Surface & Coatings Technology 203 (2009) 2003–2012

Contents lists available at ScienceDirect

Surface & Coatings Technology

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / s u r f c o a t

Predictive modeling and experimental results for residual stresses in laser hardening
of AISI 4140 steel by a high power diode laser
Neil S. Bailey, Wenda Tan, Yung C. Shin ⁎
Center for Laser-based Manufacturing, Purdue University, West Lafayette Indiana, 47907, USA

a r t i c l e i n f o a b s t r a c t

Article history: A predictive model for residual stresses induced in a laser hardened workpiece of AISI 4140 steel with no
Received 11 November 2008 melting has been developed and experimentally verified. A transient three-dimensional thermal and kinetic
Accepted in revised form 23 January 2009 model is first solved to obtain the temperature and solid phase history of the workpiece, which is then
Available online 7 February 2009
sequentially coupled to a three-dimensional stress model to predict residual stresses. The phase
transformation strains are added to the thermal strains at each time step during the heating and cooling
Keywords:
Laser hardening
cycles to obtain the resultant residual stresses in the workpiece. The importance of considering phase
Predictive modeling transformation has been explained through the comparison of the magnitudes of residual stresses with and
Phase transformation without the inclusion of phase transformation kinetics. The model predicted strong compressive residual
Residual stress stresses of about 200 MPa in the heat affected zone due to austenite-to-martensite transformation. The
4140 steel predictions matched well with the X-ray diffraction measurements.
© 2009 Elsevier B.V. All rights reserved.

1. Introduction exclusively to martensite. At lower cooling rates, the austenite will

transform to both martensite and bainite.
Laser hardening (LH) of steels is a surface enhancement process, Kou et al. [4] developed a three-dimensional finite difference heat
similar to flame and induction hardening, but with a few distinct flow model to obtain the temperature history in laser-hardened AISI
advantages. A high level of laser energy can be very precisely delivered 1018 steel. Phase transformation of pearlite and ferrite to austenite were
to a small area of the workpiece [1]. Since the heated region is so small, discussed with the help of calculated peak temperatures. They predicted
the bulk of the workpiece acts as a heat sink, causing rapid quenching. complete transformation of austenite to martensite based on the
In this way, the heat affected zone will become almost fully calculated cooling rates. However, the microstructure near the top
martensitic, producing a surface that can better withstand wear and surface of the workpiece that showed a small amount of ferrite made
corrosive environments while minimizing distortion. The phase them believe that the time allowed for carbon atoms to diffuse in
transformation will also produce high compressive stresses, giving austenite was still not sufficient. Ashby and Easterling [5], using a two-
the workpiece better fatigue strength as well [2,3]. dimensional analytical solution for the heat flow, proposed that the
Many papers [4–23] have presented the theory of laser transforma- structural changes during heating are diffusion controlled, and the
tion hardening in hypo-eutectoid steels. Normalized steel consists of extent of the change would depend on the total number of diffusive
pearlite (P) and proeutectoid ferrite (α). As shown in Fig.1 [10], when the jumps in the cycle, measured in terms of the “kinetic strength” of the
material is locally heated by the laser to a temperature above the eutectic heat cycle. Austenite to martensite transformation during cooling was
temperature, A1 (727 °C), the pearlite colonies begin to transform to assumed to be independent of diffusion, but calculated by the quench
austenite (γ), while the proeutectoid ferrite remains unaffected. rate of the cycle. They proposed that the quench rates for all heating
However, as the temperature increases further, carbon from newly cycles in the process were sufficient to convert all austenite with carbon
formed austenite diffuses into the low carbon ferrite. If kept above A1 content greater than 0.05 wt.% to martensite. Davis et al. [6] considered
temperature for a sufficient time the carbon distribution will become complete austenization in their model by estimating the time above A3
homogeneous. If the austenization temperature, A3, is exceeded, then temperature that the workpiece must remain for carbon diffusion to take
any remaining ferrite will transform to austenite [4–6,9–10]. High place. Ohmura et al. [7], in addition to numerically solving the heat flow,
heating and cooling rates (as high as ~103 K/s) have been reported to be also included carbon diffusion in austenite and solved Fick's Law of
typical with the laser hardening process. This is due to the conduction of diffusion to obtain carbon concentration. During cooling, they assumed
heat from the localized heated region into the bulk material [9]. If the that if the cooling rate is higher than the critical cooling rate, the aus-
subsequent cooling is fast enough, austenite is transformed almost tenite transforms to martensite starting at Tm-start (martensite start
temperature). If the cooling rate is lower than the critical cooling rate, the
⁎ Corresponding author. Tel.: +1 765 494 9775. austenite is transformed to fine pearlite. Jacot and Rappaz [8] developed
E-mail address: shin@purdue.edu (Y.C. Shin). a two-dimensional model for the heating cycle in laser hardening of

0257-8972/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2009.01.039
2004 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012

stress state by causing dilatation strain, represented by arrow 5.

