Some questions and answers on steels

H. K. D. H. Bhadeshia

I have been reading the Chou and Bhadeshia paper about grain control in the oxide-dispersion strengthened, mechanically alloyed steel MA957. MA957 has 14wt% Cr and it is possible to produce a little austenite
by heat treatment at 910-1010 C, because this alloy falls into the
gamma-loop for its chemical composition. Likewise, the addition of a
little nickel (0.5 wt%) is recommended.

As far as I know, the aim of introducing austenite is to refine the coarse grain structure. However, I am interested in a coarse grain structure for optimum creep porpouses. How can the presence of austenite help help improve the creep

You are right that the Chou and Bhadeshia paper deals with the production of an equiaxed, refined grain structure which at first sight contradicts the achievement of good creep strength. However, if failure is by a mechanism involving brittle fracture, for example, transgranular cleavage across the coarse grains, then a refinement of grain structure and the presence of austenite should help improve the creep life. This is by preventing premature failure by a mechanism other than creep.

I have read that raising the Ac1 temperature should lead to better creep-resistant steel because the initial tempering heat treatment can be carried out at a higher temperature.

I don't understand the mechanism. Tempering at a higher temperature
should coarsen the microstructure and hence, some of the creep life
of the material can be assumed to have been used up.

The Ac1 temperature is that at which austenite begins to form during heating at a specified rate. I have searched the literature and found that this idea comes from plots of the creep rupture strength versus the calculated Ae1 temperature, where Ae1 is the temperature where austenite forms on heating under equilibrium conditions. However, the inference that the creep strength increases with Ae1 is not firm because there are other differences between the steels, for example the chemical composition which has effects beyond just Ae1.

It is possible that a balance is required between strength and ductility. Sometimes, a weaker material obtained by tempering at a higher temperature is preferred because of its better ductility, which could actually lead to a better rupture life. The tempering temperature can be raised in line with the Ae1 temperature, because austenite should not form during tempering.

I need to determine the Ae3 temperature for a microalloyed steel experimentally. I wonder if continuous cooling at very slow rate such as 0.05 K/s will provide the answer.
Is the effect of austenite grain size very important at this slow rates?

Yes, slow cooling will give the most reliable measure of the Ae3 temperature. However, it is difficult to make a quantitative comment regarding the exact cooling rate or the effect of austenite grain size. A finer grain size is better in this context because the greater number density of nucleation sites reduces the supercooling necessary to achieve transformation.

From an experimental point of view, it is best to use two cooling rates, say one at 0.05 K/s and the other at 0.02 K/s. If the transformation temperature does not change then you can assume that it is close to Ae3. Also, use the smallest austenite grain size for the reason stated above.

You can then compare your experimental result with a phase diagram calculation (using MTDATA), done for both equilibrium and paraequilibrium. The comparison should reveal whether you are close to the truth.

One final point - in a microalloyed steel, the austenitisation temperature becomes a variable because the solubility of the microalloying element in austenite is a function of temperature.

Let me know what happens.

Do you know of any papers on the effect of the quench rate on the martensite-start temperature of steels?


Hennessey, Sharma & Ansell, Metallurgical Transactions 14A (1983) 1013

Donachie and Ansell, Metallurgical Transactions 6A (1975) 1863

Ansell, Donachie and Messler, Metallurgical Transactions 2 (1971) 2443

How does one model the formation of austenite from martensite?

The best thing to do would be to read the chapter on reaustenitisation in Bainite in Steels". There are many references there. A search on BIDS for the latest work would also be useful.

First, I have confusion for the fundamental terminology.
What are the differences between diffusional and  reconstructive transformation
mechanism and between displacive and diffusionless? Could you explain these
terms in detail?

Second, I read section2.4.3 repeatedly In 'Bainite in steels'. However it is
difficult to understand. So, I ask your help.
I learned that the surface relief due to shape change accompany bainite
reaction and the shape change is the description of invariant plane strain and
this invariant plane strain shape change has significant shear component.  If
IPS has no shear component(like case d2 in Fig.2.21), Is this called as IPS?
Can I think that IPS is not related with  transformation mechanism, either
reconstructive or displacive? that is, can IPS occur for reconstructive

The terms reconstructive and displacive are more generally applicable than diffusional and diffusionless.

The essential difference between reconstructive and displacive is that the latter is a transformation in which the parent lattice is converted into that of the product by a deformation. With reconstructive transformations, the lattice change involves mass transport in such a way that strain energy is minimised. This mass transport requires diffusion.

Thus, a transformation can be displacive but small atoms, which are located in the interstices, can diffuse. This is why "diffusionless" has a different meaning from displacive.

A general invariant-plane strain shape deformation has a shear and a dilatational component. A displacive transformation in steels will lead to an IPS with a large shear component (0.25) and is proof of a displacive mechanism.

An invariant-plane strain is one which leaves one plane undistorted and unrotated.

However, ANY plate shaped product, whether it grows by a displacive or reconstructive transformation mechanism, will show an IPS shape deformation but WITHOUT the shear component. There will simply be a dilatation normal to the habit plane. Otherwise, it cannot be a plate because there will be enormous misfit at the habit plane.

I have conducted experiments on austempered ductile cast iron
and have the following queries:

1) Deformation of austenite at temperatures in the range 316-375 centigrade accelerated the transformation.

2) Deformation of austenite at temperatures in the range 400-425 centigrade had relatively little effect on the transformation.

3) Deformation of austenite at temperatures in the range 316-375 centigrade showed little if any "mechanical stabilization."  I conclude this from the fact that the final volume fraction of retained austenite, which was determined using X-ray diffraction, was generally unaffected by deformation.  Do these results contradict the displacive mechanism of bainitic

Your experiments deal with the transformation of plastically deformed austenite to bainite in cast iron.

Let me say first of all that cast irons always contain chemical segregation so there is a difficulty in reading too much in your results. Nevertheless, I will make the following comments:

  1. Small amounts of plastic strain will accelerate transformation by enhancing the nucleation rate. So it is not surprising that transformation at low temperatures is accelerated by deformation.
  2. Transformation at a higher temperature leads to a smaller fraction of bainite via the incomplete reaction phenomenon. Therefore, it is not surprising that you observe little change when you transform at high temperatures.
  3. You have not found mechanical stabilisation, but have you used sufficient deformation. Small plastic strains also give inhomogeneous deformation.
  4. Your conclusion that you have no mechanical stabilisation from the fact that the retained austenite content did not change is quite unsafe. This is because it is the amount of austenite that existed at the transformation temperature, i.e. the residual austenite, which matters. Some of the latter will inevitably decompose to martesite.
  5. I do not think you can draw conclusions of mechanisms on the basis of the evidence presented.

In your Steels book, is there a particular source for Fig 12.7 on stainless TTT's for M23C6 and NbTiC, or did you draw it schematically from the authors cited for the various micrographs in nearby pages?  A colleague's looking at such kinetics at present.

This goes back to the first edition and I am uncertain of the source. However, you can find such diagrams in "Austenitic Stainless Steels, by P. Marshall, Elsevier Applied Science Publishers.

Why does the Ac1 Temperature lowers as the heating rate increases ?

Can you indicate whether the observed trend is significant after considering the error bars?

What is the Curie temperature of pure iron?

It is about 760 Centigrade

Why does the carbon content of austenite which is in equilibrium with graphite, increase when total carbon concentration is increased, in ductile cast iron at  900 C deg ?

This would not happen if the cast iron is simply Fe-C. However, when other solutes (such as Mn, Si) are present, there are additional degrees of freedom. The phase rule says P+F=C+2 where P is the number of phases, F the degrees of freedom and C the number of components. Let us say we have three components (Fe, Mn, C), two phases (austenite and graphite), there are then two degrees of freedom assuming constant pressure. If the temperature is also constant, there is one degree of freedom left. So changing the concentration of any component, whilst keeping the temperature constant, will alter equilibrium.

If you are using MTDATA, then try a calculation for Fe-C cast iron and you should find that the carbon concentration of the austenite does not change.

What is the difference between allotriomorphic ferrite and the forms of bainite? Are they are just different morphologies for the body-centred structure in equilibrium with Fe3C, depending upon the conditions that were applied during transformation?

No, the differences are more fundamental. Allotriomorphic ferrite grows by a mechanism in which all the atoms diffuse, so that the rearrangement of atoms associated with the transformation from austenite to ferrite does not cause strain, other than a small volume change. This is called a reconstructive transformation mechanism.

Products which grow in this way are not impeded in their growth by obstacles such as austenite grain boundaries, and hence do not have a crsytallographic shape (hence the term allotriomorph).

Bainite, on the other hand, forms by a displacive transformation mechanism in which the austenite is deformed into the new body-centred cubic crystal structure. Such a deformation in which the atoms move in a coordinated fashion cannot be sustained across grain boundaries. Bainite plates are therefore restricted to a single austenite grain in which they grow. The strain associated with the displacements is large and leads to a plate shape.

Picture of displacements associated with bainite.

When using Rosenthal's 3D solution for welding, the cooling rate along the centreline top surface (y=z=0) appears to be independent of velocity! How can we rationalise this?

Isn't this because at that point, the temperature is infinite? Rosenthal's solution assumes a point heat source.

What is the thermal expansion coefficient for the transformation from austenite to allotriomorphic ferrite?

Austenite has a larger thermal expansion coefficient than ferrite. Typical values are

Austenite 0.000021 per K
Ferrite   0.000013 per K

The exact values will depend on the chemical composition. The coefficients are weak functions of temperature.

See for example,

Scripta Metallurgica et Materialia 29 (1993) 1011-1016
Scripta Metallurgica et Materialia 27 (1992) 325-328

I have just been reading your paper in MST vol.15, p22, 1999 and have a question or two for you.

Towards the end of this paper, you write about impingement of
transformed product and the effect on nucleation and growth kinetics. I recall
that most of the theories dealt with the early stages when coalescence
was insignificant. I wonder if two effects have been evaluated in the
literature:- one is the decrease in solute concentration at the far
boundary when precipitates get large, the other (which occurs earlier in the process) is the decrease in overall nucleation rate consequential on diffusion fields being set up around stable particles. If you look at Fig. 13 in your paper then, on the left hand picture, you should draw another circle around each particle inside which the nucleation rate will be zero or very small as a result of solute depletion. Outside of that circle the nucleation rate will be finite and increasing rapidly with distance away from the particle.

There is another Avrami-type
problem of competing diffusion fields which influence the average
nucleation rate. Depletion of the solute available for nucleation and
growth is an obvious effect when the particles get large. My question
is:- have these effects been formally incorporated into the diffusion
and growth transformation theories? 

Let us define two kinds of impingement, "hard" referring to physical contact and "soft" the impingement between solute diffusion fields.

Hard impingement is dealt with by calculating an erroneous volume
fraction, the so-called extended fraction. Suppose that a phase beta is precipitating. In a time interval dt, the change in the volume fraction assuming that all particles can grow through each other and that nucleation can occur everywhere (even in transformed regions) is dV^e, where the superscript identifies the change in extended volume. The true change dV is then obtained by multiplying this by the propability that the change occurs in untransformed volume;

dV = (1-[V_beta]) dV^e

where V_beta is the fraction of beta. This assumes random nucleation but can be adapted for non-random events. This is the standard theory to deal with hard impingement and is reviewed by Christian in his Theory of Transformations in Metals and Alloys.

Soft impingement is much more difficult. If you specify the location of all the nucleation sites then it can be dealt with rigorously using finite element or similar discrete methods. However, this defeats the purpose of microstructure modelling where one wishes not to specify the starting microstructure. Consequently, what I use is the "mean field approximation" where the composition of the matrix is altered uniformly to compensate for the solute absorbed or desorbed by the precipitate. This also works well but is clearly an approximation.

Could you produce for me a few paragraphs on the use of neural networks, including words on errors and uncertainties and why this empirical method is useful.

The development and processing of materials is complex. Although
scientific investigations on materials have helped greatly in
revealing the underlying phenomena, there remain many problems
where quantitative treatments are dismally lacking. For example,
whereas dislocation theory can be used to estimate the yield strength
of a microstructure, it is not yet possible for anyone to predict the
strain hardening coefficient of an engineering alloy. It follows that
the tensile strength, elongation, fatigue life, creep life and
toughness, all of which are vital engineering design parameters, cannot even be estimated using dislocation theory.  The lack of progress in predicting  mechanical properties is  because
of their dependence on  large numbers of variables. Nevertheless, there are clear patterns which experienced metallurgists recognise and
understand. For example, it is well understood that the toughness of a
steel can be improved by making its microstructure more chaotic so that propagating cracks are frequently deflected.  It is not clear exactly how much the toughness is expected to improve, but the qualitative relationship is well established on the basis of a vast number of experiments.

Neural network models are  extremely useful in such circumstances, not
only in the study of mechanical properties but wherever the complexity of the problem is overwhelming from a fundamental perspective and where simplification is unacceptable in that it destroys the technological goal.  Good materials science has the responsibility to reach objectives in a cost and time--effective way. Any model which deals with only a small part of the required technology is therefore unlikely to be treated with respect. Neural network analysis can help resolve these difficulties whilst striving for longer term solutions. Neural network analysis  helps perceive patterns in data and problems
of enormous complexity. There are many examples where these patterns,
once revealed, have led to the development of physical models (review,
ISIJ International). The networks can be structured in such a way as
to include those physical relationships as are know. For example, in
the modelling of fatigue crack growth rates, it is appropriate to
include the Paris law and the threshold stress intensity range in the
formulation of the network. In this way, hybrid models consisting of
known or new scientific principles and the raw neural network approach
can be combined to produce meaningful models.

There have been major advances in the treatment of errors and
uncertainty in neural network analysis. Naturally, there is an
estimate of the perceived level of noise in the output, noise of the
kind which leads to a different experimental result when the
experiment is repeated, because some variable is not controlled. But
there is, in addition, the power of a Bayesian framework (MacKay). This allows the calculation of error bars representing the uncertainty
in the fitting parameters. The method recognises that there are many
functions which can be fitted or extrapolated into uncertain regions of the input space, without unduly compromising the fit in adjacent
regions which are rich in accurate data. Instead of calculating a
unique set of weights, a probability distribution of sets of weights is used to define the fitting uncertainty. This latter error estimate is particularly useful in highlighting gaps
in knowledge in systems where there are so many variables that it is
difficult to get a clear vision. For more information see the Neural Networks in Materials Science

I am doing neutron measurements and modelling of intergranular and
interphase stress development during loading.  It has been suggested that a TRIP steel would be interesting to look at because of the formation of martensite as loading progresses.

