure alkali halides.
Cations M+ are black, anions X- are white and molecular ions X 2· are shaded.

'..-wc:

Q)
..c:

100

Energy Loss, eV

100 50

Energy Loss, eV

Figure 3.4, Evaporated LiP irradiated by 100 ke V electrons for four minutes.

100

60
the created metallic Li was lost to knock-on damage. Figure 3.4c compares the Li Kedge EELS spectra of beam damaged LiF with that of a metallic Li specimen.
Figure 3.5 displays a dark field / bright field pair of the damaged area. The
diffraction pattern contains distinct rings from fcc LiF and broader bands from bcc Li
metal. The dark field / bright field pair was taken using the overlapping (220) ring of LiF
and (211) ring of Li metal. Particles of LiF range from 30 nm to 1 mm, while particles of
Li average 10 nm. The round and broad features of the Li K-edge are consistent with
those of small particles (Figure 3.4b).
A thorough understanding of radiation damage and its effects is essential for
obtaining accurate EELS measurements of lithium compounds. In addition to the
displacement of Li and halide atoms, changes in the valence of transition metals valence
can be induced by radiation damage [12]. These changes in valence can affect the near
edge structure of measured EELS spectra. Specifically with lithium metal oxides,
changes in the transition metal valence can affect the 0 K -edge in addition to the core
edges of the metal. (Section 3.4). Enough beam damage studies were performed in the
course of this work to give confidence in our ability to understand and control the
radiation damage (Figure 3.6).

Figure 3.5. TEM bright and dark field micrographs of evaporated LiF irradiated by 100
ke V electrons. Larger LiF particles (30-1 urn) are imaged from the fcc (220) ring while
metallic Li particles (lOnm) are imaged from the bcc (221) ring.

Figure 3.6. TEM bright field micrograph of evaporated Li 22 Sns irradiated by a focused
beam of 200 ke V electrons.

3.3 Metallic Lithium and Alloys
The low-loss spectrum of a lithium alloy is typically dominated by bulk plasmons.
Figure 3.7 shows the profiles of metallic Li, LiAI, and Lh2Sns with plasmon peaks at 7
eV, 15 eV, and 7 eV respectively. The plasmon-loss energy is proportional to the square
root of the free electron density. The low-loss spectra may contain multiple plasmon
peaks depending on thickness. Buried among the bulk plasmons are the Al L 23 edge at 70
eV, a Sn N 4S edge at 27 eV, and Li K-edges around 55eV. The bulk plasmons shown in
Figure 3.7 overwhelm the other useful edges. Thus, to study the core transitions, one
must use thinner specimens or deconvolute the spectra of thicker samples.
The Li K edges of the tested Li alloys have broad profiles characteristic of
transitions to continuum states. The edges shift relative to metallic Li, consistent with
their respective Fermi Energies. Figure 3.8 displays background-subtracted Li K-edges
of metallic Lithium, LiAI, and Li 22 Sns.
The background-subtracted Li K-edge displayed in Figure 3.8 is comparable to
those reported by Liu and Williams [13]. From optical density of states calculations of
the Li K edge, Bross [14] predicts a sharp peak at 56.3 eV and two broader secondary
peaks at 58.4 and 61.4 eV. These peaks are most likely attributed to Van Hove
singularities (where V'E(k)=O in the energy band structure). Multiple scattering, phonon
broadening at room temperature and the finite energy resolution of the spectrometer limit
our ability to observe the features proposed by Bross.
The large atomic ratio of Li in the Li 22SnS alloy has made this alloy attractive as
an anode material. This alloy has a voltage of 0.1 V vs. lithium and a large theoretical
capacity, 791 mAh/g, but suffers from poor cycle life attributed to mechanical failure

Li plasmon 7 eV
LiAI plasmon 15 eV
Li22 Sns plasmon 7 eV
>.
.....

'00

(])
.....

•..•••••.••.•

LiAI

,....

..... .......

......................

........

........ ,........... . .. .....

Energy Loss, eV

Figure 3.7. The low-loss spectra of metallic Li, LiAI, and Li 22 Sns.

100

."t::
C/)

Q)
+-'

AI L23 73 eV

Energy Loss, e V

Figure 3.8. Lithium K-edges of metallic Li, LiAl, and Liz2SnS.

100

66
(Section 4.2). The crystal unit cell of the LbSns phase contains 432 atoms with 16
distinct Li environments and four Sn environments. Around each Li atom, 66-85% of its
nearest neighbors will be other Li atoms. It is therefore not surprising that the Li K-edge
experiences a minimal chemical shift, <0.2 eV. This lack of chemical shift is also
consistent with the electrochemical potential relative to metallic lithium.
The B32 Zintl phase of LiAI holds great promise as an anode material. This
ordered body-centered cubic (bcc) phase has relatively fast kinetics, a voltage of 0.3 V
vs. lithium, and a theoretical specific capacity of 800 mAh/g. However, poor cycle life
attributed to a face-centered cubic (fcc) phase transition limits its utility.
We attribute the chemical shift of the LiAI Li K-edge to the nature of the Zintl
phase. This class of compounds consists of an electropositive, cation component and an
anionic component of moderate electronegativity. Nevertheless, these Zintl compounds
are not salt-like, but have metallic properties including luster and electronic conductivity.
The profile of the Li K-edge of LiAI is clearly metallic and does not show the sharp
features one expects with transitions to bound states. It is reasonable to believe Li 2p
states are pulled further from Li nucleus by the anioic Al neighbors. Thus, in a core
excitation, the Li Is electron is excited a higher energy state in LiAI than Li metal. This
could explain 2.2 eV chemical shift observed in the LiAI Li Kedge. We define the
chemical shift as the difference in edge onset taken at half the peak maximum (59 eV).
The L23 edge of Al displays a delayed maxima 10-15 ev from the threshold
(Figure 3.9). This is due to the centrifugal potential barrier when 2p electrons are excited
to final states with 1>2. The L23 edge of Al has a profile similar to metallic Al rather than
Ah0 3.

AI metallic

'." .. '.. ~r~. . . .". . . 1I............ . . .

>.
......

-00

Q)
......

100

110

120

Energy Loss, eV

Figure 3.9_ Aluminum L23 edges of metallic AI, LiAI, and Ab03-

130

This is similar to the minor modification of the carbon K-edge by Li intercalation
(Section 4.1.4). We speculate that Li in a metallic state does not significantly reconfigure
the lowest energy states of the solute atoms.

3.4 Lithium Oxides
EELS studies on lithium oxides are important for several reasons. From Section
3.2 on radiation damage we see that it is important to know the lithium oxide spectra
radiation damage of specimens. Additionally, there are numerous opportunities for
contamination during sample preparation. Above all, the removal of moisture from
sample environments is the most important step for the successful specimen preparation
[15]. A number of lithium oxides were studied in order to recognize contaminants and
understand the electronic states of lithium in transition metal oxides.
Lithium hydroxide, LiOH, is the end product of lithium exposure to moisture.
This holds true, to varying extents, for the majority of lithium compounds. Figure 3.10
displays the low-loss spectra of LiOH, Li 2C03, and LiNi0 2 . The distinct peaks at 11 e V,
18.5 eV and 30 eV of the LiOH low-loss spectra are easily discerned in beam damaged
samples. The peak at 18 eV was found to correspond to the (101) ring ofthe TEM
diffraction pattern. This interpretation is consistent with eaylLiOH EELS spectra by
Fellenger [13]. The background subtracted Li K-edge of LiOH displayed in Figure 3.11
show peaks at 59.5 eV and 64.5 eV. These sharp features can be understood qualitatively
by considering the consequence of the Li/LiOH phase transition. Energy bands narrow as
metallic Li evolves into insulating LiOH. The narrow bands correspond to a sharper
density of states profile and thus more well defined final states. Additionally, the Li K-

·00

..........

" '

.............

.........................

100

Energy Loss, eV

Figure 3.10. Low-loss spectra of lithium carbonate (LhC03 ), lithium hydroxide (LiOR),
and lithium nickel oxide (LiNiO z).

.... ····.\UNi02
~.•.

>.....

'w
.....cQ)

.....

..........................

........ .
............-........ .--.

UOH

100

Energy Loss, eV

Figure 3.11. Lithium K-edges of lithium carbonate (LizC0 3), lithium hydroxide (LiOH),
and lithium nickel oxide (LiNi02 ).

71
edge of the insulator is less susceptible to broadening from secondary plasmon scattering.
Liu et al. [15] associates the LiOH peak seen at 68 eV to the density of states around the
second N 1 symmetry point in the N direction in the Li Brillouin zone [16].
Lithium nickel oxide, LiNi0 2 , has been studied as a viable cathode material. It
has the same layered structure as LiCo0 2 (Figure 4.5) but is unstable with lithium
deintercalation due to Jahn-Teller distortion [17]. In this CuPt tetragonal distorted phase
(Section 4.2.1), six oxygen atoms are octahedrally coordinated to a central lithium atom.
The low-loss spectrum of LiNi0 2 displays a plasmon at 9.5 eV, typical of transition metal
oxides. The background subtracted Li K-edge shown in Figure 3.11 has distinct peaks at
61.5 eV and 69.5 eV similar to those seen in LiF (Figures 3.4b, 3.14). The chemical shift
and distinct profile of the LiNi0 2 Li K-edge is attributed to the difference in
electronegativity between Li and its oxygen first nearest neighbors. Structural effects
must also playa role in this Li K-edge profile. Taft and Zhu [18] have found metal KELNES to exhibit a more prominent peak at the edge in the case of octahedral as opposed
to tetrahedral oxygen.
Similarly, first nearest neighbor anions greatly determine the measured oxygen Kedges (Figure 3.12). The oxygen K-edge of LiOH has a doublet peak very similar to the
one seen in the Li K-edge. We speculate these peaks represent the same bound states
associated with the excitation ofLi Is electrons. Thus, there is likely significant
hybridization between the unfilled Li 2p and the oxygen 2p atomic orbitals.
The oxygen K-edge of LiNi0 2 has an additional prepeak at 529 eV associated
with transitions to oxygen 2p states hybridized with Ni 3d states [19]. This type of
hybridization shifts spectral weight to the lowest-energy states in a number of third row

.....

....
....••••

U2 Cf"\,..
, ~-..:s

............: .............................

:!::::

UOH

(f)

520

530

540

550

560

570

580

Energy Loss, eV

Figure 3.12. Oxygen K-edges of lithium carbonate (Li 2C03 ), lithium hydroxide (LiOH),
and lithium nickel oxide (LiNi02 ).

73
transition metal oxides [20]. The prepeak intensity is dependent on the degree of
hybridization with unoccupied 3d states and thus decreases as the number of electrons in
3d band increases [21]. The structure above 535 eV is associated with transitions to
higher energy bands of Ni 4s, 4p and oxygen 3p character.

3.5 Lithium Halides
Lithium halides can be found as components of the solid electrolyte interphase
(Section 1.1). The rock salt structures and ionic bonds of the lithium halides produce
large chemical shifts in the measured Li K-edges. The low-loss region of LiF displays
considerable fine structure between 17 eV - 20 eV due to excitons (Figure 3.13). Figure
3.14 displays the Li K-edges of evaporated lithium halides and a crushed LiNi0 2 sample.
All of these materials have Li atoms coordinated by octahedrally anions. The strongly
electronegative anions form a potential "cage" for electron scattering.
Large chemical shifts in the Li K-edge are observed in EELS spectra of lithium
halides. Relative to metallic lithium, the lowest unoccupied electronic states of Li ions
shift to higher energies as Li 2p orbitals are "pulled" toward the anion neighbors. The
band gaps of the LiF, Liel and LiBr are 14.5 eV, 9.4 eV, and 7.6 eV respectively [22].
The edge thresholds in many ionic insulators correspond to excitations to bound exciton
states within the band gap. Thus the measured chemical shift is reduced by an amount
equal to the exciton binding energy. Figure 3.15 displays a linear relation of anion
electronegativity with the measured chemical shift. In the case of the lithium halides and
theLiNi0 2 samples, the chemical shift was taken to be the difference between the
maximum of the inital transition peak and the onset of the metallic Li K -edge at 54 eV.

.......

"00

LiBr

' \.~~

..... ~

...."....

......

"""'" '.' . ....

..-.,........

......,"' .....
..... f • •" ......

Energy Loss, eV

Figure 3.13. Low-loss spectra of evaporated LiF, Liel, and LiBr.

_~ • •

100

LiNi02~,' "

Br M45 69 eV

"'- ....... ----,..,.... --..,..,-,-

~'--------~~~.~-----------.-------

Energy Loss, eV

Figure 3.14. Lithium K-edges of LiP, Liel, LiBf, and LiNi0 2 •

100

10
LiNiO 2

LiCi

>Q)
+-'

!'!:::

..c

LiF

LiBr

CJ)

..c

• LiAI

()

Slope = 2.35

o~~~~~~~~~~wu~~~~ww~~~~~wu~~

Electronegativity of anion

Figure 3.15. Measured chemical shift of the the Li K-edge versus the anion
electronegativity.

77
The chemical shifts are calculated to be 7.8 eV, 7.4 eV, 6.9 eV and 6.3 eV for LiF,
LiNi0 2 , LiCl, and LiBr respectively. The chemical shift of LiAl is taken from the
difference in Li K-edge onsets at half the peak maximum. Although LiAl is an ordered
bcc phase, polar nature of the Zintl phase justifies its use in this plot.

3.6 Electron Structure Calculations
The near edge structures of the lithium halides and LiOH were calculated using
the self-consistent FEFF8 code of Rehr et al. [23]. This method is similar to KorringaKohn-Rostocker (KKR) band structure calculations but additionally allows for variance
in the number of nearest neighbor shells (cluster size). These central atom
approximations allowed for the core hole potential to be accurately modeled [24, 25]. It
is thus possible to distinguish the contribution of specific scattering paths to the near edge
structure. Cluster sizes of six coordination shells were employed to determine the near
edge structure. Previous work on MgO has shown that six to seven shells are enough to
obtain the near edge structure of these materials [26]. Figure 3.16-3.18 shows the
measured and calculated Li K-edges for various Li halides. All of the simulated Li Kedges predict initial peaks at lower values than those measured. The calculated FEFF8
spectra shown in Figure 3.16-3.18 have been shifted to match measured spectra (LiF - 3.8
eV, LiCl- 2.8 eV, and LiBr - 2.5 eV). The difference between the measured and
calculated edge onset is most likely an underestimation of the respective band gaps of the
Li halides (LiF - 14.78 eV, LiCl- 9.4 eV, LiBr -7.6 eV) [22]. The LiBr Li K-edge is
influenced by a Br M23 edge at 69 eV. M 23 edges have massive delayed maximum which

(f)

LiF FEFF8

100

Energy Loss, eV

Figure 3.16. Lithium K-edges of evaporated LiF and calculated with multiple scattering
FEFF8 code.

·00
Q)

+-'

Liel FEFF8

100

Energy Loss, eV

Figure 3.17. Lithium K-edges of evaporated Liel and calculated with multiple scattering
FEFF8 code.

Br M45 69 eV

>.
......
·00

Q)
......

LiBr FEFF8

100

Energy Loss, eV

Figure 3.18. Lithium K-edges of evaporated LiBr and calculated with multiple scattering
FEFF8 code.

81
accounts for the slow rise in intensity. The M 23 edge probably washes out the small peak
at around 78 e V.
A great deal of work has been done to characterize the features of the Li K -edge
of LiF. Self-consistent local density calculations by Zunger and Freeman [27] accurately
predict interband transition to states having -93% Li 2p character, L1.c, L2.c, and L 3 .c.
(Figure 3.16). The methods used by Zunger and Freeman ignore the possibility of
multiplet structure in the excitons formed upon exciting the Li K-shell. Bandgaps are
commonly underestimated by local density approximations. Using a LiF6· s cluster with a
self-consistent unrestricted Hartree-Fock method, Kunz et. al. [28] calculate fine splitting
in the region from 58 eV to 62 eV. Using the Boukaert-Smaluckowski-Wigner (BSW)
notation common to energy-band theory, the I r l and 3r l state represent the Us states of
the free Li ion. The Iris and 3rlS state represent the up states ofthe free Li ion. The 3.6
eV splitting of I r I _ 3r I is speculated to arise from constraining of Li + excited states by the
electrophobic F ions [29]. In the case of the 1.0 eV splitting of Iris _ 3rlS is speculated to
arise from the orientation of the Li p-like orbitals between and parallel to bonding with
the neighboring flourines. [28]. The transition from the Li K-shell to a ls2s eS) state is
not dipole allowed. The peak at 58 eV has also been observed in XPS observations [30].

82
3.7 Ref,-:rences
[1] R. D. Leapman, L. A. Grunes, and P. L. Fejes, Phys. Rev. B 26,614 (1982).
[2] S. Ono, K. Hojou, and K. Kanaya, Proc. Fifth Int. Conf. on High Voltage Electron
Microscopy, (Jap. Soc. Electron Microscopy, Kyoto, 1977), p. 481.
[3]L. W. Hobbs, Quantitative Electron Microscopy, Proc. of the 25th Scottish Universities

Summer School in Physics, edited by J. N. Chapman and A. J. Craven (SUSSP
Prublications, Edinburgh, 1983), p. 413.
[4] D. R. Liu and D. B. Williams, Phil. Mag. B 53, 123 (1986).
[5] K. R. Zavadil and N. R. Armstrong, Surface Science, 230, 47 (1990).
[6] K. R. Zavadil and N. R. Armstrong, Surface Science, 230, 47 (1990).
[7] W. A. Hart and O. F. Beumel, Comprehensive Inorganic Chemistry, Vol. I, J. C.
Bailer, H. J. Emeleus, R. Nyholm and A. F. Trotman-Dickenson (Pergamon, Oxford,
1973), p. 331.
[8] D. G. Lord and T. E. Gallon, Solid State Commun., 36, 606 (1973).
[9] L. S. Cota, Araiza, and B. D. Powell, Surf Sci. 51, 504 (1975).
[10] Y. Tadami, K. Saiki, and A. Koma, Solid State Commun. 70, 261 (1988).
[11] F. Golek, and W. J. Sobolewski, Solid State Commun, 110, 143 (1999).
[12] L. A. J Garvie and A. J. Craven, Ultramicroscopy. 54, 83 (1994).
metallic Li references
[13] H. Fellenger, Z. Phys. 196,311 (1961).
[14] H. Bross, Proceeding from the International Symposium on X-ray Spectra and
Electronic Structure of Materials, Vol. 2, edited by A. Faessler, G. Wiech, Munich, p. 1
(1973).

83
[15] D. R. Liu, H. E. Rommal, and D. B. Williams, J. of Electron Microscopy Technique,
4, 381 (1986).
[16] H. Bross and G. Bohn, Z. Phys.B 20, 261 (1975).
[17] I. Nakai, K Takahashi, Y. Shiraishi, T. Nakagome, F. Nishikawa, J. of Solid State
Chemistry, 140, 145 (1998).
[18] J. Taft<\> and J. Zhu, Ultramicroscopy, 9, 349 (1982).
[19] P. Kuiper, G. Kruizinga, J. Ghijsen, and G. A. Sawatzky, Phys. Rev. Lett. 62, 221
(1989).
[20] G. van der Laan, J. Zaanan, and G. A. Sawatzky, Phys. Rev. 23, 4253 (1986).
[21] G. van der Laan, C. Weestra, C. Haas, and G. A. Sawatzky, Phys. Rev. 23,4369
(1981).
[22] Landolt-Bomstein: Numerical Data and Functional Relations in cience and
Technology, New Series Vol. 23, subvolume a, edited by O. Madelung, New York, p.
160 (1984).
[23] J. J. Rehr, S.I. Zabinsky, and R. C. Albers, Phys. Rev. Lett., 69, 3397 (1992).
[24] R. Brydson, D. D. Vvedensky, W. Engel, H. Sauer, B. G. Williams, E. Zeitler and J.
M. Thomas, J. Phys. Chern. 92, 962 (1988).
[25] R. Brydson, J. Bruley, and J. M. Thomas, Chern. Phys. Letts. 149, 343 (1988).
[26] L. A. J. Garvie, A. J. Craven, American Mineralogist, 79, 411 (1994).
[27] A. Zunger and A. J. Freeman, Phys. Rev. B 16, 2901 (1977).
[28] A. B. Kunz, J. C. Boisvert, and T. O. Woodruff, Phys. Rev. B 30, 2158 (1984).
[29] T. Aberg and J. L. Dehmer, J. Phys. C 6, 1450 (1973).

84
[30] J. P. Scott, S. L. Hulbert, F. C. Brown, B. Bunker, T. C. Chiang, T. Miller, and K. H.
Tan, Phys. Rev. B 30, 2163 (1984).

