Influence of successive thermal cycling on microstructure evolution of EBM-manufactured alloy 718 in track-by-track and layer-by-layer design

(1)

http://www.diva-portal.org

This is the published version of a paper published in Materials & design.

Citation for the original published paper (version of record):

Karimi, P., Sadeghi, E., Åkerfeldt, P., Ålgårdh, J., Andersson, J. (2018)

Influence of successive thermal cycling on microstructure evolution of

EBM-manufactured alloy 718 in track-by-track and layer-by-layer design

Materials & design, 160: 427-441

https://doi.org/10.1016/j.matdes.2018.09.038

Access to the published version may require subscription.

N.B. When citing this work, cite the original published paper.

Permanent link to this version:

(2)

In

ﬂuence of successive thermal cycling on microstructure evolution of

EBM-manufactured alloy 718 in track-by-track and layer-by-layer design

Paria Karimi

a,

⁎

, Esmaeil Sadeghi

a

, Pia Åkerfeldt

b

, Joakim Ålgårdh

a,c

, Joel Andersson

a

Department of Engineering Science, University West, 461 86 Trollhättan, Sweden

b_{Department of Engineering Sciences and Mathematics, Luleå University of Technology, 971 87 Luleå, Sweden} c

Powder Materials & Additive Manufacturing, Swerea KIMAB AB, 164 40 Kista, Sweden

H I G H L I G H T S

• Track-by-track and layer-by-layer sam-ples were manufactured by EBM. • Effect of successive thermal cycling

(STC) on microstructure was studied. • PDAS was ﬁner (~ 35%) in the one-time

melted area of the track than the over-lap zone in track-by-track samples. • Higher cooling rate in the bottom than

top layers was observed in all the layer-by-layer samples.

• Slightly higher hardness (~ 11%) was obtained at the bottom layers in both the single and thick walls.

G R A P H I C A L A B S T R A C T

a b s t r a c t

a r t i c l e i n f o

Article history: Received 26 May 2018

Received in revised form 31 August 2018 Accepted 19 September 2018 Available online 20 September 2018

Successive thermal cycling (STC) during multi-track and multi-layer manufacturing of Alloy 718 using electron beam melting (EBM) process leads to a microstructure with a high degree of complexity. In the present study, a detailed microstructural study of EBM-manufactured Alloy 718 was conducted by producing samples in shapes from one single track and single wall to 3D samples with maximum 10 longitudinal tracks and 50 vertical layers. The relationship between STC, solidification microstructure, interdendritic segregation, phase precipitation (MC, δ-phase), and hardness was investigated. Cooling rates (liquid-to-solid and solid-to-solid state) was estimated by measuring primary dendrite arm spacing (PDAS) and showed an increased cooling rate at the bottom compared to the top of the multi-layer samples. Thus, microstructure gradient was identified along the build direction. Moreover, extensive formation of solidification micro-constituents including MC-type carbides, induced by micro-segregation, was observed in all the samples. The electron backscatter diffraction (EBSD) technique showed a high textured structure in〈001〉 direction with a few grains misoriented at the surface of all samples. Finer microstructure and possibility of moreγ″ phase precipitation at the bottom of the samples resulted in slightly higher (~11%) hardness values compared to top of the samples.

Keywords:

Electron beam melting Alloy 718

Track by track Layer by layer

Successive thermal cycling Microstructure evolution

1. Introduction

Electron beam melting, as a powder-bed fusion (PBF-EB) technique

in theﬁeld of additive manufacturing (AM), is capable of directly

manufacturing three dimensional (3D) parts from a computer aided ⁎ Corresponding author.

E-mail address:paria.karimi-neghlani@hv.se(P. Karimi).

https://doi.org/10.1016/j.matdes.2018.09.038

Contents lists available atScienceDirect

Materials and Design

(3)

design (CAD)ﬁle [1–3]. EBM is aimed to manufacture lightweight, and geometrically complex parts as well as rapidly produce small series re-quired by aerospace and automotive sectors [4]. During EBM process, metallic powder particles are rapidly irradiated by a high energy elec-tron beam in a spot/linear scanning strategy to provide track-by-track and layer-by-layer deposition of new material [5].

In the EBM process, the electron beam is scattered over a powder bed where the electron beam energy is absorbed by the powder parti-cles in successive steps of sintering and melting of each layer, as well

as re-melting of the solidi_{ﬁed underlying layers. The rest of the energy}

conducts to the underlying layers [6,7]. This way of repetitive processing

is a thermo-physical process known as“successive thermal cycling

(STC)_{” which can induce thermally-activated diffusion processes like}

grain growth, phase transformation, and precipitation, etc. Therefore, a heterogonous microstructure for a multi-layer part is obtained due to the non-equilibrium condition of molten materials within such rapid process [8]. In fact, STC directly affects the temperature gradient within the melt pool and cooling rate from liquid-to-solid and

solid-to-solid states that signiﬁcantly inﬂuence on the microstructure [6].

Thus, establishing a correlation between the thermal history and micro-structure characteristics of an EBM-as built material is highly educative. In addition, the foundation of the process physics in EBM is heat trans-formation, which includes a high-intensity and localized movement of a heat source as well as a rapid cooling. Therefore, estimation of the cooling rate can be an essential tool to understand the evolution of

mi-crostructural features such as solidi_{ﬁcation morphology, precipitation of}

phases, and grain orientations [9].

In general, the EBM melt pool formation consists of the following physical phenomena: absorption of the electrons kinetic energy when interacting with the powder material, which generates heat and X-rays, heat transfer (radiation, conduction under vacuum and convection during cooling down by helium injection), evaporation and ejection of the feedstock powders or molten materials [10]. A variety of methods, e.g., thermocouples, and high resolution near infrared (NIR) cameras have been proposed to measure the thermal history during the AM pro-cesses, e.g., laser metal deposition (LMD), selective laser melting (SLM),

and EBM [11–15]. In one approach, thermocouples have been used to

measure the temperature at selected points on parts in LMD [11,16]. However, the thermocouples have limited applications in EBM as they must be attached to bottom of the build plate before the pro-cess begins, and can only measure temperature near the build plate [17].

Sizeable literature is available on the microstructural

characteriza-tion of EBM parts made of Alloy 718 [5,18–21]. Most of the experiments

on Alloy 718 have been conducted on bulky parts, including either cubic samples or rod/plate geometries [6,7,22]. The others have mainly been focused on tailoring grain structure from columnar to equiaxed transi-tion by altering process parameters [19,20]. It has also been depicted that microstructure of EBM-as built parts can be manipulated by alter-ing the build temperature and process parameters such as scannalter-ing strategy, beam speed, and beam power [18,19]. The process parameters

affect the thermal gradients (G), liquid_{–solid interface velocities (R),}

peak temperatures (Tp), cooling rates (Ṫ), and number of thermal cycles,

which in turn, control the solidiﬁcation structure, crystallographic

tex-ture, and_{ﬁnal microstructure [}6]. However, microstructure evolution

of Alloy 718 processed by EBM in small scales like track-by-track or layer-by-layer has not been reported. Unfortunately, empirical mea-surements and monitoring of the thermal history during multi-layer de-position in EBM are challenging, as it contains localized heating and

rapid melting/solidiﬁcation in a contained chamber.