Do you think this is feasible? I know martensite suffers from very wide diffraction peaks due to the variation of internal strains.  Is this likely to be a problem / obscure the austenite peaks?
Do you know where I can get hold of a suitable material?  It is said that TRIP steels have fallen out of fashion somewhat!  We are restricted to performing the experiments at room temperature, and I would want to see significant further plastic deformation and transformation to martensite after Luders straining.  I don't think we can measure during Luders band formation because we'd just get a meaningless mixture of strains from the deformed and undeformed regions.

Do you have any "must read" references that you can suggest to me?

A TRIP steel, with transformation-induced plasticity would certainly be very useful in looking at the development of intergranular and interphase stresses. The transformation will to some extent act to relieve the stresses, so you will need to do a comparison against a control sample which does not transform.

I think that the problem of wide diffraction peaks can be minimised by using martensite which forms at a low temperature. I would recommend Fe-0.4C-28.1 Ni wt% which has an M_S temperature of -80 degrees centigrade. It undergoes TRIP at -44 degrees centrigrade (Journal of Materials Science 17 [1982] 383-386). There is no problem in finding austenite peaks which are well separated from those of martensite because of the different lattice spacings.

You will not be able to buy such a specific alloy, but you can get it made either at CORUS or I believe at Sheffield University. I do not have details.

You could by trial and error, or using a martensite-start temperature calculation (see MAP ) design an alloy for ambient temperature work if this is essential.

You do not have to worry about Luders bands with TRIP steels because they prevent localised necking by transformation, i.e. a high strain hardening coefficient. TRIP steels are expensive. They are therefore used little on their own. However, there is a very major market (millions of tonnes) on TRIP steels in automobile applications where only a part of the microstructure TRIPs (the remainder is allotriomorphic ferrite). Please see, for example, Materials Science and Engineering A, Volume A273-275, 1999. This contains numerous references and papers about TRIP steels. You rely for your safety during an automobile collision on these steels. 

What are the typical applications of iron and nickel-based oxide dispersion strengthened, mechanically alloyed metals?

PM2000, which is an iron-based alloy, is used as follows:

         - in furnace construction as shields or carrier systems
         - in glass industry as stirrer or plunger in molten glass
         - combustion of waste materials
         - thermocouple protection tubes
         - high temperature testing equipment
         - Burner tubes
         - automotive applications in diesel engines

PM1000, which is a nickel-based alloy, is used as follows:

         - rotating discs for glass fiber production
         - HT Screws and Fasteners
         - Face sheets in thermal protection panels
         - Space and aerospace engineering in general

Both of these alloys are made by Plansee GmbH, Germany

Where in the Bainite Book can I find information on X-ray evidence for the inhomogeneous distribution of carbon in the residual austenite?

The evidence is on page 70 of Bainite in Steels.  Work by Matas and Hehemann (1961) showed two sets of lattice parameters for austenite because some of the austenite had a larger carbon concentration. The consequence of this heterogeneous distribution of carbon in austenite is that the low-carbon regions frequently decompose to martensite on cooling from the bainite transformation temperature to ambient. In such cases, the carbon concentration determined by measuring the lattice parameter of the retained austenite will give a high value which is misleading because the low carbon regions have decomposed to martensite. If this misleading value is then used to estimate the fraction of bainitic ferrite, than the latter is overestimated.

You suggested that I grab a picture from your website which is of an
AFM (I think) of a steel surface showing a wedge of transformed
material.  Indeed, I obtained that figure and have installed it in
my book.

I am writing for two things:

1) explicit permission to use it.

2) A precise explanation of what is being shown and how it
was obtained, what the material is and what its life history is,

The picture shows the displacements caused at a free surface when a sample of austenite is transformed to bainite. You can find all the necessary details in Materials Science and Technology, 12 (1996) 121-125, by Swallow and Bhadeshia

The colour version of  picture the picture has not been published, only a black and white version in the above reference. You therefore do not need permission from the journal; you have my permission to reproduce it in your book.

Here's a basic question which is puzzling us. Creep resistant steels are severely tempered before service. What is the purpose of this tempering treatment?

We think that the only logical answer is that tempering is
necessary for stress relief, in which case it must be done at temperatures where iron atoms can move substantial distances.

There are rumours that the tempering stabilises the microstructure. But this does not make sense, because tempering at ever higher temperatures must coarsen the structure and hence reduce the overall life.

Please can you help?

"Severe" tempering improves toughness and ductility. In thin section materials (e.g. boiler tubes and pipes) this is achieved through tempering for fairly short times at high temperatures e.g. 770C/2 hrs is typical for NF616. In thicker section materials (e.g. turbine rotor forgings and castings) a similar degree of tempering is achieved through longer periods at lower temperatures (longer times being necessary to achieve uniform temperature distribution).
This tempering also allows post-weld heat treatment to be carried out under similar conditions without significantly modifying the pre-welded properties of the welded components and in weldments the severity of the tempering is probably more critical, not just in terms of relaxing residual stresses, but also in tempering
the brittle microstructures present in weld metal and HAZ. Without such tempering acceptable levels of toughness, crack growth resistance and stress corrosion resistance in off-load conditions would not be achieved.

"Severe" tempering certainly reduces creep strength in the short term (and even in the longer term at lower temperatures) but evidence from long term creep programmes shows that after long term at the highest (and therfore most critical) operating temperatures there is little or no advantage of higher proof strength - the creep rupture curves converge after long time. Evidence from the metallography shows that the higher proof strength conditions are less stable than the lower proof strength conditions - they start with higher dislocation density and smaller sub-grains but on ageing/creep testing these parameters rapidly converge with the structures observed in the lower proof
strength conditions. 

Is it necessary for iron atoms to be able to diffuse long distances in order to relieve residual stresses? I refer particularly to the 600-700 degrees centigrade heat treatments given to power plant steels.

This is a good question for which I do not have a clear answer. I would have thought that the diffusion is essential. What other mechanism is there to relieve an elastic stress by heat treatment in the context of these alloys?

Creep resistant steels are severely tempered before service. What is the purpose of this tempering treatment? 

Different heat treatments give variations in  mechanical properties (creep rupture strength, toughness) during short-term service at elevated temperatures up to about 10,000 hours. However, they do not substantially affect the  long term behaviour of the steel since the the creep curves tend to merge. Another requirement is that  material should fulfill the manufacturing specifications since a very hard material makes life difficult during fabrication.

Creep-resistant power plant steels are severely tempered before service. For example, 700 degrees centigrade for many hours. The tempering is necessary for stress relief and for post-weld heat treatment. Tempering at a lower temperature leads to a larger initial creep strength but a sharper fall, although there is no deterioration of the long term properties. Is this a fair conclusion? Because I don't see why the long term properties should be affected. Also, why worry about stability (i.e. the initial sharp fall in creep strength) if the long term properties are unaffected?

Your conclusion is reasonable although there is some uncertainty over whether the creep strengths of high and low proof strength materials cross-over or simply converge at very long times.
I don't worry about instability in lightly tempered  materials (which is different from the long term instabilities which lead to formation of Laves and Z phases, for example) -  heavily tempered materials offer no long term advantage.

We wish to improve the oxidation resistance of 9Cr1Mo type power plant steels. Would alloying with aluminium help?

If you are aiming at forming a potective alumina scale, you would need to add at least around 5 wt%, the level which is added to the Fe20Cr ferritic steels (Fecralloy and the ODS alloy MA956). I think that 5 wt% Al additions to 9Cr steels will lead to severe problems in manufacture and welding.

However it is still quite difficult to form alumina scales at low
temperatures, say up to 700C. This is because the diffusion of Al is slow and the only way to provide oxidation protection is to give the material a high temperature pre-oxidation treatment in air at 1200C. We have used Fe20Cr5Al steel retorts in our creep and corrosion furnaces and while the material operates rather well at 1000-1300C in aggressive environments, the retorts began to show catastophic failure (oxidation of Fe) at intermediate temperatures 600-800C. We cured the problem with a high temperature pre-oxidation treatment.

For the 9Cr steels, we see the main development towards improved oxidation resistamce in steam being enhanced Cr levels and increased spinel-forming additions. Co and Mn seem to be promising. Co especially is an austenite stabiliziing addition which expands the austenite field in the composition direction with no significant reduction in the austenite-ferrite transformation temperature, so that a good tempered martensitic microstructure can be achieved with up to 12 wt% Cr and 3 wt% Co. Our initial steam oxidation testing of such steels has shown much improved performance.

At high levels of Al additions, there may as you suggested be a problem with AlN formation. This would decrease the amount of N available for the precipitation of the fine nitrides and carbonitrides which are essential for high creep strength. 

I am a bit confused about the actual differences between Ac1
(heating) and Ae1 (equilibrium),and also why are we more interested in Ac1 than Ar1 in the context of tempering heat treatments?

The Ae1 temperature is the lowest temperature at which austenite can form on heating. It represents transformation under equilibrium conditions.

The Ac1 temperature is the lowest temperature at which austenite can form on heating at a specified heating rate. Ac1 is usually higher than Ae1, but tends towards Ae1 as the heating rate tends to zero.

The Ar1 temperature is that at which all of the austenite has decomposed during cooling of a steel sample from the austenite phase field. Ar1 is usually lower than Ae1, but tends towards Ae1 as the cooling rate tends to zero.

The "c" in Ac1 is from the French word "Chauffage", meaning heating. The "r" in Ar1 is from the French word "refroidissment" meaning cooling. The "e" in Ae1 stands for equilibrium.

Tempering heat treatments are conducted by taking a steel at ambient temperature and heating it to the appropriate tempering temperature. That is why Ar1 is not relevant, neither is Ae1 since the heating rate is unlikely to be that consistent with the achievement of equilibrium.

This is a question about the tempering of "boiler grade steels", i.e. the sort of alloys used to construct the steam-generating boilers for power plant. One example is the classical 2.25Cr1Mo steel.

After welding these steels are generally given a post-weld heat treatment to reduce the hardness of welded joint. If, on the other hand, the joint is not heat-treated but put into service at elevated temperatures in the as-welded condition, then how does the service stress effect (a) the service life of the joint and (b) the evolution of the microstructure?

There are generally two reasons for post-weld heat treatment for the case you describe. As you say, the first is to reduce the hardness in critical regions of the heat-affected zone of the weld.

The second reason is to relieve the inevitable residual stresses that build up as the weld solidifies and then cools non-uniformly.

Any residual stress due to welding will add to the applied stress when the component is put into service. The design engineer then has to limit the amount of applied stress in order to avoid overloading the component. If this is ignored then the component may fail prematurely.

Stresses are known to accelerate tempering reactions. For example, the evolution of the carbide phases is faster in the gauge length of a creep tensile specimen than in the unstressed grips of the tensile specimen. This is discussed on page 378 of Bainite in Steels under the subtitle Creep Tempering. Naturally, if the tempering processes are accelerated then so is creep damage.

It is possible to find many steels with a manganese concentration either less than 2 wt% or greater than 12 wt%. It seems that intermediate concentrations of manganese are not popular. Why is that? 

The low manganese steels are usually ferritic. The major role of the manganese in ferritic steels is to improve hardenability and to getter sulphur to form manganese sulphides.

One class of steels containing 12 wt% of manganese is the Hadfield alloys, but these are austenitic because they also contain a high carbon concentration of 1 wt%. The Hadfield steels have a very high work hardening coefficient and hence are abrasion resistant with major applications in the railway industries.

There are many austenitic stainless steels where manganese is used as a cheaper substitute for nickel. This was a major issue during the second world war when supplies of nickel were scarce.

With low alloy steels, for example of the kind used to make buildings and bridges, a manganese concentration much above 2 wt% is detrimental because firstly, the hardenability increases so much that the steels cannot be welded without generating brittle martensite (see carbon equivalent equations).

Manganese also segregates strongly during solidification and is primarily responsible for the banded microstructures seen in structural steels. Concentrations larger than about 2 wt% are also detrimental in this respect. These are the reasons why ferritic steels with greater than 2 wt% of manganese are not to be found.

Austenitic stainless steels have some 8 wt% of nickel, so its substitution with manganese requires large concentrations. Martensitic stainless steels on the other hand have sufficient hardenability with just 1 or 2 wt% of manganese, so like ferritic alloys then need not have higher concentrations.

A good book on the subject is Manganese Stainless Steels published by the Manganese Centre, 17 Avenue Hoche, 75008 PARIS, France. I believe this is available free from the Centre, who also provide a vast quantity of other literature on manganese steels. This particular book was published in 1986 so I do not know if it is still available.

What are the chemical compositions of Alloys PM1000 and PM2000, which are both mechanically alloyed, oxide dispersion strengthened; they are manufactured by Plansee GmBH

The nominal composition of PM1000 (wt%) is

Cr      20
Fe       3
Ti     0.5
Al     0.3
yttria 0.6
Ni     balance

An actual compositin of PM2000 (wt%) is:


This is a query about phase diagram calculations. We would like to include a stored energy term, say 50 J/mol in ferrite, and then recalculate equilibrium with a carbide. In other words, if we raise the entire free energy curve of ferrite by 50 J/mol, how does the equilibrium composition change?

Is this possible to do with MTDATA? 

There is as yet no direct way of doing this but here are three suggestions:

1) Change the basic thermodynamic data by editing the .mpi file (NOT

2) Use the new Assessment module to interactively add Gibbs energy to the appropriate unary data items in the phase (NOT RECOMMENDED - the Assessment module is very complex).

3) Make use of the VdP term in the Gibbs energy model for a phase that
allows correction of the Gibbs energy at elevated pressures.

I would recommend using the latter because:

a) the VdP term is not that far removed physically from the nature of the addition you wish to make;
b) the change can be made interactively or programatically;

c) under "normal" conditions the additional Gibbs energy is zero.

The way the change is made is to modify all the unaries for the phase to now include a pressure dependent contribution to the Gibbs energy (see MTDATA Utility Module manual). You therefore have a database of your own containing the data for the pressure dependent phase. When you search for your system you arrange for your data to be selected instead of the data from the original database.

To add Gibbs energy you simply set the pressure. For example if you added a constant 1.0e-6m3 molar volume to each unary you would create a 50J/mole addition via a "set p=50e6 !" command. This means that you could recalculate equilibira also as a function of added Gibbs enegry if you wished by stepping P. By default, P=101325Pa, the Gibbs energy addition would then be 0.101J.

My question is about environmental friendliness about steel. As described in textbook, it is very significant effect on strength of steel to use alloying elements as Mo, V, W etc, for example the formation of alloy carbides secondary hardening. But such alloying elements are harmful to human body and make it difficult to reuse the steel. So I think we must develop the properties of steel without alloying elements. What do you think about environmental friendliness about steel in future?

