Anisotropic fatigue properties of Alloy 718 manufactured by Electron Beam Powder Bed Fusion

In this study, Alloy 718 specimens manufactured by Electron Beam Powder Bed Fusion process are subjected to two di ﬀ erent post-treatments to have di ﬀ erent microstructural features. Low cycle fatigue testing has been performed both parallel and transverse to the build direction. EB-PBF Alloy 718 exhibits anisotropic fatigue behaviour; the fatigue life is better along the parallel direction compared to the transverse direction. The anisotropy in fatigue life is related to the anisotropy in the Young's modulus. The pseudo-elastic stress vs. fatigue life approach is presented as a potential solution to handle anisotropy in fatigue life assessment of additively manufactured engineering components.


Introduction
The interest in additive manufacturing (AM) of metals has consistently grown among both industrial and academic research groups across the world in the last decade.Metal AM technology is still maturing and evolving, yet the high interest is primarily due to design related advantages offered by AM for the low-volume-sector.With AM, particularly powder bed fusion (PBF), the design space has expanded considerably enabling manufacturing of topologically optimized structures, lattice structures and other generative designs easier and cost efficient.AM is poised to expand rapidly in the aviation industry, with applications such as new parts and repairs [1].AM processes inherently have complex physics that often result in anisotropic and/or location specific microstructures, which are different from cast and wrought microstructures of the same alloy [2].While the design advantages of AM are obvious, the mechanical behaviour and performance of the AM material need to be characterized and understood in depth, in relation to the microstructure, before AM parts could be used extensively in critical applications.With increasing part complexity and criticality of AM parts there is an urgent need in a thorough understanding of the fatigue properties.It is imperative considering that more than half of all the failures in aircraft components have been fatigue related [3].
Alloy 718, since its introduction in 1950s, has evolved into the most utilized superalloy in the industry [4].It is an iron-nickel-based superalloy, a sub-class of nickel-based superalloys, that is precipitation strengthened.In Alloy 718, γ" is the primary strengthening precipitate, but it also has γ' precipitates that contribute to the strength.The alloy also has δ phase, that forms at the expense of γ", which is often precipitated in a controlled manner for grain refinement and improved notch sensitivity.Excessive amount of δ phase is detrimental for the mechanical performance of the alloy.Other phases such as Laves, niobium carbide (NbC) and titanium nitride (TiN) can also exist in the alloy depending on the processing route [5,6].
Electron beam based PBF (EB-PBF) process has been successful in processing a variety of materials including superalloys such as Alloy 718, Alloy 247, Alloy 282, Alloy 625 and CMSX-4 [7][8][9][10].Alloy 718, with its status as the workhorse superalloy, is the material on which most of the EB-PBF research has been focused on so far.However, fatigue research on EB-PBF processed Alloy 718 is limited [11][12][13][14][15].Only one published research, so far, is on low cycle fatigue (LCF) properties of EB-PBF processed Alloy 718 [11], in which it has been demonstrated that the columnar microstructure of EB-PBF processed Alloy 718 exhibits anisotropic behaviour under both monotonic and cyclic loading conditions.In fact, only limited research is available on LCF of Alloy 718 processed by any AM technique [16][17][18][19][20][21].Apart from the evaluation of fatigue performance using rotating bending and bending fatigue, which the Metallic Materials Properties Database Development and Standardization (MMPDS) discourages for the purpose of design and analysis of structures in aerospace systems [22], other fatigue studies on PBF Alloy 718 has been focused on high cycle fatigue performance evaluating the influence of surface characteristics due to part orientation [23][24][25][26], texture [27], notches [28,29] and defects [28,30].
The aim of this work, therefore, is to evaluate the room temperature LCF performance of EB-PBF processed Alloy 718 and relate it to the microstructural characteristics such as phase constitution, texture etc.For this purpose, EB-PBF processed Alloy 718 is subjected to two different post-treatments and tested parallel and transverse to the build direction.Furthermore, a pseudo-elastic stress approach is evaluated and presented to handle the anisotropic behaviour exhibited by the columnar microstructure.

