Low-temperature synthesis of silicon anodes from biosilica via AlCl₃-assisted magnesiothermic reduction

Abstract

Silicon is a high-capacity anode material, yet its scalable production from sustainable precursors requires low-temperature and controllable synthesis routes. Diatom-derived SiO₂ provides an abundant biogenic feedstock, but its conversion to silicon by magnesiothermic reduction (MgTR), typically conducted at 600–900 °C, is limited by the highly exothermic nature of the reaction, which induces local overheating, promotes side-phase formation, and often results in incomplete SiO₂ reduction. Here, we elucidate the reaction pathway of AlCl₃-assisted MgTR as a strategy to decrease synthesis temperature and improve reduction efficiency. By correlating the heating ramp rate, isothermal hold time, and salt-to-silica ratio with phase evolution and the crystalline silicon fraction, we identify the parameters governing oxygen abstraction and Si formation. Time-resolved in situ synchrotron X-ray diffraction provides direct insight into the reaction mechanism, revealing the early formation of metallic Al, the transient formation of MgAl₂Cl₈ as an intermediate, and the subsequent crystallization of Si concurrent with the consumption of metallic Al, thereby suggesting that Al acts as an effective reducing agent. Silicon formation proceeds within a chloride-rich molten phase and is achieved at temperatures as low as 250–300 °C. The crystalline silicon fraction is primarily dictated by heating conditions and AlCl₃ content, with optimized parameters maximizing the Si fraction while suppressing inactive byproducts. Electrochemical evaluation of the graphite-SiO_x electrode blends demonstrates enhanced reversible capacity relative to graphite together with moderate cycling stability, confirming the electrochemical activity of the synthesized material. Overall, this work unveils the mechanistic framework of AlCl₃-assisted MgTR and provides synthesis guidelines for the low-temperature conversion of diatom biosilica into silicon-based anode materials.

---

1 Introduction

The global market for silicon (Si) anodes is forecast to expand rapidly, from USD 327–357 million in 2024 to USD 10–21 billion by 2034, corresponding to compound annual growth rates above 40% across multiple analyses.^1–3 This growth is driven largely by the demand for electric vehicles and the pursuit of higher energy density in lithium-ion batteries (LIBs). Current industrial implementations are based on blending Si into graphite electrodes, providing specific capacities in the range of 1500–2500 mAh g⁻¹, a clear improvement over pure graphite (372 mAh g⁻¹) but still below the theoretical capacity of Si (3579 mAh g⁻¹).⁴ Pushing Si content beyond current industrial limits is a technological necessity,^5–8 demanding scalable, low-temperature, and cost-efficient synthesis pathways to produce battery-grade Si from sustainable feedstock.⁹

Natural SiO₂ sources such as diatom frustules, clays, zeolites, sand, and agricultural residues offer low-cost and sustainable feedstock for silicon anodes, with inherent advantages including hierarchical porosity and high surface area.^10–13 These features facilitate Li-ion transport and help buffer the large volume changes of Si during cycling, making biogenic templates highly attractive for advanced anode design.^14–18 However, conventional carbothermal reduction, the established industrial route for silicon production, requires high processing temperatures of about 1900 °C, which destroys the intrinsic nanostructure and morphology of biogenic templates, thereby limiting its suitability for nanostructured anode design. Magnesiothermic reduction (MgTR, eqn (1)) has been proposed as a moderate-temperature alternative (600–900 °C), successfully applied to diverse SiO₂ sources including diatom frustules, clays, zeolites, sand, and agricultural residues.^19–27 Nevertheless, MgTR remains challenging to control: its strongly exothermic nature induces local hot spots that promote sintering and collapse of fragile nanostructures, while incomplete reduction or local Mg excess/deficiency leads to electrochemically inactive byproducts such as Mg₂Si and Mg₂SiO₄.^28–31 These drawbacks lower the Si yield, complicate purification, and ultimately compromise electrochemical performance.

2Mg(s) + SiO₂(s) → Si(s) + 2MgO(s) (1)

The use of low-melting AlCl₃ (T_m ≈ 192 °C) has recently attracted attention as a means to reduce the effective reaction temperature in metallothermic reductions. Owing to its low melting point, AlCl₃ introduces a transient molten-salt environment at temperatures far below conventional MgTR conditions, which can homogenize reactant distribution, facilitate mass transport, and suppress localized overheating. In aluminothermic and zincothermic reductions, this approach has enabled silicon formation at substantially reduced temperatures (200–250 °C).^32–36 In the context of magnesiothermic reduction, however, the function of AlCl₃ remains unclear. Rather than acting as an inert matrix, several studies indicate that the salt actively reshapes the reaction pathway. Reported intermediates such as MgCl₂, AlOCl, and MgAl₂Cl₈ suggest that chlorine-mediated redox chemistry occurs during MgTR.^37,38 In particular, the in situ formation of metallic Al (eqn (2)) has been proposed, implying that reduction may proceed via a secondary aluminothermic step instead of direct Mg–SiO₂ interaction. Concurrently, the formation of MgAl₂Cl₈ has been attributed to a cross-reaction between residual AlCl₃ and MgCl₂ (eqn (3)). Despite these hypotheses, direct experimental evidence clarifying the sequence of intermediate formation and their thermodynamic hierarchy is still lacking.

3Mg + 2AlCl₃ → 3MgCl₂ + 2Al (2)

MgCl₂ + 2AlCl₃ → MgAl₂Cl₈ (3)

More recently, Je et al.³⁹ proposed a reaction mechanism based on ex situ X-ray diffraction and density functional theory (DFT) calculations, in which metallic Mg dissolves in AlCl₃ to form metal-AlCl₃ complexes that act as reactive intermediates for chlorine transfer. These complexes would destabilize the Si–O bonds in SiO₂ and enable oxygen abstraction via internal Cl⁻ transfer, leading to the formation of elemental Si along with soluble byproducts such as MgCl₂ and AlOCl, rather than solid MgO. These chloride-based byproducts can be removed using mild acid treatments, offering significant advantages in product purification and scalability.

The AlCl₃–Mg system also enables SiO₂ reduction to proceed at significantly lower temperatures (∼250–300 °C), conditions generally associated with reduced particle sintering, enhanced preservation of the original silica nanostructure, and suppression of undesired Mg₂Si and Mg₂SiO₄ formation. In practice, however, protocols based on this reaction commonly rely on HF etching as a standard post-treatment step to remove residual SiO₂, indicating that the reduction frequently remains incomplete. This dependence represents a major limitation: HF processing introduces substantial environmental and safety concerns and can partially dissolve or structurally alter the newly formed silicon, thereby compromising the structural advantages targeted through low-temperature synthesis.

A rigorous elucidation of the reaction pathway governing AlCl₃-assisted MgTR is essential to enable rational control over the process. However, uncovering this pathway is intrinsically challenging because the reduction proceeds within a molten-salt environment, where reactive intermediates nucleate and evolve dynamically and are difficult to capture. Consequently, the AlCl₃-assisted MgTR route remains mechanistically ambiguous,^40–42 hindering the establishment of predictive design rules necessary to fully exploit this low-temperature strategy for silicon synthesis. Notably, existing interpretations rely almost exclusively on ex situ characterization, which does not provide access to the temporal sequence of intermediate formation and phase evolution during the reduction.

