Content uploaded by Sugeun Jo
Author content
All content in this area was uploaded by Sugeun Jo on Aug 15, 2024
Content may be subject to copyright.
Page 1 of 21
Dynamic surface phases controlling asymmetry of high-rate lithiation and
delithiation in phase-separating electrodes
Bonho Koo1,7, Jinkyu Chung1,7, Juwon Kim1,7, Dimitrios Fraggedakis2, Sungjae Seo1,
Chihyun Nam1, Danwon Lee1, Jeongwoo Han1, Sugeun Jo1, Hongbo Zhao2, Neel Nad-
karni2, Jian Wang3, Namdong Kim4, Markus Weigand5, Martin Z. Bazant2,6,8, Jongwoo
Lim1,8
1Department of Chemistry, Seoul National University, Seoul 08826, South Korea
2Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139,
USA
3Canadian Light Source Inc., Saskatoon, Saskatchewan S7N 0X4, Canada
4Pohang Accelerator Laboratory, Jigokro-127-beongil, Pohang, Gyeongbuk 790-834, South Korea
5Helmholtz-Zentrum Berlin (HZB), Albert-Einstein-Straße 15, Berlin 12489, Germany
6Department of Mathematics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
7These authors contributed equally to this work.
8Corresponding author. E-mail: jwlim@snu.ac.kr, bazant@mit.edu
Electronic Supplementary Material (ESI) for Energy & Environmental Science.
This journal is © The Royal Society of Chemistry 2023
Page 2 of 21
Supplementary section 1. Operando scanning transmission x-ray microscopy
(STXM) platforms
1-1. Electrochemistry chip fabrication
Electrochemistry chips in the microfluidic cell were microfabricated through the
following procedure. First, a silicon wafer (4-inch diameter and 200 μm thickness) was
cleaned with NH4OH/H2O2 at 50℃ (10 min), HCl/H2O2 at 70℃ (10 min) and diluted HF
(50:1, 1 min). After cleaning, low stress SiNx (150 nm thickness) was deposited on the
wafer through low pressure chemical vapor deposition method. Then, standard photoli-
thography was conducted on the backside of Si/SiNx wafer to define X-ray transparent
SiNx membrane window.
The defined SiNx window pattern on the backside was etched using inductively
coupled plasma etcher to expose the Si for KOH wet etching. After removing the pat-
terned photoresist (PR) in acetone, Si-exposed wafers were dipped into 1:3 weight ratio
(KOH:DI water) KOH aqueous solution at 80℃ to etch Si and create SiNx membrane
window (450×20 μm). The chip dimension is 4.85×6.35 mm. Photolithography was con-
ducted on the topside of the window-defined wafer to pattern current collector (CC) pad
on the electrochemistry chip.1 The photolithography process was followed by spin coat-
ing, soft bake, UV exposure, post exposure bake using a negative tone PR (SU-8 2000.5).
1-2. Particle loading on the Electrochemistry chip
As-synthesized single crystalline LFP particles were dispersed in isopropyl alco-
hol (IPA), and ultrasonication was served to avoid aggregation. Then, dispersed particles
in IPA were drop-casted and spin-coated on the working electrode on the Electrochemis-
try chip. Aggregated particles on the chip were stamped out by polydimethylsiloxane
(PDMS) to make a single layer of LFP particles on the electrode. The number of LFP
particles on the working electrode in the chip was approximately estimated through opti-
cal microscope (104~105). The LiFePO4 particles were coated with carbon through chem-
ical vapor deposition method while the PR was carbonized simultaneously, forming car-
bon current collector.
Page 3 of 21
1-3. Operando STXM experiment in microfluidic cell
Supplementary Fig. 1 Schematics of operando STXM experiment. a, (Left) Schemat-
ics of beam path of synchrotron x-ray, the microfluidic cell, and STXM optics for oper-
ando imaging. (Right) Schematic cross-sectional view of the microfluidic cell.