Therefore, this study aimed to investigate the EBM process from the

perspective of solidiﬁcation phenomena and phase transformation

pro-cess encountered for Alloy 718. In the present work, the heterogeneous microstructure observed in EBM-manufactured Alloy 718 was studied in single deposited tracks and/or layers. The effect of STC was explored through examination of primary dendrite arm spacing (PDAS), Nb-rich

phase constituents, and hardness proﬁle data.

2. Experimental procedure 2.1. Materials

The feedstock material was a plasma atomized Alloy 718 powder provided by Arcam (Arcam AB, Mölndal, Sweden). The chemical

com-position of the powder supplied by Arcam AB is given inTable 1. The

powder was mainly spherical in shape with a fewﬁne satellite particles

attached to the surface of coarse particles, seeFig. 1a. A small amount of

partially sintered powder particles was also observed as the powder

used in this study was a mixture of a virgin and aN10 times recycled

powder (Fig. 1a). The particle size distribution of the virgin feedstock

powder was reported to 45–105 μm; and the used powder was slightly

larger (~33%) giving a powder size distribution (PSD) of 44–140 μm due

to the presence of partially sintered particles (Fig. 1b). Table 1

Chemical composition of the Alloy 718 powder.

Element Ni Co Cr Mo Ti Mn Nb P Ta Al Fe Si S C

wt% 54.11 0.04 19.0 2.99 1.02 0.12 4.97 0.004 b0.01 0.52 Bal. 0.06 b0.001 0.03

(4)

2.2. EBM procedure

An Arcam A2X EBM machine (Arcam AB, Mölndal, Sweden) was

uti-lized to build up 26 samples in three groups given inTable 2.

In theﬁrst two groups, the aim was to understand the effect of STC in

two different directions (towards X and Z axis) on the microstructure

during the EBM process, seeFig. 2(a–b). The third group was aimed to

investigate the effect of wall thickness on the cooling rate. During the course of the present study, the length of the tracks was kept constant at 20 mm. A software version of 4.2.201 in the EBM machine was uti-lized to manufacture the samples. A standard linear melt theme with deactivation of power calculation function and a unidirectional scan

strategy (seeFig. 2c) was implemented. The sintering and melting

pro-cess began once the preheat temperature of 1023 °C was reached. A con-stant beam current of 7 mA and scanning speed of 1.25 m/s with a gun-accelerating voltage of 60 kV were used throughout the builds. The cur-rent and speed were chosen from standard hatch parameters for a 20

× 20 × 20 mm3_{cube. The layer thickness and line offset were kept}

con-stant at 75 and 125μm, respectively, during the build.

As shown inFig. 2, before deposition of the main samples, an

addi-tional substrate (stand) was built using the Arcam standard process pa-rameters for Alloy 718. The motivation of using a stand are; i) to prevent dilution effects which contribute to changes in melting temperature of

the powder, ii) to provide aﬂat bed with correct height for building

tracks, and iii) to ease the collection of the samples during the powder recovery system (PRS) process, used to remove the sintered powders attached to the as-built parts. The gap between each individual stand

was 5 mm, and the height of each stand 8 mm, seeFig. 2.

2.3. Characterization of microstructure

Two cross-sections from each sample were evaluated. All the sam-ples were sectioned perpendicular to the scanning direction at 7 mm from the start point and end point of each melted track/tracks, see Fig. 3. The samples were gradually ground using SiC paper up to 1200 grit. The ground samples were then polished down using a 0.05 μm SiC suspension. The polished samples were etched electrolytically using oxalic acid (10 wt%) at room temperature with 3 V for 5 to 10 s

to reveal the phase constitutes such asδ-phase and MC-type carbides.

First examination of the microstructural features was performed using a light optical microscope (LOM) (Olympus-BX60M, Tokyo, Japan) to measure the penetration depth of the electron beam and melt pool width. The type of defects such as powder induced pores Table 2

Three groups of samples manufactured by the EBM process.

Group Condition Designation Number of samples

I Track-by-track melting (or single layers), seeFig. 2a T1L1a

, T2L1, T3L1,…, T9L1, T10L1 10 II Layer-by-layer melting (or single walls), seeFig. 2b T1L1, T1L2, T1L3, T1L4, T1L5, T1L10, T1L25 and T1L50 8 III Equal number of tracks (ﬁve tracks) but with different number of layers T5L1, T5L2, T5L3, T5L4, T5L5, T5L10, T5L25 and T5L50 8

a

T = track and L = layer.

Fig. 2. a) Schematic of the stands and CADﬁle of the track-by-tack samples, b) schematic of the stands and CAD ﬁle of the layer-by-layer samples, c) scanning strategy in coordination system, and d) full image of all samples on the build plate.

stand

20 mm

7 mm

Cutting planes

Single track

(5)

and process induced defects (e.g., lack of fusion, solidiﬁcation shrinkage

porosities) was also inspected. A_{ﬁeld emission scanning electron}

mi-croscope (FE-SEM) (ZEISS EVO 50, Cambridge, UK) equipped with en-ergy dispersive spectrometry (EDS) was used to examine the

solidiﬁcation mode, corresponding phase constituents and also

charac-terize the chemical composition of the phases. Grain orientation map-pings of samples (T1L1, T5L1, T7L1, T1L5, T1L10, and T5L50) were obtained by an electron backscatter diffraction (EBSD) system (GAIA3-TESCAN, Cambridge, UK), operating with an accelerating voltage of 20 kV, to analyze grain nucleation, formation and misorientation of

elongated grains. The step size of the EBSD mapping was set to 2μm.

To predict presence of phases in the microstructure, a statistical analysis was carried out using a thermodynamic software program, JMatPro ver-sion 10.2 (Guildford, UK) [23]. An equilibrium phase diagram and the physical properties of Alloy 718 were obtained from JMatPro. Vickers microhardness (HV) measurements were performed in different loca-tions (top, middle, and bottom) of the samples using a microhardness tester (HMV-2, Shimadzu, Kyoto, Japan) with an applied load of 500 g

and a dwell time of 15 s to provide hardness proﬁle in each sample.

The indent size of Vickers indenter was about 50μm and one

indenta-tion with appropriate spacing with the edges was performed in the mid-dle of each track.

