IB Paper 3: Materials

Materials Processing:
Handout 1, Lectures 9-11
Heat Treatment of Metal Alloys

Dr Graham McShane
gjm31@cam.ac.uk

Michaelmas 2023

Image courtesy of K-H Rendigs, Airbus

Image courtesy of Alcoa

Contents:
1. Introduction………………………………………………………………..p.2
2. Heat treatment of aluminium alloys..............................p.4
3. Heat treatment of steel……………….................................p.13
4. Diffusion in heat treatment processes………………………..p.26

1
1. Introduction
Materials processing refers to all the steps required to take a raw material to a finished
product. In weeks 5-8 we will apply the fundamental principles of materials
thermodynamics and diffusion learnt in weeks 1-4 to explain how manufacturing
processes can be used to manipulate the microstructure and properties of materials.

The aim is to develop a more detailed understanding of how key process variables
interact to deliver the required properties, for different classes of material:

Process variables (inputs): Microstructure: Material properties (outputs):

• material type • phases present • stiffness (Young’s modulus)
• composition • spatial distribution • yield strength ( 𝜎𝑦 )
• thermal history • grain size / shape • fracture toughness ( 𝐾𝐼𝐶 )
• deformation history • dislocation density

The typical range of processing steps was introduced in Part IA. The following processes
will be considered in Part IB:

Lectures 9-11: Heat Treatment

• Heat treatment of metal alloys:
Al alloys, steels
strength, toughness
• Diffusion in heat treatment

Lectures 12-14: Shaping Processes

• Casting of metals:
alloy choice
cast microstructures
• Deformation processes for metals:
rolling, forging loads
work hardening, annealing
Image: Ashby, Shercliff & Cebon,
Lectures 15-16: Materials Engineering, Science,
Processing & Design
• Polymer processing
• Creep: metals at high temperatures

2
Example application: improving fuel efficiency and reducing emissions in aerospace
• High strength lightweight alloys = less material, less weight (weeks 5-6 content)
• Alloys with higher operating temperatures = more efficient engines (week 8 content)

Airbus A380 wing box (Image: K-H Rendigs, Airbus)

The lectures in weeks 5-8 will make use of the following previous content:

Part IB: materials thermodynamics and diffusion (weeks 1-4). Underpins microstructure
development in materials (phases present, proportions, compositions, spatial distribution).
(a) The equilibrium phases: (b) Phase transformations:
- minimise Gibbs free energy (𝐺) - temperature and time dependence
- composition and temperature dependence - thermodynamic driving force Δ𝐺
- atomic rearrangement by diffusion
T1 T2
T1

T2

T3
T4
eutectoid point, T3 T4
0.8 wt% C
0.3 wt% C

Part IA: microstructural origin and manipulation of properties.

• Property definitions: e.g. strength (yield stress, 𝜎𝑦 , or hardness) and toughness.
• Metals: microstructural basis of yielding, dislocations, hardening mechanisms.
• Polymers: role of molecular structure in stiffness and strength.

3
2. Heat treatment of aluminium alloys
Aluminium (Al) alloys belong to the family of ‘light alloys’ (which includes titanium and
magnesium). They play an important role in the design of lightweight components and
structures. Aluminium has a density and stiffness about 1/3 that of steel:

𝜌 = 2700 kg/m3, 𝐸 = 70 GPa

However, pure aluminium is soft (low strength): 𝜎𝑦 = 40 MPa
For structural applications, the strength can be increased by a factor of 10 through
alloying and heat treatment.

High strength aluminium alloys are particularly important for weight reduction in
transport applications (automotive, aerospace, rail, marine).

formed Al extrusions

die-cast Al joints

Alcoa/Audi
spaceframe Image courtesy of Alcoa

Jaguar F-Type

The most important hardening mechanism for high strength Al alloys is precipitation
hardening – pinning of dislocations by strong, second phase particles. Heat treatment
processes can be used to manipulate the precipitation phase transformation to
produce strong, tough microstructures.

Recall from IA the key points for precipitation hardening:

• The yield stress is inversely proportional to the obstacle spacing.
• It also depends on the strength of the obstacle.

‘Weak’ dislocation pinning ‘Strong’ dislocation pinning

4
2.1 Composition range for heat treatable Al alloys
Aluminium can be combined with a variety of other elements, resulting in a large
family of alloys. We will use Al-Cu alloys here as an example, as it is commonly heat
treated to improved its strength.

The phase diagram can be used to identify suitable alloy compositions for heat
treatment, and critical temperatures for the heat treatment process. The key
requirements for heat-treatable alloys are:
• At high temperature: single phase solid solution (high solubility of alloy additions)
• At low temperature: two-phase region (low solubility of alloy additions)
• A solvus line (solubility limit) is crossed on cooling, driving precipitation of the
second phase, which is required for precipitation hardening.

Al-Cu phase
diagram
(Al)

single phase
solid solution

solubility
two-phase limit
region

The fall in solubility of the (Al) phase on cooling is the key feature of the phase diagram
being exploited for precipitation hardened alloys. Recall from Teach Yourself Phase
Diagrams that slow cooling allows time for the equilibrium microstructure to form:

• Below the solvus line, a two-phase mixture has the lowest

Gibbs free energy, and so a phase transformation occurs:
(Al) → (Al) + CuAl2
• Cu is rejected from the single phase (Al) solid to form CuAl2
precipitates. The driving force for precipitation is the
reduction in Gibbs free energy. Diffusion enables the
transformation to occur.
• Precipitates nucleate most easily heterogeneously on grain
boundaries (more space in the Al lattice, faster diffusion).
The largest precipitates will occur here. Precipitates can
also nucleate homogeneously within the grains.
5
However, this slow-cooled microstructure provides ineffective precipitation hardening:
• Precipitates are large and spaced far apart. Precipitate coarsening is driven by Gibbs
free energy, as it reduces the interface area (and interface energy penalty) per unit
precipitate volume. Slow cooling gives time for this coarsening to occur.
• They provide ineffective obstacles to dislocations: 𝜎 ∝ 1 / obstacle spacing
𝑦

2.2 Rapid cooling of Al alloys (quenching)

Consider next the effect of increasing the cooling rate of the alloy, from the single-
phase solid solution. Recall from earlier lectures that phase transformation rate
depends on temperature and time, as follows:
T
• The overall rate of transformation depends on the
rates of nucleation and growth. TE
• This in turn depends on the undercooling:
Δ𝑇 = 𝑇𝐸 − 𝑇
• Small Δ𝑇 → small Δ𝐺 → low thermodynamic
driving force.
• Large Δ𝑇 → low thermal energy → rearrangement of
atoms by diffusion is slow.
• The peak rate occurs at an intermediate temperature. T

It is convenient in materials processing to invert the rate TE

axis and show instead the time taken for a certain
fraction to transform (‘C-curve’). The plot of fraction
transformed for a given temperature and time is known
as a Time Temperature Transformation (TTT) diagram.

