Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Journal of The Electrochemical

Society

OPEN ACCESS You may also like


- Observational Support for Massive Black
Galvanostatic Intermittent Titration Technique Hole Formation Driven by Runaway Stellar
Collisions in Galactic Nuclei
Reinvented: Part II. Experiments Andrés Escala

- Observational Signature of Tightly Wound


Spirals Driven by Buoyancy Resonances
To cite this article: Stephen Dongmin Kang et al 2021 J. Electrochem. Soc. 168 120503 in Protoplanetary Disks
Jaehan Bae, Richard Teague and
Zhaohuan Zhu

- Galvanostatic Intermittent Titration


Technique Reinvented: Part I. A Critical
View the article online for updates and enhancements. Review
Stephen Dongmin Kang and William C.
Chueh

This content was downloaded from IP address 131.179.156.10 on 10/05/2023 at 08:29


Journal of The Electrochemical Society, 2021 168 120503

Galvanostatic Intermittent Titration Technique Reinvented: Part II.


Experiments
Stephen Dongmin Kang,1,z Jimmy Jiahong Kuo,1,2 Nidhi Kapate,1 Jihyun Hong,3
Joonsuk Park,1 and William C. Chueh1
1
Stanford University, Stanford, California 94305, United States of America
2
Northwestern University, Evanston, Illinois 60208, United States of America
3
Korea Institute of Science and Technology, Seoul, Republic of Korea

Following a critical review of the galvanostatic intermittent titration technique in Part I, here we experimentally demonstrate how
to extract chemical diffusivity with a modified method. We prepare dense bulk samples that ensure diffusion-limitation. We utilize
the scaling with t relax + τ − trelax (trelax: relaxation time; τ: pulse duration), avoiding problems with composition-dependent
overpotentials. The equilibrium Nernst voltage is measured separately using small porous particles. This separation between the
diffusion measurement and the titration procedure is critical for performing each measurement in a reliable setting. We report the
chemical diffusion coefficients of LixNi1/3Mn1/3Co1/3O2 and their activation energy. We extract ionic conductivity and compare it
with total conductivity to confirm ion-limitation in chemical diffusion. The measurements suggest that the time scale for diffusion
in typical Li-ion battery particles could be much shorter than that of the intercalation/deintercalation processes at the particle
surface (Biot number less than 0.1).
© 2021 The Author(s). Published on behalf of The Electrochemical Society by IOP Publishing Limited. This is an open access
article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/
by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/
1945-7111/ac3939]

Manuscript submitted September 10, 2021; revised manuscript received October 16, 2021. Published December 3, 2021.

The conventional practices1–5 of the galvanostatic intermittent temperature for layered oxides.7,8 One must therefore find optimum
titration technique (GITT) have produced drastically inconsistent recipes that consolidate the powder material while preserving the
chemical diffusion coefficients in the literature. The inconsistency not composition.
only ranges over many orders in magnitude but also shows conflicting Here we demonstrate the consolidation of NMC111 as a model
dependencies on composition. In Part 16 of this series contribution, we composition. We start with commercially available powder of NMC111
have identified multiple systematic error sources responsible for the porous agglomerates (Toda) with a size of approximately 10 μm,
inconsistency. We have also proposed how to eliminate or mitigate consisting of many sub-micron primary particles. We pulverize the
such errors by modifying the GITT method. The modification is largely agglomerates by mechanical ball milling. The powder was placed in a
two-fold: preparation of extremely diffusion-limited samples and zirconia jar together with 5 mm zirconia balls, and the jar was rotated
utilization of the trelax + τ − trelax time scale during relaxation (τ on a planetary ball mill (PM 100, Retsch) at 400 rpm for a total active
is pulse duration; trelax is time with 0 defined at current interruption). In milling time of 80 min. The pulverized powder was placed in a graphite
this sequel contribution, we experimentally demonstrate how to die, padded with flexible graphite foil to accommodate thermal
implement our modified GITT method, following the steps outlined in expansion and contraction. The graphite die was rapidly heated in an
Part 1.6 Using LiNi1/3Mn1/3Co1/3O2 (NMC111) as a model system, we induction furnace (ramp rate >50 K min−1) under Ar atmosphere while
obtain chemical diffusion coefficients (chemical diffusivities), their applying uniaxial pressure (40–50 MPa). We found that the sintering
activation energies, and ionic conductivity as a function of Li temperature was in the range 750 °C–800 °C for this material. Thus, we
composition. hot-pressed at 800 °C for a minimal amount of time (10 min) to
minimize material loss. This process limited the oxygen loss (perhaps
Preparation of Closed-Pore Bulk Samples keeping oxygen trapped inside the bulk), allowing for a post-annealing
recovery in stagnant air (800 °C, 1 h). We could not detect Li loss under
Enlarging the sample dimension is the most effective way to these conditions.
ensure that experiments are conducted in a diffusion-controlled Bulk samples consolidated with this recipe had high exterior-
regime. When the experiment is conducted using liquid electrolytes, volume density while maintaining identical crystallographic char-
the effective sample size depends on whether liquid can penetrate the acteristics of the starting powder material. The consolidated sample
sample. In porous materials, the effective length scale subject to shown in Figs. 1a–1b had an exterior-volume density of 93%
diffusion could be much smaller than the exterior dimensions of the (measured using the exterior dimensions of the sample) while
active material. The ideal sample is therefore a bulk sample with a having an Archimedes density of 95% (measured using a He gas
density high enough to prevent macroscopic penetration of the liquid pycnometer, which yields a value equivalent to the upper bound of
through open pores. Liquid penetration into bulk samples can be an Archimedes measurement), relative to the crystallographic
evaluated by comparing the exterior-volume density (apparent theoretical value. The measurements indicate that 5% of the volume
density) with Archimedes density. Density measured with the corresponds to closed-pores, while only 2% of the volume corre-
Archimedes method is also called “true” density, but this termi- sponds to open-pores exposed on the surface. This measurement is in
nology could be misleading in the present context because an contrast to the near 100% Archimedes density reported from some
Archimedes density that is 100% of the theoretical value could cold-press sintering processes.9,10 For reference, our attempts on
indicate an open-pore structure, which is an undesirable sample cold-press sintering (while not introducing sintering-aid impurities
geometry. and limiting the maximum temperature to 800 °C) typically resulted
The challenge in consolidating electrochemical battery materials in NMC111 pellets with 100% Archimedes density and 70%
is to achieve high exterior-volume density (>90%) while avoiding exterior-volume density. The powder XRD investigation shown in
the loss of volatile species. Oxygen (and also Li for some Fig. 1c indicates that the hot-press and post-annealed samples have
compositions) tends to have high vapor pressures near the sintering lattice parameters and Ni-Li antisite defects identical to that of the
starting powder, indicating minimal material loss. Note that more
aggressive pressing or sintering conditions11,12 have been reported to
z
E-mail: Stephen.D.Kang@gmail.com alter the diffraction patterns.12
Journal of The Electrochemical Society, 2021 168 120503

