Increase of martensite start temperature after small deformationof austenite

B.B. He a, W. Xu b, M.X. Huang a,n

a Department of Mechanical Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, Chinab ArcelorMittal Global R&D Gent, Technologie park 935, Zwijnaarde B9052, Belgium

a r t i c l e i n f o

Article history:Received 24 January 2014Received in revised form24 April 2014Accepted 30 April 2014Available online 9 May 2014

Keywords:Ms temperatureDeformationGNDSubgrainsMartensitic transformation

a b s t r a c t

It is generally considered that martensite start (Ms) temperature is decreased after plastic deformation ofaustenite if the previously applied load is retrieved. This can be explained by the dislocation stabilizationmechanism. The present work performed systematic deformation and dilatometer experiments toinvestigate the effect of plastic deformation on Ms temperature. It is found that Ms temperature firstincreases at small strain and then decreases at large strain. Since the dislocation stabilization mechanismcan only predict the decrease of Ms temperature after plastic deformation, a new mechanism is thusproposed to describe this new interesting finding. That is, the increase of Ms temperature is due to thepile-up of geometrically necessary dislocations at austenite grain boundaries, while the decrease of Mstemperature is caused by the formation of subgrains in austenite grain interior.

& 2014 Elsevier B.V. All rights reserved.

1. Introduction

An important feature of martensitic transformation (MT) insteels is that it has a certain transformation starting temperature,namely, the martensite start (Ms) temperature. Ms temperature isalso an important parameter for the thermal/mechanical treat-ment of steels. For example, the fabrication of carbide-free bainite(CFB) steel and quenching and partitioning (Q&P) steel, which arenew advanced high strength steels (AHSS) for automotive applica-tions [1,2], requires the knowledge of Ms temperature in advance[3,4]. Ms temperature depends on the chemical composition [5],grain size [6], cooling rate [7], applied elastic stress [8] and defects[9]. It is reported that certain kinds of austenite grain boundariesmay provide the potent nucleation site [9,10]. On the other hand,the role of dislocation on MT is more controversial. The growth ofpre-existing martensite embryos requires the nucleation of newdislocation loops at the embryo/austenite interface and the glideof these dislocation loops can extend to the interface [11]. How-ever, the large amount of dislocations generated by a large plasticdeformation of austenite prior to the MT may stabilize the glissileembryo–austenite interface, leading to a decrease of Ms tempera-ture. This is known as the dislocation stabilization mechanism andhas been studied intensively in literature [12,13]. Nevertheless, theeffect of small deformation on Ms temperature is much less

investigated and not yet well understood. A recent report showedthat Ms temperature continuously decreased with the increase ofamount of deformation [14]. But the experiment suffered fromdecomposition of austenite into bainite during the deformationand quenching process so that the carbon content of retainedaustenite could be changed due to carbon partitioning or carbideformation, which would affect Ms temperature. Careful experi-ments should be designed to avoid austenite decompositionduring the deformation and quenching processes so that the effectof deformation on Ms temperature can be accurately measured.Thus, the present work aims to carry out well-designed systematicdeformation-dilatometer experiments to investigate the effect ofboth small and large deformations on Ms temperature, excludingthe effect of austenite decomposition.

2. Experimental procedures

Steel with a chemical composition of Fe–0.2C–1.5Mn–2Cr (inwt%) was employed as a model material for the present work. Thematerial was cast by levitation casting and hot rolled to a finalthickness of 6 mm. Cylindrical dilatometry samples with a lengthof 10 mm and a diameter of 5 mm were prepared from the hot-rolled sheets along the rolling direction. Dilatometry tests wereperformed in a Bähr, 805A/D deformation dilatometer. The Ac1 andAc3 temperatures were found to be 762 and 795 1C, respectively,with a heating rate of 10 1C s�1. The samples were homogenized at900 1C for 300 s. The heating and homogenization were carried

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/msea

Materials Science & Engineering A

http://dx.doi.org/10.1016/j.msea.2014.04.1080921-5093/& 2014 Elsevier B.V. All rights reserved.

n Corresponding author. Tel.: þ852 28597906; fax: þ852 28585415.E-mail address: mxhuang@hku.hk (M.X. Huang).