3. Results and discussion 3.1. Surface morphology

Fig. 4(a1–a2) shows resulting gaps and distortion of melted tracks in the T1L1 and T2L1 samples. The absence of a continuous melt track

ver-iﬁed that the applied beam energy input was not enough and caused

un-molten/partial melting of powder particles. Possible explanations is that the EBM process is rather complex and involves many different

physical phenomena [5]: absorption of the electron beam in the powder

bed, melting and solidi_{ﬁcation, dynamics of the melt pool, wetting of}

the solid powder particles by the melt, conductive, radiative and con-vective heat conduction within the powder bed and in the melt pool, capillary effects, and gravity, and others. [2,8,24]. The melt pool gener-ated by the electron beam is highly dynamic. The movement is driven by high surface tension in combination with low viscosity of the molten metals [8,25]. As a result, the consolidated surface shows a stochastic

nature, i.e. it is irregular and island-shaped. The insufﬁcient energy

input was most likely attributed to a high scanning speed, which

re-sulted in a too low energy input, according to Eq.(1)[26,27]. By adding

three tracks next to each other inFig. 4(a3), a continuous and uniform

solid layer was obtained, most likely due to enough overlap between

the tracks which creates sufﬁcient energy input (in resemblance to

Eq.(1)) to_{ﬁll all the gaps. Moreover, there were some irregularities}

and semi ball-shaped drops in T1L1 to T5L1, seeFig. 4(a1–c1), assumed

to be caused byﬂuctuations of the beam movement during melting of

the powders.

Energy input J

mm3

¼_{scanning speed mm}_ð voltage kV_=s_{Þ line offset μm}ð Þ current mA_ð _{Þ layer thickness μm}ð Þ _ð _Þ ð1Þ

Fig. 4(b1–b4) shows top view of the layer-by-layer samples (T1L1–

T1L4). The observed gaps in the single track (T1L1) shown inFig. 4

(b1), this indicates that the layer-by-layer samples were not adequately solid along the wall and a few gaps was observed in some layers. For in-stance, in the samples with two and three layers, the gaps were clearly visible. In addition, the width of the four-layer sample was locally narrower in some regions, most likely due to an unstable melt pool which leads to the poor bonding between the layers. Therefore, the Fig. 4. SEM images (BSE mode) from the top view of the samples showing track stability, gaps, ball formation, distortions, and voids; a1) T1L1, a2) T2L1, a3) T3L1, b1) T1L1, b2) T1L2, b3) T1L3, b4) T1L4, c1) T5L1, c2) T5L3, c3) T5L4, and c4) T5L10.

Fig. 5. LOM images from the cross section of the samples showing the microstructure in the overlap zone and the one-time melted area of the track, to better visualize the melt pool boundaries after etching, black dashed lines are drawn in theﬁgure; a) T1L1, b) T3L1, c) T7L1, d) schematic of the size of melt pool width and beam penetration depth, and e) schematic of melt pool overlaps.

(6)

(7)

presumption can be drawn that the source of the gaps in the layer-by-layer samples (too low energy input) is identical to the source in the track-by-track samples. Moreover, it can also be rationalized that, the chance of gap formation can be higher in the layer-by-layer samples than the track-by-track samples. In the same amount of energy input, smaller particles were completely melted due to the smaller volume/ mass ratio, but the coarse particles were partially melted at the track edges. Consequently, the coarse particles generated more corrugated

edges. Afterﬁrst melting, the track-by-track sample was exposed to

the powder particles only at a side edge. Another side of the edge was contacting to the previously melted solid face. Thus, it had less chance to form the wavy edges compared to the layer-by-layer sample. In the layer-by-layer sample, two edges of the track were always contacting to the powder particles. So, it had higher chance to form corrugated and discontinuous track.

In the samples with_{ﬁve number of tracks but with variation in}

num-ber of layers (T5L1, T5L3, T5L4), the top surface of the samples still

showed the ball formation, gaps or voids within the tracks, seeFig. 4

(c1_{–c3). In the present study, the beam parameters remained constant}

in all the builds, so the balling phenomenon can be as a result of either the capillary instability in the dynamic melt pool owing to Marangoni effects, or lack of dilution with the stand, leading to a melt pool spheroidization to minimize free energy [28]. These balling effects are

however not observed by adding 10 layers, seeFig. 4(c4), where heat

accumulation is believed to help generating enough energy input to smoothen the top surface.

3.2. Analysis of melt pool geometry and solidiﬁcation morphology

The LOM images inFig. 5(a–c) showed a trace of the melt pool

boundaries in the track-by-track samples. The electron beam penetra-tion depth and width of the tracks were measured which are

schemat-ically presented inFig. 5d. The melt pool width and melt pool depth

were measured as 250–390 μm and 40–110 μm, respectively for tracks,

seeFig. 5(a–c). With a consolidated layer thickness of around 75 μm,

re-melting occurred in maximum two underlying layers with measured

energy input of ~36 J/mm3_{according to Eq.}_{(1). Moreover, overlap}

be-tween the tracks was around 30_{–50% of each track, meaning that a}

part of the track was remelted twice or three times by addition of the next tracks (Fig. 5e).

It was noted in the high magni_{ﬁcation SEM image (}Fig. 5a) that the

columnar dendrites grew epitaxially from the stand and were generally

orientated along the〈001〉 crystallographic direction, parallel to the

build direction. Most of the primary dendrites were parallel to each other in the cross section of one single track; however, due to the dy-namic thermal gradient in the un-equilibrium condition of the melt

pool, some misorientations (N15°) among the growing dendrites were

observed (Fig. 5a). By addition of the adjacent track, a new melt pool was formed in which a part of this new melt pool was the overlap

zone. The previous directionally solidiﬁed dendrites in the overlap

zones were remelted and their directions changed towards the center of the new melt pool in the new added track. The change of the dendritic growth direction in the overlap zone can be attributed to severe change

of the thermal gradient/heat_{ﬂux direction during solidiﬁcation, see}

Fig. 5(b–c). Thus, the dendritic growth direction in the overlap zones, e.g. zones 1 and 2, was changing about 90° towards either the next

track or the scanning direction as shown at higher magniﬁcation in

Fig. 5c.

Primary dendrite arm spacing (PDAS) was measured by image anal-ysis (IA) technique using ImageJ software [29]. The SEM micrographs of

the polished samples with horizontalﬁeld width of 45 μm were used to

quantify the length between the tips of two adjacent dendrite arms. More details about PDAS measurements using ImageJ software can be found in a previous works [30,31]. After 15 and 10 measurements of the PDAS in the one-time melted area and overlap zone, respectively, seeFig. 6(a–b), it was observed that the mean value of PDAS was slightly lower (around 35%) in the one-time melted area of the track (~1.15 ±

0.27μm) compared to the overlap zones (~1.77 ± 0.46 μm). The overlap

zone was scanned twice by the electron beam resulted in a higher heat

content (e.g., zones 1 and 2 inFig. 5c) compared to the one-time melted

(8)

area of the track. Therefore, the heat dissipation was expected to be slower in the overlap zones.

In the layer-by-layer samples shown inFig. 7(a–b), addition of

layers, like the third orﬁfth layer, the melt pool boundaries are visible.

However, in the 10-layer sample, the melt pool boundaries at the bot-tom layers are no longer visible which presumably is caused by STC in

the bottom layers during the building process. As shown inFig. 7(a–

d), in the interior area of the layer-by-layer samples, epitaxial growth of the columnar cellular-dendritic growth was observed in the build di-rection, most likely due to the direction of thermal gradient. However,

in the exterior area adjacent to the surface of the samples,ﬁne features

were observed, which seemed to be new grains emerging due to the surface grain nucleation mechanism [32]. The melt pool surface can be undercooled thermally to induce surface nucleation by reduction or re-moval of the heat input as form of heat radiation.

