PEM Fuel Cell

The one-reaction power plant

A Proton-Exchange Membrane fuel cell is a galvanic cell. Like every battery you have ever touched, it has an anode, a cathode, an electrolyte separating them, and an external circuit that carries the current. What makes it a fuel cell rather than a battery is that the reactants — hydrogen and oxygen — flow in continuously from outside, so the device delivers power for as long as you keep feeding it. There is nothing inside to charge, and nothing to deplete.

The overall reaction is the simplest equation in physical chemistry:

H₂ + ½O₂ → H₂O    ΔG° = −237.1 kJ/mol    ΔH° = −285.8 kJ/mol

The Gibbs free-energy change ΔG sets a thermodynamic ceiling on the electrical work the cell can extract. Dividing by 2F (two electrons per H₂, F = 96485 C/mol) gives the reversible cell potential

E°_cell = −ΔG° / nF = 237 100 / (2 × 96 485) ≈ 1.229 V

which is the open-circuit voltage you would measure if the cell were perfectly reversible and you drew zero current. The ratio ΔG/ΔH ≈ 0.83 — the maximum fraction of the fuel's enthalpy you can extract as electricity. Everything between that 83 percent ceiling and the ~50 percent you observe in a real automotive stack is irreversibility, and engineering a PEM fuel cell is largely the project of pushing those irreversibilities as far down as possible.

Anatomy of a single cell

Slice a PEM cell perpendicular to its membrane and you see five distinct layers, sandwiched between two bipolar plates. Reading from the anode side toward the cathode:

• Anode bipolar plate with serpentine flow channels machined into the face. The channels supply H₂ across the active area; the plate body is a graphite composite or stamped stainless steel and carries the electronic current to the next cell in the stack.
• Anode gas-diffusion layer (GDL). A few hundred microns of carbon-fibre paper or cloth, optionally microporous-layer (MPL) coated. The GDL spreads incoming H₂ laterally so it reaches the catalyst even between flow-channel ribs, and conducts electrons back to the bipolar plate. Hydrophobic PTFE treatment helps liquid water move the right way.
• Anode catalyst layer. Platinum or platinum-ruthenium nanoparticles (2-5 nm) dispersed on high-surface-area carbon support, mixed with Nafion ionomer binder. Total Pt loading 0.05-0.1 mg/cm² on the anode side. This is where the hydrogen oxidation reaction (HOR) happens.
• Proton-exchange membrane. 15-25 µm of Nafion (or modern reinforced equivalents like Gore-Select). Conducts protons; blocks electrons and gases.
• Cathode catalyst layer. Heavier Pt loading (0.2-0.4 mg/cm²) because the oxygen reduction reaction (ORR) is the kinetically sluggish one. Same carbon-supported nanoparticles + ionomer recipe.
• Cathode gas-diffusion layer. Spreads incoming air, carries product water out, returns electrons from the external circuit to the catalyst.
• Cathode bipolar plate. Mirror of the anode plate, with air-side flow channels.

The three middle layers — catalyst-coated membrane and the two GDLs — are pre-bonded into a single component called the membrane-electrode assembly (MEA). The MEA is the cell. Everything else is plumbing.

Anode half-reaction: hydrogen splits on platinum

Hydrogen gas dissociatively adsorbs on a Pt surface, then surrenders its two electrons:

H₂ → 2 H⁺ + 2 e⁻     E° = 0.000 V  (anode, by definition)

The kinetics are fast — the exchange current density i₀ for HOR on Pt is roughly 10⁻³ A/cm², so the activation overpotential at typical operating currents (1 A/cm²) is only a few tens of millivolts. This is why the anode is not the bottleneck and why anode Pt loadings can be kept low.

HOR kinetics are catastrophically poisoned by even ppm-level CO. Carbon monoxide adsorbs preferentially on Pt active sites, displacing H₂ and dropping the active surface area. If the H₂ comes from clean electrolysis you are safe; if it comes from steam-methane reforming you have to either polish the reformate to <0.2 ppm CO or add ruthenium to the anode catalyst (Pt-Ru), which oxidises adsorbed CO to CO₂ at lower potentials than pure Pt can.

The membrane: Nafion's nanoscopic water highway

The proton produced at the anode is in fact not bare — it travels as a hydrated H₃O⁺, or as Zundel (H₅O₂⁺) and Eigen (H₉O₄⁺) cations, hopping from water molecule to water molecule by the Grotthuss mechanism. Nafion's job is to provide a continuous network of hydrated, fixed-charge channels for that proton to navigate.