Reverse effects are also possible, although usually of a much smaller
magnitude. Large strains can change the energy in the material,
slightly altering the temperature. This is illustrated by arrow 2. Latent
heat from solid phase transformations can also alter the temperature
field, represented by arrow 4. Large strains can also assist in phase
transformations, called strain-induced transformation, represented by
arrow 6. And finally, chemical concentrations, such as carbon content
in low carbon steels, will be affected by, and have an effect on all three
of these phenomena. Such extensive coupling of all these complex
processes creates difficulties in numerical modeling of laser hardening
of steels; essentially, all of these phenomena ought to be solved
simultaneously if all six coupling effects and carbon concentration are
to be considered.
Peng and Ericsson [2] and Solina et al. [26] have reported the effect of
phase transformation on residual stress states. Both have shown that
Fig. 1. Microstructural changes in hypo-eutectoid steel during heating (adopted from [10]). thermal residual stresses in laser hardening treatments are tensile and
phase transformation-induced residual stresses (austenite to marten-
site) are compressive. The generation of residual stress depends not only
hypo-eutectoid steels that combined all the steps, from dissolution of on the temperature field to which it has been subjected but also on the
pearlite, the transformation of ferrite into austenite and carbon microstructural changes that have taken place. Lacarac et al. [27]
homogenization to the grain growth of austenite. Their calculation for compared residual stresses in steels with and without phase transfor-
carbon diffusion starts at eutectoid temperature, and stops when the mations. They developed a non-dimensional finite element approach for
domain is entirely austenite and the distribution of carbon satisfies a long bars of infinite length, and determined the effect of non-
given homogenization criterion. Komanduri and Hou [11,12] used a dimensional parameters and non-dimensional metallurgical time, on
thermal analysis based on a moving heat source method to predict laser residual stresses and microstructure. As per their analysis, uniform
hardening of AISI 1036 steel. Hardened regions were determined by peak microstructures result in lower residual stresses when compared to
temperatures and temperature interaction times. Costa et al. [13] non-uniform microstructures. In laser hardening of medium carbon
developed a method that uses finite element software ABAQUS to steels, Yang and Na [28] emphasized the large influence of transforma-
calculate heat transfer and phase transformation in a laser powder tion-induced plasticity on the residual stress distribution. In their finite
deposition system. After solidification and cooling, the volume fraction element model, the volume change of austenite and martensite
of martensite is assumed to be 100%. Upon subsequent heating from transformation was considered by thermal dilatation, for which the
additional powder tracks, martensite tempering was calculated based on equivalent linear thermal expansion coefficients were adopted. In their
peak temperature. Yanez et al. [14] and Tobar et al. [15] studied laser article, they mentioned transformation-induced plasticity in both uni-
transformation hardening of steel using an analytical solution for a semi- axial and multi-axial stress states. However, it is not very clear how they
infinite media followed by a finite element analysis through ANSYS. The implemented the multi-axial transformation-induced plasticity strain
microstructural changes were calculated based on the Johnson–Mehl– rate equation in their numerical scheme. Transformation-induced
Avrami and Koistinen–Marburger equations. Studies by Miokovic et al. plasticity was shown to significantly influence the residual stresses,
[16,17] presented a systematic analysis on the effects of heating and but its dependence on the yield stress of austenite, phase fraction of
cooling rates on the resultant phase formation using a pin type structure austenite, and the volumetric dilatation was not mentioned. Grigoryants
with laser heating on the top. A coupled heat conduction and phase et al. [29] used the X-ray diffraction method to measure residual stresses
transformation model with temperature dependent parameters was in pure iron and steel samples treated with a continuous CO2 laser
used to calculate the resultant phase formation and hardness. Leung et al. irradiation with varying power. Compressive stresses were observed
[18,19] developed an exact temperature solution to a 2D quasi-steady when the material was taken through phase transformation to
state laser hardening situation with a custom beam profile. The phase martensite, but tensile stresses were observed wherever significant
transformation was predicted by a constant phase-change temperature. residual austenite was found. Using the commercial finite element
Lusquinos et al. [20] used ANSYS to predict temperature cycles in laser package SYSWELD, Zhan et al. [30] predicted residual stresses in a melt-
hardening of AISI 1045 and predicted the case depth based on peak hardened workpiece, wherein microstructural changes were based on
temperature. A recent model by Tani et al. [21] predicts the 2D phase the Johnson–Mehl–Avrami equations. It is however unclear if stress
transformation and hardness profile after laser hardening using a finite calculations included phase transformation induced strains. Wang et al.
difference method to calculate both the heat transfer and carbon [31] also used SYSWELD to predict microstructure and residual stresses
diffusion. A digitized photomicrograph of the steel was used to simulate in a 3D powder deposition system with stainless steel. The model used to
an initial microstructure. When this model was extended to 3D [22] the
carbon homogenization and initial microstructure were neglected and
an Avrami kinetic approach is used to allow for faster computation.
Another expansion of the model [23] included the effect of steel
softening due to tempering phenomena.
Temperature field, stress state, and microstructure are all strongly
coupled together, each affecting the other. These effects are part of a
phenomenon that has been termed metallo–thermo–mechanical
coupling [24,25], schematically illustrated in Fig. 2. As explained by
the authors [24,25], temperature fields affect the stress state through
thermal strains caused by temperature gradients. This phenomenon is
illustrated by arrow 1 in Fig. 2. Temperature fields also affect
microstructural changes according to phase transformation kinetics,
represented by arrow 3. Microstructural changes, in turn, affect the Fig. 2. Metallo–thermo–mechanical coupling in processes involving phase transformation.
 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012 2005

calculate metallurgical strains is not presented in detail. Şimşir and Gür

[32] used commercial finite element software MSC.Marc to determine
residual stresses considering thermal and phase transformational
strains applied to quenching of a steel cylinder. The phase transforma-
tions were calculated using the Johnson–Mehl–Avrami–Kolmogorov
kinematic model. The mechanical model considered thermal strains,
volumetric dilatational strains due to phase transformations and
transformation-induced plasticity strains.
This paper presents a method for predicting residual stresses in a laser
hardened workpiece, considering thermal strain and the phase transfor-
mation effects of volumetric dilatation strain and transformation-induced
plasticity strain. Using the finite volume method (FVM), temperature and
microstructure are simultaneously solved in a transient 3D thermal/
kinetic model based on the work by Skvarenina and Shin [9] and Lakhkar
et al. [33] and further improved to include the effects of latent heat of
phase transformation (arrow 4 in Fig. 2). The resultant microstructure is
then used to solve for the final 3D hardness field, and the 3D temperature
and microstructure histories are then used in ABAQUS, a commercial Fig. 3. A 2D slice from the 3D digital microstructure depicting phase and interface cells
during heating (adopted from [9]).
finite element code, to generate the residual stress state.