Elements such as Mo, V, W in solid solution or as compounds in steel are not harmful to the environment. They have been used safely for more than 100 years. There is, however, a move all over the world to reduce the alloying elements in steel in order to streamline production by decreasing the variety of steels. Thus, the "Supermetal" project in Japan (Professor Masahashi was a part of this before he joined Tohoku University) has one of its aims to reduce alloying elements in steels.

The main environmental issue with steel is the high consumption of fuels of relatively high quality. The average energy used per tonne of steel is about 21 GJ, in Japan about 18 GJ. Furthermore, there are CO_2 emissions which build up in the atmosphere and give the global warming phenomenon. On the positive side, steel is the most recycled material (more than 50%) and this proportion is growing.

My subject in a master's course is about the evaluation of austenitic stainless steel welds with ultrasonics. In austenitic welds, it is impossible to detect the faults with ultrasonics  because the welds have great crystal anisotropy, so that ultrasonic velocity also has great anistropy. So we attempt to simulate the propagation of ultrasonic waves through the crystallographically textured weld. For engineers to study the field of welding, what do you think that we could come up with as an evaluation technique for such welds?

This is a very interesting subject. There is one way of eliminating the growth of columnar grains in the weld, which give rise to the anisotropy. This is by inoculating the weld metal with particles of TaN, which cause considerable nucleation from the liquid. Thus, the anisotropic structure is replaced by equiaxed grains and a much better ultrasonic response.

Simulation is also an excellent idea, but you would need to have details of the crystallographic texture in the weld metal in order to do the simulation reliably. So a great deal of work is probably needed to first be able to predict the texture. 

This question is about the contents of your book "Steels". I have a question on Chapter 12. Austenitic stainless steels such as 304 have resistance to intergranular corrosion. And, chromium has a great effect on this. But Chromium forms chromium-rich carbides, which can be the main cause of intergranular corrosion. So,I think there is an inconsistency in the design of such steels.  And then, I have an idea that increasing cromium could reduce the depletion of chromium even if precipitation occurs. Is this way of thinking reasonable? 

A critical concentration (about 11-12 wt%) of Cr is needed to produce a coherent, adherent oxide film, which regenerates spontaneously when damaged, on the surface of the stainless steel. If, however, there is carbon then Cr-rich carbides form at the grain boundaries, locally deplete Cr and these depleted regions are the focus of intense corrosion.

To avoid this, small concentrations of Ti or Nb are added to combine preferentially with the carbon and hence avoid any depletion of Cr. Such steels are called stabilised stainless steels. See page 259 of "Steels". Alternatively, very low carbon stainless steels are made, for example, AISI 316L, the "L" designating low carbon.

Of course, the overall Cr concentration can also be increased as you suggest, but then it would also be necessary to increase nickel if the austenitic phase is desired. This can be expensive.

I'm researching about the liquid metal embrittlement (LME) of austenitic stainless steel (type 304) by zinc. From some papers, LME happens when 304 steel is stressed and contact with molten zinc above 1023K, and it leads intergranular cracking. thus, I think there are four factors cause of LME. They are molten zinc, temperature, stress, and grain boundary. At first I tried to research the effect of temperature and stress. At the experiment, I used tensile test sample of 304 steel with zinc on its surface,and give it fixed stress, and thermal history(heated 50K/sec.and cooled 30K/sec.). Maximum temperatures were between 1023 and 1123K, and tensile stress was changing around the yield stress at maximum temperature. At the result, there was no cracking in appearance, if the tensile stress was lower than the yield stress of maximum temperature. But other samples were cracked.

After that, I tried to find zinc in the 304 steel by BSE observation, and I was able to find it penetrated through the grain boundaries, even if there was no cracking. From this result, I suppose the LME process is firstly molten zinc is penetrating through grain boundaries, and then some reaction cause of cracking happen. I think this reaction is forming inter-metallic compound. Now I'm planning next experiment. That's EPMA for other elements and,so on. And I have some questions about my research. There are many elements in the 304 steel.  (1) What element do you think reacted with zinc? and how reacted?  (2) What kind of experiment do you interested to do next?  (3) If you think another supposition, how do you think the process of LME of 304 steel? 

Your experiments are very clear, that the mere penetration of zinc into the grain boundaries does not lead to fracture under stress, since you did not observe fracture in many cases where Zn penetrated the interfaces. So, it seems reasonable to conclude that there must be some reaction product which actually causes embrittlement.

Microanalysis of the fracture surface will certainly help identify a reaction product, although there may be difficulties in avoiding overlap of information with the matrix away from the grain boundary. Assuming that the reaction of Zn with iron is similar in austenitic and ferritic steels, there should be Fe_x Zn_y compounds, many of which are known to be brittle.

Perhaps an X-ray diffraction experiment where you expose the fracture surface to X-rays will help identify the compounds. X-rays penetrate only a few microns.

Another experiment could be to encourage reaction by heat-treatment at elevated temperature and then see if fracture is more probable.

As to mechanisms, I am not an expert but the following are possible:

1) The element reduces surface energy and hence encourages fracture. This is unlikely in your case because you get Zn penetration but not always fracture.

2) As you suggest, brittle compound formation. This is well known in the galvanising of ferritic steels for the automobile industries.

3) The situation may become complicated if your alloy is not fully austenitic, but contains some ferrite. Koseki (TETSU TO HAGANE-JOURNAL OF THE IRON AND STEEL INSTITUTE OF  JAPAN, 1993, Vol.79, No.9, pp.1108-1114) has done some work to show that Zn embrittlement is intensified if the ferrite ratio at the grain boundary is reduced.

Dr Toshihiro Koseki is a friend of mine at Nippon Steel and it might be worth e-mailing him for some advice. His e-mail is

Finally, I am not clear about the significance of the applied stress. You imply that if the applied stress is kept below the yield stress at the maximum temperature, then there is no cracking. Thus, plastic deformation must be necessary to induce cracking. So if you take a sample which has not cracked but has Zn penetration, then futher plastic deformation should induce fracture.

Fig.3.16 of "Steels" by Honeycombe & Bhadeshia is the Fe-C phase diagram. In the diagram austenite resolves ferrite and cementite at 723 C. Then ferrite and cementite form pearlite.  Your textbook says that growth of pearlite depends on the diffusion of carbon in  austenite. Frankly speaking, which does carbon of austenite diffuse easily, ferrite or cementite? I don't know that my question is answered in your textbook. In case of that, I should apologize to you. But I could not understand about it. I wish you kindly give me good advice. And I could not well understand about the forming of bainite. Anyway I must study more about the diffusion. I can't have good image in my brain about it. 

Pearlite is a mixture of ferrite and cementite, and forms directly from austenite (Fig. 3.10). Its average composition is the same as that of the austenite, so the only diffusion that is needed is parallel to the austenite/pearlite interface. See figure 3.12. Diffusion is not necessary inside either ferrite or cementite. Bainite is quite different; it forms without diffusion, but the carbon then escapes from the bainite to precipitate carbides. 

Why has VC such stability at high temperature compared with other alloy carbides? 

Perhaps there is an explanation in the concepts of chemical bonding. Cottrell, in his book "Chemical Bonding in Transition Metal Carbides",  has been able to explain many of the observed trends in the stability, crystal structure and stoichiometry of the carbides of transition metals in terms of chemical bonds. He points out that Ti, Zr and Hf, which in the periodic table are elements near the beginning of the long periods, form very stable MC carbides but the affinity for carbon diminishes further along the rows of the periodic table. A part of the reason for this is that more electrons have to be accommodated for elements further along the rows, so antibonding states are progressively filled thereby reducing the bond order. This is analogous to the hydrogen-helium example given earlier, but does not completely explain the trend because the maximum bond order occurs with Cr, Mo and W and we know that carbides of these elements are less stable. 

With MC carbides, the metal has to sacrifice four electrons to form the bonds with carbon. Titanium has exactly the right number so that its d-orbitals are left empty on forming TiC. This is not the case with VC, since vanadium has an additional d-electron which forms a V-V bond. The electrons in the two kinds of bonds, V-C and V-V mutually repel, leading to a reduction in the stability of VC when compared with TiC. This problem becomes larger along the row of the periodic table until MC carbide formation becomes impossible or unlikely. 

Why do vanadium carbide (VC) precipitates form as a pair of "wings" at the right and left sides of sperical NbX as two wings, not four or six wings? 

This is a difficult question because VC precipitates with a Baker Nutting orientation of which there are three variants (if you do not count positive and negative equivalents). It could be that once a particular variant starts to form, its diffusion field prevents the formation of other variants

In Chapter 9 of "Steels" by Honeycombe and Bhadeshia, the tempering of martensite is explained by
dividing the process into 4 stages. In the textbook. I use, it is divided into 3 stages and stage 3 and stage 4 are combined. This question is trifling but if there is reason, please teach me.

All these divisions are a little arbitrary. The important point is the first the carbon leaves solid solution to precipitate a transition carbide such as epsilon; then any retained austenite may decompose; all the carbon leaves solution and cementite precipitates; finally the ferrite grain structure itself coarsens and recrystallises.

I have never studied strengthening mechanism like solid solution strengthening, and I would like to understand this mechanism in detail. How is the strengthening mechanism affected by controlling the orientation for Fe3C in pearlite using rolling technique and directionally solidified technique ?  How is such a practical technique?          

The strength of pearlite is often expressed in terms of the interlamellar spacing. If the interlamellar spacing is "d", then the strength is said to vary with the inverse square root of d. This is like the Hall-Petch equation for grain size strengthening.

Pearlite cannot be produced by directional solidification because it grows in the solid state from austenite. This is why it is called a "eutectoid" rather than a "eutectic" reaction. However, for research purposes, the pearlite can be forced to grow in a particular direction by imposing a sharp temperature gradient.

The alignment of pearlite along the deformation direction is most
prominent in wires which are drawn either as the pearlite forms or in the pearlitic condition.

Should the precipitation of vanadium carbide in austenite retard the subsequent transformation to ferrite?

I believe that the carbide forms at the most potent nucleation sites in the austenite and hence reduces the number density available for ferrite nucleation. 

Yes, there is evidence that your suggestion is correct.

An experimental  (Fe-W-C) alloy was used to study the effect of the precipitation of carbides at austenite grain
surfaces, on the subsequent formation of allotriomorphic ferrite. It was found that the removal of potent austenite
grain boundary nucleation sites by the precipitation of carbides retards the allotriomorphic ferrite transformation.
This effect is found to be most pronounced at small undercoolings below the equilibrium transformation
temperature. At larger undercoolings where nucleation is not as much of a rate-limiting factor, the depletion of
solute from the austenite, due to carbide precipitation, results in an acceleration of ferrite formation.

Materials Science and Engineering A, Vol. A157, 1992, pp. 101-106. 

Do you have anything published or public on bainitic rail steels? I would like to include some comments on this development in a short review that I am preparing for a symposium on retained austenite in the Heat Treating Society meeting in St Louis in October. Hope you can help.

Yes, there is an article published in "Science in Parliament" on the bainitic rail steel.

Bundy (1965 Journal of Applied Physics)  gives the triple point of iron as 110 kbar 500 degrees Celsius. He cites Johnson, Stein & Davis (1962 Journal of Applied Physics) as his source. However, this 1962 paper quotes 775 Kelvin and 115 kbars as the triple point conditions. As this represents a difference of about 5% in pressure I was wondering which is the more accepted value?

Unless Bundy states a reason for citing the particular value (for example, that there was a mistake in Johnson et al.), the original paper must be taken to be nearer the truth.

Does widmanstatten austenite result from a displacive transformation, in which only interstitial elements diffuse during nucleation and growth?

I ask you this question, because the nature of this transformation could be similar to the transformation which takes place in semi-ferritic stainless steels. If this is the case, this would be consistent with  the fact that delta-ferrite and austenite were found to have similar substitutional composition. 

There is not enough evidence to comment with authority on the mechanism of Widmanstatten austenite formation in the kind of steels you are interested in, or even in lower alloy steels. For example, I do not know of any work which has revealed the shape deformation. I do not know of any work where the lengthening rate of the austenite plates has been measured. There are no atom-probe data to look at potential short-range composition changes.

All you can say is that it frequently forms as plates and that there sometimes is no bulk partitioning of substititutional solutes. However, there is no proof of a paraequilibrium mechanism, because local-equilibrium growth at high supersaturations can also be consistent with no long-range substitutional solute partitioning.

To summarise, there is a lot of speculation about mechanisms but no conclusive evidence.

What you should do, is assume a mechanism, and do calculations to predict the microstructure and see if they work. This may also help identify critical experiments, just like the enormous amount of revealing work that has been done on the austenite to ferrite transformation.

During your visit to Belgium last month I asked you a question about the difference between bainite and martensite since both transformations are displacive. As I understood the growth of both phases is the same mechanism, but nucleation goes faster and diffusionless for martensite (higher undercooling).

Now I have another question: As the transformation mechanism is the same for both materials, does it mean that phenomena of tempering of martensite in the (150 - 400C zone f.e.) are exactly of the same type as what happens to bainite immediately after formation in the same temperature region? I hope to get an interesting answer from you.

The first part isn't quite precise: both bainite and martensite grow without diffusion. Martensite also nucleates without diffusion. Carbon is partitioned during the nucleation of bainite, so it can occur at a smaller driving force compared with martensite.

Yes, the phenomenon of tempering is essentially the same for both martensite and bainite. Thus, all the characterisitics of carbide precipitation, such as the detailed crystallography and the formation of metastable phases, are identical for bainite and martensite. There is a chapter on the tempering of bainite in my book where I make comparisions.

I was also wondering if you might know (or know of someone that could help) what are the best ways to identify Bainite in a HSLA Steel.  For this particular case I am expecting a mostly martensitic matrix with bainite.   

Assuming that you do not want to indulge in detailed microscopy etc. here is a very effective way of deducing the microstructure in steel samples where, because of their lean alloy content, it becomes difficult to distinguish martensite and bainite using optical microscopy.

You say that you might have a mixture of martensite and bainite. Measure the hardness; we shall call the value of this hardness HV1.

Take a sample of your steel, say 5 mm diameter rod, 1 cm length. Heat it to 900 C for 5 minutes and plunge it into cold water. This will produce a fully martensitic speciment. Carefully measure the hardness (HV2), avoiding any regions near the surface where there may be decarburisation.

If HV2 > HV1 then you have bainite in your microstructure. 

It would be interesting to define the extent of the heat affected zone of a weld in a stainless steel, in order to be able to quantify the presence of ferrite as a percentage. Any ideas?

Any definition of a heat affected zone must in general be subjective since there is no boundary other than the size of the component to restrict the flow of heat. Therefore, you have to decide, for example, that any region which reaches a temperature that is less than say 873 K is neglected. The specific details must depend on the problem. In your case you are interested in delta ferrite formation so the limiting temperature must be much greater than 873 K.