In this work, we investigate the AlCl₃-assisted MgTR of diatom-derived SiO₂ obtained from industrially cultured diatoms¹³ by combining systematic ex situ characterization with time-resolved in situ synchrotron X-ray diffraction. This integrated approach enables the establishment of a robust mechanistic and parametric framework for the reduction process, elucidating how key synthesis parameters—including heating ramp, salt content, and isothermal holding time—govern the crystalline Si fraction and phase composition. Finally, we show the electrochemical performance of the resulting SiO_x/graphite electrode blends, thereby providing design principles for the scalable synthesis of sustainable, high-performance Si anodes from nanostructured silica.

2 Experimental

2.1 Materials characterization

Scanning Electron Microscopy (SEM) and Energy-Dispersive X-ray Spectroscopy (EDS) were performed on a Zeiss Sigma 300 instrument to examine particle morphology and elemental composition of both the precursor and reduced samples. To minimize charging effects, all specimens were sputter-coated with a thin Au layer prior to imaging.

Ex situ X-ray diffraction (XRD) was carried out on a PANalytical X'Pert PRO diffractometer equipped with Cu Kα radiation (λ = 1.5406 Å). Rietveld refinement was employed to determine the phase composition and to calculate weight fractions of the post-reduction products. For quantification of the crystalline silicon fraction in the acid-washed SiO_x, the powders were mixed with a known amount of corundum (Al₂O₃) as an internal standard, which allowed quantification of the amorphous content.

Nitrogen physisorption was conducted using a Micromeritics ASAP 2020 system. Samples were degassed at 150 °C for 12 h prior to analysis. The specific surface area (SSA) was determined by the Brunauer–Emmett–Teller (BET) method, and the pore size distribution was obtained from the adsorption branch using the Barrett–Joyner–Halenda (BJH) model.

2.2 Molten salt-assisted magnesiothermic reduction

Amorphous biosilica was obtained from industrially cultured diatom microalgae (Nitzschia sp.,¹³ Swedish Algae Factory), cultivated under controlled conditions to ensure high purity, morphological uniformity, and reproducibility. Magnesium powder (99%, <75 µm, Sigma-Aldrich) was used as the reducing agent, and anhydrous aluminum chloride (AlCl₃, 99%, Sigma-Aldrich) served as the molten-salt.

Reduction reactions were carried out by thoroughly mixing diatom-SiO₂ with Mg and AlCl₃ in defined molar ratios. The SiO₂ : Mg molar ratio was fixed at 1 : 2, while the SiO₂ : AlCl₃ ratio was varied between 1 : 5 and 1 : 10 to investigate the influence of salt content. For each ex situ experiment, the total reactant mass was 200 mg. The powders were manually ground using an agate mortar and pestle inside an Ar filled glovebox until a homogeneous mixture was obtained. The mixtures were transferred to stainless-steel crucibles with lids and placed in a tube furnace under continuous Ar flow. Temperature programs were systematically varied to study the effect of heating parameters. Heating rates of 2, 5, and 10 °C min⁻¹ were employed, followed by isothermal holds at 250 °C for 12–24 h. Representative images of the precursor mixture inside the reactor and the resulting reaction products are provided in Section S2.1 of the SI. After thermal treatment, the products were washed with 1 M HCl, filtered, rinsed with DI water and ethanol, and dried at 80 °C under vacuum overnight.

Samples are denoted using the format R–T–S, where R corresponds to the heating ramp rate (2, 5 or 10 °C min⁻¹), T indicates the isothermal holding time (12 or 24 h), and S denotes the salt content (L-low or H-high). Details of the reaction parameters are summarized in Table 1.

Table 1 Sample IDs and synthesis parameters for AlCl₃-MgTR of diatom-SiO₂. Reactant ratios (SiO₂ : Mg : AlCl₃) are given in moles

Sample ID	SiO2 : Mg : AlCl3	Ramp rate (°C min^−1)	Hold time (h)	Salt content
R5-T12-SL	1 : 2 : 5	5	12	Low
R5-T24-SL	1 : 2 : 5	5	24	Low
R5-T12-SH	1 : 2 : 10	5	12	High
R5-T24-SH	1 : 2 : 10	5	24	High
R2-T24-SH	1 : 2 : 10	2	24	High
R10-T24-SH	1 : 2 : 10	10	24	High

---

2.3 In situ synchrotron XRD of diatom-SiO₂ AlCl₃-MgTR

In situ synchrotron powder XRD measurements were performed at the BM01 station of the Swiss-Norwegian Beamline (SNBL) of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). A monochromatic X-ray beam with a wavelength of 0.706 Å and a beam size of 100 × 80 µm was employed. Diffraction patterns were recorded using a Dectris Pilatus 2 M direct photon counting area detector. Data acquisition time was 10 s. The powder mixtures, consisting of diatom-SiO₂, magnesium and AlCl₃, were sealed in sapphire capillaries (0.7 mm outer diameter) under an inert atmosphere and heated in situ using a furnace available at the beamline,⁴³ which was previously calibrated by heating a Si standard and calculating the thermal expansion of the crystal lattice. A LaB₆ powder standard was used for the detector calibration and the instrumental resolution function calculation. Two-dimensional diffraction images were integrated into one-dimensional patterns using the Dioptas software.⁴⁴ Sequential Rietveld refinements of the integrated diffraction data were carried out using FullProfApp⁴⁵ to analyze the evolution of crystalline phases during the reduction process. The temperature profile was as follows: a ramp of 5 °C min⁻¹ to 250 °C + 30 min holding time at this temperature, followed by heating ramp to 350 °C + 30 min holding time and finally further heating up to 400 °C + 30 min holding time. The sample was then cooled down at 10 °C min⁻¹ to room temperature. Raw datasets corresponding to this experiment are available at the ESRF data portal.⁴⁶

2.4 Electrochemical characterization

Electrode slurries were prepared using a Thinky planetary mixer. For the graphite–SiO_x electrodes, the slurry composition consisted of 94 wt% active material (90 wt% graphite, C-NERGY KS6L, and 10 wt% synthesized SiO_x), 2 wt% carboxymethyl cellulose (CMC) binder, 2 wt% conductive carbon (C65), and 2 wt% styrene–butadiene rubber (SBR) in DI water. To enable direct comparison, two baseline electrodes were also prepared with identical additive contents and the same mixing protocol: (i) a pure graphite electrode containing 94 wt% graphite as the sole active material and (ii) a graphite–SiO₂ electrode with 90 wt% graphite and 10 wt% pristine diatom-SiO₂.

The slurries were cast onto copper foil using a doctor blade with a gap height of 50 µm, followed by drying at 60 °C under vacuum overnight. Circular electrodes with a diameter of 16 mm were punched out and further dried at 120 °C under dynamic vacuum in the antechamber of an Ar-filled glovebox prior to assembly into coin-type half cells. Electrode mass loadings were 1.05 mg ⋅cm⁻². Each cell consisted of a working electrode (graphite, graphite–SiO₂, or graphite–SiO_x), a lithium metal foil counter/reference electrode, and a Whatman GF/A glass fiber separator. The electrolyte was 1 M LiPF₆ dissolved in a 1 : 1 (v/v) mixture of ethylene carbonate (EC) and diethyl carbonate (DEC).

