Austenite–martensite transformation
7.3.3 Austenite–martensite transformation
7.3.3.1 Crystallography of martensite formation
Martensite, the hardening constituent in quenched steels, is formed at temperatures below about 200 ◦
C. The regions of the austenite which have transformed to martensite are lenticular in shape and may easily be recognized by etching, or from the distortion they produce on the polished surface of the alloy. These relief effects, shown schematically in Figure 7.21, indicate that the martensite needles have been formed not with the aid of atomic diffusion but by a shear process, since if atomic mobility were allowed the large strain energy associated with the transformed volume would then be largely avoided. The lenticular shape of a martensite needle is a direct consequence of the stresses produced in the surrounding matrix by the shear mechanism of the transformation and is exactly analogous to the similar effect found in mechanical twinning. The strain energy associated with martensite is tolerated because the growth of such sheared regions does not depend on diffusion, and since the regions are coherent with the matrix they are able to spread at great speed through the crystal. The large free energy change associated with the rapid formation of the new phase outweighs the strain energy, so that there is a net lowering of free energy.
⫺ 1 0 ⫹ 1 (a)
A B (b)
Figure 7.21 Schematic diagram of the observed shape deformation produced by a martensite plate. (a) Contour lines on an originally flat surface. (b) Section of the surface through AB. Vertical
scale much exaggerated (after Bilby and Christian, 1956; courtesy of the Institute of Materials, Minerals and Mining).
Mechanical properties II – Strengthening and toughening 415 Direct TEM observations of martensite plates have shown that there are two main types of marten-
site, one with a twinned structure (see Figure 7.22), known as acicular martensite, and the other with
a high density of dislocations but few or no twins, called massive martensite. In contrast to the pearlite transformation, which involves both a redistribution of carbon atoms and
a structural change, the martensite transformation involves only a change in crystal structure. The structure cell of martensite is body-centered tetragonal, which is a distorted form of a body-
centered cubic structure, and hence may be regarded as a supersaturated solution of carbon in α-iron. X-ray examination shows that while the c/a ratio of the bct structure of martensite increases with increasing carbon content, the curve of c/a ratio against composition extrapolates back to c/a =1 for zero carbon content, and the lattice parameter is equal to that of pure α-iron (Figure 7.23).
From the crystallographic point of view, the most important experimental data in any martensite transformation are the orientation relations of the two phases and the habit plane. In steel, there are three groups of orientations often quoted; those due to Kurdjumov and Sachs, Nishiyama, and Greninger and Troiano, respectively. According to the Kurdjumov–Sachs relation, in iron–carbon alloys with 0.5–1.4% carbon, a {1 1 1} γ plane of the austenite lattice is parallel to the {1 1 0} α plane of the martensite, with a
α axis of the latter; the associated habit plane is {2 2 5} γ . In any one crystal there are 24 possible variants of the Kurdjumov– Sachs relationship, consisting of 12 twin pairs, both orientations of a pair having the same habit plane. However, for general discussion it is usual to choose one relation which may be written
γ axis of the former parallel to a
(1 1 1) γ α with [1 ¯1 0] γ
In the composition range 1.5–1.8% carbon, the habit plane changes to ≈ {2 5 9} γ with the orientation relationship unspecified. This latter type of habit plane has also been reported by Nishiyama for iron–nickel alloys (27–34% nickel), for which the orientation relationship is of the form
(1 1 1) γ α with [1 ¯2 1] γ
Outline of block of austenite which has transformed to martensite
Slip direction Second
(2 2 5) g habit
Surface of
austenite
Traces of [1 1 2] M
crystal
Austenite Habit planes (a)
(b)
Figure 7.22 (a) Formation of a martensite platelet in a crystal of austenite. (b) The inhomogeneous twinning shear within the platelet (after Kelly and Nutting, 1960; courtesy of the Royal Society).
416 Physical Metallurgy and Advanced Materials
0 2 4 6 8 At. % carbon
Figure 7.23 Variation of c and a parameters with carbon content in martensite (after Kurdjumov, 1948).
However, Greninger and Troiano have shown by precision orientation determinations that irrational relationships are very probable, and that in a ternary iron–nickel–carbon alloy (0.8% carbon, 22% nickel), (1 1 1) γ is approximately 1 ◦ from (1 0 1) α with [1 ¯2 1] γ approximately 2 ◦ from [1 0 1] α , and is associated with a habit plane about 5 ◦ from (2 5 9).