Chemically, Nafion is a perfluorinated backbone resembling Teflon, with periodic side chains terminating in sulfonic-acid groups (-SO₃H). When wet, those acidic groups dissociate, leaving anchored -SO₃⁻ sites and free H⁺. The hydrophobic backbone and hydrophilic side groups microphase-separate into a sponge-like structure with 2-4 nm water-filled channels. Inside those channels protons hop along chains of water molecules at room-temperature mobility close to that of bulk acid.

The relevant material parameter is the proton conductivity σ. At 80 °C and 100 % relative humidity, σ ≈ 0.1 S/cm. Drop the humidity to 30 % and σ falls roughly an order of magnitude. Drop the temperature below freezing and the water in the channels solidifies — which is why automotive PEM stacks include cold-start heating circuits and why the "freeze-thaw cycles" specification is a serious durability target.

Cathode half-reaction: the slow side

At the cathode, the returning electrons and protons reduce molecular oxygen:

½ O₂ + 2 H⁺ + 2 e⁻ → H₂O    E° = +1.229 V (vs SHE)

This reaction is the limiting step of every low-temperature acid-electrolyte fuel cell. The exchange current density i₀ for ORR on Pt is roughly 10⁻⁸ A/cm² — five orders of magnitude smaller than HOR. To pull useful current you need to overpotential the cathode by ~0.3 V, which is the single biggest chunk of voltage you lose.

Catalyst-research effort for PEM cells is therefore overwhelmingly directed at the cathode: Pt-alloy catalysts (Pt-Co, Pt-Ni) with engineered surfaces that bind oxygen intermediates a little less tightly, shape-controlled nanoparticles exposing the most active crystal facets, and various non-precious-metal alternatives based on iron- or cobalt-doped nitrogen-coordinated carbons. None of the non-Pt families has yet matched a Pt-alloy at automotive durability targets, which is why Pt is still the default and the Pt loading is still the dominant catalyst-cost lever.

The polarisation curve: where the voltage goes

If you plot cell voltage vs. current density, you get a curve that bends down through three regions:

V(i) = E_rev − η_act(i) − iR_ohm − η_conc(i)
     = 1.23 − [a + b·log(i/i₀)] − iR_ohm − (RT/nF)·ln[1 − i/i_lim]

The terms, in order:

• η_act — activation overpotential. The voltage you must give up to push the ORR forward. Tafel form b·log(i/i₀); on Pt b ≈ 70 mV/decade. Dominates at low current.
• iR_ohm — ohmic loss. The product of cell current density and the area-specific resistance, which sums membrane proton resistance, contact resistances and bipolar-plate bulk resistance. Linear in i; dominates the mid-current region.
• η_conc — concentration overpotential. At high current the cathode catalyst sees a depleted O₂ supply; the voltage collapses as i approaches a mass-transport limit i_lim. Sets the right-hand wall of the operating window.

The cell power density is i·V, so it has a clean maximum at the "knee" of the curve. Automotive stacks sit a little to the left of peak power to keep durability and efficiency reasonable — typically 0.65-0.75 V at 1.0-1.5 A/cm². A 200 cm² active-area cell at 0.7 V × 1.2 A/cm² delivers 168 W, so a 100 kW stack needs roughly 600 such cells — or, more practically, 300-400 larger cells.

Building a stack: 1 V per cell, 300 V from the bus

A single MEA gives you 0.7 V. The traction motor wants 350 V. The fix is to stack cells in series, sharing bipolar plates between adjacent cells so that each plate is anode for one cell and cathode for the next:

Vehicle	Cells	Stack power	Stack voltage	Tank pressure	Range
Toyota Mirai (gen 2)	330	128 kW	~300 V	700 bar	~650 km
Hyundai Nexo	440	95 kW	~310 V	700 bar	~666 km
Honda Clarity FC	360	103 kW	~320 V	700 bar	~589 km
Plug Power GenDrive forklift	~80	10 kW	~48 V	350 bar	8 hr shift
Hydrogen city bus (typ.)	~400	150 kW	~600 V	350 bar	~450 km

The fuel cell never powers the wheels alone — every commercial FCEV is a hybrid. The stack feeds a battery buffer (typically 1-3 kWh in a passenger car, 30+ kWh in a bus) which absorbs regenerative braking and supplies transient peak power that the stack would otherwise have to follow. Operating the stack at near-constant load extends its life dramatically, because the worst durability stress in a PEM cell is potential cycling between idle and full power — which oxidises and dissolves the Pt nanoparticles.

