Operating stability, degradation and material-to-system evidence

A useful cell changes during operation

Interphase growth, corrosion, dissolution, mechanical damage or electrolyte decomposition can alter response. A short stable measurement does not establish cycling or storage life. Diagnose changes at the relevant chemical and mechanical interfaces.

Consider an electrode that still contains a charge-storing solid but delivers less charge on a later discharge. Some mobile inventory may have become trapped in unwanted products. Some active solid may have lost contact, or an increased transport or interfacial resistance may make the voltage limit arrive sooner. These explanations can coexist. A capacity measurement reports the accessible response under its procedure, not a direct count of the surviving active atoms.

An interface is a boundary between phases; an interphase is a material region between them. At a sufficiently reducing negative electrode in a carbonate-electrolyte lithium-ion cell, electrolyte breakdown can form a solid-electrolyte interphase, abbreviated SEI. A useful SEI restricts electronic transfer to the liquid while permitting lithium-ion transport. That selectivity can slow further breakdown without preventing the intended electrode reaction. A formed film is therefore not automatically a failure.

Continued interphase formation can consume cyclable lithium and electrolyte and increase resistance. Expansion and contraction of an active phase can disrupt a surface film and expose fresh material; renewed reaction then makes more interphase. This is a chemical–mechanical feedback, not merely a thick coating added to an otherwise unchanged electrode. The first figure isolates this possible pathway and a separate loss-of-contact event; it is not a universal sequence or a geometrical model of SEI growth.

Electrode and electrolyte materials provides the connected cell roles and the meaning of accessible charge. Ionic, electronic and mixed transport provides the carrier distinction. Here the question is how those functions change with operating history, rather than how much charge a pristine powder could store.

A loss curve does not identify which material changed

A current collector can also change. Corrosion can oxidise metal into dissolved species and undermine its conducting section or adhesion to the electrode. For example, comparisons of fresh and reclaimed aluminium collectors in particular lithium-ion electrolytes found passivation breakdown, local corrosion and altered electrode response. This does not make aluminium universally unsuitable: the applied potential, electrolyte and prior surface state belong to that observation.

Dissolution can couple the two electrodes through their shared electrolyte. Dissolved transition-metal species can migrate and alter reactions elsewhere. In one investigated carbonate/salt system, manganese(II) promoted electrolyte decomposition both in the liquid and near the negative electrode. The harmful species need not first become metallic manganese. Chemical identity and solvation, meaning the surrounding solvent/ion environment, matter as well as the amount transported.

Mechanical damage can remove an electronic contact even while liquid still reaches a particle. Conversely, a connected particle can retain electronic contact while some lithium becomes inaccessible. In a studied all-solid-state aluminium-anode cell, microscopy and diffraction identified aluminium–lithium alloy mounds; capacity loss accompanied lithium trapping even though the mounds retained electrical contact with underlying aluminium. Oxidation-related obstruction of diffusion was a proposed explanation. This is distinct from simply assuming that every visible mound is disconnected material.

Separate three diagnostic questions: has the amount available for the intended transfer changed; can the relevant material still exchange ions and electrons; and can that exchange occur quickly enough before the chosen endpoint? A terminal charge curve alone generally cannot answer all three. Pair it with relevant chemical, structural and transport evidence, keeping specimen preparation and state matched. An image of cracks alone neither counts lost capacity nor proves the dominant cause.

Time at a state and repeated cycling are different histories

Cycle ageing concerns changes during a declared repeated charge/discharge procedure. Calendar ageing concerns changes over elapsed storage or rest time at stated conditions. A stored cell can continue reacting internally while its external current is zero. Temperature, state of charge, potential history and elapsed time therefore remain relevant when no full cycles are counted. State of charge is a stated measure of progress through an accessible charge range; its reference convention must be reported.

A reference performance test interrupts storage to measure response under a fixed procedure. It is useful, but it is also an exposure: charging and discharging can change composition, dimensions and interphases. In silicon-containing-cell calendar experiments, changing the interval between such tests helped separate time-associated fade from additional disruption during the tests. Do not treat diagnostic cycles as if they were chemically and mechanically invisible.

For a comparison, retain cell and specimen identity, initial conditioning, elapsed hours or days, temperature history, state during rests, charge and discharge currents, rests between steps, voltage endpoints and the event counted as a cycle. Record the charge passed as well as cycle number. Ten shallow cycles and ten deep cycles need not pass the same charge or visit the same states. Nor can two cells with different rest durations be assumed to have the same calendar exposure merely because their cycle counts match.

