PEM Electrolyser and Fuel-Cell Testing Fundamentals: Control, Sensing, Polarisation and EIS

Four states that must not be confused

Open-circuit observation draws no intentionally imposed external current, although small measurement-input or leakage currents can exist. The observed voltage is not necessarily zero.

A 0 V full-cell command is an active voltage-control request. On a generating cell, a capable load would attempt a short-circuit-like state, subject to its limits. It is not an instruction to disconnect.

A zero offset from recorded OCV, where the software offers that reference, means the recorded baseline voltage—not zero terminal voltage. If the cell changes afterwards, holding that recorded value is not the same as observing its new open-circuit state. Check the actual software reference. The cited software documentation illustrates this terminology; its EFM experiment is not a PEM test method. [20]

Output off or software stop has the behaviour defined by the actual instrument and station. A command of zero current, zero voltage and an output-disconnect state are not interchangeable. Confirm the documented state separately from the procedure for gas isolation, cooling, pressure management and safe disconnection. [19]

PEM water electrolysis: electricity to hydrogen

Water is split into hydrogen and oxygen. Protons move through the membrane from the oxygen-evolving side towards the hydrogen-evolving side, while the external circuit supplies the electrical energy. The half-reactions, written on a two-electron basis, are: [3, 5]

Anode, oxidation: H2O → ½O2 + 2H+ + 2e−

Cathode, reduction: 2H+ + 2e− → H2

Overall: H2O → H2 + ½O2

In normal electrolysis, the anode is positive and the cathode negative.

PEM hydrogen–air fuel cell: hydrogen to electricity

Hydrogen is supplied to the anode and an oxygen-containing stream to the cathode. Protons pass through the membrane; electrons pass through the external load. The reactions are: [4, 5]

Anode, oxidation: H2 → 2H+ + 2e−

Cathode, reduction: ½O2 + 2H+ + 2e− → H2O

Overall: H2 + ½O2 → H2O

During normal generation, the anode is negative and the cathode positive. Anode always denotes oxidation and cathode reduction; their electrical polarities are not universal.

The MEA provides the membrane/catalyst core. State whether your use of “MEA” also includes adjoining diffusion or transport layers. Flow fields, transport layers, current collectors, seals and assembly conditions contribute to the complete device being measured. A full-cell result is therefore not simply an intrinsic catalyst-activity measurement. [11]

Why 1.23 V and 1.48 V answer different questions

For the hydrogen/oxygen reaction at approximately 25 °C with liquid water at unit activity and gases on the stated standard-state basis, the reversible voltage is approximately 1.229 V. It relates to the reaction Gibbs energy. The corresponding thermoneutral voltage is approximately 1.48 V and relates to reaction enthalpy. Temperature, pressure and chemical activities change the relevant thermodynamic values. [5]

These are explanatory reference quantities, not conditioning voltages or protection limits. At useful current, irreversible losses generally raise the electrolyser’s required voltage above its reversible value and lower a generating fuel cell’s delivered voltage below its reversible value. Several losses can overlap; the difference cannot automatically be assigned to one electrode. [11]

Nor does a compatible electrical controller make the same physical cell reversible. The cell and supporting system must be designed and qualified for both environments before reverse operation is contemplated.

Two electrochemical electrodes can use four leads

A full cell has two electrochemical electrodes. A four-wire, or Kelvin, arrangement uses separate current-carrying and voltage-sensing leads. The voltage belongs to the chosen sense locations, rather than being inferred at the instrument end of the high-current cables. Internal membrane, contact and electronic contributions within that boundary remain part of the measurement. [11, 12]

Separate sense leads do not create a four-electrode electrochemical experiment, remove the cell’s internal resistance or recover electrical energy dissipated in external current leads. Mark the sense points on the configuration record so that comparisons use the same boundary.