Water management: the engineering core of a PEM stack

Every PEM design decision orbits the water balance. Nafion needs to stay wet. Cathode product water generation accelerates with current. Air flow drags water out. Stack temperature controls whether that water is vapour or liquid.

Two failure modes bound the operating window:

• Drying. If inlet air is too dry or stack temperature too high, Nafion dehydrates, ohmic resistance climbs, the cell goes into a death spiral of more heat → drier membrane → more ohmic loss.
• Flooding. If inlet air is too saturated or stack temperature too low, liquid water condenses in the GDL pores and blocks O₂ from reaching the catalyst. Concentration overpotential explodes; the cell stalls.

The fix is a coordinated system: a membrane-based humidifier on the inlet air, ribbed serpentine flow fields that sweep liquid water by shear, hydrophobic-coated GDLs, careful selection of operating relative humidity (typically 80-100 % RH inlet at 70-80 °C), and dynamic control of stoichiometric flow ratios (typically 1.5× stoichiometric H₂, 2× stoichiometric air). The whole subsystem is called the "balance of plant" and it is often two-thirds of stack volume.

PEM vs. its alternatives

PEM is one fuel-cell technology among several. The choice depends on application:

Type	Electrolyte	T_op	Catalyst	Typical use
PEMFC	Nafion polymer	60-80 °C	Pt / Pt-alloy	Cars, trucks, forklifts, portable
SOFC	Y-stabilised ZrO₂	700-1000 °C	Ni cermet	Stationary CHP, grid backup
MCFC	Molten Li/K carbonate	~650 °C	Ni	Utility-scale stationary
PAFC	Liquid H₃PO₄	~200 °C	Pt	Hospitals, data centres
AFC	Aqueous KOH	50-200 °C	Ni / Ag	Apollo / Shuttle, niche
DMFC	Nafion polymer	50-90 °C	Pt-Ru / Pt	Portable electronics, sensors

PEM wins for transportation because of its high power density, rapid start-up (seconds, not hours), and tolerance of frequent load transients. SOFCs are better at extracting every joule from a fuel but take an hour to come up to operating temperature — fine for a stationary CHP unit, useless for a forklift. PEM's weakness is its requirement for clean H₂ and noble-metal catalyst; SOFCs can run on natural gas internally reformed.

PEM vs. battery vs. ICE

An FCEV competes with a BEV and an ICE car. Where each wins:

Metric	BEV	FCEV (PEM)	ICE
Tank-to-wheel efficiency	~80 %	~50 %	25-30 %
Wells-to-wheel (from grid)	~70 %	~30 %	~15 % (refining + combustion)
Refuel/recharge time	20-60 min DC fast	~5 min	~3 min
Energy density (storage)	~250 Wh/kg (pack)	~1500 Wh/kg (700-bar H₂)	~12 000 Wh/kg (gasoline)
Cold-weather range	−30 % at −10 °C	−10 % at −10 °C	−5 %
Infrastructure cost / station	~$100 k DC fast charger	~$2-3 M 700-bar H₂ station	~$500 k gas station
Tailpipe	None	Water	CO₂ + NOx + particulates

The pattern is clear: BEVs are dominant for light-duty cars where round-trip efficiency and plug-in convenience win. FCEVs are competitive for heavy-duty (trucks, buses, forklifts, ferries) where high energy density and fast refuelling matter more than wells-to-wheels efficiency. ICE retains a niche where the existing fuel-distribution network is the deciding factor and emissions regulation is lax.

Where the $ live: a 2026 stack cost breakdown

An 80 kW automotive PEM stack at automotive production volumes (≥ 500 000 units/year) breaks down approximately as:

• Bipolar plates (~32 %) — stamped stainless or graphite composite, with corrosion-resistant coatings.
• Membrane (~10 %) — reinforced Nafion or PFSA equivalent at $300-500/m².
• Catalyst + Pt (~25 %) — dominated by Pt at ~$3000/kg metal. 30 g of Pt in an 80 kW stack ≈ $90 of metal, but processing and ionomer push the catalyst-layer cost much higher.
• GDLs + MPL (~10 %) — carbon-fibre papers with PTFE and microporous coatings.
• Frames, gaskets, seals, end-plates (~13 %) — non-trivial because seal failures cause crossover.
• Assembly + QC (~10 %) — leak testing, conditioning, polarisation-curve verification.