Specimen manufacturing
Specimen blanks were manufactured, in the form of cylindrical rods and cuboidal blocks, using an Arcam A2X Electron Beam Melting (EBM) system running EBM Control software V4.2.76.All the individual specimen blanks were bundled into a group within EBM Control, to be molten together rather than as individual parts.A raster scanning strategy using the Inco 4.2.76 theme, provided by Arcam, with speed function 63 and 125 µm hatch distance was implemented for the melting.The beam current and velocity for melting are controlled by the heat model algorithm in EBM Control as a function of the scan length and speed function.The scanning direction was rotated by 90°e very layer and the layer thickness used was 75 µm.The build started once the preheat temperature reached 1025 °C.The build configuration used for manufacturing LCF specimen blanks is shown in Fig. 1; three such builds were manufactured to obtain the required number of specimens for the test program.The feedstock used was gas atomized Alloy 718 powder having a nominal chemical composition listed in Table 1 and particle size range of 45-106 µm.

Post-treatment
All the specimen blanks were post-treated by hot isostatic pressing (HIP) and solution treatment (ST) and ageing.HIP was carried out in a Quintus QIH-21 HIP unit, while ST and ageing were carried out in a vacuum furnace.Two different post-treatment routines as listed in Table 2 were utilized in this work to evaluate the effect of post-treatment on mechanical properties.Hereafter, PT-1 is referred to as "standard treatment" and PT-2 as "repair treatment".The shorter ageing cycle used in this investigation is based on studies on cast Alloy 718 [31].The repair treatment is a simulation of multiple repair welding cycles, a strategy that was used to evaluate effect of multiple repair welding cycles on properties of cast and wrought Alloy 718 [32].Furthermore, the two post-treatments have been chosen to include HIP to ensure that the defects resulting from the E-PBF processing do not affect the properties.The detrimental effect of different types of defects from E-PBF processing of Alloy 718 have already been investigated by the authors and published elsewhere [13,33].

Fatigue testing
Strain controlled LCF tests were performed, at room temperature in accordance with ASTM E606/E606M [34], using an Instron 8802 servohydraulic machine with 8800MT controller and LCF3 software.An Instron 2620-602 clip on extensometer was attached at the gauge section to measure the strains over 12,5mm.LCF test specimens were extracted along the build direction from the cylindrical rods and in the transverse direction from the cuboidal blocks of the LCF build Fig. 1.Button head type specimens having a gauge diameter of 6.35 mm and gauge section length 13.2 mm were manufactured from the blanks, as shown in Fig. 2. LCF tests were performed using total strain ranges between 0.5% and 2% and a strain ratio R ε = 0; six specimens were tested in the transverse direction in the repair treatment condition and seven specimens, each, were tested in the other three conditions.The straining cycle followed a triangular wave form at a constant frequency of 0.5 Hz.If the measured plastic strain was less than 0.01% after 43,200 cycles, the testing was switched to load controlled cycling at 5 Hz.A 20% drop in the peak load from that of the stabilized hysteresis loop was used as the failure criterion for the test, after which the specimens were broken Fig. 1.Build configuration for vertical and horizontal LCF specimen blanks.Note: Z is the building direction.apart by applying a tensile load to reveal the fracture surfaces.Some of the specimens, however, were kept unbroken to extract metallographic sections to investigate crack propagation path.On such specimens, metallographic sections were extracted by electro discharge machining (EDM) leaving the remainder of the specimen intact, which were then broken apart to reveal the fracture surfaces.

Fractography and metallography
Fractographic investigation was performed using an Olympus SZX9 stereomicroscope and a Zeiss EVO 50 scanning electron microscope (SEM) fitted with an Oxford XMax N 20 mm 2 energy-dispersive x-ray spectroscopy (EDS) detector.
Metallographic samples were cut using a Struers Secotom 10 precision cutting machine or by wire-EDM.Samples were hot mounted, ground, and polished using standard metallographic procedures followed by a final vibropolishing with a 0.02 µm silica suspension.Electrolytic etching was performed with oxalic acid at 3 V for 5-10 s to reveal the microstructure.Microstructure analysis was carried out using a Zeiss AX10 light optical microscope (LOM) and a Zeiss Gemini 450 field emission gun (FEG) SEM fitted with an Oxford ULTIM MAX 100 mm 2 EDS and Oxford Symmetry electron back scatter diffraction (EBSD) detector.Analysis and imaging were performed using both Back Scattering Electron (BSE) and Secondary Electron (SE) modes.EBSD data was analyzed for texture information using Aztec Crystal v1.1 software.