3 Results and discussion

3.1 Pristine diatom-SiO₂ material

Comprehensive characterization of the pristine diatom-derived SiO₂ frustules was previously reported by the authors¹³ and is summarized in Fig. S1 of the SI. The XRD pattern of the diatom SiO₂ exhibits a broad diffraction feature centered at 2θ ≈ 22.5°, characteristic of amorphous silica. Nitrogen adsorption–desorption measurements display a type IV isotherm with a pronounced hysteresis loop, indicative of a mesoporous structure. The BET specific surface area is 69.72 m² g⁻¹, with micropore and external surface areas of 23.51 and 46.21 m² g⁻¹, respectively, as determined by t-plot analysis. At relative pressures below p/p₀ = 0.2, the isotherm exhibits a concave uptake followed by a gradual slope and a sharp increase approaching p/p₀ = 1, while the desorption branch lies slightly above the adsorption curve due to capillary condensation effects. Scanning electron microscopy reveals elongated frustules (approximately 4 µm wide and 11 µm long) with a highly ordered porous architecture, where circular pores of approximately 70 nm in diameter are uniformly distributed along the frustule walls.

3.2 Effect of hold time, salt ratio and heating rate

The influence of isothermal holding time, salt content, and heating ramp rate on reaction products, silicon yield, specific surface area, and morphological characteristics was systematically investigated. XRD patterns of samples prepared with low (SL) and high (SH) salt content at a fixed heating rate of 5 °C min⁻¹, as well as SH samples subjected to different heating ramp rates (2, 5, and 10 °C min⁻¹), are shown in Fig. 1a and c, respectively. The corresponding XRD patterns after acid washing are presented in Fig. 1b and d. For clarity, samples analyzed prior to acid washing are hereafter referred to as “pre-AW”.

Fig. 1 XRD patterns of AlCl₃–MgTR products: (a and b) R5-T12-SL, R5-T24-SL, R5-T12-SH and R5-T24-SH pre-AW (a) and post-AW (b); (c and d) high salt samples R2-724-SH, R5-T24-SH, and R10-T24-SH before (c) and after (d) AW.

Phase analysis of the unwashed samples indicates the formation of typical AlCl₃-assisted MgTR reaction products under both SL and SH conditions, including crystalline Si, Al, MgCl₂·6H₂O, and AlCl₃·6H₂O. In particular, elemental Mg was not detected. After AW, the diffractograms reveal the presence of crystalline Si and a broad peak centered at 2θ ≈ 22.5°, corresponding to residual amorphous SiO₂. The peaks attributed to Al, MgCl₂·6H₂O, and AlCl₃·6H₂O are no longer observed, confirming their effective removal. A consistent low-intensity peak at 2θ ≈ 45°, marked with an asterisk, appears in all AW samples—R5-T12-SL, R5-T24-SL, R5-T12-SH, and R5-T24-SH. This peak does not correspond to any of the known reaction products present before acid washing, suggesting the formation of a minor unidentified phase.

Samples reduced at different heating rates (10, 5, and 2 °C min⁻¹) also show characteristic peaks of AlCl₃·6H₂O, MgCl₂·6H₂O, Al, and Si. After acid washing, the patterns are dominated by prominent Si peaks and the same broad feature corresponding to residual amorphous SiO₂. The sample R2-T24-SH, reduced at 2 °C min⁻¹, exhibits two additional peaks at 2θ ≈ 26.6° and 45.5° that are not associated with any previously identified phases in the pre-AW diffractograms. The samples R5-T24-SH and R10-T24-SH also display the 45° impurity peak, though the 26.6° peak is absent. These low-intensity peaks, annotated with asterisks, are not clearly discernible in the pre-AW data due to background noise. To further investigate the origin of the unidentified reflections, several control experiments were performed (see Section S2.2 of the SI). The results suggest that the reflections likely originate from residual Al- and/or Cl-containing species, although their low diffraction intensity precludes an unambiguous phase assignment.

Rietveld refinement was performed on samples pre- and post-AW for the quantification of the crystalline Si fraction and other post-reduction products. The phase composition of all samples pre-AW is summarized in Section S2.3 of the SI. After AW, weight phase percentages show a clear trend with varying parameters as indicated in Table 2. First, fixing the heating ramp to 5 °C indicates that prolonged isothermal holding time and the increased salt-to-SiO₂ ratio enhance the crystalline Si fraction, with the sample synthesized under SH conditions and 24 h holding time showing 25.52 wt% Si. Among the samples subjected to varying heating ramps under SH conditions and a 24 h holding time, R2-T24-SH displayed the lowest Si content (19.89%) followed by R5-T24-SH (25.52%) and R10-T24-SH (28.71%), which indicates that a faster heating protocol could favor the SiO₂ reduction. Moreover, analysis of post-AW samples confirmed the effective removal of the main chloride salts and metallic aluminum phases, although minor residual species cannot be completely excluded.

Table 2 Phase compositions of AlCl₃-MgTR products post-AW, determined by Rietveld refinement

Sample ID	Phase wt% post-AW
	Crystalline Si	Amorphous content
R5-T12-SL	16.70	83.30
R5-T24-SL	20.14	79.96
R5-T12-SH	19.31	80.69
R5-T24-SH	25.52	74.48
R10-T24-SH	28.71	71.29
R2-T24-SH	19.89	80.11

---

The nitrogen adsorption–desorption isotherms for the pristine SiO₂ frustules, R5-T12-SL, R5-T24-SL, R5-T12-SH, and R5-T24-SH are presented in Fig. 2a. Up to a relative pressure of p/p₀ = 0.95, both low-salt samples exhibit lower adsorption than the precursor. However, beyond this point, the isotherm of R5-T12-SL surpasses that of the precursor, achieving the highest adsorbed gas quantity among the three. All samples show type IV isotherms with hysteresis loops. The isotherms of R5-T12-SH and R5-T24-SH closely resemble those of their low-salt counterparts, confirming the presence of mesoporosity in all samples. Fig. 2b presents the SSA distribution derived from the t-plot method. All MgTR-processed samples exhibit lower SSAs compared to the pristine frustules, which display a total SSA of 69.72 m² g⁻¹—composed of 46.21 m² g⁻¹ external and 23.51 m² g⁻¹ micropore area. For R5-T12-SL, the total SSA is reduced to 44.66 m² g⁻¹ (28.38 mg² g⁻¹ external and 16.28 m² g⁻¹ microporous), while prolonging the isothermal hold time to 24 h (R5-T24-SL) further decreases both components to 26.12 m² g⁻¹ and 12.56 m² g⁻¹, respectively. These results indicate a progressive loss of surface area with extended heat treatment under low salt conditions.