2. Mathematical model
plicit finite volume scheme with a much finer spatial mesh (5 µm)
2.1. Temperature distribution and smaller time step (0.01 s) than those of the thermal model.
Since the thermal model utilizes a coarser mesh and a larger time
The transient temperature distribution in a prismatic workpiece step, the temperature and phase fractions are linearly interpolated
can be obtained by solving the three-dimensional conservation of in both time and space between the two models. For a more
energy equation for a solid. detailed presentation of the kinetic model, the reader is referred to
reference [9] by Skvarenina and Shin, but a brief explanation is
dT
ρCp − jðkjT Þ = Q r + Q L ð1Þ given below.
dt
The initial microstructure of the AISI 4140 steel consists of about
Q r represents the radiation energy from the laser and Q L represents 45% pearlite. In order to track kinetic transformations, a program
the latent heat of phase transformation. The present model solves this has been written, which takes randomly placed cells in 3D matrix
equation using an implicit finite volume scheme where specific heat and “grows” them into a digital microstructure of pearlite colonies
and conductivity are functions of temperature. Natural convection and in a ferrite volume. The program controls the average grain size,
radiation boundary conditions are appropriately employed on work- spacing of the grains, and the final pearlite mass fraction to agree
piece surfaces and are shown in Eq. (2). with published values (see Table 1). An example of a 2D slice of a
digital microstructure undergoing phase change during heating is


kðjT  nÞ = hn ðT − T∞ Þ and kðjT  nÞ = et σ T 4 − T∞4 ð2Þ shown in Fig. 3 [9]. Each grid point of the initial microstructure
is assigned a phase, either ferrite (α), pearlite (P) or ferrite/pearlite
The details of this thermal model are described in Rozzi et al., [34] interface (α/P). This 3D digital microstructure is then used by the
and Tian et al. [35], but it has been expanded to include complex kinetic model to keep track of phase transformations during pearlite
three-dimensional geometric features. dissolution, homogenization of austenite and cooling. During heating
The following conditions should also be satisfied: above the A1 eutectoid temperature, pearlite and ferrite/pearlite
interface transform to austenite (γ) and ferrite/austenite interface (α/
T ðx; y; z; tY∞ÞYT∞ ð3Þ
γ), respectively. During homogenization of austenite, ferrite/austenite
interface cells transform into austenite cells and neighboring ferrite
As time approaches infinity, the temperature in the workpiece
cells become interface cells. During cooling, austenite cells will trans-
should approach the ambient temperature, i.e., the workpiece should
form to martensite cells or bainite cells, depending on the criteria
cool to room temperature after the laser passes. The residual stress
explained below.
state is only valid in a workpiece that has cooled to room temperature.
The homogenization of austenite is assumed to be governed by
solute diffusion. Hence, Fick's 2nd law of diffusion is utilized and is
2.2. Phase transformation kinetics
solved using the finite volume method. This calculation is performed
solely over the ferrite, austenite and ferrite/austenite interface cells
Phase transformation kinetics is solved simultaneously with the
and starts and ends at the eutectoid temperature.
energy equation at each time step. The kinetic model uses an ex-
     
ACv A ACv A ACv A ACv
= Dv + Dv + Dv : ð4Þ
At Ax Ax Ay Ay Az Az
Table 1
Microstructural properties for AISI 4140 steel [33]
Insulated boundary conditions are imposed on the edges of the
Property Value
kinetic model domain, i.e., carbon cannot diffuse across a boundary.
Pre-exponential carbon in ferrite, Doα 6 × 10− 5 m2/s
The initial condition for the carbon concentration field is calculated
Pre-exponential carbon in austenite, Doγ 1 × 10− 5 m2/s
Activation energy ferrite, Q α 80 kJ/mol across the initial microstructure using the average carbon concentra-
Activation energy austenite, Qγ 135 kJ/mol tions of each phase, ferrite and pearlite.
Universal gas constant, R 8.314 J/mol K The ferrite/austenite interface cells have three additional descrip-
Initial volume fraction of pearlite, f 0.45 tors, Ciα⁎, Ciγ⁎, and fiα, representing, respectively, the carbon concen-
Critical carbon content value, Cc 0.05%
trations of ferrite or austenite at the interface (given by the phase
2006 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012

diagram) while fiα is the volume fraction of ferrite, found by using the is compared with the amount of time required to begin bainite
lever rule shown in Equation (5). formation in an isothermal process. Sheil's Additivity Rule discritizes
the process over a number of time steps, allowing for non-isothermal
γ⁎
Ci − Ci calculations. If the integral in Eq. (8) reaches a value of 1 before the
fiα = γ⁎
ð5Þ
Ci − Ciα⁎ temperature drops below the martensite formation temperature,
bainite will begin to form.
When fiα reaches zero, the interface cell is converted to austenite
and adjacent ferrite cells are transformed to interface cells. The mass t dt
∫0 z1 ð8Þ
balance for any cell i is given by, ta ðT Þ

