Hydrogen ... The Question That Would Not Go Away

Whenever we present hydrogen-based reduction to industry, the same question comes up: If you make steel with hydrogen, will the steel be full of hydrogen? Will it break? Hydrogen embrittlement is real. A few parts per million of diffusible hydrogen can turn a high-strength steel into a brittle failure waiting to happen. So the concern is legitimate. If green steel is inherently hydrogen-charged, it would be useless for safety-critical components.

Our team at the Max Planck Institute für Eisenforschung decided to measure it directly. Not model it. Not argue from thermodynamics. Put hydrogen-reduced iron into a thermal desorption spectrometer and count the atoms coming out.

Two Routes, Two Materials, One Question in Hydrogen-Based Green Steel Production

We produced iron via two hydrogen-based routes:

HyDR (solid-state direct reduction). Commercial hematite pellets were reduced in pure H₂ at 900°C for about 690 seconds, reaching 99% reduction. The product was sponge iron—porous, with 45% porosity, pores from 2.97 μm down to 269 nm, gangue inclusions (Si, Al, Ca, Mg oxides) heterogeneously distributed, grain sizes from 0.02 to 30.5 μm². Low-angle grain boundaries (dislocation arrays) dominated at 0.61 μm²/μm², high-angle boundaries at 0.19 μm²/μm². A complex defect landscape.

HPSR (plasma smelting reduction). Hematite melted and reduced simultaneously in an Ar-10% H₂ plasma at 900 mbar. The product solidified into compact iron with spherical gangue inclusions (mainly SiO₂) distributed homogeneously.

We also took the HyDR sponge and melted it in an arc furnace under pure Ar—three melting cycles, 65 seconds each—to simulate what happens when sponge iron goes into an electric arc furnace. We call this HyDR+met.

Thermal Desorption Spectroscopy: What Came Out

We used a G4 Phoenix DH instrument. Samples were heated to 800°C at three ramping rates (800, 1000, 1200°C/h) and also rapidly (within 1 minute) to the same temperature. A mass spectrometer recorded the hydrogen desorption current. Integration gave total hydrogen content.

The as-reduced HyDR sponge iron released a lot of hydrogen. The desorption spectrum showed a large integrated area, corresponding to 39.9 ± 9.0 wppm (parts per million by weight). That is roughly 40 hydrogen atoms per million iron atoms—substantial, but still a dilute solute.

The HyDR+met sample (sponge iron melted) dropped to 1.46 ± 0.50 wppm. That is a 96% reduction. The HPSR product was even lower: 0.98 ± 0.50 wppm.

For comparison, conventional blast furnace hot metal contains 5–8 wppm. With vacuum arc degassing (VAD), that drops to 1–2 wppm. So green steel after melting is already at the vacuum-degassed level. No extra step needed.

Where Did the Hydrogen Go? Trap Deconvolution in Hydrogen-Based Direct Reduction Green Steel Production Processes

We deconvoluted the desorption spectrum of the HyDR sponge into four peaks using a Gaussian fitting routine. Then we applied the Kissinger method—plotting ln(Φ/T_p²) versus 1/T_p for three ramping rates—to extract activation energies for each peak.

The results:

Peak 1 (4.3 kJ/mol): Hydrogen from the bcc iron lattice and surface iron hydroxides. Weakly trapped.

Peak 2 (15.2 kJ/mol): High-angle grain boundaries and dislocations (the dislocation arrays that constitute low-angle boundaries). This is the classic trap population in deformed or fine-grained iron.

Peak 3 (59.1 kJ/mol): Nano-pores and magnetite (Fe₃O₄) inclusions. The 1.5 wt% residual magnetite we detected by XRD acts as a strong trap.

Peak 4 (126.1 kJ/mol): Gangue oxides—SiO₂, Al₂O₃, etc. These are the strongest traps. The activation energy matches literature values for hydrogen trapped at oxide interfaces.

Quantitatively: lattice/surface contributed about 8 wppm, grain boundaries/dislocations about 12 wppm, nano-pores/magnetite about 10 wppm, gangue oxides about 10 wppm. The porous, multiphase, fine-grained microstructure of sponge iron is an excellent hydrogen sponge.

Why Melting Removes Almost All Hydrogen

Two mechanisms operate during melting.

First, degassing. Liquid iron has a finite solubility for hydrogen (Sieverts' law: solubility proportional to square root of H₂ partial pressure). In an argon-filled furnace at 1 bar, the equilibrium hydrogen content in liquid iron is below 1 wppm. The driving force for outgassing is the concentration difference between the melt and the gas. With no hydrogen in the gas phase, the melt degasses rapidly.

Second, trap elimination. Melting collapses pores. It coalesces gangue inclusions into slag, which separates from the metal. It recrystallizes the grain structure, eliminating high-angle boundaries and dislocation arrays. The strong trapping sites simply disappear. What remains is a clean, compact iron with only lattice-dissolved hydrogen at the equilibrium level.

The HPSR process achieves the same low hydrogen content directly, because the melt is exposed to the plasma arc under conditions where hydrogen dissociates, reacts, and the product water vapor leaves, but the liquid metal itself degasses continuously.

What This Means for the Embrittlement Question for Steel Made via Hydrogen-Based Direct Reduction

The fear was that green steel would ship with 40 wppm hydrogen, ready to cause delayed cracking. Our data show the opposite. After the necessary melting step—which is part of every steelmaking route, because sponge iron must be liquefied for casting—the hydrogen content is 1–2 wppm. That is indistinguishable from conventionally produced steel that has been vacuum degassed.

We see no threat of hydrogen embrittlement arising from the use of hydrogen as a reductant. Downstream hydrogen uptake—from pickling, galvanizing, corrosive service, cathodic protection—can still happen, exactly as it does for conventional steel. That is a separate problem, addressed by coatings and microstructure design. But the raw green steel is not pre-charged.

One Subtlety We Should Acknowledge

Our TDS measurements were performed two to four weeks after sample preparation. That gap was deliberate—it simulates shipping and storage times for sponge iron. Hydrogen can outgas at room temperature, especially from weakly trapped sites. So our 40 wppm value for as-reduced sponge might be a lower bound for fresh material. But that does not change the conclusion: after melting, the number drops to 1–2 wppm regardless of the initial trap population.

We also used relatively pure hematite. Industrial ores contain more gangue. Stronger oxide traps could in principle retain more hydrogen through melting if they form stable, finely dispersed inclusions that do not coalesce into separable slag. That is a question we are now testing with three different ore grades. Preliminary results suggest the effect is small—the degassing mechanism dominates—but we will report quantitatively once the data are complete.

Practical Aspects in Hydrogen-Based Direct Reduction Green Steel Production 

If you run a HyDR shaft furnace, your sponge iron will come out with 30–50 wppm hydrogen, depending on cooling rate and porosity. Do not panic. That hydrogen will not survive the electric arc furnace. It will degas to below 2 wppm within minutes of melting. If you run HPSR, your liquid iron will already be below 1 wppm directly from the plasma reactor.

No vacuum degassing step is required for hydrogen control in green steel. That is a significant cost saving compared to conventional routes. The hydrogen that goes into the process leaves as water vapor, not as a dissolved embrittling agent.

Our data are open upon request. We encourage other groups to reproduce these measurements with their own reduction setups and ore types. The community needs a robust dataset across process conditions to put the embrittlement concern to rest. We have started that effort here.