---
title: Loss of Active Material in Li-ion Batteries
url: https://www.emergentmind.com/topics/loss-of-active-material-lam
type: topic
---

# Loss of Active Material in Li-ion Batteries

Loss of Active Material (LAM) denotes the reduction, over time, of electrode material that remains available for intercalation and deintercalation. In lithium-ion batteries, it is commonly analyzed together with loss of lithium inventory (LLI, or LII), but it is not represented by a single universal observable. Depending on the study, LAM may be defined from accessible electrode capacity, from the fraction of time-invariant diffraction signal assigned to inactive phases, from the evolution of electrochemically active surface area, or from the growth of an inactive shell that shrinks an active core. Across recent work on Ni-rich NCM||graphite–SiOx cylindrical cells, large-format graphite electrodes, particle-population models, and reduced-order or porous-electrode models, LAM emerges as a multiscale degradation mode that is spatially heterogeneous and strongly coupled to SEI growth, lithium trapping, impedance rise, and stoichiometric slippage [2511.05892] [2503.06113] [2301.07892] [2102.03715] [2204.13364].

## 1. Definitions and quantitative representations

A common capacity-based definition writes LAM as the fractional loss of accessible active capacity in a given electrode,
\[
\mathrm{LAM} = 1 - \frac{Q_{\text{active, cycled}}}{Q_{\text{active, initial}}},
\]
while lithium inventory loss is written as
\[
\mathrm{LII} = 1 - \frac{Q_{\text{total, cycled}}}{Q_{\text{total, initial}}}.
\]
In this formulation, \(Q_{\text{active, initial}}\) and \(Q_{\text{active, cycled}}\) refer to the capacity accessible in a specific electrode’s active material, whereas \(Q_{\text{total, initial}}\) and \(Q_{\text{total, cycled}}\) refer to full-cell discharge capacities. Pham et al. use this distinction to separate active material loss from lithium loss in 21700-type Ni-rich NCM||graphite–SiOx cells, although the quantities are extracted indirectly through differential voltage analysis, neutron diffraction, and neutron depth profiling rather than by direct algebraic measurement [2511.05892].

A second formulation is operational and spatially resolved. In operando graphite \(\mu\)XRD, a region is treated as active if its local diffraction signature evolves during lithiation and delithiation, and inactive if the diffraction peaks in the \(1.7\text{–}1.9\,\text{\AA}^{-1}\) window remain unchanged over an entire cycle. The time-invariant component is isolated through
\[
I_{\text{inactive}}(q,y,z)=\min_t I(q,y,z,t),
\]
and a global inactive fraction can then be written as
\[
\mathrm{LAM}
=
\frac{\sum_{y,z}\sum_p A^{\text{inactive}}_p(y,z)}
{\sum_{y,z}\sum_p A^{\text{total}}_p(y,z)}.
\]
This definition makes LAM a fraction of diffracting graphite that no longer participates dynamically, rather than merely a decrease in electrode-average capacity [2503.06113].

Modeling studies introduce additional representations. In an enhanced single-particle formulation, the active specific surface area is written as
\[
a_{i,t}=a_i+a_{i,f}-a_{i,\mathrm{ina}},
\]
with fracture-generated area \(a_{i,f}\) and inactive area \(a_{i,\mathrm{ina}}\); a natural dimensionless indicator is then \(\lambda_i(t)=a_{i,t}(t)/a_i\). In a shrinking-core description of high-nickel positive-electrode degradation, particle-level cathode LAM is written as
\[
\mathrm{LAM}_{\mathrm{pe}}=1-\left(\frac{s}{R}\right)^3,
\]
where \(R\) is particle radius and \(s(t)\) is the radius of the remaining active core. In a population-based framework, LAM is not imposed as a single scalar at all; it emerges from an evolving distribution of particle “fitness” \(W\), with
\[
W=\frac{1}{r}\frac{i}{\bar{i}},
\]
so that electrode-level degradation reflects the changing participation of particles of different size and state [2102.03715] [2204.13364] [2301.07892].