Chapter 4

Specific Electrode Environments

4.1 Electron Energy Loss Spectrometry on Lithiated Graphite
4.1.1 Introduction
Lithiated graphite is the standard anode material in Li - ion rechargeable batteries
[1]. Highly crystallized graphite can intercalate one Li atoms to a composition of LiC 6.
This is equivalent to specific charge of 372 Ah kg- I[2], although in practice graphite
anodes have specific capacities of 320 -360 Ah kg-I. Graphite anodes have high voltages
of 3-4V versus the cathode, but the difference in electrochemical potential between
metallic lithium and lithiated graphite is small, about 0.01 V. The intercalation of Li into
highly crystallized graphite changes the hexagonal planes stacking sequence of the
hexagonal planes from an ABABAB to AAAAAA [3]. This change in stacking sequence
and the high chemical potential of Li in graphite suggest that better understanding
interesting of the LiC 6 interlayer states would facilitate improvements to Li- ion
electrochemical cells.
The results of numerous studies on the band structure of lithium intercalated
graphite demonstrate the difficulty in determining the degree of hybridization between Li
atomic orbitals and graphite interlayer states. Early theoretical calculations of the LiQ
band structure began with the notion of complete charge transfer of Li valence electrons
to the graphite 1t bands [4,5]. This evolved into an elegant theory of alkali-intercalated
graphite interlayer states as interacting nonorthogonal hybrid states of Li 2s and graphite
interlayer states [6,7]. This gives credence to X-ray photoelectron spectroscopy (XPS)
results by Momose et al. [8] and others [9] claiming Li to be intercalated into graphite as

86
ionic Li+. Early experimental work by Grunes et aL [10] using Electron Energy Loss
Spectroscopy (EELS) demonstrated distortions of the graphite band structure upon
intercalation of alkali metals. Hartwigsen et aL [11] used a density functional theory,
local density approximation to determine the degree of charge transfer from Li to the
intercalant host lattice to be O.Se for LiC6 and O.4e for LiCg. Further experiments using
inelastic X-ray scattering spectroscopy (IXSS) by SchUlke [12] were able to correlate
features of LiC6 spectra to band structure calculations by Holzwarth et aL [13].
We report transmission EELS measurements of the Li K-edge in intercalated
graphite and in metallic Li. We show the Li K-edge for Li in graphite resembles neutral
metallic Li rather than of Li+.

4.1.2 Experimental
LiC 6 samples were prepared by electrochemical methods [14]. Anodes were
constructed from KS 44 graphite using 6 wt% polyvinylidene fluoride (PVDF) as a
binder. A ternary mixture of aykyl carbonates, i.e., 1: 1: 1 of ethylene carbonate (EC),
diethylene carbonate (DEC), and dimethyl carbonate (DMC) with 1 M LiPF 6 was used as
the electrolyte. Carbon half-cells were subjected to 40 electrochemical charge-discharge
cycles. The carbon electrodes were left in a charged state before being washed in DMC
and vacuum dried.
X-ray diffractometry was performed with an INEL powder diffractometer using Mo Ku
radiation (11.= 0.7092 A). The LiC6 samples were crushed with a mortar and pestle in an Argonfilled glove box. Anode materials were sealed under Ar with paraffin wax in Pyrex capillary
tubes. Detection limits were determined primarily by the statistical quality of the data,

87
estimated to be about 1%. The powdered anode material was placed onto holey carbon TEM
grids. The TEM grids were immersed in Li gettered Flourinert® FC-43 before insertion into
the microscope, where Flourinert evaporated. EELS spectra were acquired at room
temperature using a Gatan 666 parallel detection magnetic prism spectrometer attached to a
Philips EM 420 transmission electron microscope. Energy resolution of the spectrometer was
about l.2 eV with a dispersion of 0.2 eV per channel. Measurements were performed with 100
ke V electrons at a collection angle of 50 mrad. Image coupling was employed to determine
LiC6 spectra. The TEM beam current was approximately 7 nA.

4.1.3 X-ray Diffraction Results
X-ray diffractometery confirmed the presence of LiC 6 and LiC 12 phases in the
graphite anode materials [15,16]. These results were consistent with color of lithiated
graphite, which depends on the predominance of LiC6 (green/gold) or LiC 12 (bluelblack)
[17]. Diffraction peaks were broader for LiC 6 than for the unlithiated KS44 graphite
samples.

4.1.4 TEM Micrographs
The transmission electron micrographs show a heterogeneous distribution of
phases in the graphite samples. The microstructure of the cycled materials ranged from
sheet-like graphite regions to dense polycrystals. As measured from dark field and bright
field images (Figure 4.2), the sheet-like regions were on the order of 100 nm. Moire
fringes were found in some of these regions indicating random orientation of overlapping
graphite regimes. The crystallites in the polycrystalline regions were very small and of

>.....
"w

.....cQ)

LiC 6

LiC 12

Graphite

I I III I I I

II I I I
20

28, degrees

Figure 4.1. X-ray pattern of KS44 and electrochemically lithiated graphite displaying
LiC6 and LiCl2 phases"

200 nm

Figure 4.2. Bright and dark field TEM micrographs of cycled LiC 6 .

90
the order of 10 nm. These polycrystalline regions could be areas of low strained,
organized carbon surrounded by highly buckled or tetrahedrally bonded carbon [15].
Electron Energy Loss Spectrometry determined the presence of intercalated
lithium in both polycrystalline and sheet-like regions. Figure 4.2 shows bright field /
dark field TEM micrographs of a region containing polycrystalline and sheet like
graphite. The characteristic Li K-edge at 52 eV is seen in both the sheet-like and
polycrystalline regions. Regions showing similar morphology and thickness were found
to be without the 52 eV peak. These are assumed to be lithium deficient regions.

4.1.5 Carbon K-edge
Figure 4.3 presents the carbon K edges of LiC6 and KS 44 graphite. A peak at
301 eV is found in LiC6 but not observed in the graphitic carbon. Disko [18] and others
[19] have demonstrated the sensitivity of the carbon K-edge to the direction of the
momentum transfer vector q. We therefore do not expect the intensity of the 1S--71t*
peak to be reliable, since it is not an average over many crystal orientations.
Nevertheless, the onset energy of the carbon K-edge onset energy and the location of the
Is--71t* peak are essentially the same for graphitic carbon and LiC 6.

4.1.6 Lithium K-edge
Figure 4.4 displays the Li K-edge obtained from intercalated LiC 6, LiF, and
metallic Li. We identify the onset of the Li K edge in the metallic sample at 55 eV,
consistent with results of Liu and Williams [20]. The broad profile of the metallic Li Kedge is consistent with the promotion of core electrons into a continuum of free electron

>:=:
(J)

270

280

290

300

310

320

330

340

Energy Loss, eV

Figure 4.3. Carbon K-edge for LiC6 and KS44 normalized by the area 50 eV after the
edge onset.

LiC6

/ '.

....>'00
....cQ)

LiF

+......

........... , ...",

Li metal

--"""--"

..........

.......
50

100

Energy Loss [eV]

Figure 4.4. Li K-edge of metallic Li, LiC6, and LiF. The pre-edge background of the
raw data was fit to an exponential and subtracted from the data.

93
states. The Li K-edge onset of LiC 6 is shifted by only 0.2 eV from the K-edge of metallic
Li and has a similarly broad profile. Our observed Li K-edge for LiC 6 lacks the peaks at
59 eV and 63 eV previously observed by Grunes et al.[10]. These peaks evidently
originate from oxidized Li as they are found in samples exposed to atmosphere, or after
long times in the microscope (Figure 3.2). The Li K-edge of LiF, similar to that reported
by Chen et al.[21], shows a strong of chemical shift and sharp features of well-defined
unoccupied states. Such large differences are expected with the large electron transfer
from Li to F.

4.1.7 Comparison to XPS
Our EELS spectra of the Li K-edge indicate that Li in LiC6 has a local electronic
structure more similar to Li metal than Li+, contrary to previous XPS studies. Extensive
XPS results confirm aLi K-edge chemical shift of about 3 eV in LiC6 with respect to
metallic Li [8,9], easily resolved from the 1.3 eV chemical shift of Li20 [22]. The mean
free path of photoelectrons measured by XPS is on the order of 1 nm. Thus, assuming
that the surface Li remains unreacted, we expect XPS to be more sensitive to Li surface
states.

4.1.8 Discussion
The similarities of Li K-edges in LiC6 and metallic Li are consistent with
observed electrochemical potentials. This imposes significant safety concerns on the
commercial use of lithium ion batteries. Measurements of electrochemical potentials in
half-cells (Figure 4.5) show LiC 6 to be less than 0.01 V above Li metal [15]. The similar

1.000
0.900
0.800

-->
(I)

0.700
0.600

••
••
••
••
0 0.400
•\

• 1st Charge

0.500

... 1st Discharge

(I)

0.300

...
.........
...
.........
...
.........
...

0.200
0.100
0.000
0.000

0.050

0.100

0.150

0.200

0.250

0.300

0.350

Capacity (Ah/g)

Figure 4.5. Initial charge and discharge curves of lithiated KS44 graphite.

0.400

0.450

95
chemical potentials of Li in LiC 6 and metallic Li suggest similar environments for
valence electrons. Electrochemical potentials and EELS Li K-edge profiles both suggest
limited charge transfer from Li to surrounding carbon. The observed carbon K edge is
little affected by intercalated Li, but effects should be diminished by the 6: 1 ratio of C to
Li, and could diminish further if there were a loss of Li from the TEM sample. The shape
of the Li K-edge, on the other hand, would be unaffected by any possible loss of Li.

4.1.9 Conclusions
We report results from transmission EELS studies on lithium intercalated and
discuss beam damage under conditions of observation. Detailed results on the C and Li
K-edges of LiC6 are presented. The carbon K-edge in LiC6 shows an energy onset of 285
eV, unchanged from the edge onset found in graphite. The near edge structure of the
LiC6 carbon K-edge has a peak at 301 eV not found in graphite. Nevertheless, the Li K
edge energy onset was found to be indistinguishable in LiC6 and metallic Li. These
EELS findings contradict the shifted Li K edges found for LiC6 by XPS. We suggest that
XPS spectra are modified by the proximity of Li to the surface of the material.

4.1.10 References
1) A. Herold, Bull Soc. Chim. France, 187,999, (1955).

2) M. S. Dresselhaus and G. Dresselhaus, Adv. Phys., 30, 139 (1980).
3) J. R. Dahn, A. K. Sleigh, H. Shi, B. M. Weydanz, J. N. Reimers, Q. Zhong, and U. von
Sacken,
in Lithium Batteries: New Materials, Developments and Perspectives, edited by G.
Pistoia (WEditor, Amsterdam, 1994), p. 1.
4) N. A. W. Holzwarth, S. G. Louie, S. Rabii, Phys. Rev. B, 26, 5382, (1982).
5) J. J Ritsko, Phys. Rev. B, 25, 6452, (1982).
6) N. A. W. Holzwarth, S. G. Louie, and S. Rabii, Phys. Rev. B, 30, 2219, (1984).
7) M. Posternak, A. Baldereschi, AJ Freeman, Wimmer, Phys Rev. Letters, 50, 761
(1983).
8) H. Momose, H. Honbo, S. Takeuchi, K. Nishimura, T. Horiba, Y. Muranaka, Y.
Kozono, H. Miyadera, Journal Power Sources, 68, 208, (1997).
9) G. K. Wertheim, P.M.Th.M. Vanattekum, and S. Basu, Solid State Communications,

33, 1127, (1979).
10) L. A. Grunes, I. P. Gates, J. J. Ritsko, E. J. Mle, D. P. DiVincenzo, M. E. Preil, and J.
E. Fische, Phys. Rev. B, 28, 6681, (1983).
11) C. Hartwigsen, W. Witschel, E. Spohr, Phys. Rev. B, 55,8,4953, (1997
12) W. Shtilke, A. Berthold, A. Kaprolat, H. J. Guntherodt, Phys. Rev. Letters, 60, 2217,
(1988).
13) N. A. W. Holzwarth, S. G. Louie, and S. Rabii, Phys. Rev. B, 28, 1013, (1983).

97
14) Smart M C, Ratnakumar B V, Surampudi S, Wang Y, Zhang X, Greenbaum S G.
Hightower
A, Ahn C C and Fultz B Journal Electrochemical Society 146, 11, (1999).
15) J. R. Dahn, Phys. Rev. B, 44, 9170 (1991)
16) M. Morita, T. Ichimura, M. Ishikawa and Y. Matsuda, J. Electrochem. Soc., 143, L26
(1996).
17) Basu, G. K. Wertheim, S. B. Dicenzo, in Lithium: Current Applications in Science,

Medicine an Technology, R. o. Bach, (ed.), John Wiley & Sons, New York (1985).
18) M. M. Disko, Analytical Electron Microscopy 1981, edited by R. H. Geiss (San
Francisco Press, Inc., San Francisco, CA, 1981), p. 218.
19) W. Schtilke, K. - J. Gabriel, A. Berthold, and H. Schulte-Schrepping, Solid State
Communications, 79, 657, (1991).
20) D. R. Liu and D. B. Williams, Phil. Mag. B, 53, 123, (1986).
21) T-C. Chen, M. Qian, and T. G. Stoebe, Journal Physics Condensed Matter, 11, 341,
(1999).
22) K. R. Zavadil and N. R. Armstrong, Surcae Science, 230,47, (1990).

98
4.2 EELS Analysis of LiCo02
4.2.1 Introduction
Currently, all large manufacturers of rechargeable lithium ion batteries use
LiCo0 2 as the cathode material. The success of cobalt in cathode materials stems from
its excellent stability under electrochemical cycling. Although other transition metal
oxides involving nickel or manganese offer economic advantages, their Ii mited cycle life
has reduced their practicality.
LiCo02 belongs to a class of lithium metal oxides, LiM02 (M = V, Cr, Co, and
Ni) that have a rhombahedral symmetry corresponding to the R-3m space group (Figure
4.6A). Lithium and cobalt order along the (111) plane form alternating cation planes
sandwiched between oxygen planes [1]. The (111) ordering distorts the lattice to
hexagonal symmetry. This form of LiCo0 2 belongs to the R3m space group with lattice
constants a "'" 2.816 A and c "'" 14.08 A [2] (Figure 4.6B). This layered crystal structure is
typically referred to as the LiCo0 2 "layered" phase. The fully discharged state of the
cathode corresponds to LiCo0 2. With charging of the battery, lithium is deintercalated
from these planes, leaving vacancies behind. Over charging leaves Co0 2, which is
particularly susceptible to cobalt dissolution into the electrolyte [3] or oxygen loss [4].
It is commonly agreed that the transition metals in LiM0 2 cathodes change their
valence to accept the charge of the intercalated Li atom. However, the degree to which
transition metal ions change their valence and thus the amount of charge compensated
oxygen atoms, is open for debate. In the strict ionic charge transfer model, LiCo0 2 ions
have integral charges of Li+, C03 +, and 0 2+. For each lithium atom deintercalated a

o Li

.1
I.:0

o 0

B)
Figure 4.6. A) Schematic diagram of CuPt, ordered rocksalt structure of LiCo0 2 •
Lithium and Co cations order along the (111) planes [24]. B) An equivalent depiction
empathizing the O-Li-O-Co-O-Li-O layers in LiCo0 2 showing ..... ABCABC. ..... stacking
[2]. The c axis of the R 3 m space group is perpendicular to the Li, Co, and 0 planes.

100
cobalt atom would arise with a formal charge of Co4+. Thus in the strict ionic formalism,
oxygen atoms play only a minor role in lithium deintercalation but retain a constant
charge of 0

This is a gross simplification and there is substantial evidence that the

bonding within LiCo0 2 has significant covalent character. This evidence arises from
comparison of calculated and experimentally observed ligand-field effects [5-8].
There is anecdotal evidence suggesting oxygen atoms playa greater role in charge
compensation than previously suspected. The chemical potential of Li in the cathode
changes when substitutions for oxygen and cobalt are made. The cathode voltage
depends significantly on the transition metal involved (Table 1.2). However, greater
changes in voltage can be observed with anion substitutions. I

n the case of LiCo0 2 the

voltage drops considerably when oxide is substituted for sulfur (2.04 Volts) or selenium
(1.46 Volts) [9]. The fact that the voltage of the cathode is more influenced by anion
substitution brings the conventional role of the transition metal into question.
My work investigates the electronic states of cobalt and oxygen in cycled
cathodes by EELS spectroscopy. The first experimental evidence of charge
accommodation by oxygen atoms with lithium intercalation is presented here.

4.2.2 Experimental
The half-cell studies were conducted in three-electrode, O-ring-sealed, glass cells
containing spiral rolls of LiCo0 2 , lithium counter electrodes and lithium reference
electrodes separated by two layers of porous polypropylene (Celgard 2500). The LiCo0 2
cathodes used for these studies were fabricated with 6 wt% polyvinylidene fluoride
(PVDF) binder and 10% Swanigan Carbon Black conductive diluent. The carbonate-

101
based solvent ethylene carbonate (EC), dimethyl carbonate (DMC) and 1,2dimethoxyethane carbonate (DME) (1: 1: 1), containing 1 M LiPF6 salt, was purchased
from Mitsubishi Chemicals (battery grade) with less than 50 ppm of water.
Electrochemical measurements were made using an EG&G PotentiostatiGalvanostat
interfaced with an IBM PC, using EG&G Softcorr 352 software. Electrochemical
charge-discharge measurements and cycling were performed with a battery cycler
manufactured by Arbin Instruments, College Station, TX. The cycling tests were done at
current densities of 0.25 mAlcm 2 for Li intercalation and de-intercalation to the cut-off
potentials of 2.5 and 4.0 V vs. Li, respectively. Cells were held at open circuit voltage
for 15 minutes between the charge-discharge steps.
The electrodes underwent 5 charge-discharge cycles and a final charge before the
analytical measurements. The LiCo0 2 electrodes were washed repeatedly in (DMC) and
dried in an argon-filled glove box. X-ray diffractometry (XRD) was performed on
cycled LiC002 with an INEL CPS-120 powder diffractometer using Mo Ka radiation (A=
0.7092 A). The anode materials were sealed in Pyrex capillary tubes under Ar, using
paraffin wax. Detection limits were determined primarily by the statistical quality of the
data, and estimated to be about 3%.
Transmission Electron Microscopy (TEM) was performed with a Philips EM 420
instrument. TEM samples were mixed with Flourinert™ and crushed with a mortar and
pestle, in an Argon-filled glove box. The powdered mixture of anode material was placed
on a holey carbon microscope grid. Electron Energy Loss Spectroscopy (EELS) was
performed on the TEM samples using a Gatan 666 parallel detection magnetic prism
spectrometer attached to the Philips EM420 TEM. Energy resolution of the spectrometer

102
was approximately 2.0 eV, with a dispersion of 0.5 eV per channel. EELS measurements
were performed with diffraction coupling using 100 ke V electrons and a spectrometer
collection angle of 11 rnrad. The TEM beam current was approximately 7 nA.

4.2.3 Electrochemical
A large capacity fade is observed in the cells during the formation cycles (Figure
4.7), although the reversible capacity was close to what one would expect (- 120 mAh/g).
We attribute the irreversible capacity loss to the build up of a SEI and decomposition of
LiCo0 2. Dahn et al. [10] has reported the following reaction upon heating:

"IJ .•'

LiCoOry- + -6CO,04
+-6Dry, 240°C.

Li~ cCoOry ~ -

4.1

The possibility of cobalt extraction from the cathode, and subsequent deposition on the
anode, was dismissed as cells were not subjected to the high voltages associated with this
process (4.5 V) [11]. The stresses induced by cycling promote cation disorder and
rnicrocracking of LiCo0 2 particles [12]. This is consistent with the broadened peaks
observed by XRD (Section 4.2.4) and the morphologies observed by TEM (Section
4.2.6.1).
The electrochemical discharge curves of Figure 4.8 have continuous profiles
consistent with those of layered LiCo0 2 [2], as opposed to the two step profile of the
spinel LiC002 [13-15]. The observed changes in slope of the discharge curves are
consistent with documented two-phase regions [2, 11, 16].

103

0.160
0.140 '"0.120

--.=c
.-

:::::::--

----

0.100

------.,
-+-A

II(

('Qi

-8

0.080

-.-C
-.-0

c.
0.060 ('Qi

0.040
0.020
0.000

Cycle Number

Figure 4.7. Initial cycles of cells discharged to LiCo0 2 (A), Li o.62 C002 (B), Li o.7SC00 2
(C), and Li o.62 CoO z (D).

104

4.50 , - - - - - - - - - - - - - - - - - - - - - - - - - - - - .

4.00

3.50

3.00

IIIJ

2.50

• B
,. C

• D
2.00 +---+----+----+-----1---.f----+----I-----l
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
0.160

Capacity (Ah)

Figure 4.8. Cells discharged to LiCo02 (A), Lio 62C002 (B), Lio7S C00 2 (C), and Lio62 C002
(D).

105

4.2.4 X-Ray Diffractometry
Figure 4.9 displays XRD patterns of cathodes discharged to various states of
charge. The results confinn the lithium concentrations derived from the discharge curves
(Figure 4.8) and suggest microstructural damage from cycling as discussed below.
Lithium extraction leads to a decrease of interatomic distance within the ab plane
and an increase in interplanar distance between the ab planes [15,16]. This phenomenon
is clearly seen by the decrease in (003) Bragg angle in Figure 4.9. The increase in
interplanar distance is attributed to electrostatic repulsion between the layers of oxygen
atoms sandwiching the lithium planes. Upon Li deintercalation, the (003) peak is known
to disappear near x=0.93, as a new peak grows. This defines a first order phase
transition, 0.75 < x < 0.93, to a monoclinic phase which differs by lattice parameter and
Li concentration [2]. The measured XRD peaks are broadened due to small grain sizes
and nonunifonn strain distribution ..

4.2.5 EELS Analysis
A number of cobalt oxide standards were investigated and compared with cycled
cathodes. CoO has a rocksalt crystal structure with the Co+2 ions octahedrally
coordinated by 0 atoms. C03 0 4 is a spinel with two thirds of the cobalt atoms occupying
octahedral interstices (Co+\ the rest tetrahedral ones (Co+\ Bremsstrahlung isochromat
spectroscopy (BIS) has shown cobalt in LiCo0 2 to be trivalent with a low-spin ground
state configuration, six electrons in the t2g orbitals (dxy , dyz , d zx ) and zero electrons in the
e g orbitals (d x2 -y2 , dz2 )[17].

