Skip to content
LumenaryAcademy
Processes 75 min

Processing and heat treatment

How making a part sets what is inside it — casting and solidification, forming and recrystallisation, annealing, quenching and tempering of steel, sintering of ceramics, and the tempers of aluminium alloys.

Code tested · 5AI-draftedAwaiting review by a materials scientist

Your guides: Chloe · The Materials Scientist, Margaret · The Tester, Gideon · The Failure Analyst

MotivationH

One alloy, many materials

Two bars of the aluminium alloy 6061, used for bicycle frames, boat fittings and machine parts, have the same composition: about 1 % magnesium and 0.6 % silicon in aluminium. In the annealed condition, called O, one bar begins to yield at about 55 MPa and can be bent easily. The other has been solution treated, quenched and aged, the T6 condition, and yields at about 275 MPa, five times higher. Nothing was added; the difference is the arrangement of the same atoms, set by what was done to the metal. Steel shows the same thing more dramatically: one carbon steel can be soft enough to draw into wire or hard enough to scratch glass, depending on its heat treatment.

Materials scientists summarise this as a chain: processing sets structure (phases, grain size, defects, precipitates), structure sets properties, and properties set performance in service. The previous three lessons supplied the tools: diffusion to say how fast atoms move, phase diagrams to say where they are heading, and phase transformations to say how they get there. This lesson applies them to the main ways materials are made and treated: casting, forming, heat treatment of steel and of aluminium, and sintering of ceramics. Lumenary ME's manufacturing processes describes the same processes from the machine and the designer's side; this lesson looks inside the material.

Note

Everything in this lesson happens in foundries, forges and heat-treatment plants, built around serious hazards: molten metal at over 600 °C (and far hotter for steel), which can explode violently on contact with water or damp tools; furnaces and their atmospheres; quench oils that can catch fire; and heavy, hot loads. Workers there wear heat-resistant protective clothing and face shields, and follow written procedures. None of it is a home activity. The simulations below are the safe way to experiment.

ComputationC

Casting and solidification

Casting, pouring liquid metal into a mould shaped like the part, is one of the oldest ways of shaping metal and still among the most important. Lost-wax (investment) casting, in which a wax model is coated in a ceramic or clay shell, melted out and replaced by metal, produced the Chola bronzes of South India and the brass plaques of the Kingdom of Benin, and it makes the turbine blades of jet engines today. Sand casting makes engine blocks and pump housings; pressure die casting makes millions of thin aluminium and zinc parts; and most of the world's steel is continuously cast into slabs and billets that are then rolled.

When metal freezes against a cold mould wall, it forms a characteristic grain structure. A thin chill zone of small, randomly oriented grains forms where the wall cools the metal fastest. Some of these grains are oriented so that they grow fastest along the direction of heat flow; they outgrow their neighbours and form long columnar grains reaching inwards. In the centre, where the remaining liquid is barely undercooled, crystals broken off from the columnar front or formed on particles grow as roughly equal-sized equiaxed grains. Within each grain the solid grows as dendrites, tree-like crystals whose arms follow particular crystal directions, because a bump on a growing interface pokes into liquid that is more undercooled than the liquid at the flat front, so it grows faster than its surroundings and the flat front breaks up into branches.

Castings cool too fast for the solid to stay in equilibrium with the liquid. Solute atoms diffuse quickly in a liquid but very slowly in a solid, so the solid that froze first keeps its composition while the liquid around it grows richer. The Scheil equation describes the limit of no diffusion in the solid and complete mixing in the liquid. If kk is the partition coefficient, the ratio of solute in the solid to that in the liquid at the interface, then when a fraction fsf_\text{s} has frozen the liquid has the composition CL=C0(1−fs)k−1C_\text{L} = C_0(1 - f_\text{s})^{k - 1}, and the solid freezing at that moment has kCLkC_\text{L}. The block compares this with equilibrium for an aluminium alloy with 4 wt% copper.

