Martensit

8/10/2019 Martensit

1/68

Martensitic Transformations in Steels

A 3D Phase-field Study

HEMANTHA KUMAR YEDDU

Doctoral Thesis

Stockholm, Sweden 2012

8/10/2019 Martensit

2/68

ISRN KTH/MSE12/12SE+METO/AVHISBN 978-91-7501-388-6

MaterialvetenskapKTH

SE-100 44 StockholmSWEDEN

Akademisk avhandling som med tillstnd av Kungl Tekniska hgskolan framlggestill offentlig granskning fr avlggande av teknologie doktorsexamen i material-vetenskap fredagen den 15 juni 2012 klockan 10.00 i sal B2, Brinellvgen 23, Ma-terialvetenskap, Kungl Tekniska hgskolan, 10044 Stockholm.

Hemantha Kumar Yeddu, June 2012

Tryck: Universitetsservice US AB

8/10/2019 Martensit

3/68

iii

The highest education is that which does not merely give us informa-tion but makes our life in harmony with all existence. RabindranathTagore, Nobel Laureate (1913).

8/10/2019 Martensit

4/68

8/10/2019 Martensit

5/68

v

Dedicated to the most wonderful persons in my life Mom and Aunty

8/10/2019 Martensit

6/68

8/10/2019 Martensit

7/68

8/10/2019 Martensit

8/68

8/10/2019 Martensit

9/68

8/10/2019 Martensit

10/68

x PREFACE

Respondents contribution to the papers

Supplement-1: The respondent performed all the work and wrote the entire

paper with inputs from J. gren (J..), A. Borgenstam (A.B.), A. Malik(A.M.) and G. Amberg (G.A.).

Supplement-2: The respondent performed all the work and wrote the entirepaper with inputs from J.. and A.B.

Supplement-3: The respondent performed all the phase-field part of the workas well as wrote major part of the paper with inputs from J.. and A.B. Theab initio part of the work is performed by V.I. Razumovskiy (V.I.R.) withinputs from P.A. Korzhavyi (P.A.K.) and A.V. Ruban (A.V.R.).

Supplement-4: The respondent contributed to the model development as wellas to drafting the paper with inputs from J.., A.B., A.M. and G.A.

Supplement-5: The respondent performed all the work and wrote the entirepaper with inputs from J.. and A.B.

Supplement-6: The respondent performed all the work and wrote the entirepaper with inputs from J.., A.B. and P. Hedstrm (P.H.).

This work has also resulted in the following presentations

Invited talk on Role of Plasticity during martensitic microstructure evolutionin steels: A 3D phase field study, presented at Plasticity 2012, San Juan,PR, USA, January 2012.

Invited talk on Phase Transformations in Materials - A mesoscale studyon steels using phase-field approaches, presented at Los Alamos NationalLaboratory, Los Alamos, USA, January 2012.

Oral presentation on 3D Phase Field Modeling of Martensitic MicrostructureEvolution in Steels, presented at Phase Transformations Conference (PTM2010), Avignon, France, June 2010.

Poster presentation on Physical Parameters Determination From 3D PhaseField Simulations of Martensitic Transformations in Steels, presented atTMS Annual Meeting, San Diego, USA, February 2011.

Poster presentation on 3D Modeling of Martensitic Transformations in Steelsby Integrating Thermodynamics, Phase Field Modeling and Experiments,at 1st World Congress on Integrated Computational Materials Engineering(ICME), Seven Springs, PA, USA, July 2011.

Poster presentation on Effect of martensite nucleus potency on the marten-sitic transformation in steels: A Phase Field study, presented at ICOMAT2011, Osaka, Japan, September 2011.

8/10/2019 Martensit

11/68

Contents

Preface ix

Contents xi

1 Introduction 11.1 Aim of the present work . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Martensite 32.1 Morphology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32.2 Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42.3 Plastic deformation aspects . . . . . . . . . . . . . . . . . . . . . . . 6

2.4 Nucleation and growth . . . . . . . . . . . . . . . . . . . . . . . . . . 62.5 Thermodynamic aspects . . . . . . . . . . . . . . . . . . . . . . . . . 72.6 Crystallography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

3 Phase-field method 13

4 Phase-field modeling of martensitic transformations 174.1 Morphology and transformation . . . . . . . . . . . . . . . . . . . . . 184.2 Plastic deformation aspects . . . . . . . . . . . . . . . . . . . . . . . 244.3 Nucleation and growth aspects . . . . . . . . . . . . . . . . . . . . . 24

4.4 Thermodynamic aspects . . . . . . . . . . . . . . . . . . . . . . . . . 254.5 Crystallography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5 Summary of results 275.1 Morphology and transformation . . . . . . . . . . . . . . . . . . . . . 275.2 Plastic deformation aspects . . . . . . . . . . . . . . . . . . . . . . . 365.3 Nucleation and growth aspects . . . . . . . . . . . . . . . . . . . . . 385.4 Thermodynamic aspects . . . . . . . . . . . . . . . . . . . . . . . . . 415.5 Crystallography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

6 Concluding remarks and future prospects 47

xi

8/10/2019 Martensit

12/68

8/10/2019 Martensit

13/68

8/10/2019 Martensit

14/68

8/10/2019 Martensit

15/68

Chapter 2

Martensite

2.1 Morphology

Depending on the alloy composition, martensite forms in the shape of laths, i.e.ruler shaped units, or plates, i.e. lenticular shaped units, as shown in Fig. 2.1.Carbon content of the alloy plays a major role in determining the martensite mor-phology. It has been experimentally observed that martensite forms in the shapeof laths in low carbon steels and in the shape of plates in high carbon steels [15].

Figure 2.1: Light optical microscope images of the microstructure of (a) lathmartensite in an IF steel with very low carbon content of 0.0049 wt% C [16] (b)plate martensite in Fe1.86 wt% C alloy [17].

In the case of lath martensite, several parallel martensite units are formed ad-jacent to each other [15]. It has been reported that lath martensite forms in blocksand packets in an austenite grain as shown in Fig. 2.1 [16,18]. The strength andtoughness of martensitic steels, with lath morphology, are strongly related to packet

3

8/10/2019 Martensit

16/68

4 CHAPTER 2. MARTENSITE

and block sizes [18,19].In the case of plate martensite, non-parallel martensite units are formed [15].

Moreover, a distinct feature in the case of plate martensite is the midrib, which isreported to be the first forming unit consisting of many transformation twins [20,21].On either side of the midrib there exist twinned regions, with some twins that extendfrom the midrib, and untwinned regions, with dislocations [20,21].

2.2 Transformation

During rapid cooling (quenching) of austenite, athermal martensitic transformationbegins at the martensite start temperature (Ms). Thereafter, the volume fractionof martensite increases with decreasing temperature and finally, the transformationis completed on reaching the martensite finish temperature (Mf). In the case ofstress assisted martensite, the martensite start temperature at a given stress level is termed Ms.

Fig. 2.2 shows the microstructural images obtained during an in situ observa-tion, by using laser scanning confocal microscope, of plate martensite formation bymeans of athermal transformation during rapid quenching of a high carbon steelwith a composition of Fe-0.88%C-4.12 %Cr, expressed in mass % [22]. The realtime and temperature can be seen in the upper left corner. Slightly above theMs, one can see untransformed austenite grains in Fig. 2.2a. Slightly below theMs, martensitic transformation initiates in a heterogeneous manner, i.e. nucleationoccurs at different sites as can be seen in Fig. 2.2b. One can see, in Fig. 2.2b,the untransformed austenite grains as well as partly transformed austenite grainswith martensite plates. At a temperature close to Mf, one can see an almostcompletely transformed steel with complex martensitic microstructure containingseveral martensite plates oriented in different directions, as shown in Fig. 2.2c.Thus it can be understood that the martensite volume fraction increases with de-creasing temperature.