106

8.0

8.5

9.0

9.5 10.0

+-'

Li .62Coq

Q)
+-'

Li .75Coq

Li .87Coq
110

104
003

20
28, degrees

Figure 4.9. X-ray patterns of cycled LiCoO z.

107
4.2.5.1 Cobalt M 23 , Li K-edge
Figure 4.10 displays the background-subtracted Li K and Cobalt M 23 edges of
CoO, C0 30 4 , and LiCo0 2. The cobalt M 23 edge onset begins at 60 eV and is evidently
affected by the valence and the coordination of cobalt atoms. The plasmon-like profile of
the cobalt M23 transition is characterized as a major delayed edge and displays limited
fine structure [18]. The CoO edge corresponds with the expected profile while the edge
of C0 30 4 is broader and shows more structure. The broad peak at 74 eV is most likely
attributed to the +3 formal charge on two thirds of the cobalt atoms.
The Cobalt M 23 edge washes over much detail of the Li K-edge in LiCo02
spectra. It is clear that the near edge structure of the LiC002 edge is more similar to that
of C0 30 4 than CoO. Again the distinct peak at 75 eV is attributed to the Co+3 formal
charge. The most notable distinction of the LiCo0 2 spectra is the peak at 61.5 e V. This
peak is most likely the contribution of the Li K-edge. It is prudent to compare the spectra
to the Li K-edge of LiNi0 2. LiNi0 2 has the same crystal structure as LiCo0 2but is not
influenced by other overlapping edges (Figure 3.11). The sharp peak at 6leV of the
LiNi0 2 Li K-edges matches well with the observed peak in the LiCo0 2 spectra. Similar
to LiNi0 2, the Li K-edge chemical shift is due to the six electronegative oxygen
neighbors.

4.2.5.2 Oxygen K-edge
Figure 4.11 displays the oxygen K-edges of some standard samples. Previous
investigations on the EELS spectra of transition-metal oxides and full multiple-scattering
calculations have identified the origins of four distinct peaks in the oxygen K -edge [19].

108

Co M2,3 60 eV

+-'

Energy Loss, eV

Figure 4.100 Li K and cobalt M23 edge of CoO, C0304, and LiCo0 2 standards.

100

109

>·00
Q)

520

530

540

550

560

570

Energy Loss, eV

Figure 4.11. Oxygen K-edge of CoO, C0 3 0 4 , and LiCo0 2 standards.

580

110

Labeled a in Figure 4.11, the near edge structure at the threshold is attributed to
transitions to oxygen states p hybridized with cobalt 3d states [17, 20]. Peak b reflects
transitions to oxygen 2p states in the cobalt 4sp band [21]. The structure of peak b is a
function of intrashell multiple scattering within the first oxygen coordination shell.
Features in the oxygen K-edge of MgO were found to arise by a similar mechanism [22].
Peak c arises from intershell multiple scattering from outer-lying oxygen coordination
shells. Peak d is primarily the result of single-scattering events from the first oxygen
coordination shell, and hence is the onset of the extended energy -loss fine structure
(EXELFS) region. Defining R as the first-neighbor oxygen-oxygen distance, the energy
position of peak d follows a linear dependence with IIR2 [19].
There are a number of similarities between the spectra of C0 30 4 and LiCo0 2
while the spectrum of CoO is significantly different. All three spectra agree with
published X-ray absorption spectroscopy results [17]. In the CoO spectrum, peak a (530
eV) is replaced by a peak at 535 eV (Figure 4.11). Structurally, one might expect
comparable oxygen K-edge spectra from the CoO and LiCo0 2 in which oxygen atoms
occupy nearly comparable lattice sites. In this case however, it is the difference in
oxygen bonding which determines the near edge structure of the oxygen K-edge.
The near-edge structure at onset of the oxygen K-edge is determined by the
occupancy of oxygen 2p states. In the purely ionic model, oxygen would have an
electronic configuration of I s22s22p6, and thus no available states for Is electrons to be
promoted into. Covalency increases the number of available oxygen 2p states, and thus is
measured by the intensity at the threshold of the EELS oxygen K-edge. We therefore
conclude that there are oxygen 2p states in LiCo02 and C0 30 4 , which are filled in CoO.

III

This is supported by Hamiltonian cluster calculations of LiCo0 2 and CoO [17]. A
Co06 cluster in octahedral symmetry was employed to determine the ocupancy of cobalt
3d states hybridized oxygen 2p states. Hybridization arises with charge transfer from
oxygen to cobalt. This leaves holes in the oxygen p band represented by the L states.
Tables 4.1 and 4.2 display the occupation number of ground state and ionized state. The
first ionization state of CoO represents those resulting from doping with Li while the first
ionization state of LiCo0 2 can be associated with lithium deintercalation. States with
significantly more L character are found in the ground state of LiCo0 2 compared to
CoO. This increasedL character in LiCo0 2 bonds is consistent with the larger density of
unoccupied oxygen 2p states measured by the EELS oxygen ~edges.
Table 4.1 Occupation numbers of ground state and for electron removal for CoO [17].
GS
Removal
d7
d L
d9 L2

0.79
0.20
0.01

d6
d L
d8L2
d9 L3

0.19
0.62
0.18
0.01

Table 4.2 Occupation number of ground state and for electron removal for LiCo02 [17].
GS
Removal
d6
d L
d8Ll.

d L

0.47
0.44
0.09
0.00

d5
d L
d7 j}
d8j}

0.17
0.51

0.29
0.03

We conclude that there is significant hybridization with the C03+ ions and their
oxygen nearest neighbors. Thus 0 2p states available to 0 Is electrons during core

112
excitations in LiCo0 2 and C0 30 4 are occupied in CoO. The oxygen K-edge pre-peak is
observed in a large number of transitional metal oxides including SC203, Ti0 2, Th03,
V0 2, V 20 3, Mn02, Fe203, Fe304, and CuO [19-21].

4.2.5.3 Cobalt L 23 edge
Figure 4.12 shows the Cobalt L 23 edges of some standard oxides. From calculated
X-ray absorption spectra, one expects multiplet structure in the measured Ledges.
Limits to EELS resolution result in the broadened peaks, although the breadth and shift of
the CoO L3 peak is consisted with calculations using atomic multiplet theory and cubic
crystal field approximations [23]. With Co+3 and Co+ 2 atoms contributing to the C0 30 4
spectrum, it is not surprising to observe peak intensities weighted toward lower energies
in C0 30 4 relative to LiCo0 2. The L2 edge of C0 30 4 and LiCo0 2 is more intense than that
of CoO. This could be attributed to the larger number of available cobalt 3d states in
Co+3 compared to Co+2.

4.2.6 Cycled Cathodes
Cathodes were cycled under relatively passive conditions though they still
incurred a significant amount of electrochemical cycle\ing-induced damage. Particles
displayed varying degrees of damage suggesting differences in depths of discharge on a
local scale.
Significant microstructural damage can occur with delithiation. The observed
crystal structure of LiCo0 2 is Cupt structure with ABC stacking order. Recently, Klein
et al. [1] succeeded in electrochemically delithiating LiCo02 to form C002, which was

113

·00
Q)

775

780

785

C0304-/
LiCo02

790

795

Energy Loss, eV

Figure 4.12. Co L23 edge of CoO, C0 30 4 , and LiCo0 2 standards.

800

114
found to have a Cdl2 structure with AAA stacking. First-principal total energy
calculations confirm these stacking orders as the lowest energy configurations for these
systems [24]. It is reasonable that the first order phase transformation to a monoclinic
phase [2] and the accompanied changes in stacking order can cause significant
microstructural damage.

4.2.6.1 TEM Micrographs of Cycled Cathodes
Figure 4.13 displays two characteristic microstructures observed in the cycled
electrodes. The cathode material was composed of polycrystalline regions, with grains
on the order of tens of nanometers, and highly crystalline regions with grains on the order
of hundreds of nanometers. The size and population of nanocrystals is consistent with
the peak broadening observed in XRD analysis (Section 4.2.4).
The (101) diffraction spots, displaying hexagonal symmetry, are the most
prominent features of the crystalline LiCo02 diffraction pattern (top of Figure 4.13). In
addition, the inner two spots are from (003) planes. The spot highlighted by the arrow in
Figure 4.13 is from the (220) planes of the C03 0 4 spinel phase. The C030 4 is a MgAJz04
type spinel with lattice parameter a =8.08 A and is distinct from the LiCo0 2 synthesized
at lower temperatures (LT-LiCo0 2 ) [18]. The (220) spinel diffraction spots, displaying
hexagonal symmetry, have been observed in a number of cycled cells.
We attribute the elliptical shape of the strong diffraction spots to a contribution
from the spinel phase. With the exception of the (220), all of the low order spinel
diffractions overlap with those allowed in LiCo0 2 [25]. These spots may have
contributions fromLiCo0 2 (101) and the spinel (311) diffractions, which virtually

115

Figure 4.13. Bright field / dark field TEM micrographs of characteristic microstructures

of cycled LiCo0 2 cathodes.

116

overlap. X-ray data of the cycled cathodes do not exhibit a spinel (220) peak, which
suggests that the spinel crystallites are limited to a surface layer. The actual C030 4 phase
most likely contains numerous defects and cation substitutions. The presence of a C0 30 4
phase is best explained through the reaction described by Dahn (Section 4.2.3). As Li
leaves the cathode, oxygen is evolved and in extreme cases the ordered rocksalt of
LiCo0 2 collapses into a spinel of C0 30 4 (Equation 4.1). This transition is more likely to
occur on the surface where the cathode is exposed to the electrolyte.
These results are consistent with those of Wang, et al. [25] who also observe
evidence of electrochemical cycling-induced disorder. They found that the electron
diffraction patterns of cycled cathodes of LiCo0 2 and LiAICo0 2 exhibit a new family of
(220) reflections indicative of the cubic spinel phase. They also report that this evolution

in microstructure is undetectable by X-ray diffraction.

4.2.6.2 EELS Spectra of Cycled Cathodes

Figures 4.14 - 4.16 display background-subtracted EELS spectra of cycled
LiCo02 at various states of charge. The overall features are similar to those observed in
the measured standards with some broadening attributed to a microstructure damaged by
electrochemical cycling. Large thin areas (roughly 0.36 /lm2) were illuminated for EELS
analysis to maximize signal-to-noise ratios. The regions used to acquire spectra
contained a majority of polycrystalline material along with some large grains. This made
it difficult to obtain EELS spectra with good signal to noise ratio. Spectra were
deconvoluted using the Fourier-log method (Section 2.3.1) to correct for thickness
affects.

117

LiCoq

>.
.....
·w
.....cQ)

Li O.75 Cog

Energy Loss, eV

Figure 4.14. Li K and cobalt M 23 edges of cycled Lieo0 2 _

100

118

Li O.62 C002

Int
en
sit

520

530

540

550

560

Energy Loss, eV

Figure 4.15. 0 K-edges of cycled Lieo0 2 •

570

580

119

After 5 Cycles

Li O.62 Coq

>·00

775

780

785

790

Energy Loss, eV

Figure 4.16. Cobalt L 23 edges of cycled LiCo0 2 •

795

800

120
The region containing the cobalt M23 edge (broad onset at 60eV) and Li K-edge
(61 eV) is shown in Figure 4.14. From LiNi0 2 spectra (Section 3.4) and spectra from the
cobalt oxide standards (4.2.5.1), we conclude the sharp peak at 61 eV is the Li K-edge
contribution. It is therefore not surprising to observe this peak to diminish with reduced
Li concentration (Figure 4.14).
There is a striking difference in the oxygen K-edge of the cycled LiCo0 2 (Figure
4.15) from that observed in the LiCo0 2 standard (Section 4.2.5.2). Most notably the fine
structure of the oxygen K-edge is lost to broadening and the pre-peak at 530 eV (peak a)
is reduced in all the cycled samples. There may also be a small shift to higher energies of
the peak at 540 eV (peak b). The most plausible explanation is that there is significant
contribution to the spectra from oxygen atoms not associated with the pristine LiCo0 2
phase. The polycrystalline regions are not observed in the LiCo0 2 starting material and
are most likely the result of cycling damage. Oxygen atoms in these regions likely
occupy a variety of different microstructural sites. These sites could include cobalt
deficient regions, regions of the SEI, or sites at grain boundaries. Oxygen atoms in these
regions could contribute to the intensity oxygen K-edge spectra above the 532 eV but not
to the pre-peak at 530 eV associated the oxygen 2p peaks hybridized with cobalt 3d
states. These issues aside, the intensity of the oxygen K-edge pre-peak increases with
delithiation.
Figure 4.16 displays the cobalt L 23 edges for the cycled LiCo02. A small increase
in peak intensity can be observed, particularly in the L3 peak, as lithium is deintercalated.
This can be attributed to the increase of unfilled 3d states as cobalt ions increase their
valence state to accommodate the absence of lithium charge.

121

4.2.7 Discussion
Though it is difficult to decouple structural and chemical effects, we speculate
that reduced lithium concentration increases the number of available oxygen 2p states.
The presence of lithium vacancies requires charge rearrangement around the neighboring
cobalt and oxygen atoms. It has traditionally been accepted that in lithium metal oxides
(LiM0 2 ) the valence of the transition metal compensates for the charge on the
intercalated Li. The theoretical boundaries for Li concentration in an electrode of

Li xCo0 2 are O:Sx:Sl. Using a simple ionic charge transfer model, the fully lithiated
material has ions with charges of Li+, C03+, and 0 2-. In the Co0 2 end member the ions
have charges of Co4+ and 0 2-. This simple model incorrectly leads to the conclusion that
the oxygen valence is virtually unaffected by Li intercation.
An increase in the number of oxygen 2p states, deduced from the intensity of peak

a, implies charge transfer from the oxygen to cobalt. Thus the oxygen loses a significant
amount of charge as lithium deintercalates.
From another perspective, as lithium intercalates into the cathode, charge
accumulates around oxygen atoms, to shift energy bands downward and lower the
electronic energy. This could explain the shift in the peak at 540 e V (peak b) of the
oxygen K-edge to lower energies (Figure 4.15). This peak is associated with transition to
oxygen 2p states associated with the conduction band. A lowering of these bands with
lithium intercalation would explain the observed shift of this peak to lower energies.
The above argument is consistent with the model of oxygen charge compensation
proposed by Ceder et al. [9,26]. Simulations using the soft pseudopotential method [27]
in the local spin density approximation (LDA) demonstrate that oxygen atoms of the

122

stoich iometric
depleted Li

:t::

C/)

530

535

540

545

Energy Loss (eV)

Figure 4.17 Simulation of oxygen K-edge in LiCo0 2 •

550

123
cathode accommodate charge with Li intercalation. Althougth there was significant
charge rearrangement around the cobalt atoms, little net charge gain was observed.
Bonding within the lattice was found to become increasingly covalent with lithium
deintercalation. There is a metal-insulator transition that accompanies this two-phase
region [28].
This behavior is supported by calculated energy loss spectra using the ab initio
multiple scattering program FEFF8. Figure 4.17 shows a plot of simulated oxygen Kedges for stoichiometric (x = 1.0) and depleted (x =0.66) Li xCo0 2. These simulated
edges were convoluted with a gaussian function and normalized to the main peak at 544
eV to more closely represent the experimental data. These calculations show a decrease
in the pre-eak of the oxygen K-edge with Li intercalation. This decrease indicates a
change in the covalent bonding between the Li and 0 atoms. This simulation predicts a
positive 1 eV shift of the pre-peak with increasing Li concentration, which is not
observed in the experimental data.

4.2.8 Conclusions
The results of EELS studies of electrochemically cycled LiCo0 2 are presented.
EELS measures the 0 2p component of the unoccupied states. The energy of these
unoccupied states depends on Co - 0 hybridization. EELS measurements of LiCo0 2
standards reveal a prepeak in the oxygen K-edge at 530 eV. For a given composition
and phase, the intensity of this prepeak is a measure of covalency in LiCo0 2. The
intensity of the prepeak increase with reduced lithium concentration (0.67 Li x Co0 2. Thus the unoccupied 02p states are filled with Li additions. These findings

124
support the model for oxygen compensation of lithium charge in lithium transition metal
oxides. Significant microstructural damage to LiCo0 2 from electrochemical cycling was
observed by TEM.

125

Table 4.3. Normalized areas of oxygen K-edge peaks from Figure 4.15.

X of LiXCo02

Peak A

PeakB

Peak A I Peak B

0.62

0.30

0.66

0.45

0.75

0.20

0.53

0.38

0.87

0.23

0.67

0.33

1.00

0.04

0.44

0.08

Errors represent uncertainties in the fitting procedure, which arise in the case of strongly
overlapping sub-spectral components.

Table 4.4. Table 4.3. Normalized areas of cobalt L-edge peaks from Figure 4.16.

X of LixC002

L2/ L 3

0.62

0.11

0.33

0.34

0.75

0.07

0.25

0.27

0.87

0.08

0.26

0.31

1.00

0.05

0.20

0.23

126
4.2.9 References
[1] G. G. Amatucci, J. M. Tarascon, and L. C. Klein, J. Electrochem. Soc. 143, 1114
(1996).
[2] J N. Reimers and J. R. Dahn, J. Electrochem. Soc. 139,2091 (1992).
[3] G. G. Amatucci et aI., J. of Power Sources, 69, 11 (1997).
[4] Y. Gao and J. R. Dahn, Solid State Ionics, 84, 33 (1996).
[5] S. Sugano, Y. Tanabe, and H. Kitamura, in Multiplets of Transition Metal Ions in

Crystals (Academic, New York, 1970).
[6] J. Owens and J. H. M. Thronley, Rep. Prog. Phys. 29, 675 (1966).
[7] M. Tsukada, J. Phys. Soc. Jpn. 49, 1183 (1980).
[8] M. Tsukada, H. Adachi, and C. Satoko, Prog. Surf. Sci. 14, 113 (1983).
[9] G. Cedar t aI., Nature, 392, 694 (1998).
[10] J. R Dahn. E. W. Fuller, M. Obrovac, and U. von Sacken, Solid State Ionics, 69, 265
(1994).
[11] G. G. Amatucci, J. M. Tarascon, and L. C. Klein, Solid State Ionies, 83, 167 (1996).
[12] H. Wang, Y.-1. Jang, B. Huang, D.R.. Sadoway, Y. -M. Chiang, J. Pow. Sour. 81-82,
594 (1999).
[13] A. Van der Ven and G. Ceder, Phys. Rev. B 59, 742 (1999).
[14] Zachau-Christiansen et aI., Solid State Ionics, Diffusion & Reactions, 40/41, 580
(1990).
[15] M. M. Thackeray, Prog. Solid State Chern. 25, 1 (1997).
[16] T. Ohzuku and A. Ueda, J. Electrochem. Soc. 141,2972 (1994).
[l7] J. van Elp et aI., Phys. Rev. B 44, 6090 (1991).

127
[18] C. C. Ahn and O. L. Krivanek" EELS Atlas, Gatan Inc, Warrendale, PA (1983).
[19] H. Kurata, E. Lefevre, C. Colliex, and R. Brydson, Phys. Rev. B 47, 13763 (1993).
[20] P. A. van Aken, B Liebscher, V. J. Styrsa, Phys. Chern. Minerals, 25, 494 (1998).
[21] F. M. F. de Groot, M Grioni, J. C. Fuggle, J. Ghijsen, and G. A. Sawatzky, Phys.
Rev. B 40, 5715 (1989).
[22] P. Rez, X. Weng, and H. Ma, Microsc. Microanal. Microstruct, 2, 143 (1991).
[23] F. M. F. de Groot, J. C. Fuggle, B .. Thole, G. A. Sawatzky, and H. Petersen, Phys.
Rev. B, 42, 5759 (1990).
[24] C. Wolverton and Alex Zunger, , Phys. Rev. B, 57, 2242 (1998).
[25] H. F. Wang, Y. 1. Jang, B.Y. Huang, D.R. Sadoway, Y.M. Chiang, Journal of Power
Sources, 82, 594 (1999).
[26] M. K. Aydinol, A. F. Kohan, and G. Cedar, Phys. Rev. B 56, 1354 (1997).
[27] G. Kresse and J. Furthrnuler, Cornput. Mat. Sci. 6, 15 (1996).
[28] H. Tukarnoto and A. R. West, J. Electrochern. Soc.l44, 9 (1997).