## 2. Experimental observables and measurement strategies

LAM is usually inferred rather than observed directly. The recent literature combines electrochemical, structural, and morphological measurements to identify whether capacity loss reflects inactive material, cyclable-lithium loss, or both.

| Technique | Primary observable | LAM interpretation |
|---|---|---|
| DVA/ICA | Peak positions such as MaxHi, MinHi, and full-curve shifts | Peak movement to lower capacity indicates LAAM; additional curve shift relative to the peaks indicates LLI |
| Neutron diffraction / neutron depth profiling | NCM lattice parameters, LiC\(_{12}\)/LiC\(_6\) transitions, depth profiles \(c_{\mathrm{Li}}(x)\) | Reduced NCM lattice excursion and earlier graphite staging transitions indicate loss of active capacity; depth profiles reveal Li depletion or trapping |
| Operando \(\mu\)XRD | \(I(q,y,z,t)\) and the time-minimum inactive component | Resolves inactive graphite phases in 2D and tracks their C-rate dependence |
| CT / holotomography | Inter-layer distance, porosity, particle agglomeration | Constrains whether fade is dominated by macroscopic deformation or microscopic inactivation |

In the DVA framework used by Pham et al., LLI manifests as a horizontal shift of \(dV/dQ\) curves relative to specific phase-transition peaks, whereas LAM, specifically active anode material loss (LAAM), manifests as a shift of the phase-transition peaks themselves to lower capacity. Their high-SOC charge peak MaxHi is assigned to the LiC\(_{12}\)\(\rightarrow\)LiC\(_6\) transition, and the corresponding discharge minimum MinHi to LiC\(_6\)\(\rightarrow\)LiC\(_{12}\). After aligning curves to MaxHi or MinHi, the movement of these peak positions in absolute capacity reflects LAAM, and the additional shift of the full voltage curve relative to them reflects LLI [2511.05892].

The same study uses in operando neutron diffraction to quantify cathode-side LAM from the NCM \(a\)-parameter excursion,
\[
\Delta a=a_{\mathrm{SOC}=0\%}-a_{\mathrm{SOC}=100\%},
\]
with an effective metric
\[
\mathrm{LAM}_{\mathrm{NCM}}^{(a)}\approx 1-\frac{\Delta a_{\mathrm{SOH70}}}{\Delta a_{\mathrm{SOH100}}}.
\]
From SOH100 to SOH70, \(\Delta a\) decreases by 34%. Neutron depth profiling complements this by showing that, at discharge, total Li per unit area in NCM is reduced by 38% at the IC position and 29% at the IE position, while Li on the anode side increases strongly, indicating that lithium has migrated away from active cathode storage [2511.05892].

Operando graphite \(\mu\)XRD extends this logic from averaged electrode behavior to voxel-resolved inactivation. By fitting time-invariant diffraction intensity with pseudo-Voigt peaks and mapping inactive phase fractions over \(y\), \(z\), and time, the method distinguishes inactive graphite, LiC\(_{30}\), LiC\(_{12}\)-LiC\(_{18}\), and LiC\(_6\). This makes LAM a directly mappable spatial field rather than a lumped degradation number [2503.06113].

## 3. Electrode-specific mechanisms and spatial heterogeneity

Recent experiments show that LAM is rarely spatially uniform. In graphite-based negative electrodes, inactivation is strongly depth-dependent and frequently concentrated near the separator. In the large-format graphite study, inactive phases are concentrated near the separator side of the aged negative electrode; unlithiated graphite has its highest inactive fraction at the separator interface, whereas stage 2/2L often peaks in the mid-thickness. The same study distinguishes permanently disconnected “dead” graphite, which shows no evolution even at C/20 with long equilibration, from kinetically limited “slow” graphite, which appears inactive at C/5–C but can partially re-enter operation at C/20. Ex situ half-cell lithiation of aged graphite after C/20 plus a C/100 hold yielded 26.7% LiC\(_6\), 41.0% LiC\(_{12}\), and 32.3% unlithiated graphite, demonstrating that a substantial fraction remains inaccessible even under gentle conditions [2503.06113].

