Thermit Welding by the Numbers
Chapter 8 of 10 · 10 min read
Alloying the weld metal
See what plain thermit steel is made of, how punchings, manganese, nickel and modern alloy additions change it, and why a weld's composition must suit the steel it joins.
Plain thermit makes nearly pure iron. That suited some of the early uses, but a weld that is softer or weaker than the parts it joins becomes the weak point, and on a rail it wears into a dip under the wheels. From the beginning, therefore, thermit charges for welding steel carried additions: steel punchings to add metal, manganese to deoxidise and strengthen, nickel for strength and toughness, and in modern rail portions, carbon and alloying elements chosen to match the rail. The weld metal’s final composition also depends on how much of the parts melts into it.
This chapter looks at the composition and strength of plain thermit steel in 1921, the additions used then and now, the effect of dilution from the parts, and what the crusher-frame weld needs.
Plain thermit steel
Viall gave the average analysis of thermit steel in 1921 as about 0.05 to 0.10% carbon, 0.08 to 0.10% manganese, 0.09 to 0.20% silicon, 0.03 to 0.04% sulphur, 0.04 to 0.05% phosphorus and 0.07 to 0.18% aluminium. That is a very low carbon steel, close to soft iron. He gave its average tensile strength as about 61,000 pounds per square inch, about 420 MPa, with an elastic limit of a little over half that, about 34,000 pounds per square inch or 234 MPa, and noted that the strength could be varied by adding other elements.
Hart made a related point in 1910: thermit steel has a low carbon content, so rapid cooling will not harden it much. That makes plain thermit steel forgiving, but it also limits its strength. A low carbon weld cannot be hardened or strengthened much by heat treatment either, so whatever strength the weld needs has to be built in through its composition. Hart summarised the tests he had collected as showing that the ultimate strength of thermit steel in practice could be estimated at over 30 tons per square inch, which is about 414 MPa if the American short ton is meant or about 463 MPa for the long ton.
The varieties of 1921
Viall listed three commercial varieties of thermit:
- plain thermit, aluminium and iron oxide alone, used where the thermit was mainly a heat source, such as joining pipes and remaking broken necks on rolling-mill rolls and pinions;
- railroad thermit, plain thermit with additions of nickel, about 1% manganese and mild steel punchings, used for steel welds;
- cast-iron thermit, plain thermit with about 3% ferrosilicon and 20% mild steel punchings, used for welding cast iron.
The additions in railroad thermit raised the strength of the steel. The ferrosilicon in cast-iron thermit added silicon, which, as Weldability by the Numbers explains, helps keep carbon in grey cast iron as graphite rather than as hard, brittle carbide.
Punchings and manganese in the early repairs
Hart’s recorded repairs show how the additions were used. The stern frame repair used 416 pounds of thermit with 35 pounds of mild steel punchings and 8 pounds of metallic manganese. The locomotive frame test used 75 pounds of thermit with 12 pounds of punchings and 1½ pounds of manganese.
This is an illustration of the arithmetic, using Hart’s yield of half. The stern frame charge would produce about 416 × 0.5 = 208 pounds of thermit steel plus 35 pounds of melted punchings, about 243 pounds in all, so the punchings were about 14% of the steel and the manganese about 8 ÷ 243 ≈ 3.3% of it before any losses. The locomotive frame charge would produce about 37.5 + 12 = 49.5 pounds of steel, with punchings about 24% of it and manganese about 1.5 ÷ 49.5 ≈ 3.0%.
| Stern frame | Locomotive frame test | |
|---|---|---|
| Thermit | 416 lb (189 kg) | 75 lb (34 kg) |
| Mild steel punchings | 35 lb (16 kg) | 12 lb (5.4 kg) |
| Metallic manganese | 8 lb (3.6 kg) | 1.5 lb (0.7 kg) |
| Expected steel, thermit plus punchings | about 243 lb | about 49.5 lb |
| Punchings as share of steel | about 14% | about 24% |
| Manganese added, as share of steel | about 3.3% | about 3.0% |
Manganese is a strong deoxidiser, so part of it would have been lost to the slag rather than ending up in the steel. Even so, the additions show the intent: to turn soft thermit iron into a stronger, sounder steel, and to add metal from the punchings so that less thermit was needed.