Every discharge in the four-cycle interval below is listed; charging inputs and elapsed calendar times are not supplied. The records are invented to teach the accounting, with no measurement uncertainty supplied, not measurements of the named research materials or a validated lifetime model. Matching a procedure permits the stated output comparisons but need not match every relevant exposure. Four cycle results cannot establish a degradation law, population reliability or safe operating limits.

Use matched performance boundaries

State loading, rate, temperature, voltage limits, cycle definition and mass basis. Retention as a percentage can hide a low starting capacity; high efficiency can coexist with gradual loss. Compare full trajectories and the intended service requirement.

Charge capacity Q_out is the magnitude of charge delivered during the defined discharge, in ampere-hours here. One Ah is 3,600 C. Define retention after cycle n as R_n = Q_out,n/Q_out,1 for the same evaluation procedure and the stated first-cycle reference. It compares a cell with its own starting response. It is not the fraction of every original component remaining intact.

Give the original A/B question actual starting values: A delivers 0.400 Ah in cycle 1 and 0.380 Ah in cycle 4; B delivers 1.000 Ah initially and 0.900 Ah in cycle 4. Thus A retains 95% and B 90%, but B still delivers more charge. Both are compared as complete individual cells, with no cell masses supplied. These results do not rank intrinsic material capacity per gram or complete-cell energy per kilogram.

The table supplies every discharge in four successive charge/discharge cycles at the same controlled temperature of 25 °C, charge/discharge current magnitude 0.200 A, terminal endpoints 3.00 and 4.00 V, and the same declared conditioning and rest procedure. At each upper endpoint the charge step ends without a separate voltage hold; discharge ends at the lower endpoint. These are fictional evaluation conditions, not a recommended protocol. The supplied charge-weighted mean loaded discharge voltage V̄_out equals delivered energy divided by delivered charge. A nominal or open-circuit voltage cannot replace it.

A common absolute current does not establish an equal rate relative to initial capacity: I/Q_out,1 is 0.500 h⁻¹ for A and 0.200 h⁻¹ for B. Their discharge durations also differ. The table therefore compares outputs of two individual cells under this common procedure; it does not isolate an intrinsic resistance to degradation or establish matched calendar exposure. That question would need appropriately matched states, rates and time histories, with relevant replication.

Use W_out,n = V̄_out,n Q_out,n, giving Wh when V is multiplied by Ah. This is the finite electrical account W_out = ∫V dq, the sum of voltage times small delivered-charge increments. It does not require the voltage to stay constant during discharge. Mean voltage is explicitly supplied; the lesson does not reconstruct a voltage trace from one average. The second figure compares absolute charge and own-reference retention on separately labelled scales.

A cycle charge ratio is not an energy or lifetime ratio

For one declared charge/discharge pair, define coulombic efficiency here as CE = Q_out/Q_in using positive charge magnitudes at the same cell terminals. Q_in is the charge supplied during the stated charging step. Report endpoints, starting state and any rests: charge left in a different ending state, redistribution, side reactions and measurement error can all affect interpretation of the balance. The ratio does not by itself identify which chemical reaction consumed charge.

In a separate fictional pair, Q_in = 1.000 Ah and Q_out = 0.998 Ah. CE is 99.8%. The imbalance is 0.002 Ah = 7.20 C. This is an observed terminal-charge difference for that pair, not automatically a measured amount of one degradation product or a fixed irreversible loss repeated in all later cycles.

Energy efficiency uses energies at the same boundary: η_E = W_out/W_in. With supplied charge-weighted mean charging voltage 3.80 V and discharge voltage 3.50 V, input energy is 3.80 Wh and output is 3.493 Wh. Thus η_E ≈ 91.92%, despite CE = 99.8%. Equivalently, η_E = (Q_out/Q_in)(V̄_out/V̄_in) for these defined means. The lower discharge voltage changes the energy account without changing the charge units.

High CE can coexist with gradual loss of accessible charge, and it does not certify unchanged contact or resistance. Nor is multiplying initial capacity by CE once per cycle a general fade model: capacity-limiting inventory, electrode balance, side reactions and endpoints may evolve. The four-cycle A/B record supplies Q_out and voltage, but no Q_in or charging energy, so its CE and round-trip energy efficiency cannot be calculated from that table.

Use the finite trajectory and keep the system boundary

Add the four recorded discharge energies rather than multiplying the last energy by four. A delivered 5.3275 Wh and B 13.0575 Wh over those four discharges. This counts delivered electrical energy, not energy generated without recharging; each charge step requires a separate input. No energy before cycle 1 or after cycle 4 is included.

