Tempered Martensite

H. K. D. H. Bhadeshia


Tempering is a term historically associated with the heat treatment of martensite in steels. It describes how the microstructure and mechanical properties change as the metastable sample is held isothermally at a temperature where austenite cannot form. The changes during the tempering of martensite can be categorised into stages. During the first stage, excess carbon in solid solution segregates to defects or forms clusters within the solid solution. It then precipitates, either as cementite in low-carbon steels, or as transition iron-carbides in high-carbon alloys. The carbon concentration that remains in solid solution may be quite large if the precipitate is a transition carbide. Further annealing leads to stage 2, in which almost all of the excess carbon is precipitated, and the carbides all convert into more stable cementite. Any retained austenite may decompose during this stage. Continued tempering then leads to the coarsening of carbides, extensive recovery of the dislocation structure, and finally to the recrystallisation of the ferrite plates into equiaxed grains.

This is a useful description but it is revealing to consider first, the factors responsible for driving the process in the first place.

Deviation from Equilibrium

Tempering is a process in which the microstructure approaches equilibrium under the influence of thermal activation. It follows that the tendency to temper depends on how far the starting microstructure deviates from equilibrium. It is interesting therefore to consider how metastable a material can be, before dealing specifically with martensite. Turnbull characterised metastability in terms of the unit RTm where R is the universal gas constant and Tm is the absolute melting temperature. This coarse unit is a measure of the thermal energy in the system at the melting temperature; it represents a large amount of energy, typically in excess of 20,000 J mol-1.

Table 1: Degree of metastability
Metastable structure RTm
Supersaturated solutions 1
Amorphous metal 0.5
Modulated films, nanostructures 0.1

Supersaturated solutions are prominent in this list and the extent of metastability depends both on the excess concentration and on the equilibrium solubility. It can be demonstrated that excess carbon which is forced into solution in martensite is the major contributor to the stored energy of martensite.

The calculations presented in Table 2 show the components of the stored energy of martensite in a typical low--alloy martensitic steel Fe-0.2C-1.5Mn wt%. It is necessary to define a reference state, which is here taken to be an equilibrium mixture of ferrite, graphite and cementite, with a zero stored energy. Graphite does not in fact form because it is too slow to precipitate; the effect of replacing the graphite with cementite is to increase the stored energy by some 70 J mol-1.

When transformations occur at low temperatures, it is often the case that substitutional elements like manganese and iron cannot diffuse during the time scale of the experiment, whereas carbon is still mobile. The transformation then happens in such a way that the Fe/Mn ratio is maintained constant whilst the carbon redistributes subject to this constrain, until its chemical potential becomes uniform. This is known as paraequilibrium. Unlike the equilibrium state, because the iron and manganese atoms are trapped during transformation, their chemical potentials are no longer uniform. This adds a further 315 J mol-1 to the stored energy.

When bainite forms, the transformation mechanism is displacive, there is a shape deformation, which leads to an additional 400 J mol-1 of stored energy. Since there is no diffusion during transformations, but the carbon partitions following growth, the total stored energy is that for the paraequilibrium state added to the strain energy term, giving a net value of 785 J mol-1.

Martensite is not only a diffusionless transformation, but it frequently occurs at low temperatures where its virgin microstructure is preserved. Even the carbon remains trapped in the product crystal. Furthermore, the strain energy term associated with martensite is greater at about 600 J mol-1 because the plates tend to have a larger aspect ratio (thickness/length). There may also be twin interfaces within the martensite plates, which cost about 100 J mol-1. The trapping of carbon inside the martensite adds a further 629 J mol-1, which makes the total stored energy in excess of 1700 J mol-1!

Table 2: Stored energies of a variety of microstructures
Phase Mixture in Fe-0.2C-1.5Mn wt% at 300 K Stored Energy / J mol-1
1. Ferrite, graphite and cementite 0
2. Ferrite and cementite 70
3. Paraequilibrium ferrite and paraequilibrium cementite 385
4. Bainite and paraequilibrium cementite 385+400=785
5. Martensite 385+600+100+629=1714
6. Mechanically alloyed ODS metal 55

The stored energy becomes even larger as the carbon concentration is increased (Figure 1).