2026 industry target is < $40/kW at scale; current production hovers around $60-90/kW. The reduction path is thinner Nafion, less Pt at the cathode, cheaper stamped-metal plates, and amortising R&D over more units sold.

Durability: 5,000 hours and counting

Automotive durability targets are 5,000 operating hours with <10 % power degradation. The failure modes are well catalogued:

• Pt nanoparticle Ostwald ripening. Small Pt particles dissolve preferentially, larger ones grow. Active surface area shrinks; cathode overpotential climbs. Driven by potential cycling.
• Carbon-support corrosion. At high cathode potentials, the carbon black supporting the Pt oxidises (C + 2H₂O → CO₂ + 4H⁺ + 4e⁻). Especially severe during start-stop and reverse-current events.
• Membrane chemical degradation. Radical attack by H₂O₂ intermediates breaks the Nafion backbone. Mitigated with radical scavengers (cerium oxide) in modern formulations.
• Membrane mechanical fatigue. Hydration cycling swells and shrinks Nafion; over thousands of cycles micro-cracks form, eventually leading to H₂ crossover and a hard failure.
• GDL hydrophobic-coating loss. PTFE migrates and degrades; the GDL becomes wettable, flooding worsens.

Hybridisation with a battery buffer, which lets the stack avoid large transient swings, is the single most effective lifetime-extension measure. Modern automotive stacks meet the 5,000-hour target; bus stacks (which run nearer constant load) exceed 25,000 hours in field data.

Where you find PEM fuel cells in the wild

• Passenger FCEVs. Toyota Mirai (gen 1: 2014, gen 2: 2020), Hyundai Nexo, Honda Clarity FC. Volumes are small — tens of thousands per year globally — and concentrated in markets with H₂ refuelling networks (California, Japan, South Korea, Germany).
• Hydrogen city buses. Operating in Aberdeen, London, Pau, Cologne, Shanghai, Beijing and dozens of other cities. Duty cycle (predictable route, daily depot fill, slow steady speed) is ideal for PEM. ~150 kW stacks; range 400-500 km.
• Materials-handling forklifts. Plug Power has put > 70,000 PEM-cell forklifts into Walmart, Amazon, BMW, Home Depot warehouses. Lithium-ion forklifts have to recharge for 8 hours; H₂ fills in 3 minutes, runs all day. Single best PEM commercial application.
• Heavy trucks and long-haul. Hyundai XCIENT (200 kW PEM) in service in Switzerland; Hyzon, Nikola, and Daimler/Volvo Cellcentric developing semi-truck classes. The weight advantage over batteries is decisive at long range.
• Backup power and telecom. Stationary 5-50 kW PEM gensets replace diesel for cell towers and data-centre auxiliaries, particularly where regulated emissions or noise rule out engines.
• Marine and aviation niches. Hydrogen-fuelled ferries (San Francisco's Sea Change, Norway's pilot routes), drone fuel cells for extended flight time, early-stage aviation programs (ZeroAvia, H2Fly).

Common pitfalls

• Quoting 83 % "thermodynamic efficiency" as the real number. The 83 % is ΔG/ΔH at standard conditions. Real cells operate at 0.7 V vs 1.48 V (thermoneutral), giving ~47 % cell-level efficiency; system efficiency including auxiliaries is lower still.
• Mixing up open-circuit and operating voltage. OCV is ~0.95-1.0 V (less than 1.23 V because of mixed potentials and crossover). The useful operating point is 0.6-0.75 V; nobody runs a stack near OCV because there is no current.
• Treating H₂ "green" without qualifying the source. 96 % of today's commercial H₂ is grey (steam-methane reforming, ~10 kg CO₂ per kg H₂). Green H₂ requires renewable-electrolysis input. The fuel cell is only as clean as its hydrogen supply chain.
• Ignoring stoichiometry. Cathode air is supplied at typically 2× the stoichiometric requirement to manage water; H₂ is recirculated at near 1× stoichiometric with a periodic purge. Both the parasitic blower power and the purge losses cut into system efficiency.
• Cold-start denial. A PEM stack frozen below 0 °C cannot start cleanly — ice in the catalyst pores blocks current. Automotive systems include heating bands, glow plugs, and antifreeze loops, and the cold-start protocol is its own subspeciality.