The same principles govern precipitation in the aluminium alloy:

TE for (Al) as the equilibrium
phase: solvus line

(Al) stable

(Al) + CuAl2

6
Note that TTT diagrams apply to isothermal transformations only (i.e. with the specimen
held at a fixed temperature). In practical heat treatment processes, continuous cooling is
more likely to occur. There is an analogous continuous cooling transformation (CCT)
diagram, where the C-curves show the fraction transformed for a fixed cooling rate.

(Al) stable • A key piece of information shown

by the CCT diagram is the critical
cooling rate (CCR).
• This is the slowest cooling rate that
just avoids the C-curves.
(Al) + CuAl2 • Cooling at this rate or faster means
that no precipitation phase
transformation occurs. There is
insufficient time or temperature
for nucleation of precipitates.

Material cooled faster than the critical cooling rate will therefore remain a single phase
solid solution at room temperature:

Above TE (above the solvus line): Room temperature:

quench

(Al) (Al)
𝑑𝑇
≥ CCR
single phase 𝑑𝑡 single phase
solid solution supersaturated solid solution
(stable; equilibrium) (metastable; not equilibrium)

This is not the equilibrium state: it is a “metastable” micro-structure. There is more Cu

trapped in the Al lattice than the equilibrium amount, and it is therefore known as a
supersaturated solid solution (SSSS).

This results in a material with effective solid solution hardening. It has a higher yield
strength than the slow-cooled alloy with the coarse precipitates. However, this is still not
the optimum microstructure for maximum strength.

7
2.3 Age hardening
A higher yield strength than either slow-cooled or quenched alloys can be achieved by
a process known as age hardening. The aim of this heat treatment process is to
achieve precipitation hardening, but compared to slow cooling we want:
• precipitates that are smaller and more closely spaced,
• more effective obstacles to dislocations.

The process consists of three steps:

(1) solution heat treatment
• dissolve alloying elements, to (2) quenching
achieve single phase solid solution • cool faster than the CCR, forming a
supersaturated solid solution (SSSS)
temperature

(3) ageing
• re-heat, allowing time for
precipitation to occur from the
metastable SSSS

time

There are two options for the ageing step:

• artificial ageing: re-heat for a few hours at around 150-200°C
(which still keeps the material within the two-phase region on
the phase diagram),
• natural ageing: leave for several days at room temperature.

During the ageing step, precipitates nucleate and grow. The size and spacing of the
precipitates can be controlled by varying the ageing time and temperature.

With increasing ageing time, the precipitates coarsen: average size and spacing ↑

• Driven by Gibbs free energy: coarsening

reduces the (Al) / precipitate interface
area (interface energy penalty).
• Higher ageing temperature means:
- more rapid coarsening
- but a lower equilibrium fraction
of precipitates (from the phase diagram).

8
 As the precipitates grow, the effectiveness of the precipitation hardening changes:

• When the precipitates are small,

dislocations easily shear through
them. They are weak obstacles.
• As they grow, the shearing resistance
increases with the radius of the
precipitate, until the dislocation precipitate shearing
breaking angle falls to zero. radius ↑ ⇒ strength ↑
• Typically: strength ∝ radius1/2

• As the precipitates get large, the

spacing also gets large (the volume
fraction of precipitates initially
increases but then remains fixed).
• Strength is now dominated by
dislocations “bypassing” obstacles
(breaking and reforming of the
dislocation bypassing
dislocation)
• Strength ∝ 1 / spacing ∝ 1 / radius radius ↑ ⇒ strength ↓

These two effects trade off as the precipitates grow: precipitate shearing dominates
the strength at first, then bypassing. A peak in resistance is seen when the two
mechanisms swap over. The strength of the material varies with ageing time as shown:

shear stress to
overcome obstacle

bypassing
shearing

precipitate radius, r

• Artificial ageing: yield stress rises to a ‘peak aged’ condition, and then falls to a
soft ‘overaged’ state (which is similar to the slow cooled material).
• Natural ageing: the low temperature means slow diffusion, and precipitates
remain small; yield stress rises more slowly, to an intermediate plateau value. 9
Experimental evidence from transmission electron microscopy (TEM) of high-strength
aerospace Al-Zn-Mg-Cu alloy (artificially aged):
(a) initial precipitation (b) peak aged (c) over aged

(note scale: Source: Poole W.J., Shercliff H.R. and Castillo T.,
cf. diameter of Al atom = 0.28 nm) Mat. Sci. Tech. 13, Nov. 1997, pp.897-904

Note that in natural ageing, the microstructure evolution stops at a point similar to (a)
above. There is insufficient thermal energy for ageing to continue, and the material
remains “underaged” (but with useful strength, and with lower processing cost).

2.4 Age hardening: further microstructural details

We will look more closely now at the details of this age hardening process. This is
important to understand why some aluminium alloys age harden in this way, but some
do not (the latter are known as “non-heat-treatable” alloys).

• A first look at the TTT diagram

suggests there would be no = (Al)
precipitation during ageing – the
time to produce CuAl2 is very
long at ageing temperatures.
• However: metastable (non-
equilibrium) precipitates can
form at these low temperatures
(even at room temperature).
• There is another set of C-curves
governing the precipitation of
these metastable phases.