Figure 1. Consolidation of NMC111 powder into a closed-pore bulk sample. (a) Photograph of a consolidated bulk pellet with a diameter of 12 mm. Density
measurements indicate that the majority of the pores are closed pores. (b) TEM bright field image of the consolidated bulk sample. (c) Synchrotron powder XRD
comparing the starting powder (above) to that obtained from grinding the bulk sample (below). Powder from the bulk sample was diluted with silica powder to
limit the total absorbance of the specimen. Parameters obtained from Rietveld refinement were identical for both samples (R 3̄ m, a = 2.864 Å; c = 14.25 Å; O
position (z) = 0.259; Ni defects on Li site = 3%).

Minimizing the time and temperature of the hot-press and post- relaxation experiments as well. We confirmed with a three-electrode
annealing in air were important in preserving the sample composi- cell that the counter electrode contribution was negligible in this
tion after the consolidation process. After the 10 min 800 °C hot bulk-sample setting using a Li metal counter, which is attributed to
press, surface XRD from the as-pressed pellet indicated lattice the minimal surface area of the bulk sample dominating the overall
parameter shifts and peak broadening. However, after post-annealing cell impedance.
in air at the identical temperature, we were able to recover the
composition profile. Other possible routes of consolidation include:
GITT Experiment
hot-pressing in an oxygen environment (although with challenges
due to thermal expansion problems of oxygen-compatible dies); Pulse-relaxation experiment.—The pulse-relaxation experiment
using sintering aids to facilitate consolidation without pressure;12 for measuring chemical diffusivity is designed according to the
increasing the ramp rate to minimize the sintering time.13 principles established in Part I.6 Each current pulse is designed to
The Li composition of the bulk was controlled electrochemically have a minimal magnitude, a duration (τ) that is short enough to
using 316 stainless steel coin cells. The consolidated bulk samples ensure minimal perturbation of the surface composition, but still
were cut into smaller flat pieces typically with a thickness less than long enough to see the time-dependency originating from the bulk
0.5 mm. The flat sample was sandwiched between porous carbon process. Whether the pulse duration is within an acceptable range
sheets and placed in a coin cell. A stack of glass fiber filter sheets has to be later checked after chemical diffusivity values are extracted
was used as the separator, also serving as a compression element from the measurement.
instead of a wave spring. Li metal was used as the counter electrode, Here, we demonstrate an example GITT pulse-relaxation se-
and an ethylene carbonate/diethylene carbonate (1:1 vol.) mixture quence (Fig. 2). Starting with a homogenized bulk sample, a
with 1 M of LiClO4 salt was used as the electrolyte. This lithiation current pulse is applied (10 min), followed by a long
construction allowed the coin cell to be disassembled safely without relaxation period (5 h). This current pulse is extremely short
breaking the bulk sample. After delithiating the bulk sample by a considering the sample geometry: our sample thickness (typically
target capacity, the sample was retrieved from the cell, rinsed several about 0.5 mm) is 50 times larger than the size of agglomerate
times in dimethyl carbonate, and annealed in Ar at 85 °C–95 °C for particles used in porous electrodes, so the current pulse of 10 min is
7+ days (measured temperature; not setting temperature). The equivalent to a diffusion duration of 0.24 s in an agglomerate particle
delithiation-annealing process was repeated in small steps (5 to 10 (diffusion time is length squared). In both the lithiation curve (violet
percent of the total Li in one step) to avoid cracking of the bulk in Fig. 2a) and relaxation curve (turquoise in Fig. 2a), two distinct
sample. The identical coin cell structure was used for pulse- regions are observed. This behavior resembles the “early transient
Journal of The Electrochemical Society, 2021 168 120503