Materials Science & Engineering A 609 (2014) 141–146

out under a vacuum condition. The critical cooling rate forobtaining a fully martensitic microstructure was verified fromthe continuous cooling experiments to be 10 1C s�1. The isother-mal holding experiments were performed to search for the bestdeformation temperature to avoid the decomposition of austeniteduring deformation. The time-temperature-transformation (TTT)diagram was measured and is shown in Fig. 1(a) where the typicalC curves for the ferrite/pearlite and bainite domains are observed.The incubation time is based on the 5% transformed volumefraction. Fig. 1(a) shows that the decomposition of austenite at550 1C has the longest incubation time, indicating that 550 1C isthe best temperature to carry out the plastic deformation ofaustenite. Such low deformation temperature of austenite canminimize the dynamic recovery of dislocations, leading to a morenoticeable effect of deformation on Ms temperature.

The temperature-deformation program employed in the pre-sent study is schematically illustrated in Fig. 1(b). As it shows, thesample was heated with a heating rate of 10 1C s�1, followed by anisothermal holding at 900 1C for 300 s and then quenched to550 1C with a cooling rate of 30 1C s�1. This cooling rate is highenough to avoid the decomposition of austenite. After an isother-mal holding of 5 s at 550 1C the sample was immediatelyquenched to room temperature with a cooling rate of 100 1C s�1

to capture the Ms temperature. The samples were reheated andquenched three times to avoid any effect of initial microstructureand to achieve a better sample accommodation and a precisedilatation measurement. The variation of Ms temperature amongthese three thermal cycles was about 71 1C. The varied amount ofdeformation (compression) was applied at 550 1C with a strainrate of 1 s�1 during the fourth circle and then immediately

quenched to room temperature. The resultant Ms temperatureafter deformation was compared with the Ms temperature fromthe third quenching of the same sample to avoid sample inhomo-geneity. The only difference between thermal cycles and thedeformation cycle was the applied deformation. For microstruc-ture analysis, the deformed samples were cut into two parts alongthe radial direction and were observed at the center of the crosssection using a scanning electron microscope (SEM) at 5 kV (LEO1530) and electron backscattered diffraction (EBSD) at 20 kV (HKLChannel 5). The samples for SEM observation were prepared bymechanical polishing down to 1 μm and then slightly etched with2% nital solution for 15 s. The samples for EBSD measurementwere vibratory polished with colloidal silica after the mechanicalfinish of 1 μm. A step size of 0.15 μm was used for EBSD measure-ment. The mechanical properties of martensite transformedfrom the deformed austenite were characterized by measuringthe Vicker's hardness with an applied load of 10 kg at roomtemperature.

3. Results and discussions

Fig. 2(a) shows the dilatation curves for the cycles with andwithout a deformation of 4.7%. The dilatometer curves werearranged vertically to give a proper comparison. The definition ofMs temperature is based on the tangent method [15]. Fig. 2(a) shows that the Ms temperature during the third quenchingwas 391 1C, which was increased to 402 1C after a deformation of4.7%. The increase of Ms temperature (�11 1C) is much larger thanthe variation of Ms temperature among different thermal cycles,showing an obvious increase of Ms temperature after a smalldeformation of 4.7%. The net change of Ms temperature (ΔMs) isadopted here because the Ms temperature was only comparedbetween the cycles with and without deformation. The positivevalue of ΔMs indicates an increase of Ms temperature afterdeformation, while the negative one shows a decrease of Mstemperature after deformation. The ΔMs with respect to thedifferent applied strains is summarized in Fig. 2(b). Each datapoint represents an average of two independent tests. The errorbar is a standard deviation. As Fig. 2(b) shows, ΔMs increases withstrain initially up to a peak value of ΔMs at a strain of 9.4%, withthe maximal ΔMs of about 20 1C. Then, ΔMs decreases with anincrease of strain. It is noted that the zero point of ΔMs wasaround the deformation of 22%. With further straining, the ΔMsbecame negative and decreased with the increase of deformation,indicating that Ms temperature was suppressed by large deforma-tion, which is consistent with reports in literature [12]. Thecorresponding engineering stress–strain curves for differentamounts of deformation are shown in Fig. 2(c). The austenitewas continuously work hardened with the increase of appliedstrains. Fig. 2(d) shows that the average Vicker's hardness of theproduct martensite increases with the increase of applied strains.This may be due to the continuous refinement of martensitemicrostructure after an increase of applied plastic deformationon austenite.