Microstructure
In the present study, the mechanical test specimens were extracted by machining, and therefore only the microstructure representative of the test volume is presented here.The microstructure, after the two post-treatments, is presented in this section.The microstructure investigation was performed on several metallographic samples extracted from both the cylindrical and the cuboidal specimens.The microstructure was identical in both the cylindrical and the cuboidal specimens.The grains were columnar parallel to the building direction having an average grain width of 192 ± 69 µm and length in the order of millimeters spanning several layers.EBSD grain orientation mapping indicated a strong 〈1 0 0〉 texture along the building direction and random orientation distribution in the transverse directions.The EBSD inverse pole figure (IPF) showing the texture and the grain width information, representative of all the material conditions investigated, is presented in Fig. 3; the texture and the grain width were similar in all the investigated conditions, but are not presented here for brevity.Oxide (rich in Al, Ti) inclusions were present both in spherical form (< 10 µm) and in shapes with high aspect ratio (widths in 20-250 µm range and thickness < 10 µm) Fig. 4. The inclusions were randomly distributed in the investigated metallographic sections, with the high aspect ratio type lying parallel to the layers.Gas porosity (< 50 µm) were distributed randomly throughout the microstructure.All these microstructural features are consistent with reported literature on EB-PBF processing of Alloy 718 [14,35].
After the standard treatment, acicular δ phase was present at the columnar grain boundaries and at intra-granular sites as in Fig. 5(a) and (b).The δ phase particles at the grain boundaries, in general, were relatively smaller in size and higher number density compared to the intra-granular sites.Carbides (NbC) were present both at the grain boundaries and at intra-granular sites as seen in Fig. 5(c) in the form of vertically aligned strings along the building direction.The strengthening precipitates are shown in Fig. 5(d).After the repair treatment, however, acicular δ phase was spread evenly throughout the material.The δ phase particles were larger in both size and quantity compared to the standard treatment (compare Fig. 5(b) and Fig. 6(a)), consistent with four more hours of treatment in the δ phase precipitation temperature regime for Alloy 718.Correspondingly, the strengthening precipitates were smaller in size compared to the standard treatment (compare Fig. 5(d) and Fig. 6(c)).The grain boundary δ phase particles were smaller than the intra-granular ones, similar to the standard treatment (see Figs.

First cycle properties
The first loading cycle was started in the tensile direction and hence was used to evaluate the yield strength as well as the Young's modulus of the different material conditions; the results are presented in Table 3 along with the number of specimens used for the evaluation.The yield strength was evaluated using the data from specimens that experienced a plastic strain of at least 0.2% during the first loading cycle.Since the γ matrix in Alloy 718 has an FCC structure, that usually does not exhibit strain rate dependence, the properties from the first cycle at different strain ranges can be treated as equivalent to monotonic properties.The properties are anisotropic, with clear differences between the two directions, as expected for a columnar microstructure.The Young's modulus was ~30% lower in the parallel direction than the transverse direction.The strong 〈1 0 0〉 texture along the parallel direction is responsible for the lower modulus [36].The yield strength was higher for the standard treatment compared to the repair treatment in both the directions.The formation of δ phase consumes the amount of niobium available for the formation of γ" strengthening precipitates [5], and therefore the repair treated condition that has significantly higher amount of δ phase consequently has lower yield strength.

Cyclic stress evolution
Cyclic stress evolution and mid-life hysteresis loops for the different material conditions tested are presented in Figs.7 and 8 for a selection of strain ranges.The figures indicate that the cyclic properties are also anisotropic, following the same trend as the monotonic properties.The cyclic stress evolution curves show that, in both post-treatment conditions, a higher stress range is required along the transverse direction than the parallel direction to achieve a specific applied strain range.Such a difference is, as expected, consistent with the difference in modulus between the directions.The mid-life hysteresis loops show that at lower strain ranges the stress response is either fully elastic or undergoes elastic shakedown, whereas, at higher strain ranges there is significant cyclic plasticity.Such a response correlates well with the monotonic properties evaluated from the first loading cycle; the material with higher modulus and lower yield strength experiences higher cyclic plasticity and vice-versa.Similar anisotropic cyclic plasticity behaviour has been reported for EB-PBF Alloy 718 at 650 °C [11].
The cyclic stress evolution curves show that at lower strain ranges the materials exhibit a pronounced level of cyclic saturation is until failure.However, at higher strain ranges the materials undergo limited cyclic hardening and then continuously soften until failure.Wrought Alloy 718 also exhibits such stable cyclic response at lower strain ranges and initial hardening followed by softening at higher strain ranges [37].In the present study, there is no difference in terms of cycle dependent softening or hardening between the material conditions with different amounts of δ phase.Such cycle dependent softening is a typical behaviour of precipitation strengthened materials due to shearing of strengthening precipitates [38].
The stress ratio (R σ ), based on true stresses, for the first cycle and the mid-life cycle are presented in Table 4 for a few selected strain ranges.R σ is calculated based on true stress, instead of engineering stress to account for the instantaneous change in area during cyclic loading, to evaluate the bias in stress response to the biased applied strain.In the first cycle, in general, there is a tensile bias in the stress  response (R σ > −1) corresponding to the tensile bias in the applied strain range (R ε = 0).With increasing applied strain range, the bias tends towards symmetry; however, in the transverse direction symmetry is surpassed and even a compressive bias is attained.Such a response indicates the existence of a tension-compression asymmetry, which is typical of anisotropic materials such as single crystal nickel superalloys [39,40].This change in the stress response from tensile bias towards symmetry (and eventually to a compressive bias in the transverse direction) is due to the increasing cyclic plasticity corresponding to the increase in the applied strain range.Accordingly, the transverse direction that has higher Young's modulus, and therefore higher cyclic plasticity, exhibits a faster change in the bias compared to the parallel direction.Furthermore, the stress ratio for the mid-life cycles indicate that, in general, there is cycle-dependent mean stress relaxation.