10
Metastable precipitates have a different crystal structure to equilibrium CuAl2 :

Al lattice precipitate
• The precipitate is coherent, i.e. its crystal lattice
matches that of the Al closely. This reduces the
interface energy penalty to nucleation.
• Because slip planes are continuous across the
interface, dislocations easily cut through these
precipitates (the “shearing” mechanism).
• They are weak obstacles, with a high dislocation
breaking angle.
coherent

As coarsening proceeds, coherency increasingly strains the lattice:

Al lattice precipitate • Overall, it becomes energetically favourable for the

precipitates to transform through a whole
sequence of metastable phases, eventually
reaching equilibrium CuAl2 , with a progressive loss
of coherency at the interface.
• This is the result of a tradeoff between the
interface energy penalty (coherent vs incoherent)
and the volumetric Gibbs free energy penalty (of
incoherent metastable precipitates vs equilibrium CuAl2).

The precipitates at the peak-aged state will be fine metastable precipitates, with larger
equilibrium CuAl2 precipitates occurring in the over-aged material.

Non-heat-treatable Al alloys:
• Some alloys, due to their
composition, do not form these
metastable precipitates at lower
temperatures.
• Precipitates only form with very
slow cooling, which would give
coarse precipitates, providing
negligible hardening. Even a
modest quench will retain a 100%
super-saturated solid solution.
• So, they rely on solid solution and
work hardening.
• Examples include: Al-Mg (‘5000
series’), Al-Mg-Mn (‘3000 series’). 11
2.5 Summary of the learning outcomes: heat treatment of Al alloys
After completing section 2 you will be able to do the following:

1. Use the phase diagram to identify suitable composition and temperature ranges for
the heat treatment of an Al alloy.
2. Use TTT and CCT diagrams to explain the differences in microstructure and
properties between slow cooled and quenched aluminium alloys.
3. Describe the evolution of the microstructure and properties during age hardening.
4. Explain the difference between heat treatable and non heat treatable Al alloys.

2.6 Quiz W5.1: heat treatment of Al alloys

Samples of a heat treatable aluminium alloy are initially held at high temperature until
they become a single-phase solid solution.

Sample A: cooled slowly

Sample B: quenched
Sample C: quenched and aged

Which of the following statements are correct (more than one can be correct)?

1. B has a higher yield strength than A, because there is solid solution hardening
instead of precipitation hardening.

2. C has a higher yield strength than B, because there is precipitation hardening

instead of solid solution hardening.

3. C has a higher yield strength than A, because there is precipitation hardening

instead of solid solution hardening

Additional worked examples: heat treatment of Al alloys

For further explanation, a video will be available with the following worked examples:
• Tripos Question 2016 Q3, which helps with Examples Paper 4 Q1.
• A discussion of Examples Paper 4 Q1(b-c), and non-heat-treatable alloys.

Now attempt: Examples Paper 4, Q.1

12
3. Heat treatment of steel
Steels (Fe-C alloys) are the most widely used class of engineering alloys. In this section
we will consider how heat treatment can be used to manipulate their microstructure
and mechanical properties. Steels are an important engineering alloy due to their
stiffness, strength and low cost, though this comes with a high density:

𝜌 = 8000 kg/m3, 𝐸 = 210 GPa

Mild steels (0.1-0.2 wt% C) are ubiquitous structural materials. Alloying and heat
treatment can be used to produce extremely hard steels, useful for cutting tools, wear
resistant gears etc.

alloy steel, heat treated

mild steel
up to
𝜎𝑦 ≈ 300 MPa
𝜎𝑦 ≈ 2000 MPa

Images: Corus media gallery

The key region of the phase diagram for the heat treatment of steels is shown below.
Typical steel compositions lie in the range ~0.1 - 2 wt% C. ‘Mild steels’ have a low C
concentration, ~0.1-0.2 wt% C. In what follows, we focus on hypo-eutectoid steels:
those with composition <0.8 wt %.

single phase region

two phase region

Like Al alloys, the heat treatment of steels involves moving between single and two-
phase regions of the phase diagram. However, unlike the precipitation phase trans-
formation in Al alloys, steel involves a three-phase reaction:
• FCC Austenite ( γ ) → BCC Ferrite ( α ) + Fe3C

13
3.1 Slow cooled steel
The microstructures of slow cooled steels were introduced in
Teach Yourself Phase Diagrams. A brief recap is provided
here. Slow-cooled steels are referred to as normalised. Hot-
rolled steel is often in this condition.
Image: Corus media gallery

Microstructure evolution on slow cooling for a Fe – 0.4 wt% C steel:

(A) the material enters the two-phase

α + γ region at around 800°C;
– the ferrite α rejects carbon into the
austenite γ;

(B) at the eutectoid temperature, the

austenite (now containing 0.8wt% C)
transforms to the microstructure
pearlite, containing the phases ferrite
α + cementite Fe3C.

A B

The properties of normalised hypo-eutectoid steels are determined by the carbon

content:
• The carbon content affects the proportion of ferrite and pearlite: the level rule,
applied just above the eutectoid temperature, gives the proportion of austenite,
which transforms into pearlite on further cooling.
• Cementite is very hard, and the pearlite structure obstructs dislocation motion
effectively. So, the higher the pearlite fraction, the harder the material.
• Low carbon steels have only a small proportion of pearlite, and have a yield stress
around 200 MPa, which is combined with high ductility. This is the lower end of
the strength range.
• Normalised eutectoid steel (100% pearlite) has a yield stress around 450 MPa, but
with lower ductility. This is the upper end of the strength range. 14
3.2 TTT diagrams for hypo-eutectoid steels
Just like Al alloys, cooling rate has a significant effect on the microstructure and
properties of steel. We can use the TTT diagram to assess the influence of temperature
and time on the phase transformations. The TTT diagram for 0.4 wt% C, hypo-
eutectoid steel is shown below, alongside the phase diagram for comparison:

(i)
(ii)

(iii)

(iv) Fe-C phase diagram

Note: T axes aligned

The various transformations (i)-(iv) shown on the TTT diagram are now described in turn:

(a) Diffusion-controlled transformations (higher temperatures):

austenite, γ → ferrite, α + cementite, Fe3C (equilibrium phases)
TTT diagram has the conventional C-curve shape for diffusional phase transformations
at higher temperatures, with the following details (for hypo-eutectoid steels):

(i) above the “nose” of the C-curves, and above the “A1 temperature”, 723°C:
• ferrite forms first (from austenite grain boundaries)
• final microstructure: ferrite + austenite (proportions given by the tie line)

(ii) above the “nose” of the C-curves, and below the “A1 temperature”, 723°C:
• ferrite forms first (from austenite grain boundaries)
• transformation switches from ferrite to pearlite at the carbide line;
• final microstructure: ferrite + pearlite

Microstructure: A quench-and-hold close to (and

below) the A1 temperature is essentially very similar
to slow continuous cooling, as discussed above.
Shown right is the final microstructure for case (ii).