There is an excellent book by Grong, Metallurgical Modelling of Welds, second edition, Institute of Materials, which you could use to calculate the flow of heat and hence the peak temperatures reached in the heat affected zone as a function of the distance from the fusion boundary.

The literature shows that there is a model to quantify the formation of delta ferrite in stainless steels, in the heat affected zone of a weld. However, to use this requires an estimate of the Ac4 and Ar4 temperatures. How can we obtain these?

So presumably, the model you mention is not kinetic, but simply an experimental measurement or a phase diagram calculation.

Both Ac4 and Ar4 require a kinetic model based on firm thermodynamic information. There is no easy solution, you need to generate these data.

I am working on the welding of stainless steels which may contain delta ferrite. I did a three hour heat treatment in a furnace at 1673 K, to reveal whether a set of three stainless steels would form delta ferrite. These experiments showed that delta ferrite does form in some cases. However, when I tried the same experiments in a thermomechnical simulator, delta ferrite did not form because of the short dwell time at high temperatures. My conclusion is that the hold time is important. Can you comment?

This of course makes sense since you must allow time for samples to reach equilibrium.

However, you should not do a three hour experiment in a thermomechanical simulator at 1673 K because the sample will degrade and the machine itself may trip and be ruined.

You also need to understand what the purpose of the furnace experiment is. Surely you were attempting to verify phase diagram calculations rather than simulate a heat affected zone. After all, the temperature pulse experienced by a zone like that is brief. The simulator experiments are more representative of the HAZ.

Can you provide me with some information about the acicular structures in microalloyed steels?

I have written a review on the subject, which you can download from the world wide web.

Note that the term "acicular" strictly means needle like, whereas there are no needle like ferrite phases in steels. What is seen to be a needle on planar sections is in fact a plate in three dimensions. The correct terminology is therefore "plate-like".

You may also find the following reference useful:

"Models for Acicular Ferrite", International Trends in Welding Research, eds. S. A. David and J. M. Vitek, ASM International, Ohio, USA, 1992, pp. 213-222. H. K. D. H. Bhadeshia 

This question is regarding the possibility of creating transparent steel. How can we make visible light refract through steel and reduce its opaque nature. Does it depend on the crystalline nature of the substance, if so, why diamond can refract light and steel can not. Does this have anything to do with the energy of light interacting with the electrons in outer orbit of atoms in that substance.

thanking you,


Good question. The reason why metals are opaque is because of the metallic bond. The valence electrons delocalise and the resulting electron energy levels are closely spaced, such that visible light can cause electronic transitions. Thus, the light is absorbed.

By contrast, the gaps between the bands in diamond, where the bonding is covalent, are very large and hence diamond is transparent.

A crystalline structure per se is not necessary for transparency since, for example, window glass is transparent but has an amorphous structure. It is whether the visible light can cause electronic transitions and hence be absorbed.

If iron is finely divided, it will eventually undergo a metal to insulator transitions. An insulator has large band gaps, so it should at the same time become transparent.

Most of FCC alloying elements (like Cu, Ni, Mn, etc) stablise the austenite phase in steels, but  aluminum stablises ferrite. Is there any particular reason for the observed behaviour of Al?

This is difficult to answer. The concept of a particular element stabilising a particular phase is weak. For example, chromium is supposed to stabilise ferrite, but it nevertheless reduces the martensite-start temperature.

Why a particular element behaves in the way it does can be answered in a number of ways. Thus, if it reduces the free energy of a phase A relative to another phase B then it stabilises phase A relative to phase B. The free energies can be measured or in rare cases can be calculated using electron theory. But even then, I do not find it easy to picture how a particular element works. 

Why does the Ac1 Temperature decrease as the heating rate increases ?

I do not fully understand this question. Have you got a specific example? In general, the Ac1 temperature should increase with the heating rate. 

What is the hardness of PM1000 and PM2000, the mechanically alloyed nickel and iron, before and after typical recrystallisation heat treatment?

PM1000 has a hardness of 550 HV before and 250 HV after recrystallisation.

PM2000 has a hardness of 400 HV before and 290 HV after recrystallisation.

How does nitrogen affect the fatigue properties of a weld metal with the following composition?

Fe-0.77Si-1.76Mn wt %, 600 p.p.m of oxygen, 140 p.p.m. of nitrogen

The steel you talk about is essentially ordinary mild
steel. The main detrimental effect of nitrogen is that
it causes strain age hardening, particularly in the root
pass of a weld. This causes embrittlement.

In your case, failure is by fatigue. Therefore, there
could be many reasons other than nitrogen which
contribute to failure. The main reason could be bad weld
design from the point of view of stress concentrations
and the build up of residual stresses due to welding.
You also have a very high oxygen concentration - fatigue
failure can initiate at inclusions.

Please find below a list of papers on the detrimental
effect of nitrogen.

T W Lau M M Sadowski T H North G C Weatherly,
Effect of
nitrogen on properties of submerged arc welds
Science and Technology 4 (1988) 52-61

D J Abson
Influence of current type and polarity on the
WI Research Report 292/1985, Novemeber

R B Oldland
Al and N effects on the micro and props of
single pass SMA welds,
Australian Welding Research,
December 1985, pp. 31-43

B. Ahlblom, U. Bergstrom, N E Hannerz I Werlefors
Influence of welding parameters on N & micros. of SMA
WI conference residuals and impurities in steel,
paper 49 1986

T Lau, M M Sadowsky, T H North and G Weatherly
Effect of
Nitrogen on the toughness of HSLA weld deposits
metallurgy of structural steels, ed. J Y Koo, TMS AIME.,

B Ahlblom, U. Bergstrom, N E Hannerz and I Werlefors
influence of welding parameters on nitrogen content
WI Conf. Residuals and impurities in steel, 1986,
paper 49, 1-12

O Grong, A O Kluken and B Bjornbakk
Effect of N on weld
metal toughness in self-shielded flux cored..
Joining and
Materials 1 (1988) 164-169

D J Abson
Microstructure and mechanical properties of
vertical-up C Mn Ni..
Welding Institute Res. Rep.
7931.01/86/544.3 1-30 (1987)

T Boniszewski
Self shielded arc welding
W.I. Conf.
Advances in Joining & Cutting, 31 Oct 1990, paper 36

M Kocak, B I Petrovski, E Richter and G M Evans
of Ti and N on the fracture properties of weld metals
- Vol. 3, Materials Engineering, ASME 1994, 277-289

B Petrovski and G Evans
Effect of N, Ti and strain ageing
on the toughness of weld metals
EUROJOIN 2, florence, May

G M Evans
Effect of aluminium and nitrogen on Ti-B
containing steel welds
American Welding Journal 76 (1997)
welding alloying effects

I read 'Bainite in steels'. However I couldn't understand something. So, I email to you.

In page.30, That the dislocation density of bainitic ferrite increase as the  transformation temperature is reduced is written. This confuse me as compared  with paragraphs of page 29. Why does the dislocation density of bainite increas?

I think, as Written in page 29, If temperature decreases, the strength of parent and  product phase increases, and then plastic relaxation is difficult to occur. Thus there is low density of dislocations.
I can't combinated two contexts of page 29 and page 30. Could you help me?

There are two temperature dependent effects. First, as you say, the strength of both the austenite and ferrite increases as the temperature is lowered. This makes it easier to accommodate the shape deformation elastically and hence should lead to a reduction in the dislocation density.

However, the second effect is that there is dynamic recovery. Recovery occurs more readily at high temperatures, and leads to a reduction in the dislocation density.

It follows that there will at first be an increase in the dislocation density as the temperature is reduced and recovery is minimised. However, there will then be a sharp reduction in the dislocation density when the shape deformation is elastically accommodated at a sufficiently low temperature.

You may find it useful to have a look at a recent paper.

In your book 'Bainite in steels', you say that the post-transformation carbon content of  bainitic ferrite tends to be significantly higher than equilibrium. Is this attributed to dislocations (i.e. interstitial trapping around  dislocation) within bainitic ferrite?

I understand from the book that carbon contents of bainitic ferrite after  transformation is not equal to that during transformation, due to fast carbon  diffusivities though the bainitic ferrite grows diffusionless.  Is this right?


Yes, the measured carbon concentration of bainitic ferrite is far greater than expected from equilibrium. You can find evidence for this in Bainite in Steels, page 38. Alternatively, look at the original paper, where the retention of excess carbon in the bainitic ferrite was attributed to the trapping of carbon at dislocations. Indeed, you can can find there, direct evidence for the trapping of carbon at dislocations.

On your second point, although bainite grows without diffusion, it forms at temperatures where the carbon that is not trapped at dislocations can escape rapidly from the supersaturated ferrite into the residual austenite. This process of ferrite decarburisation has been modelled in terms of the diffusivities in ferrite and austenite and the various limiting concentrations.

I have a question for you. It is about the different modes of joining for metals.  Could you help me by giving a small description of the processes of  soldering



so that I can translate your answer into French for students in Canada?

The American Welding Society defines welding as "a materials joining process used in making welds". This is clearly not an intellectually satisfying definition! The definitions below are from J. F. Lancaster, The Metallurgy of Welding, Brazing and Soldering, The Institute of Materials, London, 1965. 

Fusion welding is where the two edges or surface to be joined are heated to the melting pint and, where necessary, molten filler metal is added to fill the joint gap. The filler metal is usually not very different from the materials being joined. Such welds have three zones which are important: the fusion zone, the unmelted heat-affected zone whose microstructure is altered due to the heat from the welding process, and the unaffected parent plate. This definition is material independent, for example, polymers can be welded in this way.

It is possible to obtain welds without fusion, for example by bringing together clean metallic surfaces into intimate contact. Such solid state welding can be a prolem in the construction of space stations.

Both brazing a soldering involve a filler metal but with a much lower melting temperature than the components being joined. To obtain a satisfactory brazed or soldered joint, it is necessary for the filler to wet the component, to spread and to be drawn into the joint by capillary action.

When the solidus temperature of the filler metal is below 500 centigrade, the process is termed soldering, whereas when it exceeds this arbitrary temperature, the process is called brazing.

One important point is that whereas in welding the integrity of the joint depends critically on the mechanical properties of the filler metal as measured in isolation, the mechanical properties of brazes or solders are very poor if tested on their own. However, in brazing or soldering the filler metal thickness is kept very small so that the component itself constrains the braze and prevents it from failing. This is why it is vital to minimise the thickness of such joints.

I hope this is useful. Please send me the French version when you have done it. 

Do you know a good, fairly recent, review of the theory of
precipitation in the solid state? Unwisely, I have thrown out most of my vast collection of reprints and ageing has meant that I cannot remember where the good papers are! I do remember a good review by Ken Russell of about 15-20 years ago but cannot recall where it was published. No doubt more has been published since then. Just give me a lead and I'll do the hard work.

The very best review on the general theory of precipitation is Christian's 1975 Theory of Transformations in Metals and Alloys, Pergamon Press, Oxford.

Russell's papers that I know about are also general:
boundary nucleation kinetics, Acta Metallurgica 17 (1969) 1123-1131;
Linked flux analysis of nucleation in condensed phases,
Acta Metallurgica 16 (1968) 761-769; Nucleation - Mostly Homogeneous and in solids Solid-Solid Phase Transformations, TMS AIME Warrendale Pennsylvania  pp.371 - 398
1982;  Precipitation at Interphase Boundaries,
Metall. Trans. A 9A (1978) 363
(with Aaronson).

But I cannot recall any review of the  theory of precipitation in steels. There are many reviews of experimental observations, for example by Honeycombe.

I hope this is helpful.

In a  MIG/MAG welding-wire (low alloyed steel) the amount of surface slag is greater than with another similar wire. I have analysed the wires and found that there is a slightly difference in sulphur content  between the wires.  In the weld metal the difference is
smaller. I have read about the Marangoni effect and how it effects the
circulation in the weld pool. Do you think that this effect may be
"responsible" for the differences seen in slag distribution, and in that case is the small difference observed large enough? Or do you have any other suggestion? There is a slight difference also in Mn content. Do you think this is significant for the formation of surface slag?

There is a simple way of telling whether the Marangoni effect is important. In austenitic stainless steels, the Marangoni effect was first noticed because of large changes in the penetration. You could measure the weld cross-sections and look for significant differences.

My own opinion is that this is not the issue. You are talking about the slag layer rather than the weld profile. Also, both the welds you talk of contain, I think, sufficient sulphur at concentrations greater than 0.01 wt %.

Unfortunately, I know VERY little about slags so I can't help much more, but I would suggest you look at the weld profiles just to eliminate the Marangoni effect.

         Thanks for your help.
         I have additional question.

         If there is IPS shape deformation without the
         shear component and the product is not
         plate-shaped, Can I say that the transformation
         mechanism of the product is not displacive?

If there is an invariant-plane strain shape change which has no shear component then the mechanism of transformation is ambiguous. A plate which forms in this way could grow by reconstructive or displacive transformation ASSUMING that there is sufficient atomic mobility at the temperature where it forms.

One important point - whatever the mechanism of transformation, any shape change will lead to a strain in the surrounding material. This strain energy is important and its consequences should not be ignored.

Is there any advantage if the solidification is first as ferrite in case of weld metal solidification. What may be the advantage if on the other hand the first forming phase during solidification is austenite.

It is known that welds which solidify in a fully austenitic state have a greater tendency to hot-cracking and microfissuring. This can be avoided by promoting primary solidification to delta-ferrite. This is partly to do with the different solubilities of impurities in austenite and in ferrite.

Dear Sir.

There is usually a 'incubation period' in the TTT diagram. Why does this period appear? Can this period be regarded as the time for nucleation?

No, the incubation period cannot be regarded as a reflection of nucleation alone. All TTT diagrams measure the fraction of transformation, which is a combination of nucleation and growth. Therefore, the incubation time corresponds to the time taken to detect a given fraction of transformation. That fraction depends on the resolution of the experimental technique. Typically, 5% transformation.

I want to know the mechanism of carbide accompanying the reaction of bainitic ferrite and why bainite reaction is slow even though it grows diffusionless.

In the 'Bainite in steels', I don't know the meaning of Fig.6.1.
And I can't understand the sentence that 'This indicates that the efeect of solute additions on the nucleation of Widmanstatten ferrite and bainite is more than just thermodynamic in the section 6.1.2.

In upper bainite, the carbides precipitate from carbon-enriched austenite. In lower bainite, there is additionally, some precipitation from supersaturated ferrite.

At the temperatures where bainite and martensite form, the carbides grow by a paraequilibrium displacive mechanism. This means that they have the same Fe/X atom ratio as the matrix phase, where X is a substitutional solute. However, carbon must diffuse so they grow at a rate which is controlled by the diffusion of carbon.

