Operando X-ray diffraction of lithium trapping in silicon-anode half-cells during extended cycling

Introduction

Silicon remains one of the most intensively studied negative-electrode materials for lithium-ion batteries because it stores far more lithium than graphite. Its practical difficulty is equally familiar: large volume change, fracture, unstable interphase growth, loss of electronic contact, and incomplete lithium recovery during delithiation all reduce cycle life and initial coulombic efficiency [6,7,8,9]. The crystalline-to-amorphous transformation, formation of crystalline Li15Si4 at deep lithiation, and stress-coupled transport have been documented for many silicon forms, from powders to nanowires and thin films [1,2,3,10,11,12,13].

Lithium trapping in silicon electrodes is not a single mechanism. Some lithium remains in crystalline Li15Si4 or amorphous LixSi after the cell is nominally charged. Some becomes immobilised in SEI products. Some resides in silicon domains that are no longer electrically connected to the current collector. Recent studies have used isovalent structural design, multi-edge X-ray scattering, NMR, and interfacial spectroscopy to separate these contributions and to show that high irreversible capacity cannot be assigned only to one phase [4,5,14,15,16,17].

Operando X-ray diffraction is valuable because it follows phase changes during cycling rather than after relaxation or disassembly. In silicon, diffraction has already revealed the disappearance of crystalline Si, emergence of amorphous LixSi scattering, and formation of crystalline Li15Si4 under sufficiently deep lithiation [1,2]. The challenge is that much of the important material is amorphous, nanostructured, or interfacial. Diffraction is therefore a phase-resolved but incomplete lithium-inventory tool. It must be paired with electrochemical accounting and post-mortem measurements if the word trapping is to mean more than a residual peak.

This study uses operando synchrotron XRD to track silicon nanoparticle half-cells over extended cycling under two lower-cutoff protocols. The central question is whether avoiding crystalline Li15Si4 by stopping at 80 mV meaningfully reduces trapped lithium, and whether the remaining loss can be seen in amorphous diffraction features. We do not claim that XRD directly observes every trapped lithium species. Instead, we use diffraction to constrain the phase-resolved part of a broader lithium-inventory model.

Electrode preparation and cell design

Electrodes contained 70 wt% commercial crystalline silicon nanoparticles, 15 wt% conductive carbon, and 15 wt% sodium carboxymethyl cellulose/styrene-butadiene rubber binder blend. The silicon primary-particle diameter was 105 +/- 35 nm from transmission electron microscopy, with agglomerates up to 1.8 micrometres after slurry drying. The areal silicon loading was 1.16 +/- 0.05 mg cm-2, corresponding to a theoretical silicon capacity of approximately 4.1 mAh cm-2 if fully lithiated. Practical capacities were limited deliberately to reduce catastrophic pulverisation during operando testing.

Windowed pouch half-cells used lithium metal counter electrodes, glass-fibre separators, and 1 M LiPF6 in EC/EMC with 10 wt% fluoroethylene carbonate. FEC was included because interphase control is known to be central to silicon-anode stability, and silicon nanotube work has shown how SEI control can dominate cycling outcome [18]. The X-ray windows were 25 micrometre Kapton on the silicon side and 50 micrometre aluminium-laminated polymer on the lithium side. Cell compression was maintained by a spring-loaded holder that also fixed the X-ray path length.

Two voltage protocols were compared. The deep-lithiation group used 10 mV to 1.0 V versus Li/Li+. The restricted group used 80 mV to 1.0 V. Both used three formation cycles at C/20 followed by cycling at C/5, with periodic C/20 diagnostic cycles every 25 cycles. The current was referenced to 3000 mAh g_Si-1 rather than to the full theoretical capacity. This choice avoided extreme end-of-lithiation conditions while still allowing Li15Si4 formation in the 10 mV group.

Three operando cells and six conventional coin-cell repeats were prepared for each protocol. The operando cells had slightly lower absolute capacity than the coin cells because of additional window dead volume and compression differences. Capacity trends and coulombic efficiencies, however, were reproducible within the reported uncertainty. All values in the article are reported per gram of silicon unless otherwise stated.

