Thermal treatment and evolving microstructure

Heating is a route through possible structures

A steel becomes much harder to deform permanently after quenching, yet a trial that needs both strength and tensile ductility rejects that state. A later tempering step lowers its measured proof stress but allows it to pass both requirements. Why can an additional treatment improve the useful result even while one property decreases?

A treatment can dissolve particles, permit precipitation, change phases or modify a deformed grain structure. Quenching can suppress some transformations; subsequent ageing can permit others. The useful sequence depends on the alloy system and starting state.

Follow one representative batch of AISI 1045 medium-carbon steel, containing 0.45% carbon by mass. AISI 1045 is the steel grade designation. Its room-temperature starting structure here contains ferrite and cementite. Ferrite is an iron-rich solid phase with a body-centred cubic crystal structure; cementite is the carbide Fe₃C. Some of these two phases occur in alternating thin layers within pearlite colonies. Pearlite is a two-phase microstructural constituent, not a third phase.

The records below are original, invented teaching data for this material context. The temperatures and times describe a stipulated local history; separate observations supply its structural states. They are not measured AISI 1045 properties, an industrial treatment specification or a kinetic model. The useful reasoning is how the starting state, path, mechanisms and final evidence fit together.

Heating changes where the carbon can reside

Austenitising means producing austenite, the iron-rich phase with a face-centred cubic structure, by heating into the appropriate range. For this example, a local specimen temperature of 850 °C is maintained for 20 min. At the end of that hold, the intended region is stipulated to be fully austenitic, with the initial cementite dissolved and carbon distributed through the austenite. At this high-temperature state, austenite can accommodate the batch’s carbon in solid solution, unlike the low-carbon ferrite of the starting structure. “Fully austenitic” is an independently established condition here, not a conclusion from 850 °C alone.

As austenite forms and grows, the original ferrite–cementite arrangement is consumed. Carbon from dissolving cementite redistributes through the growing austenite by diffusion. Dissolution does not remove carbon from the specimen: under this example’s closed-carbon condition, the overall 0.45% mass fraction is unchanged while its location and the crystal structure change.

Sufficient temperature and time are needed for the intended dissolution and redistribution. Heating too briefly can leave undissolved carbides or local composition differences; heating excessively can also change grain size or damage the surface. Composition, starting particle dimensions and the actual local temperature history matter. In a real open atmosphere, oxidation and loss of carbon near a surface can violate the example’s closed-carbon assumption.

Quenching can create a new phase

Quenching is rapid cooling relative to the transformations one wants to limit. In this steel example, it restricts the time available for carbon redistribution and formation of ferrite–carbide products such as pearlite. But it does not prevent every structural change. As the austenite cools far enough, martensite can form by a coordinated change of the crystal: neighbouring atoms move relative to one another without the long-range diffusion needed to redistribute carbon into separate ferrite and cementite regions.

The newly formed martensite initially inherits carbon from the parent austenite rather than redistributing it into separate ferrite and carbide regions. Carbon in interstitial sites distorts the lattice; carbon-containing martensite can have a body-centred tetragonal structure. The transformation also produces a highly defected arrangement, often with fine laths in this kind of steel. Distortion and obstacles to dislocation motion contribute to its large resistance to permanent deformation.

This is a distinct route from the preceding precipitation example: a suitable aluminium alloy can retain a supersaturated matrix phase on quenching, whereas the steel considered here forms martensite from austenite. Neither “rapid cooling” nor “metal” specifies a universal product. The amount of martensite depends on the alloy and path, and some austenite can remain at room temperature, called retained austenite. Carbon can also begin to rearrange during cooling; “diffusionless transformation” does not mean all later carbon motion is forbidden.

The martensitic shape change and volume change, together with unequal thermal contraction in different regions, can generate residual stresses and distortion. Residual stress remains after external loads are removed because adjacent regions constrain one another. A quenched piece can therefore be internally stressed even when nothing is pulling on it. An untempered condition may also crack or have inadequate fracture resistance, so high proof stress alone is not a complete acceptance test.

Tempering develops the quenched state

Tempering reheats a martensitic steel below the austenite-formation range for the chosen alloy and conditions. This example uses a local 450 °C hold for 60 min, followed by cooling to the common test temperature. The temperature is stipulated to be below austenite formation here. This is not a second austenitising operation and does not simply reverse the first quench.

Tempering permits further carbon redistribution to defects and into carbide particles. With appropriate time and temperature, carbides evolve and the dislocation arrangement can recover, meaning that some defects rearrange or are removed. The matrix becomes less carbon-supersaturated and its internal strains and obstacles change. The final tempered martensitic microstructure is a ferritic matrix containing carbides and an inherited, evolving morphology; “tempered martensite” is not the name of one newly created equilibrium phase.

