Silicon Anode: How It Works and Why It Swells

A silicon anode is the negative electrode of a lithium-ion cell in which silicon, rather than graphite alone, stores the lithium. Silicon can hold roughly nine times more lithium per gram than graphite (a theoretical 3579 mAh g⁻¹ against about 360 to 372 mAh g⁻¹), but it expands by around 300% or more when fully lithiated, and that swelling is the root of most of its failure modes.

This page explains how a silicon anode works, why the numbers quoted in papers differ, how the swelling turns into lost capacity, and which reporting metrics let a reader judge a claim. Everything below is drawn from three open-access, peer-reviewed reviews published in 2025 and 2026, listed under Sources with the date they were read (2026-11-12). It is a concept explainer for researchers and lab staff, not a materials-purchasing guide.

Short answer

Silicon is attractive as an anode material because it alloys with lithium instead of just hosting it between layers. Fully lithiated to the Li₁₅Si₄ phase, its theoretical capacity is 3579 mAh g⁻¹, more than nine times the value for graphite (Nepal and Dhakal, 2026). The cost is a volume change of up to about 300% at full alloying, against under 15% for graphite. That expansion cracks particles and the surface film (the solid electrolyte interphase, SEI), consumes lithium on every cycle, and lowers the first-cycle efficiency and the long-term Coulombic efficiency that determine how many cycles a full cell can survive.

Definition and scope

In a lithium-ion cell the anode is the electrode that takes up lithium during charging. Graphite is the established anode material: lithium ions slot between its carbon layers, and the lattice changes little. Silicon works differently. Lithium ions react with silicon to form lithium–silicon alloy phases, which is why the capacity is so much higher and also why the lattice must physically make room for the added atoms.

Three related terms are worth separating, because the literature uses them loosely:

• Silicon anode usually means an electrode in which silicon is the main or a major active material, whether as nanoparticles, a silicon–carbon composite or a silicon oxide (SiOx).
• Graphite–silicon composite or blend means a graphite anode with a fraction of silicon added. Nepal and Dhakal (2026) describe graphite–silicon composites as an important current industrial approach, while noting that cell-level energy densities approaching 500 Wh kg⁻¹ remain challenging with graphite-dominant silicon blends.
• Silicon-based anodes for solid-state batteries pair silicon with a solid electrolyte instead of a liquid one. Zhang et al. (2025) review this case separately; the silicon–solid-electrolyte combination is the subject of that review.

For context on a different high-capacity chemistry that also fades by a well-understood mechanism, see the CASRAI guide to the lithium-sulfur battery.

How a silicon anode works

Lithiation and the capacity figure

During charging, lithium enters the silicon and forms alloy phases up to Li₁₅Si₄, which corresponds to the 3579 mAh g⁻¹ theoretical capacity quoted in all three reviews. Zhang et al. (2025) also give 4200 mAh g⁻¹ for the fully lithiated phase; the paper writes that phase as Li₂₂Si₄, which is usually written Li₂₂Si₅ elsewhere, so the safest wording is “the fully lithiated phase, 4200 mAh g⁻¹”. Which of the two numbers a paper uses depends on which alloy phase it treats as the end point.

Why it swells

Packing that much lithium into silicon enlarges the lattice. Nepal and Dhakal (2026) give an expansion of up to about 300% at full alloying to Li₁₅Si₄. Zhang et al. (2025) state more than 300%, up to 300 to 400%, and list 420% in their Table 1. Pavlovskii et al. (2026) say up to 300% in the introduction and up to about 400% for bulk silicon in a later figure. These are not contradictions to be resolved by picking one; they reflect different silicon forms, different end-point phases and different measurement conditions, so the number should always be quoted with its source and basis.

The failure chain: expansion, SEI, efficiency

The reviews describe one connected sequence rather than several independent problems.

1. Particles crack. Repeated expansion and contraction stress the silicon. For crystalline silicon, Zhang et al. (2025) report that when the particle diameter is below 150 nm the particles neither crack nor pulverize; amorphous silicon has a larger critical size of about 870 nm.
2. The SEI breaks and re-forms. The SEI is the thin film that forms on the anode surface from electrolyte decomposition. Because it forms on silicon in an expanded state, it cracks when the particle contracts, exposing fresh silicon that reacts again and builds new SEI (Zhang et al., 2025). Nepal and Dhakal (2026) describe the SEI on silicon as highly unstable due to its dynamic nature: thinner and more inorganic on lithiation, thicker and more organic on delithiation.
3. Lithium is lost. Each SEI regeneration consumes lithium and electrolyte. This shows up as a low initial Coulombic efficiency (ICE), the share of lithium recovered in the first cycle, and as a Coulombic efficiency (CE) below the level a full cell needs. Commercial graphite reaches an ICE of about 90 to 94% (Nepal and Dhakal, 2026).
4. Cycle life falls. Nepal and Dhakal (2026) note that 1000 full-cell cycles typically requires a CE of about 99.98%; 500 and 250 cycles demand above 99.95% and 99.9% respectively, and values below 99.8% are generally impractical.