Operando XRD measurement

Synchrotron diffraction was performed in transmission geometry at 24.8 keV with a beam size of 0.5 mm x 0.5 mm. Two-dimensional patterns were collected every 90 s during the first three cycles and every 300 s during later diagnostic cycles. During ordinary C/5 cycling between diagnostic blocks, diffraction was collected for one full cycle every fifth electrochemical cycle. A CeO2 standard was used for detector-distance and wavelength calibration, and the copper current collector peak provided an internal check for drift. The scan plan follows the same operando philosophy as combined X-ray tomography and diffraction studies of lithium-ion cells, but trades three-dimensional spatial mapping for higher time resolution on the silicon electrode [20].

Background subtraction was more difficult than in ex situ powder diffraction because the pouch cell contained Kapton, separator, electrolyte, lithium metal, and current collector contributions. We measured empty-window, separator-electrolyte, and lithium-only reference cells and fitted them as constrained background components. The silicon-electrode contribution was then analysed over three regions: crystalline Si Bragg peaks, crystalline Li15Si4 peaks, and broad amorphous LixSi scattering between 1.7 and 2.6 inverse angstroms.

Peak areas were not interpreted as absolute phase fractions without calibration. For crystalline Si and Li15Si4, integrated areas were normalised to reference patterns measured from stopped cells with known lithiation states. For amorphous LixSi, we used the centroid and integrated excess intensity of the broad feature as a relative state-of-lithiation marker. This approach follows the spirit of prior XRD, PDF, and NMR studies, which show that silicon lithiation proceeds through amorphous intermediates that are only partly represented by sharp Bragg peaks [1,4,5].

Radiation damage was checked by cycling two cells with the beam parked off-electrode except during diagnostic cycles. Their capacity fade and diffraction evolution matched continuously illuminated cells within experimental scatter. The total dose was therefore not a dominant ageing mechanism under the beam conditions used here. This check is essential for operando battery diffraction, where the measurement can otherwise become part of the degradation process.

Lithium-inventory model

Lithium trapping was quantified by combining electrochemical charge balance with diffraction-derived phase markers and post-mortem spectroscopy. The electrochemical inventory gives cumulative irreversible capacity after correcting for lithium-metal counter-electrode polarisation and formation-cycle electrolyte wetting. XRD assigns part of that loss to residual crystalline Li15Si4 and to the delayed return of amorphous-LixSi scattering. XPS and 7Li NMR assign additional loss to SEI-associated lithium and broad diamagnetic lithium environments, respectively.

The model has three trapped-lithium reservoirs: crystalline Li15Si4 detectable by XRD, amorphous Li-Si retained after delithiation inferred from the amorphous scattering centroid and NMR, and SEI/inactive lithium measured by XPS and inventory residual. A fourth category, electrically isolated silicon-rich domains, is inferred when crystalline or amorphous silicon signal remains but no longer participates in charge balance during subsequent cycles. This category is necessary because fracture and loss of electrical contact are well-established degradation modes in silicon electrodes [10,11,12,13].

Uncertainty was propagated by Monte Carlo sampling of peak-fit covariance, background-subtraction coefficients, electrochemical integration error, and spectroscopy calibration uncertainty. We report inventory partitions with relatively wide confidence intervals because amorphous Li-Si and SEI lithium overlap spectroscopically. The goal is not to assign every lithium atom perfectly, but to determine whether the two voltage protocols produce meaningfully different trapping pathways.

The inventory model is intentionally conservative about diffraction. Residual Li15Si4 is counted only when its strongest reflections exceed a three-sigma threshold and show the expected relative peak positions. Broad amorphous scattering is not treated as proof of lithium trapping unless it persists at the charged state and correlates with charge imbalance. This prevents over-reading ordinary reversible amorphisation as irreversible loss.

First-cycle phase evolution

During the first lithiation, crystalline Si peaks broadened and declined continuously as voltage fell below 170 mV. By 90 mV, less than 8% of the initial crystalline Si integrated intensity remained in both protocols. This irreversible amorphisation is consistent with early silicon diffraction studies and with the accepted crystalline-to-amorphous lithiation mechanism [1,2,3]. The loss of Bragg intensity occurred before most of the delivered capacity, indicating that crystalline order disappeared while lithium insertion continued in amorphous LixSi.

In the 10 mV protocol, Li15Si4 reflections appeared during the final part of lithiation at 43 +/- 7 mV and grew until the lower cutoff. Their positions and relative intensities matched the reference pattern used in earlier silicon-anode work [1,2]. In the 80 mV protocol, no Li15Si4 peak exceeded the detection threshold during formation. The amorphous scattering centroid still shifted strongly, showing that substantial lithiation occurred without resolving crystalline Li15Si4.