python
# Solidification of an Al - 4 wt% Cu casting alloy: equilibrium (lever rule) against
# the Scheil model (no diffusion in the solid, complete mixing in the liquid).
k = 0.17          # partition coefficient C_solid / C_liquid at the interface, about 0.17 for Cu in Al
C0 = 4.0          # wt% Cu in the alloy
C_E = 33.2        # wt% Cu in the liquid at the eutectic temperature, 548 C
C_max = 5.65      # wt% Cu: the most the aluminium-rich solid can dissolve, at 548 C

print("                  wt% Cu in the solid             wt% Cu in the liquid")
print("fraction solid   equilibrium   Scheil (new layer)        Scheil")
for fs in (0.0, 0.2, 0.4, 0.6, 0.8, 0.9):
    CL_scheil = C0 * (1 - fs) ** (k - 1)
    CS_eq = k * C0 / (1 - fs * (1 - k))          # lever rule with straight-line liquidus and solidus
    print(f"{fs:14.1f}   {CS_eq:11.2f}   {k * CL_scheil:18.2f}   {CL_scheil:11.2f}")

# Under Scheil, the last liquid reaches the eutectic composition and freezes as eutectic.
fL_eutectic = (C_E / C0) ** (1 / (k - 1))
print(f"\nScheil: the liquid reaches {C_E} wt% Cu with {100 * fL_eutectic:.1f} % of the casting still liquid,")
print("        which freezes as Al + Al2Cu eutectic between the grains")
print(f"lever rule: {C0} wt% Cu is below the {C_max} wt% solubility limit, so no eutectic at all")

# Coring: copper content across one cell, from its centre (first to freeze) to its edge.
print("\ncopper across a cell, centre to edge (Scheil):")
for fs in (0.0, 0.25, 0.5, 0.75, 0.9, 1 - fL_eutectic):
    print(f"  {fs:5.2f}  {k * C0 * (1 - fs) ** (k - 1):5.2f} wt%  " + "#" * round(4 * k * C0 * (1 - fs) ** (k - 1)))
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

At first the two models agree: the first solid has only 0.68 wt% copper, because kk is small. But under Scheil conditions the liquid grows steadily richer, reaching 27 wt% copper by 90 % solid, and the last 7.8 % of the casting freezes as eutectic, Al + Al₂Cu, between the grains, even though the phase diagram says a 4 wt% alloy should be a single phase. The cores of the grains are copper-poor and their edges copper-rich, which is coring or microsegregation. It is why castings are often homogenised, held at a high temperature for hours to let diffusion even out the composition, before they are worked or heat treated. The same equation describes how dopants spread along a silicon crystal pulled from its melt (semiconductor materials).

Two other casting defects come from shrinkage: most metals contract by a few per cent as they freeze, so unfed regions develop shrinkage porosity, and gases dissolved in the liquid come out as gas porosity. Grain refiners (phase transformations) give fine equiaxed grains that feed and resist cracking better. At the other extreme, the single-crystal nickel-superalloy turbine blades of jet engines are cast by withdrawing the mould slowly from a furnace so that heat flows in one direction, and a spiral "grain selector" at the root lets only one grain through: a blade with no grain boundaries at all, which resists creep at temperatures close to its melting point.

ComputationC

Forming, recovery and recrystallisation

Most metal parts are shaped in the solid by rolling, forging, extrusion or drawing. What happens inside depends on the temperature relative to the melting point, measured by the homologous temperature T/TmT/T_\text{m}, with both in kelvin.

  • Cold working, well below about 0.4 Tm0.4\,T_\text{m}, multiplies dislocations, which tangle and block each other. The metal becomes stronger and harder but less ductile (work hardening), and stores energy in its dislocations. Wire drawing and cold-rolled sheet use this to make strong products.
  • Hot working, above about 0.50.5 to 0.6 Tm0.6\,T_\text{m}, lets the metal restore itself as fast as it is deformed, so it stays soft and very large shape changes are possible. Steel slabs are hot-rolled at around 900 to 1200 °C.

The boundary depends on the metal: lead, with Tm=600T_\text{m} = 600 K, is "hot worked" at room temperature, while tungsten worked at 1000 °C is still being cold worked.

