Thermite Welding: Joining Rails With a River of Molten Iron

The reaction that brews its own steel

Thermite welding runs on a single, blunt piece of chemistry: aluminum wants oxygen more than iron does. The charge is a stoichiometric powder blend of iron(III) oxide (Fe₂O₃, the fully oxidised form found in red rust) and aluminum metal powder, in a mass ratio of roughly 3:1. When it is hot enough, the aluminum tears the oxygen away from the iron in the classic aluminothermic reaction:

Fe₂O₃ + 2 Al → 2 Fe + Al₂O₃   (ΔH ≈ −851 kJ/mol)

The thermodynamic justification is the Ellingham diagram: the line for Al→Al₂O₃ sits well below the line for Fe→Fe oxides across all practical temperatures, meaning aluminum oxide is far more stable. Aluminum therefore reduces iron oxide spontaneously once activated, and does so with a huge negative enthalpy — about 3.9 MJ per kilogram of thermite. Crucially, that energy is released into a very small mass of product, so it has nowhere to go but temperature. In practice the melt runs at roughly 2200–2500 °C, on the order of 1000 °C above iron's melting point of 1538 °C; the theoretical adiabatic temperature is higher still (about 2900 °C) and is capped there because the products themselves begin to boil, while a real crucible bleeds heat into its lining, the slag, and radiation. The products come out as a puddle of superheated liquid iron and a raft of liquid alumina (Al₂O₃) slag. Because the oxidizer is chemically bound inside the iron oxide, the reaction needs no air — it will run in a sealed crucible, underwater, or in a vacuum. Some formulations use magnetite instead: 3 Fe₃O₄ + 8 Al → 9 Fe + 4 Al₂O₃.

The trackside process, step by step

A field rail weld is a carefully choreographed casting operation. The rail ends are cut and aligned across a gap of about 25 mm (roughly one inch) — wide enough for molten metal to fully melt back both faces rather than merely wetting them. Then:

• Mold up: Two halves of a pre-formed refractory sand mold, shaped to the exact rail profile (head, web, foot), are clamped around the gap and luted with sealing paste. The mold defines the weld collar and the risers where excess metal and slag collect.
• Preheat: A propane–oxygen torch heats the rail ends to roughly 800–1000 °C for several minutes. This is not optional: cold steel would chill the incoming iron and freeze it before fusion, producing a cold weld. Short-preheat (SkV) and preheat-free (SoWoS) portions shorten or eliminate this step by carrying extra reaction heat in the charge.
• Charge and ignite: A single-use crucible holding a measured 5–13 kg thermite portion sits above the mold. It is lit with a magnesium/barium-peroxide sparkler — a match cannot start it because ignition requires about 1000–1200 °C. The reaction runs for roughly 20–30 seconds.
• Tap and pour: A fusible self-tapping thimble in the crucible floor melts through exactly when the reaction completes, releasing the molten steel downward into the mold. The iron floods the gap, melts back both rail ends, and fuses them as one casting; the lighter alumina slag floats up into the risers.
• Shear and grind: After about 3–4 minutes of solidification the mold is stripped, and a hydraulic weld shear trims the still-hot, plastic riser flush with the railhead. A profile grinder then dresses the running surface to within a fraction of a millimeter so the wheel never feels the joint.

Why the numbers come out where they do

Every parameter in the process is a consequence of getting a full-penetration fusion weld in a massive steel section with a fixed budget of chemical heat. The superheat is the whole point. If the iron merely reached 1538 °C it would arrive already freezing; the ~1000 °C of superheat is the thermal reserve that melts back the cold rail faces and drives fusion, not just adhesion. That is also why portion mass scales with rail cross-section — a heavy 60 kg/m rail needs more molten metal (and more delivered joules) than a light tram rail to avoid being quenched by its own thermal mass.

The slag separation is pure density physics. Molten iron has a density near 7 g/cm³; liquid alumina is only about 3 g/cm³. The alumina therefore floats cleanly, so a well-designed crucible and riser system delivers metal first and traps the slag above the weld — provided the tap timing is right. Poor timing is a classic route to slag inclusions. The weld metal itself is not plain iron: the portion carries deliberate alloy additions of carbon, manganese, silicon, and sometimes chromium so that the cast composition matches pearlitic rail steel (around 0.7 % C, e.g. grade R260). Match the chemistry and you match the hardness and wear rate; miss it and you build a soft spot or a brittle one that batters or cracks under traffic. Finally, the ~25 mm gap is a compromise: wide enough that the flood of iron remelts a healthy volume of parent metal for sound fusion, narrow enough that the fixed heat budget can still fully penetrate the foot — the thin, wide section the metal reaches first, the most chill-prone part of the profile and a common origin for weld defects.