Figure 1: The free energy due to the trapping of carbon in martensite, as a function of its carbon concentration. The results are for a temperature of 473 K.

free energy

Virgin Microstructure

The virgin microstructure obtained immediately after quenching from austenite consists of plates or laths of martensite which is supersaturated with carbon. In the vast majority of steels, the martensite contains a substantial density of dislocations which are generated during the imperfect accommodation of the shape change accompanying the transformation. The plates may be separated by thin films of retained austenite, the amount of untransformed austenite becoming larger as the martensite-start temperature MS is reduced.

(a) Transmission electron micrograph of as-quenched martensite in a Fe-4Mo-0.2C wt% steel. The mottled contrast within the plates is due to a high density of dislocations. (b) Corresponding dark-field image showing the distribution of retained austenite.

Carbon Atoms

Carbon is an interstitial atom in ferritic iron, primarily occupying the octahedral interstices. There are three such interstices per iron atom. At a typical concentration of 0.4 wt% or about 2 at%, less than 1% of these interstices are occupied by carbon. Furthermore, there is a strong repulsion between carbon atoms in nearest neighbour sites. This means that carbon atoms almost always have an adjacent interstitial site vacant, leading to a very high diffusion coefficient when compared with the diffusion of substitutional solutes. In the latter case, the substitutional vacancy concentration is only 10-6 at temperatures close to melting, and many orders of magnitude less at the sort of temperatures where martensite is tempered. It follows that carbon diffuses much faster than substitutional atoms (including iron), as illustrated below.

carbon in ferritic iron diffusion in iron

(a) A carbon atom in an octahedral interstice in body-centered cubic iron.

(b) The ratio of the diffusivity of a substitutional atom to that of carbon in body-centered cubic iron.

Given that carbon is able to migrate in martensite even at ambient temperature, it is likely that some of it redistributes, for example by migrating to defects, or by rearranging in the lattice such that the overall free energy is minimised.

Precipitation of Iron Carbides

In high-carbon steels, the precipitation of excess carbon begins with the formation of a transition carbide, such as ε (Fe2.4C). ε-carbide can grow at temperatures as low as 50oC. Indeed, most of the iron carbides can precipitate at low temperatures, well below those associated with the motion of substitutional solutes. This is because they grow by a displacive mechanism which does not require the redistribution of substitutional atoms (including iron); carbon naturally has to partition. This corresponds to a process known as paraequilibrium transformation in which the iron to substitutional solute ratio is maintained constant but subject to that constraint, the carbon achieves a uniform chemical potential.

Martensite is said to be supersaturated with carbon when the concentration exceeds its equilibrium solubility with respect to another phase. However, the equilibrium solubility depends on the phase. The solubility will be larger when the martensite is in equilibrium with a metastable phase such as ε carbide. Some 0.25 wt% of carbon is said to remain in solution after the precipitation of ε-carbide is completed.

Although most textbooks will begin a discussion of tempering with this first stage of tempering, involving the redistribution of carbon and precipitation of transition carbides, cementite can precipitate directly. This is particularly the case when the defect density is large. Trapped carbon atoms will not precipitate as transition carbides but cementite is more stable than trapped carbon.

(a) Transmission electron micrograph of martensite in a Fe-4Mo-0.2C wt% steel after tempering at 190oC for 1 hour. The carbon has in this case precipitated as fine particles of cementite. (b) Corresponding dark-field image showing the distribution of retained austenite, which has not been affected by the tempering.

Decomposition of Retained Austenite

tempered martensite

Tempering at higher temperatures, in the range 200-300oC for 1 h induces the retained austenite to decompose into a mixture of cementite and ferrite. When the austenite is present as a film, the cementite also precipitates as a continuous array of particles which have the appearance of a film.