When a cold-worked metal is heated (annealed), three processes follow in turn. In recovery, dislocations rearrange into lower-energy walls and some annihilate; strength falls a little and internal stresses relax. In recrystallisation, new strain-free grains nucleate, usually at old grain boundaries, and grow, consuming the deformed structure, driven by its stored dislocation energy; the metal becomes soft and ductile again, and its kinetics follow the Avrami equation of the previous lesson. Then comes grain growth: the new grains coarsen to reduce the total area of grain boundary, whose energy is the only driving force left.

Grain growth is easy to simulate with the Potts model introduced for this purpose by Anderson, Srolovitz and co-workers in 1984. Each site of a lattice carries a grain orientation; the energy is the number of neighbouring pairs with different orientations, a stand-in for boundary area; and a Monte Carlo step lets each site try to adopt a neighbour's orientation, accepted when it does not raise the energy and occasionally when it does. The block runs it on a 96 × 96 lattice and measures the grain size the way a metallographer does, by the mean linear intercept: the length of test lines divided by the number of boundaries they cross.

python
import numpy as np

rng = np.random.default_rng(11)
L, Q = 96, 64                 # lattice of L x L sites; Q possible grain orientations
s = rng.integers(0, Q, size=(L, L))
SHIFTS = [(-1, -1), (-1, 0), (-1, 1), (0, -1), (0, 1), (1, -1), (1, 0), (1, 1)]   # 8 neighbours

def sweep(spins, kT=0.3):
    """One Monte Carlo step per site.  Each site proposes the orientation of a random
    neighbour and accepts it if the number of unlike neighbours (the boundary energy)
    does not rise, or with probability exp(-dE/kT) if it does.  The four sublattices
    of every other row and column are updated in turn: no two of their sites are neighbours."""
    n = L // 2
    for a in (0, 1):
        for b in (0, 1):
            P = np.pad(spins, 1, mode="wrap")              # periodic edges
            nbrs = np.stack([P[1 + a + di:1 + a + di + L:2, 1 + b + dj:1 + b + dj + L:2] for di, dj in SHIFTS])
            here = spins[a::2, b::2]
            proposal = np.take_along_axis(nbrs, rng.integers(0, 8, size=(1, n, n)), axis=0)[0]
            dE = (nbrs != proposal).sum(axis=0) - (nbrs != here).sum(axis=0)
            accept = (dE <= 0) | (rng.random((n, n)) < np.exp(-np.maximum(dE, 0) / kT))
            here[accept] = proposal[accept]                # 'here' is a view into spins

def mean_intercept(spins):
    """Mean linear intercept in lattice spacings: line length per boundary crossing,
    along rows and columns, as in the metallographer's line-intercept method."""
    crossings = np.sum(spins != np.roll(spins, 1, axis=0)) + np.sum(spins != np.roll(spins, 1, axis=1))
    return 2 * spins.size / crossings

print(" MC steps   mean intercept d   d^2 / steps")
steps, sizes, done = [], [], 0
for target in (10, 25, 50, 100, 200, 300, 400):
    while done < target:
        sweep(s)
        done += 1
    if done == 50:
        early = s.copy()
    d = mean_intercept(s)
    steps.append(done)
    sizes.append(d)
    print(f"{done:9d}   {d:16.2f}   {d * d / done:11.2f}")

slope = np.polyfit(np.log(steps[2:]), np.log(sizes[2:]), 1)[0]
print(f"\ngrowth exponent from 50 steps on: d ~ t^{slope:.2f}  (parabolic growth: 0.5)")
print("\nthe same 40 x 40 patch after 50 and after 400 steps (every other row; one letter per orientation):")
letters = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789#$%&*+=?"
for row in range(28, 68, 2):
    print("".join(letters[v] for v in early[row, 28:68]) + "    " + "".join(letters[v] for v in s[row, 28:68]))
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The mean intercept grows from 3.8 to 20.6 lattice spacings over 400 steps, and from 50 steps on it follows d∝t0.51d \propto t^{0.51}, close to the parabolic law d2−d02=ktd^2 - d_0^2 = kt expected when boundaries move at a speed proportional to their curvature. The two snapshots show the mechanism: small grains with few sides shrink and vanish, and their neighbours take over their space.

