See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/222403174
Formation of nanostmctured steels by phase
transformation
ARTICLE in SCRIPTA MATERIALIA · OCTOBER 2004
Impact Factor: 3.22 · DOI: 10.1016/j.scriptamat.2004.06.006
CITATIONS
READS
33
29
3 AUTHORS, INCLUDING:
Carlos Garcia-Mateo
Spanish National Research Council
140 PUBLICATIONS 2,059 CITATIONS
SEE PROFILE
Available from: Carlos Garcia-Mateo
Retrieved on: 10 February 2016
Formation of Nanostructured Steels
by Phase Transformation
T. Yokota, C. G. Mateo and H. K. D. H. Bhadeshia
University of Cambridge
Department of Materials Science and Metallurgy
Pembroke Street, Cambridge CB2 3QZ, U.K. www.msm.cam.ac.uk
Introduction
The purpose of this paper is to develop thermodynamic criteria which define the finest grain
structures that can be achieved in steels, and to see how far current technology is from achieving
the minimum possible grain size.
In a pure material, a grain boundary is a region where the adjacent crystals do not match
perfectly and hence is associated with an excess free energy per unit area (σ). In alloys, there
may also be a difference in chemical composition of the abutting crystals and that of the
boundary may in turn differ from that of the bulk crystals. These differences contribute a
chemical component to σ.
For an equiaxed polycrystalline grain structure, the grain boundary surface per unit volume
(SV ) is related to the grain size L (mean lineal intercept) by the equation SV = 2/L. The
stored energy per unit volume due to the grain boundaries is then
∆GV = σSV = 2σ/L
(1)
Grain size refinement is an important method for improving both the strength and toughness
of engineering alloys. In the context of steels, the stable phase at ambient temperature is
generally ferrite (α), which is generated by transformation from austenite (γ). If ∆Gγα
V is the
free energy change per unit volume when austenite transforms into ferrite, then
γ
α
|∆Gγα
V | ≥ σSV − σSV
(2)
which for equiaxed austenite–grains becomes
|∆Gγα
V |≥
2σγ
2σα
−
Lα
Lγ
(3)
It follows that the smallest ferrite grain size that can be achieved is when all of Gγα
V is used
up in creating α/α grain boundaries, so that
min
Lα
=
2σα
+ 2σγ /Lγ
|∆Gγα
V |
(4)
The term 2σγ /Lγ arises when the formation of ferrite destroys the austenite grain boundaries,
thus providing a thermodynamic gain which adds to the driving force for transformation.
There are a number of interesting insights to be gained from this equation. A reduction in
the austenite grain size should always lead to a corresponding reduction in the ferrite grain
size; however, the magnitude of the change depends also on |∆Gγα
V |, i.e. on the undercooling
at which the γ → α transformation occurs. The austenite grain size becomes less important
at large undercoolings. Equation 4 is not correct when dealing with anisotropic austenite
grains because of their greater surface to volume ratios [1,2,3]; the relationship between SV
and L is then different from the one assumed above [4]. Mechanical deformation prior to phase
min
transformation therefore enables smaller values of Lα .
It has been assumed here that σγ is constant, but crystallographic texture may lead to a
reduction in stored energy if low Σ orientations are introduced by thermomechanical processing.
Random, high–energy austenite grain boundaries are ideal for the achievement of the smallest
ferrite grains.
As shown below, these elementary equations can be used to study the grain sizes achieved in
industrial steels.
Allotriomorphic Ferrite
Most engineering steels have a microstructure dominated by allotriomorphic ferrite; the steels
are produced under conditions of continuous cooling transformation. A large variety of data
exist in the published literature [5–25], for steels in roughly composition categories: Fe–0.15C–
0.2Si–0.7Mn and Fe–0.15C–0.2Si–1.5Mn wt%. However, the published data are in general
incomplete. Thus, the transformation temperature Ar3 was in each case calculated empirically
[24,25]:
Ar3 ≃ 910 − 80xMn − 20xCu − 15xCr − 55xNi − 80xMo − 40 log{Ṫ }
(5)
where x is the concentration in wt% and Ṫ the cooling rate in K s−1 . The equation is based
on steels which have been deformed by rolling to greater than 50% reduction in a temperature
regime where the austenite does not recrystallise, with cooling rates in the range 1–10 K s−1 .