The bainite reaction is not slow, the growth of bainite occurs three orders of magnitude faster than would be permitted by the diffusion of carbon, and very very many orders of magnitude faster than permitted by the diffusion of substitutional atoms.

However, the growth rate is not as fast as the speed of sound in the metal. This is because the shape deformation induces plastic deformation in the adjacent austenite, which damps the motion of the interface.

Notice that even martensite need not, and frequently does not, grow fast. This is not a requirement for a displacive mechanism of transformation.

Referring to the next part of your question, Widmanstatten ferrite does not form at a fixed undercooling below the Ae3 temperature. Instead, the extent of undercooling increases as the stability of austenite is increased (i.e. when the Widmanstatten-start temperature W_S is decreased). Therefore, it is not possible to explain changes in W_S by postulating a fixed undercooling below Ae3; the effect of alloying elements is therefore not purely thermodynamic. Kinetics become more difficult at lower temperatures, so if an alloying element depresses Ae3, W_S is depressed even more.


I read section 6.1.2 in the book 'Bainite in Steels'.
You say that equation(6.1) can be regarded as a universal nucleation function and used to predict the highest temperature for displacive transformation. I cannot understand something. Do you intentionally omit the terms about the effect of alloying elements? I cannot understand well.

No, there is no disregard of alloying elements. Solutes affect the free energy of transformation. This is how the influence of alloying elements manifests in the calculation of transformation temperatures.

In a plain carbon steel, the martensite-finish temperature M_f drops with increasing carbon content faster than the martensite-start temperature M_s does. Why?

The martensite-start temperature has fundamental meaning whereas the martensite-finish temperature is more arbitrary. Thus, from the Koistinen and Marburger equation, the fraction of martensite (f) is given by

1-f = exp{-0.011[M_s-T]}

where T is a temperature below M_s. According to this, there will always be some retained austenite. Therefore, M_f can only be defined according to specified fraction of martensite - it does not define the end of the martensitic transformation.

Unfortunately, most papers and textbooks do not consider this and simply quote an M_f. Therefore, I would not take trends in M_f seriously.

I would like to know what the term "topologically close-packed" structure means. I found that, it is also called as "geometrically close-packed structure", mostly abbrevated as TCP or GCP respectively. As the name says, has it got something to do with the spatial relations like volume and density.

A close-packed structure is one in which the atoms touch all the near neighbours, in such a way that the density is maximised. The cubic close-packed structure is (austenite in steel, or copper are good examples), where in the {111} plane each  atom touches six neighbours along the three <110> directions in that plane. These close-packed planes are then stacked in a sequence which repeats every three layers (...ABCABCABC...), such that each atom in a close-packed layer touches three in the plane above and three below. This gives the maximum packing density for hard spheres of uniform size.

A topologically close-packed structure (or GCP) is one in which the stacking and arrangement of the host atoms is like that in a truely close-packed structure (like Cu), but the atoms do not in fact touch. This allows other species of atoms to occupy the spaces.

For example, in certain oxides, the oxygen atoms are in a close-packed arrangement with metal atoms in the interstices. However, the atoms are not close-packed since they may not touch each other.

I came across your program on the MAP site which  predicts stress/strain data for the hot torsion test. I am currently doing a PhD in the area of low temperature torsion. I was wondering if the program also works at low temperatures or has it been written exclusively for high temperature deformation?

The MAP_NEURAL_HOT_TORSION deals with torsion tests on austenite in steels. It follows that the lowest temperature that it could be applied to for your steel, is where the steel is fully austenitic.

Yes, but it does seem from the literature that the difference between the M_s and M_f temperatures increases quite substantially as the carbon concentration increases. Of course, although this is predicted by the Koistinen and Marburger equation, I don't understand the physical basis. Can you help?

The physical justification of the Koistinen and Marburger equation has been given by Magee (Phase Transformations, ASM, Metals Park, Ohio, 1970, eds H. I. Aaronson and V. F. ZacKay, page 118). The term within the exponential contains a rate of change of driving force with temperature. This term depends on alloy and hence the difference between M_s and "M_f" will have an alloy dependence.

Do some calculations using the Magee theory and see if they are consistent with the reported variations in M_s-M_f.

O.O. Miller ("Influence of austenitizing time and temperature on
austenite grain size of steel" , Transactions of the ASM, 1951, pp. 260.) analised the evolution of the prior austenite grain size (PAGS) in steels with a very wide range of carbon content (from low carbon to high carbon steels), with Mn ranging from 0.2 wt-% to 0.9 wt.-%, Si content about 0.2%, and some of them alloyed with Mo, Cr and V.
Miller plots PAGS vs austenitising temperature for constant austenitising time. My question is the following. As austenitising temperature increases PAGS increases, as expected. However, there is a particular temperature at which a decrease of PAGS is detected. For instance, in a particular steel, at 950C-5h a PAGS of -2 ASTM is determined and at 1150C-5h, the PAGS decreases to 1 ASTM. In both cases the PAGS is homogeneous. This fact is detected in all the steels studied in the above mentioned work. Do you you know the reason of this behaviour?

This is strange. However, he does seem to offer a possible explanation on page 273, which I paraphrase as follows:

At low temperatures the small inclusions cause heterogeneous grain growth with a few large grains dominating the field of observation. As the temperatur is raised, the inclusions dissolve giving uniform grain coarsening and hence a finer grain structure. Then, a further rise in temperature simply increases the scale of uniform coarsening.

Why was it that some people apply two tempering heat treatments to creep-resistant power plant steels after normalising?

Here are possible reasons:

  1. After austenitisation and cooling, the steel may contain retained austenite whose decomposition can influence dimensional stability. Therefore, before machining, an initial temper is given at about 500 centigrade to induce the decomposition of the austenite. A second tempering treatment may then be given after some other fabrication process. However, this does not explain the very high temperatures used in many double tempering treatments.
  2. An initial temper to soften the material for fabrication, followed by a stress-relief heat treatment.
  3. A second temper may be required in order to stress relieve after a welding operation.
  4. A historical accident, the reason for which has been forgotten!

In Joe Robsons thesis he refers to the martensitic structure of a 10Cr power plant steel as having lath boundaries and block boundaries, with the block boundaries containing a number of laths (p.28). Are the block boundaries he refers to prior austenite grain boundaries?

No, an austenite grain is partitioned into several packets (blocks), each of which contains parallel laths with the same habit plane trace on a two-dimensional section.

The laths in a packet tend to be in similar crystallographic orientation so from the point of view of cleavage fracture, the packet is the unit of fracture rather than the lath.

There is a good review on this subject by T. Maki, Materials Science Forum vols. 56-58 (1990) 157-168. This particular issue of the journal is in fact the proceedings of ICOMAT 89, where ICOMAT stands for "International Conference on Martensitic Transformations. In fact, Maki has written extensively on this subject.

If a plate shape precipitate grows by a reconstructive mechanism and produces IPS shape deformation, would you expect an increase or decrease  in the lengthening and thickening growth rates when the pre-transformed matrix is deformed significantly?   

When a plate grows by a reconstructive mechanism, it does result in an invariant-plane strain shape deformation, but there is no shear component. There is diffusion which eliminates the shear. Therefore, any defects in the parent phase are eliminated as the plate grows. This adds to the driving force and hence accelerates transformation. This is akin recrystallisation.
There are no circumstances I can think of in which a reconstructive transformation would be retarded by first deforming the parent phase.

I am currently writing up my 4th year thesis on the strain ageing of drawn pearlitic wire under the supervision of Prof. George Smith. George suggested that I ask you what value you would assign to the
ferrite/cementite interface energy. The only value I have found so far is 0.7J/m^2.

Puls and Kirkaldy, who did a lot of work on the pearlite reaction, quote a value of the ferrite/cementite interfacial energy in Metall. Trans. 3 (1972) 2777-2796. Someone has borrowed my copy of this paper otherwise I could quote the value to you.

Based on the microstructural evolution,  please explain
the prediction of high temperature strength of neural

The microstructure of creep resistant steel is determined by
chemical composition and by heat treatment. The heat
treatment includes service conditions. Hence, all the
information about microstructure is included implicitly in
this set of variables, which can therefore be used to
completely define the problem. Of course, the number of
variables is large. And it should be large because each
component of heat treatment and composition has a role.
This is obvious from physical metallurgy. The neural
network has had a liberating effect on materials science in
that it allows one to perceive the effect of a large number
of variable.

Another approach is to include the microstructure directly
in the list of variables (see for example, Materials Science
and Technology, Vol. 11, 1995, 1046-1051.). And every
aspect of microstructure can be included to reveal the
detailed role of each component. The trouble with this
approach is that there are few data available where the
relationship between microstructure and properties is well
or completely characterised.

Finally, where physical relationships are known, they can
be embedded in a neural network. Generally, however,
physical models are simply inadequate when it comes to
making quantitative predictions, because they do no address
real complexity. Thus, there is no model other than one
using neural networks, for the prediction of creep rupture
strength. And this is in spite of the fact that there has
been an incredible amount of research over a very long
period of time.

Your work involves the modelling of creep strength with a large number of variables. Given the old principle that "one can fit an elephant with a fourteen parameter equation", it striked me that
a "forty variable" model for stress rupture requires some

An elephant should not be modelled as a sphere. An
elephant truly requires a large number of variables to
describe it properly. Thus, creep rupture strength is not a
simple property - I estimate that a total of 114 variables
affect the creep rupture strength of a ferritic steel. This
is known from physical metallurgy experiments. However, we
have not yet succeeded in including as many as 114, but we
are making progress.

I suppose that your question also implies that there may be a problem with overfitting. The way in which this is avoided is described in ISIJ International, Vol. 39, 1999, pp.

I've read "Bainite in Steels". Do you think that the nucleation of bainite is occurred by operational nucleation mechanism(pre-existing embryo model) and
nucleation rate is controlled by the movement of the transformation dislocation? I've read that the activation energy barrier is proportional to the chemical free energy change. If so, I think the nucleation rate for bainite decreases as the transformation temperature decreases and thus, transformation start time for isothermal transformation
does also decrease. However, I can not see this phenomenon yet. Am I misunderstanding the contents of book?  And I tend to describe the behavior of bainite reaction by using Avrami equation. I want to know the effect of precipitation of carbide on the kinetics. Could you give advice to me about the my work? I'm waiting for you help.

The evidence for the nucleation of bainite is that the activation energy is that for the movement of the interface (i.e. the interface exists). The only barrier to nucleation is the thermally activated motion of the interface rather than a heterophase fluctuation as in classical theory.

You are not correct in assuming that because the activation energy is proportional to the driving force, the barrier increases as the driving force increases. Indeed, you will see from equation 6.11 or from equation 6.35, that the activation energy for interface motion decreases as the driving force Delta GCHEM becomes more negative.

G* = Go* ...... + A Delta GCHEM
where A is a constant.

Go* is a positive constant, and as Delta GCHEM becomes more negative (i.e. the driving force increases), G* decreases.

For theory on carbide precipitation, see Materials Science and Technology, Vol. 6, 1990, pp. 592-603.

Please can I have copies of the two images showing an organised and
a chaotic set of ferrite plates?

The pictures were published in "Reaustenisation in Steel Weld Deposits"  Proceedings of an International Conference on Welding Metallurgy  of Structural Steels,   published by The Metallurgical Society of the AIME,   Warrendale, Pennsylvania.   Edited by J. Y. Koo, 1987, pp. 549-563.    J. R. Yang and H.K.D.H. Bhadeshia 

You can download the images from the following locations:

Chaotic plates >

I observed a cellular substructure in the as-welded specimen of 9 wt% Ni steel. But heat treating the all-weld specimen at 1000 deg C for 1 h and quenching in water lead to a decrease in the substructure. I couldnt understand this. Will seggregation be removed if the weld is austenised completely. And also that, there is a drecrease in the hardness of the water quenched specimen from that of the as-welded one. I think this should be due to the presence of retained austenite.

The cellular structure that you observed at first is due to chemical segregation. Although the segregation is hardly affected by the heat treatment at 1000 C for 1 h, the microstructure is.

The original microstucture was due to solidification. When you take the sample into the austenite phase field, and then quench, it becomes martensitic. The martensite forms in all regions so the microstructure appears more homogeneous. However, if you conduct microanalysis, you will still see the original segregation pattern.

To eliminate segregation requires heat treatment at about 1200 C for 3 days.

I cannot explain the change in the hardness without more details. How much did it change by?

Following is the detail regarding the 9 wt% Ni weld :

                                Hardness / VHN

As welded                           373

Water Quenched                      333

Should this not be attributed to the presence of retained austenite, due to quenching from 1000 C when the sample was in fully austenised condition.

The difference in hardness is substantial. It could be due to the presence of retained austenite in the as-welded sample, but you would need to establish that using X-ray diffraction. Then you would need to have an explanation of why the as-welded sample contains more austenite than that reheated and quenched. The cause is not entirely obvious to me.

You calculated that the diffusion distance of an iron atom in three weeks at 150 C is about 1.0E-17 m. Could you tell me the same data for 190 C for 2 weeks?. I have tried myself to get that data from Fig. 2.10 in Bainite in steels, but the graph only can tell for temperature higher than 300 C.

The self-diffusion coefficient of iron in austenite is

D = 4.9×10-5 ×exp{-Q/RT} m2  s-1

where Q = 283926 J mol-1. These data are obtained from the Handbook of Chemistry and Physics, CRC Press, 57th edition, 1976-1977.

The mean diffusion distance is 2(Dt)0.5. Therefore, for a heat treatment of 3 weeks at 453 K, the diffusion distance is about 5.6×10-17 m.

For a heat treatment of 2 weeks at 463 K, the corresponding diffusion distance is
1.6×10-12 m

Shouldn't the temperature be 423 K instead of 453 K in the first part?

Yes, sorry, this is just a typographical error, 423 K was used to get the result in the first part.

I get 5.6e-16 for the second part of the calculation.

Once again, you are right, sorry!

To be able to determine the stability of austenite which forms during
post-weld heat treatment (say 650 degc for 5 min) we need to know the
chemical composition of austenite as a function of time (since the
transformation may not be completed after 5min). Then one uses the
chemical composition of the newly formed austenite to determine Ms.  But how can we predict the change in austenite chemical composition as a function of time ?

Assume first that there is local equilibrium at the interface as austenite forms, at 650 Centigrade. Assume also that the diffusivities of the components in the steel are of similar magnitudes. In that case, the composition of the austenite will not change with time, simply the volume fracation.

If you now remove the assumption about diffusivities, but maintain local equilibrium at the interface, then in an alloy steel, the tie-line determining the composition of the austenite and ferrite at the interface, will change as transformation progresses. (see tie-line shifting, Progress in Materials Science, 1985, vol. 29, page 347.