Grain size matters because fine grains make a metal both stronger and tougher (the Hall–Petch relation of elasticity, plasticity and strengthening). Steelmakers therefore fight grain growth. Microalloyed steels contain a few hundredths of a per cent of niobium, titanium or vanadium, which form fine carbonitride particles that pin grain boundaries (Zener pinning); combined with rolling schedules that recrystallise the austenite repeatedly and then cool it under control, this thermomechanical processing gives the fine-grained, strong, weldable steels of pipelines, bridges and ships.

Worked examplesH

Annealing, quenching and tempering steel

Steel heat treatments all begin by austenitising, holding the steel above its A₃ temperature (or, for hypereutectoid steels, above A₁) until it is austenite, and then differ in how it is cooled, as the figure shows.

85072755020time (schematic)temperature (°C)A₃A₁austenitisequenchtemperfurnace cool: annealair cool: normalisequench and temper
Schematic temperature–time paths for heat treatments of a 0.4 wt% carbon steel. After austenitising above A₃, slow cooling in the furnace (annealing) gives coarse pearlite and ferrite; cooling in still air (normalising) gives finer pearlite and a refined grain size; quenching gives martensite, which is then reheated below A₁ and held (tempering).
  • Full annealing: cool slowly in the furnace. The austenite transforms high on the C-curves to coarse pearlite (and ferrite or cementite), giving the softest, most machinable condition.
  • Normalising: cool in still air. The transformation happens lower, giving finer pearlite and smaller grains: stronger and tougher than annealed steel, and more uniform after forging or welding.
  • Hardening: quench in water, brine, oil, a polymer solution or high-pressure gas, fast enough to miss the C-curves and form martensite.
  • Tempering: reheat the quenched steel to between about 150 and 650 °C. Carbon leaves the martensite as fine carbides, internal stresses relax, and hardness falls while toughness rises. Low tempering temperatures suit tools and bearings, which need hardness; higher ones suit shafts, bolts and springs, which need toughness.

Whether a quench produces martensite depends on how fast each part of the section cools, and the inside of a thick bar cools much more slowly than its surface, because heat must be conducted out through the steel. The block solves the heat equation across a long round bar, the same mathematics as diffusion, for three diameters and two quenchants, with rounded properties for hot steel and illustrative effective heat-transfer coefficients. It compares the centre's average cooling rate from 800 to 500 °C with two critical rates: about 200 K/s for the plain eutectoid steel modelled in phase transformations, and 10 K/s for an illustrative alloy steel.

python
import math
import numpy as np

# Quenching a long steel bar: radial heat conduction (finite volumes, explicit), with heat
# transfer to the quenchant at the surface.  Rounded property values for hot steel.
k, rho_c = 30.0, 7850 * 600.0          # W/(m K) and J/(m^3 K)
T_start, T_bath = 850.0, 40.0          # C
quenchants = {"agitated water": 3000.0, "oil": 700.0}   # effective h, W/(m^2 K): illustrative

def quench(radius, h, n=30):
    """Return the times (s) at which the centre and the outermost cell cool from 800 to 500 C."""
    dr = radius / n
    r_face = np.arange(n + 1) * dr                     # cell faces, 0 .. radius
    vol = math.pi * (r_face[1:] ** 2 - r_face[:-1] ** 2)   # per metre of bar
    G = k * 2 * math.pi * r_face[1:-1] / dr            # conductance between neighbouring cells
    U = 2 * math.pi * radius / (1 / h + dr / (2 * k))  # last cell to the bath
    cap = rho_c * vol
    dt = 0.4 * np.min(cap / (np.r_[0, G] + np.r_[G, U]))
    T = np.full(n, T_start)
    t, marks = 0.0, {}
    while T[0] > 450:
        q = G * (T[:-1] - T[1:])                       # heat flow outwards across inner faces
        dQ = np.r_[0, q] - np.r_[q, U * (T[-1] - T_bath)]
        T = T + dt * dQ / cap
        t += dt
        for name, i in (("centre", 0), ("surface", -1)):
            for level in (800, 500):
                if (name, level) not in marks and T[i] <= level:
                    marks[(name, level)] = t
    return {name: (marks[(name, 800)], marks[(name, 500)]) for name in ("centre", "surface")}

