Die approbierte Originalversion dieser Diplom-/ Masterarbeit ist in der Hauptbibliothek der Tech- nischen Universität Wien aufgestellt und zugänglich. http://www.ub.tuwien.ac.at The approved original version of this diploma or master thesis is available at the main library of the Vienna University of Technology. http://www.ub.tuwien.ac.at/eng
76
Embed
Diplomarbeit - TU Wien · 2019. 12. 11. · Technische Universität Wien Diplomarbeit On the Electrochemical Properties of Li-ion Conducting Li 7 La 3 Zr 2 O 12 and its Utilization
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Technische Universität Wien
Diplomarbeit
On the Electrochemical Properties of Li-ion Conducting
Li7La3Zr2O12 and its Utilization in All Solid State Li-ion
Batteries
Ausgeführt am
Institut für Chemische Technologien und Analytik der
Technischen Universität Wien unter Anleitung von
Univ.-Prof. Dipl.-Phys. Dr. rer. nat. Jürgen Fleig
Projektass. Dipl.-Ing. Stefan Smetaczek
durch
Joseph Ring, BSc.
15. März 2019
Die approbierte Originalversion dieser Diplom-/ Masterarbeit ist in der Hauptbibliothek der Tech-nischen Universität Wien aufgestellt und zugänglich.
http://www.ub.tuwien.ac.at
The approved original version of this diploma or master thesis is available at the main library of the Vienna University of Technology.
http://www.ub.tuwien.ac.at/eng
Abstract
In this work, the impact of eld stress on Li7La3Zr2O12 (LLZO) polycrystalline pel-
lets and single crystals was investigated. Voltages of up to a few V were applied
using metal electrodes at elevated temperatures in dierent gas atmospheres (argon
and air). The electrochemical processes induced by eld stress were investigated
by a combination of impedance spectroscopy and chemical analysis. Local con-
ductivities were determined using circular thin lm micro-electrodes prepared by
photolithography and ion-beam etching. Laterally resolved elemental compositions
of the samples were measured by laser-induced breakdown spectroscopy (LIBS).
The correlation between changes in electrical properties and chemical composition
is discussed.
Moreover, electrochemical properties of Li containing oxide thin lm electrodes
on LLZO were investigated by means of impedance spectroscopy and DC-cycling.
In order to prepare such electrodes (e.g. LiMn2O4 and Li4Ti5O12) a RF sputter-
ing device was built and tested for dierent materials. Inuence of the substrate
temperature on the properties of the thin lms were investigated. All solid state
cells were built with Ta-doped LLZO single crystals and polycrystals and reversible
electrochemical (dis)charging of the oxide electrodes was demonstrated.
1
Acknowledgments
The contribution of all LLZO single crystals by S. Ganschow1 and S. Berendts2
is gratefully acknowledged. Polycrystalline pellets were prepared by the group of
D. Rettenwander3, whose contribution to this work is also gratefully acknowledged.
Samples of alumina powders (CT1200 SG and CT3000 SG), courtesy of Almatis
Germany, were used for the fabrication of heating elements.
1Leibnitz Institute for Crystal Growth, Germany2Berlin University of Technology, Germany3TU Graz, Austria
2
Contents
1 Introduction 5
1.1 A Brief Introduction to Lithium ion Batteries . . . . . . . . . . . . . 5
The discharging reaction can be reversed by an external bias, recharging the battery
back to charged state.
At rst glance, the working principle of LIBs appears simple. In practice, how-
ever, various other processes can take place inside a LIB, which need to be consid-
ered. In order to make a useful energy storage device, a LIB needs to meet certain
criteria: High energy densities (gravimetric and volumetric), long cycle life with
8
Figure 1.1: Schematic illustrating the working principle of lithium ion batteries.Upon discharge, Li-ions (red spheres) traverse the inside of the battery throughthe electrolyte (blue), oxidizing the graphite anode (left) and reducing the layeredoxide cathode (right). Electrons pass through the current collectors (hatched bars)and the resistive load, doing useful work. Recharging of the battery is achieved byapplying an external bias, driving the reaction in reverse direction. The pictureelements are not drawn in correct scale.
little degradation, high charging and discharging rates, high energy eciency, good
stability (thermal, mechanical and chemical), and safety. These requirements impose
high demands on cell design and materials properties, making LIBs sophisticated
devices. Obviously, cathode and anode materials need to have a high, respectively
low potential vs. Li/Li+ in order to make a high voltage cell. Further, lithiation
and delithiation of the materials needs to be reversible over many cycles. Layered
compounds are widely used as electrode materials because they can be (de)lithiated
by (de)intercalation of Li. Open cell voltages of modern LIBs are as high as 5 V [8],
but energy densities of LIBs have to be increased further for mobility applications.
Cathode materials must contain at least one other element, which can be re-
versibly reduced, giving a high reduction potential. Layered transition metal oxides
are the most suitable materials for this. Since these compounds contain not only
Li, but numerous other atoms per formula unit, the energy density is fairly low.
9
Further, most cathode materials cannot be fully delithiated, because the resulting
compound would decompose irreversibly into more stable phases, decreasing the en-
ergy density even more. Anode materials, on the other hand, should have a low
reduction potential. Naturally, Li metal has the lowest possible reduction potential
versus Li/Li+ (0 V). Since there are no atoms other than Li inside Li metal, the
energy density of Li metal is very high. However, certain problems (dendrite forma-
tion, reactions with electrolyte) arising from Li metal anodes are hard to overcome.
As a result, graphite is the most commonly used anode material of today's batter-
ies, because it has comparably good characteristics, but does not suer from the
problems mentioned above.
The electrolyte must be a good Li-ion conductor (σLi > 10−4 S/cm), but a poor
electronic conductor (σe < 10−10 S/cm). Furthermore, it needs to be stable against
chemical reactions with the electrodes. Aqueous electrolytes e.g. cannot be used
in high-voltage devices, because water is not stable at voltages exceeding 1.23 V.
Presently, most LIBs use Li-salts (such as LiPF6 or LiClO4) dissolved in organic,
aprotic solvents (e.g. ethylene carbonate, dimethyl carbonate) as an electrolyte.
These electrolytes have high ionic conductivities (> 10−2 S/cm) and practically no
electronic conductivity. Organic electrolytes, however, are ammable, which can
lead to devices bursting into ames or even exploding in case of failure [9]. This
safety hazard is a major obstacle for the application of LIBs in electric vehicles and
other technology elds. However, ammability is not the only aw of liquid organic
electrolytes. Li metal cannot be used as an anode material in combination with liquid
organic electrolytes. This is partly because they are not chemically stable against Li
metal, which can lead to irreversible capacity losses. Moreover short circuits due to
Li-dendrite formation may occur. The issues imposed on liquid, organic electrolytes
raise the question for alternative electrolytes, such as ionic liquids, solid polymer
electrolytes and inorganic solid state electrolytes. Much research eort has been
focused on solid state, inorganic Li-electrolytes [1012], because they are promising
to overcome the problems of organic-based electrolytes. Further, inorganic solid
electrolytes could even enable the use of Li metal as an anode material and high
voltage cathodes within one cell. In recent years, great progress was made [1316]
in the search for highly Li-conductive (> 10−3 S/cm) ceramic electrolytes, making
10
a big step towards the utilization of solid state electrolytes in LIBs.
1.2 Solid Li Ion Conducting Electrolytes
Most research on solid Li electrolytes is focused on crystalline materials. It is noted,
however, that highly conductive amorphous Li electrolytes have been reported [17
19]. Most solid Li electrolytes are either oxide [10, 20, 21] or sulde [18, 22, 23] based.
In general, sulde based electrolytes exhibit higher Li conductivities, whereas oxides
are chemically more stable. Both types of electrolytes, however, can reach high Li-ion
conductivities (> 10−4 S/cm) at room temperature as well as wide electrochemical
stability windows (exceeding 5 V) [24], making them promising electrolyte materials
for all solid state type LIBs.
Beside high ionic and low electronic conductivity, electrochemical stability of the
electrolyte against the electrodes is a crucial requirement for its implementation in
a LIB. Electrochemical stability is usually referred to the stability against oxidation
or reduction of the electrolyte by the cathode or anode, respectively. Oxidation
of the electrolyte can occur if the electrolyte contains electrons in higher energy
states than the lowest vacant electronic state inside the cathode. This can, from a
thermodynamic point of view, result in the transfer of electrons from the electrolyte
into the cathode material (i.e. oxidation of the electrolyte). Naturally, the electron
transfer is accompanied by chemical reactions, which are irreversible and undesired.
Similar to the oxidation, the electrolyte can be reduced by the anode. This case
requires a vacant electronic state inside the electrolyte and an electron with higher
energy inside the anode. Both, reduction and oxidation reactions are undesired
because they result in an irreversible capacity loss. Further, safety hazards can result
from uncontrolled chemical reactions between electrolyte and electrode materials.
Compatibility of the electrolyte and the electrode materials is therefore an important
aspect of cell design.
The compatibility of electrolytes and electrode materials can be described by a
simple picture (see Fig. 1.2). The highest occupied electronic state and the lowest
unoccupied electronic state of the electrolyte mark the limits of its electrochemical
stability window. In order to make a thermodynamically stable cell, the occupied,
11
respectively unoccupied, electronic states of the electrode materials have to lie in-
side this window. If this is not the case, undesired chemical reactions can occur,
as described above. This means, that the electrochemical stability window of the
electrolyte limits the voltage of the cell. The width as well as the energetic location
of the stability window is therefore an important material property for electrolytes.
Figure 1.2: Schematic energy diagram of a LIB. The electrochemical stability win-dow (Eg) of the electrolyte (orange) spreads from about 0.5 V to around 4 V (withrespect to Li metal). The cell drawn in this schematic is thermodynamically unsta-ble, because the electrode materials do not lie within the stability window. However,the formation of an electronically insulating interface layer (SEI) can stabilize the cellkinetically, preventing the cell from further degradation due to energy mismatches,thus enabling high open-circuit voltages (VOC).
When considering an organic electrolyte, the stability window is dened by the
molecular orbitals (HOMO and LUMO) of the organic molecules of the solvents. For
crystalline solids, this corresponds to the valence band maximum and conduction
band minimum, respectively. Ideally, an electrolyte has a wide electrochemical sta-
bility window, that extends from less than 0 V to a voltage beyond 5 V. However,
even when energetic mismatches between the electrolyte and electrode materials
12
occur, the cell can become stable. This is the case, when the reaction between elec-
trolyte and electrode leads to the formation of a stable solid/electrolyte interface
layer (SEI) [25], which has good ionic but negligible electronic conductivity, thus
preventing the oxidation/reduction from proceeding further, without blocking the
Li-ions from passing between electrolyte and electrode. The formation of an SEI,
of course, is accompanied by a capacity loss, but when a stable SEI is formed, this
loss does not recur after the rst charging of the device. The formation of an SEI
can therefore extend the electrochemical stability window of an electrolyte, enabling
higher cell voltages. When the SEI is chemically or mechanically unstable, however,
a continuous capacity loss takes place, reducing the cycle life of the cell. The impor-
tance of SEIs for the cell performance makes the interface between electrolyte and
electrodes a very interesting research subject [2630]. Some researchers even argue,
that ion transport across an electrode-electrolyte is more important than the bulk
ionic conductivity of the electrolyte [31].