The microstructure of product martensite transformed from theparent austenite with a prior deformation of 4.7% is shown inFig. 3(a). Several large martensite blocks were observed. Interest-ingly, the one marked with red arrow had a straight block boundary,across the austenite grain and halted at the prior austenite grainboundary (PAGB). The morphology of this martensite block mayindicate that its propagation was not hindered by the smalldeformation of 4.7%. Therefore, from the observation in Fig. 3(a),it may be concluded that a small deformation of 4.7% did not affectthe growth of martensite. With the increase of applied strains, themartensite blocks which penetrated the whole austenite grain are

Fig. 1. (a) The TTT diagram of the present steel. The ideal temperature fordeformation of austenite without decomposition is 550 1C. (b) A schematicillustration of temperature program for the present steel (�3 means that suchtreatment repeats 3 times). The red line represents the applied deformation at550 1C during the fourth cycle. (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

B.B. He et al. / Materials Science & Engineering A 609 (2014) 141–146142

seldom observed. This can be confirmed from the SEM images asshown in Fig. 3(b) and (c) for the strains of 9.4% and 18%,respectively. Fig. 3(d) shows that the microstructure of the marten-site blocks transformed from the austenite with a large deformationof 25.7% was significantly refined. The martensite blocks markedwith red arrows propagated from the austenite grain boundary,halted within the austenite grain and formed a sharp end. It canalso be observed that many martensite blocks initiated from thegrain interior, other than PAGB. The martensite blocks typicallyaccompanied curved boundaries. The martensite blocks werefurther refined after the applied deformation of 36.2%, as shownin Fig. 3(e). From the morphological analysis, it seems that a largedeformation of 25.7% and 36.2% strongly affected the growth of themartensite blocks.

Fig. 4(a) and (b) are the EBSD band contrast maps for themartensite transformed from the prior austenite with deformationof 4.7% and 25.7%, respectively. The band contrast maps confirmedthe observation from SEM images (Fig. 3) that the large plasticdeformation of 25.7% significantly refined the martensite micro-structure, while the small plastic deformation of 4.7% had almostno influence on the growth of the martensite.

The carefully designed experiments in the present work canensure that the deformation of austenite at 550 1C will not resultin any decomposition of austenite during deformation andquenching. The enormously long incubation time (�1800 s) forthe present alloy at 550 1C as shown in Fig. 1(a) can minimize thedecomposition of austenite into bainite during deformation. It canbe confirmed from the decomposition kinetics of deformedaustenite at 550 1C that this long incubation time remained, andthe detailed experiments will be reported elsewhere. By consider-ing the high strain rate of 1 s�1 as employed in the presentexperiment, it can be concluded here that no deformed austenite