# Critical cooling rates (K/s) for a fully martensitic centre: the plain eutectoid steel of the
# model in the previous lesson, and an illustrative alloy steel with much higher hardenability.
critical = {"plain": 200.0, "alloy": 10.0}

print("bar      quenchant         800->500 C: near surface   centre   centre rate   centre martensitic?")
print("                                                                            plain    alloy")
for d_mm in (10, 25, 50):
    for name, h in quenchants.items():
        m = quench(d_mm / 2000, h)
        s0, s1 = m["surface"]
        c0, c1 = m["centre"]
        rate = 300 / (c1 - c0)
        verdict = ["yes" if rate >= critical[steel] else "no" for steel in ("plain", "alloy")]
        print(f"{d_mm:3d} mm   {name:15}   {s1 - s0:20.1f} s   {c1 - c0:5.1f} s   {rate:7.0f} K/s"
              f"   {verdict[0]:>5}    {verdict[1]:>5}")
Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

The plain carbon steel's centre does not harden in any of these bars: even the 10 mm bar in agitated water reaches only 132 K/s at its centre. (Real water quenches remove heat faster than this average estimate during the first fraction of a second, while the water boils violently at the surface, so thin plain-carbon parts do harden at and near the surface.) The alloy steel, whose C-curves sit far to the right, hardens through everywhere except at the centre of the 50 mm oil-quenched bar, which cools at only 6 K/s. This mass effect is why designers choose steels by their hardenability for the section size, and why large parts are made from alloy steels even though the alloying elements cost more.

Tempering depends on both temperature and time, because it is diffusion. John Hollomon and Leonard Jaffe showed in 1945 that the hardness after tempering depends, approximately, on the single parameter P=T(C+log⁡10t)P = T(C + \log_{10} t), with TT in kelvin, tt in hours and C≈20C \approx 20 for many steels. Equal PP means equal effect. One hour at 600 °C gives P=873.15×20=17 463P = 873.15 \times 20 = 17\,463; at 550 °C the same PP needs log⁡10t=17 463/823.15−20=1.215\log_{10} t = 17\,463/823.15 - 20 = 1.215, about 16 hours, and at 650 °C only about 5 minutes. Heat-treatment plants use this to convert a recipe from one furnace schedule to another.

Many parts need a hard surface and a tough core. Besides the carburising and nitriding of diffusion, induction hardening heats only a thin surface layer of a medium-carbon steel with a high-frequency magnetic field and quenches it at once, and laser and electron-beam hardening do the same with a beam.

Worked examplesH

Sintering ceramics

Ceramics such as alumina and zirconia melt at over 2000 °C and are brittle, so they can be neither cast easily nor shaped by deformation. Instead they are made from powders: a fine powder is pressed, cast as a slurry or injection-moulded into shape, giving a fragile green body that is perhaps 50 to 60 % dense, which is then fired at a high temperature, typically 0.6 to 0.8 of its absolute melting point, until the particles bond and the pores close. This is sintering, and it is how spark-plug insulators, zirconia dental crowns, silicon nitride bearing balls, porcelain and tiles, and the tungsten carbide tools of machine shops are made.

The driving force is the reduction of surface area: a powder has an enormous surface, and atoms at a surface have higher energy than those inside. Where two particles touch, a neck forms and grows, fed by atoms diffusing in by several paths, as the figure shows.

surface diffusion: neck grows, no shrinkagegrain-boundary diffusion:densifieslattice diffusion from the boundary (dashed): densifiesneckcentres approach as material leaves the boundary
Two powder particles sintering. Atoms move into the concave neck because its curvature lowers their chemical potential there. Surface diffusion (and evaporation and condensation) only moves material along the surface: the neck grows but the particle centres stay the same distance apart. Diffusion from the grain boundary between the particles, along the boundary or through the lattice, removes material from between them, so the centres approach and the body shrinks and densifies.