1.2.1 The Garnet-Type Material Li7La3Zr2O12
Presently, many dierent solid Li-electrolytes are known [24]. Due to their high
chemical stability and good ionic conductivity, oxide materials are among the most
promising candidates for the utilization in LIBs. Owing their exceptionally high
bulk ionic conductivity, garnet-type Li-oxides have received special attention [16,
3244] since they were introduced to the LIB community in 2003 by Thangadurai
et al. [45].
Most research eorts on garnet-type electrolytes have been focused on Li7La3Zr2O12
(LLZO) and variations (structure, Li-content, doping, substitution) of it. The garnet
structure of LLZO is complex. It can be described by coordination-polyhedra, i.e.
LaO8 dodecahedra and ZrO6 octahedra, that share common edges. The interstice of
this three-dimensional framework can accommodate a maximum of 9 Li+-ions per
formula unit, distributed among the possible tetrahedral and octahedral Li-sites.
At room-temperature, pure LLZO is only stable in its tetragonal crystal structure
(s. Fig. 1.3), which is a slightly distorted form of the cubic garnet structure. The
tetragonal structure belongs to space group No 142 (I41 / acd) and shows a rather
low ionic conductivity (around 10−6 S/cm) at room temperature [32]. In this struc-
13
Figure 1.3: Crystal structure of tetragonal Li7La3Zr2O12. The solid box indicatesthe unit cell. The three dierent types of Li-sites are shown on the right. Graphicreproduced with permission of Elsevier [32].
ture, two thirds of the tetrahedral sites (16 out of 24) are vacant, while all of the
octahedral sites (48) are occupied by Li-atoms, resulting in a total of 56 occupied
(out of a total of 72) Li-sites per unit cell. Note, that one unit cell corresponds
to four complete formula units of LLZO (Li7La3Zr2O12). Partial occupation of the
Li-sites is an important prerequisite for high ionic conductivity, since ionic motion
within a crystal is dependent on the presence of vacant sites for the transported
ion. Ions can then move from their initial site to a vacant site in close proximity,
leaving a new vacancy behind. This ionic motion has to be activated, since there
is an energetic barrier between the two neighboring sites, that has to be overcome.
The probability of a jump from one site to a neighboring site, and hence the ionic
conductivity σion can be related to the activation energy (Ea):
σion ∝ exp(− Ea
kBT)
The height of this activation barrier is inuenced by the local crystal structure (i.e.
bond lengths, ionic radii, etc.). The relative number of vacant sites is, of course, an-
other important factor for ionic conductivity. The mechanism of ionic motion inside
14
the complex structure of LLZO is not only inuenced by the relative occupation of
the Li-sites, but also by the distribution among the dierent sites (tetrahedral or
octahedral). In LLZO, tetrahedral and octahedral Li-sites are interconnected to a
3D network. The Li coordination-polyhedra share common, triangular faces. Each
tetrahedral site is surrounded by four octahedral sites, whereas each octahedral site
connects two dierent tetrahedral sites. As mentioned before, the bulk ionic con-
ductivity of tetragonal LLZO is fairly low, compared to the cubic polymorph (>
10−4 S/cm). This dierence is attributed to small but signicant dierences of the
crystal structure. While both structures have the same basic garnet network, the
distribution of Li-ions among the Li-sites in cubic LLZO is dierent and not as or-
dered as in tetragonal LLZO. This leads to a far higher ionic conductivity in the
cubic polymorph of LLZO. A detailed discussion on the ion migration mechanism
in garnets would be beyond the scope of this thesis, thus the reader is referred to
the literature [34, 36, 46, 47]. As mentioned before, the cubic garnet-phase of LLZO
is no stable at room temperature. However, the cubic phase can be stabilized by
doping with various elements [33, 4852]. Doping with aliovalent atoms can also
change the Li-content of the LLZO, further inuencing the conductivity. The high-
est conductivity yet (1.2 · 10−3 S/cm), amongst LLZO garnets, was achieved by
Rettenwander et al. by doping LLZO with Al 3+ and Ga3+.
Beside the high ionic conductivity of LLZO, the high chemical stability is a key
property for its applicability in LIBs, especially for high voltage cells. LLZO has
wide electrochemical stability window (i.e. band gap) of more than 5 V [33, 53]. The
energetic location of the stability window, with respect to desired cathode/anode
materials, is an important aspect as well. Several works on the stability of LLZO
indicate, that LLZO is stable against Li metal [34, 5356]. The interface between
LLZO and Li metal anodes, however, can give a large resistance. This can be
attributed to poor mechanical contact between the two materials, but small contact
area is not the only contributor to the large interface resistance. LLZO does not
exhibit chemical stability against air (i.e. oxygen, moisture and CO2) [37, 43, 57,
58]. The degradation of LLZO in ambient atmosphere leads to the formation of a
Li2CO3 layer. If such a layer is present prior to electrode deposition, it leads to a
high interface resistance between LLZO and the electrode material. Thus, special
15
care must be taken to prevent this, making the processing of LLZO laborious.
The high resistance of the interface between LLZO and cathode materials, such
as LiCoO2 (LCO), is believed to be a key issue for solid state LIBs [31, 59, 60].
Despite its wide stability window, LLZO forms a resistive reaction layer with LCO at
elevated temperatures [6163]. Sintering or annealing of solid state cells at elevated
temperatures, however, is a common strategy to achieve good contact between the
dierent materials. Recent studies showed, that thermally induced inter-diusion
occurs at temperatures as low as 300 C [64], while other researchers claim, that
solid state reactions between LLZO and LCO do not occur below 700 C [63]. The
reported reaction products include La2Zr2O7, LaCoO3, La2CoO4, La2Zr2O7 and
Li2CO3, if CO2 is present in the atmosphere. The resulting reaction layers show
high area specic resistance of several kΩ · cm−2 [64]. Several approaches to this
problem, such as lower processing temperatures or addition of an interlayer, are
suggested. The critical role of the electrolyte-cathode interface presses for extensive
research, to overcome the problems induced by interface reactions.
A reliable synthesis of highly conductive LLZO is an essential prerequisite for
larger scale production. However, reports on conductivities of LLZO with nominally
equal or very similar composition, diverge [62, 65, 66]. This scattering could be due
to dierent synthesis conditions. There are several dierent synthetic approaches for
the production of LLZO (e.g. solid state reaction, mixed precursor method, sol-gel
methods, etc.). Most synthetic routes to LLZO electrolytes include a sintering or
annealing step for densication of the material. The elevated temperature during
these steps can lead to the evaporation of relatively volatile LiO2. The resulting
Li-loss is often compensated by adding excess Li-containing salts (such as LiO2 or
Li2CO3) prior to the sintering step. Elaborate studies [42, 44] by Wachter-Welzl
et al. on nominally equal LLZO samples showed scattering over almost two orders
of magnitude (2 · 10−5 − 8 · 10−4 S · cm−1) in bulk ionic conductivity. Further,
variations in ionic conductivity within single samples were observed by means of
microelectrode impedance spectroscopy. Detailed chemical analysis showed varia-
tions in the elemental composition of the samples. However, no direct correlation
between local conductivity and composition was apparent. Point defects, such as
oxygen vacancies can also aect the Li-ion conductivity. Kubicek et al. observed
16
signicant oxygen vacancy concentrations in doped cubic LLZO by secondary-ion
mass spectroscopy (SIMS) [41]. Oxygen vacancies can inuence the stoichiometry
as well as the the crystal structure of LLZO. Both eects can have an impact on the
conductivity, thus, a complex relationship between oxygen vacancy concentration
and ionic conductivity was observed. Interestingly, at 350 C, an unexpectedly high
oxygen tracer diusivity (diusion coecient as high as 8.2 · 10−12 cm2 · s−1) was
observed. Despite the great research eorts, the complex interplay of stoichiometry,
defect concentration and conductivity remains unclear.
1.3 Motivation and Aim
The cubic modication of LLZO exhibits excellent properties for a solid state Li
electrolyte. Processing of the material is challenging. Uncontrolled variations of
stoichiometry during synthesis have inuence on the conductivity. The underlying
relation between stoichiometry and Li-ion conductivity is not fully understood. To
get a deeper understanding, the preparation of samples with well dened stoichiom-
etry is crucial. Due to the high temperature steps in most synthetic routes, this is
hardly achievable. Hence, a tool for well-dened modications of the composition
within LLZO samples is desirable.
One aim of the experiments was thus the controlled manipulation of local stoi-
chiometry of Ta-doped LLZO single crystals by application of eld stress. Stoichiom-
etry gradients in solid ionic conductors can be induced by so-called Wagner-Hebb
polarization, referring to the rst observations of this eect by C. Wagner [67] and
M. Hebb [68]. After polarization of a single crystal, by means of metal electrodes lo-
cally resolved conductivity measurements as well as chemical analysis was performed
to check for resulting changes in stoichiometry and conductivity. Results from these
experiments could help to understand the relationship between local conductivity
and stoichiometry in more detail.
These experiments came along with other phenomena, such as electron conduc-
tion due to ion blocking electrodes or chemical decomposition of LLZO upon high
eld stress. To get a better understanding, further eld stress experiments were
conducted under varying conditions (atmosphere, temperature, geometry, etc.). All
17
these experiments are described in Part I (chapters 2 to 3).
The second part of the thesis deals with oxide thin lm electrodes on LLZO.
There is still very little agreement on the properties of the interfaces between typical
LIB electrodes (e.g. LiCoO2, LiMn2O4, Li4Ti5O12) and LLZO. Model thin lm
oxide electrodes on single crystalline LLZO may serve as excellent systems for a
more systematic study of these interfaces and the charging-discharging behavior.
Therefore such model systems were prepared and rst measurements were performed
to show the practicability of this approach. In order to deposit such electrodes, an
novel sputtering device was constructed and used.
18
Part I
Eects of Field Stress on the
Electrochemical Properties of the
Solid State Electrolyte LLZO
19
Chapter 2
Experimental
2.1 Sample Preparation
All experiments were conducted with Ta-doped LLZO (LLZTO) single crystals, cour-
tesy of S. Ganschow1 and S. Berendts2. LLZTO single crystals with the nominal
composition Li6La3ZrTaO12 were grown by the conventional Czochralski technique.