Fig. 2 N₂ physisorption and microstructural characterization of post-AW AlCl₃–MgTR products: (a–d) adsorption–desorption isotherms and corresponding external and micropore surface area contributions for R5-T12-SL, R5-T24-SL, R5-T12-SH, R5-T24-SH, R2-T24-SH, R5-T24-SH, and R10-T24-SH. (e–h) TEM images of selected samples: (e) R5-T24-SL, (f) R2-T24-SH, (g) R5-T24-SH, and (h) R10-T24-SH.

The high-salt samples, R5-T12-SH and R5-T24-SH, exhibit similar micropore areas of 15.06 and 15.44 m² g⁻¹, respectively—lower than the precursor's value. However, in contrast to the low-salt series, the external surface area increases with extended holding time: R5-T24-SH exhibits 37.73 m² g⁻¹ compared to 26.05 m² g⁻¹ for R5-T12-SH. This inverse trend suggests that salt content critically influences the evolution of surface area during MgTR.

In terms of BET surface area, both R5-T12-SL and R5-T24-SL show reduced values (44.66 and 38.68 m² g⁻¹, respectively) compared to the pristine frustules. A corresponding decline is observed in cumulative mesopore surface area, which decreases from 39.21 m² g⁻¹ in the precursor to 15.38 m² g⁻¹ and 16.98 m² g⁻¹ for the 12 h and 24 h samples, respectively. The average mesopore diameter increases significantly upon processing—49.02 nm and 29.49 nm for R5-T12-SL and R5-T24-SL, respectively—suggesting coalescence or collapse of smaller pores during reduction, particularly under prolonged heating. Under high salt conditions, R5-T12-SH exhibits a total SSA of 41.10 m² g⁻¹ and a mesopore surface area of 13.47 m² g⁻¹. Extending the hold time to 24 h increases both values to 53.17 m² g⁻¹ and 28.65 m² g⁻¹, respectively—an opposite trend to that observed in the low-salt series.

Fig. 2c presents the nitrogen isotherms of samples synthesized at different heating ramp rates. All exhibit type IV isotherms with hysteresis at p/p₀ > 0.9. Fig. 2d displays the corresponding SSA distribution, revealing that micropore area remains relatively constant (16 m² g⁻¹) across all ramp rates, while external surface area increases with faster heating: 24.82, 37.73, and 43.25 m² g⁻¹ for R2-T24-SH, R5-T24-SH, and R10-T24-SH, respectively. This trend indicates that a faster heating ramp promotes external surface development during reduction. BET and BJH data (Table 3) further support this observation: total SSA increases with ramp rate, reaching 59.35 m² g⁻¹ for R10-T24-SH, compared to 53.17 m² g⁻¹ and 40.76 m² g⁻¹ for R5-T24-SH and R2-T24-SH, respectively. Similarly, cumulative mesopore surface area increases from 15.23 m² g⁻¹ (2 °C min⁻¹) to 28.65 m² g⁻¹ (5 m² g⁻¹) and 33.54 m² g⁻¹ (10 °C min⁻¹). In contrast, the average mesopore diameter decreases with increasing ramp rate, from 39.94 nm at 2 °C min⁻¹ to 23.44 nm at 5 °C min⁻¹ and 24.17 nm at 10 °C min⁻¹. This suggests that faster heating leads to finer mesopores and greater external surface area.

Table 3 N2 physisorption results for samples synthesized at varying hold times, salt contents and heating ramp rates

Sample ID	SSA [mg^2 g^−1]	Cumulative mesopore surface area [mg^2 g^−1]	Avg. mesopore diameter [nm]
SiO2 frustules	69.72	39.21	15.61
R5-T12-SL	44.66	15.38	49.02
R5-T24-SL	38.68	16.98	29.49
R5-T12-SH	41.10	13.47	38.90
R5-T24-SH	53.17	28.65	23.44
R10-T24-SH	59.35	33.54	24.17
R2-T24-SH	40.76	15.23	39.94

---

Fig. 2e–h present TEM analysis of selected post-AW samples, illustrating the influence of salt content and thermal parameters on the microstructure and phase composition. Fig. 2e shows HAADF-STEM images of the R5-T24-SL sample, synthesized with low salt content. This sample consists of three main components: amorphous SiO₂ frustules, amorphous SiO_x, and crystalline Si. Compared to the other samples, it contains a higher fraction of residual SiO₂ and a lower relative content of SiO_x (x < 1) and crystalline Si. STEM-EELS/EDS mapping confirms the coexistence of these phases, and EDS quantification of SiO_x particles in the mapped region yields an average composition of 74 at% Si, 23 at% O, 2.3 at% Cl, and 0.8 at% Al. By contrast, Fig. 2f displays bright-field TEM images of the R2-T24-SH sample, prepared with high salt content and the slowest ramp rate. This sample contains only amorphous SiO₂ particles and crystalline Si particles, with no evidence of composite domains (i.e., Si nanocrystals embedded in an amorphous matrix). Elemental mapping (EELS and EDS) of the red-framed region confirms Si and O as the main constituents, along with surface-localized carbon, while Mg, Al, and Cl are absent.

Fig. 2g shows the results for the R5-T24-SH sample, prepared with a high salt content and a moderate heating ramp. The sample contains amorphous SiO₂ frustules, amorphous SiO_x (x < 1), and crystalline Si. The magnified region reveals the presence of small Si nanocrystals embedded within an amorphous SiO_x (x ≪ 1) matrix, indicative of partial composite formation. Fig. 2h presents bright-field TEM images of the R10-T24-SH sample, synthesized at the highest ramp rate. Similar to the slow-ramp counterpart (R2-T24-SH), this sample contains only amorphous SiO₂ and crystalline Si particles, with no evidence of composite domains. Elemental maps from the red-framed STEM region confirm Si and O as the dominant elements, with only trace amounts of Al and no detectable Mg or Cl. Taken together, these observations highlight that composite formation is highly sensitive to the heating ramp and salt content, with intermediate conditions favoring the partial embedding of crystalline Si within an amorphous SiO_x matrix. At this point, it is worth noting that the apparent preservation of the precursor morphology is associated with the fraction of the material that remains unreacted or only partially reduced. In contrast, regions where crystalline Si particles are formed do not retain the original morphology. This behavior agrees with previous studies on AlCl₃-assisted reductions, where silica undergoes structural disintegration prior to Si crystallization.^39,47,48

3.3 Reaction mechanism by in situ synchrotron XRD

Due to the low Si yield obtained after the AlCl₃-MgTR, time resolved in situ synchrotron XRD measurements were performed to elucidate the reaction mechanism. The same reactant stoichiometries as those used in the previous sections (SiO₂ : Mg : AlCl₃ molar ratios of 1 : 2 : 5 and 1 : 2 : 10) were employed to investigate the effect of the SL and SH concentrations.

An initial in situ experiment was conducted to define appropriate measurement conditions and temperature protocols, as shown in Figure S4 of the SI. The sample was first heated to 220 °C and held isothermally, followed by a second isothermal step at 250 °C and a final heating stage to 300 °C, where the temperature was maintained for 2 h. Analysis of the diffraction data revealed that an isothermal step of 30 min was sufficient for phase evolution to stabilize at each temperature. Moreover, a markedly higher reaction progress was observed between 250 and 300 °C, identifying this interval as critical for the reduction process. Since the in situ measurements are specifically designed to investigate the reaction dynamics and phase evolution, the thermal protocol initially explored the same synthesis temperature used in the ex situ experiments (250 °C) and was subsequently extended stepwise to 400 °C, with isothermal holding times of 30 min, in order to expand the explored temperature range and gain further insights into the reaction pathway. Nevertheless, the key mechanistic events of the reduction occur within the temperature range investigated under ex situ conditions.