Only the paths that take atoms from the grain boundary between particles make the body shrink: grain-boundary diffusion and lattice diffusion from the boundary (and viscous flow, in glasses). Surface diffusion and evaporation–condensation only reshape the pores. Sintering proceeds in stages: necks grow; the pores become a connected network of channels along grain edges, which shrink; finally the pores pinch off into isolated voids that slowly vanish. Over the whole process a part typically shrinks by 15 to 20 % in each direction, which designers must allow for. The constant enemy is grain growth, which is also driven by surface energy: if grain boundaries sweep past pores faster than the pores can shrink, the pores are stranded inside grains and never close. A famous solution was the addition of a fraction of a per cent of magnesia to alumina, which slows boundary migration enough for alumina to sinter to full density, translucent enough for the arc tubes of sodium-vapour street lamps.

Engineers speed it up in several ways. Liquid-phase sintering adds a small amount of a phase that melts at the sintering temperature and wets the particles, such as cobalt in tungsten carbide, so that atoms move through the liquid. Hot isostatic pressing applies gas pressure, of the order of 100 MPa, at temperature, which closes pores in ceramics, metal powders and castings. Spark plasma sintering, or field-assisted sintering, passes a large pulsed current through the powder and its die, heating it in minutes. And cold sintering, reported by Clive Randall's group at Penn State in 2016, densifies some ceramics below 300 °C with a little water or another transient liquid and pressure, opening the way to co-firing ceramics with polymers and metals.

ExplanationH

Heat treatment of aluminium alloys

Aluminium alloys come in two kinds. The non-heat-treatable ones, such as the 1xxx (nearly pure), 3xxx (manganese) and 5xxx (magnesium) series, are strengthened by solid solution and work hardening. The heat-treatable ones, the 2xxx (copper), 6xxx (magnesium and silicon) and 7xxx (zinc and magnesium) series, are strengthened by precipitation hardening (phase transformations), and their treatment is read straight off the phase diagram.

02004005486600245.658temperature (°C)liquidαα + θ (Al₂Cu)α + liquid(1) solution treat, about 520 °C(2) quench(3) age, about 170 °Csolvuscopper (wt%)
The aluminium-rich corner of the aluminium–copper diagram, schematic, with the treatment of an alloy with 4 wt% copper. Key points are at their assessed positions: aluminium melts at 660 °C, the eutectic is at 548 °C, and the solid dissolves at most 5.65 wt% copper. (1) Solution treat in the single-phase α field, between the solvus and the solidus; (2) quench to room temperature, trapping the copper in a supersaturated solution; (3) age at a moderate temperature so that fine precipitates form.

The window for solution treatment is narrow: hot enough to dissolve the copper (above the solvus), but not so hot that the grain boundaries, where the last eutectic liquid of the casting froze, start to melt (below the solidus and the eutectic). Industrial furnaces for aerospace parts must hold their temperature within a few kelvin, and their thermocouples and controls are calibrated and audited to standards such as SAE's AMS2750. The quench must be fast, because slow cooling lets coarse, useless precipitates form at grain boundaries; thick sections are hard to quench, and a fast quench distorts thin ones, so the quenchant is a compromise.

The condition of an aluminium alloy is written as a temper after the alloy number, in the designation system maintained by the Aluminum Association (ANSI H35.1):

TemperMeaning
Fas fabricated, no control
Oannealed: softest, most formable
H1x, H2x, H3xstrain-hardened (non-heat-treatable alloys), with the second digit giving the degree of hardening
T4solution treated, quenched and naturally aged at room temperature
T6solution treated, quenched and artificially aged to peak strength
T7solution treated, quenched and overaged, trading some strength for better resistance to stress-corrosion cracking and more stable dimensions

The aircraft alloy 7075 is a well-known case for T7: in the peak-aged T6 condition it can suffer stress-corrosion cracking in some environments, so parts that face that risk are deliberately overaged to tempers such as T73, accepting lower strength.

ExplanationH

Processing in industry and research

Simulation of processes. Foundries simulate mould filling, solidification and defects before pouring metal, with commercial packages such as MAGMASOFT and ProCAST; forging and rolling are simulated by finite elements (Lumenary ME's finite element method); and heat-treatment simulations combine the heat equation, transformation kinetics and mechanics to predict hardness, distortion and residual stress. Integrated computational materials engineering chains these with the thermodynamic and kinetic models of the previous lessons, and the US Materials Genome Initiative, among similar programmes in Europe and Asia, has funded the databases and tools that make it possible (see multiscale modelling).