It is known that solidi_{ﬁcation morphology is affected by a}

combina-tion of temperature gradients (G) in the liquid at the solidiﬁcation front,

and the solidiﬁcation rate (R) [33]. In the microstructures presented in

Fig. 8, formation of the columnar dendritic structure with veryﬁne secondary arms can be explained by a change in G and R [8]. The Fig. 7. LOM images of layer-by-layer samples illustrating the melt boundaries and cellular-dendritic structure, a) T1L3, b) T1L5, and c) T1L10, and d) SEM image (BSE mode) showing new grain formation in the exterior area of T1L25.

(9)

solidiﬁcation condition was changing during the solidiﬁcation of the melt pool and the G and R were spatially and temporarily in the melt pool. It is reported that the magnitudes of G and R vary inversely

along the solidiﬁcation boundary. The G increases and the R decreases

when traversing from the melt pool centre to the bottom of the melt pool whereas G decreases and R increases at the top surface of the melt pool, in accordance to the literature [5,24,34,35]. Thus according

to this theory as shown inFig. 8d, a cellular structure with no secondary

arms is expected at the start of the solidiﬁcation. However, as seen in

Fig. 8(a–c), the solidiﬁcation morphology was mainly dendritic with

veryﬁne secondary dendrite arm spacing including random existence

of cellular structures. Moreover, the observed solidiﬁcation morphology

was similar along the build direction up to 50 layers. The dendritic struc-ture was observed all over the sample, which was expected to be af-fected by STC during repetitive layer addition. No change in the

solidi_{ﬁcation morphology can likely be attributed to the small geometry}

of the investigated specimens.

3.3. Microstructural characterization of defects

One of the common defects observed in all samples was a band of possible shrinkage pores located along the melt pool boundaries,

which potentially can be caused by insufﬁcient liquid available to

com-pensate the shrinkage during the solidiﬁcation, seeFig. 9(a–c). The

pri-mary reason for such behaviour can be the presence of a small melt pool

(with very low amount of liquid) being solidiﬁed extremely fast as a

re-sult of the rapid temperature cycling in the EBM process. Moreover, by a

closer look in the band of shrinkage pores shown inFig. 9d, it was found

that the shrinkage pores were located in the region perpendicular to the

layers and mainly in the interdendritic areas. The dominant mechanism that induced such shrinkage pores is believed to be the residual stresses

caused during the rapid solidi_{ﬁcation [}36]. The shrinkage pores form

during cool-down phase of the molten top layers which is hindered by the underlying previously processed layers [3]. Moreover, the surface cracks particularly in the edges of the added tracks were observed which can be due to two reasons. Firstly, it can be because of the macro-scopic defects such as delamination resulted from low energy input. Secondly, it can be due to the longitudinal shrinkage stresses acting on the tracks. Such cracks are similar to the transverse cracks observed in welding, which are perpendicular to the direction of the track, see Fig. 9(e–f) [37,38].

3.4. STC effect on the grain formation mechanism

The EBSD inverse pole_{ﬁgure (IPF) colour maps from cross sections}

of three track-by-track samples (T1L1, T5L1, and T7L1) along the build

direction (Z-axis, seeFig. 10d) are presented inFig. 10(a–c). Overall,

the epitaxial solidiﬁcation initiated from the partiality re-melted grains

in the stand was attributed to alignment of the thermal gradient with

the build direction with high degree of texture in〈001〉 crystallographic

orientation perpendicular to the scanning direction. This behaviour is typical of an AM processed Alloy 718 [22], where the red colour

corresponded to the〈001〉 crystallographic grain orientation in the IPF

colour legend, as shown inFig. 10d. Upon rapid cooling from the melted

state, the growing grains align themselves with the steepest tempera-ture gradients which results in columnar shaped morphology. Typically,

grain structure in AM process is similar to directionally solidiﬁcation

process in which a positive thermal gradient should be imposed at the

(10)

growth front to let the columnar grains to grow. This is obtained where heat is extracted through the solid, in the opposite direction to the growth direction. The temperature gradient in the liquid ahead of the solid to liquid interface is positive. In this case, columnar grains will be

arranged parallel to each other in arrays [22]. Newﬁne and randomly

orientated grains were also observed to nucleate at the top surface most likely due to undercooled thermal condition caused by radiation of heat from the surface which led to a reduction in the thermal gradient at top surfaces [5,11].

As indicated inFig. 10(b–c) by hatched squares from samples

con-taining multiple tracks side-by-side, parts of the grains from the previ-ous track in the overlap zone was partially re-melted where the

direction of the solidiﬁed grains changed towards the adjacent track.

This can be due to the re-melting of the previous track where a temper-ature gradient is induced from repetitive melting. The grains grew epi-taxially towards the centre of the new melt pool, where the thermal gradient was around 90° to the build direction, which was always in an identical direction for the next adjacent tracks. As a result, the grain direction rotated towards the centre of the adjacent tracks as shown

in the black hatched square inFig. 10, which is also in accordance to

lit-erature [39]. Vrancken et al. [40] reported similar grain orientations in the overlap zones for SLM-processed materials.

There were some islands of grains with different orientations, most

likely in the overlap zones (points 1 and 2 inFig. 10), with their sizes

being coarser than that of the grains at the surface of the tracks (point 3 inFig. 10). The potential reason for nucleation of these grains can be

the partially melted grains in the overlap zones which detach

them-selves from the solid–liquid mixture surrounding the melt pool. These

partially melted grains in the melt pool can act as nuclei site for the for-mation of new grains [32]. Coarsening of these grains at overlap zones can be explained by referring to the cooling rate, which was probably lower in the overlap zone. As expected, the energy input content in the overlap zone was higher than that in the one-time melted area of the track, and therefore the slower cooling rate in the overlap zone pro-motes formation of the slightly coarser grains in these regions

com-pared to theﬁne surface grains [39].

The EBSD analysis of the layer-by-layer samples both in single walls and thick walls (T1L5, T1L10, and T5L50) revealed predominantly

co-lumnar grain structure along the Z-axis, seeFig. 11. It is clear that the

grains were generally coarser in the interior part of the walls where

di-rections mainly were〈001〉 compared to those at the edges with 〈101〉

and〈111〉 directions. This was observed in all the layer-by-layer

sam-ples with various numbers of layers. At the bottom of the samsam-ples;

T1L5, T1L10, and T5L50, seeFig. 11(b–d), most of the grains, grown

ep-itaxially from the partially remelted grains, starts from the grains in the stand which were orientated along the build direction. In the subse-quent added layers, grains grew owing to very sharp thermal gradient along the build direction; therefore, elongated and columnar grains

were observed along this direction, seeFig. 11(b–d). The observed ﬁne

grains in the top and edge surfaces of walls were mainly due to

undercooled thermal condition as discussed previously, seeFig. 11(b–

d) [5].

Fig. 10. EBSD orientation maps in the build direction (YZ plane); a) T1 L1, and b) T5 L1, c) T7 L1, and d) sample coordinate system and inverse poleﬁgure colour legend.