For researchers selecting compatible single-cell hardware, ScienceGears MEA test cells provide a relevant starting point for matching the cell hardware, flow-field arrangement and measurement boundary to the experiment. For acidic PEM water electrolysis, do not infer anode-side materials compatibility from the fixture name: verify all wetted plates, conductive coatings and seals against the oxygen-evolving environment and the planned operating conditions. [21]

A reference electrode provides a different measurement

In a three-electrode experiment, working-electrode potential is measured against a reference while current passes between working and counter electrodes. Introducing a reference into an operating membrane cell needs a suitable geometry and interpretation. Full-cell voltage alone does not reveal both individual electrode potentials. [11]

Use the actual manufacturer’s terminal assignments and approved grounding arrangement. Check differential input range, input-to-ground/common-mode limits and channel isolation separately; a small measured cell voltage does not establish a safe input-to-ground voltage. Labels such as WE, CE, RE, sense, high and low are not universal wiring instructions. Retain protective earth and do not move leads on an energised reactive assembly. [19]

Record three different kinds of acceptance evidence

This separation is a planning framework. Fortiner and colleagues provide a specific electrical-short check within an ambient-pressure benchtop PEMWE procedure for cells no larger than 25 cm². That is a source to examine for an applicable setup, not a universal voltage or pass/fail threshold. [9]

Do not proceed on a suspected short or failed containment result. Follow the approved investigation procedure. An ordinary resistance reading across a hydrated or reactive cell is not a substitute for a validated integrity method.

Figure 2. Full-cell sensing and three separate checks. Conceptual measurement boundaries and three independent kinds of acceptance evidence. Instrument-chain verification, containment and actual-cell electrical integrity answer different questions. Use the applicable procedures; no generic ohmmeter, OCV-only, self-discharge or leak-test recipe is implied. The three-electrode sketch illustrates the reference principle, not a validated reference placement inside a PEM MEA.

A sequence that preserves the comparison

The exact order and conditioning programme remain source-dependent. Lickert and colleagues’ round-robin work shows why preparation, interruptions, thermal control and reconditioning must be documented when laboratories compare PEMWE data. Applying the same nominal equipment or waiting the same number of hours is not enough to establish the same state. [10]

An OCV record may be a useful observation where the method calls for it. It is not a complete performance or containment test. In particular, an electrolyser’s OCV depends on its state and gas history; do not replace the actual protocol with one universal expected value.

Before starting, save the method file and record the cell/MEA identity, active-area definition, current convention, sense points, ranges, environmental channels, sample timestamps, averaging window, fault response and raw-data location. Unresolved fields mean the plan still needs work; a blank limit is not unrestricted permission.

Define a reporting window, not just a dwell time

A useful electrical stability metric is a fitted slope over a specified trailing window. For voltage, express that slope in units such as mV/min. Its acceptance threshold and window must suit the method and the required comparison. The electrical result also needs acceptable environmental and instrument conditions: a small voltage slope while inlet temperature drifts is not automatically a matched measurement.

Once the required conditions are met, calculate the selected-window mean and the defined measure of variability. Record the number of readings and sample interval. Keep the preceding transition and any excluded windows in the raw dataset. A fixed dwell can be part of a validated protocol, but elapsed time alone does not demonstrate that another setup has reached the intended state. [9, 10]

Worked example: two accepted windows and one drifting window

Synthetic data only. The following original ScienceGears teaching model uses a 5 cm² cell, three arbitrary 60-second steps and one-second sample spacing. The current values are 2.5, 5.0 and 7.5 A. The assumed current readback equals the commanded current; that equality is an exercise assumption, not a property of real measurements.

Use the final ten readings of each step: times 50–59, 110–119 and 170–179 seconds. For this exercise only, accept an absolute fitted voltage slope no greater than 1 mV/min and assume all environmental and instrument checks pass. Neither the time intervals nor the threshold is a recommended PEM operating value.

Figure 3a. Raw staircase: commanded current. Synthetic command: 2.5, 5.0 and 7.5 A over three arbitrary 60-second steps. The underlying record contains 180 samples at 0–179 s. Readback equals command only by assumption. These values are not a test programme.

Figure 3b. Raw staircase: voltage and reporting windows. Synthetic voltage response for the teaching model. Candidate windows contain samples at 50–59, 110–119 and 170–179 s. The invented drift limit is 1 mV/min; environmental and instrument acceptance are assumed. The third window fails. Ten one-second-spaced samples span nine seconds.

Only the first two window means belong in the accepted polarisation dataset under that invented rule. The third trace and its rejection reason remain in the record. Extending a line through its final value would conceal that the planned comparison was not achieved.