Now assume that local equilibrium is not maintained at the moving interface. In that case, there is an infinite number of possibilities. However, you might assume that austenite forms by paraequilibrium transformation. The subsequent enrichment of austenite in nickel can be dealt with by a simple diffusion model (see for example, Bainite in Steels, equation 4.4).

My advice is that until you have experimental proof that the composition of the austenite is not constant during the heat treatment at 650 C, take it to be the composition given by the phase diagram, i.e., the assumptions outlined in paragraph 1 of this answer.

Many researchers have studied phase transformation kinetics using the dilatometer.

Can this technique distinguish the Widmanstatten ferrite from allotriomorphic ferrite?

If it is possible, How can we distinguish both phases from the dilatometer data  under isothermal condition?

No, I do not think this is possible. A dilatometer only monitors dimensional changes, which for polycrystalline specimens should be more or less identical for both phases.

Could I know what is   "Creep feed grinding". I found this term in the following paper

H.K.D.H. Bhadeshia.,"Neural Networks in Material Science", ISIJ Internatioanl, vol. 39 (1999), No.10, pp966-979

Conventional grinding involves a grinding wheel reciprocating rapidly (say 40 mm/s) over the workpiece as it gradually lowers to it's final depth of cut. Creep-feed grinding is a process where a formed grinding wheel is plunged into the workpiece, slowly (say 3 mm/s) producing a finished part in a single pass.

See animation

I am a mechanical engineering student at the Peninsula Technikon in South Africa.  I am currently involved in a study of welding metallurgy.  The objective of this study is to control the microstructure of a welded steel structure.

In this study I will be using SYSWELD software and the Finite element method package.

My problem is that I do not have  access to the latest material related to my study so it is not easy to do my work. I therefore need some information pertaining to the control of weld microstructures.  If possible can you suggest some websites that could be useful to me regarding this subject matter.

Here are a few references which you can obtain from the Institute of Materials, 1 Carlton House Terrace, London SW1Y 5DB, U. K.:

Mathematical Modelling of Weld Phenomena,
       Published by the Institute of Materials, 1993, pp. 109-182.
       Edited by H. Cerjak and K. E. Easterling.
       H. K. D. H. Bhadeshia and L. E. Svensson

Mathematical Modelling of Weld Phenomena - II,
       Published by the Institute of Materials, 1995, pp. 71-118.
       Edited by H. Cerjak,
       H. K. D. H. Bhadeshia

Mathematical Modelling of Weld Phenomena - III,
       Published by the Institute of Materials, 1997, pp. 229-284.
       Edited by H. Cerjak and H. K. D. H. Bhadeshia.

Mathematical Modelling of Weld Phenomena - IV,
       Published by the Institute of Materials, 1998, pp. 235-246, 302-320.
       Edited by H. Cerjak and H. K. D. H. Bhadeshia.

You can also look at the Materials Algorithms Project web site.

The topics that are also in my study are the following: weld thermal
  cycle, heat affected zone.  Also I would like to more about mathematical modelling of weld microstructures.  Thank you very much.

All of these topics, and more, are discussed in detail in the books I described earlier. 

My query is concerned with the modelling of austenite formation from
martensite and retained austenite. The retained austenite has a composition given by equilibrium with delta-ferrite at a very high temperature.

I have a problem regarding the Cr profile in the model I try to do.
I have calculated the Cr content of austenite and ferrite in a Fe-13wt%Cr binary alloy. Under local equilibrium condition the Cr concentration at the interface is increasing for both phases. For instance, at 940 degc 8% of austenite is predicted to be stable and its Cr content is about 11.3 % while ferrite's Cr content is about 13.1 %. However, the initially present film of retained austenite has composition which is given by the composition of the austenite at high
temperature. Therefore the composition of the retained austenite is
initially 13%Cr.

Consequently, I don't know how to draw the profile and how to calculate the supersaturation coefficient in this particular case. Which assumptions should I make ?

On the other hand, do you have any suggestions regarding the values I
should assume for:

a) the total area of the ferrite/austenite interface (c*c)
b) the number of  particle of austenite per unit of volume (Nv) c) the minimum detectable thickness increase Delta am

You are currently working on a model which does not allow diffusion within the growing phase. Therefore, you have to assume that the concentration in the austenite at the interface is that appropriate for the transformation temperature, and not the one for the high temperature. There will therefore be a profile within the austenite which is stepped at the position where austenite growth started.

A more sophisticated model, which can be done as the next step using finite difference, could allow for diffusion in both the growing and diminishing phases

(a) You can esitmate the total ferrite/austenite interface per unit volume (S_V) from the parent grain size, assuming that all
the parent phase grain boundaries are covered with a thin layer of austenite.  Thus, S_V=2/L where L is the mean lineal intercept defining the grain size of the parent phase.

(b) The number of particles per unit volume is not relevant because you use S_V, and because there is no nucleation.

(c)  A dilatational strain of 0.0001 is detectable, so a volume strain of 0.0003 is detectable.

In creep resistant steels, it is possible to
find many alloys which have Cr at around 2.25 wt% or slightly higher, and  then in the range 9-12 wt%.

However, it is rare indeed to find intermediate concentrations such as 5 wt% alloys. Why is that?

There are some alloys with intermediate levels e.g 3CrMo and 5CrMo. However these are not widely used so you are broadly correct.

Additions of Cr to mild steel improve hardenability, creep strength, oxidation and erosion/corrosion properties. These low alloy steels such as 2.25%CrMo steels are generally ferritic/bainitic in structure.My impression is additions of Cr beyond about 2.5% don-t give any further significant advantage until you reach 9% when martensitic structures become possible due to suppression of the
bainite reaction . This then allows significantly greater hardenability and creep strength (with other alloy

But there must be some other reason for not using the 5 wt% alloys because not all alloys are used at temperatures where the oxidation resistance has to be good. For example, the 2.25Cr1Mo alloy has sufficient oxidation resistance at 565 C.

Above about 2.5Cr creep strength doesn't improve further unless you add something thing else. At 9-12Cr you start reaping the benefits of corrosion resistance and martensitic transformation on air cooling.  A 5CrMoV was tried in the 60's but there were major cracking  problems welding it and it was abandoned. However, I do not have references to support this, it is simply a rumour about an alloy called "Rex 500".

The rupture properties of 5CrMo are given in BSI public discussion document PD6525 Part 1 (1990). Rupture strength is lower than that of 2.25CrMo. The CrMo steels depend on Mo2C for creep strength. Maybe higher Cr destabilises this in favour of M23C6?

That is very interesting because at the moment our model, presumably incorrectly, predicts a monotonic increase in the creep strength with the Cr concentration.

Following your suggestion, we shall also try and investigate changes in the carbide precipitation sequences as a function of the Cr concentration.

Hi, I work for Battelle and am looking for a person or a book that could tell me about welded joints at high temperatures.  Specifically, I'd like to calculate when the welded joint in my experiment will fail/fracture due to the application of high temperature for a length of time.  I realize creep will play a significant role.  May you please send me graphs, informations, or provide me with a name of a person or reference.  I hope you can help.

Design of Creep-Resistant Steel Welds", Trends in Welding Research, eds S. A. David, T. DebRoy, J. A. Johnson, H. B. Smartt and J. M. Vitek, ASM International, Ohio, 1999, pp. 795-804. By H. Bhadeshia

"Modelling the Creep Rupture Strength of Ferritic Steel Welds" Science and Technology of Welding and Joining, Vol. 5, 2000, 81-90.   D. Cole, C. Martin-Moran, A.G. Sheard, H. K. D. H. Bhadeshia and D. J. C. MacKay 

For computer programs specifically for doing the calculations you require, see the review on the subject.

Does the thermal conductivity of a material decrease with increase in grain size. ? Is so would you mind letting me know the mechanism by which this happens ?

Thermal conductivity is determined by the rate at which phonons can transfer heat along a crystal. The phonons are scattered by barriers which determine their mean free path. Grain boundaries scatter phonons and hence reduce thermal conductivity.

Do you think a magnetic NDT method could be used for nickel-base
alloys? I've read that antiphase domain boundaries pin domain walls
but I don't know how these evolve over time in, e.g., a power station
using Ni alloy components instead of steel. 

The antiphase boundaries are between domains of structural order, rather than magentic order. And nickel is only weakly ferromagnetic with a Curie temperature of just 380oC. Of course there are nickel alloys which are ferromagnetic, but I am fairly sure that the nickel-base superalloys of the type used for elevated temperature applications are not ferromagnetic. Therefore, I would not have thought that you could use the magnetic method for nickel as you do for steel.

I am a student of Science Material in Ferdowsi univer. of Mashhad Iran. Please send me information about "316l Stainless steel".
Thank you.

  316 is a molybdenum-containing austenitic stainless steel with a specified chemical composition

0.08C max, 0.03N, 17-19Cr, 10-14Ni, 2-4Mo and 1.5Mn wt%

316L on the other hand, has a lower carbon concentration of 0.03 wt% maximum to prevent sensitisation.

Some of the links did not appear properly in your answer.

Sorry!. For computer programs capable of doing the calculations you require, see The Materials Algorithms Library

You may also be interested in a review on the subject.

It would be greatly appreciated if you advise me about following basic questions about bainite:

1) How should I get Delta-G for bainite nucleation ? Supercooling from T_0?

2) How should I calculate carbon concentration in bainite and at interface  bainite /austenite?

3) How should I calculate size and growth of bainite?

Bainite nucleation is defined from the point where the magnitude of GMAX, which is the driving force for nucleation, exceeds the magnitude of GN, which is the universal nucleation function.

The activation energy for nucleation is then proportional to the difference between these two quantities.

Our model for bainite considers the overall transformation kinetics to be controlled by the nucleation process. Once nucleation happens, the growth of a sub-unit is fast. Furthermore, each sub-unit only grows to a limited size as plastic accommodation of the shape change stifles the growth of an individual unit. Therefore, we assume that each nucleus transforms a fixed volume of austenite. In this case, it is not necessary to consider the growth kinetics at all.

In our model, the bainite grows without diffusion, the partitioning of carbon occurring subsequent to growth. Therefore, the compositions of the austenite and bainite at the interface are identical.

We have measured the amount of austenite by X-ray diffraction and I don't understand the results. To give one example: X-ray measurements suggest that the austenite content the metal changes from 3%  to 6.6% after aging 48h at 100 C and to 10.6% after aging 16h at 250C. We also measure a small decrease in "ferrite content" and hardness after heat treatment.

Is it possible that martensite reverts to austenite!? or do I have to look for explanations related to experimental problems with X-ray diffraction.

What confuses me further is that hardness and magnetic response points in the same direction as X-ray measurements.

Any comment would be appreciated.

I am afraid this does not seem possible, because the diffusion distance at these temperatures is incredibly small, and it is unlikely that austenite reverts by a displacive mechanism.

In your X-ray measurements, do you use a single austenite peak and a single ferrite peak, because if you do, then there will be a strong effect of texture which can vary between samples. Have you used the same sample before and after tempering? Have you removed the surface layer by chemical polishing before X-ray analysis?

In general, X-ray measurements of austenite content are best done using three peaks from each of austenite and ferrite, to average out texture effects.

One thing you might do, is a blind test in which you give two samples from the same condition for X-ray analysis to check the reproducibility of the results.

The hardness must of course change due to heat treatment, but I am not sure why the magentic response points in the same direction.

I would be grateful if you could keep me in touch because we also do many X-ray measurements.

If austenie that has been sufficiently homogenized, is then deformed at temperatures below the recrystallization temperature, is it likely that substitutional solutes or carbon conentrations may exhibit some type of segregation, either by volume diffusion to or away from dislocations or a sweeping up type process where as the dislocation is gliding it sweeps up substitional solutes? 

Yes, you would expect misfitting atoms to locate preferentially at defects.

   This is a question about the book "Bainite in Steel"

On page 9 at Section 1.2.4 "Thermodynamics" in Fig 1.4. To' temperature is explained as the temperature at which austenite and ferrite with strain energy have the same free energy.

  On the other hand in the text (page 10) Tom temperature is explained as the temperature at which austenite and martensite (i.e. supersaturated tetragonal 'ferrite') have the same free energy.
     Could you explain me the relation between To' and Tom.

Let us assume that ferrite has a body-centered cubic crystal structure but martensite in steel has a body-centered tetragonal crystal structure. TO is the temperature at which ferrite and austenite of the same chemical composition have the same free energy. On the other hand, TOM is the temperature where austenite and tetragonal-martensite of the same composition have the same free energy.

The difference is that with ferrite, the carbon atoms are distributed at random on the three different sublattices of octahedral holes in the body-centered cubic lattice. When the carbon atoms are Zener ordered on just one of these sublattices they cause tetragonality.


Is the soulte content in ferrite that I get from MTDATA same as the
solute content in ferrite that is in equilibrium with precipitate.

The solubility of a solute in a phase can only be defined with respect to another phase. For example, the solubility of carbon in ferrite which is in equilibrium with cementite is quite different from the case where it is in equilibrium with graphite.

When you did the MTDATA calculations, you allowed a number of phases and components to exist. Of these, the calculations may have predicted one phase other than ferrite to be stable under the conditions specified. The solubility of carbon in ferrite will then be with respect to equilibrium with that other phase.

My question is concerned with which is the most importnat factor to achieve ultrafine ferrite grain size via reconstructive transformation from austenite deformed at large undercooling, very low temperature deformation, very large degree of supercooling, or dynamic transformation occuring during deformation? I am puzzled.

Some reserches described that first impingement of such ultra ferrite grains formed during dynamic trasformation, followed by intergranular Fe3C precipitation is a key factor for such ultra-refinement. However,
in our experiment (motivated by you), using steels with relatively high quench hardenabilty, reheating up to 750 C after deformatuib at low temperature (570 C) in austenite field also allows ultrafine ferrite grains to be formed successfully and reproducibly despite the fact that supercooling appears not so large and also dynamic transformation is unlikely in this process. The same result is obtainable using 0.2C-0.83Mn-2.0Cr wt% steel as well as 0.2C-2.0Mn wt% steel. Based on this result, I believe that low temeperature deformation probably plays a vital role for such refining. Also,I reconfirmed that low temperature deformation causes microbands instead of dislocation cells, and the microbands persist during reheating up to 750 C. Therfore my cuurent interpretation is that low temperature deformation causing microbands is the most important for reducing ferrite grain size down to 1 micrometer, though proper supercooling is required for making the microbands to be effective nucleation sites. I appreciate it in advance if you would kindly comment on this issue for me. Low temperature deformation??  Low temperature transformation?? or Dynamic transformation??


There are a number of well-known effects of deformation on the kinetics of reconstructive transformations. When austenite is deformed, there is an increase in the amount of austenite grain boundary surface per unit volume. See for example,

I Czinege and T Reti,
Determination of local deformation in cold formed products by a measurement of the geometric characteristics of the crystallites
Eighteenth International Machine Tool Design and Research Conference, Forming, Volume I (1977) 159-163.