128
4.3 119Sn Mossbauer Spectrometry of Li-SnO Anode Materials for Li-Ion Cells
4.3.1 Introduction
There is widespread scientific and technical interest in the high specific energy of
Li-ion cells for secondary batteries. This high specific energy density is derived from the
high cell voltage, typically 4 V (versus 1.3 V for a typical Ni-MH secondary battery).
The Li densities in the cathode and anode materials are modest, however, and it is hoped
that higher capacities of these electrodes will lead to further increases in the specific
energies of Li-ion cells. Anode and cathode materials are now subjects of numerous
investigations [1-3]. Huggins performed some early work on alloy anodes [4,5].
Recently, Ioda, et aI., of Fujifilm Celltec Co., Ltd., announced a new class of anode
material comprising a composite Sn oxide glass [6,7]. These Sn oxide glasses have a
reversible capacity of approximately twice that of carbon materials, but unfortunately
exhibit a large irreversible capacity and capacity fade after tens of charge-discharge
cycles.
Early evidence suggested that the Li inserted in the Sn oxide glass material was
ionic [7], but Courtney et aI., McKinnon and Dahn have provided convincing evidence
for the formation of metallic Sn and Sn-Li alloys during Li insertion [8-10]. The general
picture is that Li reduces the Sn oxides to metallic Sn and lithium oxides. Mao, et aI.
[11,12], has performed further studies on the mechanism of Li insertion in tin oxides and
alloys with 119Sn Mossbauer spectrometry measurements using a sealed cell. With
increasing Li concentration in the anode material, a series of Li-Sn phases were observed
by XRD [13], including Li22Sn5, which represents an increase in volume over that of
pure j3-Sn by a factor of 3.6. Courtney and Dahn argued that since the increase in

129
specific volume induces large local stresses, the cycle life of the electrode is poor when
the Sn-rich regions in the electrode are large [8-13]. The role of microstructure on the
cycle life of Sn oxide anodes remains poorly understood, however.
Althought the Fujifilm Celltec material has not yet been used for products in the
marketplace, its promise has prompted a number of investigations into other Sn and Sn
oxide materials that can be used as anodes in Li-ion cells [9-18]. Huang et al. [14] has
studied the insertion of Li into SnO, showing again that the Li served to reduce the Sn,
and a Li-Sn alloy was formed at higher Li concentrations. Here we report results of
ll9Sn Mossbauer spectrometry measurements at 11 K and 300 K on partially - and fullycharged Li-SnO anode materials. We present detailed measurements of the recoil-free
fractions (RFF.) of the anode materials, and we show that the RFF's of the Sn oxide in
the anode is anomalous, indicative of atomic-scale heterogeneities in the distribution of
Sn atoms. Similar results are reported for the ~-Sn in the anode material, although the
contribution from the ~-Sn is not definitively resolved from the Li22Sn5. We also
present results from a study on the deterioration of Li-charged anode materials and Li-Sn
alloys during long-term atmospheric exposure. During atmospheric exposure, the
selective oxidation of Li causes ~-Sn to separate quickly from the Li-Sn alloy, and the ~
Sn then oxidizes over longer times. The hyperfine parameters of the Mossbauer spectra
indicate that in the early stages of oxidation, the Sn is in small or defective oxides with
Sn4 +, but a little Sn 2+ was observed at later times.

4.3.2 Experimental

130
Electrode and Sample Preparation. - Commercial SnO powder was purchased
from Cerac Chemical. The SnO electrodes were fabricated on copper foil substrates
using 6 wt % polyvinylidene fluoride (PVDF) binder with 10% carbon black as a
conductive diluent. These electrodes were used in half-cells with Li metal as the anode
and an electrolyte of 1M LiPF6 dissolved in 30 % ethylene carbonate (EC) and 70%
dimethylene carbonate (DMC). Selected amounts of Li, varying from 0 to 6.4 mole per
mole of SnO, were titrated into SnO or extracted from LixSnO using galvanostatic
measurements at a current density of 0.020 rnA cm-2 . Here we define "x" in LixSnO as
the molar ratio of Li to the initial SnO. Figure 4.18 shows the initial charge and discharge
curves, with labels indicating the approximate state of the samples used for Mossbauer
spectrometry measurements. In what follows, these samples will be designated Li2.3SnO
and LisSnO. Similar electrodes and cells were used in measurements of cycle life using
an automatic battery cycler made by Arbin Corp., College Station, TX. Cycle life
measurements were performed on sealed full-cells with LiCo0 2 cathodes. The voltage
range was controlled to be between 3.0 and 4.1 V with a charging current density of 0.2
mA/cm2 and a discharging current density of 0.4 mA/cm 2 . Cycle life measurements
were also performed on a half-cell using a Li counter electrode. The half-cell cycle life
measurements used the same current densities for charging and discharging, but had a
voltage range between 0 and 1 V.
To characterize the Li-SnO electrode materials by X-ray diffractometry (XRD),
transmission electron microscopy, and Mossbauer spectrometry, after S cycles the
electrodes were removed from the half-cell and washed in DMC. The electrodes were
vacuum dried and the powders were scraped. All operations were performed in a glove

131
box with a pure and dry argon atmosphere. Samples for XRD were sealed in pyrex
capillary tubes under Ar atmosphere. TEM samples were prepared by crushing anode
material with a mortar and pestle under Fluorinert FC-43 by 3M. Copper backed,
amorphous carbon grids were then dipped into the anodelfluorinert suspension and
inserted into the microscope. For Mossbauer spectrometry, the anode powder was mixed
with a small amount of dehydrated boron nitride powders to ensure an overall thickness
homogeneity. The samples were pressed into pellets that were encapsulated in altuglass
sample holders and sealed with an altuglass glue. The samples comprised typically 50 mg
of anode material. To study oxidation behavior, powdered samples were also stored in
ambient laboratory air. Although the XRD and Mossbauer spectrometry samples were
exposed to ambient conditions on different continents, equivalent atmospherics were
sought by inoculating the Pasadena air with vapors from a bottle of Chateau Ie Barradis
1996.
119Sn Mossbauer spectra were obtained at room temperature (denoted "RT",
nominally 300 K) and at 11 K in transmission geometry with a spectrometer operated in
the conventional constant acceleration mode. A Ba Il9Sn0 3 radiation source with a
strength of - 10 mCi was used. Typical acquisition times were 12 h or 24 h. All ll9Sn
isomer shifts are referenced to BaSn03 at RT. The velocity scale was calibrated with a
metallic body-centered (bee) iron sample and source of 57 Co in Rh. Spectra of
commercially available powders of Sn02, BaSn03, and a 12!-tm foil of metallic [3-Sn
were also measured for use as calibration standards. Spectra of mixtures of known

l32
amounts of commercial powders of Sn02, Li22Sn5, and I3-Sn were also measured at 300
K and 11 K.
All spectra were recorded with adequate counting statistics to permit a
deconvolution of the naturallinewidth from the spectrum. This deconvolution was useful
for revealing the presence of different chemical environments of Sn atoms. The
deconvolution utilized a conventional constrained Hesse-Riibatsch method [19]. The use
of a model-independent histogram method for such an experimental purpose is new to the
best of our knowledge. Since a simple Lorentzian lineshape is deconvoluted from the
experimental spectra, in practice the deconvolution procedure is much the same as using
the constrained Hesse-Riibatsch method for determining an isomer shift distribution. The
calculation of hyperfine parameter distributions from spectra has been the subject of a
number of studies and recent reviews [19-22]. Mathematically, this class of "inverse
problems" is ill-posed. Small changes in spectra may lead to large changes in the
resulting distributions. Regularization methods are therefore required when working with
real experimental data, and regularization impairs the deconvolution procedure. d
Nevertheless, the deconvolution provides a significant and welcome improvement
in resolution. For Sn spectra, the FWHM of the dec onvoluted Lorentz line was chosen
here as a typical minimum FWHM of a single Sn spectral line, that is r=0.82 mmls. As
another consequence of the regularization procedure, small oscillations appear above the
flat zero background of the deconvoluted spectra. These oscillations are artifacts of the
deconvolution procedure. Although the smoothing parameter must be small to avoid

Regularization conditions are set by the smoothing parameter of the constrained Hesse-Rtibatsch method
[19]. For example, when a Lorentzian line with a full-width-at-half-maximum (FWHM) lL = 0.22 mmls is
removed from a calibration spectrum of metallic bcc iron (with a good signal-to-noise ratio) whose inner

133
distorting excessively the main contributions, it must be large enough to avoid the
oscillations that are driven by counting statistics. e All spectra were calculated with the
same fitting conditions. The deconvoluted spectra are normalized to unity.

4.3.3 Electrochemical Results
Figure 4.18 shows the cell voltage vs. Li capacity for the first coulometric
titration of the SnO electrode. The large irreversible capacity of about one -third the
initial capacity (1240 mAhlg) was typical of cells of Sn oxide anode materials [12-18].
This irreversible capacity for the first cycle is probably caused by several factors
including the consumption of Li atoms during the reduction of the SnO, the reduction of
the electrolyte with the formation of a solid-electrolyte interphase, and perhaps the
formation of kinetically inaccessible Li. With x>6.2, the cell voltage of 0 V indicates that
there is an electroplating of Li on the electrode.
Cycle life results are presented in Figure 4.19. The impressively high initial
capacity declines markedly after several cycles, but the deterioration slows considerably
after five cycles. After about 30 cycles the capacity was approximately 300 mAhig in
both the half-cell and full-cell cycling tests. These results are typical of those reported on
other Sn oxide anodes [10-18].

4.3.4 X-Ray Diffractometry and TEM Results

lines have a FWHM r M =0.24 mmls, a line well approximated by a Gaussian line of FWHM r G =0.13
mmls remains. This is much wider than the Lorentzian function of FWHM = 0.02 mmls that is predicted.
, Such oscillations can have only a weak effect on the integral results presented in Tables 4.3 and 4.4,
however.

l34

..--... 2.5
C/)

.:!::::

--- 2.0·

:..J
CJ)

> 1.5

'.+::'

o 1.0Jil--...................
Q)

0..

0.5

O.O~~--~--~--~~--~--~--~--~~--~--~~~~

200

400

600

800

1000

1200

1400

Capacity (mAh/g)

Figure 4.18. First pair of electrochemical charge and discharge curves on SnO anode at
0.02 mAlcm2 . Anodes for further analysis were obtained from samples charged initially
to x=2.3 and x=5, as marked approximately on the figure.

135

600

-E

500

O'J
........
.r::::.

400

'(3

ro
ro

300

200
100
o~~~----~~~----~--~~--~~~--~

Cycle Number

Figure 4.19. Capacity fade data on full cell of SnO anode with LiCo0 2 cathode with 0.43
mAlcm 2 .

136
X-ray diffractometry was performed with Mo Ka radiation (~= 0.07107 nm) using an
Inel CPS 120 powder diffractometer with a position sensitive detector. The electrodes were
composed initially of powdered SnO. Nevertheless, upon the first electrochemical titration of
Li to x=S in LixSnO, XRD showed a distinct diffraction pattern from Li22SnS alone (see
Figure 4.20).

The electrode material charged to x=2.3 showed a mixture of ~-Sn and

Li22SnS, but no SnO. Evidently the SnO is quickly reduced during Li insertion. When more
Li is inserted, the ~-Sn forms an alloy with Li metal. XRD shows this alloy to be Li22SnS
both for low and high Li insertions, but the diffraction peaks from Li 22SnS are broader for the
electrode material charged to x=2.3. Figure 4.20 also shows a significant broadening of the Xray peaks from the ~-Sn in the anode material, indicative of crystallite sizes of 10 nm or so.
There was no evidence for oxides of either Sn or Li in any XRD patterns. From the large
amounts of Sn 4+ observed by Mossbauer spectrometry, we would expect a significant amount
of diffraction from Sn oxides. The absence of diffraction peaks from Sn oxide indicates that it
is probably amorphous, and its broad diffraction pattern was lost when stripping the
diffraction pattern of the glass capillary from the measured data. Although we would expect
only weak X-ray scattering from Li oxides, perhaps near the limit of detectability, we also
suspect that the Li oxides may be amorphous or present as small particles with broadened
diffraction peaks.
Transmission electron microscopy was performed with a Philips EM430
transmission electron microscope operated at 200 keY. A bright field transmission
electron micrograph of the LisSnO material is presented in Figure 4.21. The electron
beam damage to the specimen proved to be quick, substantial, and recognizable as a
change in the shapes of the regions being examined. The image in Fig. 4.20 was acquired

137

28 (degrees)

Figure 4.20. X-ray diffraction patterns of Li 22 SnS alloy, and Sn oxide materials, LixSnO,
charged initially to x=2.3 and x=5.

138

150 nm

Figure 4.21. Bright field transmission electron micrograph of the LisSnO anode
material.

139
within a few seconds after the region was moved into the electron beam. The diffraction
pattern shows primarily ~-Sn, although some diffraction spots from Li22Sn5 were visible
in other diffraction patterns. On the other hand, this material was found to be primarily
Li22Sn5 by XRD. The conversion to ~-Sn suggests substantial oxidation of the thin TEM
sample, as discussed below. The TEM image of Figure 4.21 shows many small regions
that appear dark, likely because they are ~-Sn particles that are diffracting or more
absorbing. These regions range in size from a few nanometers to tens of nanometers. The
lighter regions are probably Li oxide.

4.3.5 Mossbauer Spectrometry Results
4.3.5.1 Mossbauer Spectra of Control Samples
At the top of Figure 4.22 are transmission Mossbauer spectra from ~-Sn, Sn02,
and Li22Sn5 acquired at 300 K (RT). Deconvoluted spectra with higher resolution are
shown below them. Also shown in Figure 4.22 are deconvoluted spectra acquired at 11
K. In general, the more oxidized the Sn, the larger the shift to negative velocities. The

Sn02 spectrum at RT comprises a doublet of equal intensity lines with IS=O ± 0.03 mm1s
and a quadrupole splitting QS= 0.58 ± 0.04 mmls. This is in good agreement with
published values: (QS=0.50 mmls [23], QS=0.61 mm1s [24], QS= 0.55 or 0.56 mmls
[25], QS= 0.56 mm1s [26], QS= 0.54 mmls [27]). The ~-Sn spectrum at RT consists of a
slightly-broadened single line (r=0.95 mm1s) with an isomer shift of 2.52± 0.03 mmls at
RT. It agrees with published values (IS= 2.56 ±0.02 mmls [24], 2.542±0.005 mmls [28]).
It is important to note that the ~-Sn isomer shift falls within the typical range (2.5-2.7

140

- - ...
....

LiZZSns
,' ,

300 K

..-.

Sn f3

--a..
(f)

300 K

11 K
-2

-1

Velocity (mm/s)

Figure 4.22. Top: transmission Mossbauer spectra of J3-Sn, Li 22 SnS' and Sn02 (RT). Below
them are deconvoluted Mossbauer spectra of J3-Sn, Li 22 SnS' and Sn02 (RT).
Middle: deconvoluted Mossbauer spectra from Sn oxide materials, LixSnO, charged to x=2.3
and x=5 (RT).
Bottom: deconvoluted Mossbauer spectra from Sn oxide materials, LixSnO, charged to x=2.3
and x=5 (11 K).

141
mmls) of isomer shifts of Sn 2+ in SnO or in SnOx (x<2) (the precise values depend on x
and on the actual oxide structures). The 13-Sn can be distinguished, however, because the
oxide spectra exhibit quadrupole splittings larger than 1 mmls, and often close to 2 mmls
[23-30]. The isomer shift increases by about 0.04 mmls when 13-Sn is cooled to 77 K
[31]. Theoretical calculations [32] shows that the isomer shift of 13-Sn falls between the
isomer shift of covalent 13-Sn, 2.021 ± 0.012 mm/s at RT [28], and that of a hypothetical
metallic fcc Sn structure. A quadrupole splitting QS=0.41±0.04 mmls is measured for 13Sn from the deconvoluted 11 K spectrum. Our data are in good agreement with results
from recent theoretical calculations [33] of QS=0.37 mmls for C-Sn and QS=0.47 mm/S
for Sn02.
Three spectra were measured from the control sample of Li22SnS: one at RT, one
at 11 K, and one again at RT after recording the low-temperature spectrum. The
spectrum of Li22SnS is mainly a single broadened line. After deconvolution, a main line
with a maximum at 1.93 ± 0.03 mm/s is observed. This line is asymmetric and a high
intensity shoulder at about 1.7S mmls is also observed. A line with a relative area of 0.12
is also observed at 2.S6 ± 0.03 mmls. The deconvo1uted RT spectrum in Fig. 4.21 also
show a much smaller broad line at about 1.0S mmls which may not be experimentally
significant, and very small oscillations that are artifacts of the deconvolution procedure.
(These oscillations are considerably stronger for the 11 K deconvoluted spectrum owing
to poorer counting statistics.) The crystallographic structure of Li22SnS is of the Li22PbS
type (fcc, a=1.978 nm, with 16 formula units). This structure contains 432 atoms per unit
cell and 80 Sn atoms in four different Sn sites with respective proportions 0.2, 0.2, 0.3,
0.3. In principle, these four chemical environments contribute up to eight independent

142
lines to the observed spectrum, but the resolution even of the deconvoluted spectra is
inadequate to identify the eight independent peaks. We can say only that the broadening
of the Li22Sn5 spectrum probably originates with the different crystallographic sites for
Sn atoms in the Li22Sn5 structure. At 11 K, the spectrum is broadened significantly and
rather symmetrically around 2.13 mmls. This suggests an increase of the quadrupole
splittings of the Sn sites when temperature is decreased. The isomer shift of
approximately 2.0 mm1s for Li22Sn5 is consistent with a-Sn. We suggest, however, that
the more negative isomer shift than that of ~-Sn may have a chemical origin involving
charge transfer between the Sn atoms and their Li neighbors.
The 300 K spectrum of Li22Sn5 measured subsequent to the 11 K spectrum is
similar to the original 300 K spectrum, but shows some differences. The main line is still
at 1.95 ± 0.03 mmls, but the shoulder at 1.75 mmls is relatively more intense than
previously. The relative area of the line at 2.56 mmls has increased to 0.19, and a smaller
line is seen at 1.25 mmls. We show below that these changes are consistent with an
oxidation of the sample.
From these results on the control samples, we present a convenient summary of
the velocity ranges in 119Sn spectra:

-1.0 to 1.0 mmls corresponds to Sn4 +
1.0 to 2.3 mmls corresponds to Sn in a Li-Sn alloy
2.2 to 3.2 mmls corresponds to ~-Sn

143
Spectral areas were integrated over these three velocity ranges to prepare Tables 4.S and
4.4 for the different anode materials. For the spectra measured at 11 K, however, the
overlap of the spectral components from ~-Sn and the Li-Sn alloy required that their
spectral components were fit to a pair of overlapping Gaussian lines. This fitting
procedure was more ambiguous than integrating the areas of well-resolved lines, so we
present error bars in Table 4.S to indicate uncertainties in the area fractions. These areas,
normalized by the recoil-free-fractions, could be used to determine the fraction of Sn
atoms in these three different chemical states.

4.3.5.2 Mossbauer Spectra of Anode Materials
Although the anode was prepared from SnO, Mossbauer spectrometry shows that
the anode material of Li2.3SnO is a mixture of ~-Sn plus a Li-Sn alloy, evidently
Li22SnS from the XRD results. The fully-lithiated anode material, LisSnO, is primarily
Li22SnS from XRD, and its Mossbauer spectrum is quite close to that of the standard
sample of Li22SnS. It is not surprising that no SnO is present after five charge/discharge
cycles, because we expect the Li to reduce SnO. Evidently the O-Sn forms with lower Li
stoichiometry, perhaps less than Li2.3SnO, and the Li22SnS forms at higher Li
concentrations, accounting for all the Sn in the sample of LiSSnO. This is approximately
consistent with trends reported for other Sn oxide anode materials [8-14]. Owing to the
low recoil-free-fraction of ~-Sn at RT, the fractions of spectral components in the
spectrum acquired at 11 K are more representative of the actual phase fractions.

144
XRD (Figure 4.20) shows that the Li22Sn5 intermetallic compound forms over a
broad range of Li insertions, at least from Li2.3SnO to Li5SnO. It seems that there are
some differences in this compound for low and high Li insertions, however. The
coulometric titration of Fig. 4.17 shows a change of voltage from 0.7 to 0.25 V between
x=2.3 and x=5.0, although a short plateau is observed at 380 mV near x=4, and can be
associated with Li22Sn5 [34]. The Mossbauer spectrum from the Li22Sn5 in the anode
material of composition Li2.3SnO seems to be shifted towards more negative velocities
than that of the material of composition of Li5SnO (Figure 4.22). Finally, the XRD peaks
from the anode material of composition Li2.3SnO are broader than those from Li5SnO
(Figure 4.20). We suggest that the Li22Sn5 that forms initially is more defective
crystallographic ally than the Li22SnS that forms after more Li insertion. The chemical
potential for Li in this more defective material is evidently lower than that of the more
perfect Li22Sn5. Perhaps the crystallographic defects originate with a sub-stoichiometry
of Li, such as Li vacancies.
The change in the fractional contributions from the Li22Sn5 and the J3-Sn after
the sample of Li5SnO was cooled to 11 K is probably not significant, since there is strong
overlap of these parts of the Mossbauer spectra. The sample of Li 2.3SnO increased
considerably its fraction of Sn 4+ after cryogenic exposure, losing its alloy component and
increasing the fraction of J3-Sn. This is consistent with oxidation of the sample, since one
of the windows on the sample package was not reliable upon cryogenic exposure. A
detailed study of the oxidation behavior of these materials is presented below.