The table displays extra digits to make the calculation traceable; they are not a claim of experimental resolution. The final ten samples span nine seconds between the first and last timestamps. Their sample standard deviations describe time-adjacent readings, not ten independent cells or the complete measurement uncertainty. Figure 3c shows only the accepted means; it does not fit a performance curve through two points. If an environmental check failed, a small slope would not rescue the point. [17]

Figure 3c. Accepted polarisation means. Only the first two means pass the teaching rule. The third trace remains in the raw record but is excluded from the accepted dataset. Error bars are within-window sample standard deviations and are small at this scale; they are not total uncertainty or independent-MEA reproducibility. No performance curve is fitted through the two points.

Compare like with like

Plot raw terminal voltage against measured geometric current density. Fuel-cell output power density can be added as pout = Vcellj. For electrolysis, that same positive-magnitude product represents electrical input density—not output or efficiency.

At matched current density and conditions, lower electrolyser voltage means less electrical input per unit charge; higher generating fuel-cell voltage means more electrical output per unit charge. Neither statement alone establishes product recovery, net system efficiency or which component caused the difference. [5, 11, 16]

For voltage-controlled steps, apply the same acceptance logic to the current response and actual operating state. A fast linear sweep is a transient experiment unless its equivalence to the intended steady-state measurement has been established. Report direction, dwell or stability rule, averaging and range changes. Do not extract one electrode’s kinetics from a complete-cell curve without justifying the other contributions.

CA: set voltage and observe current

Chronoamperometry records current with time following a change in applied potential. In the full-cell experiments here, the explicitly declared quantity is terminal voltage. A voltage-hold question is: How much current does the cell sustain at this voltage, and how does that response change with time? [1]

For the electrolyser, the source provides the current required at the approved voltage. For a generating fuel cell, a compatible load adjusts the current drawn to maintain an attainable terminal voltage. Record actual voltage, current, conditions and control-limit flags; the programmed voltage is not the entire record.

A current transient can reflect capacitive charging, reaction, transport and changing operating conditions. IUPAC describes conditions under which the Cottrell relationship applies; a porous, gas-fed membrane cell is not automatically that ideal diffusion experiment. Do not fit a convenient transient model without establishing why its assumptions are appropriate. [1, 11]

The retained synthetic CA model makes this distinction clear: after one arbitrary relative voltage step at t = 0, the voltage remains at its new level. One hypothetical current response approaches a final level from above and another from below. Figures 4a and 4b use those original functions. They are alternatives, not simultaneous predictions or measured PEM responses. A downward transient alone is not evidence of catalyst damage.

Figure 4a. CA: impose a voltage step. Original synthetic command in relative units: one step from 1.0 to 1.2 at relative time zero, then a hold. Relative time and voltage are not an operating duration or cell-voltage recommendation.

Figure 4b. CA: observe alternative current responses. Two hypothetical current functions follow the same voltage command. One approaches the final relative value from above, the other from below. They are alternative illustrations, not measured devices, mechanistic fits or proof of irreversible degradation.

CP: set current and observe voltage

Chronopotentiometry measures potential with time following a change in applied current. Here it asks what terminal voltage the complete cell requires or delivers at the prescribed current. It is not a constant-power experiment, even where both are abbreviated “CP” in different software contexts. [2]

The synthetic CP model uses one relative current-magnitude increase at t = 0 and opposite voltage changes for the two normal operating branches (Figures 4c and 4d). Each voltage change is relative to that device’s own pre-step baseline; it is not an absolute voltage. The current then remains at its new level. The model does not imply identical cell voltages, a universal transient shape or an approved load sequence.

Figure 4c. CP: impose a current-magnitude step. Original synthetic current magnitude rises from 1.0 to 1.5 at relative time zero and remains there. Each normal device branch uses a positive current magnitude; this is not a signed-current or gas-flow programme.

Figure 4d. CP: compare changes from each device’s baseline. Synthetic voltage changes after the same current-magnitude increase: positive for the electrolyser and negative for the generating fuel cell. Each is a change from its own pre-step baseline, not an absolute voltage. The curves are teaching functions, not fitted kinetics or prescribed device responses.

