How Voltage Spikes Are Helping Build the Future of Green Steel

The arc voltage fluctuates by up to 15 volts during unstable plasma conditions — a seemingly small number, but one that signals chaotic energy swings capable of derailing an entire steelmaking process designed to eliminate carbon emissions.

This isn’t just noise. It’s a fingerprint of instability in hydrogen plasma arcs, the fiery heart of a next-generation technology called Hydrogen Plasma Smelting Reduction (HPSR). If perfected, HPSR could decarbonize one of the world’s dirtiest industries: steel production, which emits more CO₂ annually than all cars and trucks on Earth combined. But today, that promise is held back by something invisible, volatile, and poorly understood — the electric arc.

In a new study, Thomas Voglhuber-Brunnmaier and colleagues at Johannes Kepler University in Austria reveal how the electrical supply system itself can destabilize the arc, turning what should be a steady plasma torch into a flickering, inefficient, electrode-gnawing beast. More importantly, they show that arc length — the physical manifestation of stability — is tightly linked to voltage. And that means we can measure instability in real time using nothing more than standard electrical sensors. From there, they demonstrate active stabilization using a fast-switching buck converter with adaptive control, achieving current stability within 0.03% variation.

This isn’t theoretical. It’s instrumentation meeting control theory in a lab-built smelting cell, where every volt and millisecond counts. The result? A blueprint for taming plasma in real time — a critical step toward green steel at scale.

The Science

Steel is the backbone of modern civilization. It builds our cities, powers our grids, and carries our goods. But traditional steelmaking relies on coking coal to reduce iron ore, releasing roughly two tons of CO₂ for every ton of steel. That’s unsustainable. Enter HPSR: a process that replaces carbon with hydrogen as the reducing agent. When hydrogen is produced via electrolysis using renewable electricity, the entire process can be nearly carbon-free.

In HPSR, hydrogen is ionized in an electric arc between a graphite electrode and a molten iron bath. The plasma provides both the energy to melt the ore and the reactive hydrogen ions to reduce it. No coal. No direct CO₂ emissions. Just heat, light, and the promise of clean metal.

But there’s a catch: plasma arcs are inherently unstable. They writhe, stretch, and snap unpredictably due to turbulence, electromagnetic forces, and thermal buoyancy. These fluctuations degrade efficiency, waste hydrogen, and erode expensive electrodes — a major barrier to commercialization.

While much research has focused on the metallurgical side, Voglhuber-Brunnmaier’s team zeroes in on an overlooked culprit: the electrical supply system. Most industrial power supplies behave like voltage sources, but arcs have negative differential resistance — meaning as current increases, voltage drops. This counterintuitive behavior can destabilize the circuit, especially when parasitic inductance and capacitance are involved.

To probe this, the team built a custom experimental setup: a transparent, four-sided chamber with welding-grade optical filters, allowing high-speed stereo imaging of the arc from two orthogonal angles

. A graphite electrode is lowered toward a steel plate, igniting the arc, while gas (argon or hydrogen mixtures) is introduced to simulate process conditions.

Crucially, they didn’t rely on standard lab power supplies. Instead, they designed a fast current controller based on a buck converter — a type of DC-DC power converter that can adjust output voltage rapidly by switching an IGBT transistor at 50 kHz. This is paired with an adaptive state-feedback controller running in real time on a Beckhoff industrial PC, sampling current and voltage every 125 microseconds.

To measure arc stability directly, they developed a stereoscopic vision system using two Raspberry Pi cameras with global shutter sensors and 16mm lenses

. By defining a region of interest (ROI), they achieved a frame rate of 100 Hz, sufficient to track arc motion. The arc is detected in both views, reconstructed in 3D via spline interpolation, and its length calculated from the centerline.

This dual instrumentation — electrical control and optical measurement — allows them to correlate what’s happening electrically with what’s happening physically in the plasma.

What They Found

The first clue came from re-examining old data. In prior experiments using a conventional thyristor-based power supply, voltage and current signals were erratic

. When plotted in the U-I plane, the arc appeared to jump between multiple operating points

.

By analyzing block-wise residuals of a simple arc model (fitting differential resistance and open-circuit current), the researchers identified brief periods of stability — just seconds long — where the arc behaved predictably. During these windows, the arc wasn’t flickering randomly; it was switching between two or more metastable states, likely due to shifts in attachment point on the melt pool.

But here’s the key insight: when the arc length changed, so did the voltage — even at constant current. A longer arc meant higher voltage. A shorter arc, lower voltage. This isn’t surprising in principle (voltage should scale with gap distance), but proving it empirically under dynamic conditions is critical.

In their new setup with active control, they stretched the arc by slowly raising the electrode while holding current steady at 18 A. The results were clear: arc voltage rose in tandem with arc length

Arc Voltage Correlates with Arc Length

Label	Value
Time (s)	285
Time (s)	300
Time (s)	315

Data from Figure 13 in the paper. Voltage tracks arc length during controlled extension at constant current (18 A).