LixFePO4/C electrode on a microfabricated chip is placed in sandwiched SiNx membrane
where thickness between the membranes is ~ 1 μm. b, Multiple galvanostatic charge/dis-
charge curves of [100]-oriented LFP/C electrode in the microfluidic cell during operando
STXM experiment.
The microfluidic electrochemical cell consists of two microfabricated SiNx mem-
branes of the Electrochemistry and the spacer chip where liquid electrolyte flows through
the gap of ~ 1 μm as shown in Supplementary Fig. 1(b). For operando STXM experi-
ments, we used 1 M LiClO4 in ethylene carbonate/dimethyl carbonate (EC/DMC, 1:1
volume ratio).
To estimate faradaic capacity of the active materials, [100]-oriented LFP, on the
microfabricated electrode, which contains the number of particles 104~105, galvanostatic
current ~1 nA was applied to the working electrode. Once the capacity was confirmed by
the first electrochemical cycling, the working electrode was charged and discharged with
various C-rates during operando STXM experiments. Galvanostatic (de)lithiations of the
LFP Electrochemistry chip showed a typical electrochemical voltage plateau, small por-
tion of non-faradaic capacity less than 5% at various C-rate, and reasonable rate-capabil-
ities as shown in Supplementary Fig. 1(c). We concluded that operando electrochemistry
STXM platform is well-functional and representable for coin cells.
Page 4 of 21
Supplementary section 2. Characterizations of [100]-oriented LFP micro-
platelet single particle.
Supplementary Fig. 2 Validation of [100]-oriented LFP single particle. a, (Left) SEM
image of as-synthesized [100]-oriented LFP particle. (Right) SEM image of the LFP par-
ticle at a 45-degree tilt. b, X-ray diffraction pattern of as-synthesized [100]-oriented LFP,
and the LFP dispersed on silicon wafer substrate. c, d Electrochemical rate-capability (c)
and cyclability (d) tests of the [100]-oriented LFP.
Synthesized [100]-oriented LFP microplatelet is about ~400 nm wide, ~1 μm long,
and ~150 nm thick as shown in SEM images in Supplementary Fig. 2(a). Powder X-ray
diffraction (XRD) patterns (orange-solid line) of the as-synthesized LFP well capture the
theoretical XRD features of LFP (black-vertical lines, JCPDS No. 04-012-5179) in Sup-
plementary Fig. 2(b).2 Le bail fit results for power XRD data of [100]-oriented LFP are
provided in Supplementary Section 3. For XRD patterns of the LFP dispersed on flat
silicon wafer substrate (blue-solid line), the (200) intensity strongly manifests, which re-
flects the synthesize LFP has highly (100) preferred orientation.
In addition, we performed rate capability, capacity retention and cyclability tests
for the synthesized LFP with LFP/Li half-cells, shown in Supplementary Fig. 2(c) and
Page 5 of 21
(d). Specific capacity of the synthesized LFP was ~160 mAh/g, the capacity was com-
pletely recovered at 0.1C after multiple rate-capability tests, and the initial capacity was
maintained over ~80 cycles at 0.1C. This strongly supports that the synthesized [100]-
oriented LFP particle is a suitable model material for investigating Li (de)insertion spa-
tiodynamics with negligible defects.
Page 6 of 21
Supplementary section 3. Refinements of power X-ray diffraction (XRD)
Supplementary Fig. 3 Le bail fit results of the synchrotron XRD patterns of as-syn-
thesis [100]-orientated LFP primary particles. Le bail fit was conducted with structure
corresponds to Pnma space group (No.62).