3.3.1 Low salt conditions. Fig. 3 presents the results of the study under SL conditions. The evolution of the diffractograms collected during the experiment, together with the corresponding heating protocol, is displayed in Fig. 3a and b, respectively. Rietveld refinement was employed to quantify the phase weight fractions throughout the experiment. However, the pronounced asymmetry observed in the AlCl₃ (020) reflection (see Fig. S5 of the SI) prevented a reliable refinement at the early stage of the reaction. Consequently, Rietveld refinement was only performed once the AlCl₃ had completely melted, as shown in Fig. 3c and d. A more detailed discussion about this issue is provided in Section 3.2 of the SI. Based on the evolution of the main Bragg reflections, the reaction mechanism can be divided into four distinct stages, separated by dashed lines and marked with different colors in the images.

Fig. 3 In situ study of AlCl₃-MgTR under SL conditions: (a) time-resolved diffraction data, (b) corresponding heating protocol, (c) refined scale factors of the detected crystalline phases and (d) corresponding phase weight fractions. Dashed lines and shaded regions delineate the distinct reaction stages. The red dashed box highlights the region where the diffractograms exhibit strong diffuse scattering.

3.3.1.1 Initial heating stage. During the first stage, Bragg peaks associated with crystalline AlCl₃ and Mg are clearly visible in the diffractograms, while the amorphous SiO₂ lacks long-range order, therefore appearing as a broad intensity feature at 2θ ≈ 10°. No evidence of reaction is observed in this stage.
3.3.1.2 Molten salt activation. The second stage is characterized by the onset of Al reflections and the decrease in intensity of Mg peaks, indicating the reaction between AlCl₃ and Mg, which leads to the formation of the mixed salt MgAl₂Cl₈ as an intermediate phase and MgCl₂ as the final product. From 100 °C onward, trace amounts of MgAl₂Cl₈ are detected. However, the formation of this phase rapidly accelerates as the temperature approaches the AlCl₃ melting point. When intensity of AlCl₃ reflections starts to decrease, a transient signal of diffuse scattering emerges, indicating that the salt is entering its molten state. Subsequently, this signal progressively decreases while MgAl₂Cl₈ rapidly crystallizes. At ≈175 °C the absence of AlCl₃ reflections indicates that the salt has either completely reacted with Mg to form Al and MgAl₂Cl₈, or has entered the molten state. Between this temperature and ∼220 °C, the formation of metallic Al continues, while the mixed salt reaches its maximum weight fraction (84.7%). Upon further heating up to 235 °C, MgAl₂Cl₈ is rapidly consumed, leading to the formation of MgCl₂, while Mg content continues to decrease and metallic Al increases once again. Simultaneously, a pronounced increase in the diffuse background intensity is observed (as marked with a dashed box in Fig. 3a), which is consistent with the formation of a molten phase. At the end of this stage, the weight concentration of crystalline phases is 57% MgCl₂, 30% Al and 13% Mg. An example of Rietveld refinement of one diffractogram collected during this stage is shown in Fig. S6 of the SI.
3.3.1.3 SiO₂ reduction. In this stage, the partial reduction of the amorphous SiO₂ precursor and the crystallization of Si take place, while the temperature profile includes three 30 min holding steps at 250, 350 and 400 °C, followed by a cooling ramp down to room temperature. During the first holding step (250 °C), Si and the final byproduct AlOCl slowly crystallize, as indicated by the evolution of their refined intensities in Fig. 3c, while metallic Al is consumed to a minor extent and MgCl₂ content remains essentially unchanged. It is worth noting that the phase weight concentration graph may not fully reflect the actual phase evolution since some phases that were not previously detected (such as molten or amorphous phases) become detectable at this stage. Therefore, for a correct interpretation, the refined scale factors and phase weight graphs (Fig. 3c and d) should be analyzed simultaneously. Upon further heating up to 350 °C, the reduction reaction accelerates, leading to a more pronounced formation of Si, AlOCl and MgCl₂, together with the total consumption of the metallic Al. In addition, once 350 °C is reached, the strong diffuse scattering signal disappears, which indicates the consumption of the molten phase formed at the end of the previous stage. From this temperature onward, no significant reaction signal is observed. Moreover, during the cooling ramp, no crystallization of AlCl₃ is detected, confirming the complete consumption of the salt during the reaction. An example of Rietveld refinement of one diffractogram collected during this stage is shown in Fig. S7 of the SI. A simple mass balance shows that the crystalline products detected by in situ XRD (Si, AlOCl and MgCl₂) cannot account for all Al and Cl introduced as AlCl₃. This indicates that a fraction of the Al/Cl inventory remains in a noncrystalline reservoir (molten or amorphous chloride species) and/or redistributes along the sealed capillary through the gas–condensed equilibrium of AlCl₃, potentially condensing in cooler regions outside the XRD sampling volume.

3.3.2 High salt conditions. Fig. 4 summarizes the in situ study conducted under SH conditions. The contour plot shown in Fig. 4a, together with the corresponding heating protocol in Fig. 4b, allows the evolution of the reaction components to be followed throughout the entire experiment. The results of the Rietveld refinement are summarized in Fig. 4c and d, where the refined scale factors and the phase weight concentration of each phase are displayed, respectively. As in the low salt study, Rietveld refinement was performed once AlCl₃ had completely melted. Unlike the SL study, the reaction under SH conditions can be divided into four distinct stages. While the overall reaction pathway is similar to that observed in the SL experiment, with the transient formation of MgAl₂Cl₈ and the temporal correlation between metallic Al consumption and Si crystallization, significant differences are found in the products formed during the reaction.

Fig. 4 In situ study of AlCl₃-MgTR under SH conditions: (a) time-resolved diffraction data, (b) corresponding heating protocol, (c) refined scale factors of the detected crystalline phases and (d) corresponding phase weight fractions. Dashed lines and shaded regions delineate the distinct reaction stages. The red dashed box highlights the region where the diffractograms exhibit strong diffuse scattering.