15
(iii) below the “nose” of the C-curves: diffusion is more inhibited
• austenite transforms directly to ferrite + iron carbide, in a fine-scale dispersion
• this microstructure is called bainite.

Microstructure: Bainite nucleates on the austenite

grain boundaries, with simultaneous growth of ferrite
and iron carbide, giving a fine-scale dispersion of iron
carbide needles within a ferrite matrix (rather than the
lamellar structure of pearlite).

(b) Diffusionless transformation (lower temperatures):

austenite, γ → martensite, α’ (metastable phase)
Quenching carbon steels to bypass the nose of the C-curves (i.e. faster than the Critical
Cooling Rate) prevents diffusive phase transformations from taking place. Instead there
is a transformation to a metastable phase called martensite (α’): case (iv) .

• Quenching produces a supersaturated solid solution of C in Fe, i.e. there is

more carbon trapped in the Fe lattice than the equilibrium amount (similar
to the Al alloy case).
• However, in contrast to Al alloys, iron is FCC at high temperatures (austenite),
but BCC at low temperatures (ferrite). Therefore the SSSS is also initially in
the ‘wrong’ atomic packing, compared to equilibrium.
• To overcome this, the material can get closer to BCC without diffusion, by a
modest distortion (straining) of the lattice. This is the martensite phase
transformation.

This lattice shape change is

achieved by shear – narrow
lattice shears
regions shear across the
austenite grain (at the speed of
growth direction
sound in the material, effectively
instantaneously), leaving a needle
C remains in
of the martensite phase within
solid solution
the austenite grain.

needle of martensite 16
This transformation takes place without
diffusion:
• there is no time-dependence on the TTT
diagram,
• the contours of fraction transformed are
horizontal lines.

However, the fraction transformed does depend on the temperature:

• Martensite begins to form at the “martensite start temperature”, Ms
• Reducing the temperature further increases the fraction of austenite that
transforms to martensite.
• The transformation is complete at the “martensite finish temperature”, MF
lower temperature: larger fraction transformed

This is for thermodynamic reasons:

• As the temperature falls, at a temperature well below A1 (see above), the Gibbs free
energy of the martensite phase becomes lower than the austenite phase: there is a
thermodynamic driving force (Δ𝐺) for the martensite phase transformation.
• A certain degree of undercooling (and hence Δ𝐺) is needed to start the
transformation. This occurs at the martensite start temperature.
• At a given temperature, the transformation of a percentage of the austenite into
martensite can reduce the free energy sufficiently to stop further transformation.
• The undercooling (and Δ𝐺) must be increased further to induce further
transformation. This completes at the martensite finish temperature.

17
 Example: reading from the TTT diagram for a hypo-eutectoid steel
Estimate the final microstructure for the hypo-eutectoid steel with the TTT diagram
shown below, subject to the following heat treatment:
• Austenitise (i.e. solution heat treat) for 30 minutes at 850˚C
• Quench to 630˚C, hold for 10 minutes, quench to room temperature.

γ → ferrite ( α )
austenite ( γ )
carbide line:
• ferrite growth stops when the
(a) (b) (c) carbide line is reached, and
pearlite growth begins
γ → pearlite ( α + Fe3C )

γ → bainite

C-curves:
• show the total % of austenite
that has transformed
• interpolate linearly,
perpendicular to the contours,
to estimate the fraction at a
given point

γ → martensite ( α’ )

(d)

Point Total % transformed Transformed phases Balance

(from the C - curves)
(a) 0% none 100% austenite
(b) 40 % 40% ferrite 60% austenite
(c) 100 % 40% ferrite + 60% pearlite none
(d) 100 % 40% ferrite + 60% pearlite none

Note that no martensite transformation occurs between (c) and (d), as there is no
austenite left at (c) to transform on quenching. It has all transformed to ferrite and
pearlite, which is unaffected by the quench.
18
3.3 Effect of carbon content on the steel TTT diagram
As the carbon content of hypo-eutectoid steel increases (between ≈0.1 and 0.8 wt% C),
there are systematic changes to the TTT diagram:

(1) The proportion of ferrite to pearlite

decreases:
• This follows the phase diagram, and the
lever rule applied just before the
eutectoid transformation (see examples
paper 1).
• At the eutectoid composition, the ferrite
region has disappeared altogether: the
carbide line on the TTT diagram merges
with the curve for the start of
transformation, which goes straight to
pearlite above the nose.

(2) The C-curves move to longer times:

TTT diagram: eutectoid steel
• The greater amount of C to redistribute
delays all of the diffusional
transformations.
T ( ˚C )
(3) The martensite start and finish 600
temperatures decrease:
• More carbon in supersaturated solid 400
Ms
solution means more lattice strain.
• This requires a greater undercooling to 200 Mf
achieve the martensite transformation.
0 0.2 0.4 0.6 0.8 1.0 1.2
wt % C

Note that for eutectoid steel, Mf is -50˚C. At room temperature, a quenched specimen
will therefore contain a fraction of retained austenite.

19
3.4 Properties of bainite and martensite
Recall that slow cooled (‘normalised’) hypo-eutectoid steels
have a yield strength dependent on the carbon
concentration (and hence pearlite fraction) in the range:
𝜎𝑦 ≈ 200 - 450 MPa
(0.1 - 0.8 wt% C)
Bainite has the ideal microstructure for effective
precipitation hardening: a fine-scale dispersion of hard
precipitates (Fe3C). Also, all grains have the same micro-
structure and strength. Yield strengths are in the range:
𝜎𝑦 ≈ 500 - 700 MPa

Martensite has a very high yield

strength (i.e. high hardness):
𝜎𝑦 ≈ 1000 - 3000 MPa
• The BCC Fe lattice is heavily
more distortion
supersaturated with carbon. This of Fe lattice =
gives solid solution hardening. harder
• This is further enhanced by the martensite
distortion of the lattice, as many
interstitial holes contain C atoms more carbon
that are too large to fit.
• The higher the C content, the
greater the distortion of the lattice,
and the higher the hardness.