On first delithiation, Li15Si4 peaks in the 10 mV cells mostly disappeared by 330 mV, but a weak residual shoulder remained at the strongest reflection position after the cell reached 1.0 V. The residual was only 1.5 +/- 0.6% of the preceding maximum and was near the detection limit, so we do not interpret it as a large trapped reservoir after cycle 1. The larger first-cycle irreversible capacity is instead dominated by SEI growth and amorphous-phase hysteresis, consistent with NMR and interfacial studies [4,5,16,17].

The first-cycle coulombic efficiency was 83.1 +/- 1.2% for the 10 mV cells and 86.7 +/- 0.9% for the 80 mV cells. The restricted cutoff therefore improved initial efficiency but did not remove the main formation loss. This is expected: avoiding Li15Si4 reduces one mechanical and phase pathway, but SEI formation and irreversible amorphisation still consume lithium.

Extended cycling response

After formation, both protocols showed gradual capacity fade and increasing hysteresis. At cycle 120, the 10 mV cells retained 61 +/- 4% of formation discharge capacity, whereas the 80 mV cells retained 74 +/- 3%. Average coulombic efficiency after cycle 20 was 99.31 +/- 0.06% for the 10 mV group and 99.55 +/- 0.04% for the 80 mV group. The difference is small per cycle but substantial over extended cycling.

The operando diffraction trends diverged strongly. In the 10 mV cells, Li15Si4 peak area during lithiation initially increased as the electrode became better wetted and then declined after cycle 35 as active material was lost. More importantly, residual Li15Si4 peak area at the charged state increased nearly monotonically. By cycle 100, the residual area was 7.8 +/- 1.1% of the preceding lithiation maximum. In the 80 mV cells, Li15Si4 remained below detection in all diagnostic cycles.

The amorphous scattering feature also changed with cycling. In both protocols, its centroid moved reversibly during lithiation and delithiation, but the charged-state centroid drifted toward the lithiated position over time. This drift was faster in the 10 mV cells. We interpret it as partial retention of amorphous LixSi and as increasing contribution from electrically isolated, incompletely delithiated domains. The interpretation is supported by post-mortem 7Li NMR, which showed broader lithium environments in aged 10 mV electrodes than in restricted-cutoff electrodes.

Residual crystalline Si intensity at the charged state did not recover after the first cycle. Instead, a weak broad Si-like feature reappeared after 80 cycles in some 80 mV cells. Cross-section imaging showed that this feature correlated with silicon-rich agglomerates near cracks in the electrode coating. We assign it to electrically isolated or poorly accessed silicon rather than to reversible recrystallisation.

Lithium trapping pathways

By cycle 120, cumulative irreversible capacity was 0.74 +/- 0.08 mAh cm-2 for the 10 mV cells and 0.43 +/- 0.06 mAh cm-2 for the 80 mV cells, excluding the initial lithiation of excess lithium metal. The inventory model attributes 43 +/- 6% of the 10 mV loss to trapped Li-Si phases, 39 +/- 7% to SEI-associated lithium, and 18 +/- 5% to electrically isolated silicon-rich domains. In the 80 mV cells, the trapped Li-Si fraction fell to 27 +/- 6%, while SEI-associated loss accounted for 49 +/- 8%.

The distinction between crystalline and amorphous trapped Li-Si is important. Residual Li15Si4 accounted for only about one third of the trapped Li-Si reservoir in the 10 mV cells at cycle 120. The remainder was inferred from amorphous scattering and NMR. Avoiding Li15Si4 therefore helps but does not eliminate lithium retention. This agrees with recent lithium-trapping work showing that structural design can reduce irreversible lithium loss, but also that multiple reservoirs contribute to low initial and extended-cycle coulombic efficiency [14,15].

SEI-associated lithium increased approximately with the square root of cycle number after formation, consistent with diffusion-limited or crack-renewal-limited interphase growth rather than a constant parasitic current. The 10 mV cells had higher SEI growth even after normalising by delivered capacity. We attribute this to larger volume change and repeated exposure of fresh silicon surfaces, consistent with negative-electrode SEI modelling and with the stress and fracture literature [10,11,12,13,19,21,22,23].

The electrically isolated fraction was the least certain reservoir. It was inferred from capacity loss, inactive silicon-rich diffraction features, and post-mortem morphology. The confidence interval is therefore wide. Nevertheless, excluding this reservoir forced the model to over-assign lithium to SEI in a way inconsistent with XPS lithium-to-fluorine and lithium-to-carbon ratios. The multi-reservoir model better reconciles electrochemistry with diffraction and spectroscopy.