. The coefficient of variation in current was just 0.03%, confirming the controller’s precision. Voltage, however, varied by several volts — directly tracking the physical extension of the plasma channel.

This means voltage isn’t just a side effect; it’s a proxy for arc length. And since arc length determines energy distribution, hydrogen utilization, and electrode wear, voltage becomes a real-time diagnostic for process health.

They also confirmed that instability isn’t just a plasma problem — it’s a system problem. When they used a simple lab power supply without sufficient series resistance, the arc wouldn’t ignite reliably. Adding a 1 Ω resistor (mimicking the resistance of a 100 mH inductor) stabilized ignition — not because of inductance, but because it satisfied the stability condition: source resistance must exceed the arc’s negative differential resistance.

The buck converter, with its large output inductor (2.5 mH) and fast adaptive control, acts like a near-ideal current source. It suppresses oscillations that arise from the arc’s negative resistance interacting with circuit parasitics. The result is a smooth, damped response even during rapid setpoint changes.

Arc Voltage Correlates with Arc Length

Label	Value
Time (s)	285
Time (s)	300
Time (s)	315

Data from Figure 13 in the paper. Voltage tracks arc length during controlled extension at constant current (18 A).

Why This Changes Things

Steel accounts for about 7–9% of global CO₂ emissions. Decarbonizing it is essential for climate goals. HPSR is one of the most promising pathways, but only if it can be made stable, efficient, and durable.

Today, most pilot projects focus on chemistry and materials. This paper shifts attention to control engineering — the hidden discipline that turns lab curiosities into industrial processes.

Consider the implications:

• Electrode wear: Unstable arcs cause localized hot spots and erratic current paths, accelerating graphite erosion. In industry, electrode consumption can account for 10–15% of operating costs. By stabilizing the arc, this technology could extend electrode life dramatically.
• Energy efficiency: Fluctuating arcs waste energy through radiation, convection, and incomplete reduction. A stable arc delivers power predictably, improving thermal efficiency and reducing electricity demand — crucial when that electricity comes from expensive green sources.
• Hydrogen utilization: Hydrogen is costly to produce and store. If plasma instability leads to poor mixing or incomplete dissociation, utilization drops. A steady arc ensures optimal conditions for hydrogen to react with iron oxide.
• Scalability: Industrial electric arc furnaces operate at tens of megawatts. Without active stabilization, scaling HPSR could amplify instabilities, not reduce them. This work proves that fast electrical control can suppress those instabilities before they grow.

But perhaps the most profound implication is measurability. In most industrial smelting systems, the arc is buried under molten metal and slag, invisible to cameras. You can’t control what you can’t measure. By showing that voltage correlates with arc length, the team opens the door to indirect sensing — using electrical signals alone to infer plasma geometry.

This transforms the control problem. Instead of relying on complex, failure-prone optical systems, future HPSR plants could use standard voltage and current sensors to estimate arc length in real time, feeding that data into a supervisory controller. That’s a game-changer for robustness and cost.

It also reframes the role of power electronics in heavy industry. Traditionally, power supplies are seen as dumb energy delivery systems. Here, they become active control actuators — capable of damping oscillations, rejecting disturbances, and shaping plasma dynamics on millisecond timescales.

What’s Next

The team has built a powerful foundation, but challenges remain.

First, the current system assumes a relatively clean, visible arc. In real HPSR reactors, the arc will be submerged under slag and metal, making optical validation impossible. The correlation between voltage and arc length must be validated under industrial conditions — perhaps using surrogate measurements or digital twins.

Second, while voltage tracks length, it doesn’t capture shape. An arc can be long and straight or long and coiled — two very different states with different heat transfer and chemical kinetics. Future work could combine voltage with higher-order electrical signatures (e.g., noise spectra, harmonic content) to infer more complex plasma geometries.

Third, the controller demonstrated here regulates current, not power or arc length. The next step is closed-loop arc length control — using voltage as a feedback signal to actively adjust current and maintain a desired arc geometry. This would require a model that maps electrical inputs to plasma length, possibly using machine learning trained on stereo data.

Finally, scaling up introduces new physics. At multi-megawatt levels, electromagnetic forces, three-phase interactions, and large thermal inertia will dominate. The 50 kHz switching used here may not suffice. But the principle remains: fast, intelligent power electronics can stabilize inherently unstable processes.

The broader lesson extends beyond steel. Plasma is used in semiconductor manufacturing, waste treatment, and fusion research — all domains where stability is paramount. This work demonstrates that by treating the power supply not as a passive source but as an active participant in system dynamics, we can tame even the most chaotic phenomena.

Green steel won’t be built in a single breakthrough. It will be forged in increments: better materials, cleaner hydrogen, smarter controls. This paper delivers one such increment — quiet, precise, and essential. It shows that the path to decarbonization isn’t just about replacing fossil fuels with renewables. It’s about rethinking how we control the processes that shape our world.

And sometimes, that control begins with a single number: the voltage across an arc.