ΔCi 6 Here dt is the length of a time-step, ta(T) (obtained from the TTT
V = ∑ SJki ð6Þ diagram in Fig. 4) is the incubation time required isothermally at
Δt k=1
temperature T for bainite to form, and t is the current cooling time of
where V = h3, S = h2, and h is the edge length of the 3D cell. In the an austenite cell in the model. Under the parameters used in this
present model, h is 5 μm. study, bainite is not formed because the cooling rate is sufficiently fast
The diffusion coefficient for carbon in austenite and ferrite is an to avoid satisfying Eq. (8), thereby avoiding formation of bainite. If
Arrhenius relation as shown in Eq. (7). Eq. (8) is satisfied under conditions other than those used in this study,
  the JMA model is used in the present thermal/kinetic model to
Qν calculate the amount of bainite that would be formed.
Dν ðT Þ = D0ν exp − ð7Þ
RT ðt Þ Once the temperature has dropped below the martensite forma-
tion temperature, Tm-start, the KM equation is used to calculate the
Ashby and Easterling [5] provide the specific coefficients for both volume fraction of martensite in the cooling process.
phases, which are listed in Table 1, with other relevant microstructural
properties of AISI 4140 Steel. fm = fγ⁎ f1 − exp½− 0:011ðTm−start − T Þg ð9Þ
Earlier references [5,6,9,10] assume that after the thermal cycle
and associated austenite homogenization has finished, all the aus- Here, Tm-start (obtained from the TTT diagram) is the temperature
tenite with carbon content greater than 0.05% transforms to mar- at which martensite begins to form, and fγ⁎ is the volume fraction of
tensite, regardless of cooling rate. However, the phase transformation austenite at the time when temperature reaches Tm-start. If the cooling
model presented here has been expanded to consider the effects of rate is sufficiently fast, i.e., no bainite is formed, all austenite that does
cooling rate on phase transformation. As the workpiece begins to cool, not change into martensite will be considered retained austenite.
austenite is transformed to martensite if the cooling rate is fast The cooling cycle uses a much coarser mesh than the heating cycle
enough or bainite if the cooling rate is slower. The phase transforma- (h = 50 µm). Every 1000 cells are grouped together in cubes of 10 × 10 × 10
tion in the cooling cycle of the laser hardening process is based on cell volumes and the average phase fraction of each cube is calculated.
Sheil's Additivity Rule, the Johnson–Mehl–Avrami (JMA) Model, and Cooling cycle calculations are then performed over each cube and the
the Koistinen–Marburger (KM) equation [36,37]. phase fraction of each cube adjusts accordingly. Temperature and phase
In order to determine if bainite formation will occur, the current are recorded at each time step. The finite element residual stress model
cooling time of a material undergoing non-isothermal transformation will use this information to calculate residual stress. Once the workpiece

Fig. 4. Time temperature transformation (TTT) diagram for AISI 4140 steel (from [38]).
 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012 2007

Table 2 and σ—(ε,T) is the temperature-dependent flow stress of the material.

Stress-strain properties by phase for AISI 4140 steel [40–43] For simplicity, plastic and elastic strains are assumed to be
Property Value independent of material phase. Thermal strains and phase transfor-
Thermal expansion coefficient of ferrite, α α 1.61 × 10− 5 1/°C mational strains, however, are strongly dependant on phase. Thermal
Thermal expansion coefficient of pearlite, α P 1.53 × 10− 5 1/°C strains exist in a workpiece wherever thermal gradients exist. The
Thermal expansion coefficient of austenite, α γ 2.20 × 10− 5 1/°C thermal strain increment can be calculated by
Thermal expansion coefficient of martensite, α m 1.15 × 10− 5 1/°C
Volume change: pearlite to austenite, ΔV/V − 0.33%
ΔeTij = α ðΔT Þδij ð18Þ
Volume change: ferrite to austenite, ΔV/V − 0.38%
Volume change: austenite to martensite, ΔV/V 1.026%
α = ∑α i fi ð19Þ