No signal of reaction is observed during the initial heating stage. When the temperature is further increased up to ∼125 °C, within the molten salt activation stage, the first indications of MgAl₂Cl₈ formation appear. At ∼170 °C, the formation of the mixed salt accelerates, leading to the crystallization of metallic Al and the partial consumption of Mg, indicating no significant difference in the activation onset temperature compared to the SL case. From this temperature to 235 °C, the mixed salt disappears from the diffractograms, while metallic Al continues to form. Mg is further consumed and a strong diffuse scattering signal emerges in the background, as indicated by the red box in Fig. 4a. Therefore, the molten salt activation stage follows essentially the same reaction pathway as in the low salt experiment, with one notable difference: no crystallization of MgCl₂ is observed by the end of this stage. The SiO₂ reduction stage also follows a similar pathway. During the first holding step, at 250 °C, Si and AlOCl reflections slowly emerge while metallic Al is progressively consumed. This process further accelerates upon heating up to 350 °C, beyond which no significant reaction signal is detected. Moreover, the scattering signal associated with the molten phase persists throughout the entire reduction stage. Upon cooling, reflections corresponding to a new crystalline structure abruptly emerge at 230 °C, while the diffuse scattering signal disappears, indicating the crystallization of a previously molten phase. Importantly, this phase is not residual AlCl₃, but is instead consistent with the MgAl₂Cl₈ structure, confirming the complete consumption of the salt during the reaction. However, its quantification is not possible since it crystallized as a polycrystalline phase rather than as a powder phase, as shown in Figure S8 of the SI. It should be mentioned that under these SH conditions, the Al and Cl supplied by AlCl₃ exceed the amount that can be accommodated in the crystalline phases detected by XRD. The excess is therefore attributed to a chloride-rich molten or amorphous reservoir that gives rise to the diffuse scattering observed at high temperature and only partially crystallizes upon cooling.

3.3.3 Reaction pathways and the mechanistic framework. The in situ studies presented above provide the first time-resolved evidence of the reaction pathway of the AlCl₃-MgTR, enabling direct tracking of the SiO₂ reduction process and revealing mechanistic features that could not be resolved by previous ex situ approaches. In particular, the early formation of metallic Al prior to Si crystallization, together with the persistence of a chloride-rich molten phase throughout the Si formation step, highlights the highly dynamic reaction environment. While a chlorine transfer mechanism was proposed in a previous study,³⁹ the present results suggest that, if operative, this pathway would involve the consumption of the metallic Al generated during the molten salt activation stage, rather than the Mg acting as the active reducing species. Notably, under both SL and SH conditions, Si crystallization ceases once metallic Al is fully consumed, despite Al not being expected to act as an effective reducing agent at such low temperatures. This behavior underscores the critical role of the chlorine-rich molten phase in lowering the effective reaction temperature and sustaining Si formation. A further mechanistic finding concerns the origin of the mixed salt MgAl₂Cl₈. Contrary to previous assumptions,³⁸ MgAl₂Cl₈ does not form via a cross reaction between AlCl₃ and MgCl₂. Instead, the in situ data demonstrate that it is the first crystalline phase to emerge during the molten salt activation stage, under both SL and SH conditions, forming directly through the interaction between AlCl₃ and Mg. This temporal sequence provides strong indications that MgAl₂Cl₈ acts as an intermediate in the generation of metallic Al.

To further probe the role of metallic Al in the reaction pathway and to detect when the reduction of SiO₂ starts, an additional in situ experiment was performed under SH conditions, using crystalline SiO₂ as the precursor (see Section S3.5 of the SI). Unlike the amorphous SiO₂, using the crystalline precursor enables direct monitoring of SiO₂ consumption through its diffraction reflections. Results from this experiment show the concurrent consumption of metallic Al and crystalline SiO₂, providing additional evidence that supports the participation of metallic Al during the reduction.

Finally, comparison between SL and SH conditions highlights the influence of AlCl₃ content on the evolution of intermediate species and final products. As previously mentioned, MgAl₂Cl₈ disappears at temperatures of approximately 235 °C, indicating a transition to a molten chloride-rich phase. This behavior is consistent with the reported AlCl₃–MgCl₂ phase diagram, which shows a pronounced depression of the melting temperature near 67 mol% AlCl₃, corresponding to the stoichiometric composition of MgAl₂Cl₈.⁴⁹ The evolution of the molten phase strongly depends on the AlCl₃ content. Under SL conditions, the system shifts away from this AlCl₃-rich composition during heating, promoting the formation of MgCl₂ and metallic Al. In contrast, under SH conditions the higher AlCl₃ content keeps the system within the AlCl₃-rich regime, enabling the recrystallization of MgAl₂Cl₈ upon cooling.

3.4 Electrochemical performance

The electrochemical performance of the synthesized SiO_x materials, corresponding to the samples characterized ex situ, was evaluated in graphite-blended electrodes and benchmarked against pure graphite and a graphite–SiO₂ (Fig. 5a and b). A wt% ratio of 10/90 (SiO_x/Gr) was selected to test the synthesized material in an industrially relevant configuration. Graphite delivered an initial coulombic efficiency (ICE) of 84.5% and a formation-cycle lithiation capacity of 384 mAh g⁻¹. Incorporation of SiO₂ slightly increased the capacity to 395 mAh g⁻¹ but reduced the ICE to 79.1%, likely reflecting irreversible Li consumption associated with conversion and alloying reactions.^14,50

Fig. 5 Voltage profile curves of (a) Gr, (b) Gr/SiO₂, (c) Gr/R5-T12-SH, (d) Gr/R5-T24-SH, (e) Gr/R2-T24-SH, and (f) Gr/R10-T24-SH electrodes for the formation cycle, cycle 10, and cycle 50. (g) Evolution of lithiation capacity against the cycle number for all electrodes (a–f).

Gr-SiO_x electrodes exhibited markedly higher initial lithiation capacities (460–518 mAh g⁻¹; Fig. 5c–g and Table 4). This capacity enhancement was accompanied by a reduction in ICE (77.6–81.7%), with samples containing higher crystalline Si fractions (R10-T24-SH and R5-T24-SH) showing the lowest ICE values, indicative of increased irreversible reactions during SEI formation and the first alloying cycle. In contrast, R5-T12-SH and R2-T24-SH, with the lowest Si content, exhibited the highest ICE (81.7%).

Table 4 Summary of electrochemical performance, for baseline electrodes and Gr-SiO_x electrodes. The Si yield of each SiO_x material is added for reference

Sample ID	Si [wt%]	ICE [%]	Cycle 1	Cycle 10	Cycle 50
Gr	—	84.52	384	330	331
Gr-SiO2	—	79.12	395	309	310
R5-T12-SH	19.31	80.28	494	350	336
R5-T24-SH	25.52	77.94	509	350	342
R10-T24-SH	28.71	77.56	518	337	324
R2-T24-SH	19.89	81.65	460	325	314

---

After 50 cycles, the reversible capacities of the Gr-SiO_x electrodes stabilized between 314 and 342 mAh g⁻¹. R5-T12-SH and R5-T24-SH retained 68.4% and 67.2% of their initial lithiation capacities, respectively, and retained capacities above that of graphite (331 mAh g⁻¹). In contrast, R10-T24-SH and R2-T24-SH exhibited faster capacity decay and dropped below the graphite baseline within 20–30 cycles, highlighting the trade-off between Si content and cycling stability. All electrodes reached coulombic efficiencies above 99.8% after 30 cycles, while graphite achieved the highest CE after 50 cycles (99.98%), closely followed by R5-T12-SH (99.93%).

Differential capacity (dQ/dV) analysis (Fig. S10 of the SI) supports these conclusions. While Gr and Gr-SiO₂ showed only graphite-related features, Gr-SiO_x electrodes exhibited additional peaks at 0.04 V and 0.45 V associated with the lithiation and delithiation of crystalline Li₁₅Si₄. These features progressively weakened and disappeared by 30 cycles, indicating early formation but rapid electrochemical deactivation of crystalline Si. The more persistent Li₁₅Si₄ signal observed for R2-T24-SH is consistent with its faster capacity decay. Overall, R5-T12-SH and R5-T24-SH deliver the most balanced performance.