The spatial signature of this inactivation is not merely binary. In pristine graphite, lithiation proceeds with a moderate separator-to-collector gradient, and nearly all pixels at a given depth follow the same staging sequence. In aged graphite, regions with \(x\approx 0\), \(0.4\), \(0.6\), and \(0.9\) coexist in the same map, and the lithiation gradient reverses: near the separator, average \(x\) becomes lower because of heavily inactive and dead graphite, while more-current-carrying active graphite remains nearer the current collector. The local deviation
\[
\mathrm{Deviation}(y,z)=x(y,z)-\langle x\rangle
\]
is strongly anti-correlated with local inactive fraction at end-of-C/5 charge, with \(R=-0.95\) and \(p\approx 3\times 10^{-235}\). A heterogeneity factor,
\[
H(t)=\frac{1}{A}\iint_A \left|x(y,z,t)-\langle x(t)\rangle\right|\,dy\,dz,
\]
shows that the active portion of the aged electrode never returns to a fully homogeneous state, even at operating points where the pristine electrode does [2503.06113].

Morphological observations support a transport-mediated contribution to LAM. Nanoholotomography showed pristine graphite porosity of about 35% and aged porosity of about 25%, with aged particles appearing larger and more agglomerated. The interpretation given is repeated expansion and contraction together with accumulated electrolyte decomposition and SEI products, leading to particle coarsening, pore filling, increased tortuosity, and regions that become effectively isolated from electrolyte [2503.06113].

In graphite–SiOx blends, SiOx contributes an additional pathway. In the 21700 Ni-rich NCM||graphite–SiOx cells, a small DVA peak near \(Q_{\mathrm{SiOx}}\approx 0.7\,\mathrm{Ah}\) is assigned to SiOx lithiation; its charge-capacity fraction is \(q_{\mathrm{SiOx}}\approx 0.181\), corresponding to \(Q_{\mathrm{SiOx,spec}}\approx 505\,\mathrm{mAh\,g^{-1}}\) for the stated blend. This peak slowly moves to lower capacity and nearly vanishes with cycling, directly indicating loss of active SiOx, which accelerates overall anode LAAM [2511.05892].

On the cathode side, Ni-rich layered oxides exhibit a different mechanism. In the neutron-diffraction study, reduced \(a\)-parameter excursion indicates loss of accessible NCM Li swing, while the total unit-cell volume change \(\Delta V\) grows larger with aging because the \(c\)-parameter contracts strongly at high delithiation. The resulting larger anisotropic volume change for a smaller Li swing enhances mechanical stress and promotes NCM particle cracking. In the particle-level high-nickel model, this degradation is formalized as an irreversible layered \(\rightarrow\) disordered spinel \(\rightarrow\) rocksalt transition that produces a growing electrochemically inactive shell around an active core [2511.05892] [2204.13364].

## 4. Mechanistic and mathematical models

A central modeling question is whether LAM should be treated as a uniform scalar loss or as an emergent consequence of heterogeneous particle populations. The population-effects framework addresses this directly by describing an electrode as a distribution \(f(t,c;r)\) of particles over filling fraction \(c\) and radius \(r\). Its governing equation is a modified Fokker–Planck equation,
\[
\frac{\partial f}{\partial t}
=
-\frac{\partial}{\partial c}(fR)
+
\frac{\partial}{\partial c}\left(D\frac{\partial}{\partial c}(fW)\right),
\]
with fitness
\[
W=\frac{1}{r}\frac{i}{\bar{i}}.
\]
This representation makes two points explicit. First, smaller particles have higher baseline fitness because of the \(1/r\) factor. Second, degradation feeds back on current distribution: high-fitness particles carry more current, degrade faster, and then lose future fitness. In this model, electrode-level degradation is formed from various contributions of particle-level degradation, especially from smaller particles, and distinct particle-scale mechanisms leave characteristic signatures in capacity-loss and voltage profiles [2301.07892].