XRD data was measured at beamline 9B at Pohang Light Source, South Korea. A
wavelength of incident X-ray was 1.5226 Å, 2θ range was 10°–131°, and the step size
was 0.01°. Le Bail fit was performed using the TOPAS software.3 Lattice parameters are
a = 10.3262(3), b = 6.0052(3), c = 4.6932(0), and Cell Volume = 291.0316(3) (Å 3) with
Rp: 7.88 Rwp: 10.27, Rexp: 5.87, GOF = 1.75. The unit cell volume 291.03 Å 3 indicates
negligible anti-site defects.4
Page 7 of 21
Supplementary section 4. Quantification of oxidation states and Li concentra-
tion through STXM at L3 absorption edge
Supplementary Fig. 4 Comparison of Li composition maps between linear combina-
tion of full spectra and 3 energy method. a, b, Li composition maps using fine energy
points (full spectra) fitting (a) and three energy points fitting (b) are presented. c, X-ray
absorption spectra in fully lithiated LFP (pink-solid line), fully delithiated LFP (violet-
solid line), and Region A in Supplementary Fig. 4(a) and (b) (white circle). The absorp-
tion spectrum of Region A is fitted by the linear combination of two end-phase absorption
spectra (pink/violet).
STXM at L3 absorption edge offers superior chemical sensitivity for measuring
nanoscale particles and local current density. This technique is particularly advantageous
for analysing primary single particles, as 200 nm thick LiFePO4 absorbs around 60% of
the incident X-rays at the Fe L-edge, while only absorbing about 1% at the Fe K-edge
used by hard X-ray TXM.5 Therefore, the Fe L-edge is highly sensitive to oxidation states,
allowing to measure fewer energy points for accurate quantification compared to the K-
edge of transition metals. In this regard, operando STXM has been employed as a cutting-
edge technique not just for various energy storage materials,6–8 but also for many catalyst
materials,9–11 significantly advancing our comprehension of nanoscale kinetics in chemi-
cal and electrochemical systems.
In this work, Li composition of [100]-oriented LFP single particles is quantified
with ~40 nm spatial and ~30 s temporal resolution through STXM image analysis at Fe
L3 absorption edge. The exposure time for each pixel is typically ranges from 1 to 3
Page 8 of 21
milliseconds, allowing for a full image scan of particles as large as 1-2 micrometres within
30 to 60 seconds. The pixelwise optical density (OD) is linearly fitted by the reference
absorption spectra at Fe L3-edge of fully-lithiated (Fe2+) and fully-delithiated (Fe3+) LFP.
Here, two different quantification methods were carried out; the pixelwise OD was fitted
by linear combination of two reference absorption spectra with 1) fine energy points (17
points), or 2) three energy points of 703, 707.5, and 710 eV. To validate two linear com-
bination methods, region A was specified, white boxes in Supplementary Fig. 4(a) and
(b). L3-edge absorption of region A is plotted as circular symbols in Supplementary Fig.
4(c), and the linearly combined spectrum (a black-dashed line) accurately fits the absorp-
tion data. Fine energy points and three energy points methods showed the sufficiently
similar Li distributions as shown in Supplementary Fig. 4(a) and (b). Thus, to increase
temporal resolution during operando STXM, the linear combination with three energy
points were used for the quantification.
Pixelwise X-ray absorptions at 707.5 and 710 eV are calculated by subtraction of
the optical density at the pre-edge energy point 703 eV from those at 707.5 and 710 eV,
respectively. The relative amount of Li-rich and Li-poor composition, x and y, is deter-
mined by the following equation,
.
(1)
Here, represents the pixelwise absorption at the energy of E, and
and
indicate
the absorption coefficients of the end-member spectra (LFP and FP) at the energy of E.
Li concentration in LixFePO4 is then calculated from
. The hue and brightness
in Li composition maps in Fig.3-5 in the main text represent and , respectively.
Page 9 of 21
Supplementary section 5. Li composition change during 0.15C (de)lithiation
Supplementary Fig. 5 Line cuts of Li composition and time-averaged composition
change in LixFePO4 single particles during 0.15C (de)lithiation. a and b, Representa-
tive particle and corresponding line cuts profile along the white arrow are plotted during
0.15C delithiation (a) and lithiation (b) respectively (c) Time-averaged Li composition
change between two consecutive Li maps during 0.15C (de)lithiation. Blue colour indi-
cates a relative scale of the composition change rates.