145

4.3.5.3 Recoil-Free Fractions: Standards
The recoil-free fraction (the efficiency of the Mossbauer effect for a 119Sn
nucleus in a given material at a specific temperature) is difficult to measure on an
absolute basis. It is much easier to measure ratios of recoil-free fractions, which can be
ratios of areas of spectral components of different phases, or ratios of areas of the same
spectral component at different temperatures. For anode materials, we prefer the method
of comparing the spectral areas of the same phase, since the phase fractions in a particular
sample remain constant at different temperatures. Before using this method with anode
materials, however, we first prepared four "standard" samples with known amounts of
Sn02 and ~-Sn powders. Standard No.1 had an approximately equal amount of Sn in
the two phases (24 mg Sn02 plus 20 mg ~-Sn). Standard No.2 had more ~-Sn (8 mg
Sn02 plus 47 mg ~-Sn), which ensured more comparable spectral areas at higher
temperatures. The other two standards were mixtures of known amounts of Li22Sn5 with
either ~-Sn or Sn02. Standard 3 was (26.52 mg Li22Sn5 plus 13.42 mg Sn02) and
Standard 4 was (15.44 mg Li22Sn5 plus 24.56 mg ~-Sn).
The measured spectral intensities of the Sn02, ~-Sn , and Li22Sn5 at temperature,
T, are the product of the recoil-free-fractions, f(T), and the atomic fractions, x(T):

IS n02(T) = xSn02 fS n02(T)

1~-Sn(T)

= x~-Sn fD-Sn(T)

(4.2)

(4.3)

146
ILi22Sn5(T) = x Li22Sn5 f Li22Sn5 (T)

(4.4)

For comparison to literature results on ~-Sn and Sn02, we define the ratio, R(T), as:

RS n0 2/Sn(T) =

isnO, (T) __ x(3-s/sno, (T)

ff3-sn (T)

(4.5)

XSnO , I f3-Sn (T)

We obtained RS n02/Sn(11K) = 1.74±O.30 and RS n02/Sn(11K) = 1.40±0.30 for
Standards Nos. 1 and 2, respectively. With values in the literature of fS n 02(llK) = 0.89
at 11K [35], and f~-Sn(lOK)= 0.715 ±0.01 [36], we expect RS n02/Sn(1lK)= 1.24, in
reasonable agreement with our results. At room temperature we find RS n02/Sn(300K) =
8.4±0.8. At room temperature, the Lamb-Mossbauer factors are reported to be:
fS n 02(300K) = 0.56 [27], fS n 02(300K)=0.473 [29], and f~-Sn(300K) = 0.04±0.01 [36],
f(300K) = 0.060±0.002 [37], although for 25 nm nanoparticles of Sn f~_
Sn(300K)=0.022±0.001 [37]. These previously published values give a range of
RS n02/Sn(300K) from 8 to 14, again in reasonable agreement with the results on our

standard samples.
The standard samples composed of Li22Sn5 plus I3-Sn, and Li22Sns plus Sn02,
were measured at RT, at 11 K, and again at RT. It was found that the sample was
unstable over periods of weeks, forming some ~-Sn when the Li reacted with the O.
Nevertheless, from the spectra measured promptly at RT, we found:

147

fbSn(RT)
RSnlLiSn(RT) = fLi22Sn5 (RT)

xLi22Sn5 IbSn(RT)
= xbSn ILi22Sn5 (RT) = 0.50 ± 0.10

xSn02 ILi22Sn5(RT)
fLi22Sn5(RT)
RLiSnlSn02(RT) = fS n02(RT) = xLi22Sn5 IS n02(RT)

= 0.60 ± 0.10

(4.6)

(4.7)

Evidently the alloy Li22Sn5 has a larger effective Debye temperature than ~-Sn, although
smaller than Sn02. The sample of Li22Sn5 plus ~-Sn, which was more stable against Li
oxidation, showed as expected at 11 K that the RFF's of Li22Sn5 and ~-Sn were similar:

fbSn(llK)
RSn/L·S
(11K) -= fLi22Sn5 (IlK)
1 n

xLi22Sn5 IbSn(11K)

= xbSn ILi22Sn5(I1K) -- 1.0 +- 0 .2

(48)

We made an effort to obtain these Debye temperatures from our standards, which are:
277 K for Sn02, 212 K for Li22Sn5, and 134 K for ~-Sn. Errors may be about ±20 K.

4.3.5.4 Recoil-Free Fractions: Anode Materials
It might be expected that the phase fractions of the Li5SnO anode material can be
determined with the recoil-free fraction information from the standard samples.
Qualitative phase fractions can be obtained from Tables 4.5 and 4.6, but we cannot
quantify them further owing to the following interesting phenomenon.

148
Mossbauer spectra from the sample of LisSnO were measured at 11 K and 300 K. We
can use Eqs. 1 and 2 to eliminate the phase fractions to compare the temperature
dependencies of the recoil-free fractions:

RS n02/Sn(RT)

IS n02(RT) IbSn(llK)

RS n0 2/S n == RS n02/Sn(llK) - IbSn(RT) IS n02(1lK)

(4.9)

Using our own experimental results from the calibration standards, we obtain RS n02/Sn
= S.4. From the data in Table 4.S for the anode material of LisSnO, we obtain RS n 02/Sn

= 1.5±0.4.
The unexpectedly different result for RS n 02/Sn and RSnlLi22Sns for the
standards and for the anode material is well beyond expected errors, and can have two
explanations. When comparing the Sn02 and the [3-Sn, either the effective Debye
temperature of the [3-Sn in the anode materials is higher than that of the [3-Sn in the
control sample, or the Debye temperature of the Sn 4 + in the anode material (nominally
Sn02) is lower than that of the Sn4 + in the control sample of Sn02. We cannot prove
one case or the other from the measurements we have performed. One argument is that
the Debye temperature of the Sn02 in the anode material is lower than that of the control
sample, because atomic-scale defects in oxide structure can suppress the Debye
temperature. Assuming the [3-Sn in the anode material to have a Debye temperature of
134 K, we obtain a Debye temperature of 146 K for the Sn 4 + in the anode material.
Another argument is that electropositive Li neighbors may increase the phonon

149

frequencies of ~-Sn. We expect that electropositive Li neighbors can alter one way or
another the phonon frequencies of ~-Sn [38], but it is not clear if the frequencies will
increase when the small ~-Sn particles are of 10 nm spatial dimensions. We consider it
less likely that the anomalous behavior of RS n 0 2/Sn originates with a stiffening of the
vibrational frequencies in the ~-Sn, which would need a Debye temperature close to that
of Sn02. Either case, however, requires that the microstructure of the anode material
contain disorder on nanometer or sub-nanometer dimensions.
In the same way, by comparing the RFF.'s of the Li22SnS alloy and ~-Sn we find
R SnlLi22Sn5 =0.2 for calibration standards, and R SnlLi22SnS =0.35 for the anode
material of LisSnO. We consider this discrepancy insignificant, however, because of the
difficulty of resolving the two overlapping contributions to the Mossbauer spectra at 11
K. We cannot reliably report an anomaly in the RFF. ratio of the ~-Sn and the Li22Sn5 in

the anode material. It is known, however, that the Debye temperature of nanoparticle Sn
is low [37], so it is possible that the Debye temperature of the Li22Sn5 in the anode
material is also suppressed.
One might expect to use the results from XRD to quantify the fractions of phases
in the anode material, but we detected no Sn02 in the XRD patterns. We offer the
following interpretation of these results on the Sn02, in the anode material. First, the
Mossbauer spectrometry tells us only that the Sn is Sn4 +, not necessarily the structure of
crystalline bulk Sn02. The hyperfine parameters of the Sn 4 + in the anode material differ
from those reported in bulk Sn02. In the next section we describe how the Sn 4 +
spectrum of the anode materials evolved to that characteristic of Sn02 when the anode

150
material was exposed to ambient air over a long period of time. The initial Sn 4 + in the
anode material is not in the expected local structure of Sn02. The absence of distinct
XRD peaks indicates that the Sn4 + oxide is either present as extremely small particles, or
as an amorphous phase (or both). The anomalous ratio of RS n02/Sn for Sn in the
electrode material tends to support the interpretation of small particles, although an
amorphous oxide could also have a low recoil free fraction, and would be less visible in
an XRD pattern.

4.3.6 Oxidation in Ambient Air
Figures 4.23 and 4.24 present XRD patterns from the Li2.3SnO and Li5SnO
materials during long-term exposure to ambient air at 300 K. All samples show the same
trends, and oxidation proceeds in two stages. The initial Li-Sn alloy undergoes a
separation into ~-Sn and Li oxide over a period of about 30 minutes for the Li2.3SnO and
several hours for the Li5SnO anode material. This process is followed by a much slower
oxidation of the ~-Sn by the formation of Sn02 observed by Mossbauer spectrometry.
Figures 4.25 - 4.26 present Mossbauer spectra from the Li2.3SnO, Li5SnO and Li22Sn5
materials, respectively, during long-term exposure to ambient atmosphere at 300 K. All
samples show the same trends, and these trends are in good agreement with the results
from XRD. The initial Li-Sn allol undergoes a separation into ~-Sn and Li oxide,
without the formation of Sn4 +. This process is followed by a much slower oxidation of
the ~-Sn by the formation of Sn4 +, which approaches the spectrum of Sn02. In the early

151

28 (degrees)

Figure 4.23. X-ray diffraction patterns from Li 2.3 SnO anode material exposed to ambient air
for various times.

f The small amount of Sn + in the starting material could be associated with oxide regions kinetically
inaccessible to Li. or a surface oxidation associated with inadvertent atmospheric exposure.

152

4 days
80 min
70 min

60 min

50 min

40 min

·00

30 min

29, degrees

Figure 4.24. X-ray diffraction patterns from LisSnO anode material exposed to ambient air
for various times.

IS3
stages of oxidation, for the three materials Li2.3SnO, LisSnO, and Li22Sns, the
hyperfine parameters of the Sn4 + differ from those of crystalline Sn02, but both samples
show a trend towards a symmetric quadrupole doublet, approaching it over 100 days or
so. Since the Sn02 formation was observed clearly by Mossbauer spectrometry, but not
by XRD, we suggest that this oxide has an amorphous structure.
The spectra of Li22SnS exposed to air for more than 100 days show two small
peaks at 1.89 mmls and at 3.61 mmls with QS= 1.72 ± 0.04 mmls and IS= 2.7S ± 0.04
mmls, which are typical of Sn 2+ in SnO [23-26,30,39]. The hyperfine parameters
depend on the structure, and the parameters found here agree with those of amorphous
Sn 2+0 [39] and of ultrafine oxidized Sn particles [2S], but not with those of crystalline
tetragonal SnO [40,41].
We observed a curious small peak at 4 mmls in the anode material undergoing
oxidation. This peak is probably real, but is not easy to understand. It could be one peak
of a quadrupole doublet from an environment with a large electric field gradient, for
instance surface SnO, Sn203 or amorphous SnO x ' Perhaps it originates with an unusual
crystal structure of Sn such as electrochemically-prepared allotropic forms of Sn "u2-Sn"
(42-45], such as fcc diamond type [42] with lattice parameters of about S.66 A for IS
about 4.10 mmls. Ref [44] gives QS=O, IS= 4.08±O.OS mmls.

4.3.7 Thermodynamics of Anode Reactions and Oxidation
The initial insertion reaction of Li into SnO is the energetically favorable reaction of Li
oxidation, which occurs with a change in chemical potential of the Li atom of about 3 eV
with respect to the Li metal reference electrode. The capacity of this reaction to consume

154

LizzSns

. .... .

SnOz

---en_.

118 days

a..

18 days
5 days
Initial
-1

Velocity (mm/s)

Figure 4.25. Deconvo1uted room temperature Mossbauer spectra from Li2.3SnO anode
material exposed to ambient air for various times.

155

SnOz

.,
43 days

.-.

(J)

1 5 days

a..

2 days
1 7 hours
Initial
-1

Velocity (mm/s)

Figures 4.26. Deconvoluted room temperature Mossbauer spectra from LisSnO anode
material exposed to ambient air for various times.

156

122 days

.CJ)

1 5 days

3 days
24 hours
Initial
-1

Velocity (mm/s)

Figure 4.27. Deconvoluted room temperature Mossbauer spectra from Li22Sn5 alloy
exposed to ambient air for various times.

157
Li is much smaller than the subsequent reaction at about 0.5 V, which involves the
fonnation of aLi-rich Li-Sn alloy, evidently the compound Li 22Sn5. Experimental
evidence for these two reactions is seen in Fig. 4.17 at the voltages of 3 and 0.4 V. Note,
however, that the first reaction of Li oxidation has a much larger capacity than expected
from the formation of Li20 or LiO. We attribute this excessive consumption of Li to the
decomposition of electrolyte, as for example in the formation of the surface-electrolyte
interphase (SEI). Neither the oxidation of Li in reacting with SnO, nor the oxidation of
Li during electrolyte decomposition, are reversible reactions in normal cell operation.
The second reaction below 0.8 V with respect to the Li reference electrode is reversible in
nonnal cell operation. In some Sn oxide anode materials the formation of a Li -Sn alloy
may occur in stages involving a number of intermediate alloy compounds [8 -14, 41], but
for our initial SnO anode we find evidence only for the formation of the Li22Sn5
intermetallic compound. From the data of Figure 4.18, it appears that when the Sn is
fully utilized in this compound, the subsequent reaction is the deposition of Li metal on
the anode. This is the third reaction shown at 0.0 V. This reaction is perhaps
electrochemically reversible, but it is likely that the electroplating of Li will cause
changes to the anode or SEI.
With the exposure of the anode material to oxygen in the ambient air, the system
is opened and new reactions occur. The first reaction is the same as occurs in the first
charging of an electrochemical cell, that is, more Li will oxidize. This reaction is favored
thermodynamically more than the oxidation of the Sn, but oxidation of the Sn is expected
once the Li is consumed. The data of Figure 4.23 - 4.27 show that the oxidation of the Li
occurs in tens of minutes, whereas the Sn oxidation occurs over tens of days. The

158

oxidation of Li is strongly favored kinetically over the oxidation of Sn, suggesting that a
higher diffusive mobility of Li atoms than Sn atoms may playa role in the kinetics of
oxidation.
The total free energy change of each reaction must be favorable. In the first
reaction of lithiation of SnO or oxidation of the anode material in air, for example, the Li
oxides must be more stable than the Sn oxides if Li oxidation is to occur. This is true,
with the difference in standard energies of formation of these compounds being about 30
kJ/mol. The chemical bonds involving Li are stronger than those involving Sn, and the

electrode reactions can be understood by consideration of the Li alone. The exception is
the final oxidation of Sn in ambient air, which does not involve further changes to any Li.
lt is interesting that per Sn atom, the standard free energies of formation of SnO and

Sn02 are not very different, being -252 and -258 kJ/mole, respectively. Mossbauer
spectrometry showed that the formation of Sn4 + in a structure related to Sn02 was the
dominant reaction during the exposure of Li22Sn5 to ambient air. With increasing time,
however, we found evidence for the formation of a small amount of Sn 2+ (as in SnO) by
the appearance of a weak Mossbauer peak at about 4 mm1s.

4.3.8 Conclusions
Mossbauer spectrometry and XRD measurements were performed on anode
materials of SnO in which Li was inserted electrochemically. These measurements were
interpreted with the aid of results on standard samples prepared from ~-Sn, Li22Sn5, and
Sn02, measured at room temperature and 11 K. In anode materials, at low Li capacities
the SnO is reduced to small particles of ~-Sn. The Sn4+ in the anode material had an

159

anomalous temperature dependence of its recoil-free fraction, indicating a severely
defective structure on the atomic scale. The lack of XRD peaks from a Sn oxide indicates
that the Sn 4 + is in an amorphous oxide, and may have small spatial features as well. With
increasing Li concentration, there was first the formation of ~-Sn as small particles of

-10 nm dimension, and Li oxide, which may have been amorphous. With more Li
insertion, a Li-Sn alloy was formed. This alloy seems to have been exclusively the
Li22Sn5 intermetallic compound. Although the Li22Sn5 develops over a range of Li
concentrations in the anode material, from XRD it appears that at low Li insertions the
Li22Sn5 is more defective than bulk Li22Sn5' From coulometric titrations, this
crystallographic ally-defective Li22Sn5 appears to have a more favorable chemical
potential for Li atoms.
Although the reactions in SnO-Li electrode materials involve all atom species, the
thermodynamic tendencies for these reactions were dominated by the chemical
preferences of the Li atoms. The voltages of these reactions show that the initial insertion
of Li into SnO is accompanied by an additional oxidation of Li, which we interpret as
electrolyte decomposition in the formation of a surface-electrolyte interphase (SEI), for
example. The irreversible capacity for this initial insertion of Li is large, being 500
mAhlg. The remaining cycles are largely reversible, with Li atoms inserting into and deinserting from metallic alloys that include Li22Sn5. While the early cycles showed an
excellent reversible capacity of greater than 600 mAhlg, the material has an unfortunately
short cycle life of perhaps 30 cycles.
We also studied the oxidation of Li-charged anode materials and Li-Sn alloys
during long-term atmospheric exposure. The oxidation tendencies involved a quick

160
selective oxidation of Li that occurred over tens of minutes. Metallic f)-Sn was observed
during this separation of the Li-Sn alloy. The f)-Sn then oxidized over longer times of
weeks, forming primarily Sn 4 + but later a small amount of Sn 2 +. The hyperfine
parameters of the Mossbauer spectra indicate that in the early stages of oxidation of the
Sn, there is a formation of amorphous, small, or highly defective oxides with Sn4 +.

4.3.9 Acknowledgments
We thank Prof. B. Malaman (Universite Henri Poincare, Nancy I) for useful discussions.

161

Table 4.S. Relative spectral areas fi (i=I,2,3) at 300 K and at 11 K for the Sn4+, LiSn alloy and Ji-Sn contributions in LisSnO.

Sample

fl (Sn4+)

f2 (Lh2SnS alloy)

f3 (/3-Sn)

300 K as-prepared

0.09

0.80

0.11

at 11 K

0.13±0.03

0.62+0.12

0.25±0.10

300 K after 11 K

0.10

0.59

0.31

Errors represent uncertainties in the fitting procedure, which arise in the case of strongly
overlapping sub-spectral components.

Table 4.6. Relative spectral areas fi (i=1,2,3) at 300 K and at 11 K for the Sn4+, LiSn alloy and Ji-Sn contributions in Li2.3SnO.

Sample

f1 (Sn4+)

f2 (alloy)

f3 <13-Sn)

300 K as-prepared

0.01

0.36

0.63

at 11 K

0.07

0.10

0.83

300 K after 11 K

0.12

0.0

0.88

162

4.3.10 References
1. Lithium Batteries, New Materials, Development and Perspectives, G. Pistoia, Editor,

Elsevier, New York, (1994).

2. Lithium Batteries, 1.-P. Gabano, Editor, Academic Press, New York (1983).
3. D. H. Doughty, B. Vyas, T. Takamura, and 1. R. Huff, eds., Materials Research

Society Symposium Proceedings 393, MRS, Pittsburgh (1995).
4. B. A. Boukamp, G. C. Lesh, and R. A. Huggins, 1. Electrochem. Soc. 128, 725 (1981).
5. R. A. Huggins and B. A. Boukamp, U. S. Patent 4,436796 (1984).
6. Y.ldota, et aI., European Patent 651,450Al (1995). Y.ldota, et aI., U. S. Patent
5,478,671 (1995).
7. Y.ldota, T. Kubota, A. Matsufuji, Y. Maekawa, T. Miyasaka, Science 276, 1395
(1997).
8. I.A. Courtney and 1. R. Dahn, 1. Electrochem. Soc. 144,2943 (1997).
9. I.A. Courtney and 1. R. Dahn, 1. Electrochem. Soc. 144,2045 (1997).
10. I. A. Courtney, W. R. McKinnon, and 1. R. Dahn, 1. Electrochem. Soc. 146,59
(1999).
11. O. Mao, R. A. Dunlap, I. A. Courtney, and 1. R. Dahn, 1. Electrochem. Soc. 145,
4195 (1998).
12. O. Mao, R. A. Dunlap, and 1. R. Dahn, 1. Electrochem. Soc. 146,405 (1999).
13. O. Mao and J. H. Dahn, J. Electrochem. Soc. 146,414 (1999).
14. Y. Wang, 1. Sakamoto, C. K. Huang, S. Surampudi, S. G. Greenbaum, Solid State
Ionics 110, 167 (1998).

163
15. T. Brousse, R. Retoux, U. Herterich, and D. M. Schleich, J. Electrochem. Soc. 145, 1
( 1998).
16. W. Liu, X. Huang, Z. Wang, H. Li, and L. Chen, J. Electrochem. Soc. 145,59 (1998).
17. S. C. Nam, Y. H. Kim, W.1. Cho, B. W. Cho, H. S. Chun, and K. S. Yun,
Electrochemical and Solid-State Letters 2, 9 (1999).
18. R. A. Dunlap, O. Mao and J. R. Dahn, Phys. Rev. B 59, 3494 (1999).
19. G. Le Caer and J. M. Dubois, J. Phys. E: Sci. Instrum. 12, 1083 (1979).
20. G. Le Caer and R. A. Brand, J. Phys. Condo Matter 10,10715 (1998)
21. S. J. Campbell and F. Aubertin in Mossbauer Spectroscopy Applied to Inorganic

Chemistry, VoI3,G. Long, Editor, Plenum Press, New York (1988).
22. D. Rancourt in Mossbauer Spectroscopy Applied to Magnetism and Materials

Science, Vol 2, G. Long and F. Grandjean, Editors, Plenum, New York (1996) p. 105.
23. B. Stjema, C. G. Granqvist, A. Seidel and L. Haggstrom, J. Appl. Phys. 68, 6241
(1990).
24. M. S. Moreno, R. C. Mercader and A. G. Bibiloni, J. Phys. Condo Matter 4, 351
(1992).
25. C. H. Shek, J. K. L. Lai, G. M. Lin, Y. F. Zheng and W.H. Liu, J. Phys. Chern. Solids
58, 13 (1997).
26. J. Isidorsson, C. G. Granqvist, L. Haggstrom and E. Nordstrom, J. Appl. Phys. 80,
2367 (1996).
27. J. L. Solis, J. Frantti, V. Lantto, L. Haggstrom and M. Wikner, Phys. Rev. B 57,
13491 (1998).