During rapid cooling, a diffusion controlled transformation of austenite to fer-rite, with a very low carbon content, does not occur due to lack of time. Howeveras the temperature is reduced below the T0 temperature, i.e. the temperaturewhere the Gibbs energies of ferrite and austenite are the same, there exists a ther-modynamic driving force available for the formation of ferrite (martensite) withthe same composition as that of austenite, i.e. occurrence of a diffusionless phasetransformation.

Martensitic transformation leads to the crystallographic transformation of facecentered cubic (FCC) austenite in to body centered cubic (BCC) martensite. Thecarbon atoms that are randomly distributed on the interstitial sites in FCC donot have time to migrate to the BCC in a random manner and hence move in acoordinated motion. This increases the tetragonality of the BCC lattice and thusthe carbon containing martensite is of body centered tetragonal (BCT) structure.The tetragonality of martensite increases with increasing carbon content [23].

8/10/2019 Martensit

17/68

2.2. TRANSFORMATION 5

Figure 2.2: In situ observation, by using laser scanning confocal microscope, ofplate martensite formation in a high carbon steel, where the real transformationtime and temperature are marked in the upper left corner [22]. (a) Microstructureat a temperature slightly above Ms. (b) Microstructure at slightly below Ms. (c)Final martensitic microstructure at a temperature close to Mf.

In addition to the increased tetragonality, the increase in carbon content alsoleads to a volume expansion, i.e. dilatation. Moreover, the solid state nature of thetransformation, that takes place by a cooperative movement of atoms, requires thetwo phases to be highly coherent and hence gives rise to a large amount of internalstresses inside the material. The internal stresses are relaxed by means of plasticdeformation, i.e. by shear deformation. Moreover, the shear deformation needsto be a lattice-invariant shear such that it does not give rise to a macroscopicallyinhomogeneous crystal structure. Hence the lattice-invariant shear should only bemicroscopically inhomogeneous [23].

Theoretically, it has also been shown that the dilatation and shear needs to beassociated with the rotation of martensite such that the combination of all threemechanisms gives rise to the experimentally observed crystallographic orientation

8/10/2019 Martensit

18/68


relationships of martensite [23]. Thus based on the above discussion, it can be un-derstood that the martensitic transformation occurs by a cooperative short rangemovement of atoms such that the crystal lattice coherency (continuity) is main-tained.

2.3 Plastic deformation aspects

As mentioned in Section 2.2, plastic (shear) deformation in the surrounding austen-ite and internally in the martensite facilitates the relaxation (accommodation) ofstresses, which are generated due to the volume and shape changes caused bythe athermal transformation. In the case of stress-assisted transformation, theplastic accommodation process is popularly known as the Greenwood-Johnson ef-

fect [2426], which can be considered to be at the origin of TRIP (Transformationinduced plasticity) phenomenon. The TRIP phenomenon has been found to beadvantageous in increasing the strength, imparted by martensite, and in increasingthe ductility, caused due to the dislocations generated by the plastic deformationduring martensitic transformation. Thus plastic deformation plays an importantrole in the martensitic transformation.

The platic (shear) deformation should not disturb the overall crystal structureand hence needs to be a lattice-invariant shear, as explained in Section 2.2. Thelattice-invariant shear deformation can be generated by the movement of perfectdislocations or partial dislocations that gives rise to slip or twinning respectively

[23], as shown in Fig. 2.3. Fig. 2.3a corresponds to a case when there is no plasticdeformation inside the martensite and hence in this case the shape changes arecompletely accommodated by deformation of the surrounding austenite. Figs. 2.3band 2.3c show two different cases of plastic deformation, i.e. slip and twinningrespectively, inside martensite.

It has been reported that the type of shearing mechanism also affects the mor-phology of martensite [15,23]. In low carbon steels, where lath martensite is formed,slip mechanism is preferred and in high carbon steels, where plate martensite isformed, twinning mechanism is preferred [23]. Thus it can be understood thatlath martensite is a dislocation-rich structure [15, 23, 27], whereas plate marten-

site contains several internal twins [15,23], as mentioned in Section 2.1. The highdislocation density of lath martensite also indicates that the major plastic defor-mation occurs inside martensite, despite its higher yield strength compared to thatof austenite. Dislocations also play a dominant role during nucleation and growthof martensite, as explained in the following section.

2.4 Nucleation and growth

Martensite nucleation occurs in a heterogeneous manner [28], as can be seen in Fig.2.2b. The nucleation initiates at a point in the austenite where there is a densestacking of dislocation arrays [2830]. The heterogeneous nature of nucleation can

8/10/2019 Martensit

19/68

8/10/2019 Martensit

20/68


martensite. At T0 temperature, where there is no chemical driving force availableas the Gibbs energies of the two phases are equal, the Gibbs energy barrier G

opposes the phase transformation. As the temperature is reduced to Ms, duringquenching, the driving force becomes large enough such that it can overcome theGibbs energy barrier and hence the phase transformation takes place. Apart fromthe chemical energy Gchem, aroused due to quenching, the austenite-martensiteinterface gives rise to an interfacial energy, i.e. the gradient energy Ggrad, and thecoherency of the two phases gives rise to internal stresses that in turn give rise toelastic strain energy Gel.

Based on the studies on nucleation and growth mentioned in Section 2.4, it canbe understood that there exist different thermodynamic driving forces that governthe different stages of the martensitic transformation. Some works have showed thata less driving force is needed for growth [40, 41] and autocatalysis [37] comparedto that required for nucleation. Thus martensite, once nucleated, could grow andalso give rise to autocatalysis above Ms temperature. However Borgenstam etal. [40] reported that the driving force for nucleation and growth are very close toeach other, although a less driving force is needed for growth compared to that fornucleation.

In the case of stress-assisted martensite formation the externally applied me-chanical energyGappl, depending on the nature of the load, either contributes to ordetracts to the thermodynamic driving force. Thus if the nature of load is favorablefor martensite formation, one can expect that Ms is higher than Ms, whereas M

s

is lower than Ms if the nature of load is not favorable for martensite formation.Moreover, for a given stress state, only those martensite units with the most favor-able orientation with respect to the applied stress are formed. This effect is knownas the Magee effect [2426].

The thermodynamic aspects of both athermal and stress-assisted transforma-tions, in terms of an internal variable that locally represents the extent of thetransformation, is schematically shown in Fig. 2.4. The internal variable can beconsidered as a parameter that corresponds to the crystal structure of the twophases, i.e. austenite (= 0) and martensite (= 1), as explained in the followingsection.

2.6 Crystallography

The common plane, between the austenite and martensite, on which martensitenucleates is called the habit plane [42]. Experimental results show that the habitplane in the case of lath martensite is close to {111} [43,44] as well as {557} [5].

As martensitic transformation occurs by a cooperative movement of atoms thattransforms the FCC structure in to the BCC or BCT structure, the habit plane, i.e.the interface between the product and parent phases must be highly coherent. Thisis also supported by the experimental observations that the habit plane is well-defined crystallographically, i.e. there exist specific crystallographic orientation

8/10/2019 Martensit

21/68

2.6. CRYSTALLOGRAPHY 9

Figure 2.4: Schematic showing the thermodynamic aspects of athermal and stress-assisted martensitic transformation. Gibbs energy curves in the absence of chemicaldriving force, in the presence of chemical driving force only, as well as in the presenceof both chemical and mechanical driving forces.

relationships between austenite and martensite.In order to understand the crystallography of martensite, phenomenological

theories were developed independently by Wechsler, Lieberman and Read and byBowles and Mackenzie [23]. Such phenomenological theories are based on the con-cept that the habit plane should be an invariant plane, i.e. undistorted and unro-tated plane [23]. Thus a deformation along the invariant plane that also satisfiesthe coherency requirement is called an invariant plane strain, where twinning is anexample of such a deformation.