decomposed into bainite during the deformation process. Thedilatation curve after a small deformation of 4.7% (Fig. 2(a)) showsthat the length of the deformed austenite continuously decreasedduring the cooling process at the regime of 450 1C, indicating thatthe cooling rate of 100 1C s�1 as employed in the present experi-ment was high enough to avoid the decomposition of thedeformed austenite at this temperature. By considering that thebainite transformation at 450 1C had the shortest incubation timeamong all other temperatures (Fig. 2(a)), it may be reasonable toconclude that no deformed austenite decomposed into bainiteduring the quenching process. So it can be confirmed that theplastic deformation only introduced dislocations and no decom-position of austenite affected MT. As the austenite grain wascontinuously work hardened after deformation (Fig. 2(c)), theincrease of Ms temperature after the small deformation canneither be explained by the dislocation stabilization mechanism[12] nor by the increased yield strength of the austenite [14]. Anew mechanism which can explain both the increase of Mstemperature at small deformation and decrease of Ms temperatureat large deformation is proposed as follows.

Ashby [16] showed that uniaxial deformation of polycrystallinematerials can lead to a pile-up of geometrically necessary disloca-tions (GNDs) at grain boundaries and the generation of randomlydistributed dislocations within the grain interior. These randomlydistributed dislocations can be rearranged to form subgrains in thegrain interior at large deformation.

According to Kaufman and Cohen [11], the martensite embryomay pre-exist at high temperature, freeze during the quenchingprocess and become super-critical at Ms temperature. As certainaustenite grain boundaries can provide the favoured nucleationsites [9,10], it is reasonable to assume that the pre-existingembryos may locate at or very close to those specific austenite

Fig. 2. (a) The dilatation curves for the cycles with and without deformation of 4.7%. (b) ΔMs with respect to the amount of deformation. Each error bar represents thestandard deviation of two independent tests. (c) Engineering stress–strain curves for the different amounts of deformation at 550 1C. (d) Vicker's hardness measurement ofthe martensite transformed from the parent austenite with different amounts of deformation. Each error bar represents the standard deviation of five indents.

B.B. He et al. / Materials Science & Engineering A 609 (2014) 141–146 143

grain boundaries. It is further assumed that the operation of somepre-existing embryos is sufficient for the dilatometer to capturethe Ms temperature. The nucleation of martensite at the subgrainboundaries may be possible at a later transformation stage such asthe transformation burst where the autocatalysis dominates butnot at the very beginning of MT (Ms) so that this scenario is notconsidered here. For the dilatation signal during MT to be macro-scopically detectable using dilatometer equipment, a certain frac-tion (χ) of nucleation sites occupied by the potent pre-existingembryos should operate together. χ is expressed as follows [17]:

χ ¼ n0exp �ΔGkT

� �ð1Þ

where n0 is a fraction of martensite embryos with critical potencyat a certain temperature (T), k is the Boltzmann constant and ΔG isthe activation energy for a heterogeneous nucleation of martensiteat the austenite grain boundary.

The potency of an embryo is defined as the minimum chemicaldriving force for a pre-existing embryo to activate [18]. A recent3-D phase field modeling showed that the potency of an embryodepends on its size, shape, orientation, dislocation density anddislocation mobility [19], confirming a previous theoretical pre-diction [20]. The embryo may be densely stacking dislocationarrays [21]. It is suggested that the embryo requires a critical

number of dislocations inside to be sufficiently potent [20] andthat the growth of embryo requires creation and expansion of newparallel dislocation loops [22]. Fig. 5(a) and (b) schematicallyillustrate the pile-up of GNDs at the austenite grain boundaryafter the small and large deformation, respectively. The orderlydistributed GNDs at the austenite grain boundary may fulfill therequirement of creation of new parallel dislocation loops for thegrowth of the martensite embryo. In other words, the potency ofan embryo may be increased due to the pile-up of GNDs at theaustenite grain boundary where the embryo is located. Corre-spondingly, the value of n0 and χ at the original Ms temperaturewill be increased due to the presence of GND induced bydeformation. The dilatation signal for detecting Ms temperatureis also linked with the average volume of initial martensite lathsgrown from some potent pre-existing embryos during quenching.Fig. 3(a) shows that a prior deformation of 4.7% on parentaustenite did not refine the product martensite blocks, whichindicates that the average volume of martensite laths was notaffected by this small deformation. It is schematically illustrated inFig. 5(a) that the initially transformed martensite lath (red rec-tangle) can still cross the whole austenite grain after the smalldeformation. It is noted here that lath martensite is a hierarchystructure which contains the packet, block, sub-block and lath. Ingeneral, the first transformed martensite laths can cross the whole