Effect of lower cutoff voltage

Lower cutoff voltage controlled whether crystalline Li15Si4 formed, but it also changed mechanical and interfacial ageing. The 10 mV protocol accessed more capacity during early cycles, reaching 2790 +/- 80 mAh g_Si-1 on formation discharge. The 80 mV protocol reached 2290 +/- 70 mAh g_Si-1. By cycle 120, however, the absolute capacity gap had narrowed because the deep-lithiation group faded faster.

Li15Si4 formation is not intrinsically irreversible. Fresh cells can form and remove crystalline Li15Si4 during a cycle. The problem is repeated formation in a porous composite electrode that is expanding, cracking, and reforming SEI. Residual Li15Si4 appears after many cycles when domains become electronically isolated or when delithiation becomes transport-limited. This interpretation is compatible with first-principles and mechanics studies showing anisotropic lithiation, stress effects, and orientation-dependent phase transformation in silicon [21,22,23,24].

The 80 mV cutoff improved coulombic efficiency and reduced residual crystalline trapping but left amorphous retention visible. This suggests that voltage-window engineering should be paired with particle-size control, binder elasticity, interphase stabilisation, and electrode architecture. A voltage cutoff alone cannot stop SEI growth on newly exposed surfaces or recover lithium from isolated domains.

The practical tradeoff is therefore application-dependent. Deep lithiation maximises early capacity but accelerates phase and mechanical degradation. Restricted lithiation sacrifices initial capacity but improves retention and lithium inventory. The operando XRD data make this tradeoff phase-specific rather than purely electrochemical.

Stress, fracture, and diffraction signatures

Diffraction peak broadening and scattering-background changes provide indirect evidence of mechanical degradation. The copper current-collector peak remained stable, ruling out significant fixture drift, while the silicon-electrode amorphous feature broadened asymmetrically after repeated deep lithiation. Cross-sections showed crack networks and coating delamination near regions with the strongest broadening. These observations are consistent with in situ stress measurements and fracture studies showing that lithiation generates large stresses in silicon films, particles, and pillars [10,11,13].

The strongest mechanical signature was not peak shift but loss of reversible scattering amplitude. During early cycles, the amorphous feature intensity followed capacity closely. After roughly 50 cycles in the 10 mV protocol, the feature continued to shift during cycling but with decreasing amplitude. This means that some material still changed lithiation state while another fraction stopped participating. Diffraction alone cannot determine whether that inactive fraction is electrically disconnected, ionically blocked, or mechanically isolated; the inventory model uses post-mortem evidence to distinguish these possibilities.

Particle-size effects are relevant. Size-dependent fracture studies show that smaller silicon particles can better tolerate lithiation strain [13]. Our commercial nanoparticle distribution included large agglomerates, and those agglomerates were preferentially associated with inactive silicon-rich features after cycling. A narrower primary-particle and agglomerate distribution would likely reduce the isolated-domain reservoir.

Comparison with post-mortem measurements

Post-mortem XPS showed thicker fluorinated and organic SEI products in the 10 mV electrodes. The Li 1s signal remained high after careful dimethyl carbonate rinsing, but XPS alone cannot distinguish LiF-rich SEI from trapped Li-Si below the surface. This is why we use XPS as an interfacial constraint rather than as a standalone lithium inventory. Interfacial X-ray studies of silicon anodes have made clear that surface chemistry and buried silicon state evolve together [16].

Solid-state 7Li NMR on delithiated electrodes showed a broad resonance assigned to diamagnetic lithium environments in SEI and disordered Li-Si. The signal was stronger and broader for the 10 mV protocol. These observations are consistent with real-time NMR and PDF studies that have identified amorphous LixSi environments and their incomplete reversal under some cycling conditions [4,5].

Focused-ion-beam cross-sections showed that the 10 mV electrodes thickened by 52 +/- 9% after 120 cycles, compared with 31 +/- 7% for the 80 mV electrodes. The deep-lithiation group also contained more through-thickness cracks and more separated carbon-binder-silicon clusters. These structural changes explain why residual Li-Si peaks can grow even as total active capacity declines: the trapped material is present, but increasingly disconnected from reversible electrochemistry.