An austenite-martensite interface could be a coherent invariant habit plane ifthe FCC and BCT are oriented in such a way that their dense-packed planes anddense-packed directions are parallel to each other. Thus there exist specific crys-tallographic orientation relationships between austenite and martensite, i.e. ori-entation (rotation) of the habit plane of the martensite crystal with respect to aparticular crystallographic plane in the austenite crystal.

There exist four main austenite-martensite orientation relationships (OR),namely Bain OR, Nishiyama-Wassermann (N-W) OR, Kurdjumov-Sachs (K-S) ORand Greninger-Troiano (G-T) OR [23,45,46]. The K-S and N-W OR describe theexperimentally observed OR of martensite, whereas the G-T OR is intermediatebetween the K-S and N-W OR [46]. The Bain OR is considered to be the basic OR

8/10/2019 Martensit

22/68


that can be used to explain the K-S and N-W OR, as will be explained below.According to the Bain OR, an FCC austenite crystal (= 0) can be transformed

to BCT martensite crystal (= 1) by applying the Bain distortion (Bain strain), i.e.a set of a compressive strain along one of the coordinate axes and two equal tensilestrains along the other two coordinate axes [23]. There exist three possibilitiesto apply such a Bain strain on an FCC crystal and hence there exist three Bainstrains that give rise to three Bain variants (1, 2, 3), as shown in Fig. 2.5. TheOR between FCC and BCT crystals, according to Bain correspondence is shown inFig. 2.6.

Figure 2.5: Schematic showing the three different Bain variants [4].

Figure 2.6: Bain orientation relation showing the correspondence between FCC and

BCT crystals [23].

8/10/2019 Martensit

23/68

8/10/2019 Martensit

24/68


Figure 2.8: A schematic of the K-S orientation relation [18].

8/10/2019 Martensit

25/68

Chapter 3

Phase-field method

The modeling, performed more than a century ago by van der Waals, of a liquid-gas system by means of density function that continuously varies over the inter-face can be considered as the first application of the phase-field method. TheGinzburg and Landaus work, approximately 50 years ago, to model superconduc-tivity using a complex-valued order parameter can be considered to be the originof the concept of the order parameters, i.e. phase-field variables. The micro-scopic (continuous) theories proposed by Landau to describe the structural changesoccurring during a order-disorder transition by using mathematical expressions,which are called Landau polynomials, are widely used in the modeling of orderingtransformations. The work by Cahn and Hilliard that proposed a thermodynamicformulation that accounts for the gradients in thermodynamic properties in het-erogeneous systems with diffuse interfaces is usually regarded as the basis for thepresent phase-field models. Khachaturyan and his co-workers, in the 80s, havedeveloped Phase-field Microelasticity theory, i.e. a phase-field theory coupled withelastic strain calculations based on the Eshelby approach, and by using Landaupolynomials to model martensitic transformations. Thereafter the order-disordertransitions have been further explored by several researchers and their co-workers,e.g. L.Q.Chen, Y.Wang, A.Artemev, A.Yamanaka, A.Finel, A.Saxena, T.Lookman,S.Shenoy, K.Bhattacharya and V.Levitas. Another significant contribution to thephase-field approach can be attributed to Langer, who developed a phenomeno-logical single phase-field model that is mainly used to model solidification. There-after the solidification-type of models are well studied by several researchers, e.g.J.Warren, B.Boettinger, R.Kobayashi, A.Karma, I.Loginova, J.Agren, G.Ambergand N.Moelans. Phase-field models are now widely used for simulating a wide rangeof phenomena, e.g. microstructure evolution, solidification, precipitate growth andcoarsening, martensitic transformations, grain growth, solid state phase transfor-mations, crack propagation and nucleation, dislocation motion [24]. More detailsabout the above mentioned history and the corresponding references can be foundin Ref. [4].

13

8/10/2019 Martensit

26/68

14 CHAPTER 3. PHASE-FIELD METHOD

Figure 3.1: A schematic of the diffuse interface approach, i.e. continuous variationin the internal properties across the interface [48].

Phase-field method is a diffuse interface approach, i.e. the internal propertiesof a system undergoing a phase transformation vary continuously over the interfacebetween the two phases. The diffuse interface approach can be explained with thehelp of schematic Fig. 3.1 that shows the advancement of a domain wall at threeinstances [48]. Each vertical rod represents the direction of magnetization of a planeof atoms, whereas each row of rods represent the domain wall at an instance. Oneither side of the center, indicated by an arrow in the middle, one can see that themagnetic spin of atoms is different. It can also be seen that the spin direction varies

continuously over the interface. From time to time, it can be seen that the domainwall (arrow) advances and also that the spin characteristics change continuouslyover the interface. Hence the phase-field method, which is based on the diffuseinterface approach, is capable of predicting the continuous variation in the internalproperties over the interface between the two phases of a system that is undergoinga phase transformation. Thus the advantage of the diffuse interface approach is thatit avoids the need to explicitly track the position of the moving phase interface ateach instance. Thus phase-field models are able to predict complex microstructuralevolutions.

During a phase transformation occurring in a material, the microstructure

evolves due to minimization of the Gibbs energy of the system. The Gibbs energy ofthe system under consideration is described based on the underlying physics of thephenomenon that has to be studied. Thus depending on the physical description ofthe system, the Gibbs energy Gcould consist of different parts, e.g. bulk (Gbulk),interfacial (Gint), elastic (Gel), or any other applied external energy (Gappl).

G= Gbulk +Gint +Gel +Gappl (3.1)

The bulk part of the Gibbs energy Gbulk corresponds to the bulk chemical proper-ties of the two phases. The interfacial part of the Gibbs energy Gint correspondsto the properties of the interface. The elastic part of the Gibbs energy Gel cor-responds to the elastic strain energy that is generated inside the material due to

8/10/2019 Martensit

27/68

15

Figure 3.2: A schematic of the phase-field approach.

the phase transformation, e.g. martensitic transformation, spinodal decomposition.The applied external energy Gappl corresponds to the energy generated due to anexternally applied force, e.g. mechanical force due to externally applied stresses,magnetic forces due to an externally applied magnetic field.

The Gibbs energy G is dependent on several internal variables (properties) thatdescribe the material characteristics, such as grain characteristics as well as thecomposition of the material. In order to model the microstructure evolution, theGibbs energy G(, c) of the system is expressed as a mathematical function thatdepends on the internal variables, and c. , a non-conserved quantity, governs thecharacteristics of the grain (crystal) andc, a conserved quantity, is the compositionof the system. Such internal variables that describe the material characteristics arefield variables that are continuous mathematical functions in time and space, i.e.(x , y , z , t) and c(x , y , z , t), and are called phase-field variables.

The values of the phase-field variables vary between 0 and 1. In the case of asingle phase-field approach used to study solidification, = 0 or 1 corresponds toeither of the two bulk phases, whereas a deviation from these two values correspondsto the interface region between the two bulk phases. In the case of multi-phase-fieldapproach used to study an order-disorder phase transformation, which gives rise toan asymmetry in the resultant crystal structure such that each crystal orientationof the product phase is distinct, one has to consider different phase-field variablesthat represent each of the above mentioned crystal orientations. Thus the numberof phase-field variables will depend on the number of crystal orientations. Eachphase-field variable will be equal to 0 in the product phase and equal to 1 in theirrespective product phases.

The temporal, i.e. time-dependent, evolution of the microstructure is trackedby solving a set of partial differential equations (PDEs). Thus the solutions of

8/10/2019 Martensit

28/68

16 CHAPTER 3. PHASE-FIELD METHOD

the two PDEs yield the values of the internal variables at each time step and ateach point in the material and thereby predicts the changes in the microstructure.A schematic of the phase-field approach is presented in Fig. 3.2. The PDE thattracks the changes in the non-conserved quantities, such as grain characteristics, isusually referred as Cahn-Allen equation, whereas the PDE that tracks the changesin the conserved quantities, such as composition, is usually referred as Cahn-Hilliardequation [24]. Moreover both the equations contain several kinetic coefficients thatgovern the kinetics of the respective phase-field variables. In the schematic Fig. 3.2,the kinetic coefficients are not presented in order to give a rather simple picture ofthe phase-field modeling concept. More detailed derivations and explanations canbe found in some of the reviews on phase-field modeling [24].