where αi is the coefficient of thermal expansion for phase i, and fi is

in the model has cooled to room temperature and phase transformations the phase fraction of phase i. When a given volume in a workpiece
are complete, the final phase fraction field is used to calculate hardness. undergoes transformation from one phase to another, the micro-
After the cooling cycle, the hardness value of a cube can be calcu- structural pattern transforms. For example, austenite changing to
lated according to the phase fractions in the cube by martensite will transform from a face-centered cubic body structure
to a body-centered cubic structure, thereby dilating the volume. The
H = ∑ fν H ν ð10Þ
strain increment due to this volumetric dilatation is given by
where fν is the fraction of phase ν in the cube, and Hν is the hardness
1 ΔV
of phase ν. The hardness of martensite, Hm, is given by Eq. (11) ΔeΔV
ij = Δ Xδij ð20Þ
3 V
c2 where ΔV/V is the percent volume change due to phase transforma-
Hm = 1667c − 926 + 150 ð11Þ
fm tion from one phase to another and ΔX is the change in phase fraction
during the time increment. According to Das et al. [40] strains are also
where c is the nominal amount of carbon in the steel (c = 0.40 wt.% for
produced when dilatational strains interact with the already-existing
AISI 4140). For regions that contain neither martensite nor bainite
stress field. These strains are termed transformation-induced plasti-
(region that does not undergo transformation) bulk material hardness
city strains and can be calculated as
is assumed to be about 240 Vickers depending on measured hardness
of the material. 5 Sij ΔV
ΔeTrP
ij = ð2 − 2Xn − ΔX ÞΔX ð21Þ
4Y V
2.3. Stress calculation
where Sij is the deviatoric stress tensor, Xn is the amount of phase
Phase transformation strain, including volumetric dilatation strain already changed during the entire cycle, and Y is the yield stress of the
and transformation-induced plasticity strain, can be added to the weaker phase which happens to be austenite during these transfor-
elastic, plastic, and thermal strains as shown in Eq. (12) [24]. mations. References [41–44] give the material properties used in
Eqs. (19)–(21) and are shown in Table 2.
Δeij = ΔeEij + Δ ePij + ΔeTij + Δ eΔV TrP
ij + Δeij ð12Þ
3. Solution scheme
Elastic and plastic strains are governed by the isotropic elasto-
plasticity constitutive model in ABAQUS [39]. It is assumed that the A flowchart showing the various steps in the computational
plastic flow is independent of strain rate and material hardening is scheme employed in this work is given in Fig. 5. A thermal and phase
isotropic. Elastic strain can be expressed as transformation analysis is first performed using a transient 3D

1 + vr vr
eEij = σ ij − σ kk δij ð13Þ
E E

where Young's modulus, E, and Poisson's ratio, νr, are both functions
of temperature. The temperature-dependent plastic strain rate can be
expressed as
 
AF : AF : AF
ΔePij = G σ kl + T ð14Þ
Aσ kl AT Aσ ij

where Ĝ is the hardening function

 
1 AF AF AF
= − + σ mn ð15Þ
G Aemn Aκ
P Aσ mn

and F is the temperature-dependent von Mises yield function

1 1 2
F= S S − σ ðe; T Þ = 0: ð16Þ
2 ij ij 3

Here, κ is the work hardening parameter where κ̇ = σijε ̇ijP, Sij is the
deviatoric stress tensor

1
Sij = σ ij − σ δ ð17Þ
3 kk ij Fig. 5. Flowchart of numerical procedures for the calculation of residual stresses.
2008 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012

Fig. 6. Schematic of laser hardening process showing the temperature survey cross-section and the location of each survey point.

thermal model simultaneously coupled with a transient 3D kinetic transformation-induced plasticity strain. User subroutine USDFLD is
transformation model that includes the effects of carbon diffusion in also used in order for UEXPAN to have access to the current stress state
austenite homogenization and latent heat of phase transformation. while the model is running. The sum of these three strains is then sent
The resulting 3D hardness profile of the heat affected region is then to ABAQUS for each stress component at each time step, where it adds
calculated from the final microstructure. The temperature history can them to the elastic and plastic strains. Once the temperature in the
then be accessed by ABAQUS [39] through the user subroutine UTEMP workpiece has become uniform and cools to room temperature, the
while the phase fraction history is accessed by user subroutine, resulting stress state is taken as the residual stress state of the
UEXPAN. Because the thermal/kinetic model and the stress model do workpiece.
not use the same discretization technique (i.e. finite volume method The sizes of the cell volumes and finite elements used in this study
vs. finite element method) and since each requires a different mesh are as follows. All three dimensions of each cell and element are
density, linear interpolation in both space and time is used by the uniform inside the fine mesh regions, giving each cell and element a
stress model when accessing the temperature and phase fractions. In perfect cube shape. Inside the fine mesh area of the thermal model,
UEXPAN, the phase fraction of each phase is used to calculate the the length of a cell volume was 0.5 mm and the time step was set to
phase-dependent thermal strain, volumetric dilatation strain, and 0.1 s. The kinetic model has a much finer mesh with a cell volume
length of 5 µm and a time step of 0.01 s. The stress model used an
element size of 0.3 mm within the fine mesh region and an automatic
time increment defined by ABAQUS.

4. Results

A rectangular block of AISI 4140 steel measuring 50 × 50 × 18 mm

was laser hardened using a Nuvonyx 4 kW high power direct diode
laser as shown in Fig. 6. This laser has a rectangular beam profile with
a size of 12 × 8 mm and was set to a power of 1 kW and traverse speed
of 2 mm/s. The laser track coincides with the center of the workpiece,
the shorter axis of the beam profile aligned with the direction of laser
travel as shown in Fig. 6. A constant absorptivity of 0.68, determined
by a calibration experiment [45], is used for the simulation. Initial
temperature of the workpiece is 27 °C.
The temperature histories of 10 survey points, shown in Fig. 6,
were recorded. As shown in Fig. 7, the temperatures at all of these
survey points go above the A1 eutectoid temperature where the
pearlite transforms to austenite. All points, except j and g, are also

Table 3
Time taken for each survey point to cool from the A1 eutectoid temperature, 727 °C, to
the Tm-start martensite start temperature, 330 °C