The starting materials were Li2CO3, La2O3, ZrO2, and Ta2O5. Carbonates and
oxides were dried and mixed in the required stoichiometry with a 10% excess of
Li2CO3. Afterwards the powders were uni-axial pelletized, isostatically pressed at
2800 kbar and nally sintered at 1373 K for 16 h in air. Capped magnesia crucibles
were used while covering the pellets with the respective LLZTO powder to avoid Li-
loss during sintering. Because of the high melting temperature, the sintered LLZTO
samples were molten by radio frequency induction heating using a 25 kW microwave
generator. An iridium seed (pulling rate 1.5 mm · h−1, rotation speed 1 rpm) was
used for the crystal growth performed under nitrogen atmosphere. An active af-
terheater was applied to adjust the temperature gradient in the set up. Thermal
insulation was established by an outer alumina ceramic tube lled with zirconia
that the LLZTO single crystal obtained has indeed the composition Li6La3ZrTaO12.
The crystals were cut into slices (about 7 by 5 mm and 1 mm thick). Since surface
degradation of LLZO in air is a known issue, all crystals were mirror-polished with
1Leibnitz Institute for Crystal Growth, Germany2Berlin University of Technology, Germany
20
SiC grinding paper (#4000) prior to any electrode deposition.
Ionically blocking gold electrodes (200 nm thickness) were deposited onto the
crystals by DC-sputtering (Baltec Med 020) at room temperature. Micro-structuring
was done by photo-lithographic techniques and subsequent ion beam etching. Sam-
ples of two, slightly dierent electrode congurations were prepared this way (s.
Fig. 2.1). The bottom face of both the crystals was completely covered with a gold
electrode. The top faces were structured to get arrays of circular microelectrodes
(100 µm diameter). Additional stripe electrodes were prepared on one sample.
Figure 2.1: Schematic illustration of the prepared samples. The bottom face ofboth the samples is covered with a single electrode. Sample (a) has large stripeelectrodes on the top face, microelectrodes are located between. On sample (b),only microelectrodes were prepared.
2.2 Field Stress Experiments with stripe electrodes
Ta-doped single crystals were subjected to electric eld stress at elevated temper-
atures (300 - 450 C) under lab atmosphere (air). The experimental setup is illus-
trated schematically in Fig. 2.2.
2.2.1 Electrochemical Methods
The polarization voltage (2.5 - 5 V) was applied via the large stripe electrodes (s. Fig.
2.1 (a)), for several hours (15 to 68 h) at elevated temperatures. The samples were
heated from below via a lab-built heating unit connected to a Linkam thermal control
unit. Since the temperature is controlled according to a thermocouple inside the
heating unit, the actual sample temperature is lower than the set temperature. Since
no exact knowledge of the actual sample temperature is required for the experiments,
all temperature indications in this work correspond to the set values, which are
21
Figure 2.2: Schematic illustration of the experimental setup. Field stress is appliedvia two opposing stripe electrodes at elevated temperature (1). Laterally resolvedconductivity measurements are done via microelectrode electrochemical impedancespectroscopy, prior to and after the polarization experiments at room temperature(2).
22
higher (about 20 - 50 K) than the actual sample temperatures. A Keithley, 2611
Source Meter unit was used to apply voltage and measure the resulting current. The
polarization experiments were done at elevated temperatures to ensure high ionic
conductivities of Li-ions and at least some mobility of O-ions [41].
Each stripe electrode was contacted together with the opposite stripe electrode.
Thus, the applied electric eld was only present in the area between the correspond-
ing electrode pairs. After a given time upon voltage the sample was cooled to room
temperature (25 C). The bias was still applied during cooling, to keep possible po-
larization eects from relaxing into the initial state of the sample. Cooled to room
temperature, the ionic mobility was considered so low, that no fast relaxation of po-
larization could take place. Between each of the stripe electrode pairs, three arrays
of circular microelectrodes were used for conductivity measurements at room tem-
perature, prior to and after the polarization experiments. These experiments were
performed by means of electrochemical impedance spectroscopy (EIS). For these
measurements, a single microelectrode was contacted as a working electrode to the
impedance analyzer, whereas the large area backing electrode (on the backside of
the crystal) was contacted as the counter-electrode. For the EIS measurements, a
Novocontrol Alpha Analyzer was used, in the frequency range from 1 MHz to 1 kHz.
Owing the relatively small area of the microelectrodes, the probed volume of the
EIS is located directly underneath the microelectrodes, which allows for laterally
resolved conductivity measurements, as has been shown by J. Fleig [69]. The tech-
nique of microelectrode-EIS (ME-EIS) was applied successfully to investigate local
conductivities in LLZO [42].
2.2.2 Chemical Analysis
After the polarization experiments and subsequent ME-EIS measurements, the ele-
mental composition of the sample was determined by laterally resolved laser induced
breakdown spectroscopy (LIBS). This technique uses a laser beam which is focused
onto the sample surface. Upon impact of the laser beam, material is ablated and
excited. Upon relaxation, the present elements emit characteristic light, which is
measured by a detector. The intensities of the emitted light can be related to rel-
ative elemental compositions within the sample. Line-scans were performed across
23
Figure 2.3: Schematic illustration of a LIBS measurement. While the laser beamis moved over the surface of the sample, material gets ablated and excited. Uponrelaxation, characteristic radiation is emitted, indicating the elemental compositionof the scanned material.
the polarization axes, since a stoichiometry gradient was expected to be present
along the axes (s. Fig. 2.2). Due to matrix eects, the LIBS measurements cannot
be used directly for absolute quantication of elemental composition. A laborious
standardizing procedure would be required for this, which would be beyond the
scope of this work. However, no absolute quantication of the elemental compo-
sition is needed for this experiment. The relative composition within the sample
yields enough information to observe a possible stoichiometry gradient, caused by
the electric eld.
2.3 Field Stress Experiments with Microelectrodes
Further polarization experiments, with more focus on the polarization current, were
done, using a dierent sample geometry (s. Fig. 2.1 (b)). Field stress (from -2 V up
to 2 V) was applied by contacting either two neighboring MEs, or one ME and the
large backside electrode at elevated temperature (300 C) and room temperature (s.
Fig. 2.4). The atmosphere was either air or argon. Using only one or two ME for
each experiment, a large number of experiments could be done on one crystal.
LIBS measurements were done after polarization, to investigate possible changes
in elemental composition due to eld stress. As only MEs were contacted during
polarization, the laser was not scanned over the sample, but focused on individual
24
MEs to obtain information about the local composition.
Figure 2.4: Schematic illustration of the polarization experiments using microelec-trodes. Two neighboring MEs (a) or a single ME and the large counter electrode(b) were contacted. Field stress was applied, measuring the resulting current.
25
Chapter 3
Results and Discussion
3.1 Results from Polarization of Stripe Electrodes
The impact of eld stress on electrochemical properties of LLZTO was investigated
by means of ME-EIS (ionic conductivity) and LIBS (elemental composition). Both
LIBS and EIS measurements were done prior to and after the polarization experi-
ments, to reveal any change caused by eld stress.
3.1.1 ME-EIS
Since most of the voltage between a ME and a large counter electrode (covering the
entire bottom face of the sample) drops very close to the ME, the diameter of the
electrode determines the probed volume beneath [69]. The resistances, measured my
EIS are mostly determined by the conductivity of a hemisphere with a radius of 2d,
where d is the diameter of the ME. Fig. 3.1 shows a typical impedance spectrum of
a ME, obtained prior to any polarization, in form of a Nyquist-plot. Fitting of the
impedance data was done with Zview software (Vers. 3.4f , Scribner Associates).
There are two distinct features in the plot, a high-frequency arc and a low-frequency
capacitive increase. The former corresponds to the charge transport in the probed
sample, which is described by a resistive element (RLLZO) and a parallel constant-
phase element CPELLZO in the equivalent circuit. The low-frequency increase is
caused by the ionically blocking gold microelectrode, described by CPEAu in the
equivalent circuit. A tting of the equivalent circuit to the impedance spectra reveals
the values of RLLZO. Since single crystals were investigated, no charge transfer along
26
Figure 3.1: EIS data (circles) of a ME, measured prior to any experiments at roomtemperature. Fitting results are shown as a dashed line. The used equivalent isshown in the inset.
grain boundaries is possible. Thus, only the local bulk ionic conductivity of the
probed sample contributes to the high-frequency arc. From the resistance of the
high-frequency arc, the local bulk ionic conductivity σLi can be calculated:
σLi =1
2 d RLLZO
(3.1)
Impedance data measured from various microelectrodes prior to polarization show
variations in ionic conductivity (s. Fig. 3.2). After the measurements, the sample
was subjected to thermal treatment (58 hours at 300 C), similar to the polarization
experiments, but without application of any bias. Then, impedance spectra were
measured from the same microelectrodes, again (s. Fig. 3.2). The sample already
shows a scattering of conductivity (2 · 10−5 - 9 · 10−5 S/cm), right after preparation.
An increase of conductivity (up to 2 · 10−4 S/cm) is apparent for most ME, whereas
other ME show no signicant changes in conductivity.
After the application of DC bias (3 V) for 15 hours at elevated temperature (400
C) and subsequent cooling to room temperature (under bias), EIS measurements
were carried out, to investigate the impact of eld stress on the conductivity. After
each polarization experiment, the microelectrodes between the corresponding stripe
27
Figure 3.2: Ionic conductivities, measured via EIS at various microelectrodes. Scat-tering of conductivity is apparent directly after sample preparation (black squares).After the thermal treatment (red diamonds), an increase in conductivity can beobserved.
electrodes were used for ME-EIS (s. Sec. 2.2). Fig. 3.3 shows data from ME-
EIS prior to and after polarization via two opposing stripe electrodes. Close to the
cathode, a slight increase in conductivity is apparent, whereas close to the anode a
decrease can be observed. This general trend was present also on further polarization
experiments, however, the magnitude of the conductivity changes was rather low in
most cases.