8/10/2019 Martensit

29/68

Chapter 4

Phase-field modeling of martensitic

transformations

The pioneering efforts in simulating the martensitic microstructure evolution couldbe attributed to Khachaturyan [5] and his co-workers [6,8,9,12]. The other signif-icant contributions to this field could be attributed to Bhattacharya [11], Saxenaand his co-workers [7,49,50], Levitas and his co-workers [10,5153]. The effect ofthe plastic deformation on the transformation, by coupling the phase-field methodwith the plasticity equation, has been studied by Tomita and his co-workers [13,14],

based on the work of Guo et al. [54].The present work is based on the work by Khachaturyan et al. [5, 8, 9]. Theplastic deformation is also included, based on the work by Yamanaka et al. [13]. Themodeling of the transformation in polycrystals is based on the work by Yamanakaet al. [14]. The phase-field as well as the continuum mechanics equations are solvedby using the finite element method, which allows a straight forward and transparentformulation of the equations.

The finite element method (FEM) is well established for the modeling of ther-momechanical deformation. The basis of the method is to solve for the responseof a complicated body to imposed constraints by dividing it into smaller elements,

whose responses are more readily evaluated. Each element has a simple shape, suchas a triangle or quadrilateral for a 2D problem, or a tetrahedron for a 3D analysis.The vertices of the elements are called nodes. The complete body is formed bya mesh of elements, joined at nodes. Associated with each element is a stiffness,which is dependent upon the constitutive law of the material within the element.By subjecting the element to a force, the overall response of the body is determinedby simultaneously solving for the node displacements such that the conditions ofequilibrium and boundary constraints are met. A detailed explanation about theFEM can be found in Ref. [55].

All the single crystal computations are performed on a mesh with 50x50x50 grid

points and by using tetrahedral finite elements, whereas the polycrystal computa-

17

8/10/2019 Martensit

30/68

18CHAPTER 4. PHASE-FIELD MODELING OF MARTENSITIC

TRANSFORMATIONS

tions are performed on a mesh with 150x150x150 grid points. The entire mathe-matical framework is solved by using femLego software, a symbolic computationaltool [56]. femLego is a set of Maple procedures and fortran subroutines that can beused to build complete fortran simulation codes for partial differential equations,with the entire problem definition done in Maple. OpenDX software is used forplotting the images.

In the following sections, a brief overview of the phase-field model developedto study the morphology and transformation aspects, associated with athermalmartensitic transformation occurring in a single crystal as well as in a polycrys-talline material and stress-assisted martensitic transformation occurring in a singlecrystal, is presented. Einstein summation convention, i.e. summation implied overrepeated indices, is used to express all the equations except for the indices repre-senting different martensitic domains.

A brief procedure explaining the multi-length scale modeling, i.e. by couplingCALPHAD, ab-initio, phase-field and experimental data, of martensitic transfor-mation is also presented. A summary of the simulation procedures to study variousphysical concepts, such as plastic deformation, nucleation and growth, thermody-namic aspects and crystallography, associated with martensitic transformation isalso presented. More details can be found in the respective supplements.

4.1 Morphology and transformation

Athermal martensitic transformation, Supplement-1In this part of the work, athermal martensitic transformation occurring in a singlecrystal of Fe0.3%C steel is considered. A diffusionless phase transformation, suchas martensitic transformation, can be modeled using the Cahn-Allen equation [57],also known as the Time-Dependent Ginzburg-Landau (TDGL) kinetic equation [5].The microstructure evolution can be simulated by predicting the time-dependentvariation of the phase-field variable, which in turn is related to the minimization ofthe Gibbs energy Gof the system with respect to phase-field variable p as:

p

t =

q=v

q=1

Lpq

G

q (4.1)

where Gq

is a variational derivative that serves as a driving force for the forma-tion of martensite denoted by the phase-field variable q that is dependent on theposition vector r, v is the total number of martensite variants and Lpq is a kineticparameter.

In the present work the Bain orientation relation is considered, as this is regardedas the basic orientation relation that can be used to explain the K-S and N-Worientation relations as explained in Section 2.6. According to Bain orientationrelation, explained in Section 2.6, the phase transformation of FCC austenite toBCT martensite gives rise to three martensite variants (orientations). Thus in

8/10/2019 Martensit

31/68

4.1. MORPHOLOGY AND TRANSFORMATION 19

order to represent the three possible orientations, three phase-field variables, i.e.1, 2, 3, need to be considered in the model and hence three phase-field equations,i.e. for p = 1, 2, 3 and v = 3 in Eq. (4.1), need to be solved at each time step. Thusthe three phase-field variables correspond to the three Bain groups (Bain variants),mentioned in Section 2.6.

From a thermodynamic point of view, as explained in Section 2.5, the Gibbsenergy consists of three parts:

G=V

Gchemv +G

gradv +G

elv

dV (4.2)

Gchemv corresponds to the chemical part of the Gibbs energy density of an un-stressed system at the temperature under consideration. Ggradv is the extra Gibbs

energy density caused by the interfaces. The transformation of cubic austenite intotetragonal martensite induces elastic strain energy density Gelv, into the material.

Chemical energy

The chemical part of the Gibbs energy density Gchemv , expressed as a Landau-typepolynomial [9,58], is given by:

Gchemv (1, 2, 3) = 1Vm

12

A 21+

22+

23

13

B 31+

32+

33

+14

C

21+22+

23

2 (4.3)

By considering the driving force Gm, i.e. the difference in the Gibbs energiesof austenite and martensite, and the Gibbs energy barrier G terms in the aboveequation, the coefficients are modified [58] as: A= 32G, B= 3A 12Gm andC= 2A12Gm. G is expressed as G =

Vm22 . Vm is molar volume, is

thickness of the interface and relates to interfacial energy as explained below.

Gradient Energy

The gradient energy density term, Ggradv as presented in [9,58] can be expressed as:

Ggradv =p=vp=1

ij(p)p

ri

p

rj(4.4)

wherer(x,y,z)is the position vector expressed in Cartesian coordinates. ij is theisotropic gradient coefficient matrix, i.e. diagonal tensor with all the elements equalto= 9

2Vm

16G

. is interfacial energy, Vmis molar volume and G is Gibbs energybarrier.

8/10/2019 Martensit

32/68


TRANSFORMATIONS

Elastic energy

The internal elastic stresses developed in the material, as explained in Sections 2.2

and 2.5, give rise to elastic strain energy density Gelv. The elastic stress can becalculated using the Microelasticity theory [5] as explained below. As mentioned inSection 2.6, martensitic transformation gives rise to a change in the crystal structurecaused by Bain strains00ij(r), which in turn induce stress-free transformation strains0ij(r) in the material. The surrounding austenite matrix exerts strain ij(r) on themartensite to resist the stress-free transformation strains and thereby induces elasticstrain elij(r), which in turn gives rise to stress ij(r), in the material. The actualstrain tensor ij(r) is linearly related to the local displacement vector u(r), whichis obtained by solving the mechanical equilibrium condition.

The elastic strain energy density term Gelv can be expressed as:

Gelv = ij(r)0ij(r)

ij(r)dij(r) (4.5)

where 0ij(r) is the stress-free transformation strain, ij(r) is the actual strain andij(r) is the stress, given by:

ij(r) = cijklelkl(r) (4.6)

wherecijklis the tensor of elastic modulii and elkl(r) is the elastic strain, given by:

elkl(r) = kl(r)0kl(r)

plkl(r) (4.7)

whereplkl(r) is the plastic strain.The actual strain ij(r) is given by:

ij(r) =12

ui(r)

rj+

uj(r)ri

(4.8)

where u(r) is the local displacement vector obtained by solving the mechanical

equilibrium condition, expressed as:

ij(r)rj

= 0 (4.9)

The displacement vector can then be used to calculate the actual strain ij(r),which in turn can be used in calculating the local elastic strain energy density Gelv.