It is also extremely brittle. For practical purposes, it can be considered to have

negligible fracture toughness (< 5 MPa.m1/2).

3.5 Quenching and tempering of carbon steels

Martensite is rarely used in its as-quenched form: the low toughness means the
component would shatter easily. Hence, it needs to be tempered: re-heated to a
temperature of order 450-600°C (still in the two-phase α + Fe3C region on the phase
diagram) to encourage diffusion.

Tempering has the following effects:

• the yield strength reduces: after tempering 𝜎𝑦 ≈ 450 - 800 MPa
• the fracture toughness significantly increases.

20
This is due to the microstructural changes that takes place during tempering:
• The carbon in supersaturated solid solution diffuses to form Fe3C precipitates.
• The Fe lattice changes from martensite to its equilibrium state: undistorted BCC.
(a) (b)
Fe – 0.9 wt% C steel:
(a) quenched (martensite)
(b) quenched & tempered
(ferrite + cementite, Fe3C)

Images: DoITPoMS micrograph library: https://www.doitpoms.ac.uk/miclib/full_record.php?id=318

As the tempering time increases, the

Fe3C precipitates coarsen, giving a
steady fall in strength.

The tempering temperature also

influences the softening rate (faster
diffusion means faster coarsening), and
the volume fraction of precipitates
formed (from the phase diagram).

The fine-scale dispersion of

carbides in tempered
martensite provides excellent
precipitation hardening. It
results in a yield strength 2-3
times greater than slow
cooled (normalised) steel, but
with comparable fracture
toughness.
temper

Component cross-sections
can therefore be reduced
(saving weight), or design quench
loads increased.

For further details see

experiment M2.

21
3.6 Hardenability
A key consideration for steels subjected to thermal cycles is the hardenability: how
easily a steel forms martensite on cooling. Measures of hardenability:

• TTT diagram: the time taken for diffusive

phase transformations to begin (e.g. ferrite,
pearlite C-curves).
higher hardenability
= longer transformation times

• CCT diagram: the critical cooling rate, the

slowest cooling rate that gives 100%
martensite (i.e. just missing the nose of the
C-curves).
higher hardenability
= slower CCR

• The largest bar diameter that will give 100%

martensite at its centre after quenching:
higher hardenability
= larger diameter

Heat flow analysis is used to relate

cooling rate and component geometry.
Standard geometries are used to
characterize the response of different
steels: e.g. long cylinders, quenched in
water, oil or air.

When do you want a steel with high hardenability?

• When you want to form martensite easily, typically when you want to quench and
temper the steel, e.g. for cutting tools (you need high hardness and toughness).
• High hardenability means a larger component can be “through hardened”, i.e.
forming martensite through the whole cross-section that can then be tempered.
• High hardenability also means a component of a given size can be cooled more
slowly (e.g. by air cooling) while still achieving 100% martensite. This gives much
lower thermal stress, reducing thermal distortion and the risk of quench cracking
(noting that martensite is brittle). 22
When do you want a steel with low hardenability?
• When you do not want to form martensite on
cooling, for example: welding.
• Fusion welding involves traversing a heat source
(flame, arc or laser) along the joint line between
two components
• The thermal cycle will form austenite in the heat
affected zone (HAZ) close to the weld metal (the
region that melts); this could form martensite on
cooling.
• In this context, forming martensite is undesirable
– it is brittle, risking weld failure by fast fracture.
high hardenability
= poor weldability

Factors affecting the hardenability of steels:

(1) Carbon concentration:

• Increasing the C content increases the hardenability, by delaying the diffusional
phase transformations (see above).
• However, the effect is relatively weak: even eutectoid steel requires a bar diameter
< 30mm (water quenched) to form 100% martensite in the centre. Heat-treatable
steel components can be considerably larger than this.

(2) Further alloying:

• This increases the
hardenability by
delaying the diffusional alloy
phase transformations. additions
• If up to 5 wt% of various
alloying elements (Ni, Cr,
Mo, V, W) are added,
these are known as low
alloy steels. hypo-eutectoid steel low alloy steel

• For the γ → α phase transformation to take place, these alloying elements need
to be rearranged by diffusion (as they have a different solubility in FCC austenite vs
BCC ferrite).
• This diffusion of substitutional solute elements takes time, and hence shifts the C-
curves to longer times.
23
High alloy steels for cutting tools:

• As well as needing high hardenability (to allow

quenching and tempering) tools for high-speed
metal-cutting often operate at high temperatures
(due to the heat generated by plastic
deformation).
• In plain C steel, the Fe3C precipitates may coarsen
(or re-dissolve), leading to softening.

To prevent this, high alloy steels are used (containing Mo, W and V – up to 20 wt% in
total). These elements perform multiple functions:
• increasing the hardenability;
• the quench and temper treatment also leads to the formation of alloy carbides
(Mo2C, W2C, VC) – these are stable at high temperatures, retaining precipitation
hardening in service;
• contribute to high temperature solid solution hardening.

3.7 Summary of the learning outcomes: heat treatment of steels

After completing section 3 you will be able to do the following:

1. Use the phase diagram to explain the microstructure, and hence properties, of slow
cooled steel.
2. Use the TTT diagram to explain the effect of transformation temperature and time
on the microstructure and properties of steel.
3. Describe the microstructure and properties that result from quenching and
tempering steel.
4. Explain what is meant by the ‘hardenability’ of steel, why it is important in steel
processing, and how it can be influenced.

24
3.8 Quiz W5.2: heat treatment of steels
Three steel samples are compared:
Sample A: 100% pearlite, hardness HV 260
Sample B: 50% pearlite and 50% ferrite, hardness HV 210
Sample C: 100% martensite, hardness HV 650
A B C

Which of the following statements are correct (more than one can be correct)?