Dark field transmission electron micrograph of martensite in a Fe-4Mo-0.2C wt% steel after tempering at 295oC for 1 hour. Only the cementite is illuminated. The film of cementite at the martensite plate boundaries is due to the decomposition of retained austenite.

Further Tempering

Tempering at even higher temperatures leads to a coarsening of the cementite particles, with those located at the plate boundaries growing at the expense of the intra-plate particles. The dislocation structure tends to recover, the extent depending on the chemical composition. The recovery is less marked in steels containing alloying elements such as molybdenum and chromium.

Bright field transmission electron micrograph of martensite in a Fe-4Mo-0.2C wt% steel after tempering at 420oC for 1 hour.

tempered martensite

The recovery of the dislocation structure and the migration of dislocation-cell and martensite boundaries leads not only to a coarsening of the plates, but also an increase in the crystallographic misorientation between adjacent plates, as illustrated in the adjacent figure. The data are from Suresh et al., Ironmaking and Steelmaking 30 (2003) 379-384.

tempered martensite

Precipitation of Alloy Carbides

diffusion distance

Alloy carbides include M2C (Mo-rich), M7C3, M6C, M23C6 (Cr-rich), V4C3, TiC etc., where the 'M' refers to a combination of metal atoms.

However, all of these carbides require the long-range diffusion of substitutional atoms. They can only precipitate when the combination of time and temperature is sufficient to allow this diffusion. The figure on the left shows the calculated diffusion distance in ferrite for a tempering time of 1 h. It is evident that the precipitation of alloy carbides is impossible below about 500oC for a typical tempering time of 1 h; the diffusion distance is then just perceptible at about 10 nm.

The alloy carbides grow at the expense of the less stable cementite. If the concentration of strong carbide forming elements such as Mo, Cr, Ti, V, Nb is large then all of the carbon can be accommodated in the alloy carbide, thereby completely eliminating the cementite.

tempered martensite tempered martensite

Fe-0.1C-1.99Mn-1.6Mo wt% quenched to martensite and then tempered at 600oC. (photograph courtesy of Shingo Yamasaki). The bright field transmission electron micrograph is of a sample tempered for 560 h, whereas the dark-field image shows a sample tempered for 100 h.

The precipitates are needles of Mo2C particles. The needles precipitate with their long directions along <100>α.

tempered martensite tempered martensite

Fe-0.1C-1.99Mn-0.56V wt% quenched to martensite and then tempered at 600oC for 560 h (photograph courtesy of Shingo Yamasaki).

The precipitates are plates of V4C3 particles which precipitate on the {100}α planes.

Severe Tempering

tempered bainite

Fe-0.98C-1.46Si-1.89Mn-0.26Mo-1.26Cr-0.09V wt% tempered at 730oC for 7 days (photograph courtesy of Carlos Garcia Mateo). This transmission electron micrograph shows large cementite particles and a recovered dislocation substructure. There are sub-grain boundaries due to polygonisation and otherwise clean ferrite almost free from dislocations. The plate microstructure is coarsened but nevertheless retained because the carbides are located at plate boundaries.

An alloy such as this, containing a large fraction of carbides is extremely resistant to tempering. The original microstructure was bainitic, but similar results would be expected for martensite.

tempered bainite

Fe-0.98C-1.46Si-1.89Mn-0.26Mo-1.26Cr-0.09V wt% tempered at 730oC for 21 days (photograph courtesy of Carlos Garcia Mateo). The optical micrograph shows some very large spherodised cementite particles. The ferrite has completely recrystallised into equiaxed grains.

An alloy such as this, containing a large fraction of carbides is extremely resistant to tempering. The original microstructure was bainitic, but similar results would be expected for martensite.


tempered martensite

Fe-0.35C-Mo wt% alloy quenched to martensite and then tempered at the temperature indicated for one hour (data from Bain's Alloying Elements in Steels). Whereas the plain carbon steel shows a monotonic decrease in hardness as a function of tempering temperature, molybdenum in this case leads to an increase in hardness once there is sufficient atomic mobility to precipitate Mo2C.