The observed electrochemical trends appear to be influenced by the structural properties and phase composition of the synthesized SiO_x material. In particular, the higher initial capacity correlates with the higher crystalline Si fraction, and it is also associated with a lower ICE. However, this enhancement cannot be exclusively attributed to the crystalline Si fraction, since contributions from the amorphous content, different surface area, and irreversible reactions occurring during lithiation of SiO₂ particles may also play an important role. The present data do not allow all these contributions to be disentangled. These results nevertheless confirm the activity of the synthesized SiO_x material when incorporated into SiGr blended electrodes.

4 Conclusions

This work provides a mechanistic and parametric framework for AlCl₃-assisted magnesiothermic reduction (MgTR) of diatom-derived biosilica and its application in silicon-based anodes for lithium-ion batteries. By systematically varying the heating ramp rate, isothermal holding time, and salt-to-silica ratio, we demonstrate that silicon yield and structural preservation are governed by the combined effects of thermal history and molten-salt chemistry.

Time-resolved in situ synchrotron X-ray diffraction provides important insights into the AlCl₃-MgTR reaction mechanism. The results show the early formation of metallic Al and later consumption and the persistence of a chlorine-rich molten phase throughout Si crystallization, enabling SiO₂ reduction at temperatures as low as 300 °C. The mixed salt MgAl₂Cl₈ is identified as the first crystalline intermediate, forming directly from the interaction of AlCl₃ with Mg and acting as a precursor to metallic Al generation. These findings support that the Si formation proceeds through the previously reported chlorine transfer mechanism, but with metallic Al acting as the effective redox-active species rather than Mg.

Ex situ structural analysis shows that extended holding times and higher AlCl₃ contents increase crystalline Si yield, while faster heating ramps promote higher external surface area and finer mesoporosity. Importantly, only intermediate synthesis conditions favor the formation of Si nanocrystals embedded in an amorphous SiO_x matrix, which provides an optimal balance between electrochemical activity and structural stability.

Electrochemical evaluation of Gr-SiO_x composite anodes confirms the electrochemical activity of the synthesized powders. Materials containing moderate crystalline Si fractions (20–25 wt%) and that preserve mesoporosity exhibit the most favorable electrochemical response, delivering CE above 99.8% after moderate cycling time. In contrast, samples with a higher crystalline Si fraction show faster capacity fading, suggesting that electrochemical performance is influenced by multiple structural factors.

Overall, this study demonstrates that the performance of biosilica-derived SiO_x is not dictated by silicon content alone, but by the interplay of crystallinity, surface area, and impurity phases controlled through synthesis parameters. By combining in situ mechanistic insights with systematic structural and electrochemical analysis, this work establishes clear design principles for AlCl₃-assisted MgTR, advancing a low-temperature and scalable route for sustainable silicon anode materials.

Author contributions

Pedro Alonso-Sánchez: conceptualization, data curation, formal analysis, investigation, supervision, methodology, validation, writing – original draft, writing – review and editing. Emilie Hvidsten Swensen: conceptualization, data curation, formal analysis, investigation, methodology, validation, writing – review and editing. Kesavan Thangaian: investigation, methodology, writing – review and editing. Per Erik Vullum: formal analysis, investigation, validation, writing – review and editing. Vadim Diadkin: resources, writing – review and editing. Fride Vullum-Bruer: funding acquisition, validation, writing – review and editing. Javier Campo: funding acquisition, supervision, validation, writing – review and editing. Ann Mari Svensson: funding acquisition, supervision, validation, writing – review and editing. Federico Cova: conceptualization, data curation, investigation, supervision, validation, writing – original draft, writing – review and editing. Maria Valeria Blanco: conceptualization, funding acquisition, investigation, supervision, validation, writing – original draft, writing – review and editing.

Conflicts of interest

There are no conflicts to declare.

Data availability

The data that support the findings of this study are openly available in the ESRF data portal at https://doi.esrf.fr/10.15151/ESRF-ES-2166874660.

Supplementary information (SI) is available. See DOI: https://doi.org/10.1039/d6ta02278b.

Acknowledgements

The authors acknowledge financial support from the SUSTBATT project (M-ERA.NET, Project No. 337463), funded by the Research Council of Norway (Project No. 315947). This work was also supported by Grant No. PCI2022-132993, funded by MCIN/AEI/10.13039/501100011033 and by the European Union “NextGenerationEU”/PRTR. Support from the Research Council of Norway to NORTEM (Project No. 197405) is also gratefully acknowledged. Beamtime at BM01 (SNBL), ESRF, through experiment MA-6477, is acknowledged. The authors thank the Swedish Algae Factory for providing the diatom SiO₂ material.