The last retention does not determine this total. For an unfamiliar comparison, replace only B’s middle discharge capacities by 0.950 and 0.920 Ah, retaining its first and last capacities and all four mean voltages. This B′ record has the same final 90% retention as B, but totals only 12.9205 Wh. Its intermediate history has changed even though both endpoints match. Each record contains all four discharges; a few diagnostic points separated by unreported operating cycles would not suffice for the same total.

At the constant 0.200 A discharge current, cycle 4 lasts Q_out/I: 1.90 h for A and 4.50 h for B. Their charge-weighted mean voltage is also their time-mean voltage here because current is constant. Both mean powers are therefore 0.200 × 3.35 = 0.670 W, while their delivered energies and durations differ. A requirement for at least 0.800 Ah per evaluated discharge selects B in all four rows, not A; a different mass, power or volume requirement needs its own evidence. Passing these four rows is not a life guarantee.

The cell-terminal boundary excludes converters, control electronics, pumps or thermal-management hardware. If a wider system consumes energy in those functions, count it at that boundary rather than silently retaining the cell-only result. A different electrode balance, electrolyte amount or compression can also change degradation, so a material improvement needs evidence in the relevant assembly. A cell history does not establish a complete pack’s delivered service or environmental consequences.

Change a condition to test an explanation, not to erase the history

Consider a separate, matched diagnostic comparison at the same temperature, starting state and terminal window. Fresh matched cells supply 1.000 Ah at both 0.200 A and 0.0200 A. Equivalently aged matched cells supply 0.800 Ah at 0.200 A and 0.980 Ah at 0.0200 A. Use separate matched cells for the two rates so that the diagnostic ordering itself is not an unrecorded difference. Their equivalence is stipulated for this teaching case; a real comparison needs replication and state checks.

The slower-rate observation recovers 0.180 Ah of the 0.200 Ah high-rate shortfall: 90% of that apparent shortage. It is evidence that much of the high-rate deficit is conditional on the measurement rate. It does not show that 90% of damaged material has healed. At lower current, smaller polarisation and more time for redistribution can allow more charge before an endpoint. This is consistent with an access or rate limitation but does not identify a unique resistor, diffusion process or interphase composition.

A useful follow-up combines a matched voltage/current history with independent chemical inventory or surface-composition evidence and structural/contact evidence. Ask whether an apparent film grew, material dissolved or became trapped, electrolyte changed, or contacts detached; distinguish observation from its proposed cause. Post-test disassembly can itself alter air-sensitive surfaces. Keep the same specimen identity and preparation record rather than treating any post-test image as the undisturbed operating state.

If a later test uses a narrower voltage window and reports 98% retention from its own first cycle, it has changed the accessible reaction range and the denominator. Re-evaluate the absolute output and relevant mechanisms under that window; do not splice its percentage onto the earlier trajectory. Operating stability is the persistence of a defined function in a defined history, supported by evidence at the material, interface, cell and intended system levels.

Worked examples

Try the reasoning

1. Does one operating cycle establish storage life?

   Show the answer and reasoning

   No.

   Calendar ageing tests a different exposure.

2. Why record voltage limits?

   Show the answer and reasoning

   They affect accessed reactions and degradation.

   Different windows can produce different apparent performance.

3. A new pair C and D starts at 0.600 and 0.900 Ah and finishes a matched evaluation at 96% and 82% retention respectively. Find the final charges. Which meets a stated minimum 0.700 Ah?

   Show the answer and reasoning

   C finishes at 0.576 Ah and D at 0.738 Ah. Only D meets 0.700 Ah in that evaluation.

   Multiply each own-reference retention by its own starting charge. The larger percentage does not reverse the starting-size difference. No material-per-mass ranking follows without the relevant masses and states.

4. B′ has discharge capacities 1.000, 0.950, 0.920 and 0.900 Ah at the table’s four mean voltages. Find its four-discharge energy and compare it with B. Can their identical first and last capacities settle the comparison?

   Show the answer and reasoning

   B′ delivers 12.9205 Wh, 0.1370 Wh less than B’s 13.0575 Wh. Identical capacity endpoints do not determine the intervening energy.

   The middle energies are 0.950 × 3.45 = 3.2775 Wh and 0.920 × 3.40 = 3.1280 Wh. Add 3.5000 and 3.0150 Wh for the endpoint discharges. This is a finite sum over all four cycles, not an interpolation through missing operating cycles.

5. In a separate pair, Q_in = 0.500 Ah at mean 4.00 V and Q_out = 0.495 Ah at mean 3.60 V. Calculate CE and terminal energy efficiency. Does 99% CE mean only 1% of input energy is unreturned?