The free energy change ∆GV at the calculated transformation temperature Ar3 was estimated
using MTDATA in combination with the SGTE database [26], for the reaction γ → γ ′ + α,
assuming paraequilibrium. The symbol γ ′ stands for carbon–enriched austenite.
min
Fig. 1 shows the variation in the ferrite grain size (Lα ) as a function of ∆GV , calculated using
equation 3 with σα = 0.6 J m2 ; given the absence of data the 2σγ /Lγ term in equation 3 was set
to zero. These calculations are presented as the ‘ideal’ curve in Fig. 1. The curve indicates that
min
at large grain sizes, Lα is very sensitive to ∆GV , i.e. the undercooling below the equilibrium
transformation temperature. However, reductions in grain size in the submicrometer range
require huge values of ∆GV , meaning that the transformations would have to be suppressed
to large undercoolings to achieve fine grain size.
Also plotted on Fig. 1 are points corresponding to measured ferrite grain sizes from the low
and high–Mn steels as described earlier; the ∆GV values are calculated for the appropriate
min
transformation temperatures. It is evident that except at the lowest undercoolings, Lα ≫ Lα .
The experimental data all represent continuously cooled steels; during transformation, the
enthalpy change leads to recalescence. The maximum temperature rise due to recalescence can
be estimated using:
∆T max ≃ ∆H/Cγ
(6)
where ∆H is the change in enthalpy associated with the phase transformation (calculated
using MTDATA for Ar3 ) and Cγ the appropriate heat capacity. The value of ∆GV was then
recalculated for the temperature Ar3 +∆T , assuming that all the enthalpy change is used up in
instantaneously heating the sample. The curves marked low and high–Mn on Fig. 1 represent
these recalescence corrected curves corresponding to the ideal case where there an enthalpy
change is dissipated isothermally at Ar3 . It is seen that the recalescence corrected curves
show better agreement with the experimental data, indicating that at large undercoolings,
Fig. 1: Plot of the logarithm of ferrite grain size versus the free energy change
min
at Ar3 . The ideal curve represents the values of Lα . The points are experimental data. The curves marked low and high–Mn represent calculated values
min
of Lα
after allowing for recalescence.
the achievement of fine grain size is limited by the need to dissipate enthalpy during rapid
transformation.
There are therefore two significant conclusions from this work. The first is that from the
form of equation 3, it is necessary to achieve very large undercoolings before submicrometer
grain sizes can be achieved. The difficulty when these large undercoolings are attained is that
the enthalpy of transformation is likely to lead to recalescence, so that the actual grain size
obtained is more than an order of magnitude greater than can be obtained in principle. It is
necessary to design some industrial process which enables isothermal transformation at large
undercoolings.
Bainite and Martensite
Strain energy dominates displacive transformations and this is reflected in their characteristically fine and plate–like microstructures. In what follows, the microstructure is discussed
primarily in terms of the plate thickness, since for a plate–shape, the mean lineal intercept is
twice the thickness, almost independent of the plate length.
The elastic strain energy per unit volume (GeV ) of a transformed region with the same shear
modulus (µ) and Poisson’s ratio (ν) as the isotropic constraining matrix is given by [27]:
due to volume change
z
}|
{
1−ν e
πt
πt
π
t
2
GV = (1 + ν)∆2 + δ 2 + (1 + ν)∆δ + (2 − ν) s2
µ
9
4l
3l
|8
{z l }
due to shear
or
t
GeV ≃ µ(s2 + δ 2 )
l
(7)
where ∆ is the uniform dilatation strain which is now known to be zero and the other strains s
and δ are illustrated in Fig. 1. t and l are the thickness and diameter respectively of the oblate
spheroid shape used to represent the martensite plate. For typical values of the parameters in
equation 7, with t/l = 0.05, GeV ≃ 600 J mol−1 , which is large, and the reason why the plates
tend to be thin.
Assuming that a martensite plate can grow unhindered across an austenite grain of size Lγ ,
the mean lineal intercept. The maximum thickness of the plate is obtained by balancing the
chemical free energy change accompanying transformation ∆Gγα against GeV , i.e.
t≃
|∆Gγα | × L
µ(s2 + δ 2 )
(8)
The plate thickness defines the mean free slip distance which is about 2t [4] and hence is the
effective grain size. With these assumptions, the size of elastically accommodated martensite
plates is related directly to the austenite grain size and the driving force for transformation.