Survey point a b c d e
Cooling time⁎ (s) 2.4 2.5 2.4 2.3 2.4

f g h i j
Fig. 7. Temperature history of various points shown in Fig. 6 relative to the A3 austenization
⁎from 727 °C to 330 °C 2.3 2.6 2.5 2.7 2.9
temperature and the A1 eutectoid temperature.
 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012 2009

structure at these subsections will be martensite with limited retained

austenite. The temperature at survey points j and g drops sufficiently
fast to avoid bainite formation. But since the temperature at these
points did not peak over the A3 austenization temperature and not all
ferrite transformed into austenite, there will be residual ferrite that
will soften the material at these points.
Fig. 8 shows a predicted hardness profile taken from the center of
the workpiece. The region consisting of only martensite and limited
retained austenite has a high hardness value of 712 Vickers. The
Fig. 8. Hardness profile of the survey area. transition region between the bulk material and martensitic region
goes through partial austenite transformation. From calculations of
heated above A3 austenization point, where the remaining ferrite also Eqs. (10) and (11), it has a lower hardness between that of martensite
transforms to austenite. At this point in time, the material at j and g and the bulk material, 712–250 Vickers.
will have a mixed microstructure of austenite and ferrite, depending Fig. 9 shows the optical micrographs obtained at different regions
on the temperature and carbon diffusion. Table 3 shows that the in the heat affected zone (HAZ) of the cross-section. The polished
temperature at all survey points drops from the A1 eutectoid tem- specimen was chemically etched with 2% Nital for about 40 s in order
perature, 727 °C, to the martensite start temperature Tm-start, 330 °C, to obtain these micrographs. Due to the high cooling rates (shown in
within 3.0 s. This cooling rate ensures missing the “nose” of the TTT Table 3), homogeneous martensite is obtained through most of the
diagram, as can be seen in Fig. 4. Consequently, the final micro- hardened region (Fig. 9a, needle like structure). The cooling rates

Fig. 9. Microstructure of the laser hardened AISI 4140 steel sample. (a) Homogeneous martensite. (b) Heterogeneous martensite. (c) Partial martensite. (d) Unaffected base material
(pearlite and ferrite).
2010 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012

Fig. 10. Comparison of predicted and measured case depth at the center of the workpiece.

Fig. 12. Residual stresses perpendicular to the laser travel along the surface of the
closer to the HAZ boundary are slower. Hence, this region (Fig. 9b) workpiece.
essentially has coarser and inhomogeneous martensite, although the
hardness value is similar. The transition region (Fig. 9c) between the
martensite and the unaffected material is softer, due to residual ferrite of the rotated angles, ψ, should always be linear. The residual stress
that didn't have time to transform into austenite. The material outside can then be determined by the slope of that relationship according to
of the HAZ (Fig. 9d) remains unaffected by heat. As shown, it consists Eq. (22),
of pearlite (darker phase) and α-ferrite (lighter phase).    
The microhardness in the HAZ was measured with a Vickers E 1 Ad
σ= ð22Þ
microhardness tester (1 kgf load and a dwell time of 20 s). Fig. 10 1 + νr d0 Asin2 ψ
shows the comparison of predicted hardness with measured hardness
into the depth at the center of the workpiece, giving a case depth of where E/(1 + νr) is an constant of 168.9 GPa and d0 is the diffraction
1.20 mm. The predicted results are in excellent agreement with the spacing in an unstressed state. The settings and constants used for the
microhardness measurements. measurements are as follows. The radiation source was Cr Kα and the
From the thermal/kinetic model, phase fractions of pearlite, ferrite, diffraction angle, 2θ, for the {211} lattice plane was set to 156.0°.
austenite, and martensite were obtained at each time step. The phase Residual stress values were predicted across the face of the
fractions were then accessed by the stress model which calculated the workpiece and into the depth of the workpiece as shown in Fig. 11.
thermal strains, volumetric dilatation strains, and transformation- Figs. 12 and 13 show predicted residual stress variation along the
induced plasticity strains at each time step of the stress model. surface and depth of the cross-section, respectively. Predicted stress
Residual stresses were measured at 2 points on the surface of the values agree very well with measured stress values.
workpiece: one measurement at the center of the laser track and From Figs. 12 and 13, it can be seen that phase transformation has a
one measurement 2.5 mm away from the center of the laser track. The significant effect on the final stress state. Thermal residual stresses are
X-ray diffraction method was used to measure the stress at these essentially tensile, while stresses due to martensite and bainite
points using a Siemens D500 X-Ray Diffractometer. Radiation incident transformations are compressive. During heating of the workpiece,
on the surface of the workpiece at a certain angle, 2θ, will diffract onto austenite transformation is accompanied by a slight volume contrac-
the collector with a certain diffraction spacing, d. As the workpiece is tion, while during cooling, martensite or bainite is accompanied by a
rotated through a series of angles, ψ (keeping 2θ constant), the large volume expansion. Thus, depending on whether thermal strains
diffraction spacing will slightly shift due to the residual stress. The or phase transformation strains are dominant, the residual stress field
relationship between the diffraction spacing and the square of the sin in a laser-hardened track can be either predominantly tensile or
compressive. As can be seen in Fig. 12, austenite-to-martensite
transformation leads to a high compressive stress region of about

Fig. 11. Residual stress value reporting scheme in the finite element model. Fig. 13. Residual stresses perpendicular to the laser travel into the depth of the workpiece.
 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012 2011