The stress-free transformation strain is given by:

0

ij(r) =

p=v

p=1

p(r)00

ij(p) (4.10)

8/10/2019 Martensit

33/68

8/10/2019 Martensit

34/68

8/10/2019 Martensit

35/68

8/10/2019 Martensit

36/68


TRANSFORMATIONS

0ij(r) =p=v

p=1

Qik( , , )Qjl( , , )p(r)00kl (p) (4.22)

where the rotation matrix Q( , , ) is given by:

Q(,,) =coscos cossin+sinsincos sinsin+cossincoscossin coscos+sinsinsin sincos+cossinsin

sin sincos coscos

(4.23)

Eq. (4.10) is replaced with Eq. (4.22) and the entire model framework explainedin the above section corresponding to Supplement-1 is updated accordingly to modelthe athermal martensite formation in polycrystalline materials.

4.2 Plastic deformation aspects, Supplements-5 and 6

The plastic deformation aspects associated with the transformation can be studiedby studying the effect of plastic relaxation rate k on the transformation. Hencedifferent simulations are performed with different plastic relaxation rates k andby considering the driving force for nucleation at Ms. The model presented in

Supplement-1 is employed to perform the simulations by considering the athermalmartensitic transformation occurring in a single crystal of Fe0.3%C steel.

It can be observed from Eqs. (4.13) and (4.14) that k controls the plastic strainrate and thereby quantitatively controls the extent of plastic deformation of the ma-terial. Orowan equation states that the strain rate is proportional to the dislocationdensity, Burgers vector and the dislocation velocity, which in turn is proportionalto the shear stress [59]. Thus by comparing Eq. (4.13) with the Orowan equation, itcan be understood that k is directly proportional to the dislocation mobility as wellas to the dislocation density. Hence by studying the effect of plastic relaxation rate,the effect of dislocation density and dislocation mobility on the embryo potency,

morphology as well as on the overall transformation can be studied.

4.3 Nucleation and growth aspects, Supplement-5

Although nucleation is not simulated in the present work, several aspects associatedwith the martensite nucleation are investigated by studying the effect of the marten-site embryo potency on the transformation. The model presented in Supplement-1is employed to perform the simulations by considering the athermal martensitictransformation occurring in a single crystal of Fe0.3%C steel. As mentioned inSection 2.4, the potency of the embryos depends on their size, shape, orientationand the dislocation density at that nucleation site.

8/10/2019 Martensit

37/68

4.4. THERMODYNAMIC ASPECTS 25

In order to study the effect of the dislocation density on the embryo potency,the effect of the model parameter, i.e. plastic relaxation rate k in Eq. (4.14), on thetransformation is studied. Thus different simulations are performed with differentvalues ofk. The relation between k and the dislocation density is explained abovein Section 4.2. In order to study the favorable size of the martensitic embryo, dif-ferent simulations are performed with different embryo sizes. In order to study thefavorable shape of the martensitic embryo, different simulations are performed byinitiating the transformation (simulation) with a pre-existing spherical martensiteembryo and with an ellipsoidal embryo. In order to study the effect of the embryoorientation on the transformation, three different simulations are performed by con-sidering a pre-existing ellipsoidal shaped embryo with different orientations, i.e. thelongest axis of the ellipsoid being oriented along the three different coordinate axes.

4.4 Thermodynamic aspects, Supplement-6

As explained in Section 2.5, there exist different thermodynamic driving forcesthat govern different stages of the transformation, such as driving forces requiredfor nucleation, growth and autocatalysis. The driving force for nucleation (Gnucl.)atMsis obtained by using Thermo-Calc software and TCFE6 database [60], whichis based on the CALPHAD method. For a given driving force Gnucl., the value ofthe plastic relaxation rate kis determined, as explained above in Section 4.3, suchthat the transformation occurs at the experimental Ms, i.e. the embryo attains

potency at Ms. The model presented in Supplement-1 is employed to perform thesimulations by considering the athermal martensitic transformation occurring in asingle crystal of Fe0.3%C steel.

The critical driving force required for growth (Ggrowth) of martensite is de-termined by starting the simulation with Gnucl. and thereafter by reducing thedriving force in a step-wise manner until the growth completely stops. In orderto study the critical driving force required for autocatalysis (Gautocat.), a similarprocedure as mentioned above is employed by ensuring that autocatalysis is avoidedat each step.

4.5 Crystallography, Supplement-6

The habit plane of martensite is determined by plotting the plane along whichthe infinitesimally small unit of martensite forms during the simulations. Therotations of martensite variants, during the simulations, are studied by observingthe deviation of the martensite units from the above mentioned habit plane. Theconcept of the invariant habit plane, as explained in Section 2.6, is also studied byplotting a 2D section image of the 3D microstructure image and the correspondingdisplacement plot. The model presented in Supplement-1 is employed to performthe simulations by considering the athermal martensitic transformation occurring

in a single crystal of Fe0.3%C steel.

8/10/2019 Martensit

38/68

8/10/2019 Martensit

39/68

Chapter 5

Summary of results

5.1 Morphology and transformation

Athermal martensitic transformation, Supplement-1

The simulation results indicate that the martensitic microstructure consists of sev-eral morphological mirrors, i.e. twinned variants marked by 1 and 2 as shown inFig. 5.1, as also observed in the experimental works [15]. It is observed from thesimulations, as shown in Fig. 5.2, that nucleation of several martensite variants,i.e. autocatalysis, occurs as a consequence of the materials pursuit to minimizethe internal stresses. It is also observed that in the absence of plastic deforma-tion and dilatation, the martensitic microstructure formation is completed earliercompared to that in the presence of plastic deformation and dilatation, as shownin Fig. 5.3. It can be understood that plastic deformation facilitates relaxation

Figure 5.1: Martensitic microstructure consisting of twinned martensite variants

[58].

27

8/10/2019 Martensit

40/68

28 CHAPTER 5. SUMMARY OF RESULTS

(a) (b) (c)

Figure 5.2: 3D simulation results with elastoplastic material and having anisotropicelastic properties [58]. Snapshots at (a) t*=0 (b) t*=40 (c) t*=230.

Figure 5.3: Top view of the simulation results of athermal martensitic trans-formation in single crystal. (a) Without dilatation and plastic deformation, atnon-dimensionless time t*=170. (b) With dilatation and plastic deformation, at

t*=230 [58].

of the stresses and hence reduces the need for minimization of stresses by meansof new martensite variant formation, i.e. autocatalysis. Thus it can be concludedthat both autocatalysis and plastic deformation are different modes of relaxing theinternal stresses generated due to the martensitic transformation.

From Fig. 5.3, it can be seen that martensite variants form in pairs, i.e. variantpairs (1, 2), (1, 3) and (2, 3), in different directions. This pattern is observed inboth pure elastic material, i.e. Fig. 5.3a, as well as in the elastoplastic material, i.e.Fig. 5.3b, and hence it can be understood that the formation of pairs of domains is

8/10/2019 Martensit

41/68


Figure 5.4: Results from athermal martensitic transformation in single crystal with

non-clamped boundary conditions, at t*=140 [58].

due to the tendency of the system to minimize its elastic strain energy. It is likelythat in a given direction, a variant combination is preferred in order to maximizethe energy minimization.

The simulation results indicate that the boundary conditions on the grain alsoplay a major role in the martensitic microstructure evolution. A clamped grainwith constrained boundary conditions gives rise to a muti-variant martensitic mi-crostructure, whereas a non-clamped grain with unconstrained boundary conditions

gives rise to formation of a single martensite variant over the entire grain, as shownin Fig. 5.4. In the case of a non-clamped system, the unconstrained boundariesof the simulation domain accommodate the elastic stresses and hence there is nourgent need for the material to form the autocatalysed martensite variants.