This increase in surface area per unit volume naturally leads to a greater number density of nucleation sites for ferrite. However, the deformation also increases the effectivness of the austenite grain boundaries as nucleation sites, so the the number of ferrite nuclei per unit area of grain boundary increases. This is because the deformation introduces steps in the interfaces.

Other defects, such as microbands introduced into the austenite by plastic deformation, are also well known to act as heterogeneous nucleation sites for ferrite.

The deformation of austenite may also cause strain-induced precipitation of carbides, which will affect the development of microstructure.

In a reconstructive transformation, the defects present in the austenite are generally eliminated by the growing ferrite. This accelerates transformation by providing an increased driving force. This in turn may lead to grain refinement.

Reducing the transformation temperature helps sometimes because the increased driving force will lead to a greater nucleation rate, but this may become complicated by the onset of other transformations.


It is very likely that the lower the temperature of deformation of austenite, the greater will be the retained defect density and hence will lead to a finer ferrite grain size. However, dynamic recrystallisation of austenite, which happens during higher temperature deformation can also help because it leads to a finer austenite grain size.

Transformation during the deformation process is complicated because some of the ferrite will then be cold-deformed.

To summarise, anything you can do to enhance the nucleation rate of ferrite should help achieve a finer grain size, including the deformation of austenite.

In general, the 0.2% proof stress (YS) is estimated from hardness as follows:

  YS= Vickers Hardness/3

However, harndess is more related to ultimate tensile strength than yield strength as it involves plastic deformation around indentation. The above formula won't  contain  work hardening parameter.

You are right that a hardness indentation is associated with considerable plastic strain. The following paper includes a conversion from hardness to strength whilst taking into account the work hardening behaviour:

J R Cahoon, W H Broughton and A R Kutzak

Determination ofyield strengthfrom hardness measurements

Metall Trans 2 (1971) 1979

You might also wish to look at a special issue of Philosophical Magazine A 74 (1996) which deals with hardness measurements.

What is the mechanism of bainitic carbide growth? 

There is a great deal of evidence that the carbides associated with these transformations form by displacive mechanism without the partitioning of substitutional solute.

The mechanism must obviously involve the  diffusion of carbon, but not of substitutional solutes (X) or  iron atoms.  Considerable experimental data  show that the carbide precipitation associated with bainite and  martensite does not lead to a partitioning of substitutional solutes. The precipitation can occur under conditions where the diffusion rates  of iron and substitutional atoms are incredibly small  compared with the rate of precipitation.

It has in fact been believed for some time
that  the cementite lattice may be generated by the deformation of the ferrite crystal structure, at a rate controlled by the diffusion of  carbon into the appropriate sites. The Fe/X ratio thus remains constant everywhere and subject to that constraint, the carbon achieves equality of chemical potential; the cementite is then said to grow by paraequilibrium transformation. The way in which the ferrite lattice could be deformed to produce the right arrangement of iron atoms needed to generate the cementite has been considered by Andrews and Hume--Rothery et al., and the subject has been reviewed by Yakel. High--resolution transmission electron
microscopy supports the idea that the carbide particles grow  by displacive transformation (Sandvik,  Nakamura and Nagakura and Taylor et al.).

In a remarkable experiment, Babu et al. have shown using the atom--probe technique that the cementite obtained by tempering martensite is forced to inherit the silicon concentration of the martensite. This is in spite of the fact that the equilibrium solubility of silicon in cementite is negligible.

The number of carbide variants which grow in any given plate of martensite decreases when the virgin martensite is tempered under the influence of the stress. This response of the
microstructure to an externally applied stress is consistent with a displacive mechanism of transformation.

To summarise, it appears that substitutional solute atoms are trapped in the cementite when the latter precipitates during the bainitic or martensitic transformations in steels. That is, the cementite forms by a paraequilibrium transformation mechanism. In silicon containing steels the free energy change associated with the paraequilibrium precipitation of cementite must be much smaller than when the cementite is free of silicon. It is probable  that this is what leads to   suppression of cementite in high--silicon bainitic or martensitic steels.

A major omission from the array of experimental evidence is that no one has as yet measured the shape deformation due to carbide precipitation. This is because the particles are so incredibly fine, but atomic force microscopy may yet prove possible as an aid to characterising the surface relief.

All the references quoted above can be found in  Bainite: Unresolved Issues

When austenite is deformed and allowed to transform to bainite, do bainitic ferrite plates nucleate intragranularly?  

I have not seen evidence for independent nucleation on the defect structure induced by deformation, but that does not mean that it does not happen.

  Thank you your answer. However, I want to ask you again.   Does "the size" mean either sheaves, lath or packet?
  From your answer "the size " correspond to sheaves.
  Is it correct?

The size here refers specifically to that of an individual plate (or lath). A sheaf or a packet is a collection of such plates.

Many thanks for your quick answer.
However, I have further question about the relation between To and To'.
I understand that To' is for "f+strain" and Tom is for martensite with ordered carbon. Then, what is To for?. What is "strain" for To'?
Please explain me again the relationship among To, To' and Tom.
With best wishes.

The T0 temperature is that at which austenite and ferrite of the same chemical composition have the same free energy. The ferrite is in this case a disordered solid solution and there is no strain.

However, for displacive transformations in particular, the shape deformation causes a lot of strain in the surrounding material. This strain has an associated strain energy which must be accounted for. In practice it leads to a reduction in the transformation temperature because the undercooling has to be larger to account for the strain energy. This gives the T'0 temperature.

T0m is as T0 but with the carbon atoms in the ferrite in a Zener ordered configuration.

I would like to get information regarding formation of austenite in carbon steels  under extremely high heating rates 10000 K/s (as encountered during laser surface hardening).  This includes mechanism involved, shift of Ac1 and Ac3 with incresed heating rates etc.. Not much literature is avlilable in this regard. Which books should i refer to to get these equations and the mechanisms involved in such rapid transformation.

Please find below a list of references on the mechanism of austenite formation. A few deal with high heating rates but I doubt if you will find anything on 10,000 K/s.

The physical metallurgy of low-C low-Alloy steels containing B
K J Irvine, F B Pickering, W C Heselwood and M Atkins
J I S I 186 (1957) 54-67
The formation of austenite in a low-alloy steel
N C Law and D V Edmonds
Metall Trans A 11A (1980) 33
The growth of austenite as related to prior structure
A E Nehrenberg
J of Metals 188 (1950) 188
The later stage of reverse transformation in low-C low alloy stee
S Matsuda and Y Okamura
Trans I S I J 14 (1974) 445
A CCT transformation atlas for commercial power plant materials
P J Alberry, B Chew T C Gilmore P Woodlands
CEGB internal report TPRD/M/1630/R87 July 1987
Intercritical reheating during welding/HAZ structure and properti
D J Sparkes
Welding Institute research report 352/1987 October 1987
Effect of cooling rate on intercritically reheated ...
C L Davis and J E King
Materials Science and Technology 9 (1993) 8-15
Metallurgical factors controlling HAZ toughness in HT80
R Yamaba H Chiba H Gokyu T Komai
IIW Doc IX 1422 86
On the reversibility of the alpha (BCC) to gamma (FCC) transfm...
W C  Ku and R P Zerwekh
Solid-Solid Phase Transformations, TMS AIME Warrendale Pennsylvan
Reaustenitisation in steel weld deposits
J R Yang and H K D H Bhadeshia
Welding metallurgy of structural steels, ed. J Y Koo, TMS AIME..
Computer simulations of the austenite/ferrite diffusional...
J Agren
Acta Metall. 30 (1982) 841-851
The martensitic transformation in the iron-nickel system
L Kaufman and M Cohen
J of Metals, October 1956, 1393-1401
Reaustenitisation from bainite
J R Yang and H K D H Bhadeshia
Phase Transformations '87, IOM, London, ed. G. W. Lorimer
Austenite memory effect in 1Cr-1Mo-0.75V(Ti,B) steel
S T Kimmins and D J Gooch
Metal Science, 17 (1983) 519-532
The kinetics and structural mechanisms of phase transformations..
V D Sadovskii
Probl. Metalloverskie Term Obrab, pp.31-52 (1956)
Laser transformation hardening of a high purity Fe-C-Cr alloy
J R Bradley and S Kim
Scripta Metall. 23 (1989) 131-136
Kinetics of Austenite formation from a spherodized ...
R R Judd and H W Paxton
Trans Met Soc AIME 242 (1968) 206-215
Effect of alloying elements on the formation of austenite and
M Hillert, K Nilsson and L E Torndahl
J I S I  January 1971 49-66
On the thermodynamics of thermoelastic martensitic transformation
R J Salzbrenner and M Cohen
Acta Metall 27 (1979) 739-748
The effect of austenite ordering on the transformation temp......
M Umemoto and C M Wayman
Metall Trans A 9A (1978) 891-897
Dislocation structures produced by reverse transformation in a ..
S Kajiwara and T Kikuchi
Acta Metall. 30 (1982) 589-598
Reversible movement of the austenite martensite interface and ..
S Kajiwara and T Kikuchi
Phil Mag A 48 (1983) 509-526
Reversibility of crystal orientation in the reverse martensitic..
S Kajiwara
Phil Mag A 39 (1979) 325-339
Reversibility of shape deformation in the reverse martensitic...
S Kajiwara
Phil Mag A 41 (1980) 403-415
Formation of austenite from ferrite and ferrite-carbide aggregate
G R Speich and A Szirmae, appendix by G R Speich and M J Richards
Trans Met Soc AIME 245 (1969) 1063-1074
Formation of austenite during intercritical annealing ...
G R Speich, V A Demarest and R L Miller
Metall Trans A 12A (1981) 1419-1428
Use of the reverse martensitic transformation and precipitation..
S Jin, D Huang, J W Morris and G Thomas
New Aspects of martensitic transformation, Japan Inst Metals 1976
Formation of austenite from lath like martensite
S Watanabe, Y Ohmori and T Kunitake
New Aspects of martensitic transformation, Japan Inst Metals 1976
Isothermal martensite in some Fe-Ni base alloys with very low..
K Shibata, M Himeno, T Kato, T Fujita and T Araki
New Aspects of martensitic transformation, Japan Inst Metals 1976
Nearly perfect shape memory effect in Fe Ni C alloys
S Kajiwara
Trans J I M 26 (1985) 595-596
On the Burgers vector of dislocations produced by a cyclic...
K Marukawa and S Kajiwara
Phil Mag A 55 (1987) 85-97
Shape memory effect in high Ni steels
S Kajiwara, T Kikuchi and N Sakuma
Proc. Int. Conf. on Mart. Trans. ICOMAT 1986 Japan Inst. Metals..
Der Orientierungszusammenhang bei der Ruckumwandlung der ...
H Kessler and W Pitsch
Acta Metall. 13 (1965) 871-874
Phase Transformation Kinetics and Hardenability of Medium-Carbon
W T Cias
Climax Molybdenum Company, Ann Arbor, Michigan
Thermoelastic martensitic transformations and shape memory
C M Wayman
Phase transformations in solids, ed. T Tsakalos, Materials ...
Some aspects of martensitic transformations, 1979
C M Wayman
Phase Transformations, Institute of Metals, Vol.1 York meeting...
Influence of annealing in the ferrite and austenite phase ...
I E Locci and G M Michal
Metall Trans A 20A (1989) 237-245
The growth of austenite into ferrite in the Fe-N system
J D Grazier, H W paxton and W W Mullins
Trans Met Soc AIME 233 (1965) 130-142
On the formation of Widmanstatten austenite
M Hillert
Metall. Trans. A 17A (1986) 741-741.
Effects of aging on the microstructure of 17-4 PH stainless steel
U K Viswanathan, S Banerjee & R Krishnan
Materials Science and Engineering A104 (1988) 181-189
Explanation of a ferrite to austenite structural memory phenomeno
J Forchelet and W Form
Mem Etud Scien Rev Metall, pp.525-538 Oct 1982
Transformation in Armco Iron by Rapid Heating
N Iguchi and K Yokota
J Japan Institute of Metals 39 (1975) 19-23.
Microstructure and mechanical properties of duplex martensitic...
V G Gorbach, J Jelenkowski and J Filipiuk
Mat Sci and Tech 5 (1989) 36-39
Furukawa NT alloys
Furukawa electric
commercial brochure
Shape memory effect related to thin plate martensite with large..
T Maki, S Furutani and I Tamura
ISIJ International 29 (1989) 438-445
Shape memory effect in ferrous alloys
T Maki and I Tamura
ICOMAT 86, Japan Institute of Metals (1986) 963-970
alpha->gamma transformation behaviour during heating from the ...
K Tsuzaki, K Yamaguchi, T Maki and I Tamura
J Japan Inst of Metals 74 (1988) 1430-1437
Recrystallisation of reversed austenite and subsequent ...
T Maki, H Morimoto and I Tamura
Trans ISIJ 20 (1980) 700-706
Austenitisation during intercritical annealing of an Fe-C-Si-Mn..
J J Yi, I S Kim and H S Choi
Metall. Trans. A 16A (1985) 1237-1245
Microstructure of welded and weld simulated 3Cr-1.5Mo-0.1V...
J M Vitek and S A David
Metall. Trans. A 21A (1990) 2021-2036
Dissolution of cementite in an Fe-Cr-C alloy Pt II:
L Hoglund, B Jonsson, Z-Ki Liu & J Agren
TRITA-MAC-0442 Sept. 1990
Structure property relationships in reheated SA steel weld metals
A O Kluken and O Grong