(11)

3.5. Estimation of cooling rate by PDAS measurement and effect of Nb concentration

Estimation of the cooling rate can be determined by measurement of

the secondary dendrite arm spacing [7]. As shown inFig. 8(a–c), the

sec-ondary arms were unable to grow enough due to the rapid solidiﬁcation

during the EBM process. Thus, if not enough secondary arms are present in the microstructure, PDAS can effectively be used for the measure-ment. In general, PDAS grow until a maximum Nb concentration is reached in the interdendritic areas [32]. However, in its solid state, owing to diffusion of Nb from the interdendritic areas to dendrite cores, coarsening of PDAS may occur. Therefore, in estimating the cooling rate by PDAS, the kinetics of Nb diffusion which depends on time, temperature, and diffusion distance must be considered. For this purpose, the Nb concentration in the dendrite cores was measured in different heights from the stand in the layer-by-layer samples and

re-sults are presented inFig. 12(a_{–b). The EDS results showed that the}

mean value of Nb concentration was ~5.2 wt% and remained almost constant in the dendrite cores along the height. Similar Nb concen-tration along the full height can be due to either the process param-eters used in this study (constant current and scanning speed) or small geometry of the samples (which led to less dwell time for

the bottom layers at temperatureN 800 °C). Therefore, it can be

as-sumed that there was no back diffusion of Nb from the interdendritic areas to the dendrite cores in the layer-by-layer samples up to 50 layers.

To rationalize the lack of the Nb back diffusion at the bottom layers during STC, it is worth considering that the back diffusion was dependent on the diffusion distance, time available for diffusion, and diffusivity of

solute in the solid [41–43]. Here, the diffusion distance was about half

of the dendrite arm spacing (~0.5–1.5 μm) and the time for diffusion

was the holding time of the bottom layers in the chamber, which was

ap-proximately 2.5 h in around 950–1000 °C (see the data obtained from

EBM logﬁles inFig. 13). It should be considered that this temperature

range was recorded by the thermocouple beneath the build plate, thus it was not precisely the temperature of the bottom layers during the EBM process.

On the other hand, the cooling rate was also attributed to the ther-mal history of the part in the EBM process, which was more complex

than the simple thermal gradient during the solidiﬁcation. In fact,

under vacuum, there are two modes of heat transfer consisting of

a) radiative loss (Eq.(2)) from the top surface and b) heat conduction

to the build plate which was very signiﬁcant (Eq.(3)), seeFig. 14a.

How-ever, during theﬁnal cool down, the helium injection into the chamber

changed the mode of heat loss from radiation to convention from the top of the part, while the bottom of the system remained conductive [44]. However, there were no process models in the literature to capture these complex transients.

Heat loss by radiation can be calculated by Eq.(2)[45]:

Q1¼ εσ T4h−T4c

A ð2Þ

where Th= hot part absolute temperature (K), Tc= cold surrounding

absolute temperature (K), A = surface area of the specimen (m2_),_σ

(Stefan-Boltzmann constant) = 5.6703 × 10−8(W/m2_K4_{), and}_{ε =}

emissivity coefﬁcient of the part.

Fig. 12. a) SEM (BSE mode) images from bottom, middle and top of the T5L25 sample, EDS point analysis (a1–a5, b1–b5, c1–c5), and b) the Nb concentration results from three different heights in the T5L25 and T5L50 samples.

(12)

Heat loss by conduction can be calculated by the Eq.(3)[45]:

Q2¼ −kA Th−Tc

dx

ð3Þ where Q = heat transfer rate (w), K = thermal conductivity (w/m·k), A

= surface area of the specimen (m2_{), and dT/dx = temperature}

gradi-ent (K/m).

Considering the physical properties of Alloy 718 (Table 3) and by

as-suming the simpliﬁed heat transition phenomenon during sample

manufacturing (Fig. 14), heat conduction to the stand was more pre-dominant than heat radiation from the top surface. For instance, in the

single wall sample with 10 layers (with a height ~735μm), heat

conduc-tion is approximately 3.2 × 102_{times more than radiation, which means}

heat radiation is negligible (according to Eqs.(2) and (3)). However,

further investigation using computational modelling can provide better comparison in the heat transition mechanisms during the process. Thus,

in general, such heat transition in_{ﬂuences the cooling rate from the}

liq-uid to solid and solid to solid state condition. Subsequently, the variation of the cooling rate along the build height affects microstructure details.

As shown in the high magni_{ﬁcation SEM image of the layer-by-layer}

sample containing 50 layers (Fig. 14b), the primary dendrite arm spac-ing increased along the increased build height, which can be due to faster heat conduction from the bottom than top layers (Fig. 14b).

In the EBM process, the build plate acted as a heat sink leading to a

unidirectional local heatﬂow [44]. During the solidiﬁcation, dendrites

grew opposite to the heatﬂow direction, which was perpendicular to

the liquid-solid interface [46]. Thus, the average values for PDAS were measured for each sample to understand the variation of the cooling rates during a single layer formation (the track-by-track samples) and along the build direction (the layer-by-layer samples). Based on the

PDAS values (by 15–25 measurements), an average and a standard

de-viation were calculated for each sample, as shown inFig. 15. Based on a

previous work [47], the cooling rate,Ṫ, during the solidiﬁcation process

of a dendritic microstructure can be estimated using Eq.(4);

λ ¼ A Ṫ−n ð4Þ

whereλ is the primary dendrite arm spacing (μm), A (μm) and n

(di-mensionless) are proportionality coef_{ﬁcients related to materials,}

which can be taken to be 122.6μm and 0.342 for Alloy 718, respectively

[48]. The use of these values was valid for the directionally solidiﬁed

(DS) materials in high cooling rate range of 3 × 104_{K/s to 3 × 10}6_K/s

[49,50].

Fig. 15a shows that the average cooling rate from theﬁrst deposited track to the last deposited one was slightly decreasing, despite the high standard deviation inherent in the EBM-manufactured materials. The slightly decreasing trend can be due to increasing the amount of the ma-terial by track addition in the samples leading to slower heat dissipation and lower cooling rate. In the layer-by-layer samples (T1LX, X: 5, 10, 25, and 50 layers) the average PDAS was lower at the bottom than at top of the sample. As a comparison, in the sample with 25 layers, PDAS was

1.96 ± 0.22_{μm for the bottom region (ﬁrst 35 μm from the stand)}

and 2.94 ± 0.31μm for the top region (1315 μm from the stand),

indi-cating that the cooling rate was higher at the bottom of the walls (Fig. 15b). The bottom layers had a direct contact with the build plate, where much higher thermal conductivity and lower temperature (about 1000 °C) were available compared to the melt pool temperature of Alloy 718. Thus, a higher cooling rate was obtained for the bottom layers [51]. Similar results were also observed for the thicker walls

con-tainingﬁve tracks (Fig. 15c). It shows that by increasing the wall

thick-ness toﬁve tracks, the same trend from the top to the bottom of the

samples was achieved. It was also noticeable inFig. 15(b–c) that by

in-creasing the wall thickness (T5L1 to T5L50), the cooling rate decreased which was due to more heat accumulation in the thicker walls than sin-gle walls.