1) A is harder than B because it has been cooled more rapidly.

2) C is harder than A because it has a higher carbon concentration.

3) C has the highest hardness because it has undergone a displacive rather than a
diffusive phase transformation.

4) A, B and C can be tempered to increase their ductility.

Additional worked examples: heat treatment of steels

For further explanation, a video will be available with the following worked examples,
which help with Examples Paper 4 Q2-4:
• Tripos Question 2008 Q2.
• Tripos Question 2016 Q4(d).

Now attempt: Examples Paper 4, Q.2-4

25
4. Diffusion in heat treatment processes
Diffusion plays an important role in heat treatment processes. Two specific examples
will be considered in this section:

• Modelling thermal history: As noted in the previous

sections, the heat treatment of metal alloys is dependent on
thermal history, in particular the cooling rates. There are
direct parallels between the diffusion of atoms (covered in
earlier lectures) and the conduction of heat. Analogous
solutions to the diffusion equations can therefore be applied.
• Surface hardening: In many practical heat treatment
problems, we want to achieve different properties for the
surface compared to the bulk of a component. For example,
to deliver improved wear resistance, without compromising
component toughness. Surface hardening processes depend
on both the diffusion of atoms and the diffusion of heat.

Before proceeding, we will recap some key diffusion results from earlier lectures. The
1D flux of atoms (atoms s-1 m-2) is given by Fick’s first law:
𝜕𝐶
𝐽 = −𝐷
𝜕𝑥
where 𝐶(𝑥) is the concentration, and 𝐷 = 𝐷0 exp(−𝑄/𝑅𝑇) is the diffusion
coefficient, which is dependent on the temperature, 𝑇. The governing equation for
unsteady 1D diffusion is Fick’s second law:
𝜕𝐶 𝜕2𝐶
=𝐷 2
𝜕𝑡 𝜕𝑥
which can be solved to find the concentration variation with position and time, 𝐶(𝑥, 𝑡).

Two solutions have been presented so far.

(1) “Impulse response”: normal (or Gaussian) distribution:

𝑛0 𝑥2
𝐶(𝑥, 𝑡) = exp −
2 𝜋𝐷𝑡 4𝐷𝑡

This has the following properties:

• It tends to a delta function as t → 0,
and has a standard deviation 2𝐷𝑡 .
• The constant 𝑛0 is set by the initial
conditions: it is the total number of atoms
per unit area (i.e. the area under the curve)
which remains constant for all values of 𝑡 .

26
 (2) “Step response”: error function (erf) distribution:
𝑥
𝐶(𝑥, 𝑡) = 𝐶1 erf
2 𝐷𝑡

This has the following properties:

• The value of erf ∞ = 1 and erf 0 = 0 .
• 𝐶 𝑥, 𝑡 tends to a step function of amplitude
𝐶1 as 𝑡 → 0, and tends to zero as 𝑡 → ∞ .
• The ‘width’ of the step is of the order 𝐷𝑡 .
• Note that this solution can be re-scaled to
the case with non-zero concentration at
𝑥 = 0 , i.e. 𝐶 𝑥 = 0, 𝑡 = 𝐶0 .
𝑥
𝐶 𝑥, 𝑡 = 𝐶0 + (𝐶1 − 𝐶0 ) erf
2 𝐷𝑡

In the following, we will make use of a third solution of the diffusion equation: when
the initial conditions are a sinusoidal function.
2𝜋𝑥
𝐶(𝑥, 𝑡 = 0) = 𝐶1 sin
𝑤

To solve this, take a trial solution of the

form:
2𝜋𝑥
𝐶(𝑥, 𝑡) = 𝐶1 𝑓(𝑡) sin
𝑤
Here, 𝑓(𝑡) is an unknown time-
dependent amplitude. Substituting this
trial solution into Fick’s second law:

2
𝜕𝐶 𝜕 2𝐶 𝑑𝑓 2𝜋
=𝐷 2 ⇒ +𝐷 𝑓=0
𝜕𝑡 𝜕𝑥 𝑑𝑡 𝑤

This has the solution: 𝑓 𝑡 = exp −𝑡/𝜏 , 𝜏 = 𝑤 2 /4𝜋 2 𝐷

Therefore, a solution to the diffusion equation with sinusoidal initial conditions is:
2𝜋𝑥 4𝜋 2 𝐷𝑡
𝐶(𝑥, 𝑡) = 𝐶1 sin exp −
𝑤 𝑤2

This solution can also be rescaled for the case where 𝐶 𝑥 = 0, 𝑡 = 𝐶0 , i.e.
2𝜋𝑥 4𝜋 2 𝐷𝑡
𝐶 𝑥, 𝑡 = 𝐶0 + (𝐶1 −𝐶0 ) sin exp −
𝑤 𝑤2

27
4.1 Modelling thermal history
So far, we have looked at the diffusion of atoms. However, by comparing this with heat
transfer by conduction, we can show that the governing equations are the same.
Therefore, the same solutions can be used, with a change of variables.

diffusion of atoms thermal conduction

𝐶 = concentration (atoms m-3) 𝑇 = temperature (K)

𝜕𝐶 𝜕𝑇
Fick’s first law: 𝐽 = −𝐷 Fourier’s law: 𝑞 = −𝜆
𝜕𝑥 𝜕𝑥
𝐽 = atom flux (atoms s-1 m-2 ) 𝑞 = heat flux ( W m-2 )
𝐷 = diffusion coefficient (m2 s-1 ) 𝜆 = thermal conductivity ( W m-1 K-1)

Conservation of mass: Conservation of energy:

net flow of atoms in = change in net flow of energy in = mass x specific

concentration x volume heat capacity x change in temperature

𝜕𝐽 𝜕𝑞
𝐽𝛿𝑡 − 𝐽 + 𝛿𝑥 𝛿𝑡 = 𝛿𝐶𝛿𝑥 𝑞𝛿𝑡 − 𝑞 + 𝛿𝑥 𝛿𝑡 = 𝜌𝛿𝑥 𝑐 𝛿𝑇
𝜕𝑥 𝜕𝑥
𝜕𝐶 𝜕𝐽 𝜕2𝐶 𝜕𝑇 1 𝜕𝑞 𝜆 𝜕2𝑇
∴ =− =𝐷 2 ∴ =− =
𝜕𝑡 𝜕𝑥 𝜕𝑥 𝜕𝑡 𝜌𝑐 𝜕𝑥 𝜌𝑐 𝜕𝑥 2