Specify the variable you are not controlling

CA needs permitted current boundaries. CP needs permitted voltage boundaries, including the relevant upper-voltage protection for an electrolyser and lower-voltage protection for a generating fuel cell. Add any other cell-specific limits and the required flow, pressure and thermal responses.

Record ramps, transition behaviour, sampling, duration and averaging. A change of control mode should begin from a compatible state under a validated transition, not an unrelated setpoint. For an apparent durability trend, state the fitted interval and how conditioning and interruptions were treated. A short-term slope is not a validated service-life prediction. A matched post-test baseline and further diagnostics are needed before stronger claims about irreversible change. [10, 13]

Choose the perturbation and preserve the DC context

EIS measures the frequency-dependent small-signal voltage/current relationship, including phase, around an operating state. Potentiostatic EIS (PEIS) perturbs voltage and measures current; galvanostatic EIS (GEIS) perturbs current and measures voltage. Either may be useful when the complete configuration supports the state and the required response. The JRC PEMWE method explicitly recognises current or voltage perturbation. [14]

Start with measured DC voltage and current plus stable, recorded temperatures, pressures and feeds. State the actual frequency list and order, range settings, integration/cycles and timestamps. “EIS at 5 A” is incomplete if the cell was still changing or the readback did not achieve the command.

With the current-into-cell convention used above, write the complex perturbations as δṼ and δĨin:

Z(f) = δṼ(f)/δĨin(f)

If current is instead defined as positive fuel-cell output, Iout = −Iin, the equivalent expression is Z = −δṼ/δĨout. Check the instrument’s definition before altering exported signs. Impedance is not the DC ratio V/I.

Make amplitude and time explicit

For a single sinusoidal voltage perturbation, one notation is:

V(t) = VDC + ΔVpeak sin(2πft)

State whether the entered amplitude is peak, peak-to-peak or root-mean-square (RMS). For the zero-mean sinusoidal AC component, peak-to-peak is twice peak, and RMS is peak divided by √2. That relationship does not describe the RMS of the complete waveform when a DC offset is included.

Illustrative headroom check. At one frequency, 10 mV peak across an assumed impedance magnitude of 5 mΩ corresponds to 2 A peak of AC current. This is arithmetic, not a recommended perturbation. For a single sinusoid, a necessary current-bound check is:

IDC − ΔIpeak ≥ Imin   and   IDC + ΔIpeak ≤ Imax

The limits must include the intended operating regime, not only a hardware maximum. Also check the resulting voltage, power and control behaviour. A multisine signal needs the extrema of the combined waveform. Passing a bound does not establish linearity, stationarity or phase accuracy.

Low frequencies also cost time. At 10 mHz, one period lasts 100 seconds; three periods would take 300 seconds before additional settling or other frequencies. Three is merely a timing illustration here. A longer spectrum is useful only if the operating state remains sufficiently consistent throughout acquisition.

A practical quality sequence

Verify the measurement chain with an appropriate approved arrangement. Choose a frequency/impedance region supported by the actual cell, instrument and cables. Compare suitable perturbation amplitudes and repeat selected measurements. Investigate meaningful changes with amplitude, clipping, overload, range switching, drift or poor repeatability before interpreting a fitted model. Keep the raw real and imaginary data.

External connections can create substantial high-frequency contributions in low-impedance cells. Franzetti and colleagues demonstrate why simply discarding inductive points can distort the inferred resistive response in their PEMWE setup. Their method is configuration-specific, not a general instruction to short a reactive cell or rewire it during operation. [12]

Read the axes before naming an arc

A common Nyquist display plots real impedance Z′ horizontally and −Z″ vertically. With a positive-up vertical axis, a capacitive response with Z″ < 0 appears above the axis; an inductive response with Z″ > 0 appears below it. An inductive feature is not automatically an artefact in every system. Bode magnitude and phase plots provide complementary frequency information. [12]

Synthetic circuit example. For a series resistance R0 followed by a parallel Rp–C combination:

Z(f) = R0 + Rp/(1 + i2πfRpC)

Using the original teaching values R0 = 20 mΩ, Rp = 80 mΩ and C = 0.2 F gives limiting real-axis intercepts of 20 and 100 mΩ. For this ideal circuit, fpeak = 1/(2πRpC) ≈ 9.947 Hz, at Z′ = 60 mΩ and −Z″ = 40 mΩ. Figure 5 uses equal impedance scales and calculates the frequency markers from the same equation. The intercepts are frequency limits, not finite sampled points. These parameters do not identify physical components in a measured MEA. For geometric-area normalisation, ZA = AZ, in Ω cm² when A is in cm².