Secondary hardening is usually identified with the tempering of martensite in steels containing strong carbide forming elements like Cr, V, Mo and Nb. The formation of these alloy carbides necessitates the long--range diffusion of substitutional atoms and their precipitation is consequently sluggish. Carbides like cementite therefore have a kinetic advantage even though they may be metastable. Tempering at first causes a decrease in hardness as cementite precipitates at the expense of carbon in solid solution, but the hardness begins to increase again as the alloy carbides form. Hence the term secondary hardening. Coarsening eventually causes a decrease in hardness at high tempering temperatures or long times, so that the net hardness versus time curve shows a secondary hardening peak.


tempered martensite

Typical time scales associated with the variety of processes that occur during tempering. The actual rates depend on the alloy composition.

Elements such as silicon and aluminium have a very low solubility in cementite. They greatly retard the precipitation of cemenite, thus allowing transition iron-carbides to persist to longer times.

Case Studies: AerMet 100

AerMet 100 is a martensitic steel which is used in the secondary-hardened condition; its typical chemical composition is as follows:

Composition of Aermet 100 in wt%
C Co Ni Cr Mo Mn Si Al Ti S P
0.23 13.4 11.1 3.0 1.2 0.03 0.03 0.004 0.013 0.001 0.003

The cobalt plays a key role in retarding the recovery of martensite during tempering, thereby retaining the defect structure on which M2C needles can precipitate as a fine dispersion. By increasing the stability of body-centred cubic iron, it also reduces the tendency of martensite to revert to austenite during tempering. The carbon concentration is balanced such that all the cementite is replaced by the much finer alloy carbides during secondary hardening.

Impurity concentrations and inclusions are kept to a minimum by vacuum induction melting and vacuum arc refining. Unlike conventional steels, the manganese and silicon concentrations are also kept close to zero because both of these elements reduce the austenite grain boundary cohesion.

The steel is VIM/VAR double-melted and forged or rolled into the final form.

The as-received steel is usually "homogenised" at 1200oC for 8 hours. This is because the cast and forged alloy contains banding due to chemical segregation. Austenitisation is at about 850oC for 1 h, followed by quenching in oil to ambient temperature and cryogenic treatment to reduce the amount of retained austenite from some 2% to less than the detection limit. The sample is then tempered in the range 500-600oC, depending on the properties required. Since the Ae1 temperature is about 485oC, thin films of nickel-rich austenite grow during tempering. The films are apparently beneficial to the mechanical properties.

The optimum combination of strength and toughness is obtained by tempering at 470oC. The as-quenched steel has a martensitic microstructure with a few undissolved MC (5-12 nm) and M23C6-type carbides (20-100 nm). The high toughness (about 160 MPa m1/2) in the as-quenched state is believed to be due to the low strength, the cleanliness of the steel and the fact that the undissolved carbides are spherical. It has been suggested that the toughness in this state can be further improved by refining the M23C6 particle size; since the steel is not used in the as-quenched condition, the significance of this result is in emphasising the need for cleanliness. Any inclusions must clearly be smaller than the M23C6 particle size-range.

Tempering at 430oC, 5 h is associated with a minimum in toughness because of the precipitation of relatively coarse cementite platelets in a Widmanstätten array. An increase in the tempering temperature to 470oC leads to the coherent precipitation of needle--shaped molybdenum--rich zones, and a peak in the strength; the precipitation occurs at the expense of the cementite particles, so the increase in strength is also accompanied by a large increase in toughness. The formation of austenite films may also contribute to the toughness.

Further tempering leads to the precipitation of M2C carbides, recovery of the dislocation substructure, and a greater quantity of less stable reverted-austenite. The austenite that forms at higher temperatures has a lower nickel concentration and its instability is believed to be responsible for the decrease in toughness beyond about 470oC tempering, in spite of the decrease in strength.

tempered martensite tempered martensite
Strength of AerMet 100 as a function of tempering temperature, the tempering time being 5 h Corresponding toughness. Both figures are based on data from Ayers and Machmeier, Metall. and Mater. Trans. A, 24 (1993), 1943.