Why the quoted numbers differ

A frequent source of confusion is that the same headline quantities are reported with different values in different papers. The table below sets out the differences found in the three reviews and what explains them. It is a reading aid, not a ranking of which paper is right.

Quantity	Values reported	Why they differ	Source
Silicon capacity	3579 mAh g⁻¹; 4200 mAh g⁻¹	3579 corresponds to Li₁₅Si₄; 4200 is given for the fully lithiated phase	Nepal and Dhakal 2026; Zhang et al. 2025
Graphite capacity	360 mAh g⁻¹; 372 mAh g⁻¹	360 is the reference value used in the 2026 metrics review; 372 is the limited specific capacity cited by Pavlovskii et al.	Nepal and Dhakal 2026; Pavlovskii et al. 2026
Silicon expansion	Up to about 300%; more than 300%; 300 to 400%; about 400% (bulk); 420%	Form of silicon, end-point alloy phase and measurement context vary between papers	All three reviews
Critical particle size	Below about 150 nm (crystalline); about 870 nm (amorphous)	Different silicon structure: crystalline versus amorphous	Zhang et al. 2025; Nepal and Dhakal 2026

A practical consequence: when comparing two papers, check whether both state the same lithiated phase and the same graphite reference value before treating a “ninefold” or “tenfold” capacity advantage as a like-for-like result.

Types of silicon anode material

The reviews group silicon anode designs by how they accommodate expansion. The data table later on this page compiles the quantitative figures; the descriptions here stay close to what each source states.

Nano-silicon

Shrinking particles below about 150 nm avoids cracking in crystalline silicon (Zhang et al., 2025; Pavlovskii et al., 2026). The trade-off is surface area: Nepal and Dhakal (2026) cite roughly 100 m² g⁻¹ for silicon nanoparticles, and Pavlovskii et al. (2026), citing Keller et al., report that ICE scaled linearly with specific surface area, while impurities also reduce ICE. Smaller particles therefore protect the structure but can cost first-cycle efficiency.

Silicon–carbon composites

Combining silicon with carbon improves structure, cycling and Coulombic efficiency (Pavlovskii et al., 2026). A yolk-shell design (Si@Void@C) leaves an internal gap so the silicon can expand without breaking the outer shell (Nepal and Dhakal, 2026). The same review lists a Si@SiOx@C dual-shell example at 91% ICE and a Si@C design at 74.9%, showing how much design details change first-cycle losses.

Silicon oxides (SiOx)

A porous SiOx matrix can buffer expansion through the Li₂O and lithium silicates that form on first lithiation (Pavlovskii et al., 2026). Nepal and Dhakal (2026) report that a SiO₂ shell itself expands approximately 100% on lithiation, so oxide is a buffer, not an expansion-free material.

Graphite–silicon blends

Adding silicon to graphite raises capacity while keeping most of graphite’s mechanical stability. This is the approach described as important in current industry, with the stated limit that blends dominated by graphite struggle to approach 500 Wh kg⁻¹ at cell level (Nepal and Dhakal, 2026).

Mitigation strategies and the metric each one targets

Because the failure chain is sequential, each mitigation can be read against the measurement it is meant to move. This cross-walk is the quickest way to check whether a claimed improvement addresses the right problem.

Strategy	What it addresses	Metric to look for	Source
Particles below about 150 nm	Particle cracking and pulverization	Capacity retention; ICE (surface area raises SEI loss)	Zhang et al. 2025; Pavlovskii et al. 2026
Yolk-shell Si@Void@C	Gives expanding silicon space inside a shell	Cycle life; ICE	Nepal and Dhakal 2026
Si/C composites	Structure, conductivity, stability	Cycling; Coulombic efficiency	Pavlovskii et al. 2026
SiOx matrix	Buffers expansion with Li₂O and lithium silicates	Cycle life; first-cycle loss	Pavlovskii et al. 2026
Binders: CMC, PAA, polyimide	Aqueous binders suited to silicon	Retention of electrical contact	Zhang et al. 2025
FEC electrolyte additive	Improves SEI stability; extended cycle life from tens to hundreds of cycles, but does not fully prevent SEI instability	Long-term CE; cycle number	Nepal and Dhakal 2026
Prelithiation	Offsets first-cycle lithium loss	ICE; full-cell energy density	Nepal and Dhakal 2026

Nepal and Dhakal (2026) observe that many of the most successful full-cell demonstrations rely on prelithiation, which also means a half-cell result without prelithiation may not predict full-cell behaviour. They call for scalable and cost-effective prelithiation strategies. On binders, the same review notes that binder properties influence stress distribution and retention of electrical connectivity, but treats binder design as outside its scope, so the binder row above rests on Zhang et al.

Where it matters for labs: a reporting checklist

For lab staff, research offices reviewing data-management plans and anyone assessing a battery paper, the useful question is whether a result reports enough to be interpreted. Nepal and Dhakal (2026) define the metrics that matter for silicon anodes; the checklist below follows their definitions.