Nickel thermit
Hart described nickel thermit, a mixture of nickel oxide and aluminium that reacts as 3NiO + 2Al → 3Ni + Al₂O₃, releasing metallic nickel. A can of it introduced into the molten thermit iron, or fired alongside it, added nickel to the weld. He wrote that one per cent of nickel increased the strength of ordinary iron by about a third, and that nickel raised the tensile strength without reducing the elastic limit. Nickel also improves toughness, which is why it appears in many modern weldable steels and welding consumables.
Dilution from the parts
The weld metal is not only thermit steel. The faces of the parts melt into it, as Chapter 6 intended, and they bring their own composition. The final weld metal is a mixture, and its composition can be estimated from the shares of each contributor, the same mixing rule applied to fusion welds in Weldability by the Numbers.
This is an illustration. Suppose the crusher frame’s cast steel contains about 0.30% carbon, the thermit steel about 0.08%, and the melted face layer of about 1.41 kg (Chapter 6) mixes into the 25.5 kg of weld metal in the gap and collar. The face metal is then about 1.41 ÷ (25.5 + 1.41) ≈ 5% of the weld metal, and the weld’s carbon content is roughly 0.95 × 0.08 + 0.05 × 0.30 ≈ 0.09%. In reality, mixing is incomplete: the weld metal next to the faces has more parent metal in it than the middle of the gap. But the estimate makes the point: with plain thermit, the weld metal would remain a soft, low carbon steel, far weaker than a cast steel frame designed for strength.
- D
- share of weld metal from melted faces, about 1.41 ÷ 26.9 ≈ 0.05
- C_thermit
- plain thermit steel, about 0.08%
- C_parent
- cast steel frame, about 0.30% (assumed)
- C_weld ≈ 0.95 × 0.08 + 0.05 × 0.30
- C_weld ≈ 0.09%
Slag and gas: what the additions cannot fix
Additions control the composition of clean weld metal. They cannot make up for slag or gas trapped in it. Hart identified tapping too early as a frequent reason thermit welds failed, and put the reason bluntly: no steel is strong if it is permeated with slag. If the crucible is tapped before the reaction is complete and the slag has separated, slag particles are carried into the mould with the iron, and the riser cannot float them all out before the weld freezes.
The timing has to be right in both directions. Too early, and slag comes through with the iron. Too late, and the iron loses superheat, with the lack of fusion described in Chapter 2. Modern self-tapping crucibles, which open automatically when the reaction has heated a closing plug through, take much of the judgement out of the timing, and procedures set the conditions for those that are tapped by hand.
Gas is the other contaminant. Moisture in the charge, the mould or the luting releases hydrogen and steam into the iron, which come out as it freezes, leaving porosity. The same dry conditions that prevent the eruptions in Chapter 3 prevent most porosity. The aluminium in thermit also acts as a deoxidiser, which is part of why a small surplus is normal (Chapter 1): it ties up oxygen in the iron that would otherwise form gas or oxide inclusions as the weld freezes.
Matching the weld to the parts
Modern thermit portions for welding steel include alloying additions chosen so that the weld metal suits the steel being joined. TWI notes that rail thermite mixtures include pellets or rods of a high-carbon alloying metal, which melt into the weld metal during the reaction. Rail steels range from standard carbon grades to harder, heat-treated head-hardened grades, and suppliers offer portions matched to each, so that the weld’s hardness and strength are close to the rail’s.