The three measurement families therefore agree at the level needed for this study. XRD identifies the phase-resolved ordered component and tracks amorphous scattering. NMR supports retained disordered lithium. XPS constrains SEI-associated lithium. Microscopy explains how active material becomes isolated. None of the methods alone would justify the full partition.

Implications for silicon-anode protocols

The data support three practical recommendations. First, lower-cutoff voltage should be reported alongside silicon loading, electrolyte, binder, and formation protocol. A silicon cell cycled to 10 mV and one cycled to 80 mV are not testing the same phase pathway. Second, capacity retention should be paired with lithium-inventory metrics. Similar retained capacities can hide different balances of SEI loss, trapped Li-Si, and inactive silicon.

Third, operando diffraction should be interpreted with humility. It is excellent for detecting crystalline Si and Li15Si4 and for following broad amorphous signatures when signal-to-background is good. It is poor at quantifying interfacial lithium and silent about lithium in X-ray-amorphous SEI unless paired with chemistry-sensitive probes. Overclaiming XRD-only lithium trapping would obscure the real complexity of silicon ageing.

For electrode design, the restricted-cutoff results argue for suppressing crystalline Li15Si4 when long cycle life and lithium inventory are priorities. However, the remaining amorphous retention and SEI growth show that voltage control is not enough. Stable binders, engineered porosity, robust SEI-forming additives, and particle-size control remain necessary. This conclusion is consistent with the broader silicon-anode literature, where the best cycling often comes from combining multiple design levers rather than from one phase-control rule [7,8,18].

Limitations

The study uses half-cells with lithium-metal counter electrodes. Half-cells are useful for mechanistic work, but lithium metal masks full-cell lithium-inventory constraints and can introduce its own interfacial artefacts. The absolute irreversible capacities should therefore not be transferred directly to graphite-free full cells or commercial silicon-graphite blends.

The electrode loading is moderate, and the operando windowed pouch geometry differs from production cells. Thicker electrodes would have larger gradients in lithiation state, stress, and electrolyte transport. The X-ray beam samples a finite region rather than the whole electrode, so spatial heterogeneity may be underrepresented even though multiple positions were measured.

The amorphous-phase analysis is semi-quantitative. Broad scattering from amorphous LixSi overlaps background and SEI contributions, and its intensity is sensitive to density and morphology. We therefore use amorphous features as inventory constraints, not as absolute phase fractions. Future work should combine operando diffraction with pair-distribution-function analysis, neutron methods, or isotope-sensitive NMR for stronger lithium quantification.

Finally, the study examines one silicon nanoparticle/binder/electrolyte formulation. Different particle sizes, prelithiation strategies, artificial SEI coatings, or silicon-oxide compositions may shift the relative importance of Li15Si4, amorphous trapping, and SEI loss. The methodology should generalise, but the numerical partition should not be treated as universal.

Conclusion

Operando synchrotron XRD reveals how lithium trapping in silicon nanoparticle half-cells evolves from formation through extended cycling. Deep lithiation to 10 mV produces crystalline Li15Si4, and residual Li15Si4 at the charged state grows with cycling. Restricting lithiation to 80 mV suppresses detectable Li15Si4 and improves capacity retention, but it does not eliminate amorphous Li-Si retention or SEI-associated lithium loss.

The main conclusion is that lithium trapping in silicon is phase-distributed and history-dependent. Crystalline Li15Si4 is a visible and important marker under deep lithiation, but it is not the whole trapped-lithium reservoir. Combining operando diffraction with charge balance, NMR, XPS, and morphology provides a more realistic inventory than any single method. Such phase-resolved accounting is necessary for designing silicon-anode protocols that balance early capacity against long-term cyclable lithium retention.

Data and code availability

The supplementary archive contains raw diffraction frames, calibration files, fitted peak tables, electrochemical logs, XPS spectra, NMR spectra, microscopy image stacks, and scripts used for peak fitting, lithium-inventory propagation, and figure generation. The beamline metadata include detector geometry, exposure times, sample positions, and background-reference scans.