300 MPa in the transverse direction (perpendicular to laser travel), P Pearlite

near the center of the laser track. Volumetric dilatation is the major QL Latent heat of phase transformation (W m− 3)
contributor in phase transformation stresses. Stresses across the Qr Power generation per unit volume of the substrate (W m− 3)
surface, as well as stresses into the workpiece depth, show a strong Qν Activation energy of generic phase ν, either α or γ phase (J
compressive stress zone, then a small tensile stress region, followed by mol− 1)
a weak compressive stress region. Compressive stresses are favorable R Universal gas constant (8.314 J mol− 1 K− 1)
towards higher fatigue strength and resistance to corrosion and wear Sij Deviatoric stress tensor (Pa)
[46]. Future work in this area could include residual stress analysis in t Time (s)
multi-track laser hardening with track overlap and residual stress ta(T) Isothermal incubation time required at time T for bainite to
analysis in a workpiece with complex geometry. form (s)
Tm-start Temperature at which martensite begins to form (°C)
5. Conclusions T(t) Temperature at time t (°C)
T∞ Initial workpiece temperature and ambient temperature
A predictive model for calculating residual stresses in a laser (°C)
hardened steel workpiece has been presented and experimentally ΔV/V Volumetric strain occurring during phase transformation
verified. A finite volume 3D transient thermal/kinetic model which ΔX Amount of austenite being transformed in a particular step
calculates the temperature history and phase change history was used (volume fraction)
to simulate the heat transfer and corresponding phase transformation Xn Volume fraction of austenite already transformed
in a 50 × 50 × 18 mm AISI 4140 steel block, while taking into account Y Yield stress of the weaker phase during phase transforma-
appropriate boundary conditions, latent heat of solid phase transfor- tion (Pa)
mation, carbon diffusion in austenite homogenization, and cooling
effects. The final hardness field was then calculated using the resulting
microstructure. The hardness and case depth of the experimental Greek symbols
workpiece agreed with the model's results. The thermal history and α Ferrite
phase fraction history were then used in a finite element residual αi Coefficient of thermal expansion of phase i (K− 1)
stress model that takes into account thermal strains and phase γ Austenite
transformation effects of volumetric dilatation and transformation- δij Kronecker's delta
induced plasticity. The model predicted high compressive stresses εt Emissivity of the surface
throughout the hardened martensitic region surrounded by a region εijE Elastic strain
of tensile stresses and finally a stress-free region. The compressive εijP Plastic strain
stress in the hardened region is mainly due to volumetric dilatation of εTij Thermal strain
the local material transforming into martensite. The magnitude of the εijΔV Volumetric dilatation strain
residual stress in the hardened region of the experimental workpiece εTrP
ij Transformation-induced plasticity strain
was measured using X-ray diffraction, which agreed with the results 2θ Incident angle of radiation during X-ray diffraction (deg)
from the model. κ Work hardening parameter
ν Generic phase in diffusion model (placeholder for α or γ phase)
Nomenclature νr Poisson's ratio
A1 Eutectoid temperature (°C) ρ Density of the material (kg m− 3)
A3 Austenization temperature (°C) σ Stefan–Boltzmann constant (W m−2 K−4)
c Nominal concentration of carbon in the steel (wt.%) σij Induced stress (Pa)
Cp Specific heat at constant pressure (J kg− 1 K− 1) —(ε,T)
σ Temperature-dependent flow stress (Pa)
Ci Carbon concentration in cell i (wt.%) Ψ Rotation angle in X-ray diffraction (deg)
Cv Carbon concentration in phase ν, either α or γ phase (wt.%)
Ciα⁎ Carbon concentration at α side of interface in kinetic model
(wt.%) Acknowledgments
Ciγ⁎ Carbon concentration at γ side of interface in kinetic model
(wt.%) The authors wish to gratefully acknowledge that this study has
Ckν⁎ Carbon concentration at ν side of interface where ν is either been partially funded by the National Science Foundation (Grant No:
α or γ phase (wt.%) IIP-0538786) and the State of Indiana through the 21st Century R&T
d Diffraction spacing in X-ray diffraction (m) Fund.
D0ν Pre-exponential diffusion coefficient of α or γ phase (m2 s− 1)
Dν Diffusion coefficient in phase ν, either α or γ phase (m2 s− 1) References
E Young's modulus (Pa)
fi Volume fraction of a given phase [1] N.B. Dahotre, Surface Engineering Series Volume 1, ASM International, Materials
Park, OH, 1998.
fiα Volume fraction of ferrite at α/γ interface [2] R.L. Peng, T. Ericsson, Scand. J. Metal. 27 (1998) 223.
fγ⁎ Volume fraction of austenite at the time when temperature [3] I. Cerny, I. Fürbacher, V. Linhart, J. Mater. Eng. Perform. 7 (1998) 361.
reaches Tm-start [4] S. Kou, D.K. Sun, Y.P. Lee, Metall. Trans. A 14 (1983) 643.
[5] M.F. Ashby, K.E. Easterling, Acta Metall. 32 (1984) 1935.
F Temperature-dependent yield function
[6] M. Davis, P. Kapadia, J. Dowden, W.M. Steen, C.H.G. Courtney, J. Phys. D: Appl. Phys.
Ĝ Hardening function 19 (1986) 1981.
h Size of cell in numerical model (m) [7] E. Ohmura, K. Inoue, K. Haruta, JSME Int J. Ser. 1(32) (1989) 45.
hn Natural convection coefficient (W m− 2 K− 1) [8] A. Jacot, M. Rappaz, Acta Mater. 47 (1999) 1645.
[9] S. Skvarenina, Y.C. Shin, Surf. Coat. Technol. 201 (2006) 2256.
Hi Hardness of phase i (Vickers) [10] R. Patwa, Y.C. Shin, Int. J. Mach. Tools Manuf. 47 (2007) 307.
Jki Solute flux from cell k to cell i in kinetic model (kg m− 2) [11] R. Komanduri, Z.B. Hou, Int. J. Heat Mass Transfer. 44 (2001) 2845.
k Thermal conductivity of the material (W m− 1 K− 1) [12] R. Komanduri, Z.B. Hou, Int. J. Mach. Tools Manuf. 44 (2004) 991.
[13] L. Costa, R. Vilar, T. Reti, A.M. Deus, Acta Mater. 53 (2005) 3987.
m Martensite [14] A. Yánez, J.C. Álvarez, A.J. López, G. Nicolás, J.A. Pérez, A. Ramil, E. Saavedra, Appl.
n Normal vector of the surface Surf. Sci. 186 (2002) 611.
2012 N.S. Bailey et al. / Surface & Coatings Technology 203 (2009) 2003–2012