164
28. J. G. Stevens and W. L. Gettis, Isomer Shift Reference Scales, Mossbauer Effect Data
Center, University of North Carolina, Asheville, NC, USA (1981).
29. M. S. Moreno and R. C. Mercader, Phys. Rev. B 50, 9875 (1994).
30. H. G. Neumann, P. Zeggel and K. Melzer, J. Non-Cryst. Sol. 108, 128 (1989).
31. G. M. Rothberg, S. Guimard and N. Benczer-Koller, Phys. Rev. B 1, 136 (1970).
32. A. Svane, Phys. Rev. Lett. 60, 2693 (1988).
33. A. Svane, N. E. Christensen, C. o. Rodriguez and M. Methfessel, Phys. Rev. B 55,
12572 (1997).
34. J. Q. Wang, I. D. Raistrick, and R. A. Huggins, J. Electrochem. Soc. 133,457 (1986).
35. M. S. Moreno and R. C. Mercader, Hyperfine Interact. 83,415 (1994).
36. C. Hohenemser, Phys. Rev. l39, A185 (1965).
37. I. P. Suzdalev, M. Ya Gen, V. I. Gol'danskii and E. F. Makarov, Soviet Phys. JETP
24, 79 (1967).
38. V. Ozolins and A. Zunger, Phys. Rev. Lett. 82, 767 (1999).
39. G. S. Collins, T. Kachnowski, N. Benczer-Koller and M. Pasternak, Phys. Rev. B 19,
1369 (1979).
40. M. Renteria, A. G. Bibiloni, M. S. Moreno, J. Desimoni, R. C. Mercader, A. Bartos,
M. Uhrmacher and K. P. Lieb, J. Phys. Condo Matter 3,3625 (1991).
41. J. Sangster and C.W. Bale, J. of Phase Equilibria 19,70 (1998).
42. V. Rusanov, T.S. Bonchev, I. Mandjukov and M. Mihov, J. Phys. F: Metal Phys. 16,
515 (1986).
43. V. Rusanov, I. Mandjukov, T.S. Bonchev and K. Kantchev, J. Phys. F: Metal Phys.
18,1311 (1988).

165
44. S.K Peneva, N.S. Neykov and KD. Djuneva, Z. Kristall. 202, 191 (1992).
45. S.K Peneva, N.S. Neykov, V. Rusanov and D.D. Chakarov, 1. Phys.: Condens.
Matter 6,2083 (1994).

166

Appendix A

Mechanical Alloying of Fe and Mg

A.1 Introduction
Magnesium alloys are extremely attractive candidates for hydrogen storage
applications since many can absorb between 3 and 8 wt. % hydrogen [1,2].
Unfortunately, all Mg alloys studied to date have been burdened with both an excessive
hydrogen stability (i.e., inconveniently low equilibrium pressures), and sluggish reaction
kinetics at service temperatures below 300°C [1,2]. During electrochemical chargedischarge cycles, Mg2Ni electrodes have exhibited disappointingly low capacities and
cyclic lifetimes [3,4]. There are recent reports of improvements in the hydriding
behavior of Mg-Ni alloys processed by mechanical alloying [5-8]. Practical Mg alloys
for either hydrogen gas storage [6-8] or electrochemical cells [5] remain elusive,
however.
The present investigation was designed to study Fe-Mg alloys prepared by
mechanical alloying. We were interested in determining if mechanical alloying could
extend the solid solubilities of either Mg in bcc Fe or Fe in hcp Mg, which in equilibrium
are negligible at any temperature [9]. We chose this alloy system so that we could
perform 57Fe Mossbauer spectrometry measurements to obtain information about the
first neighbor chemical environment of Fe atoms in bcc alloys [10]. To our knowledge
there have been no previous studie~ of mechanically alloyed Fe-Mg. Didisheim, et al.
[11] have prepared the ternary hydride Mg2FeH6 by sintering at about 500 e under 20 0

120 atm hydrogen pressure. While this hydride has an excellent 5.5 wt.% hydrogen
content, it is too stable to desorb its hydrogen at practical temperatures. Hydrogen

167
desorption of Mg2FeH6 at elevated temperatures leads to disproportionation into
elemental Mg and Fe, consistent with the very low mutual solubilities of Mg and Fe.
We found that for Mg concentrations up to about 20 at. %, mechanical alloying
can produce single-phase bcc alloys. From density and diffractometry measurements, we
found that the Mg atoms occupy substitutional sites on the bcc lattice. Mossbauer
spectrometry shows, however, that rather few 57Fe atoms have first-nearest-neighbor Mg
atoms, indicating the formation of Mg-rich regions in the bcc lattice. Similarly, we
deduce that the hcp Mg-rich alloys contain some Fe, but the Fe atoms are chistened.
Interactions of hydrogen with these metastable alloys will be the subject of future studies.

A.2 Experimental
The Fe-Mg alloys were prepared from appropriate masses of pure metal powders
of iron (99.9%, 6-9 micron) and magnesium (99.8%, -325 mesh), obtained from Johnson
Matthey, MA, USA. Ball milling was performed for 24 h with a Spex 8000 mixer / mill
using hardened steel vials and balls. Knife edges on the cap and the bodies of the vials
were used to seal the vial with a copper gasket. The ball-to-powder weight ratio was
10: 1. The vial was loaded and sealed in an argon atmosphere. X-ray diffractometry was
performed on the as-milled powders with an INEL CPS-120 powder diffractometer using
Co Ka radiation (A = 1.7902 A). The fractions of bee and hcp phases were determined
by comparing the intensities of the bec (110) and hcp (101) peaks. These peaks were
reasonably immune to the effects of overlap with broadened neighboring peaks. Thus
detection limits were determined primarily by the statistical quality of the data and we

168
estimated to be about 0.5%. The conversions of peak intensities into phase fractions were
done with the expressions [12]:

bee

0.06921 1JQ
1101 + 0.06921 110

=--~.-~.
-101

hep

1101 + 0.06921 110

(la,b)

Lattice parameters were determined using the Nelson-Riley extrapolation method [13].
Grain sizes were estimated with the Scherer approximation
0.89A
L=---B(2e) ·cose

(2)

where L is the crystal thickness, A is the wavelength, andB(28) is the full width at half
maximum for the (10 I) hcp peak and the (110) bcc peak.
Density measurements were performed with the Archimedes method using
toluene and a Mettler digital balance. The density of toluene was calibrated by density
measurements of V, Ti, and eu ingots. These measurements were compared to models of
alloy density by assuming that Mg atoms occupied either Fe sites or interstitial sites in
the bcc matrix. The modeled densities of the substitutional, Psub, and interstitial, Pinter,
alloys are:

2mFe
2mMg
Psub = (l-c~ + c 3

(3a)

2mFe
c 2mMg
Pinter = ~ + l-c a3

(3b)

169
where mFe is the atomic mass of Fe, mMg is the atomic mass ofMg, c is the Mg
concentration in at. %, and a is the measured lattice parameter.
Mossbauer spectrometry was used for measuring the effects of Mg on the
hyperfine magnetic fields (HMF) of Fe. Mossbauer spectra were obtained in
transmission geometry at room temperature with a constant-acceleration spectrometer.
The radiation source was 10 mCi of 57 Co in a Rh matrix. The method of Le Caer and
Dubois [14] was used to extract hyperfine magnetic field (HMF) distributions from the
broadened sextets of these spectra. Saturation magnetization measurements were
performed on the samples at room temperature in an applied field of 4 kOe using a
rotating sample magnetometer calibrated with different masses of Fe and Ni powders.

A.3 Results
A.3.1 X-ray Diffractometry
Figure A.l shows X-ray diffraction patterns from the ball-milled Fe-Mg alloys.
These diffraction patterns show predominantly bcc phase until the Mg composition
exceeds 20 at.%. For Mg concentrations up to 20 at.%, there is an increased broadening
and shift of the bcc diffraction peaks. This behavior suggests that Mg dissolves into the
Fe bcc lattice for concentrations less than 20 at. %. To test this hypothesis, a sample of Fe
- 15 at.% Mg was annealed at 500°C for three hours with the intent of un mixing the Mg
from the Fe. The X-ray diffraction patterns of the as-milled and annealed samples are
compared in Figure A.2. Compared to the annealed sample, the diffraction peaks from
the as-milled sample are clearly broadened and shifted in position. The insert of Figure
A.2 shows the emergence of an hcp phase in the annealed sample. The lattice parameters

170

• bee
hep

•••
>......
·00
Q)
......

• •

"'0
Q)

.. ... • • • •

co
10-

IMg201
IMg151
IMglOl

IMg051
40

100

120

140

28 (deg rees)

Figure A.I. X-ray diffraction patterns of Fe-Mg alloys. Diffraction peaks for Mg
concentrations below 20 at. % are shifted to smaller angles. For alloys of higher Mg
concentration, hcp peaks are observed and the bcc peaks shift back to the bcc Fe
positions.

171

(/)
Q)
+oJ

"'0
Q)

lo.-

100

120

29 (degrees)

Figure A.2. X-ray diffraction patterns off as-milled FessMgls ' and same material after
annealing at sooe for three hours. The insert is an enlargement of the low angle range.

l72
of the as-milled Fe-Mg alloys are shown in Figure A.3. Alloys with Mg concentration
below 20 at. % had X -ray peaks that were shifted to lower angles, corresponding to an
increase in bce lattice parameter. Figure A.3 shows how the lattice parameter of the pure
bcc Fe (initially 2.8662 A [15]) is increased with the addition of Mg, consistent with the
larger metallic radius of Mg.
For higher Mg concentrations, a Mg-rich hcp phase coexists with the Fe-rich bcc
phase. Beyond concentrations of 20 at. % Mg, the lattice parameters for the bcc phase
relax to nearly that of pure iron. Analysis of the hcp lattice parameters assumed a closepacked structure and consistently gave values less than the literature value of 3.2094 A
for pure Mg [15]. This smaller lattice parameter suggests Fe substitution onto the hcp
lattice. Figure A.4 compares the bcc and hcp phase fractions as calculated with Eqns.
1a,b as a function of the Mg concentration. The plot is nonlinear and shows a greater
fraction of hcp phase than the Mg concentration in the alloy.

A.3.2 Density
Density measurements were used to determine the nature of the site occupancy of
Mg in the bcc Fe-rich alloys. As demonstrated by Figure A.5, the experimental density
data most resemble the model for Mg dissolving substitutionally into Fe. The
experimental densities are systematically below the calculated values, possibly because
of some residual porosity in the ball milled samples.

A.3.3 Mossbauer Spectrometry

173

0.291

Pure Mg

-E

::::r

0.290

(")

0.320

-C

:....

CD
.Q)

co
:....
co

"'0
Sll
Sll
~.

0.289

--e- hcp

... bcc

a..

Sll
.-+
.-+

0.318

o·
CD

"'U

Sll
Sll

.- 0.288

CD
.-+
CD

.....J

0.316

..0

::J

-3

0.287

Pure Fe

0.286

0.314
100

Concentration Mg (at. 0/0)

Figure A.3. Lattice parameters of the bcc and hcp phases, with reference data for pure
bcc Fe and hcp Mg. Also shown with a triangle is the lattice parameter of the annealed

174

1.00
0.90
0.80
0.70

......

0.60

ctS
U.

0.50

:....

CIJ

ctS

.s:::.

a..

0.40
0.30
0.20
0.10
0.00

100

Concentration Mg (at. 0/0)

Figure A.4. Fractions of bcc and hcp phases determined from intensities of x-ray
diffraction peaks and Eqs. la,b.

175

8 ..~ ____

substitutional

('t)

E 6

>......
'00 5
Q)

measured

Concentration Mg (at.%)

Figure A.S. Density measurements of as-milled alloys, and calculated curves using Eqs.
2a,b for models of interstitial and substitutional Mg sites.

176

>en

:t:::

"'C
Q)

.t::!

-6

-4

-2

Velocity (mm/s)

Figure A.6. Mossbauer spectra of as-milled Fe-Mg alloys.

177
Some Mossbauer spectra are presented in Figure A.6, and the corresponding HMF
distributions are presented in Figure A.7. These HMF distributions were fit to two
broadened gaussian functions, one centered at 325 kG and the other at 300 ± 2 kG. The
areal ratio of these two gaussian functions is presented as a function of Mg concentration
in Figure 8. For Mg concentrations less than 20%, the peak at 300 kG increases with
greater Mg concentration. For higher Mg concentrations, this 300 kG peak decreases
considerably. The data of Figure A.8 follow closely the trend in lattice parameter shown
in Figure A.3. There is evidence by the peak at 300 kG in the ball milled Fe sample that
there may be a small contamination from Mn in the milling media. The much stronger
peaks at 300 kG observed form the Mg-containing alloys overshadow this contamination
peak, however.

A.3.4 Rotating Magnetometry
The saturation magnetization values obtained with a 4 kOe applied field are
presented in Figure A.9. For all Mgconcentrations, the alloy magnetization per atom
follows approximately a simple dilution with Mg concentration.

A.4 Discussion
A.4.1 Chemical Distributions in bce Alloys
For Mg concentrations less than 20 %, X-ray diffractometry showed that the
samples were entirely bcc phase. This suggests solubility of Mg in the bcc phase, as do
the two observations that 1) the single phase bcc alloy shows an increase in lattice
parameter with Mg concentration (Figure A.3), and 2) the measured density is more

178

60x10

----

Fe standard

-3

FeBM--......,...../\.

T ""

•.

r· ..·\ ~
: ~. ~

/1,,

18 Mg

.'!::

/ ':

..0

CO
..0
0l0-

: \

Ii:

5 Mg

a..

/ !:
!:
!:
i:

\\
\ \\

///

-'
---O~-~-~-~-~-~-~·~~··~~....~~~~~~l-~~~~~~~~'~~.J

260

280

300

320

340

HMF (kG)

Figure A.7. Hyperfine magnetic field distributions obtained from experimental spectra.

179

..... Area (300kG) / Area (325)

0.4

---0

0.3

a..

-.
or"

a..

0.2

0.1

O.O~L..J......L...1..-L..J.....L..J.....L..J.....L..J......L....L..L..J.....L..J.....L..J.....L..J......L...1..-L..J.....L..J.....L..J.....L..J......L....L..L..J.....L...1.....I

Concentration Mg (at.%)

Figure A.S. Ratio p(l)/p(O) (relative numbers of 57Fe atoms with one Mg first-nearestneighbor), obtained fitting to the HMF distributions of Fig. 7 Gaussian peaks at 300 kG
and 330 kG. Note the resemblance to the composition trend of Fig. 3.

180

-0

••

,....

-c

C'>

• •
20

Mg Concentration (at. 0/0)

Figure A.9. Saturation magnetization data for as-milled Fe-Mg alloys at room
temperature in a 4 kOe magnetic field.

100

181
consistent with a substitutional solid solution than an interstitial solid solution (Figure
A.5).
On the other hand, Mossbauer spectrometry indicates that the bcc Fe-Mg samples
are not disordered bcc solid solutions of Mg in Fe. The 57Fe HMF distributions of
Figure A.7 are typical of those from Fe alloys with nonmagnetic solutes that cause no
disturbance of the magnetization of neighboring Fe atoms (i.e., magnetic holes in the bcc
lattice). Each first-nearest-neighbor (lnn) non-magnetic solute atom causes a decrease in
the magnitude of the 57Fe HMF by about 25 kG [10]. For this reason we interpret the
peak at 325 kG as originating with 57Fe atoms with zero Mg neighbors, and the peak at
300 kG as originating with 57Fe atoms with one Mg neighbor. For a truly random solid
solution, the ratio of intensities of the two peaks should be the ratio of intensities of the
binomial probability distributions, P(N,m,c), where N is 8 for the Inn shell, m is the
number of Inn Mg atoms (0 or I), and c is the Mg concentration. The ratio
P(8,1 ,c )IP(8,0,c), is:

P(8,l,c)
P(8,0,c)

£ill
p(O)

(4)

11 71 c(l-c)
O! 8! (I-c)

(5)

182

£ill
p(O)

8c
(l-c)

(6)

For a ratio, p(l)/p(O) = 0.45, the maximum value shown in Figure A.8, the Mg concentration must be 5.3 %. This calculated concentration is smaller by a factor of 3 then
the actual alloy composition of 18 %.
This large discrepancy between Mg conceItrations indicated by Mossbauer
spectrometry and X-ray diffractometry led us to perform an ancillary experiment to
validate our interpretation of the Mossbauer spectra. In particular, if Mg atoms do not act
as magnetic holes in the bcc Fe lattice, we would not expect simple systematics for how
Mg atoms affect the HMF at neighboring 57Fe atoms. While it is improbable that the Mg
atoms themselves hold a magnetic moment, we thought that perhaps the Fe neighbors of
Mg atoms may have enhanced magnetic moments. Such a phenomenon occurs with Ni
and Co neighbors of Fe, although these elements have little chemical resemblance to Mg.
The magnetization results presented in Figure 9 are consistent with simple dilution,
which implies no magnetic moments on the Mg atoms, and little change in the magnetic
moments of those 57Fe atoms having Mg neighbors. For this reason, the 57Fe HMF
caused by Mg atoms is best be interpreted as a disturbance in the Fe 4s conduction
electron polarization that is transferred from the Mg site to neighboring 57Pe atoms [10].
Por Mg concentrations as high as 18 %, Mossbauer spectrometry shows that there
are few 57Fe atoms with Inn Mg neighbors. Nevertheless, density and diffractometry
measurements show that the Mg atoms are within the bcc phase. We are led to the
conclusion that the Mg atoms are in the bcc phase, but their distribution is heterogeneous.
Specifically, we believe the Mg atoms cluster into zones on the bcc lattice, and these

183
zones have few Fe atoms in them. The 57Fe Mossbauer spectra therefore show few Fe
atoms with Mg neighbors. There has been a previous 57Fe Mossbauer spectrometry
study of Fe/Mg artificial multilayers, prepared in thin film form by molecular beam
epitaxy [16, 17]. For thick Fe layers, the spectra reported in this previous study look
much like those of Figure A.6. There is a strong main sextet of peaks characteristic of
pure Fe, and a weaker satellite peak with an average hyperfine field that is about 90 % of
the bulk value identified with Fe atoms near the Mg interface layer [17]. The secondary
peak at 300 kG in Figure A.9 for the ball milled Fe-Mg alloys is 92% of the bulk value at
325 kG.
Chemical unmixing of Fe and Mg atoms in the bcc alloys could also involve some
segregation of Mg atoms to grain boundaries. With a grain size range of 1017 nm for
the bcc crystallites, grain boundary segregation would require only short diffusion
distances. The following experimental evidence shows that grain boundary segregtaion
does not accommodate all Mg atoms, however: 1) in spite of a careful search, we did not
find any variations in the backgrounds of the diffraction patterns between 20 - 50 degrees
28 that would be consistent with diffuse scattering from a disordered Mg structure, 2) the
density of the grain boundary would probably be more consistent with the density of Mg
metal, leading to a simple linear slope across the entire graph of Figure A.55 (without a
kink at 18%), and 3) the alloy lattice parameter is increased by 1.0 % for the 18 at.% Mg
alloy.

A.4.2 Limits of Solubility

184
Unmixing of Fe and Mg atoms on the bcc lattice of the as-milled alloys is
expected from the large difference in electronegativity and size of the Fe and Mg atoms.
Hume-Rothery rules for solubility in mechanical alloying were proposed previously [18].
The rules for a Spex 8000 mixer/mill are that 25 at. % solid solubility occurs when the
metallic radii of the elements differs by 15 %, and their electronegativities differ by less
than 0.4. The metallic radii of Fe and Mg differ by 25 % (Mg is larger), and their
electronegativities differ by 0.64 (Mg is smaller) [19]. The dissolution ofMg in our Ferich bcc alloys is therefore expected to be low.
For Mg concentrations of 20 at. % and higher, the as-milled alloys comprised twophase (bcc plus hcp) mixtures (see Figure A.I). Simultaneously, the lattice parameter of
the bcc phase decreased precipitously, although it remained larger than for pure bcc Fe.
Meanwhile, the lattice parameter of the hcp phase was decreased somewhat below its
value for pure hcp Mg. The intensity of the Inn satellite peak in the 57Fe HMF
distribution (at 300 kG) decreased significantly for the alloys with Mg concentrations
greater than 20 at.%. These changes at Mg concentrations near 20 at.% are reminiscent
of a first-order phase transformation, such as one that occurs by nucleation and growth.
We suggest that during ball milling of Fe-rich alloys there is a significant barrier, either
energetic or kinetic, to the nucleation of the hcp phase. At a composition around 20 at. %
Mg, however, some of the larger Mg-rich regions transform into the hcp structure. The
sharpness of the hcp diffraction peaks indicate hcp crystallite sizes range from 15 - 22
nm, so these Mg-rich hcp crystallites appear to have grown in size by acquiring more Mg.
Without the benefit of Mossbauer spectrometry, we have less detailed information
about the Mg-rich hcp phase. Since the lattice parameter of the hcp phase is smaller than

185

that of pure hcp Mg, we expect that there is some solubility of Fe in the hcp phase. The
curvature of the phase fraction curves in Figure 4 suggests that at Mg concentrations
above 20%, there is greater solubility of Fe in the hcp phase than there is for the Mg in
the bcc, but features of these curves are difficult to interpret quantitatively. The
Mossbauer spectra for Mg concentrations greater than 20% (Figure A.6) did not exhibit
the singlet peak previously attributed to isolated Fe atoms in a Mg matrix [17]. This
results indicates that ball milling does not produce isolated Fe atoms in the hcp Mg
matrix
Data on the lattice parameter of the Mg-rich hcp phase leads us to conclude that
Fe is dissolved in the hcp matrix. The lack of a paramagnetic peak in the Mossbauer
spectra, however, implies that the Fe atoms are clustered on the hcp lattice.