At 0.15C delithiation, line cuts of Li composition in Supplementary Fig. 5(a)
shows the full extraction of Li channels at the bottom of the particle and clear phase
boundaries between Li-rich and poor phase, representing channel-by-channel kinetics. In
contrast, during 0.15C lithiation in Supplementary Fig. 5(b), Li fills up from the surface
region, but the whole channels are not fully lithiated, showing more uniform Li distribu-
tion. The composition gradients in the [010] channels are much smoother during lithiation
than that during delithiation. Line cuts clearly reflect that lithiation occurs homogene-
ously along the [010] channels while delithiation drives strong phase separation and dom-
ino-cascade features.
Page 10 of 21
Supplementary section 6. Definition of 1) pixelated channels (Chs) and 2) the
surface/internal pixel of STXM image
Supplementary Fig. 6 A representative pixelated channel (Ch) and the surface pixels
of [100]-oriented LFP primary particle. A representative Ch (red) and the surface pixel
(violet) are specified in the LFP particle. The internal pixels are the complement of the
surface pixels within the single particle.
We termed an array of pixels along the [010] direction at ~40 nm resolution as a
pixelated channel (Ch) (red-coloured pixels in Supplementary Fig. 6). To specify Chs, Li
map frames were rotated with respect to the particle morphologies. The long edge side of
the particle was maintained parallel to the vertical axis, and the short side edge was as-
sumed to be parallel to the [001] direction based on the statistical TEM analysis.
In addition, the surface pixel was defined as a group of edge pixels at the [010]
facet within the single particles (marked in violet in Supplementary Fig. 6). The internal
pixel was defined as the remaining part, except for the surface pixels within the entire
particle pixel. Here, the pixel-wise particle boundary was confined under the following
criterion of brightness: (see Supplementary Section 4). The surface pixels
of the Chs were also defined as two edge pixels of the Chs, where the violet and red pixels
overlap in Supplementary Fig. 6.
Page 11 of 21
Supplementary section 7. Operando STXM experiment at 4C (de)lithiation
for specifying the rate determining step
Supplementary Fig. 7 Operando STXM results to identify the rate determining step.
a, Representative Li maps of a [100]-oriented LFP during electrochemical (de)lithiation
at 4C. We intentionally left the pre-delithiated core region (orange-dashed oval in (a))
after 4C lithiation to precisely locate the formation of Li-poor phase during 4C delithia-
tion. b, c Corresponding line cuts of Li composition during 4C delithiation along line 1
(b) and line 2 (c), shown in (a). d, Schematics of phase transformation pathway of a par-
ticle with the pre-delithiated core under SRL and BDL process.
Page 12 of 21
For phase-separating materials, surface-reaction-limited (SRL) and bulk-diffu-
sion-limited (BDL) process drive a sharp diffusional front of the Li composition near the
surface region.12,13 Thus, concurrent Li-rich and Li-poor domains near the surface regions,
as shown in Fig. 5(a), complicate the identification of the rate-determining step (RDS)
during high-rate (de)insertion reactions. To specify the RDS at high C-rates, we designed
operando STXM experiments of [100]-oriented LFP particles at 4C delithiation, where
the LFP particle had a pre-delithiated core region after fast 4C lithiation (Supplementary
Fig. 7(a)). With a pre-delithiated core, triggering inhomogeneity in the interfacial energy
within the particle and effectively lowering the nucleation barrier of the early transition
of Li-poor domains is possible.