Report this	Definition used in the source	Why a reader needs it
Practical gravimetric capacity	Capacity per gram of electrode material, in mAh g⁻¹; the review states practical capacities of at least 1400 mAh g⁻¹ are generally needed	Separates achieved capacity from the theoretical 3579
Mass loading	Active material per electrode area, in mg cm⁻²; to be considered alongside gravimetric capacity	Low loadings flatter results
Areal capacity	Mass loading × gravimetric capacity ÷ 1000, in mAh cm⁻²; cathode areal capacities are typically about 3 to 4 mAh cm⁻²	Shows whether the anode can match a real cathode
Initial Coulombic efficiency	First-cycle lithium loss to SEI and other irreversible reactions	Crucial for full-cell viability; graphite is about 90 to 94%
Long-term Coulombic efficiency	Ratio of discharge to charge capacity per cycle	About 99.98% is typically needed for 1000 full-cell cycles
Tap density	Mass per packed volume after tapping; graphite about 1.1 to 1.2 g cm⁻³	Determines volumetric capacity
Volumetric capacity	Tap density × gravimetric capacity, in mAh cm⁻³	High gravimetric capacity at low density can still lose at cell level

The same review argues that no single metric is sufficient and calls for one architecture that delivers an ICE of at least 90%, long-term CE of about 99.90% or better and at least 1000 mAh g⁻¹ at high tap density. When documenting such experiments, recording electrode loading, cell format, electrolyte composition (including additives such as FEC) and any prelithiation step alongside the raw cycling data supports reuse, in line with the FAIR data principles and good rigor and reproducibility practice.

Structural characterisation of silicon phases and amorphisation is often done with diffraction; see the CASRAI XRD diffraction guide for how diffractometers are set up.

Cell-level reality: energy density and lifetime

Nepal and Dhakal (2026) put conventional graphite cells at a maximum of about 315 Wh kg⁻¹, with silicon-containing designs at about 400 Wh kg⁻¹ described as realistic today and a 500 Wh kg⁻¹ target associated with the Battery500 programme. For lifetime, the same review states that graphite cells can exceed 15 years of calendar life while silicon-containing cells typically degrade within 1 to 2 years, and that automotive use requires a 10-year calendar life and a 1000-cycle lifespan. The gap between those two lifetime statements is the practical reason cycle-life data and long-term CE matter more than a headline capacity.

Common misunderstandings

• “Silicon has ten times the capacity of graphite.” The ratio depends on the phase and graphite value used. Against 3579 mAh g⁻¹ and a graphite reference near 360 to 372 mAh g⁻¹, the reviews describe it as more than nine times; using 4200 raises the ratio.
• “Silicon swells 300%.” That is the low end of a range of 300 to about 420% reported across the three reviews. Always attach the source and silicon form.
• “Nano-silicon solves it.” Small particles resist cracking but have higher surface area, and ICE scaled with specific surface area in the work cited by Pavlovskii et al.
• “A good half-cell proves a good battery.” Full-cell results often depend on prelithiation, and cycle life depends on CE at the 99.9% level, which half-cell capacity plots rarely show.
• “An additive fixes the SEI.” FEC extended cycle life from tens to hundreds of cycles in the cited work but does not fully prevent SEI instability.

FAQ

What is a silicon anode?

It is a lithium-ion battery negative electrode that uses silicon, alone or combined with carbon or oxides, to store lithium by alloying. Its theoretical capacity of 3579 mAh g⁻¹ is more than nine times that of graphite.

Why do silicon anodes swell?

Alloying lithium into silicon enlarges the lattice by up to about 300% at Li₁₅Si₄, with reviews citing up to 400% or more for some forms. Graphite expands under 15%.

Why is the initial Coulombic efficiency low?

Lithium is consumed forming the SEI and in other irreversible reactions in the first cycle, and the SEI on silicon is unstable. Surface area and impurities lower ICE further.

What particle size avoids cracking?

Zhang et al. (2025) report that crystalline silicon particles below 150 nm neither crack nor pulverize; amorphous silicon has a critical size of about 870 nm.

What Coulombic efficiency is needed for long life?

About 99.98% for 1000 full-cell cycles, above 99.95% for 500 and 99.9% for 250 (Nepal and Dhakal, 2026).

Are silicon anodes used commercially?

The 2026 review describes graphite–silicon composites as an important current industrial approach, with silicon-containing cells at about 400 Wh kg⁻¹ regarded as realistic today.

Sources

• Nepal and Dhakal, “Silicon Anodes for Next-Generation Li-Ion Batteries: Design, Metrics, and Practical Routes to 500 Wh kg⁻¹ and 1000 Cycles,” Small Science 6(8):e70347, 2026, doi 10.1002/smsc.70347. PMC13516851. Read 2026-11-12.
• Zhang et al., “Silicon-based anodes for solid-state batteries: challenges, opportunities, and multiscale strategies,” RSC Advances 15(40):33561-33585, 2025, doi 10.1039/d5ra05126f. PMC12442552. Read 2026-11-12.
• Pavlovskii et al., “Nanostructured Silicon Anodes for Lithium-Ion Batteries: Advances, Challenges, and Future Prospects,” Materials 19(2):281, 2026, doi 10.3390/ma19020281. PMC12842676. Read 2026-11-12.