Figure 2. Example of a GITT sequence on a NMC111 bulk sample. (a) A 10 min lithiation pulse (violet, bottom axis, left to right for time progression) is
applied, followed by a 5 hour relaxation (turquoise, top axis, right to left for time progression). Because of the small number of charges passed through the pulse
relative to the sample capacity, the net composition shift is negligible as indicated from the absence of net change in open-circuit voltage. (b) At least three pulse-
relaxation sequences are repeated, varying the current pulse magnitude. The slope of interest is that related to the bulk relaxation process (chemical diffusion),
rather than the initial transient related to the surface. (c) By plotting the bulk relaxation slope obtained from (b) as a function of current pulse magnitude, one can
separate the Faradaic current from parasitic contributions.

effects” discussed in Part I, and thus we attribute the initial slope to a


surface property and the later slope to a bulk property. To separate
the Faradaic current from the parasitic current, the pulse-relaxation
is repeated at least three times with varying pulse magnitudes
(Fig. 2b). By plotting how the bulk relaxation slope
dV
( d ( t + τ − t ) ) changes with the applied current pulse magnitude
relax relax
(Ictrl), as shown in Fig. 2c, we find the linear scaling factor:
eq
dV 2 v M ∂VLi 1
= (−Ictrl + Ipar ). [1]
d ( t relax +τ − t relax ) π FS ∂xLi D˜

The x-axis intercept gives an estimation of the non-Faradaic parasitic


current (Ipar), the opposite sign of the Faradaic current with a
magnitude <0.5 μA in this example case. Sample surface area S can
be estimated from the sample geometry, F is the Faraday constant,
and the molar volume vM can be taken as the volume of one chemical
unit of LixNi1/3Mn1/3Co1/3O2 if we let xLi represent the Li content x
in this formula. Then, the only factor remaining to be determined is
∂V eq
the Nernst voltage dependency on Li content ( ∂xLi ), discussed in the
Li
next subsection.

Nernst voltage determination.—The Nernst voltage derivative


eq
∂VLi
∂xLi
is needed for converting GITT measurements to chemical
diffusivity. To facilitate equilibration of the Li composition,
agglomerate particles (≈10 μm, Toda) were used rather than bulk
samples. Composite electrodes were made with a dilute content of
active material, consisting of a weight ratio of 6:2:2 for NMC111,
carbon black, and polyvinylidene fluoride (minimizing electrolyte
effects). The final thickness of the electrode was <40 μm (excluding
the aluminum current collector) after fully drying the N-Methyl-2-
pyrrolidone solvent in a vacuum oven overnight at 60 °C.
The Nernst voltage curve obtained using half-cells with GITT is
shown in Fig. 3a, and the derivative with respect to Li composition
eq
∂VLi
yields the ∂xLi
curve shown in Fig. 3b. It is seen that the obtained
charging curves are identical for the 1st, 2nd, and 3rd cycles; the
starting Li composition was “reset” after each cycle by a long 2.5 V Figure 3. Determining the derivative of the Nernst potential (a) Open circuit
hold at the end of discharge for full lithiation.14 voltages obtained using GITT on NMC111 porous electrodes at 30 °C . The
The Nernst voltage curve can also be used to calibrate the solid markers represent the first charging curve. Using two other cells from a
different electrode, GITT open circuit voltages were also measured from the
composition of the bulk samples. To confirm the validity of this second cycle (+ markers) and third cycle (× markers). For these later-cycle
approach, we estimated the Li composition of annealed bulk samples cells, the cells were discharged with a long voltage hold at 2.5 V to return the
using the open-circuit voltage (OCV) and the calibration curve in starting composition to Li fraction 1.0. 14 The curves are identical within
Fig. 3a. Then, we pulverized the sample into powder with a mortar experimental resolution. (b) Derivative of the Nernst voltage curve obtained
and pestle (in an Ar-filled glove box) and made a capillary X-ray in (a).
Journal of The Electrochemical Society, 2021 168 120503

Figure 4. Chemical diffusivity of lithium extracted from GITT measure- Figure 6. Ionic conductivity calculated from chemical diffusivity measure-
ments on NMC111 bulk samples at 30 °C . Each marker color represents a ments on NMC111 bulk samples at 30 °C . Ionic conductivity is extracted
different bulk sample. The dashed line is a guide representing the trend of the assuming an ion-limited case, Eq. 2. Plotted together is the total conductivity
compiled data. The shaded area is the 0.95 confidence interval (t-distribution) measured using a four-probe method on bulk samples, confirming the ion-
for the trend line. limited assumption.

Next, we check the validity of the semi-infinite diffusion


condition.6 By taking the upper bound of the D̃ values from Fig. 4 as
2 × 10−9 cm2 s−1, size of the thinnest sample used in the experi-
ment as 2L = 0.426 mm, and the pulse duration τ = 10 min, the
D̃τ
upper bound of the dimensionless pulse duration τ̂ = L2 evaluates to
3 × 10−3. This value is 2 orders of magnitude smaller than the
criterion for 1D planar systems (τ̂ < 0.25), confirming the correct
design of the sample and pulse duration.