New processes. Metal additive manufacturing melts powder or wire with a laser or electron beam, track by track, and so is a casting process with tiny melt pools and cooling rates of thousands to millions of kelvin per second, giving very fine, often unusual microstructures; much current research, as of 2026, is about controlling them and the defects they bring (see Lumenary ME's additive manufacturing). In ceramics, field-assisted and cold sintering are moving from laboratories into production for some electronic components.

Measurement and quality. Heat-treated parts are checked by hardness testing, by metallography (cutting, polishing, etching and examining a section under the microscope, microscopy), and for critical parts by non-destructive testing. Aerospace and automotive heat treaters are audited against written specifications for furnace uniformity, instrument calibration and records, because a part that has been heat treated wrongly usually looks exactly like one that has been treated correctly.

Common misconceptionsH

Common misconceptions

Misconception

"Annealing makes metal stronger." Annealing softens: it relieves stresses, recrystallises cold-worked grains and, for steel, produces soft pearlite. The word is sometimes confused with hardening because both involve heating.

Misconception

"Quenching hardens any metal." Quenching hardens steels with enough carbon because it forms martensite. Quenched copper or pure aluminium is no harder, and a heat-treatable aluminium alloy is at its softest just after quenching; it hardens later, during ageing.

Misconception

"Recrystallisation means melting and refreezing." Recrystallisation happens entirely in the solid, well below the melting point: new strain-free grains nucleate and grow into the deformed metal.

Misconception

"Sintering melts the powder together." Most sintering happens in the solid, by diffusion, at 60 to 80 % of the absolute melting temperature. Only liquid-phase sintering uses a liquid, and then only a small fraction of the material melts.

Exercises

Exercises

Take R=8.314R = 8.314 J mol⁻¹ K⁻¹ and temperatures in kelvin wherever a formula needs them.

1. Processing and heat treatment

Warm-upHuman

A foundry makes a casting, and then a new version with every dimension doubled, in the same alloy and the same kind of mould. By Chvorinov's rule, t=B (V/A)2t = B\,(V/A)^2, how many times longer does the larger casting take to solidify?

2. Processing and heat treatment

Warm-upHuman

Taking hot working as working above about half the absolute melting temperature, which of these are hot working? Select all that apply. (Melting points: lead 327 °C, copper 1085 °C, aluminium 660 °C, tungsten 3422 °C.)

Choose all that apply

3. Processing and heat treatment

CoreHuman

A recrystallised metal has a mean grain size of 20 µm. After 1 hour at an annealing temperature it is 40 µm. Assuming parabolic grain growth, d2−d02=ktd^2 - d_0^2 = kt, what is the grain size, in µm to the nearest whole number, after 4 hours at the same temperature?

4. Processing and heat treatment

CoreHuman

A 7xxx-series aluminium alloy forging will carry a sustained tensile load in a humid, salty environment, where stress-corrosion cracking is the main concern, and it can accept a little less than peak strength. Which temper should be specified?

Choose one

5. Processing and heat treatment

CoreHuman

During sintering, which transport mechanisms make the powder compact shrink and densify (bring particle centres closer together)? Select all that apply.

Choose all that apply

6. Processing and heat treatment

CoreComputational

Write scheil_eutectic_fraction(C0, k, C_E), which returns the fraction of a casting that is still liquid when the Scheil liquid composition CL=C0(1−fs)k−1C_\text{L} = C_0(1 - f_\text{s})^{k - 1} reaches the eutectic composition C_E; that liquid freezes as eutectic. Raise ValueError unless 0 < k < 1 and 0 < C0 < C_E.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

7. Processing and heat treatment

CoreComputational

Write equivalent_time(T1_C, t1_h, T2_C, C=20.0), which returns the time in hours at temperature T2_C (°C) that gives the same Hollomon–Jaffe tempering parameter P=T(C+log⁡10t)P = T(C + \log_{10} t) as t1_h hours at T1_C, with TT in kelvin.

Runs entirely in your browser, in an isolated worker. Nothing is sent to a server.

Further reading and resources