Also, the current, owing through the sample during polarization was measured
(s. Fig. 3.4). A rapidly decreasing current is found which reaches 1.3 µA after 5
minutes. After about 60 minutes, the current has dropped to about 1.2 µA and
then remained almost constant for many hours. The relatively high current at
the beginning at the experiment is attributed to a motion of Li-ions. In order to
compensate the gradient in electrochemical potential, induced by the electric eld,
the Li-ions move from the cathodic side to the anodic side of the sample. Assuming
the gold electrodes are blocking the Li-ions, the Li-ion current cannot be retained
for a longer time but drops. After the ionic current has largely diminished, a low
current of about 1µA is still constantly owing. This so-called steady state current is
attributed to the small, but non-zero electronic conductivity of the sample [53]. Since
the electrons can readily pass through the gold electrodes, the electronic current is
28
Figure 3.3: Ionic conductivities, measured via EIS at microelectrodes between thepolarization electrodes, before (black squares) and after polarization (blue circles).Again, scattering of the conductivity is apparent before the polarization. Afterapplication of eld stress via the stripe electrodes (left and right), conductivitieshave changed.
constant over time. In such a case the electronic conductivity can be estimated from
the voltage (U=3 V), current (I=1 µA), length (l=4 mm) and cross section (A=0.63
mm2) of the conductor:
σ =1
R
l
A=I
U
l
A(3.2)
From equation (3.2) an electronic conductivity of about 2·10−5 S/cm is obtained
at 400 C set temperature. This conductivity value, however, is too high to be
attributed to the poor electronic conductivity of LLZO.
Therefore we assume that a continuous decomposition of the sample leads to a
certain Li-ion current which contributes to the 1 µA measured. Existence of some
ionic current is supported by several features observed during and after eld stress.
After polarization, optical changes of the polarization electrodes were apparent (s.
Fig. 3.5), indicating electrochemical reactions at the electrodes. The anode shows a
rough surface after polarization. The roughness appears to be caused by gas bubbles,
possibly arising from the oxidation of oxide ions, which lift o the gold electrode
from the sample surface. On top of the cathode, a dark solid was deposited during
polarization, growing from the edge of the electrode over the entire surface of the
sample. Only a small area of the underlying metal electrode is still visible in the
29
Figure 3.4: Current measured during polarization (3 V) at 400 C. Starting at arelatively high value, the current drops drastically within minutes to a value ofabout 1 µA. a) shows the entire experiment, whereas (b) shows a magnied view ofthe rst 30 minutes.
microscope image. The dark solid most likely consists of Li2CO3, possibly mixed
with Li2O or LiOH. These are expected for cathodic reduction of oxygen in a Li
-ion (i presence of CO2 and H2O, see below). Electrochemical reactions during
polarization could also cause the irregular increase of current in the time interval
from about 6 hours to 11 hours (s. Fig. 3.4). After these irregularities, a slight
decrease of current is observed. This drop can be caused by the gradual lift-o of
the anode, due to the formation of oxygen gas, which leads to a smaller contact
area, thus increasing the resistance of the sample.
The deposition of Li salts and gas formation at the cathode and anode, respec-
tively, can be explained by the following, suggested mechanism (s. Fig. 3.6): Upon
eld stress, oxygen ions are oxidized at the anode, leaving oxygen vacancies inside
the crystal:
Anodic half reaction : OXO →
1
2O2 + V ∗∗O + 2e′
The gaseous oxygen formed upon oxidation is trapped underneath the metal an-
ode, lifting it o. The positive charge of the oxygen vacancies is compensated by
formation of negatively charged Li vacancies (at a ratio of 1:2). The Li-ions are
transported towards the cathode leaving negatively charged vacancies at the an-
ode. These excess Li-ions react with excess electrons, provided by the cathode, and
gaseous species from the ambient atmosphere (i.e. H2O, O2 and CO2), thus reducing
30
the gases and depositing LiOH, Li2O or Li2CO3, e.g.:
Cathodic half reaction : 2Li+ + 2e′ + CO2 +1
2O2 → Li2CO3
Since electrons, Li-ions and the ambient atmosphere are required for this reaction, it
can only take place at the triple-phase-boundary, i.e. the edge of the metal electrode.
Overall, the process leaves Li- and O depleted LLZO and oxygen bubbles underneath
the anode, while the cathode is covered with Li-containing salts.
Figure 3.5: Optical microscopy images of the polarization cathode (left) and anode(right) after a bias of 3 V was applied for 15 hours at 400 C. Prior to polarization,these electrodes had a smooth and reective golden surface. Magnication: 10x,scale bars: 200 µm.
3.1.2 LIBS
After the application of eld stress, LIBS measurements were carried out. The laser
was scanned across the sample in lines between the stripe electrodes (s. Sec. 2.2).
Each line was scanned twice, since the gold electrodes were still present on top of
the sample, possibly aecting the measurement of the elemental composition of the
LLZTO. The rst laser scan of each line completely ablated the gold covering the
sample. The second line-scan is not aected by the electrodes, thus the signals can
be attributed to the LLZTO sample.
The intensity of the emitted, characteristic radiation cannot be directly used for
analysis of the elemental composition. Possible variations in laser-intensity or other
inuence on the signal, can simulate variations in concentration of the elements.
31
Figure 3.6: Schematic illustration of the suggested mechanism. Oxidation of oxygenat the anode leads to oxygen gas evolution. Oxygen gas formed underneath themetal anode leads to bubbles, partially lifting o the anode. At the cathode, gaseousspecies, such as O2, are reduced, yielding Li2CO3. Overall, Li-ions are transportedthrough the sample, leaving vacant Li-sites and O-vacancies at the anodic side.Li-containing salts are formed at the triple phase boundary of the cathode.
To circumvent this problem, a reference signal is required. The element, emitting
the reference signal should have a uniform concentration throughout the sample.
Thus, any variations in signal intensity of this element can be attributed to external
eects, such as laser-intensity, laser-focusing etc. Both La and Zr are expected to be
uniformly concentrated throughout the sample. The relative La/Zr signal (s. Fig.
3.7) is scattered, but shows no signicant changes over the scanned lines. Since it is
highly improbable, that the concentrations of La and Zr are related in a way such
as to yield a uniformly scattered, relative signal, the results presented in Figure 3.7
prove, that La and Zr concentrations can be considered constant. Therefore, a Zr
signal (λ=327 nm) was chosen as a reference signal for all LIBS measurements.
The relative signals of La and O showed no signicant changes over the length
of the line-scans. However, primarily Li was expected to move through the sample,
as its ionic conductivity is highest. The relative Li intensity (λ=610 nm) of two
line-scans are shown in Fig. 3.8.
Between the two stripe electrodes, no distinct gradient in Li content is apparent.
However, depletion and enrichment of Li is apparent underneath the cathode and
anode, respectively. The shift of Li from the anode to the cathode is in agreement
with the suggested polarization mechanism and the polarization current data. The
32
Figure 3.7: La-267 signal, referenced to the Zr-327 signal of a typical LIBS line-scan.Prior to this scan, a preliminary laser scan was performed to ablate the electrodematerial. The positions of the polarization electrodes are indicated by golden stripes.
increased Li content at the cathodic side of the sample indicates, that some can
Li-ions accumulate at the cathode that do not react with the gas phase, thus locally
reducing the LLZTO.
3.2 Results from Polarization of Microelectrodes
The current, owing during polarization of microelectrodes, was measured under
various conditions (gas atmosphere, temperature, voltage), to get a deeper under-
standing of the electrochemical processes occurring during polarization of LLZTO.
Further, the impact of eld stress on the elemental composition was investigated by
LIBS measurements.
3.2.1 Polarization Current Measurements
Polarization of two neighboring MEs gives rise to a current, similar to the current
observed during polarization of large electrodes (cf. Sec. 3.1.1). A typical current
prole, obtained during repeated polarization under Ar-atmosphere at 400 C, is
shown in Fig. 3.9. The sample was polarized repeatedly with increasing bias. Be-
33
Figure 3.8: Li-610 signal, referenced to the Zr-327 signal of two LIBS line-scans.Each line was scanned twice to separate signals of the LLZO from any signals of thegold electrodes and possible precipitates on top. (1) shows the rst line scan overthe sample. After the rst scan all gold electrodes were ablated. The second scan(2) is therefore not aected by electrodes and possible precipitates.
tween each polarization period (10 minutes), the voltage was set to 0 V for 1 minute.
All curves in Fig. 3.9 show two distinct features:
i. A relatively high current (exceeding 1 nA) at the beginning of the experi-
ment, rapidly decreasing to values < 0.1 nA within seconds (depending on the
voltage). Apparently, the positive current approaches a constant, small value,
which is not reached within the time of the experiment. The magnitude of
the constant, so-called steady-state current (SSC), is directly related to the
applied voltage. As in section 3.1.1, the SSC is attributed to a small, but
non-zero electronic conductivity of the sample, since no other charge carrier is
expected ow through the sample steadily.
ii. When the voltage is set to 0 V, a negative current arises. The magnitude of
this negative current is decreasing in a similar manner as the positive current
during polarization. However, the current must approach 0 A, since no voltage
is applied to the sample. Again, the time of the experiments was not long
enough to observe the current reaching 0 A.
The results shown in Fig. 3.9 reveal that application of bias to neighboring MEs (s.
34
Figure 3.9: Typical current data, obtained during polarization of two neighboringmicroelectrodes. 100 (black), 200 (red) and 300 mV (blue) were applied for 10minutes in air at 400 C. Between each polarization period, the bias was set to 0 Vfor 1 minute. The voltage-over-time prole is shown in the inset.
Fig. 2.4 (a)), gives rise to a reversible charging in the sample. Removing the bias
leads to a reverse current, indicating, that the charge carriers, shifted during polar-
ization, migrate back towards their initial positions. Most probably the following
electrochemical processes take place: The bias causes a dierence of the electro-
chemical potential of the Li-ions in LLZO. The gradient of electrochemical potential
serves as the driving force for Li-ion migration. The shift of Li-ions compensates the
gradient in electrochemical potential, approaching equilibrium (i.e. constant electro-
chemical potential). When the bias is removed, the gradient in Li-ion concentration
is no longer compensated for, thus the gradient leads to a reverse migration, ap-
proaching the original Li-ion distribution.
Li-ions are the ionic charge carriers with the highest mobility in LLZTO. There-
fore, the reversible polarization current can be attributed to the reversible migration
of Li-ions. Further, migration of Li-ions due to eld stress was shown in experi-
35
ments described above (s. Sec. 3.1). Since the experiments were conducted under
Ar atmosphere, reactions with the gas phase are highly improbable. Indeed, no de-
composition reactions (cf. Sec. 3.1.1) were observed during any of the polarization
experiments (voltages up to 3 V). Thus, in the true steady state situation (not yet
reached here) only electrons are expected to ow. This is also in accordance with
the much lower currents found in these experiments compared to those in Sec. 3.1.
3.2.2 Inuence of the Atmosphere
The formation of Li2CO3 shows, that reactions of LLZTO with the gas phase can
occur when eld stress is applied via blocking electrodes (s. Sec. 3.1.1). Therefore,
the chemical composition of the gas phase aects the polarization in some way. To
get a better understanding of the inuence of the gas phase on the polarization of
LLZTO, polarization experiments were done under Ar and air atmosphere at 400
C. Both under Ar and air atmosphere, the same voltage prole was applied to a
ME, whereas the large backing electrode served as the counter electrode (s. Fig. 2.4
(b)). The resulting current data are displayed in Fig. 3.10.