Figure 5. Nyquist axes and the limits of a single-RC example. Synthetic R₀ + (Rₚ parallel C), with R₀ = 20 mΩ, Rₚ = 80 mΩ and C = 0.2 F. Exact markers are computed at 100 Hz, 9.947 Hz and 1 Hz. Open markers at 20 and 100 mΩ denote limiting intercepts, not finite-frequency observations. Both axes use equal impedance scales. These circuit parameters are not independently identified components of a real PEM cell.

HFR is not automatically membrane resistance. High-frequency resistance must be extracted from a sufficiently resolved resistive response, accounting for the measurement boundary and parasitics. It is not necessarily the real part at the single highest sampled frequency. State whether it came from an intercept, extrapolation or model. Contacts and other internal contributions may remain; do not label all of it “membrane resistance” without supporting evidence. [11, 12]

Separate consistency from physical interpretation

The Kramers–Kronig relationships connect the real and imaginary parts under the relevant linearity, causality, time-invariance and mathematical convergence conditions. Lin-KK checks and residuals can help assess consistency, alongside independent checks on the measurement chain, perturbation sensitivity, drift and repeatability. A pass does not select one unique equivalent circuit or prove a degradation mechanism. [15]

Fortin and colleagues provide a useful PEMFC example: changing humidity and cathode stoichiometry changed the measured spectra. Their model assignments belong to their controlled comparison. The lesson is to inspect the operating environment before interpreting a larger impedance feature as lasting catalyst damage; multi-sine acquisition is not required for this basic reasoning. [13]

The compact EIS record should therefore retain six groups: DC state; control and range; AC request; frequency/timing; validity checks; and interpretation assumptions. This is the bridge between a plotted spectrum and a reproducible measurement.

Area-specific resistance and ohmic correction

For resistance R in ohms and geometric area A in cm²:

RASR = RA   [Ω cm²]    and    ΔVohmic = ImagR = jRASR

Multiply resistance by area, not divide it. Use a resistance that matches the relevant operating state and sensing boundary. With positive current magnitudes, a simple post-processed correction has opposite directions:

Electrolyser: ViR-free = Vmeasured − ImagR

Fuel cell: ViR-free = Vmeasured + ImagR

Original teaching example. For 5 A, 20 mΩ and 5 cm², the assigned ohmic drop is 0.100 V and ASR is 0.100 Ω cm². The earlier 1.90 V electrolyser value becomes 1.80 V after this correction; the 0.65 V fuel-cell value becomes 0.75 V.

That correction removes the assigned ohmic term. It does not automatically remove transport losses or isolate one electrode’s kinetics. Preserve both datasets and explain the resistance extraction. Post-processing is also different from real-time feedback compensation; a corrected literature plot is not an instruction to enable aggressive online compensation. [11]

Actual electrical energy uses raw terminal data

For the stated sensing boundary, calculate electrolyser input as ∫VcellImagdt, and generating fuel-cell output as ∫VcellIoutdt. Use raw terminal voltage and aligned measured current—not the iR-corrected curve. Extending to system efficiency requires accounting for the relevant pumps, heaters, cooling, gas handling and power conversion. [5]

Charge is not measured hydrogen

For a defined interval of ordinary, unidirectional electrolysis, Q = ∫Imagdt. The water-splitting stoichiometry gives nH₂,theoretical = Q/(2F) and nO₂,theoretical = Q/(4F). At an assumed constant 5 A for one hour, Q = 18,000 C and theoretical hydrogen is approximately 0.093279 mol using rounded F = 96,485 C/mol. Do not integrate absolute current through a reversal or a different diagnostic and label it all hydrogen-producing charge.