Stress-assisted martensitic transformation, Supplement-2

The results indicate that the role of stress states in martensitic microstructureevolution is very crucial. The microstructure evolution is significantly affected dueto the anisotropic loading conditions, as shown in Fig. 5.5. The different anisotropicloading conditions significantly affect the microstructures, as shown in Fig. 5.6.Based on the microstructures obtained under different loading conditions, it can besummarized that the variants that minimize the externally applied strain energyand maximize the net driving force are favored, i.e. only variants with a favorableorientation under a given stress state are formed, viz. Magee effect [2426].

The comparison of the von Mises equivalent plastic strain plots with the vonMises equivalent stress plots indicates that the transformation induced plasticity(TRIP) is an accomodation process to minimize the stresses aroused due to thephase transformation, viz. Greenwood-Johnson effect [2426]. Fig. 5.7 shows astage during the microstructure evolution where the martensite variants have notyet hit the (111) plane, whereas Fig. 5.8 shows a subsequent evolution stage when

8/10/2019 Martensit

42/68


Figure 5.5: Top view of microstructure evolution under anisotropic loading

condition-8c, mentioned in Section 4.2 of Supplement-2, at different time steps(a) t*=10 (b) t*=30 (c) t*=120.

Figure 5.6: Results from the simulations of stress-assisted martensitic transforma-tion in single crystal, under anisotropic loading conditions (8a - 8f) mentioned in

Section 4.2 of Supplement-2. Snapshots at t*=120.

8/10/2019 Martensit

43/68


Figure 5.7: Microstructure, equivalent stress and equivalent plastic strain plotsobtained under uni-axial tensile loading at t*=85. (a) Side view of microstructure.(b) Top view of the microstructure plotted on (111) plane. (c) Equivalent stresscorresponding to the microstructure plotted on (111) plane. (d) Equivalent plasticstrain corresponding to the microstructure plotted on (111) plane.

the martensite variants have hit the (111) plane. It can be seen from Figs. 5.7and 5.8 that as the martensite variants hit the (111) plane from both sides, theequivalent stress inside the plane increases, as can be observed in the correspondingequivalent stress plot. Once the equivalent stress exceeds the yield limit, it intiatesplastic deformation as shown in the corresponding equivalent plastic strain plot,thus indicating the initiation of the TRIP effect.

The simulation results, as shown in Fig. 5.9, indicate that the compressivestresses applied by the martensite on austenite, due to the volume expansion causedby the transformation, leads to stabilization of austenite and thereby gives rise tohigher content of retained austenite in between the martensite variants. From thestress plot it can be seen that the boundaries of the martensite variants are in

8/10/2019 Martensit

44/68


Figure 5.8: Microstructure, equivalent stress and equivalent plastic strain plots

obtained under uni- axial tensile loading at t*=88. (a) Side view of microstructure.(b) Top view of the microstructure plotted on (111) plane. (c) Equivalent stresscorresponding to the microstructure plotted on (111) plane. (d) Equivalent plasticstrain corresponding to the microstructure plotted on (111) plane.

8/10/2019 Martensit

45/68


Figure 5.9: Top view of microstructure and the corresponding equivalent stressplotted on (111) plane, obtained under a tri-axial anisotropic loading condition.(a) Microstructure at t*=110. (b)von Mises equivalent stress of correspondingmicrostructure at t*=110.

dark blue color, indicating that they exert a compressive stress on the surroundingaustenite, which is in accordance with Refs. [23,61].

It has been observed in the simulations that the interaction between differentmartensitic variants could also lead to higher internal stresses in the martensiticinterfaces and thereby gives rise to plastic accommodation, which is in good agree-ment with the results of Marketz and Fischer [62]. By comparing the microstructurewith the stress plot in Fig. 5.10, it can be observed that the interfaces betweendifferent martensitic variants are highly stressed regions and thereby lead to plasticdeformation in the interface regions, as can be seen in the plastic strain plot.

Multi-length scale modeling of martensitic transformations,Supplement-3

The results indicate that a physically based multi-length scale model to simulate themartensitic transformation in practically important materials such as stainless steelsis possible. The ab initio calculations of the elastic properties indicate that stainlesssteels are elastically anisotropic, as shown in Fig. 5.11. The simulations predict theathermal martensite formed in stainless steels to be of lath-type morphology, whichis in good agreement with experimental results [63]. It is observed that, most often,all three martensite variants, i.e. all three Bain variants (groups), form in pairs as

8/10/2019 Martensit

46/68

8/10/2019 Martensit

47/68


Figure 5.11: Calculated characteristic surfaces showing the Youngs modulus as afunction of crystallographic direction in a random FM Fe17%Cr7%Ni alloy. Thevalues on the color scale and on the axes are in GPa.

shown in Fig. 5.12. This result is different compared to the result obtained in thecase of carbon steel, where only two out of the three Bain variants are formed inany given direction, as shown in Fig. 5.3b.

The simulation results of the stress-assisted martensite formation in stainlesssteels under different stress states are similar to those obtained in the case of carbonsteels, as presented in Supplement-2.

Athermal martensitic transformation in polycrystalline material,Supplement-4

The simulation results show that impingement of martensite variants on the grainboundaries gives rise to higher stresses along the grain boundary. Such high stressesare relaxed by martensite nucleation in the adjacent grain, viz. autocatalysis. Theresults are in good agreement with the experimental observations that grain bound-aries are probable sites for martensite nucleation [64], as shown in Fig. 5.13. It isalso observed that the orientation of a martensitic variant is governed by the orien-tation of the grain, which is also in agreement with experimental results [64]. It canbe seen from Fig. 5.13 that the martensite variants, that grow in one grain, hit the

8/10/2019 Martensit

48/68

8/10/2019 Martensit

49/68

8/10/2019 Martensit

50/68


(a) (b)

Figure 5.15: Results from simulations with (a) low plastic relaxation rate (b) highplastic relaxation rate [65].

A lower plastic relaxation rate results in a microstructure containing of manymartensite variants inside a single grain as shown in Fig. 5.15a. A higher plasticrelaxation rate results in a microstructure consisiting of a single martensite domaingrowing across the entire grain as shown in Fig. 5.15b [65]. The above results arein good agreement with the findings of Levitas et al. [35].

The plastic relaxation ratekalso affects the growth rate of martensite, as shownin Fig. 5.16. Each sub-figure corresponds to snap shots, at t = 50, from differentsimulations with different k. It can be seen that under given conditions, Viz.driving force and dimensionless time (t = 50), the growth rate of a martensitedomain increases with increasing k. Ask is directly proportional to the dislocationmobility, as explained in Section 4.2, it can be understood that a higher dislocationmobility gives rise to a higher growth rate of the domain [65]. This is in goodagreement with earlier observations that lath martensite grows with the aid ofdislocation glide [33,34].

The von Mises equivalent plastic strain plot, as shown in Fig. 5.17, indicates

that the major plastic deformation occurs in the martensite compared to that ofaustenite [65]. As a higher plastic strain implies a higher dislocation density, itcan be understood that martensite is a dislocation-rich structure, which is in goodagreement with the experimental observations of Morito et al. [27].

The plastic relaxation rate also affects the transformation by affecting the em-bryo potency, as explained in the following section.

5.3 Nucleation and growth aspects, Supplement-5

The effect of plastic relaxation rate on the embryo potency is shown in Fig. 5.18.The results indicate that for a given driving force there exists a critical plastic

8/10/2019 Martensit

51/68

5.3. NUCLEATION AND GROWTH ASPECTS 39

Figure 5.16: Results, at t = 50 and mapped on (0 1 0) plane, from differentsimulations with different k values, as mentioned below each sub-figure [65].

relaxation rate kcrit, below which the embryo is impotent, i.e. reverse transforma-tion from martensite to austenite. It can also be seen from Fig. 5.18 that as theplastic relaxation rate is higher than the critical value, the embryo becomes potentand grows. As the plastic relaxation rate is directly proportional to the disloca-tion density, it can be construed that in order for the embryo to become potent, acritical dislocation density is needed at the nucleation site [66]. The dependence ofembryo potency on the dislocation density is in good agreement with earlier reportby Cohen [28].