Fig. 3.10 shows currents, which are typical for polarization of MEs (cf. Fig. 3.9).
The red and blue curves dier slightly, indicating that the atmosphere has an eect
on the polarization current. During polarization under air, the current is generally
lower in value, and decreasing faster than the current measured in Ar atmosphere.
Further, the current measured under air is still decreasing after 10 minutes of po-
larization, whereas the current under Ar even shows a slight increase after about 7
minutes. The current in Ar after 10 minutes (about 50 pA at 1 V) is in reasonable
agreement with that for two microelectrodes (Fig. 3.9, about 25 pA at 300 mV).
Assuming that steady state current is electronic, we may use the corresponding re-
sistance (R = 1 V50 pA
= 20 GΩ) for calculations on electronic conductivity at 400 C.
From the spreading resistance formula (Eq. 3.1) we get 2.5·10−9 S/cm.. The neg-
ative currents, observed when turning o the applied bias, show an opposite trend.
The negative current under air is generally higher in value, than the current under
Ar. The faster decrease of current observed during polarization under air might
be caused by chemical changes of the sample. Although no distinct evidence for a
decomposition of the sample (i.e. bubble formation or precipitation) was apparent
36
Figure 3.10: Polarization current, resulting from experiments carried out underAr (blue) and air atmosphere (red) at 400 C. The solid lines show results frompolarization with 100 mV, the dashed lines show results from polarization with 1000mV.
at the electrodes, of the LLZO sample could occur during polarization under air.
Since surface degradation of LLZO in air is a known issue [57], this is a possible
explanation for the steady decrease in conductivity. To avoid undesired degradation
eects, most proceeding experiments were carried out under Ar atmosphere.
The current, indicated by the solid red line in Fig. 3.10, becomes negative after
about one minute. Since a positive voltage of 100 mV is applied to the sample,
a negative current, owing against the electric eld is not consistent with basic
physical concepts. Therefore, the slightly negative current values (> -10 pA) are
attributed to measurement errors, possibly arising from thermoelectric phenomena
in the measurement setup.
3.2.3 Inuence of the Temperature
Both ionic and electronic conductivities are dependent on temperature. Generally,
ionic conductivity in solids is enhanced at higher temperature, because the activation
energy for ion migration can be overcome more easily. LLZTO is a wide-band-gap
37
semiconductor, therefore thermal excitation of electrons might also lead to enhanced
electronic conductivity at higher temperature. Therefore, a generally higher current
is expected, when polarization is carried out at higher temperature. Fig. 3.11 shows
current measurements from similar polarization experiments, conducted at room
temperature (25 C) and 400 C under Ar atmosphere.
Figure 3.11: The black lines, solid (100 mV), dashed (200 mV) and dotted (300mV), indicate the measured current during polarization at room temperature. Thecorresponding results obtained at 400 C are indicated by red lines. A generalincrease of current at 400 C is apparent.
The comparison in Fig. 3.11 shows, that the measurements at 400 C are sub-
jected to a distinct, negative oset (s. Sec. 3.2.2), whereas the data obtained at
room temperature show barely any current-oset. This temperature dependence
suggests, that thermoelectric eects contribute to the current-oset. The basic fea-
tures of the current curves are the same, independent of temperature. However, the
red curves do not approach a steady-state during the course of the experiments (10
minutes), whereas the black curves appear constant after about 5 minutes of polar-
ization. This indicates, that a larger number of charge carriers (possibly Li-ions or
O-ions) is shifted reversibly during polarization at higher temperature. Assuming
that the steady state current (not yet reached) corresponds to electron transport
and the additional decaying current is due to Li-ions, we can roughly estimate the
amount of Li transported during the experiment.
38
We may assume a steady state current of 10 pA for 300 mV and subtract this
from the measured current. Then we integrate∫ t1t0I(t)·dt. This gives 1.8·10−8 C and
thus about 1.1·1011 Li-ions. Assuming the density of LLZTO is about 5.3 g/cm3 [70]
this would correspond to a depletion of Li in LLZTO by about 3.7·10−3 per formula
unit in a region of 1 µm beneath the microelectrode (i.e. in ca. 0.8 ·10−8 cm3).
Hence, only a slight stoichiometry gradient builds up.
3.2.4 Results from a Broader Voltage Range
Polarization experiments with MEs were carried out over broader voltage range
(from -2 V up to 2 V) under Ar atmosphere. The voltage was increased/decreased
incrementally, holding the voltage at each step for a given time period (typically
1 hour). After the maximum/minimum voltage of ±2 V was reached, the voltage
was decreased/increased again, retracing the steps. Typical voltage increments were
between 20 mV and 100 mV.
Results from polarization of a ME contacted versus the large backing electrode,
are displayed in Fig. 3.12. Voltage was increased from 0 V to 2 V (100 mV incre-
ments) and decreased to 0 V (Ar atmosphere, 300 C). Currents are muhc lower
than at 400 C. Part (a) shows selected current curves of the increasing part of the
experiment, whereas the decreasing half is shown in (b). Current curves, measured
during polarization with 1100 mV (dark green) and 1200 mV (light green) have
atypical shape. A high current at the beginning decreases rapidly. The curves do
not, however, approach a constant value within the course of polarization time (1
hour). Strangely, the current starts increasing slightly after few minutes.
When the voltage is increased to 1300 mV, a distinct change of the current curve
is apparent: The initially high current decreased over several minutes. Further, the
current drops down below the current from 1200 mV polarization after about 45
minutes and keeps decreasing afterwards. Subsequent current curves (1400 mV -
2000 mV) have the same curve shape again. However, values of the current curves
are declining, despite the increasing voltage.
When the voltage is decreased (b) again, the curves seem to approach a constant
value from below. When voltage is turned o after polarization (c.f. Fig. 3.9), a
negative current, approaching zero, arises because charge carriers migrate back to-
39
wards their initial positions at 0 V. Similarly, when the voltage is decreased during
polarization, a reverse current is induced. However, when the voltage is still applied,
this revers current is overlain by the constant steady-state-current, thus approaching
a positive steady state. In contrast to Fig. 3.12 (a), the curves in Fig. 3.12 (b) seem
almost constant after 60 minutes of polarization. Further, the approached current
values are related to the applied voltage directly. Overall, the results suggest, that
currents, obtained when the voltage is decreased, are more suitable for the deter-
mination of actual steady-state current values. Still, the rather unusual behavior
might indicate that most probably a chemical decomposition process rather than the
electronic conductivity is measured at such high voltages. This is also in agreements
with the measurements shown below.
Polarization of MEs with negative voltages gives rise to dierent current curves,
as described above. Selected results from polarization with up to -2 V are displayed
in Fig. 3.13. The voltage prole was analog to the positive voltage polarizations,
described above. All other experimental parameters were identical to the positive
voltage polarization experiments.
Disregarding the negative signs of current and voltage, the results from nega-
tive polarization of MEs are similar to the results from positive polarizations: The
negative current, arising from negative polarization, is relatively high in value at
the beginning, dropping rapidly to smaller negative values, approaching a steady
state. Increasing the voltage to higher, negative values, results in higher negative
currents. When the value of the negative voltage is decreased again, some charge
carriers migrate in the reverse direction, giving rise to a positive, decreasing current.
The comparison between corresponding dashed and solid lines in Fig. 3.13 indicates
how closely the curves have approached the steady state current.
3.2.5 U-I Characteristic
Numerous polarization experiments in the voltage range from -2 up to 2 V were
carried out in a similar way as described in the previous sections. The voltage
increments and the time of each polarization step was varied, but the basic exper-
iment was unchanged. To investigate the nature of the steady-state current, which
is approached during a polarization experiment, steady-state currents of many ex-
40
Figure 3.12: Current, measured during polarization of a ME at 300 C. The voltagewas increased (a) and decreased (b) incrementally in 100 mV steps, going from 0 Vto 2 V and back to 0 V. For the sake of readability, only selected data is shown. Theblack line, indicating the current measured at 2 V, is displayed in both graphics andcan be used as a reference when comparing the two graphics.
41
Figure 3.13: Current, measured during polarization of a ME at 300 C. The voltagewas decreased (solid lines) and increased (dashed lines) incrementally in 100 mVsteps, going from 0 V to -2 V and back to 0 V. For the sake of readability, onlyselected data is shown. The solid and the corresponding dashed line of each colorapproach the same steady-state-current.
periments were compared (s. Fig. 3.14 & Fig. 3.17), revealing the relation between
polarization voltage and steady-state current. Experiments were carried out under
Ar atmosphere at 300 C. Fig. 3.14 shows the voltage range between -0.5 V and 0.5
V, measured on 4 microelectrodes. with current values in the scale of pA. Exem-
plary current-time curves for two electrodes are shown in Fig. 3.15. For plotting the
steady state I-U-curve of Fig. 3.14, the mean current values of the last 10 minutes of
each voltage step was used. Despite dierences we can state that for all microelec-
trodes the currents are in the few pA range. It is also obvious that there is a voltage
intercept present for extrapolating the curves to zero current. This oset between
50 and 250 mV might partly be due to thermovoltages but might also include a
contribution from any built-in chemical potential dierence of Li.
It is also obvious that the curves are non-linear. Assuming electronic conductivity
as the reason of the current ow, this would be not surprising. Supposed the counter
electrode were hardly polarized (due to its large size) the voltage in steady state
42
translates to a chemical potential dierence of Li beneath the microelectrode. This
should aect the local electron concentration and thus the electronic conductivity.
Figure 3.14: Steady-state current plotted against the polarization voltage, rangingfrom - 0.5 V to 0.5 V at 300 C. Each color represents a single MEs, which wascontacted for polarization. The dierent symbols indicate the dierent measurementseries.
The black boxes show hysteresis, indicating, that the stead-state values obtained
from the increasing part of the series (upper branch) are higher than the ones ob-
tained during decrease of voltage (lower branch). Thus, the steady-state was not
approached closely enough to obtain accurate steady-state current data. However,
since the hysteresis is small (< 1 pA), the data can be used as estimates. The
red diamonds (bottom half lled) show almost no hysteresis at all, suggesting that
steady-state conditions were approached closely during the experiments. Similarly,
the green circles in the negative voltage range show only small hysteresis. Extrap-
olation between the negative and positive voltage data, suggest a current oset of
about -2 pA, which can be attributed to thermoelectric eects within the measure-
ment setup (s. Sec. 3.2.2 and Sec. 3.2.3). The red diamonds (top half lled) are
43
Figure 3.15: Current, measured during polarization of two individual MEs (a) and(b). Steady state values of each voltage step were estimated by taking the averagecurrent of the last 10 minutes of each step. The steady state values of (a) correspondto the red diamonds (bottom half lled) in Fig. 3.14 whereas (b) corresponds to thegreen circles.
from a measurement series which does not include voltage steps going back towards
0 V. Thus, the reliability of the steady-state currents cannot be deduced from any
possible hysteresis. However, the voltage increments as well as the duration of each
voltage step was the same as in the positive voltage range, which gave almost no hys-
teresis. Therefore the red diamonds in the negative voltage range can be expected
to be close to the true steady-state values.