3.6. Variation of phases and Nb segregation

The combined continuous cooling transformation (CCT) and time

temperature transformation (TTT) diagrams for the speciﬁc

composi-tion of Alloy 718 used in this study were developed using JMatPro (Fig. 16a). The solidus state temperature was predicted to be ~1350 °C

with the formation ofγ matrix at the dendrite cores (Fig. 16b). By

pro-ceeding the solidiﬁcation, the heavy elements like Nb, Mo, and Ti are

de-pleted from the matrix to the liquid and subsequently formation of common carbide phases such as NbC or TiC in interdendritic were Fig. 14. a) Schematic of the simpliﬁed heat losses during layer addition in the EBM process, and b) LOM and SEM (BSE mode) images of the etched layer-by-layer sample with 50 layers and the high magniﬁcation of PDAS in the top and bottom of the sample.

Table 3

Physical properties of Alloy 718.

Thermal conductivity (W/m-k) 26.6 [43] Emissivity coefﬁcient of Alloy 718 under vacuum (1275 K) ~0.16 [44]

Th(K) ~1635

(13)

observed, seeFig. 8a [52]. The segregation of Nb, Mo, and Ti from the interdendritic regions was in agreement with the literature [52]. By

the JMatPro a thermodynamic simulation software [39] inFig. 16a, and

predicted cooling rate by measuring PDAS which was˃9 × 107

K/h,γ″

andγ′ were assumed not to be present in the matrix, however a

practical investigations by high resolution techniques need to be per-formed to conclude a rigid hypothesis.

Nb in Alloy 718, is one of the most important elements with its

seg-regation coefﬁcient (deﬁned as the composition of the interdendritic

re-gions divided by that of the dendrite cores) as high as 4.30 [53]. The

0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5

Cooling rate (×10

5

K/s)

Distance from stand (mm)

T5L1 T5L5 T5L10 T5L25 T5L50 0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5

Cooling rate (×10

5

K/s)

Distance from Stand (mm)

T1L1 T1L5 T1L10 T1L25 T1L50 100 µm T1 T2 T3 T4 T5 T6 T7 T8 T9 T10 300 µm

a)

_b)

c)

50 layers

Build direction

n

oit

ce

ri

d

dli

u

B

Build direction

300 µm

Stand

0 0.5 1 1.5 2 2.5 3 3.5 4 0 1 2 3 4 5 6 7 8 9 10 11

(

et

ar

g

nil

o

C×

1

0

5

)s

/

K

Track number

T1L1 T3L1 T5L1 T7L1 T10L1

Fig. 15. Cooling rate measurement based on the PDAS values; a) in the track-by-track samples b) in the layer-by-layer samples, and c) layer-by-layer samples withﬁve tracks.

(14)

segregation of Nb to the interdendritic areas during the solidiﬁcation was in agreement with previous studies [6,15,52]. Content of Nb in the microstructure is highly depended on both the cooling rate (liq-uid-to-solid) which was very high in the EBM process and dwell time inside the chamber (solid state). Higher cooling rate caused lower value of the Nb segregation, since Nb had less time to segregate during the rapid cooling rate in the liquid-to-solid state. Generally, by increas-ing the dwell time at high temperatures and back diffusion of Nb from the interdendritic areas to dendrite cores, the Nb segregation decreased, which was not seen in the present study (Fig. 12) [53]. For each sample,

10 images were taken at a high magniﬁcation (×4000) to examine

Nb-rich phase, based on point counting method presented in ASTM 562-08

[54]. As shown inFig. 17a in the track-by-track samples, the Nb area

fraction did not signiﬁcantly alter along the track addition direction.

By considering large standard deviations inFig. 17(b–c) for the

Nb-rich phase content which was inherent in the EBM as-built materials, the Nb-segregation did not show any clear trend at the top and bottom

of the layer-by-layer samples in both thicknesses of one andﬁve tracks.

It is noticeable that by increasing the thickness of the walls (T5L1 to T5L50), the Nb segregation increased, which can be due to lower cooling rate in the thicker walls.

3.7. Variation of hardness

Based on the size of the applied load, the indent size of about 50μm

had appropriate distance from the edges of the tracks with 250–390 μm

melt pool width. In the track-by-track samples, the range of hardness

was reported to ~360–400 HV0.5. The variation of hardness did not

show any signiﬁcant trend from the ﬁrst track up to 10 tracks. As the

cooling rate of the liquid in front of the solidiﬁcation boundary

in-creased, aﬁner dendritic solidiﬁcation microstructure was generated.

Fig. 18(a–b) illustrates the hardness proﬁles of the 1, 5, 10, 25 and 50-layer deposits after performing the Vickers micro-indentation tests. The slight difference in the hardness values of different locations can

be attributed to the resemblance in the solidiﬁcation characteristics

and also to the fact that they experienced STC after the solidiﬁcation.

Moreover, the difference in hardness at bottom layers (ﬁrst 20 μm

from the stand) compared to the top region (3200μm) in the

layer-by-layer samples (both thin and thick walls) were low (˂50 HV0.5).

The slightly higher (~11%) hardness at the bottom of these samples

can be possibly associated with theﬁner microstructure due to the

faster cooling rate and also possibility of precipitation of main

strength-ening phases (e.g.,γ″) as a result of STC. The bottom compared to top

(15)

layers during the EBM process experienced more heat from STC, since

the past solidiﬁed layers confront with an aging treatment process

which motivated theγ″ precipitate phase. It possibly resulted in

differ-ent size and fraction ofγ″ at the top and bottom regions. Measuring size

and fraction of the precipitate can provide information about the corre-lation of PDAS and precipitate size with the hardness values which are the scopes for future studies.

4. Conclusions

The effect of successive thermal cycling (STC) on the microstruc-ture, texmicrostruc-ture, phase evolution and cooling rate based on primary dendrite arm spacing (PDAS) of the EBM-manufactured Alloy 718 were investigated by manufacturing track-by-track and layer-by-layer samples. The deposited tracks became continuous solid mate-rials after adding three adjacent tracks using constant scanning speed (1.25 m/s) and current (7 mA) along the tracks. The effect of STC in the track-by-track samples displayed that the dendrite direc-tion in the overlap zone changed around 90° from the build direcdirec-tion towards either the adjacent track or scanning direction. In addition,

the primary dendrite arm spacing (PDAS) was slightlyﬁner (~35%)

in the one-time melted area than the overlap zone of a track. Owing to STC, the cross section of the EBM-manufactured samples

showed the epitaxial growth of columnar grains with strong〈001〉

texture from previous layers and nucleation of new elongated grains in the overlap zones. The difference in PDAS observed in both the

thin (one track) and thick (ﬁve tracks) layer-by-layer samples

indi-cated that the cooling rate was different in the various heights and it showed slightly higher cooling rate in the bottom than top layers. It can be concluded that STC affected the microstructure evolution. By increasing the wall thickness, the cooling rate seemed to be re-duced. The results from the Nb-rich phase measurement displayed

that STC was not signiﬁcant in different height of the layer-by-layer

samples due to the limited number of layers. The hardness proﬁle

along the walls (both the single and thick walls) was slightly higher

(~11%) at the bottom layers due to eitherﬁner microstructure than at

the top layers, or precipitation ofγ″ due to STC which are the scopes

for future investigations. In addition, in the light of this study, a path-way to produce a homogenized microstructure along the build direc-tion by changing key process variables (e.g., scanning speed, beam current) in the EBM process can be achieved.