The simulations performed to study the effect of embryo potency on the trans-formation, as mentioned in Section 4.3, indicate that the size, shape and orientationof the martensite embryo affect the transformation. It becomes very difficult for asmall embryo to sustain the very high elastic stresses generated inside the materialand hence it shrinks and disappears completely. If the embryo is considerably bigsuch that it can sustain the elastic stresses, it becomes potent and starts to grow.However the comparison of the critical size of the nucleus obtained from the sim-

8/10/2019 Martensit

52/68

8/10/2019 Martensit

53/68

5.4. THERMODYNAMIC ASPECTS 41

(a) (b) (c)

Figure 5.19: Iso-surfaces of phase-field variable (= 0.5) obtained at t* = 80 fromthree different simulations started with a pre-existing ellipsoidal embryo whosemajor axis is aligned along (a) X-axis (b) Y-axis (c) Z-axis [66]. t* denotes dimen-sionless time.

ulations with that of the size determined from earlier theoretical works [31] stillneeds an in-depth analysis.

The simulations predict the favorable shape of the embryo as an ellipsoidalshape, as can be seen in Fig. 5.18c, which minimizes the combined interfacial andelastic strain energy of the embryo [31]. The results also indicate that a favorably

oriented embryo gives rise to a higher martensite volume fraction. The results ofthree different simulations started with an ellipsoidal martensite embryo orientedin different directions is shown in Fig. 5.19.

5.4 Thermodynamic aspects, Supplement-6

As mentioned in Section 4.4, the procedures to determine the different driving forcesrequired for growth and autocatalysis are shown in Figs. 5.20 and 5.21 respectively.The results indicate that there exist three different critical (minimum) drivingforces, Viz. driving force required for nucleation (Gnucl.), driving force requiredfor growth (Ggrowth) and driving force required for autocatalysis (Gautocat.),which govern the martensitic transformation as well as the microstructure evolu-tion. The simulations predict that Ggrowth

8/10/2019 Martensit

54/68


Figure 5.20: Illustration of the approach to predict the critical driving force forgrowth [65].

Figure 5.21: Illustration of the approach to predict the critical driving force forautocatalysis [65].

8/10/2019 Martensit

55/68

5.5. CRYSTALLOGRAPHY 43

Figure 5.22: Prediction of initial habit plane [65].

5.5 Crystallography, Supplement-6

The model predicts the initial habit plane of the infinitesimally small martensiteunit to be (111) as shown in Fig. 5.22. In order to minimize the Gibbs energy ofthe system, this martensite domain rotates away from the initial habit plane as thetransformation progresses, as shown in Fig. 5.23. On the formation of self nucleatedmartensite units, the rate of rotation decreases as the energy minimization is nowachieved by means of autocatalysis. Thus the model predicts the final habit plane ofthe initially formed martensite domain, mentioned above, to be (211) [65]. Theresults are in agreement with the theoretical explanation that martensite variantstend to rotate in order to minimize the strain energy [5,9].

The rotations of martensite variants are studied in order to verify the theo-retical explanation that dilatation coupled with shear and rotation could give riseto the experimentally predicted habit planes [9], as explained in Section 2.6. Thedilatation is governed by the Bain strains, that are supplied as input parametersfor the simulations. The shear deformation is governed by the plasticity equation(4.13). The rotation is governed by the minimization of the strain energy thatyields displacement, which in turn gives rise to the rotation of the variants [5,9].

Fig. 5.24 shows a 2D section of the 3D athermal martensitic microstructure aswell as the corresponding displacement plot. In the displacement plot, red colorindicates maximum displacement and dark blue color indicates zero displacement.By comparing the two images, it can be seen that the dark blue regions in thedisplacement plot are the centers of the martensitic domains shown in the corre-sponding microstructure image. This shows that the centers of the martensiticdomains are the regions where displacement is zero. Thus the habit plane, whereinfinitesimally small unit of martensite forms, is undistorted and unrotated as the

8/10/2019 Martensit

56/68


Figure 5.23: Simulation results showing the martensite domain rotating away fromthe initial habit plane of (-1 1 1) [65]. (a) Initial habit plane att = 7 (b) Rotationat t = 45 (c) Rotation at t = 150.

Figure 5.24: Illustration to explain the invariant habit plane concept. (a) 2D sectionof the 3D athermal martensitic microstructure (b) Displacement plot of the corre-sponding microstructure shown in (a). Red color indicates maximum displacementand dark blue color indicates zero displacement.

displacement is zero. This is in agreement with the phenomenological theories ofmartensite [23].

Both the phase-field method and the Microelasticity theory [5] are based onthe continuum principles, i.e. principles that are used to describe a physical phe-nomenon by using mathematical functions that are continuous functions in spaceand time. Thus the phase-field variables and the displacement field, obtained bythe minimization of the elastic strain energy by using the Microelasticity theoryas explained in Eqs. (4.5) - (4.9) in Section 4.1, are continuous functions of thecoordinates in the lattice space. This ensures the crystal lattice continuity, i.e.

8/10/2019 Martensit

57/68

8/10/2019 Martensit

58/68

8/10/2019 Martensit

59/68

Chapter 6

Concluding remarks and future

prospects

The results obtained in this work indicate that it is possible to model and to studyseveral aspects of the martensitic transformation using the phase-field method in asatisfactory way. Although the nucleation is not simulated in this work, the sim-ulations performed to study the effect of embryo potency on the transformationindicate that the size, shape and orientation of the embryo as well as the dislo-cation density at the nucleation site govern the martensitic transformation. The

simulations predict that martensite, once nucleated, could grow as well as lead toautocatalysis, above Ms. Although this result is in agreement with experimentalresults [40], the experiments indicate that martensite could only grow up to a tem-perature that is slightly aboveMs, whereas the simulations predict that martensitecould grow well above Ms and hence needs to be studied in detail by performingmore experiments as well as simulations.

The experimental observations that the martensitic microstructures consist ofseveral twinned martensite variants are observed in the simulated microstructuresas well. Autocatalysis, i.e. nucleation of several martensite variants, plastic defor-mation as well as the non-clamped boundary conditions act as relaxation measures

of the stresses developed, due to the martensitic transformation, inside the mate-rial. The microstructures obtained in a polycrystalline material indicate that themartensite variants growing in one grain, when impinge up on the adjacent grains,could initiate martensite nucleation in the adjacent grains, where the grain bound-aries act as nucleation sites. The study of plastic deformation aspects show thatthe martensitic microstructure is significantly affected by the plastic relaxation rate,which corresponds to the dislocation mobility and density.

The model predicts the initial habit plane of martensite in agreement with theexperimental results and predicts that the variants rotate in order to minimize thestrain energy. If the rotations of various variants could also be plotted along with

the Bain variants shown in the present figures, it might be possible to predict some

47

8/10/2019 Martensit

60/68

8/10/2019 Martensit

61/68

8/10/2019 Martensit

62/68

50 CHAPTER 6. CONCLUDING REMARKS AND FUTURE PROSPECTS

Figure 6.3: Variation of mean equivalent plastic strain with martensite volumefraction obtained under applied uni-axial tensile stress.

Thus with a tone of cautious optimism, it can be concluded by saying thatphase-field method can be a useful qualitative tool in understanding several aspectsassociated with the martensitic transformations, provided the underlying physics istaken care of in an appropriate manner.