For hydrogen attributed to the same interval and declared collection boundary, the collected-product Faradaic efficiency is FEH₂,collected = 2FnH₂,collected/Q; multiply by 100 for a percentage. Account for background product, calibration, sampling-time alignment and changes in gas inventory. Water vapour, leaks, holdup and crossover affect recovery. A collection-based result is not automatically identical to the selectivity of the electrode reaction. [16]

As a conditional electrical-only illustration, unity hydrogen yield and the stated standard-state higher-heating-value (HHV) basis give approximately 1.48/1.90 = 77.9% at 1.90 V, excluding auxiliary loads and heat input. This is not measured system efficiency. An electrical-only ratio can exceed 100% below the corresponding HHV-equivalent voltage because electricity is not then the only energy input. Account for heat and reactant/product states in the complete balance. The HHV-equivalent reference voltage is not a universal operating-temperature thermoneutral value. [5]

Repeat the comparison, then close the system

A before-and-after comparison should return to the same declared temperatures, feeds, pressures, sensing boundary and preparation context. Disclose any recovery or reconditioning rather than treating it as an invisible reset. A measurement after changing conditions cannot by itself isolate irreversible change. Benchmarking studies show why interruptions and history belong in the comparison. [10]

Then follow the approved coordinated shutdown. Electrical unloading, gas management, water circulation, cooling and depressurisation must be handled as the actual installation requires. Verify the final state before disconnection or disassembly and save the shutdown record. There is no universal “turn the potentiostat off first” instruction that this article can safely supply. [18]

Distinguish time-sample variability from experimental replication

Within-hold variability concerns readings within one averaging window. Repeat runs repeat a method on the same assembly. Reassembly comparisons include rebuilding the cell. Between-MEA comparisons include independently prepared specimens. State which conditions changed, which remained fixed and which question the repeats answer; these different designs should not all be labelled measurement repeatability. [17]

Thousands of adjacent readings from one hold are not thousands of independent MEAs. A small within-window standard deviation does not establish voltage accuracy, and a high-resolution display does not supply a complete uncertainty estimate. Instrument calibration, ranges, area definition, environmental measurements and data processing may all contribute. Metrology terminology distinguishes repeated-value agreement from uncertainty in the reported result. [17]

Choose the replication and measurement quality needed for the comparison—not an arbitrary universal sample count. A few millivolts of difference require evidence that the complete experiment can resolve that difference under the stated conditions.

Best-practice reporting checklist

This checklist adapts the existing ScienceGears planning framework. It is a reporting aid, not a compliance certificate. A useful final result states what was observed, under which conditions, what conclusion follows and which alternative explanations remain.

Can one workstation test both PEM devices?

Yes, when the complete configuration supports the required source and sink operating points, sensing and diagnostics. The same software technique name does not establish this. Confirm the actual configuration before constructing the method.

Is open circuit the same as setting 0 V?

No. Open circuit is an observation without intentionally imposed external current. A 0 V terminal command is active control; a zero stored-OCV offset is a different reference again. Output-off behaviour must also be checked separately.

Should I use current or voltage control first?

Choose from the scientific comparison. Current steps make prescribed current-density comparisons direct; voltage holds answer how much current is sustained at that voltage. Both require a permitted window, acceptance criteria and limits on the uncommanded variable. [1, 2]

Does full-cell measurement require a separate reference electrode?

Not for terminal voltage and current. Individual-electrode potentials require an appropriate electrochemical reference arrangement and interpretation. Separate Kelvin leads alone do not provide that information. [11]

Can I use one dwell time for every polarisation point?

Only as part of a method whose suitability has been established for the intended comparison. A fixed dwell does not guarantee the same physical state on every setup. Retain the required stability and environmental checks. [9, 10]

What should happen to a drifting point?

Keep its raw trace, flag why the point failed and follow the pre-agreed repeat, investigation or stop procedure. Do not silently substitute the last value or change the acceptance rule after seeing which result looks best.

Is falling CA current or changing CP voltage proof of degradation?

Not by itself. Check the actual conditions and measurement state, then compare matched baselines and targeted diagnostics. The trace establishes a response; attributing it to irreversible material change requires further evidence. [10, 13]

Is the highest-frequency point the membrane resistance?

Not automatically. Determine whether a resistive contribution is resolved, how parasitics were treated and what lies within the sense points. State the HFR extraction method rather than rename one frequency point. [11, 12]

Does a good circuit fit or Kramers–Kronig pass prove the mechanism?