The blue symbols deviate most from a linear I-U relation. The current data
obtained during the positive voltage measurement series is shown in Fig. 3.16.
Starting at 100 mV, the voltage was increased in 20 mV steps. The rst polarization
period was longer (1 hour) to ensure steady-state conditions, since the rst voltage
step was the highest (from 0 V to 100 mV). The subsequent periods were shorter
(10 minutes) since the increments were only 20 mV. From Fig. 3.16, however,
it is apparent, that the 10 minute periods are insucient to obtain steady-state
conditions. Especially at higher voltages, the long time, required to achieve steady-
state conditions, makes measurements of steady-state currents very time consuming.
For obtaining an estimate of the electronic conductivity we used extrapolated slopes
analyzed for U=0 V. We get (1-2 pA)/100 mV and thus a resistance of 50...100 GΩ.
This corresponds to 0.5 - 1·10−9 S/cm. This is slightly lower than the estimate of 400
C (2.5·10−9 S/cm) which might reect the true thermal activation but could also
44
be caused by the problems caused by only apparent steady states, non-linearities
and voltage osets.
Fig. 3.17 shows selected U-I data from a broader voltage range (-2 V to 2 V).
Since higher voltages were applied to the sample, the polarization time periods were
increased (typically 1 hour) to ensure steady-state conditions. In the positive voltage
part, the linear relation between voltage and steady-state current (cf. Fig. 3.14), is
continued up to 2 V. In the negative voltage part, the U-I relation at lower voltages
(s. Fig. 3.14) is continued similarly. However, at about -800 mV, a drastic in-
crease in current is apparent. The current data up to -1 V (red circles) show almost
no hysteresis, thus indicating that steady-state conditions were approached closely
during each polarization period. Therefore, the strong increase of current at about
-800 mV cannot be attributed to insucient polarization time. Rather, it suggests a
change in conduction mechanism at about - 800 mV. At voltages below -1 V (black
boxes), high negative currents are obtained. However, the currents show irregular
uctuations and do not approach steady-state within the polarization time (1 hour).
Steady-state currents could therefore only be estimated. The mean value of the last
10 minutes of each polarization period were used as rough estimates. After the
minimum voltage of -2 V was reached, the voltage was increased again (approach-
ing 0 V). The current data shows strong deviations from the data obtained when
the voltage was decreased towards -2 V (i.e. hysteresis). This strongly suggests
irreversible chemical changes in the sample. Hence precesses may have started that
correspond to those found for macroscopic electrodes in Sec. 3.1. The still dierent
current densities for the microelectrodes (1.3·10−6 A/cm2) and macroscopic elec-
trodes (1.6·10−4 A/cm2) may be caused by either still higher voltage or the higher
temperature.
3.2.6 LIBS
The impact of ME-polarization on the local composition of the sample was investi-
gated via LIBS. 8 dierent MEs were polarized (voltages between -2 V and 2 V) for
1.5 hours under Ar atmosphere at room temperature. After polarization, the rela-
tive, local chemical composition was determined via LIBS. Comparison between the
8 MEs and 4 MEs that were not polarized, showed no signicant dierence in chem-
45
Figure 3.16: Current measurements corresponding to the blue triangles (bottom halflled) in Fig. 3.14. The distinct negative slope at the end of some polarization stepsindicates, that steady-state was not approached closely enough to obtain reliablesteady-state currents. The voltage increments were 20 mV. Since the rst polariza-tion had a greater step (from 0 V to 100 mV), the polarization time was 1 hour,whereas the subsequent polarizations were only 10 minutes each.
ical composition between any of the MEs. However, since the LIBS measurements
were not done directly after polarization of each ME, possible eects of polarization
could have been reversed before the LIBS measurements.
46
Figure 3.17: Steady-state currents between -2 V and 2 V (a) and a magnied view ofthe range from - 1V to 2 V (b) at 300 C. The black boxes indicated rough estimatesof steady-state current values.
47
Part II
Sputtering and Testing of LIB
Electrode Materials
48
Chapter 4
Construction of a Sputtering Device
A spherical vacuum chamber (240 mm diameter), equipped with several anges of
dierent diameters, serves as the basis of the device (s. Fig. 4.1 and Fig. 4.2). The
chamber is mounted on an aluminum frame. A radio-frequency (RF) sputtering
gun (Gencoa 3G Circular Magnetron) is passed through the top ange alongside
with a shutter. The gun is connected to a cooling water cycle, the RF-power source
(Seren, R 301 RF Power Supply) and gas lines (argon and oxygen). The gases are
let into the chamber directly at the sputtering target. The gas ow of argon and
oxygen can be controlled independently by mass-ow-controllers (Tylan, FC 260 for
Ar and FC 2900 M for oxygen). A turbo molecular pump (Leybold Turbovac 250 iX)
combined with an auxiliary scroll pump (Agilent Technologies, SH-110 Dry Scroll
Vacuum Pump) is connected to the chamber via one of the side-anges. Two pressure
sensors (Leybold, Ceravac CTR 100 N and Penningvac PTR90 N) are installed via
another ange to monitor the chamber pressure. Wiring for heating, thermocouples
and possible in-situ measurements are passed through another ange.
The sample stage holds an alumina sample holder which can also serve as a heat-
ing unit. The sample holder is equipped with a type S thermocouple for temperature
control. So far, the highest tmeperature used in operation was 640 C. Sputtering
crystalline thin lms often requires elevated deposition temperatures to aid crystal-
lization. At room temperature, especially ceramic materials tend to form thin lms
with poor crystallinity. The target to sample distance of the device is variable. For
the course of the experiments presented in this work, the distance between target
and sample was set to 6 cm.
49
Figure 4.1: Overview of the sputtering device with labels marking the most essentialparts.
4.1 Preparation and Characterization of Thin Films
First tests of the sputtering device were done with a copper target. Sputtering of
metals is relatively simple, since no reactive atmosphere is required. Further, the
targets can be simply prepared from sheet metal. Since metals are ductile, there is
no danger of metal targets cracking or breaking due to thermal expansion during
sputtering.
Copper was RF-sputtered onto polished YSZ substrates in 3·10−2 mbar argon
atmosphere at room temperature. The substrates were partially covered with a
shadow mask to obtain a relief after deposition. Sputtering power was between 50
W and 110 W and the deposition time was between 5 and 15 minutes. The thickness
of the thin lms was measured with a prolometer (Bruker, Dektak XT) by scanning
over the relief surface.
Targets for deposition of ceramic materials were purchased from Beijing Loyaltar-
gets Technology Co. Active materials for LIBs, i.e. the cathode material LiMn2O4
(99.9 % purity) (LMO) and the anode material Li4Ti5O12 (99.9 % purity) (LTO)
were deposited onto YSZ substrates similarly to the copper thin lms, as described
above. However, the sputtering atmosphere was dierent, since oxide materials re-
50
Table 4.1: Deposition parameters of the samples prepared for structural investiga-tions.
quire a reactive atmosphere. In pure Ar, oxides would be partially reduced during
deposition, resulting in undened stoichiometry and undesired materials properties.
Deposition was done at room temperature rst, since the heating unit had not yet
been installed at that time. Thickness of the lms was determined after deposition
as described earlier.
Further samples were prepared in order to investigate structural and electro-
chemical properties of the oxide thin lms. LTO and LMO lms were deposited
onto silicon wafers and quartz glass substrates at room temperature for structural
investigations (see Table 4.1 for deposition parameters). Crystal structures of the
samples were determined by means of powder X-ray diraction (XRD) in Bragg-
Bretano geometry (PANalytical, X'Pert Pro, Cu K-α, 45 kV/40 mA). Quartz glass is
amorphous and the used Si-wafers are cut in [700]-orientation, thus giving no Bragg-
reections in the measured angle range. After deposition and XRD measurements,
the lms samples were subjected to a thermal treatment (annealing) to improve
crystallinity. The samples were heated in air to 500-700 C for 8-12 hours. After
annealing, the structures of the samples were again measured via XRD.
All electrochemical procedures described below were done using a Novocontrol Al-
pha Analyzer-Pot/Gal 30V-2A. Samples for electrochemical characterization of the
sputtered thin lms were prepared using LLZTO single crystals (s. Sec. 2.1) and
polycrystalline LLZTO pellets of the same nominal composition. Sputtering param-
eters were similar to the parameters described in Table 4.1. Thin lms of LMO (60
nm) were sputtered symmetrically onto a polycrystalline pellet and subsequently a
gold layer (50 nm) was sputtered on top of each side. Prior to and after an annealing
step (3 h at 700 C) for crystallization, the sample was characterized by means of
EIS. Further, the sample was electrochemically cycled repeatedly.
51
Table 4.2: Deposition parameters of the asymmetric sample for electrochemicalcharacterization.
CompoundTime(min)
Power(W)
Temperature(C)
Thickness(nm)
Pressure(mbar)
pO2(mbar)
Li4Ti5O12 20 80 600 31 1.2 10−3 2.0 10−3
LiMn2O4 100 50 400 20 2.0 10−3 3.0 10−3
An asymmetrical sample was prepared using a LLZTO single crystal. LMO was
deposited onto one side and LTO onto the other side by reactive RF-sputtering
at elevated temperature. Deposition details are shown in Table 4.2. Thin gold
layers (20 nm) were sputtered on top of each side as current collectors. The sample
was characterized by EIS and DC-cycling. Film thickness was determined at a
prolometer. Comparison with the thicknesses of layers, which were deposited at
room temperature (s. Table 4.1), shows, that the deposition rates are lower at
elevated temperatures. This can be attributed to a higher density of the lms, since
crystallization and densication is promoted at elevated temperatures.
52
Figure 4.2: From top left to bottom right: Overview; side view, revealing pressuresensors and turbo molecular pump; view inside the chamber during deposition ofLTO, revealing the wiring of the sample stage; sample during deposition.
