Drier membranes unlock reliable fuel cell cold starts down to minus 18 degrees

Fuel cell electric vehicles promise zero-emission driving with long ranges and rapid refueling, but they face a stubborn winter problem: the water their own operation produces can freeze inside the stack while the vehicle is parked, blocking reactant supply and preventing the engine from starting at all. A new experimental study published in European Transport Research Review now maps out, on an automotive-sized laboratory stack, which freeze-start strategies actually work, and identifies one dominant factor: how thoroughly the membrane is dried before the stack cools down.

The research team, led by Markus Schwarz with co-authors Albert Albert and Rüdiger-A. Eichel, worked with a 10-cell proton exchange membrane (PEM) short stack, each cell with an active area of 285 square centimeters — a scale the authors argue is far more representative of vehicle hardware than the small single cells used in most prior freeze-start experiments. Crucially, the test bench was fitted with an additional cooling circuit designed to reproduce the thermal behavior of a real vehicle, allowing active cooling of the stack down to minus 18 degrees Celsius and heat-up rates that closely match those derived from vehicle data.

The physics of the problem is unforgiving. Inside a PEM fuel cell, hydrogen reacts with oxygen from air at the catalyst layer on the cathode side, producing electricity, heat and water. Liquid water and moisture end up distributed throughout the system — in the bipolar plates, the gas diffusion layers, the catalyst layers and the membrane itself. When a parked vehicle cools to ambient temperature, which takes roughly 14 hours, any residual water can turn to ice. On the next start attempt, that ice can block gas channels and pores, starving cells of hydrogen and triggering cell reversal, a damaging condition in which negative voltages force alternative reactions such as oxygen evolution and carbon oxidation to sustain the current.

The study breaks a freeze start into distinct phases: freeze preparation (a drying phase after shutdown), stack cooling, and the freeze start itself with internal heat-up. During drying, humidification is switched off and the stack is flushed with dry air and fresh hydrogen; the progress of drying can be tracked in real time through the high-frequency resistance, or HFR, typically measured at 1000 kilohertz, since a drier membrane conducts protons less readily and therefore shows higher resistance. The authors emphasize that drying is a balancing act — excessive drying raises protonic resistance and can mechanically stress the membrane and catalyst support, while insufficient drying invites failed starts and accelerated degradation.

In the experimental sequence, each test began from a defined operating point at 1 A/cm² and 65 degrees Celsius, after which current and flows were ramped down, drying began, and a brief nitrogen purge sealed both compartments. The stack was then actively cooled at 6 kelvin per minute to the target starting temperature, held for 25 minutes to ensure a homogeneous temperature distribution, and finally restarted: hydrogen to the anode, air to the cathode, an open-circuit phase, and then a potentiostatic heat-up in which the stack voltage was stepped down to a target value, forcing current through the cells and generating heat internally until the stack outlet exceeded 15 degrees Celsius — the point at which ice formation is no longer expected. From there, control switched to galvanostatic mode and the stack was warmed to its normal 60-degree operating range.

The central result concerns the HFR threshold at which drying is terminated. With low thresholds of 0.20 to 0.25 ohm·cm² per cell, the stack could not be started below about minus 10 degrees Celsius without cell voltages dropping below minus 0.1 volts — the study’s failure criterion, chosen to guard against cell reversal. Raising the threshold to 0.40 ohm·cm² per cell extended the minimum start temperature to minus 15 degrees, and the most aggressive variant, FP5 at 0.60 ohm·cm² per cell, enabled reliable starts at minus 18 degrees with no negative cell voltages; the weakest cell still delivered a minimum voltage of 0.108 volts. The explanation is straightforward: a drier membrane retains more capacity to absorb the product water generated during the cold start itself, so less water reaches the gas diffusion layers, catalyst layers and flow channels where it could freeze and cause blockages.