Activation Energies
At each Li composition, we do pulse-relaxation experiments at
three different temperatures (15 °C, 30 °C, and 45 °C) to extract the
Arrhenius activation energy for chemical diffusion. The differential
Nernst voltage curve (Fig. 3b) obtained from 30 °C can be used
identically at other measurement temperatures without introducing
meaningful errors. The temperature sensitivity of Nernst voltages
(VLieq ) at near full-lithiation (3.6 V vs Li) and half-lithiation (4 V vs
Li) is only different by 0.1–0.2 mV K−1. Therefore, the Li composi-
Figure 5. Arrhenius activation energy for chemical diffusion, extracted from ∂V eq
tion derivative ( ∂xLi ) will only change by <1% for a temperature
GITT measurements on NMC111 bulk samples. Error bars represent the Li
linear fitting errors from a log DT vs 1/T analysis. change as big as 20 K.
Activation energies are obtained by assuming an Arrhenius
diffraction sample. The Li composition was then estimated using the relation for DT, as shown in Fig. 5. This procedure is in anticipation
lattice parameters and the calibration curve reported in Ref. 14. For a that the activation energy of chemical diffusion will correspond to
bulk sample with an OCV-estimated Li fraction of 0.78, the XRD- that of a partial conductivity (ionic in this case, as will be confirmed
estimated Li fraction was 0.76. In another sample, the two estima- in the next section). Applying the Arrhenius relation to DT instead of
tions were 0.54 and 0.58, respectively. This consistency indicates the D yields activation energies about 26 meV larger.
Li composition can be inferred from OCV with reasonable un-
certainty. Insight from Measurement Results
Ionic vs electronic conductivity.—Chemical diffusion is an
Diffusivity extraction.—We do pulse-relaxation experiments ambipolar process where charge neutrality requires the concurrent
(Fig. 2) on bulk samples of various Li compositions, as prepared diffusion of two oppositely charged species. Using the chemical
through the procedures in section “Preparation of Closed-Pore Bulk diffusivity of NMC111, we can infer whether the Li ions or the
Samples”. Then, using the Nernst voltage derivative curve in Fig. 3b, electronic species are limiting for this bulk transport process. We
chemical diffusivities are extracted from Eq. 1. The results are assume that the ionic species are limiting for chemical diffusion, and
plotted in Fig. 4. Note that the relation between D̃ vs Li composition verify whether this assumption holds. When Li ions are limiting,
displayed in Fig. 4 is qualitatively different than any other literature chemical diffusivity can be expressed as (derived in the Appendix of
report compiled in Fig. 1 of Part 1.6 Part I):
Journal of The Electrochemical Society, 2021 168 120503