No. Fit quality, consistency with the relevant assumptions and physical attribution are separate questions. Overlapping processes and non-unique models require controlled comparisons and other evidence. [11, 15]

Can I use iR-corrected voltage for efficiency or copy the example settings?

Use raw terminal voltage and current for actual energy at that boundary. The numbers in this guide are teaching assumptions, not executable settings or measured product yields. Efficiency also requires its stated material and energy accounting basis. [5, 16]

When does a PEM single-cell setup need a booster or electronic load?

Use additional source or sink hardware when the base instrument alone cannot support the required current, voltage or continuous power at the actual cell operating point. Verify the complete configured system, including protected-variable limits, sensing and EIS capability, rather than assuming that a DC hold guarantees the required dynamic measurement.

Why should PEM gas flow be reported with reference temperature, pressure and dry/wet basis?

Because a volumetric flow value is not fully defined without its reference conditions and calibration basis. A humidified total stream is also not automatically equivalent to dry reactant flow, so these details are necessary for meaningful comparison between tests.

What should be defined in the method sheet before the software sequence is started?

Record the scientific question, source method and local adaptation, permitted operating window, sensing boundary, actual supporting conditions, stability and averaging rules, uncommanded-variable limits, fault response, sampling and raw-data location before running the sequence.

When should a measurement be treated as a separate diagnostic rather than ordinary full-cell performance testing?

Treat it as a separate diagnostic when the question requires a different gas environment, reaction direction, reference-electrode arrangement, electrical reference, control envelope or independent analytical evidence. Keep its preparation, limits, data and return-to-baseline procedure distinct from the normal performance record.

Related ScienceGears reading

For the physical components, start with What is a membrane electrode assembly? Keep the definitions, assembly record and measurement boundary consistent as you move from component selection to full-cell testing.

• Electrochemical Cell Selection Guide : Beaker Cell, Sealed Cell, Jacketed Cell, H-Cell and Flow Cell

• Application notes

• Potentiostats / Galvanostats

• Membrane Electrode Assembly (MEA) Test Cells

• Electrolyser Test Stations

• Fuel Cell Test Stations

Scientific notes and further reading

The workflow, acceptance framework and numerical exercises are an educational synthesis adapted from the existing ScienceGears PEM guide and planning materials. The staircase, waveform and RC examples are synthetic; no ScienceGears cell was operated to produce them. Literature examples are identified as published research, not recreated as customer or in-house results.

The references below distinguish terminology, official scope records, research methods, diagnostic studies and metrology. Catalogue access is not a review of a complete standard. Consult the applicable full procedure and equipment documentation before selecting executable values. Detailed source-access and revision notes are retained in the ScienceGears editorial record.

[1] IUPAC. Chronoamperometry. Compendium of Chemical Terminology (Gold Book), 5th online edition (2025), based on the IUPAC Recommendations 2019, published in Pure and Applied Chemistry in 2020. DOI: 10.1351/goldbook.09129. Official definition

[2] IUPAC. Chronopotentiometry. Compendium of Chemical Terminology (Gold Book), 5th online edition (2025), based on the IUPAC Recommendations 2019, published in Pure and Applied Chemistry in 2020. DOI: 10.1351/goldbook.09126. Official definition

[3] US Department of Energy. Hydrogen Production: Electrolysis. Official technical overview. Official overview

[4] US Department of Energy. Fuel Cells. Official technical overview. Official overview

[5] EG&G Technical Services, Inc. (2004). Fuel Cell Handbook, Seventh Edition. US Department of Energy, National Energy Technology Laboratory, November 2004. Chapter 2: thermodynamics, energy balance and efficiency. Handbook

[6] Malkow, T.; Pilenga, A.; Tsotridis, G.; De Marco, G. (2018). EU harmonised polarisation curve test method for low-temperature water electrolysis. European Commission JRC104045; EUR 29182 EN. DOI: 10.2760/179509. Official record

[7] Tsotridis, G.; Pilenga, A.; De Marco, G.; Malkow, T. (2015). EU harmonised test protocols for PEMFC MEA testing in single cell configuration for automotive applications. European Commission JRC99115; EUR 27632 EN. Repository release 2016. DOI: 10.2790/54653. Official record