Notably, two parameters that might have been expected to matter did not. Increasing the cathode airflow during drying, and raising the HFR threshold only slightly from 0.20 to 0.25 ohm·cm² per cell, produced no measurable improvement in the minimum start temperature. The authors caution, however, that the temperature intervals tested were coarse — jumps of 5 degrees — so small gains could have escaped detection. They also stress that drying must be homogeneous across the stack: unavoidable cell-to-cell flow maldistribution from manufacturing tolerances can leave individual cells under- or over-dried and cause premature failure even when the average HFR target is met.

The second part of the study examined how the heat-up strategy itself should be run. Varying the cathode mass flow by plus or minus 0.10 grams per second around a baseline of 0.75 grams per second had only moderate effects: heat-up rates ranged from 0.26 to 0.30 kelvin per second, and total water production stayed within a narrow band of roughly 9.75 to 10.60 grams. Varying the stack voltage during potentiostatic operation proved far more consequential. Lowering the target 10-cell stack voltage from 5.25 volts to 2.75 volts raised the heating rate from 0.26 to 0.32 kelvin per second and, counterintuitively, reduced total water production until the 15-degree mark from 10.3 grams to 8.8 grams — even though the instantaneous water production rate was higher for the low-voltage case. The reason is efficiency: at lower voltage the fuel cell is less efficient, converting more of its energy into heat, so the stack defrosts faster and accumulates less total water before ice is no longer a threat.

The trade-off is electrical power. The lowest-voltage strategy, FS8 at 2.75 volts, delivered the lowest average electric power of the tested variants at 547 watts, while the highest power, 752 watts, was recorded for the intermediate FS3 configuration. The authors argue this trade-off is acceptable for real vehicles, citing vehicle-level analyses indicating that power demand immediately after a cold start is low. They also note that once the stack passes 15 degrees and control switches to galvanostatic operation, careful current ramp-up and elevated cathode stoichiometry remain essential to prevent local flooding, hydrogen starvation and the carbon corrosion that global starvation can cause on anode catalyst layers.

The study’s broader context is the growing role of fuel cell vehicles in decarbonization scenarios, particularly for long-range, rapid-refueling and heavy-duty applications. The U.S. Department of Energy requires unassisted fuel cell starts at temperatures as low as minus 30 degrees Celsius and targets 50 percent of rated power within 30 seconds of a start request at minus 20 degrees — targets the present methodology has not yet reached, since testing stopped at minus 18 degrees. One cited modeling study predicted around 1,241 freeze-start events over a 15-year service life for a fuel cell vehicle operated in Canada, underscoring how frequently the capability is exercised in cold climates. The authors are explicit about their limitations: the 10-cell short stack cannot capture all thermal, flow-distribution and control effects of full-size automotive systems, the parameter space was limited, and durability was assessed indirectly through startability criteria and voltage behavior rather than long-term degradation testing. Future work, they write, should extend the method to minus 30 degrees and validate the short-stack approach against a complete fuel cell system. Within its scope, however, the message is clear: combine aggressive but controlled drying with low-voltage potentiostatic heat-up, and a fuel cell stack can pull itself out of a deep freeze without external heaters, without cell reversal, and with less water left behind to freeze.

Subject of Research: Experimental investigation of freeze-start strategies for automotive-sized PEM fuel cell short stacks under vehicle-relevant thermal conditions

Article Title: Freeze start strategies for fuel cell electric vehicles: balancing durability, reliability and practicability – an experimental approach on automotive-sized short stacks

Article References: Schwarz, M., Albert, A., & Eichel, R.-A. (2026). Freeze start strategies for fuel cell electric vehicles: balancing durability, reliability and practicability – an experimental approach on automotive-sized short stacks. European Transport Research Review, 18(1), Article 83. https://doi.org/10.1186/s12544-026-00845-x

Image Credits: AI Generated

DOI: 10.1186/s12544-026-00845-x

Keywords: fuel cell electric vehicles, PEM fuel cell, freeze start, cold start, sub-zero temperatures, membrane drying, high-frequency resistance, potentiostatic operation, short stack, cell reversal, hydrogen mobility, degradation