𝑐 = specific heat capacity (J kg-1 K)

𝜆/𝜌𝑐 = 𝑎 = thermal diffusivity (m2 s-1)

The governing equations for mass and thermal diffusion have the same form. The
solutions are therefore interchangeable, with the change of variables:
• concentration 𝐶 for temperature 𝑇
• diffusion coefficient 𝐷 for the thermal diffusivity 𝑎 (note that these have the same
units, m2 s-1)

28
4.2 Example: thermal history of a quenched plate

Consider quenching a steel plate of thickness 𝐿,

from a uniform initial temperature 𝑇1 , into a
large bath of liquid at temperature 𝑇0 .
• If we assume perfect heat transfer, the
boundary condition at the surfaces is:
𝑇 = 𝑇0 for 𝑥 = 0, 𝐿 for all times 𝑡 ≥ 0
• Also, assume the plate is large, so heat flow
is 1D: 𝑇(𝑥, 𝑡).

The aim is to find a solution for 𝑇(𝑥, 𝑡) . This in turn can be used to provide information
on cooling rates, and hence the likely phase transformations.

Case 1 – thick plate limit

If we assume that dimension 𝐿 is much larger than the distance over which heat can
diffuse (i.e. for short times, a low thermal diffusivity material or a very thick plate), we
can treat the plate as effectively infinite in thickness.
• Recall that a ‘representative’ diffusion distance for atoms in time 𝑡 is given by: 𝐷𝑡
• The equivalent distance for thermal diffusion is given by: 𝑎𝑡
• So, the thick plate limit is reasonable when: 𝐿 ≫ 𝑎𝑡

The initial conditions are therefore a step

function, with amplitude (𝑇1 − 𝑇0 ), and
𝑇 𝑥 = 0, 𝑡 = 𝑇0 .

The correct solution for this problem is

therefore the error function, scaled to these
initial conditions:
𝑥
𝑇 𝑥, 𝑡 = 𝑇0 + (𝑇1 − 𝑇0 ) erf
2 𝑎𝑡

We can use this to calculate the cooling rates in

the plate (e.g. to see where martensite may
form). From the materials data book:
𝑑 2
erf 𝑍 = exp −𝑍 2
𝑑𝑍 𝜋
𝑇1 − 𝑇0 = 100 °C
𝜕 𝑑 𝜕𝑍
Taking 𝑍 = 𝑥/2 𝑎𝑡 and [ ] = 𝑑𝑍 :
𝜕𝑡 𝜕𝑡

𝜕𝑇 𝑥, 𝑡 𝑇1 − 𝑇0 𝑥 𝑥2
=− exp −
𝜕𝑡 𝑡 2 𝜋𝑎𝑡 4𝑎𝑡

29
Case 2 – finite width plate
A different solution is required if the
thermal diffusion distances are
significant compared to the plate
thickness:

𝐿~ 𝑎𝑡

The way to solve this is to model the initial conditions as the first half wavelength of a
square wave function. This can then be expressed as a Fourier series: a sum of
sinusoidal functions. Each term in that sum can be solved using the sinusoidal solution
outlined above.

= +

wavelength
+
higher modes

The equivalent square wave parameters for the plate quenching problem are:
• wavelength 2𝐿 (but only over the domain of one half wavelength: 0 < 𝑥 < 𝐿),
• mean temperature of 𝑇0 ,
• mean-to-peak amplitude scaled to the temperature difference (𝑇1 −𝑇0 ) .
Hence, the initial conditions can be described by the Fourier series in 𝑥 :
4 𝑛𝜋𝑥
𝑇 𝑥, 𝑡 = 0 = 𝑇0 + 𝑇1 − 𝑇0 ෍ sin
𝑛𝜋 𝐿
𝑛=1,3,5,…

(Maths Databook – note the change of variable from time 𝑡 to a Fourier series in 𝑥)

30
To find a solution to the diffusion equation for these initial conditions, consider each
sinusoidal term in the summation separately. Term 𝑛 has initial conditions :

4 𝑛𝜋𝑥
𝑇𝑛 𝑥, 𝑡 = 0 = 𝑇1 − 𝑇0 sin
𝑛𝜋 𝐿
Next, find a solution that satisfies this initial condition. As before, take a trial solution:
4 𝑛𝜋𝑥
𝑇𝑛 𝑥, 𝑡 = 𝑇1 − 𝑇0 𝑓𝑛 𝑡 sin
𝑛𝜋 𝐿

Here, 𝑓𝑛 (𝑡) is the unknown time-dependent amplitude of mode 𝑛. Substituting this

trial solution into the diffusion equation:

𝜕𝑇𝑛 𝜕 2 𝑇𝑛 𝑑𝑓𝑛 𝑛𝜋 2
=𝑎 ⇒ +𝑎 𝑓𝑛 = 0
𝜕𝑡 𝜕𝑥 2 𝑑𝑡 𝐿
𝑡 𝐿2
This has the solution: 𝑓𝑛 𝑡 = exp − 𝜏𝑛 = 2 2
𝜏𝑛 𝑛 𝜋 𝑎
If each term is a solution to the diffusion equation, then the sum will also be a
solution. The end result is therefore:

4 𝑛𝜋𝑥 𝑛2 𝜋 2 𝑎𝑡
𝑇 𝑥, 𝑡 = 𝑇0 + 𝑇1 − 𝑇0 ෍ sin exp −
𝑛𝜋 𝐿 𝐿2
𝑛=1,3,5,…

This result gives some insights into the plate quenching problem:

(1) The time constant 𝜏𝑛 for each mode is

dependent on the parameter 𝑳𝟐 /𝒂 .
Therefore, reducing the plate thickness 𝐿
will significantly increase cooling rates.