Continuous heating transformation of bainite to austenite
J R Yang and H K D H Bhadeshia
Materials Science and Engineering A A131 (1991) 99-133
Nucleation of austenite during intercritical annealing of a...
S W Thompson, G S Farr and P R Howell
Phase trans. in ferrous alloys, eds A R Marder & J I Goldstein...
The stability of reverted austenite in 6% Ni steel
H Haga
Trans I S I J 13 (1973) 141-144
Increased fracture toughness in a 300 grade maraging steel ...
S D Antolovich, A Saxena and G R Chanani
Metall. Trans. 5 (1974) 623-632
The formation of austenite
H W Paxton
Transformations & Hardenability in Steels (1967) Climax Mo.....
Theoretical study of the rates of dissolution of carbide ....
R C Hudd and C Bodsworth
R. Thomas & Baldwins Ltd. Internal Report 432 Sept. 1964
Study of the mechanical properties of intercritically annealed..
C Bodsworth P Moore and T Flavell
R Thomas & Baldwins LTd. Internal Rep. 433, Sept 1964
Micorstructure and local brittle zone phenomena in high....
B C Kim, S Lee, N J Kim and D Y Lee
Metall. Trans. A 22A (1991) 139-149
Influence of high-C martensitic island on CTOD value of weld..
S Aihara and T Haze
Nippon Steel Corporation report, 25 January 1988
Metallurgical factors controlling HAZ toughness in HT50 steels
T Haze and S Aihara
IIW Doc IX-1423-86
Influence of toughness and size of local brittle zone on ...
T Haze and S Aihara
7th Int. Conf. Offshore Mechanics & Arctic Engng. Houston, Texas
Influence of local brittle zone on HAZ toughness of TMCP steels
S Aihara and K Okamoto
Texas 1990 Meeting, Microalloying International
Ultra-high-strength metallic fibre
Anonymous technical brochure
Kobelco technology review No.8, June 1990
Recyrstallisation & formation of austenite in deformed ...
M Tokizane, N Matsumura, K Tsuzaki, T Maki and I Tamura
Metall. Trans. A 13A (1982) 1379-1388
Secondary hardening ultrahigh-strength steels
G R Speich
Innovations in ultrahigh strength steel technology, ed Olson et a
Shape memory effect and related transformation behaviour in ...
S Kajiwara and T Kikuchi
Acta Metall Mater 38 (1990) 847-855
Microstructure and local brittle zone phenomena in high...
B C Kim, S Lee, N J Kim and D Y Lee
Metall. Trans. A 22A (1991) 139-149
Recent developments in Fe-based shape memory alloys
T Maki
Proc. 1st Japan Int. SAMPE Symp. (1989) 225-230
Effects of some alloying elements on the transformation of ...
R B G Yeo
Trans. Metall. Soc. AIME 227 (1963) 884-890
shape memory and related phenomena
C M Wayman
Progress in Materials Science 36 (1992) 203-224
Development of low activation martensitic steels
J C Brachet, A Castaing and C Foucher
Progress report, CEA/DECM/SRMA Saclay, France, June 1994
Machine induced phase transformation in a maraging steel
F Habiby, T N Siddiqui, H Hussain, M A Khan, A ul Haq, A Q Khan
Materials Science and Engineering A159 (1992) 261-266
Chemical heterogeneities as a strengthening factor in martensite
I Dlouhy
Scripta Metall. et Mater. 26 (1992) 1581-1586
Reaustenitisation in steel welds
J R Yang and H K D H Bhadeshia
Welding Metallurgy of Structural Steels, ed. J Y Koo, TMS-AIME,
The bainite to austenite transformations
J R Yang and H K D H Bhadeshia
Phase Transformations '87, Instiute of Metals, London, ed. ...
Reaustenitisation experiments on some high strength steel welds
J R Yang and H K D H Bhadeshia
Materials Science and Engineering A118 (1989) 155-170
Complete reaustenitisation in multirun steel weld deposits
R Reed and H K D H Bhadeshia
Recent Trends in Welding Science and Technology, eds. S A David..
Conitnuous heating transformations of bainite to austenite
J R Yang and H K D H Bhadeshia
Materials Science and Engineering A A131 (1991) 99-113
Effects of heat treatment cycle on equilibrium between ...
H Khaira, A K Jena and M C Chaturvedi
Materials Science and Engineering A161 (1993) 267-271
Carbide precipitation and the reversion of martensite to ...
H Smith and D R F West
Metals Technology 1 (1974) 295-299
Morphological stability of gamma/alpha interface formed ...
M Ichinose, F Togashi, K Ishida and T Nishizawa
Metallurgical and Materials Transactions A 25A (1994) 499-508
Prediction of tensile properties of intercritical HAZ
O M Akselsen, G Rorvik and A O Kluken
Materials Science and Technology 10 (1994) 75-80
(in Japanese)
K Abiko
Scientific American (Jap. edition) 23 (1993) 20-29
Mathematical model coupling phase transformations and ....
S Denis, D Farias and A Simon
ISIJ International 32 (1992) 316-325
The austenite transformation in ferritic ductile cast iron
J M Chou, M H Hon and J L Lee
Materials Science and Engineering A158 (1992) 241-249
Predicting case depth in tempered steels hardened via laser proc.
M Y Wei and C Chen
Materials Science and Technology 10 (1994) 69-73
Effect of post-weld heat treatment on heat affected zone...
D J Sparkes, N Bailey and T G Gooch
Materials Science and Technology 6 (1990) 1215-1226
Morphology of cementite decomposition in an Fe-Cr-C alloy
Z K Liu and J Agren
Metall. Trans. A 22A (1991) 1753-1759
Partial austenitization within flow zone when cutting a low ....
J L Hau-Bracamonte
Metals Technology 8 (1981) 447-450
Theory for reaustenitisation from ferrite/cementite mixtures in..
C Atkinson, T Akbay and R C Reed
Acta Metall. et Materialia 43 (1995) 2013-2031
Assessment of resistance of low-alloy steels to reheat ...
A G Glover, W K C Jones and A T Price
Metals Technology, June 1977, 326-332
Reversion of martensite to austenite in an Fe-16Cr-12Ni alloy
T H Coleman and D R F West
Metal Science 9 (1975) 342-345
In situ observation of the alpha-gamma transformation in high....
K Sadamori and K Abiko
Ultra high purity base metals, UHPM-94, Japan Institute for ...
Austenite formation in manganese partitioning dual-phase steel
E Navara, B Bengtsson and K E Easterling
Materials Science and Technology 2 (1986) 1196-1201
Austenitization kinetics of pearlite and ferrite aggregates...
D P Datta and A M Gokhale
Metall. Trans. A 12A (1981) 443-450
Kinetics of ferrite toa austenite transformation in a high strength low alloy steel containing Ti and V
S K Jayaswal and S Gupta
Z. Metallkunde 83 (1992) 809-819
Laser pulse heat treatment: application to reaustenitisation from ferrite/cementite mixtures
R C Reed, Z Shen, T Akbay and J M Robinson
Materials Science and Engineering A A232 (1997) 140-149
Texture of austenite at high temperature in SULC cold-rolled steel
N Yoshinaga, K Kawasaki, H Inoue, B C De Cooman and J Dilewijns
Modern LC and ULC Sheet Steels for Cold Forming: Processing and Properties, Aachen, Germany (1998) 569.
Determination of reaustenitisation kinetics in a Fe-0.4C steel using dilatometry and neutron diffraction
R Reed, T Akbay, Z Shen, J. M. Robinson and J H Root
Materials Science and Engineering A A256 (1998) 152-165
The effects of cooling rate after austenitising on creep rupture strength of low C-2.25Cr1.6W-V-Nb steel
N Komai and F Masuyama
Translated from CAMP-ISIJ 11 (1998) 432-435
Phase analysis of surface layers irradiated with high--energy electron beam using Mossbauer spectroscopy
S J Choi, S J Kwon, S H Choo and S Lee
Materials Science and Engineerin A A265 (1999) 206-216
Prediction of as-quenched hardness after rapid austenitisation and cooling of surface hardened steels
T Reti, G Bagyinszki, I Felde, B Vero and T Bell
Computational Materials Science 15 (1999) 101-112

I want to distinguish between martensite and bainite in the optical microscopy. I have known that this is difficult. Could you recommend etchant by which bainite is distinguished from martensite in the mixture? And could you recommend another good methods to me?

In general, the best way is to use transmission electron microscopy because the scale of the microstructure is much smaller than can be resolved using optical microscopy.

However, if you are determined to only use optical microscopy then etch the sample lightly using nital. The martensite, assuming that it has not been tempered, will appear light. The bainite will appear dark. This is because the bainite contains fine precipitates of cementite, which provide lots of interfaces where an etchant can attack.

In the following actual images, the optical micrograph is marked "a". The dark-etching regions are bainite whereas the light background is martensite. Higher resolution transmission electron micrographs are also provided in the same image.

Of course, the hardness of bainite is also less than that of untempered martensite, so you could also use that as a distinguishing feature.

I have never heard about boundary diffusion of carbon as a control growth mechanism for pearlite formation. Have you heard about those studies?.

The most recent paper I have on this subject is by K. Hashiguchi and J. S. Kirkaldy, Scandinavian Journal of Metallurgy, Vol. 13 (1984) 240-248.

They conclude that the carbon flux contribution by boundary diffusion is about 2-2.6 times greater than by volume diffusion when transformation is at "intermediate temperatures".

What is the deformation mechanism of metastable austenite at temperatures 500-700 C?

I am not completely sure of the context of this question but ordinary plastic deformation is by a dislocation mechanism. However, it the austenite is metastable and the temperature is sufficiently low then it is possible that phase changes can contribute via transformation plasticity. You may find useful references if you search on "ausforming" in which austenite is deformed prior to transformation.

I'm studying the isothermal transformation kinetics of steel using dilatometer system. I want to draw the TTT diagram. However, I have no idea. How I can determine the transformation-start time? I'm analysing the data using Origin. But because of scatters of data, I cannot determine the start time exactly. Could you help me?

I assume that you have, for a given temperature, a curve in which you plot the dilatation against the isothermal transformation time. There is a certain amount of error associated with your experiments. Make an assessment of the error. Then say that the transformation starts when the dilatation exceeds the experimental error in your dilatation system.

Referring to your book "Bainite in steel", p.27, section l.19, you say that  "It is a well known effect that the size decreases with carbon concentration."  Does "the size" indicate sheaves, lath, packet or something else? 

The size refers to the grain size of bainite (see first line of that paragraph). Bainite has a thin plate shape so the size is given by a mean lineal intercept, which is approximately equal to twice the plate thickness (see last paragraph on page 26.)

There is a recent paper on weld metal microstructures where the authors avoid using the term bainite since "considerable confusion
surrounds the definition of the term bainite". Furthermore, they claim that in these ultra-low-carbon weld metals there is not enough C to form carbides? "Thus, this microstructure does not satisfy the microstructural definition of bainite, which defines it as a nonlamellar eutectoid decomposition product. This microstructure is identified as "lath ferrite" in...."

Isn't bainite a more appropriate term than lath ferrite?

Much of the so-called "confusion" is exaggerated; there is now more known about bainite than any other transformation in steels or indeed in any metal.

The "microstructural definition" of bainite is ill-conceived; it falsely implies that anything without carbides is not bainite. It is well known that bainite forms in two steps, first with the growth of supersaturated ferrite and then the precipitation of carbides. These steps can be separated by controlling the chemical composition and heat treatment.

The idea that bainite is a non-lamellar eutectoid has no justification. First, the growth of bainite is displacive with a large shear strain. This is in contrast to the reconstructive transformation to pearlite where there is only a volume change. Secondly, there is no partitioning of substitutional solutes with bainite, even on an atomic scale. Pearlite growth always occurs with the partitioning of all solutes. Nothing about the transformation can be predicted by assuming that bainite is a non-lamellar eutectoid.

I am studying bainite transformation in pressure vessel steel, SA508-cl.3 (0.2C-1.34Mn-0.86Ni-0.49Mo wt% Steel) using dilatometer system. I observed that total length change increases as isothermal temperature decreases. I designed thermal cycle as following,
slow cooling (10 C/s) after isothermal holding because I wanted to observe another transformation after isothermal holding. I could not observe any length change during slow cooling below 480 C. However, total length change increases below this temperature. I could not understand this inconsistent phenomenon between two cases. Could you help me? 

In your experiment, you first partly transformed the austenite into bainite. In the second stage, on continuous cooling below 480 C, you expected therefore to see more transformation.

You say you did not observe the length change due to transformation in the second stage.

The figure below shows a typical dilatometer curve during continuous cooling from the austenite phase field. I want you to focus attention on the dashed line marked ab. If the expansion coefficient that you determine from line ab is greater than that for a fully ferritic microstructure, then there is transformation happening along the dilatometer curve parallel to line ab. The rate of transformation is not large enough to give you an overt blip in the curve, but is reflected in the slope of the curve.  You can measure the expansion coefficient of ferrite by tempering the sample at 600 C and then cooling it.

I am trying to model the transformation of austenite, into a mixture of bainitic ferrite and cementite during isothermal heat-treatment. I am assuming that the bainite at first forms without diffusion but the carbon then partitions into the residual austenite, eventually to precipitate as cementite. I am interested particularly in multicomponent steels.

How can I deal with changes in the susbtitutional solute concentration when the cementite precipitates from austenite?

How can I deal with the changes of each susbtitutional solute in both bainite and austenite?

The model you assume is, I think, the correct model for the bainite transformation.

It is well-established that during the bainite transformation, cementite precipitates from austenite by a paraequilibrium mechanism. See for example, Materials Science and Engineering A, Vol. A273-275, 1999, 58-66 for a recent discussion of this topic. Therefore, you do not need to consider the partitioning of substitutional solutes during cementite precipitation - you simply need to calculate the driving force for precipitation assuming paraequilibrium rather than equilibrium.

Your second question also must recognise that the formation of bainitic ferrite does not involve the diffusion of substitutional atoms, even on the finest conceivable scale. It should therefore be possible to model the entire transformation assuming paraequilibrium.

Fig. 12.2 in your book (Bainite in steels): why does hardness increase after an initial decrease, for the case of 0.26Cwt%? What is the difference of hardness between upper and lower bainite. What morphology is harder?

Third, it has been known that CCT diagram can present the effect of prior austenite grain size on the transformation start temperature and that CCT diagram must include experimental variables such as austenitising temperature, grain size, etc. However, for high carbon equivalent steels(that is, transformation rate for high temperature phase such as polygonal ferrite and pearlite is sluggish), grain growth can occur for the slow cooling rate conditions of CCT experiment before bainite or martensite transformation starts. If this statement is right, the grain size which included in the  CCT diagram is not right? Coud you tell me the method to do continuous cooling transformation tests and avoid the grain growth phenomenon during cooling?  

The figure you ask about is from Lyman and Troiano:

The explanation, from Lyman and Troiano's paper, is as follows: as the temperature is raised above 440 C, the amount of bainite obtained during isothermal transformation decreases so the amount of martensite in the final microstructure increases. This gives the increasing hardness as the temperature rises above 440 C.

When the transformation temperature is reduced below 440 C, bainite is the predominant phase (there is almost no martensite). Therefore, the hardness is mainly a reflection of the fact that the strength of bainite increases as the transformation temperature is reduced.

Bainite produced at lower temperatures is harder. Hence lower bainite is harder. The length scale of the microstructure decreases as the transformation temperature is reduced.

There should not be any significant change in the austenite grain size during cooling from the austenitising temperature. This is because the effect of temperature decreases exponentially as the temperature is reduced. I do not know of any evidence which suggests changes in the austenite grain size during cooling, as long as the austenitising temperature is high enough and the austenitisation time is sufficient. You can test this by quenching the sample from the austenitisation temperature and comparing against one which has been cooled slowly.