53
Chapter 5
Results and Discussion
5.1 Film Thickness and Deposition Rates
First deposition rates were determined for sputtering copper at room temperature
in pure Ar atmosphere. The measurement results from the prolometer are shown
in Fig. 5.1. Naturally, longer deposition times and higher sputtering power results
in thicker lms. Roughly, deposition rates can be obtained by extrapolation of the
blue circles and black boxes. However, the lines passing through either one of the
data sets do not intersect the axes at 0/0 but rather intersect the time-axes at about
1 or 8 minutes. Since no shutter had been installed at the time of deposition, it was
not possible to sputter the target with closed shutter prior to deposition. Sputtering
with closed shutter for several minutes is a common procedure in many sputtering
processes to ensure the sputtering target is cleaned of any surface impurities that can
aect the deposition. Since this was not possible at the time of the rst sputtering
experiments, the oset in Fig. 5.1 can be attributed to the missing sputtering
procedure with closed shutter. It is apparent, that the oset value is clearly higher
at lower sputtering power. This results suggests, that detrimental surface impurities
of the copper target are removed faster at higher sputtering power.
Sputtering oxide thin lms generally results in lower sputtering and deposition
rates. Further, oxide materials require a reactive atmosphere for sputtering to cir-
cumvent possible reduction reactions of the sputtered material. Results from pro-
lometer measurements of the oxide lms are shown in Fig. 5.2. LTO deposition
rates at 80 W sputtering power are clearly higher than deposition rates of LMO at
54
Figure 5.1: Film thickness of various Cu thin lms deposited onto polished YSZsingle crystals. Sputtering power as well as the deposition time were varied.
Table 5.1: Deposition rates of Cu and the two oxide materials LMO and LTO.
CompoundPower(W)
Deposition Rate(nm min−1)
Pressure(mbar)
Cu 50 27 3 10−2
90 39 3 10−2
100 41 3 10−2
110 43 3 10−2
LTO 80 50 2 10−3
LMO 50 20 3 10−3
50 W. The blue circles, indicating the LTO lm thickness, show a linear dependence
on the deposition time, as expected.
Deposition rates for Cu, LTO and LMO were estimated from the lm thicknesses
and corresponding deposition times, assuming a linear relationship. The estimated
deposition rates are shown in Table 5.1. The deposition rate of LTO is unexpectedly
high, even higher than all measured deposition rates of Cu. This could be explained
by a lower total pressure during deposition, since higher pressure during sputtering
leads to more collisions between the sputtered material and the gas atmosphere,
thus reducing the deposition eciency. Further, the deposition rates presented in
55
Figure 5.2: Film thickness of various LMO and LTO thin lms deposited onto aSi wafer and quartz glass substrate respectively. Sputtering power as well as thedeposition time were varied. All lms were deposited at room temperature.
Table 5.1 are rough estimates, relying on 3 or less data points for each rate. More
data is required to obtain reliable deposition rates.
5.2 Crystal Structure
The crystal structure of the oxide thin lms was investigated by means of XRD.
The measurements were done in Bragg-Bretano geometry, scanning from 15 to 70
2θ at a step size of 0.020. Each step was scanned for 160 seconds. Since the
thickness of characterized lms is in the nanometer range, long scanning times were
necessary to obtain enough signal from the thin lms. As discussed earlier, thin
lms deposited at room temperature tend to have amorphous or poorly crystalline
structures. Therefore, thin lms are often subjected to thermal treatments after
deposition. X-ray diractograms of the LMO and LTO lms are presented in Fig.
5.3 and Fig. 5.4, respectively. Each gure includes diraction data from the lms as-
deposited and post-annealing. From both gures, a distinct increase of crystallinity
due to the annealing is apparent.
The as-deposited data, presented in Figure 5.3 (black line) shows no dened
56
Figure 5.3: X-ray diractograms of the LMO thin lm, deposited on quartz glass.The black line shows the data from the as-deposited scan, whereas the red lineindicates the data obtain post-annealing. A distinct increase of crystallinity dueto the thermal treatment is apparent. Due to crystallization of the quartz glasssubstrate, reexes of crystalline quartz (*) are visible.
reexes and a large hump in the region at about 20, typical for amorphous samples.
After the annealing (red line), several reexes are apparent in the diractogram.
The reexes marked with an asterisk correspond to the crystal structure of quartz,
whereas the reexes marked with the letter "s" correspond to the crystal structure
of the spinel phase of LiMn2O4. The quartz reexes can be explained by partial
crystallization of the substrate during annealing. Recrystallization of glasses at
elevated temperatures is a well-known phenomenon. Further, oblique inclusions were
apparent inside the quartz glass substrate after annealing. The other reexes can
be clearly attributed to the desired spinel phase of LiMn2O4. However, despite the
long scanning time, the reex intensities are low compared to the amorphous hump,
57
Figure 5.4: Diractograms of the LTO sample, as-deposited (black) and post-annealing(red). The sample shows no sharp reexes as-deposited but only one broadreex at about 43 . Post-annealing, more reexes are apparent and the reex atabout 43 is sharper than before. However, the few reexes could not be assignedto a single phase denitely.
therefore not all reexes of the desired LMO phase are apparent in the diractogram.
For a complete structural analysis of such thin samples, special XRD techniques,
such as grazing-incidence XRD would be required, which is beyond the scope of this
work.
The diraction data of the LTO sample is shown in Fig. 5.4. Only one, broad
reex is apparent in the as-deposited diractogram, whereas post-annealing, 4 re-
exes are visible. Again, the annealing step increased the crystallinity of the sample,
however, no denite structural assignment could be done. The reexes could be as-
signed to various phases of LTO. The weak and broad reexes do not allow for a clear
structural assignment. Possibly a mixture of various titanate phases was present in
the lm. Thicker lms or more advanced XRD techniques would be required for an
unambiguous structural investigation.
58
5.3 Electrochemical Properties of Sputtered Thin
Films
A symmetrical sample, consisting of a polycrystalline LLZTO pellet (1 mm thick,
cross section about 50 mm2) with LMO thin lms (60 nm) and Au contacts (100
nm) deposited onto both sides was prepared (details s. Sec. 4.1). Further, an
asymmetrical sample, consisting of an LLZTO single-crystal (7 mm2, 1 mm thick),
sputtered with LMO (20 nm) and LTO (31 nm) thin lms and Au contacts (20 nm)
was electrochemically characterized via EIS and DC-cycling.
5.3.1 Impedance Spectroscopy
The symmetrical polycrystalline LMO sample was characterized via EIS from 1 MHz
to 100 mHz as deposited and after an annealing step, since the lms were deposited at
room temperature. Fig. 5.5. shows impedance spectra of the as-deposited sample,
measured at room temperature and at 7.8 C as well as an impedance spectrum
of the annealed sample, measured at room temperature. Since the ionic mobility
of the sample is enhanced at higher temperatures, the overall impedance of the
sample is lower at room temperature. The as-deposited measurements show two
semicircles. At room temperature, the minimum between the high frequency and
low frequency semicircle is located at about 3 · 105 Ω. At 7.8C, the two semicircles
partially overlap, giving a relative minimum at about 1.6 ·106 Ω. The high frequency
semicircles can be attributed to the ionic conductivity of the electrolyte, since the
capacitance of the feature is in the range of 10−11 F. With the dimensions of the
electrolyte (50 mm2 by 1 mm), a relative electric permittivity of about 20 can
be estimated, which ts in the general range of oxide materials. For as-deposited
electrodes, the measured resistance of the electrolyte is high and therefore the specic
conductivity low (< 1 µS/cm), which is far lower than expected from a LLZTO
electrolyte. The low conductivity might be caused by bad electric contact between
the electrode materials and the electrolyte. The low frequency features of the spectra
can be attributed to processes at the electrode-electrolyte interfaces. However, the
corresponding capacitance (about 10−7 F) is lower than expected for a chemical
capacitance (i.e. (de)intercalation of Li-ions). Post annealing, the overall impedance
59
of the sample decreased signicantly. The onset of the electrolyte semicircle (s.
inset) is at about 2 kΩ, which corresponds to an electrolyte conductivity of about
0.1 mS/cm, a typical value for LLZO materials. The increased conductivity of
the electrolyte is attributed to better electronic contact between the electrodes and
the electrolyte, since the thermal treatment is not expected to change the specic
conductivity of the LLZTO. The low frequency feature consists of a shoulder and an
arc which does not show any tendency to reach the real axis. Rather, this electrode
arc seems to be followed by an additional arc with probably very high capacitance.
The asymmetric LTO/LLZO/LMO sample was characterized by EIS at room
temperature under argon atmosphere (s. Fig. 5.6). The displayed data was ob-
tained in the frequency range from 1 kHz to 1 mHz. At about 1.6 kΩ, a minimum
is apparent in the high frequency region (s. inset), indicating the impedance of the
electrolyte. Given the dimensions of the electrolyte, this corresponds to a conductiv-
ity of about 1.2·10−4 S/cm. A second semicircle arises from the minimum, ranging
up to 7·107 Ω in the low frequency part. This feature is attributed to processes
at the electrode/electrolyte interfaces due to its small capacitance, which is about
6·10−8 F (for 7 mm2 sample size). Then an additional capacitive contribution seems
to be present (vertical spike). A preliminary equivalent circuit (s. Fig. 5.6) t leads
to 7·10−5 F ·cm−2 which is realistic for a chemical capacitance of a 20 nm layer when
electrodes are in a charging state where large voltage increments correspond to small
charge transfer (e.g. close to LixMn2O4, x = 1). Since the cell was deposited from
a target with x=1, this is a plausible explanation for the small capacitance of the
electrodes.
5.3.2 DC Cycling
The annealed polycrystalline symmetric LMO/LLZTO/LMO sample was further
characterized by potentiostatic DC-cycling. The sample was cycled at room tem-
perature under air. The results are shown in Fig. 5.7. Positive currents arising from
positive voltages are apparent, which are decreasing after the voltage change. De-
creasing the voltage again corresponds to a discharging of the sample, which results
in negative currents. Like the charging currents, the discharging currents are de-
creasing in magnitude at constant voltage. The currents measured during charging
60
Figure 5.5: EIS spectra of the LMO/LLZO/LMO sample measured as-deposited atroom temperature (red) and 7.8 C (blue) and after an annealing step (black) atroom temperature. The equivalent circuit used for estimation of the capacitance ofthe low frequency feature is displayed. A magnied part of the diagram is shown inthe inset.
and discharging of the symmetrical cell are decreasing over time because the migra-
tion of Li-ions and electrons from the anodic to the cathodic side of the sample leads
to a partial compensation of the external electric eld. When voltage is applied, the
Li electrochemical potentials inside anode and cathode are shifted. Due to this gra-
dient, Li ions (and electrons) migrate from anode to cathode, thus compensating the
electrochemical potential gradient. Therefore, the driving force for ion and electron
migration decreases over time. The charging/discharging curves show values in the
range of µA, which is far higher than the current curves obtained from polarization of
LLZTO with blocking electrodes (cf. Part I). Since Li-ions can be (de)-intercalated
out of and into the electrodes relatively easy, ion migration and thus charge transfer
is achieved more easily compared to blocking cells. The resistance estimated from a
voltage of 1 V and a current of 0.12 µA is ca. 8 MΩ and thus, larger than the value
found in the impedance measurements for 1.5·105 Hz (43 kΩ). This also indicates a
61
Figure 5.6: EIS spectrum of the LTO/LLZO/LMO sample measured at room tem-perature under argon. A magnied view of the high frequency region is shown inthe inset.
further process at low frequencies in the impedance (see above). The charge own
in 1 hour is approximately 2·10−4 C corresponding to about 1.3·1015 Li atoms. A
60 nm LMO layer on a 50 mm2 LLZTO contains about 6·1016 Li ions and thus a
signicant amount of charging has occurred already after 1 hour for 1 V and even
more for 1 hour at 2 V (another ca. 5·1015 Li atoms). The subsequent discharging
(at 1 V and short circuiting via 0 V) seems to be strongly hampered since almost no
current ows. However, since negative currents arise during these discharging steps,
it can be concluded that some chemical charging has occurred during the charging
steps at 1 V and 2 V. Possibly the open circuit voltage is still low (due to a plateau
in the discharge curve) and the driving force for discharge is very small.