Author contributions

As theﬁrst Author, Paria Karimi has performed all the experimental

characterization, analyzed all the results, designed the structure of all articles, and had the main responsibility in writing the articles. Co-authors contributed in formulating concepts and ideas, planning the structure and article editing.

Acknowledgments

The authors would like to thank Dr. Anders Snis from Arcam AB and Mr. Jonas Olsson from University West for sharing their knowledge in running the EBM machine. The authors also would like to thank Mr. Dunyong Deng from Linköping University and Mr. Stefan Gustafsson from Chalmers University for helping in some of the EBSD mapping. The funding from the "Simulation and Control of Material affecting Pro-cesses (SiCoMap)" research group, "KK foundation", and "SUMAN-Next" project is highly acknowledged.

References

[1] C.T. Sims, Norman S. Stoloff, William C. Hagel, Wiley: Superalloys II: High-Temperature Materials for Aerospace and Industrial Power - Chester T. Sims, Nor-man S. Stoloff, William C. Hagel, 1987.

[2] J. Karlsson, M. Norell, U. Ackelid, H. Engqvist, J. Lausmaa, Surface oxidation behavior of Ti–6Al–4V manufactured by Electron Beam Melting (EBM®), J. Manuf. Process. 17 (1) (2015) 120–126.

[3] E. Sadeghimeresht, P. Karimi, P. Zhang, R. Peng, J. Andersson, L. Pejryd, S. Joshi, Iso-thermal Oxidation Behavior of EBM-Additive Manufactured Alloy 718, (presented at the 9th International Symposium on Superalloy 718 and Derivatives, Pittsburgh, Pennsylvania, USA) 2018.

[4] V. Petrovic, J.V.H. Gonzalez, O.J. Ferrando, J.D. Gordillo, J.R.B. Puchades, L.P. Griñan, Additive layered manufacturing: sectors of industrial application shown through case studies, Int. J. Prod. Res. 49 (4) (Feb. 2011) 1061–1079.

[5] H.E. Helmer, C. Körner, R.F. Singer, Additive manufacturing of nickel-based superal-loy Inconel 718 by selective electron beam melting: processing window and micro-structure, J. Mater. Res. 29 (17) (Sep. 2014) 1987–1996.

[6] W.J. Sames, K.A. Unocic, R.R. Dehoff, T. Lolla, S.S. Babu, Thermal effects on micro-structural heterogeneity of Inconel 718 materials fabricated by electron beam melt-ing, J. Mater. Res. 29 (17) (Sep. 2014) 1920–1930.

[7] M.M. Kirka, K.A. Unocic, N. Raghavan, F. Medina, R.R. Dehoff, S.S. Babu, Microstruc-ture development in electron beam-melted Inconel 718 and associated tensile prop-erties, JOM 68 (3) (Mar. 2016) 1012–1020.

[8] M.M. Kirka, P. Nandwana, Y. Lee, R.R. Dehoff, Solidiﬁcation and solid-state transfor-mation sciences in metals additive manufacturing, Scr. Mater. 135 (Supplement C) (Jul. 2017) 130–134.

[9] B. Cheng, Y.K. Chou, Thermal Simulations for Cooling Rate Mapping in Electron Beam Additive Manufacturing, Nov. 2015, V02AT02A013.

(16)

[10]I. Yadroitsev, I. Smurov, Selective laser melting technology: from the single laser melted track stability to 3D parts of complex shape, Phys. Procedia 5 (Jan. 2010) 551–560.

[11]A. Segerstark, Additive Manufacturing Using Alloy 718 Powder: Inﬂuence of Laser Metal Deposition Process Parameters on Microstructural Characteristics, 2015.

[12] R.D. Murphy, E.C. Forrest,‘A Review of In-situ Temperature Measurements for Addi-tive Manufacturing Technologies’, presented at the NCSL International Workshop & Symposium, Saint Paul, MN, United States, 2016.

[13] B. Lane, S. Moylan, E. Whitenton, L. Ma, Thermographic measurements of the com-mercial laser powder bed fusion process at NIST, Rapid Prototyp. J. 22 (5) (2016) 778–787.

[14] S. Price, K. Cooper, K. Chou, Evaluations of temperature measurements in powder-based electron beam additive manufacturing by near-infrared thermography, Int. J. Rapid Manuf. 4 (1) (Jan. 2014) 1–13.

[15] P. Karimi, T. Raza, J. Andersson, L.-E. Svensson, Inﬂuence of laser exposure time and point distance on 75-μm-thick layer of selective laser melted alloy 718, Int. J. Adv. Manuf. Technol. (Sep. 2017) 1–9.

[16] A. Gasser, G. Backes, I. Kelbassa, A. Weisheit, K. Wissenbach, Laser additive manufacturing: Laser Metal Deposition (LMD) and Selective Laser Melting (SLM) in turbo-engine applications, Laser Tech. J. 7 (2) (Feb. 2010) 58–63.

[17] J.C. Heigel, B.M. Lane, Measurement of the Melt Pool Length During Single Scan Tracks in a Commercial Laser Powder Bed Fusion Process, Jun. 2017, V002T01A045.

[18] M.M. Kirka, F. Medina, R. Dehoff, A. Okello, Mechanical behavior of post-processed Inconel 718 manufactured through the electron beam melting process, Mater. Sci. Eng. A 680 (Jan. 2017) 338–346.

[19] R.R. Dehoff, M.M. Kirka, W.J. Sames, H. Bilheux, A.S. Tremsin, Site speciﬁc control of crystallographic grain orientation through electron beam additive manufacturing, Mater. Sci. Technol. 31 (8) (Jun. 2015) 931–938.

[20]R.R. Dehoff, M.M. Kirka, F.A. List, K.A. Unocic, W.J. Sames, Crystallographic texture engineering through novel melt strategies via electron beam melting: Inconel 718, Mater. Sci. Technol. 31 (8) (Jun. 2015) 939–944.

[21] C. Körner, H. Helmer, A. Bauereiß, R.F. Singer, Tailoring the grain structure of IN718 during selective electron beam melting, MATEC Web of Conferences, vol. 14, 2014, p. 08001.

[22] A. Strondl, M. Palm, J. Gnauk, G. Frommeyer, Microstructure and mechanical proper-ties of nickel based superalloy IN718 produced by rapid prototyping with electron beam melting (EBM), Mater. Sci. Technol. 27 (5) (May 2011) 876–883.

[23] JMatPro, Sente Software Ltd., Surrey Technology Centre, United Kingdom,http:// www.sentesoftware.co.uk/jmatpro.aspx.