[8] IEC (2025). IEC TS 62282-7-1:2025, Fuel cell technologies — Part 7-1: Test methods — Single cell performance tests for polymer electrolyte fuel cells (PEFC). Edition 3.0. Official catalogue

[9] Fortiner, S.; Ouimet, R.; Young, J. L.; Bender, G.; Carmo, M.; Ayers, K. (2025). A harmonized protocol to assess the single-cell performance of proton exchange membrane water electrolyzers. Frontiers in Energy Research 13, 1549219. DOI: 10.3389/fenrg.2025.1549219. Open-access article

[10] Lickert, T.; Fischer, S.; Young, J. L.; et al. (2023). Advances in benchmarking and round robin testing for PEM water electrolysis: Reference protocol and hardware. Applied Energy 352, 121898. DOI: 10.1016/j.apenergy.2023.121898. Publication

[11] Gerhardt, M. R.; Pant, L. M.; Bui, J. C.; Crothers, A. R.; Ehlinger, V. M.; Fornaciari, J. C.; Liu, J.; Weber, A. Z. (2021). Method—Practices and Pitfalls in Voltage Breakdown Analysis of Electrochemical Energy-Conversion Systems. Journal of The Electrochemical Society 168, 074503. DOI: 10.1149/1945-7111/abf061. Publication · Author-repository text

[12] Franzetti, I.; Pushkarev, A.; Chan, A.-L.; Smolinka, T. (2023). Parasitic Effects in Impedance Spectrum of PEM Water Electrolysis Cells: Case Study of High-Frequency Inductive Effects. Energy Technology 11, 2300375. DOI: 10.1002/ente.202300375. Publication · Author-laboratory text

[13] Fortin, P.; Gerhardt, M. R.; Ulleberg, Ø.; Zenith, F.; Holm, T. (2022). Multi-Sine EIS for Early Detection of PEMFC Failure Modes. Frontiers in Energy Research 10, 855985. DOI: 10.3389/fenrg.2022.855985. Open-access article

[14] Malkow, T.; Pilenga, A.; Tsotridis, G. (2018). EU harmonised test procedure: electrochemical impedance spectroscopy for water electrolysis cells. European Commission JRC107053; EUR 29267 EN. Repository release 2019. DOI: 10.2760/8984. Official record

[15] Schönleber, M.; Klotz, D.; Ivers-Tiffée, E. (2014). A Method for Improving the Robustness of linear Kramers–Kronig Validity Tests. Electrochimica Acta 131, 20–27. DOI: 10.1016/j.electacta.2014.01.034. See also the authors’ KIT Lin-KK methodological explanation. Publication · Authors’ Lin-KK explanation

[16] Kempler, P. A.; Nielander, A. C. (2023). Reliable reporting of Faradaic efficiencies for electrocatalysis research. Nature Communications 14, 1158. DOI: 10.1038/s41467-023-36880-8. Open-access article

[17] Joint Committee for Guides in Metrology (2012). International vocabulary of metrology—Basic and general concepts and associated terms (VIM), third edition. JCGM 200:2012. See accuracy, precision, reproducibility and measurement-uncertainty definitions. Official vocabulary

[18] US Department of Energy. Safe Use of Hydrogen. Official safety overview. Official overview

[19] Stanford Research Systems (2025). EC301 Potentiostat/Galvanostat User Manual, revision 1.3, 14 March 2025. Selected control, output-state, measurement-range and grounding sections. Configuration-specific documentation; not a universal connection method. Manufacturer manual

[20] Gamry Instruments. Framework help: Electrochemical Frequency Modulation setup parameters. DC Voltage and open-circuit-reference fields only. Cited for software reference terminology—not as a PEM procedure or a recommendation to use EFM. Official software documentation

[21] Prestat, M. (2023). Corrosion of structural components of proton exchange membrane water electrolyzer anodes: A review. Journal of Power Sources 556, 232469. DOI: 10.1016/j.jpowsour.2022.232469. Peer-reviewed materials assessment.

[22] Chemours. Teflon™ Fluoropolymers Product Selection Guide, electrical properties of PTFE (typical volume resistivity >10¹⁸ Ω cm). Manufacturer materials data; values are not a specification for a particular test-cell assembly. https://www.teflon.com/en/products/product-selection-guide