In the supplied record, tempering reduces proof stress and increases elongation after fracture. That result is consistent with a changed obstacle and defect state, but it is not a universal monotonic law. Alloy-carbide precipitation can produce secondary hardening in some steels; some tempering conditions can reduce toughness. Carbide size and distribution, retained austenite and grain-boundary state all matter. A measured elongation is not itself a toughness measurement.

The first graph follows the local specimen temperature through the whole cycle. S is the starting state; H marks the end of the austenitising hold; Q is sampled after the quench, before reheating; T is sampled after tempering and final cooling. Structural labels are supplied observations and interpretations of the specified case. The temperature line alone does not measure them.

A surface record cannot stand in for the centre

The austenitising hold lasts 25 − 5 = 20 min; the tempering hold lasts 95 − 35 = 60 min. The 850-to-25 °C quench lasts 0.20 min = 12 s. Its average cooling-rate magnitude is 825/12 = 68.75 °C/s. This is a temperature difference per time, not a temperature divided by time measured from an arbitrary zero. A temperature change of 1 °C equals a change of 1 K, so the magnitude is also 68.75 K/s.

An interval average does not establish an instantaneous rate throughout cooling. The supplied near-surface path takes 5 s to cross from 600 to 300 °C: its average over that chosen window is 300/5 = 60 °C/s. These temperatures are a comparison window, not universal transformation-start or finish boundaries. The different 68.75 and 60 values describe different intervals of the same path.

Now consider a second, larger workpiece whose near-surface region and centre both begin fully austenitic at 850 °C. Their original idealised cooling records are plotted below. The near-surface trace is the expanded quench used above. The centre takes 25 s to cross the same 600-to-300 °C window, giving 12 °C/s. It spends five times as long in that window, even though both locations eventually reach 25 °C.

This longer centre passage gives diffusion-involving transformations a different opportunity. It does not by itself specify a phase fraction or a critical cooling rate. For the stipulated example, separate post-quench examinations find predominantly martensitic structure near the surface and a mixture including ferrite–carbide products at the centre. Those observations complete the inference; dividing 60 by 12 would not predict five times as much martensite.

Later, both regions are stipulated to reach the same 450 °C temperature for 60 min and receive the same final cooling. That tempering step can alter their existing structures, but it cannot be assumed to turn the centre’s already formed ferrite–carbide products into martensite without a new austenitising-and-cooling route. Matching the later hold therefore does not erase a different earlier path. A single surface temperature sensor and a furnace setpoint are insufficient evidence for a uniform treatment.

Test the useful final condition

The next table is a separate original, invented final-property record from matched small coupons of the same batch. S, Q and T correspond to the first cycle’s stated states; the coupons are not the nonuniform larger workpiece. Their preparation and observed structures are supplied independently of the property numbers. Each state is tested at 25 °C in the same direction, at the same tensile rate and with the same specimen geometry. Destructive tensile tests use separate matched coupons.

The 0.2% proof stress, σ₀.₂, is the engineering stress where the tensile curve meets a line parallel to its initial elastic slope and offset by strain 0.002. Engineering stress uses the original cross-sectional area. A value in MPa is a resistance to this specified permanent-deformation endpoint, not Young’s modulus, hardness or fracture toughness. One MPa is 10⁶ Pa, or one N/mm².

Elongation after fracture is 100(L_f − L₀)/L₀, where L₀ is the original gauge length and L_f is the gauge length after fracture with the pieces fitted together. All coupons here use L₀ = 50 mm and the same convention. T has L_f = 56 mm, giving 12%; Q has L_f = 51.5 mm, giving 3%. These gauge-dependent endpoints describe tensile ductility, not an allowable component operating strain.

A trial screen asks for σ₀.₂ of at least 750 MPa and elongation of at least 10%. S passes the elongation requirement but not the strength requirement; Q passes strength but not elongation; T passes both. Tempering changes 1050 MPa to 800 MPa, a drop of 250 MPa or about 23.8% relative to Q. Elongation rises by 9 percentage points. Calling the largest proof stress the “best treatment” would miss the stated purpose.

These exact teaching numbers carry no claim about measurement uncertainty, batch variability or service reliability. In real evidence, keep the sampling location, preparation, complete local path, surface condition, structure-identification method and property-test conditions. Matching hardness alone does not establish the same carbide distribution, residual stress, fracture resistance or fatigue behaviour. Test the properties and locations relevant to the actual decision.

Avoid universal recipes

A quench may introduce residual stress as well as preserve a desired structure. A later tempering or ageing step has a specific purpose. Do not transfer temperatures or times between alloys merely because both are called steel or aluminium alloy; confirm material-specific treatment and resulting properties.

Annealing, tempering and precipitation ageing name different purposes and starting states, not interchangeable hot holds. In an anneal of a cold-worked metal, recovery can rearrange or remove dislocations, while recrystallisation replaces highly deformed regions with new grains having a much lower dislocation density. Such a change can occur without changing the crystal phase. “Annealing” covers several treatments; state which structural change is intended.