Reduced-order electrochemical models often encode LAM through active area rather than active volume. In the ESPM formulation, the fracture area evolves as
\[
\frac{da_{i,f}}{dt}=k_i,
\]
the inactive area evolves as
\[
\frac{da_{i,\mathrm{ina}}}{dt}
=
\beta_i\bigl(a_i+a_{i,f}-a_{i,\mathrm{ina}}\bigr),
\]
and the remaining active area is \(a_{i,t}=a_i+a_{i,f}-a_{i,\mathrm{ina}}\). LAM then enters the Butler–Volmer overpotential through
\[
\eta_i=\frac{RT}{\alpha F}\sinh^{-1}\!\left(\frac{I}{2A_{\mathrm{cell}}a_{i,t}L_i i_{0,i}}\right),
\]
and also enters the film resistance because
\[
R_{\mathrm{film}}=\frac{L_{\mathrm{SEI}}}{a_{n,t}A_{\mathrm{cell}}L_n\kappa_{\mathrm{SEI}}}.
\]
As \(a_{n,t}\) decreases, local current density per active area rises, overpotentials increase, and the relative contribution of lithium plating increases for a fixed applied current profile [2102.03715].

The positive-electrode shell model for high-nickel cathodes introduces a moving core–shell interface \(s(t)\) and defines
\[
\mathrm{LAM}_{\mathrm{pe}}=1-\left(\frac{s}{R}\right)^3.
\]
Lithium conservation at the moving interface and a phenomenological phase-boundary law connect the growth of a degraded shell to particle stoichiometry. A shell resistance
\[
R_{\mathrm{shell}}=\frac{\rho\,\delta}{A},
\qquad
\eta_{\mathrm{shell}}=\rho\,\delta\,j_{\mathrm{p,ave}},
\]
with \(\delta=R-s\), couples structural degradation to cell polarization. This model predicts that LAM reduces positive-electrode active volume, while the shell resistance adds an extra overpotential penalty during operation [2204.13364].

Across these models, LAM is therefore not a single mechanistic primitive. It can represent disconnection, pore blockage, kinetic freezing, structural phase transformation, surface reconstruction, or the cumulative effect of these processes on the subset of material that still exchanges lithium reversibly.

## 5. Coupling with lithium inventory loss and capacity fade

LAM seldom occurs in isolation. In the 21700 Ni-rich NCM||graphite–SiOx study, differential voltage analysis, neutron diffraction, and neutron depth profiling jointly indicate both active anode material loss and severe lithium inventory loss. On the cathode side, Li concentration in NCM at discharge falls by 29–38% depending on position, and the NCM \(a\)-parameter excursion decreases by 34%. On the anode side, total Li measured by NDP increases by 93% at IC and 116% at IE, with a sharper and broader surface peak that is attributed to thicker SEI and more Li consumed in SEI with cycling. The combined interpretation is that lithium leaves NCM irreversibly and is deposited or trapped on the anode side, mostly as SEI and possibly trapped Li in SiOx; capacity fade at 80–70% SOH is therefore attributed primarily to LLI accompanied by substantial active anode material loss rather than dominant cathode LAM or macroscopic structural collapse [2511.05892].

That balance is system-dependent. In the high-nickel positive-electrode degradation model, a special case with \(c_s=c_{\mathrm{p,b}}\) gives \(\mathrm{LLI}_{\mathrm{tot}}>0\) but \(\mathrm{LLI}_{\mathrm{cyc}}=0\), so capacity fade is entirely controlled by positive-electrode LAM. When \(c_s>c_{\mathrm{p,b}}\), cyclable-lithium loss shifts the stoichiometry range of the negative electrode but, in that model, does not directly dominate usable discharge capacity fade; positive-electrode LAM still dominates, while additional shell resistance accelerates the apparent loss of capacity at finite current [2204.13364].

The ESPM analysis of second-life batteries reaches a related conclusion from a different route. Increasing the anode inactive-area evolution coefficient \(\beta_n\) reduces discharge capacity, raises \(R_{\mathrm{film}}\), and promotes a positive feedback in which reduced active area increases local current density, favors plating, and thereby accelerates further loss of cyclable lithium and resistance growth. This framework was introduced specifically to capture both linear and non-linear capacity fade, with early-life behavior dominated by gradual SEI growth and later-life acceleration associated with plating and its coupling to LAM [2102.03715].