[15] M.J. Tobar, C. Álvarez, J.M. Amado, A. Ramil, E. Saavedra, A. Yánez, Surf. Coat. Technol. [31] L. Wang, S.D. Felicelli, P. Pratt, Mater. Sci. Eng., A 496 (2008) 234.
200 (2006) 6362. [32] C. Şimşir, C.M. Gür, Comput. Mater. Sci. 44 (2008) 588.
[16] T. Miokovic, V. Schulze, O. Vöhringer, D. Löhe, Mater. Sci. Eng., A 435–436 (2006) 547. [33] R.S. Lakhkar, Y.C. Shin, M.J.M. Krane, Mater. Sci. Eng., A 480 (2008) 209.
[17] T. Miokovic, V. Schulze, O. Vöhringer, D. Löhe, Acta Mater. 55 (2007) 589. [34] J.C. Rozzi, F.E. Pfefferkorn, F.P. Incropera, Y.C. Shin, Int. J. Heat Mass Transfer 43
[18] M. Leung, J. Phys. D: Appl. Phys. 34 (2001) 3434. (2000) 1409.
[19] M.K.H. Leung, H.C. Man, J.K. Yu, Int. J. Heat Mass Transfer 50 (2007) 4600. [35] Y. Tian, B.X. Wu, Y.C. Shin, ASME Trans., J. Manuf. Sci. Eng. 130 (2008) 031013.
[20] F. Lusquiños, J.C. Conde, S. Bonss, A. Riveiro, F. Quintero, R. Comesaña, J. Pou, Appl. [36] S.H. Kang, Y.T. Im, Metall. Mater. Trans. A 36 (2005) 2315.
Surf. Sci. 254 (2007) 948. [37] P.R. Woodard, S. Chandrasekar, H.T.Y. Yang, Metall. Mater. Trans. B 30 (1999) 815.
[21] G. Tani, L. Orazi, A. Fortunato, G. Campana, G. Cuccolini, Proc. SPIE Int. Soc. Opt. Eng. [38] H.E. Boyer, A.G. Gray, Atlas of Isothermal Transformation and Cooling Transforma-
6454 (2007) 645404. tion Diagrams, ASM International, Metals Park, OH, 1977.
[22] G. Tani, L. Orazi, A. Fortunato, G. Campana, A. Ascari, Proc. SPIE Int. Soc. Opt. Eng. [39] ABAQUS, ABAQUS/Standard user's manual for version 6.5.1, Hibbitt, Karlsson and
6985 (2008) 69850A. Sorensen Inc, 2004.
[23] G. Tani, L. Orazi, A. Fortunato, CIRP Ann. 57 (2008) 209. [40] S. Das, G. Upadhya, U. Chandra, Proceedings of the First International Conference
[24] T. Inoue, Handbook of Residual Stress and Deformation of Steel, ASM International, on Quenching & Control of Distortion, ASM International, Chicago, 1992, p. 229.
2002, p. 296. [41] L. Jin, Doctoral Thesis, Purdue University, August (2001).
[25] S. Denis, P. Archambault, C. Aubry, A. Mey, J.C. Louin, A. Simon, J. Phys. IV 9 (1999) [42] J.Z. Zhao, C. Mesplont, B.C. De Cooman, Mat. Sci. Eng. A 332 (2002) 110.
Pr9–323. [43] M. Narazaki, G.E. Totten, G.M. Webster, Handbook of Residual Stress and Deformation
[26] A. Solina, M. de Sanctis, L. Paganini, A. Coppa, J. Heat Treat. 4 (1986) 272. of Steel, ASM International, 2002, p. 248.
[27] Lacarac, D.J. Smith, C. McMahon, M.J. Tierney, Proceedings from the 1st International [44] H.K.D.H. Bhadeshia, Handbook of Residual Stress and Deformation of Steel, ASM
Surface Engineering Congress and the 13th IFHTSE Congress, 7–10 October, Columbus, International, 2002, p. 3.
OH, 2002, p. 578. [45] Y.S. Touloukian, C.Y. Ho, Thermo physical Properties of Matter—The TPRC Data
[28] Y.S. Yang, S.J. Na, J. Heat Treat. 9 (1991) 49. Series by Purdue University, IFI/Plenum Data Corp., New York, 1970.
[29] A.G. Grigor’yants, A.N. Safonov, V.S. Maiorov, A.F. Baskov, G.P. Ivashov, Met. Sci. [46] D.I. Pantelis, E. Bouyiouri, N. Kouloumbi, P. Vassiliou, A. Koutsomichalis, Surf. Coat.
Heat Treat. 29 (1978) 691. Technol. 161 (2002) 125.
[30] H.X. Zhan, Y. Wang, C.W. Li, T. Han, B. Han, W.M. Zhao, Opt. Laser Technol. 41 (2009)
251.