7.3.3.2 Mechanism of martensite formation
The martensite transformation is diffusionless, and therefore martensite forms without any inter- change in the position of neighboring atoms. Accordingly, the observed orientation relationships are
a direct consequence of the atom movements that occur during the transformation. The first sugges- tion of a possible transformation mechanism was made by Bain in 1934. He suggested that, since √ austenite may be regarded as a body-centered tetragonal structure of axial ratio
2, the transforma- tion merely involves a compression of the c-axis of the austenite unit cell and expansion of the α-axis. The interstitially dissolved carbon atoms prevent the axial ratio from going completely to unity and,
depending on composition, the c/a ratio will be between 1.08 and 1.0. Clearly, such a mechanism can only give rise to three martensite orientations whereas, in practice, 24 result. To account for this, Kurdjumov and Sachs proposed that the transformation takes place not by one shear process but by
a sequence of two shears (Figure 7.24), first along the elements (1 1 1) γ γ , and then a minor shear along the elements (1 1 2) α
α ; these elements are the twinning elements of the fcc and bcc lattice respectively. This mechanism predicts the correct orientation relations, but not the correct habit characteristics or relief effects. Accordingly, Greninger and Troiano in 1941 proposed a different two-stage transformation, consisting of an initial shear on the irrational habit plane which produces the relief effects, together with a second shear along the twinning elements of the martensite lattice. If slight adjustments in spacing are then allowed, the mechanism can account for the relief effects, habit plane, the orientation relationship and the change of structure.
Further additions to these theories have been made in an effort to produce the ideal general theory of the crystallography of martensite transformation. Bowles, for example, replaces the first shear of the Greninger–Troiano mechanism by the general type of homogeneous deformation in which
Mechanical properties II – Strengthening and toughening 417
Figure 7.24 Shear mechanisms of Kurdjumov and Sachs. (a) Face-centered austenite with {1 1 1} γ in horizontal plane. (b) Body-centered tetragonal martensite (α). (c) Cubic ferrite (α) (after Bowles and Barrett, 1952).
the habit plane remains invariant, i.e. all directions in this plane are unrotated and unchanged in length. However, in all such cases the problem resolves itself into one of determining whether a homogeneous strain can transform the γ-lattice into the α-lattice, while preserving coherency at the boundary between them. The homogeneous strain does not do this, so that some reasonable additional type of strain has to be added.
This shear can occur either by twinning or by slip, the mode prevailing depending on the composition and cooling rate. Between carbon contents of 0.2% and 0.5% the martensite changes from dislocated martensite arranged in thin lathes or needles to twinned acicular martensite arranged in plates. In the martensite formed at low C contents (e.g. Fe–Ni alloys) the thin lathes lie parallel to each other, with a {1 1 1} γ habit, to form pockets of massive martensite with jagged boundaries due to the impingement of other nearby pockets of lathes. The inhomogeneous shear produced by deformation twinning occurs on {1 1 2} planes in the martensite, so that each martensite plate is made up of parallel twin plates of thickness 2–50 nm. By operation of such a complex transformation mode with
a high index habit plan the system maintains an invariant interfacial plane. Because of the shears involved and the speed of the transformation it is attractive to consider that
dislocations play an important role in martensite formation. Some insight into the basic dislocation mechanisms has been obtained by in situ observations during either cooling below M s or by straining, but unfortunately only for Ni–Cr austenitic steels with low stacking-fault energy (i.e. γ ≈ 20 mJ m −2 ). For these alloys it has been found that stacking faults are formed either by emitting partial dislocations with b
10 oped, the corresponding diffraction pattern revealed cph ε-martensite. On subsequent deformation
418 Physical Metallurgy and Advanced Materials
(b) Figure 7.25 Electron micrographs showing: (a) contrast from overlapping faults on (1 1 1); A is
(a)
extrinsic and B is intrinsic in nature; (b) residual contrast arising from a supplementary displacement across the faults which is intrinsic in nature for both faults A and B (after Brooks, Loretto and Smallman, 1979).
or cooling, regions of ε-martensite transform rapidly into bcc α-martensite and, indeed, the only way in which α-martensite was observed to form was from an ε nucleus.