Sluggish kinetics near full lithiation?.—It has been argued in


σ ⎛ v ∂V ⎞ the literature that the sluggish kinetics near full lithiation in layered
eq
σ ∂μ
D˜ = i2 Li = i ⎜ M Li ⎟, [2] oxides is due to the slow chemical diffusion in those
F ∂c Li F ⎝ F ∂xLi ⎠ compositions,19–21 apparently supported by selected (conventional)
GITT measurements compiled in Fig. 1 of Part I.6 Although this
where σi is the ionic conductivity from Li ions, μLi is the chemical view has been refuted through the demonstration of reaction-
potential of neutral Li species, and cLi is the volumetric density of Li limitation for standard particle sizes,14 questions remain whether
species. Using this equation, ionic conductivity is calculated from sluggish chemical diffusion can be completely ruled out. Our
the chemical diffusivity and plotted in Fig. 6. To test the validity of chemical diffusivity measurements shown in Fig. 4 reveal that
the ion-limitation assumption, we also measure the total conductivity chemical diffusion is not the slowest near full-lithiation. Although
of bulk samples. We use a four-probe method to exclude contribu- ionic conductivity does seem to trend toward a minimum near full-
tions from contact resistances15 (two-probe measurements under- lithiation, the enhancement factor (thermodynamic factor) in Fig. 3b
estimate electronic conductivity by orders of magnitude). The dominates the trend of chemical diffusivity in this composition
measured bulk total conductivity is plotted together in Fig. 6. range. This observation is consistent with what is expected from
Total conductivity is found to be much larger than the ionic thermodynamics. Although Fick’s law seemingly suggests that
conductivity at all Li compositions where D̃ was measured. This concentration gradients drive diffusion, it is the chemical potential
confirmation validates the conversion of chemical diffusivity to ionic gradient that constitutes the real thermodynamic driving force. Near
conductivity using Eq. 2. It also indicates that the total conductivity full-lithiation, a given concentration gradient translates to a much
measurement is practically equivalent to an electronic conductivity larger driving force (hence the term “enhancement factor”). Overall,
measurement. Chemical diffusion in NMC111 is therefore limited our chemical diffusion measurements provide further evidence that
by ionic rather than electronic conduction. chemical diffusion is not responsible for the sluggish kinetics and the
limited access of capacity (i.e. kinetic capacity loss19,21,22) near full-
Is the activation energy from ion hopping?.—From the many lithiation. Meanwhile, charge transfer resistance of the intercalation/
factors that contribute to Li chemical diffusivity in NMC, which one deintercalation reaction plummets exponentially as Li composition
determines the temperature dependency? From the partial conduc- approaches full occupancy,14 further suggesting that the sluggish
tivity investigation, we can first rule out the contribution from kinetics at this composition is charge-transfer limited. In the next
electronic conductivity, leaving only the factors in Eq. 2. Then, we section, we quantitatively assess the rate-limiting step.
∂μ
can also rule out ∂c Li because of its minimal temperature dependency
Li
Is chemical diffusion a limiting process?.—Of high interest in
(see subsection “Activation Energies”). Therefore, the activation
the battery research community is whether chemical diffusion is a
energy in chemical diffusion can be attributed to that of ionic
limiting factor for the electrochemical reaction of active particles.
conductivity.
Inter-particle inhomogeneity discovered in porous electrodes14
The Nernst-Einstein relation between conductivity and self-
indicates that Li insertion and desertion in typical battery particles
diffusivity suggests that the activation energy be analyzed with
are limited by the surface rather than bulk processes. However, the
D̃T or σiT: uncertainty in diffusivity values and particle geometries have made it
zF 2 difficult to understand whether the limiting kinetic process would
D˜ (T ) ∝ σi (T ) = Dc, [ 3] change by tuning the particle size or surface properties. We address
RT this question using our measured chemical diffusivity values.
where z, D, c are the formal charge, self-diffusivity, and volumetric The ratio between a surface reaction rate and bulk diffusion rate
concentration of the species responsible for the ionic conduction. is a convenient metric for quantitatively assessing interface- and
The activation energies in Fig. 5b can now be attributed to the sum bulk-limited regimes in transport problems. If we let L describe the
of the formation energy of an ionic species and the diffusion barrier size of the bulk in a planar system and h describe a mass transfer
for that same species. coefficient at the surface (in the units of length per time), the surface
If we decide to describe the regular Li ions, then cLi is virtually reaction rate can be expressed as h/L (units of inverse time). The rate
fixed with respect to temperature, and the activation energy can be for diffusion can be expressed as D/L2 (units of inverse time), where
hL
entirely attributed to the self-diffusivity of regular Li ions (D Li+ ). D is diffusivity. The ratio of these rates is the Biot number, Bi = D ,
However, this designation does not offer much mechanistic insight, a dimensionless metric. By expressing h in terms of area-specific
because it is simply a nominal value ignorant of the mobile defect exchange-current density j0 (as in the Butler-Volmer (BV) relation),
species. Describing the mobile defect responsible for the Li ion self- we can use Bi for electrochemical reactions. In the linear BV regime
diffusivity provides better insight, but this process requires knowledge for Li intercalation,
about the mobile defect (which is often a minority species). From
computational studies16–18 it has been suggested that divacancies are j0 L ∂VLieq
responsible for most of the ionic conductivity in layered oxides. The Bi = . [4]
D˜ Li RT ∂c Li
simplest treatment of this case is to count the divacancy concentration,
say cVV, treat it as z = − 2, and account for the self-diffusivity of the Here, diffusivity has been replaced with the chemical diffusivity, VLieq
divacancies DVV. The activation energy for ion hopping in this is the Nernst voltage, and cLi is the concentration of the neutral Li
divacancy description has been calculated to be a monotonically species in the units of moles per volume. To apply Bi to porous
decreasing function with respect to delithiation, primarily correlated electrode measurements, in which mass-specific exchange current is
with the inter-slab distance between layers. However, the measured measured, we reformulate Bi by assuming spherical particles of
activation energy for chemical diffusion is not monotonic upon radius r:
delithiation, and the trend is opposite from Li fraction 1.0 to 0.7. If
one were to reconcile this observation with the divacancy model, this j0 ′MM ∂VLieq
non-monotonic trend of activation energy indicates that the formation Bi = r 2 . [ 5]
3D˜ Li RT ∂xLi
energy of divacancies, rather than that of ion hopping, dominates the
activation energy at Li fraction from 1.0 to 0.7. Alternatively, a Here, j0′ is in the units of current per mass, MM is the molar mass of
different defect model might be necessary to describe diffusion. the chemical formula corresponding to xLi = 1, and xLi is the Li
Overall, explaining chemical diffusion only with ion hopping does fraction (selecting r as the length scale for Bi results in a similar Bi
not seem to be justified based on our experimental measurements. criteria as in a planar problem).
Journal of The Electrochemical Society, 2021 168 120503

method also explicitly accounts for the chemical profiles developed


during the current pulse.
Summarizing the steps for chemical diffusivity measurements:

• Prepare bulk samples dense enough to prevent electrolyte


penetration and large enough to ensure diffusion limitation.
• Electrochemically set the bulk composition to the target
composition and anneal the sample to ensure homogenization.
• Confirm that the bulk sample preserves the chemistry, crystal
structure, and phase of interest.
• Repeat pulse-relaxation experiments for a given composition
and extract the long time-scale relaxation slopes through a V vs
trelax + τ − trelax analysis.
• Pulse duration should be kept short, but long enough to observe
the bulk-transport limited relaxation regime. As in Fig. 2a, the
portion representing the “bulk” should be visible in the relaxation
curve.
• A series of pulse-relaxation measurements should be conducted
Figure 7. Biot number for NMC111 agglomerate particles, comparing with varied pulse magnitude or direction, so that one can use the
different assumptions about the effective particle radii. The lower and upper scaling relation in Eq. 1.
limits of the effective particle size are determined by the physical size of the • To determine the Nernst voltage as a function of composition,
primary and agglomerate particles, respectively. Because of partial electro- a separate sample maximizing surface area and minimizing particle
lyte penetration, the realistic effective size is in the intermediate range. A size should be prepared (i.e. ensure short equilibration times). Pulse-
large Biot number (red shade) indicates that the kinetics are limited by
relaxation experiments should be repeated, similar to conventional
chemical diffusion, whereas a small number (green shade) indicates that the
kinetics are limited by the intercalation/deintercalation reaction. We use the titration measurements, not to extract time-dependency but to only
2nd-cycle mass-specific exchange current density as reported in Ref. 14, extract the equilibrium voltages at each composition.
where j0′ is measured with an LP40 electrolyte in the limit of a thin • Combining the Nernst voltage derivative and the bulk relaxa-
(≈20 μm) and dilute (40 wt.% active mass) electrode. tion slopes, extract the chemical diffusion coefficient.
• With the measured diffusivity, iteratively check for the validity
of the experimental design, including diffusion limitation in the bulk
sample and pulse duration.
Using the j0′ values for the Li intercalation/deintercalation reaction,
reported from porous electrodes made with agglomerate particles,14 Chemical diffusivity, ionic conductivity, and activation energies
we calculate the Biot number with three different assumptions about of NMC111 were measured using this modified GITT scheme.
the effective particle radius: r = 0.1 μm (approximate primary particle Measurement results appear to be consistent with a recent study on
radius); r = 5 μm (approximate secondary particle radius); r = 0.5 μm inter-particle inhomogeneity14 indicating that chemical diffusion is
(intermediate, assuming partial electrolyte penetration). The curves not the most limiting process in battery electrodes. In other words,
plotted in Fig. 7 show that even under the unlikely assumption23–25 the time scale for the intercalation/deintercalation reaction at the
that no electrolyte penetration or particle cracking occurs during the surface is longer than that of the bulk diffusion time scale (Bi < 0.1).
forming cycle (the r = 5 μm assumption), the kinetics would not have The remaining questions include how the polycrystalline micro-
been controlled by diffusion near full lithiation (Li fraction near 0.9 or structure such as grain size or grain boundary structure influences the
above). For a realistic assumption about the effective radius, kinetics bulk chemical diffusion. The methodology proposed and demon-
in these agglomerate particles would be either reaction-limited or mix- strated in this work could help study more advanced questions37–42
controlled in the entire composition range. We note that mix- about chemical diffusion in electrochemically intercalating
controlled kinetics also behave qualitatively similar to reaction-limited materials.
cases in terms of inter-particle inhomogeneity development.14
Therefore, we conclude that interface engineering26–34 either on the Acknowledgments
particle side or electrolyte side26,27 should be an effective strategy for
improving kinetics and rate capability. Both sides of the interface This work was supported by the Toyota Research Institute
control the exchange-current density.35,36 Our analysis also suggests through the Accelerated Materials Design and Discovery program.
that many studies aiming to improve the kinetics with the intention of The authors would like to thank: Devi Ganapathi and Xiao Cui for
engineering bulk diffusion properties may have succeeded due to a contributing to improving the clarity of the manuscript; Kipil Lim
correlation between bulk and surface properties. for assistance in XRD.
ORCID
Summary and remarks.—We have demonstrated that preparing Stephen Dongmin Kang https://orcid.org/0000-0002-7491-7933
dense bulk samples is an effective strategy for ensuring diffusion- Jimmy Jiahong Kuo https://orcid.org/0000-0002-3434-1984
limitation and improving the reliability of chemical diffusion Nidhi Kapate https://orcid.org/0000-0002-5286-3663
measurements. On the other hand, the titration process for measuring Jihyun Hong https://orcid.org/0000-0001-7210-2901
the equilibrium Nernst voltages works better with smaller particles William C. Chueh https://orcid.org/0000-0002-7066-3470
(as in porous electrodes for battery devices) where the time for
equilibration is much shorter. Because these two measurements References
(transient diffusion vs equilibration) have opposite sample geometry
requirements, it is easier to optimize two different types of samples 1. W. Weppner and R. A. Huggins, “Determination of the kinetic parameters of
mixed-conducting electrodes and application to the system Li3Sb.” J. Solid State
for each measurement. Chem., 22, 297 (1977).
For the transient diffusion measurement, we have demonstrated 2. W. Weppner and R. A. Huggins, “Electrochemical investigation of the chemical
how one can take advantage of the relaxation time variable diffusion, partial ionic conductivities, and other kinetic parameters in Li3Sb and
Li3Bi.” J. Electrochem. Soc., 124, 1569 (1977).
trelax + τ − trelax . This approach avoids the non-constant over- 3. W. Weppner and R. A. Huggins, “Electrochemical methods for determining kinetic
potential contribution to transient voltages. This relaxation analysis properties of solids.” Annual Review of Materials Science, 8, 269 (1978).
Journal of The Electrochemical Society, 2021 168 120503