Continuous welded rail and where thermite fits

Thermite welding is one of two technologies that made continuous welded rail (CWR) practical, ending the bolted fishplate joints responsible for the old "clickety-clack" — joints that were weak spots, maintenance sinks, and fatigue crack starters. Most CWR is produced by flash-butt welding in a plant or aboard mobile welding trains: an electric-resistance process that flashes and then forges the rail ends together under pressure with no filler, giving very consistent, low-defect joints. But flash-butt machines need heavy equipment, power, and clear access. That is exactly where thermite wins: it is portable, needs no electricity, and can close the final field gaps between long welded strings, and reach turnouts, crossings, tight curves, and in-track repairs where a big machine cannot go.

Once welded into a continuous ribbon, the rail can no longer expand and contract freely at joints, so thermal force is managed instead: CWR is de-stressed and fastened at a rail neutral temperature (typically ~21–38 °C, region-dependent) so the steel sits in modest tension when cold and modest compression when hot. A thermite weld must therefore not only be strong in tension but must survive decades of thermal cycling and wheel loading. The joint is, in effect, a small casting expected to perform like forged rail. The technique dates to Hans Goldschmidt, who developed the aluminothermic reaction in the 1890s (patented 1895); the first rail welds followed around 1899 on tramway track in Essen, Germany, and the company he founded — today Goldschmidt Thermit / Elektro-Thermit — along with Pandrol (formerly Railtech) still supplies the portions and molds worldwide.

How thermite welds fail, and how they are tested

Because a thermite weld is a field casting, it inherits every casting defect. The common failure modes are gas porosity (from moisture in the mold or portion — a wet mold is dangerous both for the weld and for the operator, since water flashing to steam can spatter molten iron), shrinkage cavities and hot tearing as the casting contracts, slag inclusions from bad tap timing, and incomplete fusion / cold welds from inadequate preheat. Any of these acts as a stress raiser and a nucleus for fatigue cracking under repeated wheel passes, most often in the railhead-to-web fillet or the weld foot. In service, thermite welds statistically show a higher break rate than flash-butt welds, which is why they are held to strict procedures and inspection.

The work is governed by real standards — EN 14730 in Europe (aluminothermic welding of rails, including welder and procedure qualification) and AREMA practice in North America. Quality is enforced by qualifying the weld procedure and the portion, not just inspecting after the fact: preheat time and temperature, gap, and alignment are fixed by the approved kit. Finished welds are checked by visual and dimensional inspection, ultrasonic testing and dye-penetrant / magnetic-particle methods for internal and surface flaws, and periodically by slow-bend and fatigue tests on sacrificial welds pulled from qualification lots. Rail-flaw detector cars later re-scan welds ultrasonically in track. The overriding principle is that a defect the size of a pinhole, hidden in the foot of a weld carrying 30-tonne axle loads, is a rail break waiting for a cold winter night.

Beyond rails: variants, cousins, and history

Goldschmidt's original goal in the 1890s was not welding at all — he was trying to produce carbon-free metals by reducing their oxides with aluminum, and the intense heat of the reaction was almost a by-product he then turned into a joining process. The chemistry generalizes: swap the iron oxide for another metal oxide and you get that metal molten. Copper thermite (copper(II) oxide + aluminum) is the basis of exothermic welding for electrical grounding — the CADWELD process fuses copper grounding conductors and rail bonds in a graphite mold, forming a permanent, low-resistance fused metallurgical bond that will not loosen or corrode like a bolted lug. Thermite has also long been used to weld or repair very large sections that are impractical to move — ship propeller shafts, thick pipe, and heavy castings — and, notoriously, as a military incendiary (the WWII-era thermite bomb), because a burning thermite puddle at 2500 °C will melt through steel plate and cannot be smothered with water.

Against its welding cousins, thermite is the outlier that carries its heat as chemistry rather than electricity. Arc welding and electron-beam welding deliver energy electrically — as an arc plasma and as a focused electron beam; friction-stir welding joins in the solid state, and flash-butt melts only a thin interface layer that the forge upset then squeezes out — both without added filler. Thermite alone brings its own molten filler, poured like a foundry casting, which is exactly what lets it work at the end of a rail line with nothing but a crucible, a mold, and a bag of gray powder — and exactly why its quality lives or dies on preheat, timing, and a dry mold.

Two ways to weld continuous rail: portable aluminothermic (thermite) versus electric flash-butt.

Attribute	Thermite (aluminothermic)	Flash-butt welding
Heat source	Chemical (Fe₂O₃ + Al) — self-supplied	Electric resistance / arcing across the joint
Filler metal	Yes — cast weld, molten iron poured in	None — interface melted, then expelled by the forge upset; no filler added
Power needed	None (portable, works trackside)	Heavy machine, hundreds of kW of supply, hydraulic clamps
Joint mechanism	Fusion casting into a refractory mold	Flashing + forge pressure (hundreds of kN) to upset the joint
Typical use	Field joints, turnouts, repairs, awkward spots	Plant or mobile welding train, long tangent runs
Defect rate	Higher — porosity, inclusions, cold welds possible	Lower — more consistent, fewer voids