(2) The time constant 𝜏𝑛 is also very sensitive

to the mode number 𝒏 (with 𝑛 = 1, 3, 5, …).
The higher modes will decay much faster.
So, the solution can (after a short period) be
well approximated by just the first term:

4 𝜋𝑥 𝜋 2 𝑎𝑡
𝑇 𝑥, 𝑡 ≈ 𝑇0 + 𝑇1 − 𝑇0 sin exp − 2
𝜋 𝐿 𝐿

31
4.3 Surface hardening
Many steel components used in machinery (e.g. gears,
bearings, tool edges) are subjected to high surface
stresses, due to friction and sliding contact. Surface
engineering is used to harden the surface of the
component (‘case hardening’) to increase its hardness
and wear resistance – without damaging the strength
and toughness of the underlying steel

This can be achieved using a combination of the following two processes, which both
rely on diffusion:

(1) Carburising:
This technique involves immersing the steel at high temperature in a carbon-rich
atmosphere (such as molten cyanide or nitrate salts).
• Carbon diffuses into the surface of the
component, prior to any heat treatment.
• The higher C content increases the
hardenability, making it easier to form a
surface layer of martensite, which may
then be tempered.
• The higher C content will also make the
martensite harder, due to the greater
supersaturation with carbon.

The microstructure of a steel (0.15 wt% C) that has been carburised and then
quenched is shown below:
extra carbon has diffused into the surface

Higher wt% C at surface:

• higher hardenability
• forms martensite on
cooling

Image: © DoITPoMS Micrograph Library,

Lower wt% C: University of Cambridge (Dr R F Cochrane)
• low hardenability
• high proportion of ductile ferrite
32
(2) Transformation hardening:
This involves imposing a heat source to the surface, such that a thin layer is austen-
itised, and then cooling it quickly to form a layer of martensite (which can then be
tempered).
• Surface heating may use a traversing flame, laser, or electron gun; or high frequency
induction coils are used to induce surface eddy currents.
• Air cooling may be sufficient to form martensite – or a water-quench can be used.

We will use the diffusion analysis to model process (1): carburising.

• Consider a steel component with initial
carbon concentration 𝐶0 . C (x = 0, all t) = Cs
• The component is placed in a carbon rich
environment at high temperature, which
results in a higher surface concentration, 𝐶𝑠 . time increasing
• The surface concentration 𝐶𝑠 is assumed to
remain constant (i.e. large carbon supply).
• Carbon will diffuse into the surface of the
steel component over time, at a rate
dependent on temperature and time. C (all x, t = 0) = C0

The initial conditions are therefore an inverted step function, with amplitude (𝐶𝑠 − 𝐶0 ),
shifted by an initial concentration 𝐶0 .
• The solution for these initial conditions was introduced earlier in the course (the pre-
deposition stage in Si doping): 𝐶 𝑥, 𝑡 is a function of 1 − erf 𝑥/2 𝐷𝑡 .
• Adding the offset 𝐶0 to that solution, and scaling the amplitude to (𝐶𝑠 − 𝐶0 ), gives:

𝑥
𝐶 𝑥, 𝑡 = 𝐶0 + 𝐶𝑠 − 𝐶0 1 − erf
2 𝐷𝑡

33
We can use this solution to calculate the time required to achieve a particular ‘case
depth’ for the case hardening process, i.e. a particular thickness of hardened layer:

• The hardness of the steel will increase

with the carbon concentration, 𝐶.
• Hence, for case hardening, it is 𝐶 𝑥 at the minimum
necessary to specifying a minimum carburising time
concentration 𝐶min at a specified
depth into the material: the case 𝐶min
depth 𝑥case .
• This will set the depth and minimum
hardness of the hardened layer.
𝑥case

Substituting these values into the solution above, the time 𝑡case to achieve this
case depth is given by the relationship:

𝑥case 𝐶min − 𝐶0
erf =1−
2 𝐷𝑡case 𝐶𝑠 − 𝐶0

Before this can be solved for 𝑡case , we need to evaluate the concentrations 𝐶𝑠 and 𝐶0 .

• The concentration 𝐶0 is the initial solubility range

composition of the steel being carburising temp.
hardened.
• Assuming a carbon rich
environment, 𝐶𝑠 will be the
maximum amount of carbon that
can be accommodated in solid
solution at the carburising
temperature.
• This is set by the solubility limit of
the steel at the carburising
temperature, from the phase
diagram. 𝐶0 𝐶𝑠

The only unknown now is 𝑡case . Tables in the Materials Data Book give values erf 𝑌 vs
𝑌. Hence, 𝑥case /2 𝐷𝑡case can be evaluated. After working out the diffusion
coefficient 𝐷 = 𝐷0 exp(−𝑄/𝑅𝑇) at the carburising temperature 𝑇, the time 𝑡case can
be found.

34
4.4 Summary of the learning outcomes: diffusion in heat treatment
After completing section 4 you will be able to do the following:

1. Describe the analogy between thermal and atomic diffusion, and how to switch
diffusion equations and solutions from one to the other via a change of variables.
2. Compare characteristic diffusion distances and times for thermal and atomic
diffusion.
3. Use appropriate solutions to the 1D diffusion equations to analyse thermal
diffusion in heat treatment processes.
4. Describe the surface hardening processes of carburisation and transformation
hardening, and analyse them using solutions to the 1D diffusion equations.

4.5 Quiz W6.1: diffusion in heat treatment

A large steel plate of thickness 𝑤 is taken from a furnace and quenched in water. The
steel has thermal diffusivity 𝑎 and diffusion coefficient 𝐷 for carbon in austenite.

Which of the following must be true in order to use the error function (erf) is to model
the temperature vs depth below the surface, 𝑇 𝑥, 𝑡 (more than one can be correct)?

1) 𝑤 ≈ 𝑎𝑡

2) 𝑤 ≪ 𝑎𝑡

3) 𝑤 ≫ 𝑎𝑡

4) 𝑤 2 /𝑎 ≪ 𝑤 2 /𝐷

Additional worked examples: diffusion in heat treatment

For further explanation, a video will be available with the following worked examples:
• Tripos Question 2017 Q5, which helps with Examples Paper 4 Q5.
• Tripos Question 2012 Q6, which helps with Examples Paper 4 Q6

Now attempt: Examples Paper 4, Q.5-6

35