4. A. Honders and G. H. Broers, “Bounded diffusion in solid solution electrode 24. S. Oswald, D. Pritzl, M. Wetjen, and H. A. Gasteiger, “Novel method for
powder compacts. Part I. The interfacial impedance of a solid solution electrode monitoring the electrochemical capacitance by in situ impedance spectroscopy as
(Mx SSE) in contact with a M+-ion conducting electrolyte.” J. Solid State Ionics, indicator for particle cracking of nickel-rich ncms: part i. theory and validation.”
15, 173 (1985). J. Electrochem. Soc., 167, 100511 (2020).
5. A. Honders, J. M. der Kinderen, A. H. van Heeren, J. H. W. de Wit, and G. 25. E. Trevisanello, R. Ruess, G. Conforto, F. H. Richter, and J. Janek, “Polycrystalline
H. Broers, “Bounded diffusion in solid solution electrode powder compacts. Part II. and single crystalline ncm cathode materialsquantifying particle cracking, active
The simultaneous measurement of the chemical diffusion coeffcient and the surface area, and lithium diffusion.” Adv. Energy Mater., 11, 2003400 (2021).
thermodynamic factor in LixTiS2 and LixCoO2.” J. Solid State Ionics, 15, 265 26. B. Wen, Z. Deng, P.-C. Tsai, Z. W. Lebens-Higgins, L. F. J. Piper, S. P. Ong, and
(1985). Y.-M. Chiang, “Ultrafast ion transport at a cathodeelectrolyte interface and its
6. S. D. Kang and W. C. Chueh, “Galvanostatic intermittent titration technique strong dependence on salt solvation.” Nat. Energy, 5, 578 (2020).
reinvented: part 1. a critical Review.” J. Electrochem. Soc., 168, 120504 (2021). 27. W. Xue et al., “Ultra-high-voltage Ni-rich layered cathodes in practical Li metal
7. X. Hou, K. Ohta, Y. Kimura, Y. Tamenori, K. Tsuruta, K. Amezawa, and batteries enabled by a sulfonamide-based electrolyte.” Nat. Energy, 6, 495 (2021).
T. Nakamura, “Lattice oxygen instability in oxide-based intercalation cathodes: a 28. J. Zhang, J. Zhang, X. Ou, C. Wang, C. Peng, and B. Zhang, “Enhancing high-
case study of layered LiNi1/3Co1/3Mn1/3O2.” Adv. Energy Mater., 11, 2101005 voltage performance of ni-rich cathode by surface modification of self-assembled
(2021). nasicon fast ionic conductor LiZr2(PO4)3.” ACS Applied Materials & Interfaces, 11,
8. P. M. Csernica et al., “Persistent and partially mobile oxygen vacancies in Li-rich 15507 (2019).
layered oxides.” Nat. Energy, 6, 642 (2021). 29. J. Zhu, Y. Li, L. Xue, Y. Chen, T. Lei, S. Deng, and G. Cao, “Enhanced
9. R. Amin, D. B. Ravnsbæk, and Y.-M. Chiang, “Characterization of electronic and electrochemical performance of Li3PO4 modified Li[Ni0.8Co0.1Mn0.1]O2 cathode
ionic transport in Li1-xNi0.8Co0.15Al0.05o2 (NCA).” J. Electrochem. Soc., 162, material via lithium-reactive coating.” J. Alloys Compd., 773, 112 (2019).
A1163 (2015). 30. P. Liu, L. Xiao, Y. Chen, and H. Chen, “Highly enhanced electrochemical
10. R. Amin and Y.-M. Chiang, “Characterization of electronic and ionic transport in performances of LiNi0.815Co0.15Al0.035O2 by coating via conductively LiTiO2 for
Li1-xNi0.33Mn0.33Co0.33o2(NMC333) and Li1-xNi0.50Mn0.20Co0.30O2(NMC523) as a lithium-ion batteries.” Ceramics International, 45, 18398 (2019).
Function of Li Content.” J. Electrochem. Soc., 163, A1512 (2016). 31. Y. Cao, X. Qi, K. Hu, Y. Wang, Z. Gan, Y. Li, G. Hu, Z. Peng, and K. Du,
11. E. J. Cheng, K. Hong, N. J. Taylor, H. Choe, J. Wolfenstine, and J. Sakamoto, “conductive polymers encapsulation to enhance electrochemical performance of Ni-
“Mechanical and physical properties of LiNi0.33Mn0.33Co0.33O2 (NMC).” J. Eur. rich cathode materials for Li-ion batteries.” ACS Applied Materials & Interfaces,
Ceram. Soc., 37, 3213 (2017). 10, 18270 (2018).
12. J. Zahnow, T. Bernges, A. Wagner, N. Bohn, J. R. Binder, W. G. Zeier, M. T. Elm, 32. A. Zhou, X. Dai, Y. Lu, Q. Wang, M. Fu, and J. Li, “Enhanced interfacial kinetics
and J. Janek, “Impedance analysis of ncm cathode materials: electronic and ionic and high-voltage/high-rate performance of licoo2 cathode by controlled sputter-