The population-effects model adds a diagnostic layer: because different degradation mechanisms affect fitness \(W(t,c,r)\) differently, they leave different shapes in \(V(Q)\) and in cycle-by-cycle capacity loss. This suggests that electrode-level voltage and capacity trajectories can, under appropriate modeling assumptions, be used to infer which microscopic LAM mechanisms are dominant [2301.07892].

## 6. Diagnostic ambiguities, phantom attribution, and methodological limits

A recurrent misconception is that low-rate voltage-curve decomposition yields an unambiguous measure of true LAM. Recent work shows that this is not generally correct. Degradation-mode analysis based on pseudo-OCV curves is biased by two non-degradation contributions: the SOC-dependent instantaneous ohmic drop \(IR_{\Omega}(\mathrm{SOC})\) and intrinsic charge–discharge hysteresis, especially in graphite–SiO\(_x\) negative electrodes. If these are not treated explicitly, the optimizer reallocates the resulting vertical offsets and branch differences into apparent electrode scale and shift parameters, generating “phantom LAM/LLI” [2512.19773].

The quantitative magnitude of this bias can be large. In the LG M50T case, IR correction lifted the low-rate discharge pOCV by \(+13\text{–}27\,\mathrm{mV}\) with ageing; without correction, PE-LAM was increasingly under-diagnosed, reaching \(-8.80\%\) relative error at late life, LLI had a median suppression of \(-3.07\%\), graphite LAM showed a median inflation of \(+17.68\%\), and Si-LAM a median suppression of \(-6.44\%\). In the lower-resistance P45B case, simply changing the voltage window from \(2.5\text{–}4.2\) to \(3.0\text{–}4.2\,\mathrm{V}\) under-reported Si-LAM by \(13.61\) percentage points, and on a harmonized \(3.0\text{–}4.2\,\mathrm{V}\) window the discharge branch recovered \(14.38\) percentage points more Si-LAM at end of life than the charge branch. The recommended protocol is to correct only the instantaneous ohmic term, harmonize the voltage window, and base quantitative attribution on the discharge branch; negative or anomalous component LAMs on charge are treated as allocation artefacts rather than recovery [2512.19773].

Endpoint-slippage analysis has an analogous problem. Between cycles \(k-1\) and \(k\),
\[
D_{\mathrm{slip},k}=Q_{c,k}-Q_{d,k},
\qquad
C_{\mathrm{slip},k}=Q_{c,k}-Q_{d,k-1},
\]
and LAM modalities such as \(\mathrm{LAM}_{\mathrm{liNE}}\), \(\mathrm{LAM}_{\mathrm{deNE}}\), \(\mathrm{LAM}_{\mathrm{liPE}}\), and \(\mathrm{LAM}_{\mathrm{dePE}}\) each produce characteristic combinations of charge-endpoint slippage, discharge-endpoint slippage, and capacity loss. The framework is additive, so in principle one can subtract the LAM- and impedance-induced contributions from measured slippages to isolate parasitic reduction and oxidation. In practice, this requires knowledge of the average Li content of disconnected domains at the moment of disconnection, and the study emphasizes that this quantity is often “unknown and perhaps even unknowable” [2508.02878].

The methodological implication is not that LAM is inaccessible, but that its attribution is conditional on the diagnostic framework. Diffraction-based inactive-fraction mapping, active-area models, voltage-curve fitting, endpoint slippage, and depth profiling do not all measure the same object. They emphasize different operational definitions of what it means for active material to be lost: structural inactivity, kinetic non-participation, electrochemical isolation, or geometric loss of accessible active phase. Taken together, these studies indicate that rigorous LAM analysis requires both a definition matched to the measurement and explicit treatment of coupled modes such as LLI, hysteresis, and impedance rise.

Source: https://www.emergentmind.com/topics/loss-of-active-material-lam