References

1. Misra, S. et al. In situ X-ray diffraction studies of (de)lithiation mechanism in silicon nanowire anodes. ACS Nano 6, 5465-5473 (2012).
2. Obrovac, M. N. & Christensen, L. Structural changes in silicon anodes during lithium insertion/extraction. Electrochem. Solid-State Lett. 7, A93 (2004).
3. Obrovac, M. N. & Krause, L. J. Reversible cycling of crystalline silicon powder. J. Electrochem. Soc. 154, A103 (2007).
4. Key, B. et al. Real-time NMR investigations of structural changes in silicon electrodes for lithium-ion batteries. J. Am. Chem. Soc. 131, 9239-9249 (2009).
5. Key, B., Morcrette, M., Tarascon, J.-M. & Grey, C. P. Pair distribution function analysis and solid state NMR studies of silicon electrodes for lithium ion batteries: understanding the (de)lithiation mechanisms. J. Am. Chem. Soc. 133, 503-512 (2011).
6. Beaulieu, L. Y., Eberman, K. W., Turner, R. L., Krause, L. J. & Dahn, J. R. Colossal reversible volume changes in lithium alloys. Electrochem. Solid-State Lett. 4, A137 (2001).
7. McDowell, M. T., Lee, S. W., Nix, W. D. & Cui, Y. Understanding the lithiation of silicon and other alloying anodes for lithium-ion batteries. Adv. Mater. 25, 4966-4985 (2013).
8. Obrovac, M. N. & Chevrier, V. L. Alloy negative electrodes for Li-ion batteries. Chem. Rev. 114, 11444-11502 (2014).
9. Chan, C. K. et al. High-performance lithium battery anodes using silicon nanowires. Nat. Nanotechnol. 3, 31-35 (2008).
10. Sethuraman, V. A., Chon, M. J., Shimshak, M., Srinivasan, V. & Guduru, P. R. In situ measurements of stress evolution in silicon thin films during electrochemical lithiation and delithiation. J. Power Sources 195, 5062-5066 (2010).
11. Lee, S. W., McDowell, M. T., Berla, L. A., Nix, W. D. & Cui, Y. Fracture of crystalline silicon nanopillars during electrochemical lithium insertion. Proc. Natl Acad. Sci. USA 109, 4080-4085 (2012).
12. Liu, X. H. et al. Anisotropic swelling and fracture of silicon nanowires during lithiation. Nano Lett. 11, 3312-3318 (2011).
13. Liu, X. H. et al. Size-dependent fracture of silicon nanoparticles during lithiation. ACS Nano 6, 1522-1531 (2012).
14. Zhu, B. et al. Minimized lithium trapping by isovalent isomorphism for high initial Coulombic efficiency of silicon anodes. Sci. Adv. 5, eaax0651 (2019).
15. Hernandez Bertran, M. A. et al. Understanding the irreversible lithium loss in silicon anodes using multi-edge X-ray scattering analysis. Chem. Mater. 37, 3648-3660 (2025).
16. Cao, C., Shyam, B., Wang, J., Toney, M. F. & Steinrueck, H.-G. Shedding X-ray light on the interfacial electrochemistry of silicon anodes for Li-ion batteries. Acc. Chem. Res. 52, 2673-2683 (2019).
17. Gao, H. et al. Parasitic reactions in nanosized silicon anodes for lithium-ion batteries. Nano Lett. 17, 1512-1519 (2017).
18. Wu, H. et al. Stable cycling of double-walled silicon nanotube battery anodes through solid-electrolyte interphase control. Nat. Nanotechnol. 7, 310-315 (2012).
19. Christensen, J. & Newman, J. A mathematical model for the lithium-ion negative electrode solid electrolyte interphase. J. Electrochem. Soc. 151, A1977 (2004).
20. Pietsch, P., Hess, M., Ludwig, W., Eller, J. & Wood, V. Combining operando synchrotron X-ray tomographic microscopy and scanning X-ray diffraction to study lithium ion batteries. Sci. Rep. 6, 27994 (2016).
21. Kim, M., Yang, Z. & Bloom, I. Review: the lithiation/delithiation behavior of Si-based electrodes: a connection between electrochemistry and mechanics. J. Electrochem. Soc. 168, 010523 (2021).
22. Pan, J. et al. Effects of stress on lithium transport in amorphous silicon electrodes for lithium-ion batteries. Nano Energy 13, 192-199 (2015).
23. Ding, B., Wu, H., Xu, Z., Li, X. & Gao, H. Stress effects on lithiation in silicon. Nano Energy 38, 486-493 (2017).
24. Chan, M. K. Y., Wolverton, C. & Greeley, J. P. First principles simulations of the electrochemical lithiation and delithiation of faceted crystalline silicon. J. Am. Chem. Soc. 134, 14362-14374 (2012).