As depicted in the schematics of Supplementary Fig. 7(d), with delithiation being
limited by the surface reaction, Li in the core region is extracted because the nucleation
barrier of the Li-poor phase at the pre-delithiated core is lower than that at the Li-rich
surface region, and the expansion of a single domain at the core has a much lower inter-
facial energy. In contrast, if delithiation was limited by bulk transport, the composition at
the surface region would rapidly become Li-deficient and gradually become zero while
the entire reaction surface was activated. The existence of the pre-delithiated core enabled
precise localisation of the delithiation domains and determined the RDS between the SRL
and BDL.
Our observations from Supplementary Fig. 7(a) indicate that the pre-delithiated
core expands to the surface region, while the region becomes more delithiated. Line cuts
of Li composition in Supplementary Fig. 7(b) and (c) support that Li ions are mainly
extracted at the core region, and Li composition at the surface region remains populated
by Li during overall 4C delithiation. This result indicates that bulk transport in the [010]
channels in LFP is fast enough to keep up with the 4C reaction, and the characteristic
reaction timescale at 4C is much greater than the diffusion timescale in LFP. Thus, we
conclude that our operando STXM results are within the regime of the SRL process.
Page 13 of 21
Supplementary section 8. Li composition change during ~1 h relaxation right
after 7C delithiation
Supplementary Fig. 8: Representative Li map from operando STXM during ~1 h
relaxation at of LixFePO4 after 7C delithiation. a, Li composition maps of
LFP primary particles during relaxation at of LixFePO4 after 7C delithiation.
Li maps of composition change during the relaxation are shown at the right (red: -0.3 and
blue: +0.3). Li diffusion more occurs from Li-poor to Li-rich regions towards phase-sep-
aration during relaxation. b, Corresponding line cuts of Li composition along line 1 and
2 in (a).
We performed operando STXM at 7C delithiation and relaxation immediately af-
ter the delithiation. Supplementary Fig. 8(a) shows the Li composition maps and their
changes during ~1 h of relaxation, where the increase and decrease in Li concentration
during the relaxation are marked with blue and red, respectively. Notably, the Li compo-
sition changes preferentially in the surface region. The Li concentration at the delithiated
surface region is reduced more during relaxation, indicating a further development of the
Li-poor phase. In contrast, the Li composition in the lithiated surface region at t = 0 in-
creases more during relaxation. Line cuts of the Li composition along line 1 and line 2 in
Fig. 8(a) are plotted in Supplementary Fig. 8(b), supporting the nucleation of the stable
phases near the surface region during relaxation.
Page 14 of 21
Supplementary section 9. Uniformity coefficient calculation
Supplementary Fig. 9 Uniformity coefficient (UC) calculation of [100]-oriented LFP.
a, Standard deviations of pixelwise Li compositions within the particle Li maps as a func-
tion of mean Li composition of the particle. UC = 0 and 1 imply fully phase-separated
and uniform solid-solution phases, respectively. b, Li composition maps during 0.15C
and 4C delithiation process corresponding to (a). c, UC of the particles during operando
STXM as a function of C-rate. The value in parentheses indicates the number of particles
in the same electrochemical process.
We define uniformity coefficient (UC) to describe how uniformly Li composition
distributes within the particles during the overall (de)lithiation reaction.7 First, pixelwise
standard deviation (SD) of Li composition is calculated, plotted as a function of of
the particles in Supplementary Fig. 9(a). Then, SD is fitted by the following equation (2)
where represents the standard deviation curve of fully phase-separated parti-
cle at Li composition .
.