In the quench-and-temper steel route, martensite is the prepared state and tempering changes its carbon, carbide and defect arrangement below austenite formation. In precipitation ageing of a suitable solution-treated and quenched alloy, a supersaturated matrix develops solute-rich clusters and precipitates; no martensitic precursor is required. These mechanisms can share features such as diffusion or precipitation without making the treatment names synonyms.

To interpret a proposed schedule, first ask what material and starting structure it acts on, which phases and rearrangements are possible, and what each stage is meant to preserve or change. Then check the actual temperature path within the specimen and the completed structure and properties. The complete evidence supports a treatment for its tested conditions; the family name or the hottest temperature does not.

Worked examples

Try the reasoning

1. Does every quench produce the same phase?

   Show the answer and reasoning

   No.

   Available transformations depend on the alloy and conditions. In this steel, quenching forms martensite while limiting other products. In a suitable solution-treated aluminium alloy, quenching can instead retain a supersaturated matrix. Neither result follows from the word quench alone.

2. Why test after the complete treatment?

   Show the answer and reasoning

   Intermediate states may differ from the final state.

   The property must describe the state used. Q and T have different proof stress and tensile ductility even though T passed through Q. A test of Q cannot be relabelled as a test of the final tempered condition.

3. A third local record first reaches 600 °C at 6 s and 300 °C at 21 s. Calculate the cooling-rate magnitude over that window and compare its duration with the plotted near-surface and centre records. Can you assign a martensite fraction?

   Show the answer and reasoning

   The duration is 15 s and the average magnitude is 20 °C/s. That duration is three times the near-surface 5 s and 0.60 times the centre 25 s. No martensite fraction is determined.

   The same 300 °C change divided by 15 s gives the interval average. Neither average cooling rate nor a time ratio is a phase-fraction law. Composition, initial austenite and the full path, plus transformation evidence or a justified model, are needed.

4. A centre region forms ferrite and pearlite during cooling. A report claims that its later subcritical tempering hold must produce the same martensitic structure as a region that was already martensitic. What step is missing?

   Show the answer and reasoning

   The centre has not been re-austenitised and cooled along a suitable martensite-forming path.

   Tempering acts on the state that arrives at the hold. In the stipulated subcritical conditions, it does not create the missing austenite precursor. Equal later temperature and time cannot be assumed to erase the earlier diffusional products.

5. After the larger workpiece’s complete treatment, matched local coupons give near-surface proof stress 800 MPa and elongation 12%, but centre proof stress 620 MPa and elongation 18%. Both locations must meet 750 MPa and 10%. Is the workpiece demonstrated to pass? May the two proof stresses simply be averaged to qualify the centre?

   Show the answer and reasoning

   No. The centre fails the proof-stress requirement even though it passes elongation. An arithmetic average cannot qualify that local requirement.

   The centre deficit is 750 − 620 = 130 MPa. The near-surface excess is 800 − 750 = 50 MPa and does not cancel it. These are separate local material results; no component load-sharing model or constitutive averaging rule is supplied. The differing paths and structures provide a reason to inspect both locations.

6. A new matched 50 mm gauge coupon, tested after the complete cycle, has fitted final length 55.5 mm and proof stress 780 MPa. Does it meet the 750 MPa / 10% screen? Does that mean it can safely operate at 11% strain?

   Show the answer and reasoning

   It has 11% elongation after fracture and meets both screen values. No safe operating strain is established.

   100 × (55.5 − 50)/50 = 11%. Both 780 ≥ 750 and 11 ≥ 10 hold. A destructive after-fracture gauge endpoint is not a recoverable strain, a design strain limit or a toughness result.

7. Two tempered specimens have the same measured hardness, but one reached its tempering temperature rapidly and the other slowly. Must their carbide population, retained austenite, residual stress and fracture resistance all be the same?

   Show the answer and reasoning

   No. One matched property does not identify the full structural state or the other properties.

   Different paths can activate rearrangements in different orders and leave different spatial distributions. Compare the actual local paths, structure observations and the particular final properties required; the equality of hardness supplies no unique inverse reconstruction.

8. In the stipulated closed-carbon example, cementite dissolves during austenitising and carbides later appear during tempering. Has the steel necessarily gained and then lost carbon overall?

   Show the answer and reasoning

   No. Carbon can redistribute between the matrix and carbide regions while the total mass fraction remains 0.45%.

   Phase and composition distributions change without changing the bulk carbon inventory under the stated closed-carbon condition. A real surface can exchange carbon with an atmosphere, so that condition must be checked rather than assumed.

What to carry forward

• Thermal treatment is material-specific structural control.
• Evaluate the complete schedule and its side effects.
• Quenching can suppress carbon-redistributing routes while a diffusionless martensitic transformation proceeds.
• Tempering, annealing and precipitation ageing act on distinct prepared states and need their own purpose and evidence.
• Local thermal histories and independently identified structures connect a schedule to the final condition; one sensor or one hardness value cannot supply the whole state.