partial conductivities and the influence of microstructure.” ACS Appl. Energy coating with a nanoscale Li4Ti5O12 Ionic Conductor.” ACS Applied Materials &
Mater., 4, 1335 (2021). Interfaces, 8, 34123 (2016).
13. C. Wang et al., “A general method to synthesize and sinter bulk ceramics in 33. Y. Chen, K. Xie, C. Zheng, Z. Ma, and Z. Chen, “Enhanced li storage performance
seconds.” Science, 368, 521 (2020). of LiNi0.5Mn1.5O4Coated 0.4Li2MnO3-0.6LiNi1/3Co1/3Mn1/3O2 cathode materials
14. J. Park et al., “Fictitious phase separation in Li layered oxides driven by electro- for li-ion batterie.” ACS Applied Materials & Interfaces, 6, 16888 (2014).
autocatalysis.” Nat. Mater., 20, 991 (2021). 34. K. Nie et al., “Realizing long-term cycling stability and superior rate performance
15. J. J. Kuo, S. D. Kang, and W. C. Chueh et al., (2022). of 4.5 V LiCoO2 by aluminum doped zinc oxide coating achieved by a simple wet-
16. A. Van der Ven, “Lithium Diffusion in Layered LixCoO2.” Electrochem. Solid- mixing method.” Journal of Power Sources, 470, 228423 (2020).
State Lett., 3, 301 (1999). 35. D. Fraggedakis, M. McEldrew, R. B. Smith, Y. Krishnan, Y. Zhang, P. Bai, W.
17. A. Van der Ven, G. Ceder, M. Asta, and P. D. Tepesch, “First-principles theory of C. Chueh, Y. Shao-Horn, and M. Z. Bazant, “Theory of coupled ion-electron
ionic diffusion with nondilute carriers.” Physical Review B, 64, 184307 (2001). transfer kinetics.” Electrochimica Acta, 367, 137432 (2021).
18. K. Kang and G. Ceder, “Factors that affect Li mobility in layered lithium transition 36. M. Doyle, T. F. Fuller, and J. Newman, “Modeling of galvanostatic charge and
metal oxides.” Physical Review B, 74, 094105 (2006). discharge of the lithium/polymer/insertion cell.” J. Electrochem. Soc., 140, 1526
19. H. Zhou, F. Xin, B. Pei, and M. S. Whittingham, “What limits the capacity of (1993).
layered oxide cathodes in lithium batteries?” ACS Energy Lett., 4, 1902 (2019). 37. J. C. Fisher, “Calculation of diffusion penetration curves for surface and grain
20. W. Yin et al., “Structural evolution at the oxidative and reductive limits in the first boundary diffusion.” J. Appl. Phys., 22, 74 (1951).
electrochemical cycle of Li1.2Ni0.13Mn0.54Co0.13O2.” Nat. Commun., 11, 1252 38. M. Leonhardt, J. Jamnik, and J. Maier, “In situ monitoring and quantitative analysis
(2020). of oxygen diffusion through schottky-barriers in SrTiO3 bicrystals.” Electrochem.
21. A. Grenier, P. J. Reeves, H. Liu, I. D. Seymour, K. Märker, K. M. Wiaderek, P. Solid-State Lett., 2, 333 (1999).
J. Chupas, C. P. Grey, and K. W. Chapman, “Intrinsic kinetic limitations in 39. C.-C. Chen, L. Fu, and J. Maier, “Synergistic, ultrafast mass storage and removal in
substituted lithium-layered transition-metal oxide electrodes.” J. Am. Chem. Soc., artificial mixed conductors.” Nature, 536, 159 (2016).
142, 7001 (2020). 40. G. Gregori, R. Merkle, and J. Maier, “Ion conduction and redistribution at grain
22. J. Kasnatscheew et al., “The truth about the 1st cycle Coulombic effciency of boundaries in oxide systems.” Prog. Mater. Sci., 89, 252 (2017).
LiNi1/3Co1/3Mn1/3O2 (NCM) cathodes.” Phys. Chem. Chem. Phys., 18, 3956 41. K. K. Adepalli, J. Yang, J. Maier, H. L. Tuller, and B. Yildiz, “Tunable oxygen
(2016). diffusion and electronic conduction in SrTiO3 by dislocation-induced space charge
23. P.-C. Tsai, B. Wen, M. Wolfman, M.-J. Choe, M. S. Pan, L. Su, K. Thornton, fields.” Adv. Funct. Mater., 27, 1700243 (2017).
J. Cabana, and Y.-M. Chiang, “Single-particle measurements of electrochemical 42. C.-C. Chen, Y. Yin, S. D. Kang, W. Cai, and W. C. Chueh, “Electro-chemo-
kinetics in NMC and NCA cathodes for Li-ion batteries.” Energy & Environmental mechanical charge carrier equilibrium at interfaces.” Phys. Chem. Chem. Phys., 23,
Science, 11, 860 (2018). 23730 (2021).

You might also like