Notes and references

1. P. Research, Silicon Anode Battery Market Size, Share and Trends 2026 to 2035, https://www.precedenceresearch.com/silicon-anode-battery-market.
2. R. Zope, Silicon Anode Battery Market Size, Share and Growth Forecast 2026–2033, https://www.persistencemarketresearch.com/market-research/silicon-anode-battery-market.asp.
3. Market.us, Silicon Anode Battery Market, https://market.us/report/global-silicon-anode-battery-market/.
4. M. N. Obrovac and L. Christensen, Electrochem. Solid-State Lett., 2004, 7, A93.
5. P. Schweigart, W. Hua, P. A. Sánchez, C. Lian, I. Nylund, D. Wragg, S. Y. Lai, F. Cova, A. M. Svensson and M. V. Blanco, Small, 2025, 21, 2406615.
6. P. Alonso-Sánchez, W. Hua, K. Thangaian, P. E. Vullum, J. T. A. Karlsen, A. M. Svensson, F. Vullum-Bruer, J. Campo, F. Cova and M. V. Blanco, Small, 2025, 21, e2504704.
7. M. Gautam, G. K. Mishra, K. Bhawana, C. S. Kalwar, D. Dwivedi, A. Yadav and S. Mitra, ACS Appl. Mater. Interfaces, 2024, 16, 45809–45820.
8. E. Kim, H. An, I. Kang, C. Lee, M. An, S. Chae and Y. Son, J. Power Sources, 2025, 632, 236314.
9. M. V. Blanco and M. R. Palacin, J. Mater. Chem. A, 2025, 13, 21421–21435.
10. N. Liu, K. Huo, M. T. McDowell, J. Zhao and Y. Cui, Sci. Rep., 2013, 3, 1919.
11. J. Liu, P. Kopold, P. A. van Aken, J. Maier and Y. Yu, Angew. Chem., Int. Ed., 2015, 54, 9632–9636.
12. M. V. Blanco, V. Renman, F. Vullum-Bruer and A. M. Svensson, RSC Adv., 2020, 10, 33490–33498.
13. K. Thangaian, W. Hua, J. T. Aga Karlsen, I.-E. Nylund, S. Nilsson, T. Ericson, M. Hahlin, A. M. Svensson and M. V. Blanco, ACS Sustainable Resour. Manage., 2024, 1, 767–777.
14. W. Hua, P. E. Vullum, K. N.-N. Hjelseng, J. Hamonnet, P. Alonso-Sánchez, J. Zhu, Z. Hegedüs, J. R. Zuazo, F. Cova, A. M. Svensson and M. V. Blanco, Energy Environ. Mater., 2025, 8, e70074.
15. W. Hua, I.-E. Nylund, F. Cova, A. M. Svensson and M. V. Blanco, Sci. Rep., 2023, 13, 20447.
16. K. Thangaian, A. Gaarud, I.-E. Nylund and M. V. Blanco, ACS Sustainable Resour. Manage., 2024, 1, 2284–2293.
17. B. Campbell, R. Ionescu, M. Tolchin, K. Ahmed, Z. Favors, K. N. Bozhilov, C. S. Ozkan and M. Ozkan, Sci. Rep., 2016, 6, 33050.
18. J. R. Szczech and S. Jin, Energy Environ. Sci., 2011, 4, 56–72.
19. N. Wan, L. Wang, S. Li, L. Shen, F. Xi, J. Lu, Z. Tong, X. Chen and W. Ma, Small, 2026, 22, 2412705.
20. N. Kim, H. Park, N. Yoon and J. K. Lee, ACS Nano, 2018, 12, 3853–3864.
21. A. Darghouth, S. Aouida and B. Bessais, Silicon, 2021, 13, 667–676.
22. Z. Lin, P. Sun, C. Zhou and Z. Z. Fang, ACS Omega, 2025, 10, 473–483.
23. T. Autthawong, O. Namsar, A. Yu and T. Sarakonsri, J. Mater. Sci.: Mater. Electron., 2020, 31, 9126–9132.
24. L. A. September, N. Kheswa, N. S. Seroka and L. Khotseng, RSC Adv., 2023, 13, 1370–1380.
25. A. Gaarud, K. Thangaian, P. Alonso-Sánchez and M. V. Blanco, Adv. Sustain. Syst., 2025, 9, 2500117.
26. X. Li, P. Yan, B. W. Arey, W. Luo, X. Ji, C. Wang, J. Liu and J.-G. Zhang, Nano Energy, 2016, 20, 68–75.
27. K. Thangaian, T. Ericson, P. E. Vullum, P. Alonso-Sánchez, A. C. Svarverud, A. M. Svensson, F. Vullum-Bruer, M. Hahlin and M. V. Blanco, J. Power Sources, 2025, 641, 236837.
28. W. Choi, J. Bae, H. Kim, C. Son, M. Karuppaiah and J. K. Lee, Electrochim. Acta, 2024, 498, 144687.
29. W. C. Cho, H. J. Kim, H. I. Lee, M. W. Seo, H. W. Ra, S. J. Yoon, T. Y. Mun, Y. K. Kim, J. H. Kim, B. H. Kim, J. W. Kook, C.-Y. Yoo, J. G. Lee and J. W. Choi, Nano Lett., 2016, 16, 7261–7269.
30. B. Zhang, F. Wang, J. Chen, B. Li, K. Liu and Q. Han, Silicon, 2022, 14, 8409–8416.
31. P. Alonso Sánchez, K. Thangaian, O. A. Øie, A. Gaarud, M. Rodríguez Gomez, V. Diadkin, J. Campo, F. H. Cova and M. V. Blanco, ACS Appl. Energy Mater., 2025, 8, 2249–2259.
32. N. Lin, Y. Han, J. Zhou, K. Zhang, T. Xu, Y. Zhu and Y. Qian, Energy Environ. Sci., 2015, 8, 3187–3191.
33. J. Niu, D. Shen, Z. Ren, R. Zhang, D. Xia, Y. Yang, W. Dong and S. Yang, ChemistrySelect, 2025, 10, e02379.
34. M. Cai, Z. Zhao, J. Qu, Q. Ma, X. Qu, L. Guo, H. Xie, D. Wang and H. Yin, J. Mater. Chem. A, 2021, 9, 21323–21331.
35. Z. Zhao, M. Cai, H. Zhao, Q. Ma, H. Xie, P. Xing, Y. X. Zhuang and H. Yin, ACS Sustain. Chem. Eng., 2022, 10, 5035–5042.
36. Z. Zhao, M. Cai, Y. Zhao, H. Xie, Y. X. Zhuang and H. Yin, ACS Appl. Nano Mater., 2023, 6, 502–511.
37. X. Lin, A. Li, D. Li, H. Song and X. Chen, ACS Appl. Mater. Interfaces, 2020, 12, 15202–15210.
38. N. Lin, Y. Han, L. Wang, J. Zhou, J. Zhou, Y. Zhu and Y. Qian, Angew. Chem., Int. Ed., 2015, 54, 3822–3825.
39. M. Je, J. C. Kim, J. Kim, S. Kim, S. Ryu, J. Ryu, S. K. Kwak and S. Park, Advanced Science, 2025, 12, 2412239.
40. Z.-W. Zhou, Y.-T. Liu, X.-M. Xie and X.-Y. Ye, Chem. Commun., 2016, 52, 8401–8404.
41. X. Wan, Z. Tang, J. Chen, Y. Xue, J. Zhang, X. Guo, Y. Liu, Q. Kong, A. Yuan and H. Fan, Chem. Lett., 2019, 48, 1547–1550.
42. C.-H. Zheng, G.-P. Zhang, S.-S. Wang, A.-Q. Mao and D.-L. Fang, J. Alloys Compd., 2021, 875, 159974.
43. K. P. Marshall, H. Emerich, C. J. McMonagle, C. A. Fuller, V. Dyadkin, D. Chernyshov and W. v. Beek, J. Synchrotron Radiat., 2023, 30, 267–272.
44. C. Prescher and V. B. Prakapenka, High Press. Res., 2015, 35, 223–230.
45. O. Arcelus, J. Rodríguez-Carvajal, N. A. Katcho, M. Reynaud, A. P. Black, D. Chatzogiannakis, C. Frontera, J. Serrano-Sevillano, M. Ismail, J. Carrasco, F. Fauth, M. R. Palacin and M. Casas-Cabanas, J. Appl. Crystallogr., 2024, 57, 1676–1690.
46. P. Alonso, M. V. Blanco and E. H. Swensen, In-situ One-Pot Synthesis of SiOx/C Anodes by SAXS/WAXS Measurements, 2028.
47. G. Song, J. Ryu, J. C. Kim, J. H. Lee, S. Kim, C. Wang, S. K. Kwak and S. Park, Commun. Chem., 2018, 1, 42.
48. J. Ryu, J. H. Seo, G. Song, K. Choi, D. Hong, C. Wang, H. Lee, J. H. Lee and S. Park, Nat. Commun., 2019, 10, 2351.
49. M.-A. Einarsrud, H. Justnes, E. Rytter and H. Øye, Polyhedron, 1987, 6, 975–986.
50. M. Khan, X. Ding, H. Zhao, Y. Wang, N. Zhang, X. Chen and J. Xu, J. Electron. Mater., 2022, 51, 3379–3390.

---