The following are pictures of the landing gears for the Airbus Industrie A330 and A340 passenger aircraft. This is the largest landing gear assembly in commercial service, presumably to be superceded by the A380. The critical components are made from tempered martensite.

DSC00259 DSC00256
DSC00257 DSC00258

Case Studies: Creep-Resistant Steels

Creep resistant steels must perform over long periods of time in severe environments. The typical service life is over a period of 30 years, at tempertures of 600°C or more, whilst supporting a design stress of 100 MPa. They are therefore required to resist both creep and oxidation. Their microstructures must clearly be stable in both the wrought and welded states. To resist thermal fatigue, the steel must have a small thermal expansion coefficient and an high thermal conductivity; ferritic steels are much better than austenitic steels with respect to both of these criteria.

The conditions described above correspond to low strain rates and relatively low temperatures. The mechanism of creep then involves the glide of slip dislocations. Diffusion-assisted dislocation climb in necessary for continued deformation when the glide process is obstructed, for example by the presence of precipitates in the glide plane. An applied stress assists the climb process via a force which tends to push the dislocation onto a parallel plane, such that it can by-pass the particle.

Dislocation creep of this kind can be resisted by introducing a large number density of precipitates in the microstructure. This basic principle leads to a large variety of heat--resistant steels. The ones with the lowest solute concentrations might contain substantial quantities of allotriomorphic ferrite and some pearlite, but the vast majority have bainitic or martensitic microstructures in the normalised condition. After normalising the steels are severely tempered to produce a "stable" microstructure consisting of a variety of alloy carbides in a ferritic matrix. The known precipitates are illustrated in the adjacent; they determine the microstructure and are crucial in the development of creep strain.


Case Studies: Pipes

Steels pipes for the extraction of oil require high-strength, resistance to hydrogen and H2S attack, fracture toughness and the ablility to be made as seamless pipes. Such pipes are frequently connected using threaded joints and are made by quenching and tempering. Those which serve in highly corrosive environments are secondary hardened (heat treated at a very high temperatures) whereas others are tempered at temperatures around 400°C. Click on the picture on the right to see how the pipes are made using a mandrel piercing mill.


Temper Embrittlement

Whereas tempering is frequently necessary to reduce the hardness of martensite and increase toughness, the heat-treatment can lead to embrittlement when the steel contains impurities such as phosphorus, antimony, tin and sulphur. This is because these impurities tend to segregate to the prior austenite grain boundaries and reduce cohesion across the boundary plane, resulting in intergranular failure.

There are three kinds of embrittlement phenomena associated with quenched and tempered steels, each of which leads either to a minimum in the toughness as a function of tempering temperature, or to a reduction in the rate at which the toughness improves as the tempering temperature is increased:

Role of Prior Austenite Grain Boundaries

Temper embrittlement phenomena are most prominent in strong steels where the applied stress can reach high magnitudes before the onset of plasticity. This is because strong steels are based on microstructures which evolve by the displacive transformation of austenite. Ordinary steels are ferritic or pearlitic; both of these phases can grow by reconstructive transformation across austenite grain boundaries. In doing so, they destroy the structure that exists at those boundaries and remove them as potential sources for the segregation of impurity atoms such as phosphorus.

By contrast, the coordinated motion of atoms accompanying displacive transformations cannot be sustained across austenite grain boundaries. Therefore, Widmanstätten ferrite, bainite, acicular ferrite and martensite are all confined by austenite grain boundaries. A vestige of the austenite grain boundary ( prior austenite grain boundary therefore remains in the microstructure when the transformations are displacive. This is illustrated schematically in the figure below, which shows austenite grain boundaries as hard barriers to martensite (α') whereas the allotriomorphs of ferrite (α) are able to consume the austenite boundaries on which they nucleate, by growing into both of the adjacent grains.