8/10/2019 Martensit

63/68

Acknowledgements

First of all I would like to express my sincere gratitude to my principal supervisorProf. John gren for giving me this opportunity to work with one of the intriguing

and exciting research topics in the area of Physical Metallurgy, for his supervisionand for all the interesting discussions. I would also like to thank my anothersupervisor Assoc. Prof. Annika Borgenstam for the critical assessment of my workthat has greatly added value to the work. I want to express my thanks to all theother collaborators, Prof. Gustav Amberg, Mr. Amer Malik, Dr. Peter Hedstrm,Dr. Vsevolod Razumovskiy, Dr. Pavel Korzhavyi and Prof. Andrei Ruban. Myspecial thanks to Prof. Emer. Mats Hillert for the interesting discussions and ideas.Thanks to Dr. Lars Hglund and Dr. Minh Do Quang for all the help with thecomputer related aspects. I would also like to thank Prof. Hasse Fredriksson, whomI always consider as my first teacher in the field of Materials research. Thanks to

Prof. Staffan Hertzman for the motivation to work on stainless steels. My thanksto all the other colleagues at MSE that have made all these years pleasant andenjoyable.

I would like to thank the VINN Excellence Center Hero-m, Vinnova, SwedishIndustry and KTH Royal Institute of Technology for providing finance for perform-ing this work. Thanks to SNIC for providing the computer resources as well asto the STT foundation, Wallenberg foundation and Jernkontoret for providing thetravel grants to attend various conferences.

I would also like to thank Prof. A.G. Khachaturyan, Dr. Avadh Saxena, Dr.Turab Lookman, Dr. Alfonse Finel, Prof. Greg Olson, Prof. S.Q. Shi and Prof.

Kaushik Bhattacharya for all the discussions on martensite and phase-field mod-eling, during different conferences, which have greatly helped me in understandingthis subject.

Finally, I would like to express my deepest gratitude to the two most wonderfulpersons in my life, my mother and aunt, for all their support and encouragementin all the phases of my life. Thanks to my best friend Sudheer for his support andfriendship. Thanks to my friends, Sathees, Kavitha and Mitra for their supportand making my life in Sweden quite pleasant. A special thanks to Reshu for themotivation to complete this work on the martensitic transformation in a martensiticmanner.

51

8/10/2019 Martensit

64/68

8/10/2019 Martensit

65/68

Bibliography

[1] P.C. Clapp, Journal De Physique IV 5 (1995) C8-11.

[2] W.J. Boettinger, J.A. Warren, C. Beckermann, A. Karma, Annual Review ofMaterials Research 23 (2002) 163.

[3] L.Q. Chen, Annual Review of Materials Research 32 (2002) 113.

[4] N. Moelans, B. Blanpain, P. Wollants, CALPHAD 32 (2008) 268.

[5] A.G. Khachaturyan, Theory of Structural Transformations in Solids, DoverPublications Inc., New York (2008).

[6] Y. Wang, A.G. Khachaturyan, Acta Materialia 45 (1997) 759.

[7] A. Saxena, A.R. Bishop, S.R. Shenoy, T. Lookman, Computational MaterialsScience 10 (1998) 16.

[8] A. Artemev, Y. Wang, A.G. Khachaturyan, Acta Materialia 48 (2000) 2503.

[9] A. Artemev, Y. Jin, A.G. Khachaturyan, Acta Materialia 49 (2001) 1165.

[10] V.I. Levitas, D.L. Preston, Physical Review B 66 (2002) 134206.

[11] K. Bhattacharya, Microstructure of Martensite - How it forms and how itgives rise to the shape-memory effect, Oxford University Press Inc., New York(2003).

[12] Y. Wang, A.G. Khachaturyan, Materials Science and Engineering A 438-440(2006) 55.

[13] A. Yamanaka, T. Takaki, Y. Tomita, Materials Science and Engineering A 491(2008) 378.

[14] A. Yamanaka, T. Takaki, Y. Tomita, International Journal of Mechanical Sci-ences 52 (2010) 245.

[15] G. Krauss, A.R. Marder, Metallurgical Transactions 2 (1971) 2343.

53

8/10/2019 Martensit

66/68

54 BIBLIOGRAPHY

[16] S. Morito, X. Huang, T. Furuhara, T. Maki, N. Hansen, Acta Materialia 54(2006) 5323.

[17] G. Krauss, Materials Science and Engineering A 273-275 (1999) 40.

[18] S. Morito, H. Tanaka, R. Konishi, T. Furuhara, T. Maki, Acta Materialia 51(2003) 1789.

[19] S. Morito, H. Yoshida, T. Maki, X. Huang, Materials Science and EngineeringA 438-440 (2006) 237.

[20] A. Shibata, S. Morito, T. Furuhara, T. Maki, Scripta Materialia 53 (2005) 597.

[21] A. Shibata, S. Morito, T. Furuhara, T. Maki, Acta Materialia 57 (2009) 483.

[22] P. Kolmskog, Work performed as part of his doctoral thesis, To be published.

[23] Z. Nishiyama, Martensitic Transformation, Academic Press, New York (1978).

[24] M. Cherkaoui, M. Berveiller, X. Lemoine, International Journal of Plasticity16 (2000) 1215.

[25] F.D. Fischer, G. Reisner, E. Werner, K. Tanaka, G. Cailletaud, T. Antretter,International Journal of Plasticity 16 (2000) 723.

[26] R.F. Kubler, M. Berveiller, P. Buessler, International Journal of Plasticity 27(2011) 299.

[27] S. Morito, J. Nishikawa, T. Maki, ISIJ International 43 (2003) 1475.

[28] M. Cohen, Proceedings of International Conference on Martensitic Transfor-mations (1986) 16.

[29] M. Suezawa, H.E. Cook, Acta Metallurgica 28 (1980) 423.

[30] J.W. Brooks, M.H. Loretto, R.E. Smallman, Acta Metallurgica 27 (1979) 1839.

[31] M. Cohen, Transactions Metallurgical society AIME (1958) 171.[32] G.B. Olson, M. Cohen, Journal of Less-Common Metals 28 (1972) 107.

[33] K.E. Easterling, A.R. Thlen, Acta Metallurgica 28 (1980) 1229.

[34] G. Thomas, B.V.N. Rao, Proceedings of Conference on Martensitic Transfor-mations, Kiev (1978) 57.

[35] V.I. Levitas, A.V. Idesman, G.B. Olson, E. Stein, Philosophical Magazine A82 (2002) 429.

[36] E.S. Machlin, M. Cohen, Transactions AIME 191 (1951) 746.

8/10/2019 Martensit

67/68

8/10/2019 Martensit

68/68

56 BIBLIOGRAPHY

[58] H.K. Yeddu, A. Malik, J. gren, G. Amberg, A. Borgenstam, Acta Materialia60 (2012) 1538.

[59] D. Hull, D.J. Bacon, Introduction to Dislocations, fifth ed., Elsevier Ltd.,Oxford (2011).

[60] J.O. Andersson, T. Helander, L. Hglund, P.F. Shi, B. Sundman, CALPHAD26 (2002) 273.

[61] I. Karaman, H. Sehitoglu, H.J. Maier, M. Balzer, Metallurgical and MaterialsTransactions A 29A (1998) 427.

[62] F. Marketz, F.D. Fischer, Modelling and Simulation in Materials Science andEngineering 2 (1994) 1017.

[63] B.P.J. Sandvik, H.O. Martikainen, V.K. Lindroos, Scripta Metallurgica 18(1984) 81.

[64] R. Naraghi, P. Hedstrm, A. Borgenstam, Steel Research International 82(2011) 337.

[65] H.K. Yeddu, A. Borgenstam, P. Hedstrm, J. gren, Materials Science andEngineering A 538 (2012) 173.

[66] H.K. Yeddu, A. Borgenstam, J. gren, Journal of Alloys and Compounds, In

Press (2012), doi:10.1016/j.jallcom.2012.01.087.[67] S. Morito, Y. Edamatsu, K. Ichinotani, T. Ohba, T. Hayashi, Y. Adachi, T.

Furuhara, G. Miyamoto, N. Takayama, Journal of Alloys and Compounds, InPress (2012), doi:10.1016/j.jallcom.2012.02.004.

Documents

Martensit