The mean size of the plates within a single austenite grain will be somewhat smaller than predicted because of the way in which the plates geometrically partition the parent phase. Notice
how the shape deformation plays a vital role in determining the scale of the microstructure,
with the shear strain dominating the equation.
Plastically Accommodated Plates
A displacive transformation is accomplished by the translation of a glissile interface, whose
motion can be halted by defects in the austenite. Dislocations are created when the shape
deformation is accommodated by plastic relaxation of the surrounding austenite [28,29]. The
growing plate then brings itself to a halt before it collides with a hard obstacle such as an
austenite grain surface. This is vividly illustrated by the sub–unit mechanism of bainite [30,31],
in which an individual sheaf consists of a myriad of much smaller plates, each of which grows
to a limited size which can be much smaller than that of the austenite grain.
Equation 8 indicates that during unhindered growth, thermoelastic equilibrium requires that
the plates are coarser at low temperatures where |∆Gγα | is largest. This trend contradicts
observations because plastic relaxation prevents the bainite from achieving thermoelastic equilibrium. Since austenite is weaker and dynamic recovery more likely at high temperatures,
the plates in fact become thicker at elevated transformation temperatures. It is found that an
increase in the strength of austenite by 100 MPa leads to approximately a 0.2 µ m decrease in
the thickness of bainite plates [32,33]; the intrinsic effects of ∆Gγα and temperature are found
to be much smaller in comparison (Fig. 2). By obtaining bainite in a strong austenitic steel at
a large driving force, it has been possible to obtain plates which are between 20–40 nm thick,
resulting is steels which are 2.5 GPa strong whilst at the same time exhibiting respectable
toughness [34]. The reason why the thickness tends to level out beyond a strength of about
120 MPa in Fig. 2b is not understood, but investigations show that it is not because the plates
become elastically accommodated.
Recrystallised Plates
When equiaxed grain structures are required, they can be generated by deformation followed
by recrystallisation but in the context of current trends, this does not lead to a sufficiently fine
grain structure. The extent of deformation needed also limits the form of the final product.
Tsuji and co–workers [37,38] have developed a process which avoids these difficulties by starting with a martensitic microstructure, which already has a high stored energy. Martensite can
therefore be recrystallised without any additional deformation. However, Tsuji et al. discovered that a small amount of plastic strain (≃ 50%) applied to a martensitic microstructure
Thickness / nm
0
150
200
250
300
350
Temperature / °C
Fig. 2: (a) Stereologically corrected thickness of bainite plates as a function
of the transformation temperature [35]. (b) An example of the variation in
thickness as a function of austenite strength. The calculations represent the
latest analysis using the method described in [36Singh].
introduces three kinds of heterogeneities in the martensite, which have the effect of subdividing
it. These are fine dislocation cells, irregularly bent laths and kinked laths and in each case the
defect results in large local–misorientation, thereby providing ideal sites for the generation of
new grains during annealing.
Defining grain size in terms of regions with a critical level of misorientation, the original
martensite was judged to have a grain size of 3.2 µ m. Its subdivision by heterogeneous deformation, led by the same definition, to an equiaxed grain size of about 180 nm after annealing,
a reduction in scale by a remarkable factor of 18.
SUMMARY
The challenge in nanostructured steels is to obtain very fine grain size, high strength and
toughness, in bulk samples and at a cost which is reasonable. It appears, however, that recalescence is likely to limit the size of the allotriomorphic ferrite grains that can be obtained
during continuous cooling transformation. On the other hand, displacive transformations lead
to fine microstructures as a natural consequence of their atomic mechanisms. Commercially viable nanostructured–steels based on bainite and martensite are already at an advanced stage of
exploitation. Nevertheless, a fundamental understanding of the role of plastic accommodation
of the shape deformation, in limiting the scale of the microstructure, is lacking.
REFERENCES
1. I. Czinege and T. Reti: Eighteenth International Machine Tool Design and Research Conference, Forming I (1977) 159–163.
2. Y. V. Leeuwen, S. Vooijs, J. Sietsma and S. van der Zwaag: Metall. and Mater. Trans .
(29A) 1998
2925–2931.