A.S Conclusion
We performed mechanical alloying of Fe and Mg powders with a wide range of
compositions. Alloys with less than 20 at.% Mg were single phase bcc with an enlarged
lattice parameter. For Mg concentrations from 20 - 95 at. %, the alloys were a mix of bcc
and hcp phases. For the single-phase bcc alloys, density measurements indicated that the
Mg substituted onto bcc sites. Magnetization measurements indicated that the alloy
saturation magnetization followed an approximately linear dilution with the concentration
of Mg. Mossbauer spectra of the Fe-Mg alloys showed relatively few Fe atoms with Mg
first-nearest-neighbors, certainly far fewer than would be expected of disordered solid
solutions with Mg concentrations of up to 20 at.%. We conclude that the Mg is

186
distributed heterogeneously on the bcc lattice in Mg-rich zones. Likewise, we found that
some Fe was dissolved heterogeneously on the hcp lattice og Mg-rich alloys.

187
A.6 References
1. 1.1. Reilly, Z. Phys. Chern. NF 117, 155 (1979).

2. K. Yvon, in Encyclopedia of Inorganic Chemistry, edited by R. B. King (Wiley, New
York, 1994), p.l40l.
3. N. Cui, B. Luan, H. K. Liu, H. 1. Zhao, and S. X. Dou, 1. Power Sources 55, 263
(1995).
4. N. Cui, B.Luan, H. J. Zhao, H. K. Liu, and S. X. Dou, J. Alloys Compounds 233, 236
(1996).
5. Y.-Q. Lei, Y.-M Wu, Q.-M. Yang, J. Wu, and Q.-D. Wang, Z. Physk. Chern. 183,379
(1996).
6. M. Y. Song, J. Mater. Sci. 30, 1343 (1995).
7. L. Zaluski, A. Zaluska, 1. O. Strom-Olsen, J. Alloys Compounds 217, 245 (1995).
8. S. Orimo and H. Fujii, J. Alloys Compounds 232, L16 (1996).
9. W. G. Moffatt, Handbook of Binary Phase Diagrams, Vol. 3, (Genium Publishing,
Schenectady, New York, 1984).
10. H. H. Hamdeh, J. Okamoto, and B. Fultz, Phys. Rev. B 42, 6694 (1990). Brent
Fultz, in Mossbauer Spectroscopy Applied to Magnetism and Materials Science Vol. I, G.
J. Long and Fernande Grandjean, eds., (Plenum Press, New York, 1993) p. 1-3l.
11. J.-J. Didisheim, P. Zolliker, K. Yvon, P. Fischer, 1. Schefer, M. Gubelmann, and A.
F. Williams, Inorg. Chern. 23, 1953 (1984).
12. L. B. Hong and B. Fultz, J. Appl. Phys. 79, 3946 (1996).
13. H. P. Klug and L. E. Alexander, X-ray Diffraction Procedures (Wiley-Interscience,
New York, 1974) p. 594

188
14. G. Le Caer and J. M. Dubois, J. Phys. E: Sci. Instrum. 12, 1083 (1979).
15. P. Eckerlin, H. Kandler, Landolt-Bornstein Numerical Data and Functional
Relationships in Science and Technology, Vol. 6 Structure Data of Elements and
Intermetallic phases, Spring-Verlag, New York 1971
16. T. Shinjo and T. Takada in Industrial Applications of the Mossbauer Effect, G. J.
Long and J. G. Stevens, eds., (plenum, New York, 1986), p. 739.
l7. K. Kawaguchi, R. Yamamoto, N. Hosoito, T. Shinjo, and T. Takada, J. Phys. Soc.
Japan 55,2375 (1986).
18. C. Bansal, Z. Q. Gao, L. B. Hong, and B. Fultz, J. Appl. Phys. 76,5961 (1994).
19. E. T. Teatum, K. A. Gschneider, Jr., and J. T. Waber, Compilation of Calculated
Data Useful in Predicting Metallurgical Behavior of the Elements in Binary Alloy
Systems, Los Alamos Laboratory Report LA-4003 (1968).

189

Appendix B

Kinetics of Hydrogen Diffusion in LaNis_xSnxAlloys

B.I Introduction
The durability of LaNis-based electrodes during electrochemical cycling is
generally improved by a partial substitution of La or Ni with suitable solutes. Various
additives such as Nd [1], Ti [2], Zr[3] and Ce [4], for La and Co[I], Mn, Al [5] and Si [1]
for Ni have been found to be successful substituents for lowering the absorption plateau
pressures and/or improving the cycle life. Sakai et al. [5] performed a rigorous evaluation
of several other elements as ternary solutes in LaNis_xM x. The cycle life was found to
increase in the order Mn < Ni < Cu < Cr < Al < Co. In all the above ternary alloys, the
improvement in the cycle lifetime is unfortunately accompanied by a decrease in the
hydrogen absorption capacity, long activation, or slow kinetics. The use of Sn as a partial
substituent for Ni in LaNi s' on the other hand, was found to reduce the absorption plateau
pressure and minimize the hysterisis, while retaining most of the high absorption capacity
of the binary alloy [6]. Furthermore, the Sn substituent was found to result in a 20-fold
increase in the cyclic lifetime in thermal cycling [7] and a charge-discharge cycle life
comparable to a multi -component, misch metal based alloy [8]. Additionally, the kinetics
of hydrogen absorption-desorption appear to be more facile upon Sn substitution,
indicating that favorable surface conditions are prevalent on these alloys [9]. Similar
beneficial effects were realized with Ge substituent also [10].
The mechanism of electrochemical hydriding-dehydriding contains a series of
steps. For example, the charge transfer reaction produces adsorbed atomic hydrogen
(Had) and OR on the electrode surface (Volmer process). The adsorbed hydrogen either

diffuses through the surface layer and grain boundaries into the bulk of the alloy to form
the hydride or combines with adjacent adsorbed hydrogen atom to form molecular

190
hydrogen (Tafel process) and thus hydrogen evolution. The performance of the metal
hydride electrode is determined by the kinetics of the process occurring at the
electrode/electrolyte interface as well as of hydrogen transport within the bulk of the
alloy. The charge transfer process is the rate-determining step for electrodes containing
small particles, while the hydrogen diffusion dominates for larger particles [11].
Reliable values for the hydrogen diffusion coefficients in the LaNi 5 hydride
phases have been difficult to obtain due to complex microscopic diffusion processes [12,
13] in crystal structures where hydrogen simultaneously occupies several distinct types of
interstitial sites. Furthermore, various physical properties of activated LaNi 5H x powders
often impeded analyses and interpretations of commonly employed techniques such as
nuclear magnetic resonance (NMR) and quasielastic neutron scattering (QNS).
Performing a critical assessment of NMR and QNS methods to characterize hydrogen
diffusion in ~-LaNi 5H x when x > 6, Richter, et al. [12] concluded that the long-range
-8

diffusion coefficient at 300 K is in the range 1 - 5 x 10 cm Is. Ziichner, et al. [14]
applied the current pulse electrochemical method on single crystal a-LaNi 5H.x with x <
-8

0.07 to obtain anisotropic diffusion coefficients of 2 - 3 x 10 cm /s at 298 K.
Apparently, there is not a very large difference in hydrogen diffusion behavior between
these two phases even though the hydrogen contents vary considerably.
There have been relatively few studies [12] of the effect of substitutional alloying
on hydrogen diffusion in the AB 5 hydrides. Using NMR methods, Bowman, et al. [15]
showed that Al substitution greatly decreased hydrogen motion in the B-LaNi5-y Aly Hx with
an accompanying increase in activation energy. Zheng, et al. [16] obtained with
-II

electrochemical methods a room temperature diffusion coefficient of 3 x 10 cm /s for an

191
10

LaNi 4.25 Al 0.75 hydride electrode which was much below the value of 7 x 10- cm 1s
reported for a LaNi 4 Cu electrode by van Rijswick [11]. More recently, Zheng, et al. [17]
used a constant current discharge technique to derive a hydrogen diffusion constant of 7 x
-II

10 cm Is from an LaNi 4.]7 Sn0.24 electrode. Disc-type electrodes made with
MmNi 4.2 Al 0.5 M 0.3 (M = Cr, Mn, Fe, Co, Ni) were used by Iwakura, et al. [18] to evaluate
hydrogen diffusion coefficients in the a-phase by the electrochemical potential step
-8

method. The diffusion constants measured at 303 K ranged between 1.6 x 10 cm Is to
-8

3.2 x 10 cm Is. It should be noted that the hydrogen diffusion coefficients deduced
using electrochemical methods on alloy powders are smaller by one-to-two orders-ofmagnitude than values obtained from NMR, QNS, or electrochemical measurements on
bulk samples. [14, 18]
In this work, we study the transport kinetics of hydrogen in LaNi 5_xSnx alloys
electrochemical pulse techniques, i.e., chronoamperometry and chronocoulometry. Sn
substitution results in improved interfacial conditions for electrochemical hydridingdehydriding processes and in lower absorption pressures due to enlarged lattice volume.
It is interesting to see if these features, especially the latter, also lead to enhanced
transport of hydrogen with in the bulk of the alloy. The results from these studies,
combined with the extensive electrochemical and structural characterization already
performed on LaNis_xSn x alloys, will help us better understand the role of the ternary
solute on hydrogen diffusion.

B.2 Experimental

192
The LaNi 5_x Snx alloys were prepared by induction - melting and were subsequently
annealed in evacuated quartz ampuls at 950°C. The annealed ingots were subjected to
five hydrogen decrepitation cycles to activate the alloys. Metal hydride disk electrodes
were prepared by filling the BAS (Bioanalytical Systems) disk electrodes with the
mixture of MH powders (with 20% Ni and 5 % PTFE), of equal quantities in each case to
ensure equal surface area (0.07 cm"), charge density (5.6 g/cm\ and porosity in all the
electrodes. A NiOOH electrode formed the counter-electrode, and a Hg/HgO (0.098 V
vs. SHE) with a Luggin capillary served as the reference electrode in a three-electrode
flooded cell with 31 w% KOH electrolyte.
The scheme of experimentation includes constant current charges at 40 mAlcm

(20 rnA/g) to a capacity corresponding to 400 mAh/g, potentiodynarnic polarization
curves for estimating diffusion limiting currents, constant - current discharges at 64

mAlcm (33 mAlg) to -0.5 V vs. Hg/HgO for calculating the absorbed hydrogen
concentration in the alloy and chronoamperometric and chronocoulometric transient
measurements for determining the diffusion coefficients. The polarization experiments
were carried out with an EG&G 273 Potentiostat / Galvanostat using 352 corrosion
software. Chronocoulometric response was, however, monitored with a Nicolet storage
oscilloscope. Temperature measurements were conducted in a Tenney environmental
chamber, ±.O.l dc.
It may be difficult to distinguish the hydriding and hydrogen evolution processes
electrochemically. In a potentiodynamic polarization, for example, a smooth, almost
unnoticeable transition occurs from the hydride formation to the hydrogen evolution [2].
The concurrent hydrogen evolution thus induces uncertainties in the analysis of cathodic
polarization data, due to reduced current efficiency and changing surface conditions. In

193
order to alleviate such uncertainties, the present studies are restricted to the anodic
regime, both for the transient and steady state experiments. Besides, the slow diffusing
species during discharge is undoubtedly hydrogen within the bulk of the alloy, where as
in the charging process, hydroxyl ions in the electrolyte phase could move more
sluggishly, depending on the porosity and tortuosity of the electrode [17].

B.3 Results
B.3.1 Steady State Measurements
B.3.1.1 Potentiodynamic Polarization - Limiting currents
Figure B.1 shows the steady state polarization curves ofLaN~_xSnx alloys with
different Sn contents (x from 0 to 0.5) at a slow scan rate (0.5 mV/s) approximating
steady state conditions. The polarization curves show strong interference of mass transfer
processes on the charge transfer kinetics, as evident from the current being invariant with
an increase in the overpotential. At high overpotentials, ~ 400 m V, the discharge reaction
is limited by the rate of hydrogen transport within the bulk of the alloy. The
corresponding current, termed as diffusion limiting current and estimated from Figure
B.1 increases upon Sn substitution and decrease later at x ~ 0.2 (Figure B.2). The lowlimiting current of the binary alloy is due to the difficulty associated with its charging in a
partially-sealed cell. The limiting currents of the Sn-substituted alloys are - 500 mA/g) as
reported earlier for LaNi 5 •

B.3.1.2 Discharge Characteristics

194

-450
0.2

0eJ) -550

-.
==
==

e J)

-650
0.4

rI.l

>e
'"

-750
-850
-950

100

i, mA/g

Figure B.l. Potentiodynamic polarization curves of LaNi 5 _xSnx alloys, illustrating
diffusion-limited behavior.

1000

195

600

500
........
...- 400

t: 300
U=
QJ

e 200

:.:3 100

0.1

0.2

0.3

x in LaNis-xSnx

Figure B.2. Diffusion limiting currents of LaNis-xSn x alloys, estimated from
potentiodynamic polarization curves.

0.4

196
In order to obtain the value of hydrogen concentration required for calculating the
diffusion coefficients from the transient response below, the electrodes were discharged
after a prior, complete charge. Figure B.3 shows the discharge curves of LaNis_xSn, alloys.
The discrepancy in the electrochemical capacity in these alloys is mainly due to the
differences in their absorption plateau pressures [9]. MH alloys with high plateau
pressures (> 1 atm.) are not efficiently charged in the open cell adopted for these studies.
At high Sn contents, on the other hand, the discharge kinetics are relatively sluggish. The
capacity is thus maximum with a Sn content of 0.2-0.3.
It should be realized that the degree of hydriding is such that the MH alloy exists in
the two-phase region, i.e., in the a and ~ form. The measured diffusion coefficients
therefore represent an average value of a and ~ phase hydrides.

B.3.2 Transient Measurements
B.3.2.1 Chronoamperometry
The transient methods involve the application of a potentiostatic pulse
(corresponding to the mass transfer regime identified in the steady-state experiment) to
the electrode and monitoring the amperometric and coulometric responses. From the
dependence of the diffusion current and coulombic charge on time, it is possible to obtain
values for the diffusion coefficients, by applying appropriate diffusional equations. The
boundary conditions applicable to the present case for the outward diffusion of hydrogen
within the bulk of the alloy are i) uniform initial concentration, i.e., at t = 0, the
concentration of hydrogen is the same at any x, the distance from the interface ii) at t >
0, the concentration at large x approaches the bulk concentration and iii) the interfacial

197

1.0

0.8

Sn 0.2

,;,

.•-....

Sn 0.1

0.6

....=

=--,;,

.:.

Sn 0.3

0.4

Sn 0.4

0.2

0.0

100

150

200

Capacity, mAh/g

Figure B.3. Discharge curves of LaNi 5-xSnx alloys

250

300

198
hydrogen concentration is zero at t > 0, after the potential step. Assuming semi -infinite
linear diffusion conditions, the instantaneous current in the diffusion-limited regime may
l9

be expressed by Cottrell's equation as
11 FA

DH I12

c·

(1)

n I12 ,112

where i is the instantaneous current at time t, DH is the diffusion coefficient of hydrogen,

c* is the bulk concentration of the diffusing species, A is the area of the electrode and F
is the Faraday constant.
Figure B.4 shows the chronoarnperometric curves of LaNi s_5n, alloys on applying
a potentiostatic pulse of +400 mV vs. OCV, which corresponds to the diffusioncontrolled regime. From the slope of these curves in Figure B.4 and the concentrations
for the absorbed hydrogen obtained from the electrochemical discharge capacities using
geometric volume of the electrode (Figure B.3 and Table B.l), the diffusion coefficients
for hydrogen in LaNis_,Sn, alloys were calculated (Table B.1). The diffusion coefficients

thus calculated are 6.69 x 10- ,8.38 X 10-9 , 7.53 X 10- and 9.36 x 10- cm /s for Sn contents
of 0.1, 0.2, 0.3 and 0.4, respectively (Figure B.6, Table B.1).

B.3.2.2 Chronocoulometry

Here also, the electrode potential is shifted to a sufficiently positive value (for
oxidation) to enforce diffusion-limited conditions_ The coulometric output from the
Potentiostat is recorded as a function of time. The coulometric response may be
describes by the Cottrell equation integral form 19] such as

199

65
60

Sn : 0.2

Sn : 0.4

u=
;.:s

35
30

0.4

0.8

1.2
-J

t , sec

1.6

-J

Figure B.4: Analyses of chronoamperometric curves of LaNis_x Snx alloys.

2.4

200

2 1lFA D H I12 Cl 11!'
K I n:

(2)

where Q is the cumulative charge passed at any instant, t. A plot of Q vs. t !2 is thus
linear, the slope of which yields the diffusion coefficient.
Chronocoulometlic curves were obtained for LaNis_,Snxalloys at a potentiostatic pulse
cOITesponding to an anodic perturbation of 400 m V (Figure B.S). From the slope of these
curves in Figure B.S and the concentrations for the absorbed hydrogen obtained from the
electrochemical discharge capacities using geometric volume of the electrode (Figure B.3
and Table B.I), the diffusion coefficients for hydrogen in LaNis_xSnx alloys were

calculated (Table B.I). The diffusion coefficients thus calculated are 3.49 x 10- ,3.86 x

10-9 and 3.77 x 10-9 cm1/s for Sn contents of 0.2, 0.3 and 0.4, respectively (Figure B.6,
Table B.l).

B.4 Temperature Studies
Temperature studies were preformed on the LaNi 4s Sno.2 alloy to determine its
effective activation energy. The LaNi 4 SSn 02 alloy was chosen for its supelior cycle-life
[9] . Figures B.7 and B.8 display the chronoamperometlic and chronocoulmetric curves
respectively.

Both figures show the expected decrease in kinetics with lower

temperatures.

The diffusion coefficients calculated from Figures B.7 and B.8 are

displayed on an Arrhenius plot in Figure B.9. The activation energy of hydrogen in
LaNi 4 8Sno1 is calculated to be 0.19 eV by chronocoulometry and 0.29 eV by
chronoamperometry. The dispality in activation energy between the two measurements
is a result of the composite nature of the measured electrode (binder, conductive diluent,

201

0.3

Cf)

..0

0.4

0.2

3.5

..-

;::::!

0.1

0.5

1.5

2.5

t, sec

Figure B.S. Analyses of chronocoulometric response of LaNis_xSnx alloys.

202

l.OOE-08

C'-l

"'6
C.J

...:-

.-=

C.J

Chronoamperometry

9.00E-09

8.00E-09

7.00E-09
6.00E-09

....=

C'-l

::s

.-~

5.00E-09
4.00E-09

3.00E-09

Chronocoulometry

2.00E-09
1.00E-09

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

X in LaNis-xSnx

Figure B.6. Variation of diffusion coefficient of hydrogen in Sn-modified LaNis alloys

with the Sn content

203

60
40e

« 40
.-

30e

25C

-10 e

0.20

0.25

0.30

0.35

0.40

0.45

0.50

Figure B.7. Analyses of chronoamperometric curves ofLaNi . SIlo.2 alloy at various
4s
temperatures.

204

150
()

e> 100
co

..c
()

2.0

2.5

4.0

3.5

3.0

1/2

, sec

4.5

5.0

112

Figure B.S. Analyses of chronocoulmetric curves of LaNi4 .SSnO.2 alloy at various
temperatures.

205

(.)

10-8 I-

Chronoamperometry

Chronocoulometry

3.2

3.4

3.6

-3

3.8xl0

11T, 11K

Figure B.9. Arrhenius plot of LaNi4 .8 Sno.2 temperature. Activation energies calculated
from plot slopes are 0.19 eV by chronocoulometry (.A.) and 0.29 eV by
chronoamerpometry (e).

206
polycrystalline alloy). Thus an "effective" activation energy is measured which averages
over the various cyrstall orientations and the non-Faradic impedances. These results are
in good agreement with NMR studies by Bowman et al. [20].

B.S Conclusion

The diffusion coefficients for the diffusion of hydrogen, obtained from the
amperometric and coulometric outputs upon a potential pulse corresponding to the
diffusion limiting conditions, are in close agreement with each other, suggesting that the
analyses and the assumptions involved are reasonable. The actual values of the diffusion
coefficients are marginally lower than those obtained from NMR and QNS techniques,
but clearly larger than the values of Zheng et aI, obtained from constant-current
discharges.
Finally, with increasing Sn content in LaNis_xSn x alloys, the diffusion of hydrogen is
not hindered, whereas some sluggishness was observed in the charge transfer kinetics.
On the other hand, the diffusion of hydrogen seems to be enhanced marginally upon the
substitution of Sn for Ni. Measured "effective" activation energies correspond to
published NMR studies

207

Table B.1 : Determination of diffusion coefficients by chronoamperometric and chronocoulometric methods

x in LaNi5_,Snx

Capacity

Concen., C H

mAh/g

Chronoamp.