   Show the answer and reasoning

   CE = 99.0%; W_in = 2.000 Wh and W_out = 1.782 Wh, so energy efficiency is 89.1%. The unreturned terminal-energy difference is 10.9%, not 1%.

   Charge and energy have different weights: 0.495/0.500 = 0.990, while multiplying by the voltage ratio 3.60/4.00 = 0.900 gives 0.891. These defined ratios do not identify the energy difference solely as heat or a particular side reaction.

6. For a separate stipulated four-cycle B system account, a converter delivers 90.0% of each cell discharge energy to the load. Controls draw a further 0.200 Wh over the four cycles at that same load-side boundary. Calculate net useful delivered energy. What evidence has been assumed?

   Show the answer and reasoning

   0.900 × 13.0575 − 0.200 = 11.55175 Wh, about 11.552 Wh.

   The controls are subtracted once, after conversion, because their supplied energy is at the load-side boundary. This assumes the stated conversion fraction and control demand over the complete record, unchanged cell discharge output and no other contributions. It does not establish the charging input, a round-trip system efficiency or a buildable pack.

7. The final old-window B discharge is 0.900 Ah. A new-window record starts at 0.700 Ah and later retains 98%. Does its larger retention demonstrate larger remaining charge than the old-window result?

   Show the answer and reasoning

   No. The new-window remaining charge is 0.686 Ah, less than 0.900 Ah, and the windows access different states.

   Use 0.700 × 0.98. The old and new percentages have different reference charges and procedures. Neither number by itself establishes which window gives better long-term service, energy, safety or material stability.

8. X and Y have completed the same 20 cycles, but X rested for 2 days and Y for 60 days at a higher temperature. Y now delivers less charge. Has cycling resistance alone been compared?

   Show the answer and reasoning

   No. Elapsed time, temperature and rest history are confounded with the intended comparison.

   Retain those differences and make an appropriate matched follow-up; do not relabel the entire change as a per-cycle loss. Diagnostic cycles and the stored state must also be recorded. More points on the same confounded trajectory do not remove the confounding.

9. A separate C/D comparison uses 0.300 A for initial capacities 0.600 and 0.900 Ah. Compare I/Q_initial. What current would give D the same initial normalised rate as C, and would its old capacity automatically remain valid at that changed current?

   Show the answer and reasoning

   The rates are 0.500 h⁻¹ for C and about 0.333 h⁻¹ for D. D would need 0.450 A to match C’s initial ratio. Its output at that new current needs a new observation.

   Divide A by Ah to get h⁻¹, then multiply 0.500 h⁻¹ by D’s 0.900 Ah. Matching this initial ratio alone still does not match temperature, window, ageing time, loading or the evolving rate relative to later capacity. It does not predict performance at an untested current.

10. An aged electrode still has an electronic path to its collector, yet chemical analysis finds trapped cyclable species and response falls. Does maintained contact rule out degradation? Would one crack image establish the lost-capacity fraction?

    Show the answer and reasoning

    No to both. Maintained electronic contact does not establish reversible inventory or fast ionic access, and an image alone does not quantify the electrochemical loss.

    Distinguish inventory, connectivity and rate. Use the matching charge/voltage record, chemistry and spatial evidence to test the specific proposed pathway. Multiple causes can operate together.

11. At cycle 4, suppose the same supplied 0.200 A and 3.35 V mean apply to A and B. Find the discharge durations and mean powers. Does B’s larger charge imply larger mean power in this record?

    Show the answer and reasoning

    A lasts 1.90 h and B 4.50 h. Both mean powers are 0.670 W; B supplies that mean for longer.

    Duration is Ah/A. Because current is constant, mean power is I times the supplied charge-weighted mean voltage. B’s larger delivered energy arises from its longer discharge here. A different load cannot be predicted by preserving the old capacity and voltage without evidence.

What to carry forward

• Track degradation and performance together.
• Keep test windows and denominators comparable.
• A passivating interphase can be useful, while its continued reaction, damage or transport change can reduce accessible function.
• Inventory loss, lost contact and rate-dependent access are different explanations that can coexist.
• Retention uses a stated reference response; charge efficiency uses input/output charge, and energy efficiency also weights voltage.
• Keep elapsed time, stored state, diagnostic perturbations and the complete operating record, not cycle count alone.
• A finite sum of measured or supplied discharge energies is not a lifetime extrapolation or a system round-trip efficiency.
• Changed-rate or changed-window observations test conditional explanations; they do not by themselves prove repair, one mechanism or safe service.