(2)
Fitting parameter A is proportional to compositional heterogeneity. Therefore, UC is de-
fined as 1-A where UC = 1 refers completely uniform Li distribution, and UC = 0 refers
complete phase separation. In Supplementary Fig. 9(b), Li maps at the top and bottom
row are representative Li maps at 0.15C and 4C delithiation, respectively. Their UCs are
~0.43 for 0.15C and ~0.64 for 4C, respectively, indicating Li extraction occurs more uni-
formly at higher C-rates. Supplementary Fig. 9(c) shows UC of the single particles
tracked in operando STXM as a function of C-rate. At lower rates, Li distribution during
delithiation shows more heterogeneity, or more phase-separation, than that during lithia-
tion. At higher rates, solid-solution phase more appears within the particles, and UC
Page 15 of 21
during both lithiation and delithiation becomes similar, which agrees with previous re-
ports.14–16
Page 16 of 21
Supplementary section 10. Calculation of exchange current density of Chs
Supplementary Fig. 10 Schematics of [100]-oriented LFP particle and the Chs
The exchange current density of Chs within the single particle is calculated
to determine precise (de)insertion reaction rates as a function of the surface Li composi-
tion . The exchange current density is calculated by the following procedures:
1) The pixelwise Li composition difference is calculated between two sequential
Li maps which were measured with the time interval .
2) is converted into ionic current density of the Ch through the following
equation (4),
(4)
where is the electron charge in Coulomb, is the maximum volume concen-
tration of Li sites in LFP (, is the time interval of the two
sequential Li maps, is thickness of the single particle, is the width of a pixel,
is the volume of a pixel, and all pixelwise are summed up within the Ch.
Here, we hypothesized that Li composition change in Ch is resulted solely
from the current applied to the two edge-sides of the Ch during the reaction time-
scale of our study.
3) By linearisation of BV equation under the overpotential limit of ~100 mV, the
exchange current density can be approximated by
Page 17 of 21
(4)
where is the gas constant, is the room temperature, is the Faraday constant,
and is the electrochemical overpotential calculated in galvanostatic voltage
graph shown in Supplementary Fig. 1(b).
4) To reduce the uncertainty arising from the time-averaged current density between
two Li maps, we calculated the normalized current density by fitting
with the following equation (5)7,17,18
(5)
is a fitting parameter, and represents the magnitude of the exchange current density.
is equivalent to. Additionally, in Fig. 4(b) in the main text, the equation
for the solid-line, which were used for a guide to the eye, is .
Page 18 of 21
Supplementary section 11. Description for 3D phase field model and 3D re-
laxation simulation
Supplementary Fig. 11 3D phase field simulation for relaxation of [100]-oriented
LFP
To understand the phase morphologies under insertion and relaxation, we per-
formed 3D phase-field simulations on a single LixFePO4 particle with dimensions of 300
nm × 300 nm × 100 nm. Our goal is to investigate the simulated phase morphologies
during insertion in the [010]-direction and compare with our experimental observations.
The eigenstrain in the crystallographic [010] direction is assumed to be linear with respect
to Li composition, with a constant chemical expansion coefficient of 0.0346, consistent
with previous computational studies.19–21 The gradient energy penalty is isotropic (equal
in all directions). For relaxation, the diffusivity in [010] direction is taken to be 10−9 cm2/s,
consistent with first-principles calculations.22 The particle is subject to zero external force
( ) and the strain fields are considered to be coherent. The set of differential
equations of our model was discretized using finite elements, and all unknowns were ap-
proximated with linear polynomial basis functions.23 For the time integration of the re-
sulting system of differential algebraic equations, we used 2nd order Gear method.
Page 19 of 21
Supplementary section 12. Representative operando STXM results during
lithiation at various C-rates
Supplementary Fig. 12 Representative particles captured via operando STXM dur-
ing lithiation at various C-rates. Images at 0.15C were obtained from Fig. 3(a) in the
main text for comparison. Scale bars, 400 nm.
Page 20 of 21
References
1. Lim, J. et al. Origin and hysteresis of lithium compositional spatiodynamics
within battery primary particles. Science 353, 566–571 (2016).
2. Li, Z. et al. [100]-Oriented LiFePO4 Nanoflakes toward High Rate Li-Ion
Battery Cathode. Nano Lett. 16, 795–799 (2016).
3. Coelho, A. A. TOPAS and TOPAS-Academic: An optimization program
integrating computer algebra and crystallographic objects written in C++: An. J.