Chronocoul.

m. moles/cc

m A .s -II"

mCoul.s

DH Chroncoul)

cm"/s

6.69 x 10-9

x 10-9

18.8

8.38 x 10-9

3.49 X 10-9

11.9

18.6

7.53 x 10-9

3.86 X 10-9

14.4

18.2

9.36 x 10-9

3.77 X 10-9

0.1

141

0.467

6.2

0.2

285

0.978

14.5

0.3

253

0.846

0.4

271

0.913

- -

ll2

DH (Chronamp)

208
B.7 References
1. G. G. Willems, Philips J. Res., 39 (Suppl. 1), 1 (1984); J. G. G. Willems and

K. H.

J. Buschow, J. Less Common Metals, 129,13 (1987).
2. T. Sakai, H. Miyamura, N. Kuriyama, A. Kato, K. Oguru, and H. Ishikawa, J. Less

Common Metals, 159, 127 (1990).
3. T. Sakai, H. Miyamura, N. Kuriyama, A. Kato, K. Oguru, and H. Ishikawa, J

Electrochemical Soc., 137,795 (1990).
4. T. Sakai, T. Hazama, H. Miyamura, N. Kuriyama, A. Kato, and H. Ishikawa, J.

Less Common Metals, 172-174, 1175 (1991).
5. T. Sakai, K. Oguru, H. Miyamura, N. Kuriyama, A. Kato, and H. Ishikawa, J. Less

Common Metals, 161, 193 (1990).
6. M. H. Mendelsohn, D. M. Gruen, and A. E. Dwight, Mat. Res. Bull, 13, 1221 (1979).
7. R. C. Bowman, Jr., C. H. Luo, C. C. Ahn, C. K. Witham, and B. Fultz, J. Alloys and

Compounds, 217. 185 (1995).
8. B. V. Ratnakumar, G. Halpert, C. Witham and B. Fultz, J. Electrochem. Soc., 141,
L89(1994); Proc. ECS Symp. Hydrogen Batteries,. 94-27, 57(1994).
1. B. V. Ratnakumar, C. Witham, R. C. Bowman, Jr., A. Hightower, and B. Fultz, J.

Electrochem. Soc. 143,2578 (1996).
10. C. Witham, B. V. Ratnakumar, R. C. Bowman, Jr., A. Hightower, and B. Fultz, J.

Electrochem. Soc. 144, (1996)'11. M. H. J. Van Rijswick, in 'Hydrides for Energy Storage', Ed., A. F. Anderssen and
A. J. Maeland, p. 261, Permagon, Oxford (1978).
12. D. Richter, R. Hempelmann, and R. C. Bowman, Jr., in_Hydrogen in Intermetallic

CompooundsJ1. Ed L. Schlapbach (Springer-Verlag, Berlin, 1992) p. 97.

209
13. C. Schonfeld, R. Hempelmann, D. Richter, T. Springer, A. J. Dianoux, J. J. Rush, T.
J. Udovic, and S. M. Bennington, Phys. Rev. B 50,853 (1994).
14. H. Ziichner, T. Rauf, and R. Hempelmann, J. Less-Common Met. 172-174,611
(1991).
15. R. C. Bowman, Jr., D. M. Gruen, and M. H. Mendelsohn, Solid State Commun. 32,
501 (1979).
16. G. Zheng, B. N. Popov, and R. E. White, J. Electrochem. Soc. 142,2695 (1995).
17. G. Zheng, B. N. Popov, and R. E. White, J. Electrochem. Soc. 143, 834 (1996).
18. C. Iwakura, T. Oura, H. Inoue, M. Matsuoka, Y. Yamamoto, J. Electroanal. Chern.
398, 37 (1995).
19. A. J. Bard and L. R. Faulkner, "Electrochemical Methods; Fundamentals and
Applications", John Wiley & Sons, Inc., New York (1980).

20. G. Majer, U. Kaess, R. C. Bowman Jr., Phys. Rev. B, 57, pp. 599, (1998)

Appendix C
C.1 Introduction

Performance of LaNi4.7SnO.3 Sealed Cells

210
Since the mid-1980's, studies on metal hydrides as anodes in nickel metal hydride
(NiMH) batteries have moved from fundamental scientific investigations to productbased industrial efforts. Metal hydrides (MH) now have widespread application as
negative electrodes in rechargeable batteries for the consumer electronics industry. The
present investigation focused on electrochemical properties of metal hydride alloys based
on LaNiS, denoted generically as "ABS" alloys.
Commercial ABS metal hydride alloys in rechargeable batteries include
substantial substitutions of other elements for La and for Ni. To reduce cost and improve
cycle life, La is replaced with mischmetal ( Mm ) which is a reduced ore of a variety of
rare earth elements. Sakai, et al. [1] studied various ternary substitutions for Ni in LaNiS,
and reported that the cycle life improves with the ternary substituents in the order Mn <
Ni < Cu < Cr < Al < Co. The beneficial effects of Co have led to large Co substitutions
for Ni in commercial ABS alloys. Higher costs and limited availability of the strategic
element Co make it worthwhile to investigate alternate elements for substitution with Ni.
Sn has proven itself as a viable candidate for promoting cycle life while retaining cell
capacity in laboratory tests.
The characteristics of LaNi4.7Sn0.3 have been well documented in laboratory
cells. Previous studies by the Caltech /JPL group on LaNis substituted with 0 - 8.3 at.%
Sn for Ni revealed trends between alloy performance and Sn composition [2,3]. Kinetics,
thermodynamics, and cycle life of LaNiS.O-xSnx (x = 0 - O.S) were determined by
electrochemical and gas-phase studies. The microstructures of these alloys were
determined by X-ray diffractometry and TEM [4-6]. Compared to other ternary alloys,
LaNiS.O-xMx (M = AI, Ga, In, Si), Sn-containing alloys showed the longest cycle life.
The optimal concentration of Sn, with respect to a balance among cycle life, capacity and

211

kinetics, was found to be in the range of LaNi5.0-xSnx, x =0.25 - 0.3. The reason for the
improved cycle life may be related to the strong chemical bonding between La and Sn
[7].

These previous results have prompted us to investigate the performance of Snsubstituted alloys in sealed cells. Here we report results of studies on the kinetics,
capacity, and cycle life of sealed cells of LaNi4.7Sno.3. The performance of these alloys,
as-cast and after annealing, is compared to a standard commercial AB5 alloy.

C.2 Experimental
The AB5 alloys used in this study were produced by vacuum induction melting of
elemental raw materials of commercial purity (99+%). The Sn-containing alloy was
melted as a 6 kg heat. Half of the ingot was annealed for 72 hours at 950°C in an argon
atmosphere. The partial ingots were then mechanically crushed to < 75 microns powder.
The oxygen content of the powder obtained from the heat-treated ingot was 0.07 wt.%.
The chemical composition as determined by Induction Coupled Plasma (ICP) Atomic
Emission Spectroscopy was LaNi4.58SnO.3, or just slightly sub-stoichiometric. The
commercial AB5 alloy used as the control has the nominal composition
MmNi3.6CoO.7AI0.4MnO.3, where Mm is La-rich mischmetal, a mixture of light rare
earth elements, with the approximate atomic ratio of LaO.53Ce0.32PrO.04NdO.l1. The
heat size was in excess of 300 kg and the ingots were mechanically crushed to < 75
microns powder. Chemical control is ±0.2 wt. % for each of the B-side elements (Co, Al
and Mn).

212
AA NiMH cells were assembled at Energizer by winding thin planar Ni(OH)2
sinter-type electrodes and pasted MH electrodes between a layer of battery -grade nylon
separator. The MH electrodes used in this study were fabricated using pulverized, < 75
microns, MH alloy powder, admixed with a small «2% ) amount of binder and solvent,
applied to thin nickel-plated perforated steel substrate, dried, calendared and sized.
Approximately 7 g of MH alloy was used per electrode. The same alloy weight was used
for all MH compositions. Electrolyte consisting of KOH with small additions of NaOH
and LiOH was added prior to cell sealing. The cell capacity was set by the positive
electrode at approximately 1000 rnAh for all of the cells assembled.
Cycling of the AA NiMH cells at Energizer was performed on automated battery
cycling equipment. Cycling conditions were 1C rate (1000 rnA) charge to a negative
change in voltage ("negative delta V") or until 38°C was reached (which ever comes
first), followed by a ClIO charge rate (approximately 2 hours) to ensure complete cell
charging and to fill a 3 hour test time window. Cells rested 2 minutes after charge and
then discharged at lC rate to 0.9 V. Following discharge, cells were given a 1 hour rest
before recharge.
The cycling of cells at JPL was carried out with an automatic battery cycler made
by Arbin Corp., College Station, TX. The cycling conditions included a discharge step at
1C rate to 0.9 V. After a two minute rest period, charging was done in two steps. Initial
charging at I C rate was performed for a maximum of one hour or to 1.6 V followed by
charge step at C/lO rate for half an hour. This procedure gave a charge/discharge ratio of
1.25 or a cutoff voltage of 1.6 V, whichever came first.
Gas pressure and analysis were performed on laboratory equipment that punctures
a cell, releases cell gases into a pre-measured volume, measures pressure using a Dynisco

213
pressure transducer, calculates internal cell pressure and sends a sample of the captured
gas to a Gow-Mac Gas Chromatograph (Series 500) for analysis. Gas pressure and
composition were measured immediately after removing the cells from charge.
Self discharge rates were carried out at JPL using the same charge and discharge
conditions described above for cell cycling. Annealed LaNi4.7Sn0.3 and
Mm(NiAICoMn)5 cells were subjected to five cycles to establish a baseline capacity.
The cells were then allowed to sit at 25°C and 45°C for several days. Their discharge
capacities were then measured and normalized to initial capacity.
AC impedance measurements were carried out with the EG&G 273 Potentiostat
and Solartron 1255 Frequency Response Analyzer, using EG&G Impedance software
388.

C.3 Results and Discussion
C.3.1 Isotherms
Figure C.l displays isotherm data of gas-phase desorption at 45°C of a typical
Mm(NiAICoMn)5 and LaNi4.7Sn0.3 in the as-cast and annealed conditions. The
isotherm of the annealed Sn alloy displays a flatter plateau and larger capacity relative to
the Mm(NiAICoMn)5. The upward slope of the isotherm for the as-cast Sn alloy can be
attributed to the chemical inhomogeneity of the material. An inhomogeneous distribution

214

1000

,.. •

100
·u . .:>

.'"t. (

...

.-I

1\,... ,.. ..... ,...+

ro
,~

..
a..

......

10
./

-,'"

~ ---~" " /'"

... -",'
.,'

~,&,'

... " " ' ' -

---... , "'47'-" '[ .:-1
A ...... "",..I"",J

F'" .' ''VU''V,""

'J :J

X, HI ABs formula units

Figure c.1. Gas phase desorption isothenns of the as-cast LaNi4.7SnO.3 alloy, annealed
LaNi4.7SnO.3, and the commercial alloy of typical Mm(NiAICoMn)5 composition
measured at 45°C.
of Sn in the as-cast alloy wi11lead to some regions favorable for hydrogen occupancy but
other regions less so.

215

C.3.2 Self Discharge
A correlation between hydride desorption pressure and self discharge ratios have
been reported by Anani, et al [8]. A lower self-discharge for annealed LaNi4.7SnO.3
alloys was suggested by the gas phase isotherms. The lower H2 pressure at greater
hydrogen to metal hydride ratios (Hlf.u. > 3) indicates the greater stability of hydrogen in
the LaNi4.7SnO.3 alloys. The flat isotherms translate into smaller gradients in the
hydrogen chemical potential and thus reduced hydrogen evolution. Results of self
discharge tests are shown in Figure C.2 As expected, we consistently find lower self
discharge in annealed LaNi4.7SnO.3 sealed cells (AA - size) as compared to
Mm(NiAICoMn)5 cells. The degree of self discharge increases with temperature, with
the LaNi4.7SnO.3 being less sensitive to this effect. Self discharge can also be influenced
by the by-products of nylon separator degradation and nitrates remaining in the positive
Ni electrode. However, these cells were fabricated with identical procedures and thus the
differences in self-discharge shown in Figure C.2 reflect the nature of the MH electrode
material.

C.3.3 Rate Measurements
It is difficult to infer negative electrode kinetics from the performance

measurements of commercial sealed cells. This is because the cells are purposely built
with an excess of negative electrode (i.e., positive limited to reduce H2 gassing

216

70.0
Mm(Ni,Co,AI,Mn)5

~ 60.0

Cl)

+-'

50.0

Cl)
C')

40.0

c::

LaNi4.7SnO.3

co
.r::.

30.0

20.0

Mm(Ni,Co,AI,Mn) 5

Cl)

CJ')

25 C

10.0
0.0

Aging, days

Figure C.2. Self discharge data showing loss in stored electrochemical capacity versus
time for annealed LaNi4.7SnO.3 alloy and commercial alloy of typical Mm(NiAICoMn)5
composition measured at 25°C and 45°C.

217

problems). Cell capacity variations are shown in Figure C.3 as a function of discharge
current up to 3C. Groups of 2 to S cells were tested to obtain these results. The higher
capacity evidenced for the LaNi4.7SnO.3 at all discharge rates, may be indicative of
improved kinetics for this alloy. Nevertheless, the difference in rate capacity of the
LaNi4.7SnO.3 and Mm(NiAICoMn)s electrodes is not substantial.

C.3.4 Pressure Measurements
Pressures of gases evolved from the charging of the sealed cells at IC rate were
measured. There were similar amounts of total gases evolved from all cells. Figure C.4
shows that the amount of H2 gas evolved from each cell increases with the time of
charge. The reduced hydrogen emission of the annealed Sn alloy follows from the
reduction in plateau pressure. The as-cast LaNi4.7SnO.3, with its more inhomogeneous
Sn distribution, showed greater hydrogen evolution.

C.3.S Cycle Life
Cycle life tests on sealed cells of as-cast and annealed LaNi4.7SnO.3 are
compared to Mm(NiAICoMn)s control cells in Figure c.S and C.6. These experiments
were conducted at JPL and Energizer under different cycling conditions and provided the
data for Figure C.S and C.6, respectively. The as-cast and annealed LaNi4.7SnO.3 cells
exhibit shorter activation times and more rapid degradation compared to control cells in
both series of measurements.
The Mm(NiAICoMn)s alloys display superior cycle life to the LaNi4.7Sn0.3
alloys. It is widely accepted [9] that the substitution of Co and Al for Ni inhibit volume

218

, ., 5
D Annealed Sn

, .'0

D As-cast Sn

o Control

.-----

I-I--

.r:.
.05
...

I-

.----I--

+""

I--

' .00

co
c.
co
U 0.95

0.90

I-

r--I-r--

.-----

I-

0.85

C/5

Discharge Rate

Figure C.3. Electrochemical capacities of AA cells measured at different rates of

discharge.

219

eo
en

LaNi4.7SnO.3
As-cast

::J:
CI)

20
Mm(NiCoAIMn)5

L-

::;,

en
en

,'"

CI)

L-

a..
eo
......
L-

eo
a..

LaNi4.7SnO.3
Annealed

Charge time, min

Figure CA. Partial pressure of H2 gas evolved from cells during different charging times

at Ie rate.

220

1 .2

.. •

• Mn(NiCoAIMn)5

1.0

.0
-.8

.r:::.

«. 0.8
>..

.f..J

ctJ
0.
ctJ

0.6

0.4

LaNi4.7SnO.3
Annealed
0.2

LaNi4.7SnO.3
As-cast

0.0

200

400

600

Cycle Number

Figure C.5. Cycle life for sealed cells containing as-cast LaNi4.7SnO.3 alloy, annealed

LaNi4.7SnO.3, and the commercial alloy of typical Mm(NiAICoMn)s composition
measured at room temperature. Two cells of each were measured at JPL.

221

1 .2

~.............~--..~~Mn(NiCOAIMn)5

1 .0

«. 0.8
>.

+""

LaNi4.7SnO.3
Annealed

0.6

ct'J

c.
ct'J

0.4

0.2

LaNi4.7SnO.3
As-cast

0.0

200

400

600

Cycle

Figure C.6. Cycle life data for sealed cells containing as-cast LaNi4.7SnO.3 alloy,
annealed LaNi4.7SnO.3, and the commercial alloy of typical Mm(NiAICoMn)s
composition. Measurements were carried out at Energizer.

222

140
'I-

LaNi4.7SnO.3
annealed

120

...

co co

+'"'

100

:::J

CI)

CI)
G)

c:: +'"'

L...
G)

+'"'

Mm(Ni,Co,AI,Mn)S

100

200

300

400

Cycle Number

Figure C.7. Cell internal resistances vs. cycle number determined by AC impedance
studies on annealed LaNi4.7SnO.3 and the commercial Mm(NiAICoMn)5.

223
dilatation during hydrogen absorption, and thereby improves cycle life. It has also been
proposed that the Ce [10] and Nd [9] present in the mischmetal promotes beneficial
passivating films. These passivating films protect the rare earth elements from the
corrosive electrolyte. We suggest that the lifetimes of LaNi4.7SnO.3 could be improved
with the substitution of mischmetal for La. Encouraging results from Ce additions have
been reported previously [11].

C.3.6 A C Impedance

The LaNi4.7SnO.3 and Mm(NiAICoMn)s sealed cells differ considerably in their
series Ohmic resistance during cycling, as shown in Figure C.7. The Ohmic resistance of
sealed cells, generated from the interfaces between electrode, electrolyte, and leads, was
measured by AC impedance spectrometry. The initial decrease seen in this parameter
could be due to the decrepitation of the metal hydride alloy particles, which results in an
enhanced surface area (and thus reduced resistance). The subsequent increase in the
resistance may be an indication of the corrosion and passivation processes, which might
be responsible for the shorter cycle life of the cells with Sn-based alloy.

C.4 Conclusions
Our studies have found LaNi4.7SnO.3 alloys to possess limited utility in
rechargeable nickel metal hydride batteries. The enhancements relative to
Mm(NiAICoMn)s in kinetics and reduction in self discharge observed in laboratory cells
have been demonstrated in sealed cells. The LaNi4.7SnO.3 cells exhibited lower self
discharge and greater capacities at higher rates of discharge compared to cells of the
standard mischmetal composition Mm(NiAICoMn)s. The LaNi4.7SnO.3 cells did not

224
display cyclic lifetimes comparable to the Mm(NiAICoMn)s cells. Hence, the
enhancements in lifetime from Sn substitution alone, which were clearly demonstrated in
previous half-cell electrode studies [3,7] and attributed to the strong chemical bonding
between La and Sn, was not found to be as effective as the combined contributions in
commercial Mm(NiAICoMn)s alloys. The next logical step is to investigate
improvements to the lifetimes of the LaNi4.7Sno.3, by substituting Ce for a portion of the
La. We expect that the surface effects from passivating films and changes to the alloy
volume expansion on hydriding [11] will improve cycle life for reasons independent of
the beneficial effects of Sn. However, substitution of Ce may reduce capacity through
formation of other impurity phases. This behavior will be the subject of future studies.

225

c.s References
[1]

T. Sakai, K. Oguru, H. Miyamura, N. Kuriyama, A. Kato and H. Ishikawa, J. Less

Common Metals, 161, 193 (1990).

[2]

R.c. Bowman, Jr., C.H. Luo, C.C. Ahn, c.K. Witham and B. Fultz, J. Alloys and

Compounds, 217, 185 (1995).

[3]

B.V. Ratnakumar, C. Witham, R. C. Bowman, Jr., A. Hightower, and B. Fultz, 1.

0/ Electrochem. Soc., 143,2578 (1996).
[4]

M. Mendelsohn, D. Gruen, A. Dwight, Inorg. Chem. 18,3343 (1979).

[5]

J. S. Cantrell, T.A. Beiter and R.C. Bowman, Jr., J. Alloys and Compounds,

207/208,372 (1994).

[6]

M. Wasz, R.B. Schwarz, S. Srinivasan and M.P.S. Kumar, in Materials/or

Electrochemical Energy Storage and Conversion - Batteries, Capacitors, and Fuel Celis,

Symposium Proceedings of The Materials Research Society, San Fransico, CA Vol. 93,
pp. 237-242, Spring (1995).
[7]

C. Witham, R. C. Bowman, Jr., B.V. Ratnakumar, and B. Fultz, in ABS Metal

Hydride Alloys/or Alkaline Rechargeable Cells., Proceedings of The Eleventh Annual

Battery Conference on Applications and Advances, Long Beach, CA (Inst. Electrical
Electronic Engs., Piscataway, NJ), (1996), pp. 129-134.
[8]

A. Anani A. Visintin, S. Srinivasan, A. J. Appleby, J. J. Reilly, and J. R. Johnson

Electrochem. Soc. Proc. 92-5, p. 105 (1992).
[9]

J. J. G. Willems, Philips J. Res., 39 (Supp!. 1), 1 (1984); J.J.G. Willems and K.H.

J. Buschow, J. Less Common Metals, 129, 13 (1987).
[10] T. Sakai, T. Hazama, H. Miyamura, N. Kuriyama, A. Kato and H. Ishikawa, J. Less
Common Metals, 172-174, 1175 (1991).

226
[11]

G. D. Adzic, J. R. Johnson, J. J. Reilly, J. McBreen, S Mukerjee, M. P. Sridlar

Kumar, W Zhang, S. Srinivasan, J. Electrochem. Soc. 142,3429 (1995).