Matching matters in both directions. A weld much softer than the rail wears faster under wheels and forms a dip, which increases the impact loads from every passing wheel. A weld much harder or less ductile than the rail can be brittle and crack. The supplier’s portion, qualified with the rail grade under the infrastructure manager’s requirements, sets the balance.
For the crusher frame, the engineer asked the contractor three questions about the portion:
- Which cast steel grade, strength and carbon equivalent is it designed to join, and does that match the frame’s material?
- What strength and toughness does its weld metal achieve, and on what evidence?
- Does the parent steel’s composition call for controlled cooling or post-weld heat treatment (Chapter 6)?
What evidence looks like
The answers to the engineer’s questions come from test evidence, not from the label on the portion. For rail portions, the supplier qualifies each portion and procedure for each rail grade under the requirements the infrastructure manager applies, such as those in AS 1085.20, with tests on welds made under controlled conditions. For a one-off heavy repair, the contractor may have test records from similar welds, or may need to make a trial weld on a representative block of similar steel, cut it up and test it. Typical evidence includes a chemical analysis of the weld metal, tensile tests across the weld, a hardness survey from parent metal through the heat-affected zone into the weld metal, a macro-etched section showing the extent of fusion, and sometimes bend or impact tests. Weld Repairs That Hold explains how procedure qualification works for repair welding generally. Chapter 10 of this book looks at what can be tested on the actual weld once it is made.
After-heating
Hart advised that a thermit weld in a locomotive frame be given several hours of annealing by whatever means were available, if the joint was to be as tough as the frame. Annealing was less necessary for a thermit joint than for oxy-acetylene and other welds, because of thermit steel’s low carbon, but chilled steel was never as tough as annealed steel, and tests suggested that after-heating gave a more even-grained, tougher joint. A lesser version of the same benefit could be had by leaving the mould in place until the joint had cooled, which could take several hours on large parts.
The modern equivalent is the controlled cooling or post-weld heat treatment set by the procedure for the materials involved. For a heavy cast steel frame, leaving the mould on and insulating the joint so that it cools slowly is the minimum the engineer would expect; whether a full heat treatment is needed depends on the material, which is why the frame’s grade matters.
A failure: the soft weld that dipped
This is an illustration. The quarry siding’s rail welds were inspected as part of investigating the failed weld in Chapter 2. One of the other welds showed a pronounced dip in the running surface, with the rail head worn several millimetres lower across the weld than on either side, and the sleepers beside it working loose.
The rail contractor’s records showed that the siding had been relaid some years earlier with heavier, harder rail, but that one weld had been made with a portion intended for the previous, softer rail grade. The weld metal was softer than the new rail, so it wore faster. The dip grew, the wheels struck it harder, and the impact loads loosened the fastenings and sleepers around it.
The weld was cut out and replaced with a portion matched to the rail grade, the fastenings and sleepers were restored, and the quarry added a requirement to its rail maintenance contract that weld records state the rail grade and the portion type used. The incident also showed the engineer that the siding’s records needed to be audited, which Chapter 9 describes.
Exercise: more parent metal
Suppose a poorly controlled preheat and a very hot pour melt three times as much of the crusher frame’s faces, about 4.2 kg, into the same 25.5 kg of weld metal. Using the same carbon contents, estimate the dilution and the weld’s carbon content. Then estimate the manganese content of the weld metal, if the thermit steel carries 0.10% manganese and the cast steel 0.70%.
The dilution is 4.2 ÷ (25.5 + 4.2) ≈ 0.14. The carbon content is about 0.86 × 0.08 + 0.14 × 0.30 ≈ 0.11%, and the manganese content about 0.86 × 0.10 + 0.14 × 0.70 ≈ 0.18%.
Even with three times the dilution, plain thermit steel would leave the weld metal far leaner in carbon and manganese than the cast steel frame. Dilution cannot be relied on to strengthen a weld; the strength has to come from a portion designed with the right additions, confirmed by the supplier’s test evidence.