3. S. B. Singh and H. K. D. H. Bhadeshia: Mat. Sci. and Technology 14 (1998) 832–834.
4. C. Mack: Proc. Cambridge Philosophical Society 52 (1956) 246.
5. I. Kozasu, T. Shimizu and K. Tsukada: Trans. ISIJ 12 (1972) 305.
6. I. Kozasu, C. Ouchi. T. Sampei and T. Okita: Microalloying 75, ed. by M. Korchinsky
(1975) 120.
7. C. Oouchi, T. Ookita and S. Yamamoto: Tetsu to Hagane 67 (1981) 969.
8. C. Oouchi, T. Ookita and S. Yamoto: Trans. ISIJ 22 (1982) 668.
9. K. Amano. T. Hatomura, M. Koda, C. Shiga and T. Tanaka: Accelerated Cooling of Steel,
ed. P. Southwick, Pittsburgh ( 1985) 349.
10. L. J. Cuddy: Accelerated Cooling of Steel, ed. P. Southwick, Pittsburgh (1985) 235.
11. N. Nagao and K. Kunishige: Accelerated Cooling of Steel, ed. P. Southwick, Pittsburgh
(1985) 463.
12. M. lmagumbai and M. Nagumo: Accelerated Cooling of Steel, ed. By P. D. Southwick,
Pittsburgh (1985) 297.
13. T. Abe, K. Tsukada and I. Kozasu: Tetsu to Hagane 74 ( 1988) 505.
14. P. D. Hodgson, M. R. Hickson and R. K. Gibbs: Mat. Sci. Forum 284–286 (1988) 63.
15. A. Takahashi and M. Lino: ISIJ Int. 36 (1996) 235.
16. M. Fujioka, Y. Abe and Y. Hagiwara: Proc. of the lst symp. on Super Metal, JRCM,
Tokyo (1998) 205.
17. Y. Adachi, T. Tomida and S. Hinotani: Proc. of the lst symp. on Super Metal, JRCM,
Tokyo (1998) 215.
18. S. Toozuka, O. Umezawa, K. Tsuzaki and K. Nagai: Mater. Sci. Forum 284–286 (1998)
225.
19. S. Torizuka, O. Umezawa et. al. : Solid–Solid Phase Transformation ‘99 eds M. Koiwa, K.
Otsuka and T. Miyazaki, JIM (1999) 1533.
20. P. D. Hodgson, M. R. Hickson and R. K. Gibbs: Scr. Mater. 40 (1999) 1179.
21. M. R. Hickson,R. K. Gibbs and P. D. Hodgson: ISIJ Int. 39 (1999) 1176.
22. W. Y. Choo: ISUGS ( 2001) 2.
23. S. Torizuka, K. Nagai, S. Genda and Y. Kogo: ISUGS (2001) 218.
24. C. Ouchi, T . Sampei and I. Kozasu: Trans. lSIJ 22 ( 1982) 214.
25. T. Abe, K. Tsukada and I. Kozasu: Tetsu to Hagane 74 ( 1988) 101.
26. MTDATA: National Physical Laboratory, Teddington, U. K. (2003)
27. J. W. Christian: Acta Metallurgica 6 (1958) 377–379.
28. H.K.D.H. Bhadeshia and D.V. Edmonds: Metall. Trans. A 10A (1979) 895–907.
29. E. Swallow and H. K. D. H. Bhadeshia: Mat. Sci. and Techn. 12 (1996) 121–125.
30. R. F. Hehemann: Phase Transformations, ASM, Metals Park, Ohio, USA (1970) 397–432.
31. H. K. D. H. Bhadeshia: Bainite in Steels, 2nd edition, Institute of Materials, London
(2001) 1–458.
32. L.–C. Chang and H. K. D. H. Bhadeshia: Materials Science and Technology 11 (1995)
874–881.
33. S. B. Singh and H. K. D. H. Bhadeshia: Materials Science and Engineering A A245 (1998)
72–79.
34. F. G. Caballero, H. K. D. H. Bhadeshia, K. J. A. Mawella, D. G. Jones and P. Brown:
Materials Science and Technology 18 (2002) 279–284.
35. C. G. Mateo, F. G. Caballero and H. K. D. H. Bhadeshia: ISIJ International 43 (2003) in
press.
36. N. Tsuji, R. Ueji, Y. Minamino and Y. Saito: Scripta Materialia 46 (2002) 305–310.
37. R. Ueji, N. Tsuji, Y. Minamino and Y. Koizumi: Acta Materialia 50 (2002) 4177–4189.