Because straining or cooling can be interrupted during the in situ experiments it was possible to carry out a detailed analysis of the defect structure formed prior to a region becoming recognizably (from diffraction patterns) martensitic. In this way it has been shown that the interplanar spacing across the individual stacking faults in the austenite decreased to the (0 0 0 1) spacing appropriate to ε-martensite. Figure 7.25 shows micrographs which reveal this change of spacing; no contrast is expected in Figure 7.25b if the faulted {1 1 1} planes remained at the fcc spacing, since the condition of invisibility g ·R = n is obeyed. The residual contrast observed arises from the supplementary displace-
faults and ≈ 2% of the {1 1 1} spacing. The formation of regions of α from ε could also be followed, although in this case the speed of the transformation precluded detailed analysis. Figure 7.26 shows
a micrograph taken after the formation of α-martensite and this, together with continuous observa- tions, show that the martensite/matrix interface changes from {1 1 1} to the well-known {2 2 5} as it propagates. Clearly, one of the important roles that the formation of ε-martensite plays in acting as a precursor for the formation of α-martensite is in the generation of close-packed planes with ABAB stacking, so that atomic shuffles can subsequently transform these planes to {1 1 0} bcc which are, of course, stacked ABAB (see Figure 7.27). The α-martensite forms in dislocation pile-ups where the a/6 rial increases as more dislocations join the pile-up, until the nucleus formed by this process reaches a critical size and rapid growth takes place. The martensite initially grows perpendicular to, and princi- pally on, one side of the {1 1 1} γ slip plane associated with the nucleus, very likely corresponding to the side of the dislocations with missing half-planes, since α-martensite is less dense than austenite.
7.3.3.3 Kinetics of martensite formation
One of the most distinctive features of the martensite transformation is that in most systems martensite is formed only when the specimen is cooling, and that the rate of martensite formation is negligible if
Mechanical properties II – Strengthening and toughening 419
200 0.5m
Figure 7.26 Electron micrograph showing an α-martensite plate, the austenite–martensite interface and the faults in the austenite matrix (after Brooks, Loretto and Smallman, 1979).
B A ⬍ 112⬎
bcc 110 plane
fcc 111 plane
Figure 7.27
A shear of a/6
half this shear the structure has pseudo-bcc packing (after Brooks, Loretto and Smallman, 1979).
cooling is stopped. For this reason, the reaction is often referred to as an athermal 1 transformation, and the percentage of austenite transformed to martensite is indicated on the TTT curve by a series of hor- izontal lines. The transformation begins at a temperature M s , which is not dependent on cooling rate, but is dependent on prior thermal and mechanical history, and on composition. For example, it is well established that the M s temperature decreases approximately linearly with increasing concentration of solutes such as carbon, nickel or manganese.
Speed of formation
The observation that martensite plates form rapidly and at a rate which is temperature independent shows that thermal activation is not required for the growth process. Electronic methods show that the
martensite needles form, in iron–nickel–carbon alloys, for example, in about 10– 7 s and, moreover, that the linear growth velocity is about 10 3 m s– 1 even at very low temperatures. Such observations show that the activation energy for the growth of a martensite plate is virtually zero, and that the velocity of growth approaches the speed of sound in the matrix. Sometimes a ‘burst phenomenon’ is exhibited, as, for example, in iron–nickel alloys, when the stresses produced by one martensite plate assist in the nucleation of others. The whole process is autocatalytic and about 25% of the transformation can occur in the time interval of an audible click.
1 In some alloys, such as iron–manganese–carbon and iron–manganese–nickel, the martensitic transformation occurs isothermally. For these systems, growth is still very rapid but the nuclei are formed by thermal activation.
420 Physical Metallurgy and Advanced Materials
The effect of applied stress
Since the formation of martensite involves a homogeneous distortion of the parent structure, it is expected that externally applied stresses will be of importance. Plastic deformation is effective in forming martensite above the M s temperature, provided the temperature does not exceed a critical
value usually denoted by M d . However, cold work above M d may either accelerate or retard the transformation on subsequent cooling. Even elastic stresses, when applied above the M s temperature and maintained during cooling, can affect the transformation; uniaxial compression or tensile stresses raise the M s temperature while hydrostatic stresses lower the M s temperature.
Stabilization
When cooling is interrupted below M s , stabilization of the remaining austenite often occurs. Thus, when cooling is resumed martensite forms only after an appreciable drop in temperature. Such thermal stabilization has been attributed by some workers to an accumulation of carbon atoms on those dislocations important to martensite formation. This may be regarded as a direct analog of the yield phenomenon. The temperature interval before transformation is resumed increases with holding time and is analogous to the increase in yield drop accompanying carbon build-up on strain ageing. Furthermore, when transformation in a stabilized steel does resume, it often starts with a ‘burst’, which in this case is analogous to the lower yield elongation.