Fig. 3. SEM images for the product martensite transformed from the parent austenite with deformation of (a) 4.7%, (b) 9.4%, (c) 18%, (d) 25.7%, and (e) 36.2%. Part of PAGBmarked with dashed lines may still be discernible after large deformation. The red arrows point to the martensite blocks which may be significantly refined, depending onthe amount of deformation on the parent austenite grain. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of thisarticle.)

B.B. He et al. / Materials Science & Engineering A 609 (2014) 141–146144

austenite grain and thus partition the prior austenite grain [23]. Itis noted here that the volume expansion ratio during MT may notbe affected by the plastic deformation. By considering an increasedpotency of embryos and the unchanged average volume ofmartensite laths, an increase of Ms temperature after a smalldeformation can be predicted from Eq. (1). Fig. 3(d) shows thata large deformation of 25.7% on parent austenite significantly

refined the product martensite blocks, which may indicate that theaverage volume of martensite laths was also significantlydecreased. It is schematically illustrated in Fig. 5(b) that the lengthof the martensite lath was shortened due to the formation ofsubgrain boundaries which can be formed only after a largedeformation. On the other hand, the number of GNDs at theaustenite grain boundary will be saturated after a certain amountof deformation, indicating that the increased potency of theembryo will also be saturated. So the contribution of GNDs tothe Ms temperature will be saturated at a given strain. However,the subgrain size continuously decreases with strain, resulting in acontinuous decrease of the average volume of martensite laths.When the contribution of GNDs on an increase of Ms temperatureis saturated and the effect of subgrain size on Ms temperaturebegins to play a dominant role, one can expect that a much largervalue of χ will be required for Ms temperature to be detected bydilatometer. But this larger value of χ is only available at atemperature lower than the original Ms temperature, leading toa decrease of Ms temperature with strain at large deformation, asshown in Fig. 2(b). In other words, there are two competingmechanisms. The pile-up of GNDs can increase the Ms tempera-ture while the formation of subgrains will decrease the Mstemperature.

It has been suggested that the plastic deformation of austenitecan accelerate the slow bainite transformation kinetics of nano-bainite steel [24]. The present work shows that the Ms tempera-ture can be increased after the small plastic deformation ofaustenite. Since a low isothermal holding temperature is necessaryto obtain the fine bainite lath [3], one has to make sure that theresultant Ms temperature after the deformation of austenite is stillbelow the targeted bainitic holding temperature.

4. Conclusions

In summary, by detailed deformation-dilatometer investiga-tions, the present work demonstrates that Ms temperatureincreases with strain after small deformation and decreases withstrain after large deformation. This interesting finding can beunderstood in terms of the GND at the austenite grain boundariesand the subgrains in austenite grain interior. The increasedpotency of the martensite embryo due to the pile-up of GND atthe austenite grain boundary results in an increase of Ms tem-perature after small deformation. The formation of subgrains inaustenite grain interior refines the martensite microstructure,leading to a decrease of Ms temperature after large deformation.The present work fills the research gap for the effect of smallplastic deformation of austenite on the Ms temperature. Since thecorresponding mechanism is based on a general viewpoint ofplastic deformation and martensitic transformation, the increaseof Ms temperature after the small deformation should be widelyobserved.

Acknowledgments

The authors are grateful to Dr. K. Zhu from ArcelorMittal forstimulating discussions. MH acknowledges the financial supportfrom the Research Grants Council, University Grants Committee,Hong Kong (HKU719712E and HKU 712713E) and UniversityResearch Committee of HKU (201111159053).