Appl. Crystallogr. 51, 210–218 (2018).
4. Chen, J. & Graetz, J. Study of antisite defects in hydrothermally prepared
LiFePO4 by in situ x-ray diffraction. ACS Appl. Mater. Interfaces 3, 1380–1384
(2011).
5. Li, Y. et al. Dichotomy in the Lithiation Pathway of Ellipsoidal and Platelet
LiFePO4 Particles Revealed through Nanoscale Operando State-of-Charge
Imaging. Adv. Funct. Mater. 25, 3677–3687 (2015).
6. Ohmer, N. et al. Phase evolution in single-crystalline LiFePO4 followed by in
situ scanning X-ray microscopy of a micrometre-sized battery. Nat. Commun. 6,
1–7 (2015).
7. Lim, J. et al. Origin and hysteresis of lithium compositional spatiodynamics
within battery primary particles. Science 353, 566–571 (2016).
8. Csernica, P. M. et al. Persistent and partially mobile oxygen vacancies in Li-rich
layered oxides. Nat. Energy 6, 642–652 (2021).
9. Mefford, J. T. et al. Correlative operando microscopy of oxygen evolution
electrocatalysts. Nature 593, (2021).
10. De Smit, E. et al. Nanoscale chemical imaging of a working catalyst by scanning
transmission X-ray microscopy. Nature 456, 222–225 (2008).
11. Yoo, M. et al. A tailored oxide interface creates dense Pt single-atom catalysts
with high catalytic activity. Energy Environ. Sci. 13, 1231–1239 (2020).
12. Park, J. et al. Fictitious phase separation in Li layered oxides driven by electro-
autocatalysis. Nat. Mater. 20, 991–999 (2021).
13. Fraggedakis, D. et al. A scaling law to determine phase morphologies during ion
intercalation. Energy Environ. Sci. 22–27 (2020).
14. Liu, H. et al. Capturing metastable structures during high-rate cycling of LiFePO
Page 21 of 21
4 nanoparticle electrodes. Science 344, (2014).
15. Zhang, X. et al. Direct view on the phase evolution in individual LiFePO 4
nanoparticles during Li-ion battery cycling. Nat. Commun. 6, 1–7 (2015).
16. Li, Y. et al. Fluid-enhanced surface diffusion controls intraparticle phase
transformations. Nat. Mater. 17, 915–922 (2018).
17. Zhao, H. & Bazant, M. Z. Population dynamics of driven autocatalytic reactive
mixtures. Phys. Rev. E 100, 1–11 (2019).
18. Bazant, M. Z. Thermodynamic stability of driven open systems and control of
phase separation by electro-autocatalysis. Faraday Discuss. 199, 423–463
(2017).
19. Cogswell, D. A. & Bazant, M. Z. Coherency strain and the kinetics of phase
separation in LiFePO 4 nanoparticles. ACS Nano 6, 2215–2225 (2012).
20. Bai, P., Cogswell, D. A. & Bazant, M. Z. Suppression of phase separation in
LiFePO4 nanoparticles during battery discharge. Nano Lett. 11, 4890–4896
(2011).
21. Tang, M., Belak, J. F. & Dorr, M. R. Anisotropic phase boundary morphology in
nanoscale olivine electrode particles. J. Phys. Chem. C 115, 4922–4926 (2011).
22. Nadkarni, N. et al. Interplay of phase boundary anisotropy and electro-auto-
catalytic surface reactions on the lithium intercalation dynamics in LiXFePO4
plateletlike nanoparticles. Phys. Rev. Mater. 2, 1–13 (2018).
23. Fraggedakis, D., Papaioannou, J., Dimakopoulos, Y. & Tsamopoulos, J.
Discretization of three-dimensional free surface flows and moving boundary
problems via elliptic grid methods based on variational principles. J. Comput.
Phys. 344, 127–150 (2017).