Figure: (a) Martensite (α') is hindered by austenite grain boundaries whereas allotriomorphic ferrite (α) is not. (b) Failure at the prior austenite grain boundary.

650°C Reversible Temper Embrittlement

Tempering at temperatures around 650o promotes the segregation of impurity elements such as phosphorous to the prior austenite grain boundaries, leading to intergranular failure along these boundaries. The reversibility arises because the impurity atmospheres at the grain boundaries can be evaporated by increasing the tempering temperature. Quenching from the higher temperature avoids the resegregation of impurities during cooling, thus eliminating embrittlement.

In fact, one of the tests for the susceptibility of bainitic microstructures to impurity-controlled embrittlement involves a comparison of the toughness of samples which are water quenched from a high tempering temperature (680o) with those cooled slowly to promote impurity segregation.

Studies of creep resistant bainitic steels show that phosphorus and tin, and to a lesser extent manganese and silicon, are all embrittling elements. Manganese is known to reduce intergranular fracture strength. Silicon, on the other hand, enhances the segregation of phosphorus to the austenite grain boundaries, and can itself cosegregate with nickel to the grain surfaces. There are also smaller effects due to arsenic, antimony and sulphur. The tendency for embrittlement correlates strongly with an empirical J (Bodnar and co-workers) factor:

J = Mn + Si + 104 (P + Sn)

where the concentrations of elements are in weight percent.

To summarise, the impurity-controlled temper embrittlement occurs in bainite as it does in martensite; after all, neither of these transformation products cross austenite grain surfaces and hence leave them open for impurity segregation. By comparison, reconstructive transformations products such as allotriomorphic ferrite, can grow across and consume the austenite grain surfaces, thereby removing them entirely from the final microstructure.

Finally, it is worth noting that although the science of the embrittlement is well understood, for reasons of cost, commercial steels always contain more impurities than is desirable. Steps must therefore be taken to mitigate the impurity effects, for example by alloying with molybdenum to pin down the phosphorus and prevent it from segregating.

300→350°C Temper Embrittlement

Fracture is again intergranular with respect to the prior austenite grain boundaries which become decorated with coarse cementite particles during tempering. At the same time, the grain boundaries are weakened by impurity segregation. The cementite particles crack under the influence of an applied stress and in this process concentrate stress at the weakened boundaries. These factors combine to cause embrittlement.

300→350°C Tempered-Martensite Embrittlement

This effect is common in clean steels, with fracture occurring transgranularly relative to the prior austenite grain boundaries. It is attributed to the formation of cementite particles at the martensite lath boundaries and within the laths. During tempering, the particles coarsen and become large enough to crack, thus providing crack nuclei which may then propagate into the matrix. As a consequence, untempered low--carbon martensitic steels sometimes have a better toughness than when they are tempered, even though the untempered steel is stronger. The cementite behaves like a brittle inclusion.

Both of the impurity-controlled embrittlement phenomena can be minimised by adding about 0.5 wt% molybdenum to the steel. The Mo associates with phosphorus atoms in the lattice thereby reducing mobility and hence the extent to which they segregate to boundaries. Larger concentrations of molybdenum are not useful because precipitation occurs.

In many bainitic microstructures, tempering even at temperatures as high as 550°C has only a small effect on cementite size and morphology. Consequently, the low--temperature embrittlement phenomena are not found in conventional bainitic microstructures.

Related Topics

Martensitic Phase Transformations

Transformation Hardening

Movies about Martensite and Bainite

Tempered martensitic steel band holding log in position.
Tempered martensitic steel band holding log in position.
Lucerne, Switzerland
Lucerne, Switzerland

Superalloys Titanium Bainite Martensite Widmanstätten ferrite
Cast iron Welding Allotriomorphic ferrite Movies Slides
Neural Networks Creep Mechanicallly Alloyed Theses

PT Group Home Materials Algorithms Any Valid CSS!