[24]Y. Lee, M. Nordin, S.S. Babu, D.F. Farson, Effect ofﬂuid convection on dendrite arm spacing in laser deposition, Metall. Mater. Trans. B Process Metall. Mater. Process. Sci. 45 (4) (Aug. 2014) 1520–1529.

[25] C. Körner, A. Bauereiß, E. Attar, Fundamental consolidation mechanisms during se-lective beam melting of powders, Model. Simul. Mater. Sci. Eng. 21 (8) (2013).

[26] P. Karimi Neghlani, SLM Additive Manufacturing of Alloy 718: Effect of Process Pa-rameters on Microstructure and Properties, 2016.

[27] H. Gu, H. Gong, D. Pal, K. Raﬁ, T. Starr, B. Stucker, Inﬂuences of energy density on po-rosity and microstructure of selective laser melted 17-4PH stainless steel, Solid Freeform Fabrication 2013, pp. 474–489 , Austin, TX.

[28] V. Gunenthiram, P. Peyre, M. Schneider, M. Dal, F. Coste, R. Fabbro, Analysis of laser– melt pool–powder bed interaction during the selective laser melting of a stainless steel, J. Laser Appl. 29 (2) (May 2017), 022303.

[29] C.A. Schneider, W.S. Rasband, K.W. Eliceiri, NIH image to ImageJ: 25 years of image analysis, Nat. Methods 9 (7) (Jul. 2012) 671–675.

[30] H.-M. Jia, X.-H. Feng, Y.-S. Yang, Microstructure evolution and growth orientation of directionally solidiﬁed Mg–4 wt% Zn alloy with different growth rates, Acta Metall. Sin. Engl. Lett. 30 (12) (Dec. 2017) 1185–1191.

[31]M.B. Djurdjevič, M.A. Grzinčič, The effect of major alloying elements on the size of the secondary dendrite arm spacing in the as-cast Al–Si–Cu alloys, Arch. Foundry Eng. 12 (1) (2012) 19–24.

[32] Sindo Kou, Welding Metallurgy, 2nd edition, 2002.

[33]Y. Lee, Simulation of Laser Additive Manufacturing and Its Applications, The Ohio State University, 2015.

[34] R. Trivedi, W. Kurz, Dendritic growth, Int. Mater. Rev. 39 (2) (Jan. 1994) 49–74.

[35] R. Trivedi, Y. Shen, S. Liu, Cellular-to-dendritic transition during the directional so-lidiﬁcation of binary alloys, Metall. Mater. Trans. A 34 (2) (Feb. 2003) 395–401.

[36] S. Ghosh, L. Ma, N. Ofori-Opoku, J.E. Guyer, On the primary spacing and microsegregation of cellular dendrites in laser deposited Ni–Nb alloys, Model. Simul. Mater. Sci. Eng. 25 (6) (Sep. 2017), 065002.

[37] W.J. Sames, F.A. List, S. Pannala, R.R. Dehoff, S.S. Babu, The metallurgy and processing science of metal additive manufacturing, Int. Mater. Rev. 61 (5) (Jul. 2016) 315–360.

[38] J.-W. Park, J.M. Vitek, S.S. Babu, S.A. David, Stray grain formation, thermomechanical stress and solidiﬁcation cracking in single crystal nickel base superalloy welds, Sci. Technol. Weld. Join. 9 (6) (Dec. 2004) 472–482.

[39] A. Mostafa, I. Picazo Rubio, V. Brailovski, M. Jahazi, M. Medraj, Structure, texture and phases in 3D printed IN718 alloy subjected to homogenization and HIP treatments, Metals 7 (6) (May 2017) 196.

[40] Bey Vrancken, Study of Residual Stresses in Selective Laser Melting, Université Libre, Bruxelles, 2016.

[41]J.E. Matz, T.W. Eagar, Carbide formation in alloy 718 during electron-beam solid freeform fabrication, Metall. Mater. Trans. A 33 (8) (Aug. 2002) 2559–2567.

[42] J.N. DuPont, A.R. Marder, M.R. Notis, C.V. Robino, Solidiﬁcation of Nb-bearing super-alloys: part II. Pseudoternary solidiﬁcation surfaces, Metall. Mater. Trans. A 29 (11) (Nov. 1998) 2797–2806.

[43]J.N. DuPont, The inﬂuence of solid state diffusion on microstructural development during solidiﬁcation, Defect Diffus. Forum 266 (2007) 157–169.

[44]W. Sames, Additive Manufacturing of Inconel 718 Using Electron Beam Melting: Processing, Post-Processing, & Mechanical Properties, (Thesis) 2015.

[45] O. Levenspiel, The three mechanisms of heat transfer: conduction, convection, and radiation, Engineering Flow and Heat Exchange, Springer, Boston, MA 1984, pp. 161–188.

[46] S. Li, H. Xiao, K. Liu, W. Xiao, X. Han, Melt-pool motion, temperature variation and dendritic morphology of Inconel 718 during pulsed- and continuous-wave laser ad-ditive manufacturing: a comparative study, Mater. Des. 119 (Apr. 2017) 351–360.

[47] P. Karimi, D. Deng, E. Sadeghimeresht, J. Olsson, J. Ålgårdh, J. Andersson, Microstruc-ture development in track-by-track melting of EBM-manufacMicrostruc-tured alloy 718, Pro-ceedings of the 9th International Symposium on Superalloy 718 & Derivatives: Energy, Aerospace, and Industrial Applications, Springer, Cham 2018, pp. 643–654.

[48]J.K. Tien, T. Caulﬁeld, Superalloys, Supercomposites, and Superceramics, Academic Press, Boston, 1989.

[49] T. Antonsson, H. Fredriksson, The effect of cooling rate on the solidiﬁcation of INCONEL 718, Metall. Mater. Trans. B Process Metall. Mater. Process. Sci. 36 (1) (Feb. 2005) 85–96.

[50] Y. Zhang, B. Huang, J. Li, Microstructural evolution with a wide range of solidiﬁcation cooling rates in a Ni-based superalloy, Metall. Mater. Trans. A 44 (4) (Apr. 2013) 1641–1644.

[51]X. Gong, X. Wang, V. Cole, Z. Jones, K. Cooper, K. Chou, Characterization of micro-structure and mechanical property of Inconel 718 from selective laser melting, Jun. 2015, V001T02A061.

[52]A. Strondl, R. Fischer, G. Frommeyer, A. Schneider, Investigations of MX andγ′/γ″ precipitates in the nickel-based superalloy 718 produced by electron beam melting, Mater. Sci. Eng. A 480 (1–2) (May 2008) 138–147.

[53]X. P., H. Y., G. T., W. S., Z. H., Segregation and diffusion behavior of niobium in a highly alloyed nickel-base superalloy, Trans. Nonferrous Met. Soc. China 21 (11) (Nov. 2011) 2402–2407.

[54] ASTM E562-08, Standard Test Method for Determining Volume Fraction by System-atic Manual Point Count, ASTM International, West Conshohocken, PA, 2008www. astm.org.