Reference

[1] O. Bouaziz, H. Zurob, M.X. Huang, Steel Res. Int. 84 (2013) 937–947.[2] X.C. Xiong, B. Chen, M.X. Huang, J.F. Wang, L. Wang, Scr. Mater. 68 (2013)

321–324.

Fig. 4. EBSD band contrast maps of martensite transformed from the austenitewith a prior deformation of 4.7% (a) and 25.7% (b). The large plastic deformation of25.7% significantly refined the martensite microstructure.

Fig. 5. A schematic illustration of martensite formation in an austenite grain with adeformation of 4.7% (a) and 25.7% (b). The martensite embryo is illustrated as thesolid yellow ellipse, which was surrounded by the GNDs at the austenite grainboundary after deformation. The red rectangle represents the martensite lath,which was stopped by the austenite grain boundary at small deformation (a) andsub-grain boundary at large deformation (b). (For interpretation of the referencesto color in this figure legend, the reader is referred to the web version of thisarticle.)

B.B. He et al. / Materials Science & Engineering A 609 (2014) 141–146 145

[3] F.G. Caballero, H.K.D.H. Bhadeshia, Curr. Opin. Solid State Mater. Sci. 8 (2004)251–257.

[4] J. Speer, D. Matlock, B. De Cooman, J. Schroth, Acta Mater. 51 (2003)2611–2622.

[5] K. Andrews, J. Iron Steel Inst. 203 (1965) 721–727.[6] H.-S. Yang, H.K.D.H. Bhadeshia, Scr. Mater. 60 (2009) 493–495.[7] M. Nikravesh, M. Naderi, G. Akbari, Mater. Sci. Eng. A 540 (2012) 24–29.[8] J.R. Patel, M. Cohen, Acta Metall. 1 (1953) 531–538.[9] S. Kajiwara, Metall. Mater. Trans. A 17 (1986) 1693–1702.[10] K. Tsuzaki, N. Harada, T. Maki, J. Phys. IV 5 (1995) C8-167–C168-172.[11] L. Kaufman, M. Cohen, Prog. Met. Phys. 7 (1958) 165–246.[12] M. Maalekian, E. Kozeschnik, S. Chatterjee, H.K.D.H. Bhadeshia, Mater. Sci.

Technol. 23 (2007) 610–612.[13] S. Chatterjee, H.S. Wang, J.R. Yang, H.K.D.H. Bhadeshia, Mater. Sci. Technol. 22

(2006) 641–644.

[14] Z.M. Shi, J. Li, B. Chi, J. Pu, Y.S. Zhang, M.Q. Wang, J. Shi, H. Dong, Metall. Mater.Trans. A 44 (2013) 4136–4142.

[15] Y.C. Liu, F. Sommer, E.J. Mittemeijer, Acta Mater. 51 (2003) 507–519.[16] M.F. Ashby, Philos. Mag. 21 (1970) 399–424.[17] G.B. Olson, M. Cohen, J. Phys. Colloq. 43 (1982) C4-75–C74-88.[18] G. Olson, M. Cohen, Metall. Mater. Trans. A 6 (1975) 791–795.[19] H.K. Yeddu, A. Borgenstam, J. Ågren, J. Alloys Compd. (2012).[20] M. Cohen, Proceedings of the International Conference on Martensitic Trans-

formations, 1986, pp. 16–21.[21] J. Brooks, M. Loretto, R. Smallman, Acta Metall. 27 (1979) 1839–1847.[22] M. Cohen, Metall. Trans. 3 (1972) 1095–1098.[23] S. Nambu, N. Shibuta, M. Ojima, J. Inoue, T. Koseki, H.K.D.H. Bhadeshia, Acta

Mater. 61 (2013) 4831–4839.[24] W. Gong, Y. Tomota, M.S. Koo, Y. Adachi, Scr. Mater. 63 (2010) 819–822.

B.B. He et al. / Materials Science & Engineering A 609 (2014) 141–146146