The asymmetric sample (LTO/LLTZO/LMO), was also characterized via DC
cycling under argon atmosphere at room temperature. Results from potentiostatic
measurements are shown in Fig. 5.8. Typical current proles arise from the poten-
62
Figure 5.7: Results from DC cycling of the polycrystalline symmetric LMO sample.The applied voltage (blue) is shown with the corresponding current (red line).
tiostatic cycling at both positive and negative voltages. The LTO electrode of the
cell was connected to the negative terminal of the voltage source. Therefore, a posi-
tive voltage corresponds to the usual polarity of a cell, that uses a LTO anode versus
a LMO cathode. However, negative voltages correspond to a deeply discharged state
of both electrodes, i.e. LTO depleted of Li to have less than the usual minimum 4 Li
per formula unit and LMO lled to more than the usual maximum of 2 Li per for-
mula unit. Therefore, the negative voltage steps of the potentiostatic DC cycling,
polarize the cell far below the usual minimum of LMO/LTO cells, which can be
accompanied by irreversible phase transitions and other detrimental side reactions
in the electrodes. Despite the relatively high negative voltage, the cell shows no
distinct signs of degradation caused by deep discharging.
The asymmetric cell was subjected to galvanostatic DC cycling under argon
atmosphere at room temperature (s. Fig. 5.9). The results show repeated charging
and discharging of the cell with currents of ±100 nA. Note, that 1 minute at 100 nA
corresponds to 4.8·10−2 Li per formula unit in a 20 nm layer of LiMn2O4 with a 7
mm2 cross-section. Positive current values correspond to a charging of the cell, which
require positive voltages. After each charging and discharging step, the open circuit
voltage (OCV) was measured by setting the current value to zero for 10 minutes.
The results show positive, but decreasing OCV values after positive polarization
63
Figure 5.8: Results from potentiostatic DC cycling of the asymmetricLMO/LLZTO/LTO sample at room temperature. The applied voltage (blue line)is shown with the corresponding current (red line).
and negative, increasing OCV values after negative polarization of the cell. The
high, negative voltages which are applied to meet the negative (i.e. discharging)
currents indicate that the small OCV voltage of the cell after charing, is not a strong
enough driving force to give the demanded discharging current, thus requiring an
additional driving force, i.e. negative voltage. Considering the negative OCV after
the negative polarization steps, the high negative voltages lead to a negative and
therefore inverted polarization of the cell. This inversion of polarity can only be
realized in a LMO/LTO cell, when the electrodes are discharged beyond the usual
cycling limits, being Li4Ti5O12 and Li2Mn2O4, see above. Discharging of the cell
beyond common limits is known to lead to degradation of the cell. Indeed, some
degradation of the cell is apparent in Fig. 5.9, however, the cell can be cycled
repeatedly.
After the galvanostatic cycling described above, the sample was further cycled
at 80C under Ar. Since the Li-ion transfer resistance of the interfaces and all
transport resistances are lowered, more distinct signs of reversible, electrochemical
(dis)charging of the sample were expected from cycling at an elevated temperature.
The sample was, again, subjected to galvanostatic cycling, with ± 100 nA. Between
each 10 minute (dis)charging step, the OCV of the cell was measured for 10 minutes.
64
Figure 5.9: Results from galvanostatic DC cycling of the asymmetricLMO/LLZTO/LTO sample at room temperature. The applied current (red line)is shown with the corresponding voltage (blue line). The voltage limit was set toabout 3 V. Therefore, when this limit is reached during cycling, the current dropsbelow the set value of ± 100 nA.
The results are displayed in Fig. 5.10. The last cell voltage (OCV) at 25 C was
about -1.5 V and not surprisingly the cell thus started at negative voltages (about
-1.1 V) also at 80 C. Surprisingly, however, the 100 nA current did no longer suce
to get positive voltages, as after 10 minutes.Both charging and discharging as well
as the open-circuit parts of the cycling show negative voltages. The OCV after
positive and negative current steps dier by about 0.5 V, lying in the range from
-1 V to -1.8 V. Both OCVs are increasing from cycle to cycle, indicating that some
irreversible reactions occur alongside the reversible dis/charging of the sample. The
results prove, that partially reversible electrochemical cycling of an LTO/LMO type
cell is possible in a negative voltage regime.
Cyclic voltammetry (CV) was carried out (Ar, 80 C) to further elucidate the
state of charge of the sample (sweeping rate = 0.1 mV/s). Since the sample was
found to be in a state of inverse polarity (i.e. -1.4 V) after the last measurements (s.
Fig. 5.10), the rst CV sweep started at -1.4 V, going up to 2.5 V. After the rst
sweep, the sample was cycled between 0 V and 2.5 V. The results are displayed in
Fig. 5.11. The current density arising during the rst sweep is clearly dierent from
the subsequent cycles between 0 V and 2.5 V. This is expected, since the sample is
65
Figure 5.10: Results from galvanostatic DC cycling of the asymmetricLMO/LLZTO/LTO sample at 80C. The applied current (red line) is shown withthe corresponding voltage (blue line). The voltages required to meet the set currentvalues of ± 100 nA are signicantly lower than they were at room temperature (c.f.Fig. 5.9).
changed from negative polarity back to the usual positive polarity. After the sample
is brought back to positive voltages, the cycles show little dierences and share the
same basic features. In the charging parts of the cycles, a very broad and at reaction
peak is apparent at about 1.5 V. After this feature, a steep increase of current occurs.
In the discharging parts, the curves show a broad and at reaction peak at about
1.0 V. The combination of the peaks at 1.5 V and 1.0 V during (dis)charing suggest
a reversible redox reaction taking place during cycling, which is typical for this kind
of battery. However, the steep rise of current at the end of each charging sweep
does not have a corresponding negative peak, thus indicating irreversible reactions
taking place above 2 V. Despite the distinct irreversible reactions at the end of each
sweep, the cycling of the sample shows no strong signs of degradation.
At the end of the CV measurements, the sample was left at a potential 0 V. Sub-
sequently, galvanostatic DC cycling was carried out again with the same parameters
66
Figure 5.11: Results from cyclic voltammetry of the asymmetric LMO/LLZTO/LTOsample at 80C.
as before. The results are shown in Fig. 5.12. The general shape of the voltage curve
is similar to the results from previous galvanostatic measurements (c.f. Fig. 5.10).
However, after charging the sample, a positive OCV is retained over 10 minutes.
When the current is switched to - 100 nA, negative voltages of up to -0.5 V arise.
As described earlier, this is attributed to an insucient discharging current of the
sample, which is compensated by application of negative bias to the sample. After
the negative current steps, the OCV of the sample is negative, quickly approaching
0 V. The results show clearly, that the sample was electrochemically charged, re-
versibly, at positive voltages. Thus proving, that the cell was successfully brought
back to the original polarity of LMO vs. LTO. The OCVs measured after charging
of the sample are between 0.5 and 0.8 V, increasing slightly from cycle to cycle.
Considering the redox peak from Fig. 5.11, the results suggest that the sample was
not charged long enough to reach the voltage plateau of the charging curve at about
1.5 V.
67
Figure 5.12: Results from galvanostatic DC cycling of the asymmetricLMO/LLZTO/LTO sample at 80C. The applied current (red line) is shown withthe corresponding voltage (blue line).
68
Chapter 6
Conclusion
The results from Part I show that the fast Li-ion conductor LLZTO can be sub-
jected to DC voltages of several 100 mV via blocking electrodes without irreversible
reactions occurring. For Au microelectrodes, this leads to electronic steady state
currents. These measurements enable estimates of electronic conductivities and val-
ues in the range of 10−9 S/cm are found for 300-400 C. At higher voltages, however
decomposition of the material occurs. In macroscopic samples this leads to some
local conductivity changes close to the anode. Moreover, drastic chemical changes
were found by LIBS at the anode (Li depletion) and at the cathode (Li accumu-
lation). Moreover, morphological changes take place for higher voltages: Bubble
formation at the anode due to oxygen evolution (i.e. in sum Li-oxide loss) and
growth of an additional phase at the cathode (probably Li2CO3). In experiments
with polarized microelectrodes, on the other hand, irreversible changes are reected
by irreversible current changes.
The results of the second part show that the construction of a sputtering de-
vice for oxide materials was successful. Deposition rates of various metallic (Cu)
and ceramic materials (LMO and LTO) were determined. Inuence of deposition
temperature on the structural and electrochemical properties of active electrode ma-
terials were investigated. The results show a clear correlation between crystallinity
of thin lms and their electrochemical properties. A positive eect of thermal an-
nealing on structural and electrochemical properties was demonstrated.
Sputtered thin lms were used to build all solid state cells with LLZTO sin-
gle crystals and polycrystals. Reversible, electrochemical cycling of an all solid
69
state sample (LMO/LLZTO/LTO) was demonstrated in both positive and nega-
tive voltage regimes. It was shown, that after deep-discharging of the LTO/LMO
cell, reversible electrochemical (dis)charging can be carried out over several cycles
at negative voltage. Further, the cell was successfully cycled repeatedly at positive
voltages, after the cell was subjected to deep discharging.
70
Bibliography
(1) Chen, H.; Cong, T. N.; Yang, W.; Tan, C.; Li, Y.; Ding, Y. Progress in Natural
Science 2009, 19, 291312.
(2) Leung, P.; Li, X.; Ponce de León, C.; Berlouis, L.; Low, C. T. J.; Walsh, F. C.
RSC Advances 2012, 2, 10125.
(3) Steele, B. C. H.; Heinzel, A. Nature 2001, 414, 345352.