Cyclic stress-strain curve
The cyclic stress-strain (CSS) curve is obtained by plotting the midlife stress amplitude (σ a ) and strain amplitude (ε a ).The cyclic Ramberg-Osgood model is given by where H' and n' are cyclic strength coefficient and hardening exponent, respectively.The CSS curve for the respective material conditions is presented, together with stress-strain responses during the first cycle of specimens that have significant plastic strains, in Fig. 9 and the cyclic Ramberg-Osgood model constants are listed in Table 5. Cyclic softening is observed in the standard treatment condition for both the tested directions, whereas, in the repair treatment condition the CSS curve follows the monotonic curve indicating that the material undergoes neither cyclic softening nor cyclic hardening.The difference in the softening behaviour, between the standard and the repair treatments, Note: σ YS is computed as R p 0.2 offset strength in the tensile direction.
Fig. 7. Cyclic stress evolution at different applied strain ranges.could be explained by the differences in the strengthening precipitates described in Section 3.1.In the standard treatment condition, the strengthening precipitates provide sufficient strengthening effect during the first loading cycle; however, undergo shearing, dissolution due to multiple shearing events, etc. that is typical under cyclic loading and therefore result in the softening behaviour.Whereas in the repair treatment condition, due to the extensive precipitation of δ phase, both the volume fraction and the size fraction of the primary strengthening γ" phase could be lower as noted earlier.Therefore, the strengthening effect during the first loading cycle is lower, to start with, than in the standard condition.Furthermore, the shearing of neither the relatively smaller and fewer strengthening precipitates nor the extensive amount of δ phase lead to the same magnitude of softening as in the standard treatment condition.

Strain-life relationship
The strain-life relationship is based on that the total strain amplitude is an additive partition of the elastic and plastic strain amplitudes. (2) The elastic and plastic strain amplitudes are given by Eqs. ( 3) and ( 4) respectively.
. ( ) The elastic strain amplitude is estimated from the mid-life stress amplitude and the Young's modulus using the Hooke's law.The plastic strain amplitude is, then, the difference between the applied total strain amplitude and the elastic strain amplitude estimated as above.The Fig. 8. Mid-life hysteresis loops at different applied strain ranges.

Table 4
Stress ratio (R σ ) in response to different applied strain ranges.parameters of the strain-life relationship can be obtained by linear regression of the strain components, independently, as a function of life as per ASTM E739 [41].
The strain-life relationship constants for the different material conditions are listed in Table 6 and the corresponding strain-life curves are presented in Fig. 10 together with reference wrought data [42].The data points presented as open symbols indicate crack initiation from an oxide inclusion at the surface, while the data points presented as solid symbols indicate crack initiation from the slip at the surface.Similar crack initiation from oxide inclusion, formed during EB-PBF processing, and therefore deterioration of fatigue performance has been reported [12,13].In both the directions, the fatigue performance is similar for the standard treatment and the repair treatment even though a difference was observed in the cyclic stress evolution, the cyclic plasticity, and the CSS behaviour.The apparent difference seen in the parallel direction is due to the differences in the features that initiate the fatigue crack and not represent the microstructure related differences due to the two post-treatments investigated.
Similar result, that of limited influence of simulated repair treatments on LCF properties, has been reported for cast and wrought Alloy 718 [32].Fig. 11 is a representative of samples tested along the parallel direction with an inclusion-based crack initiation and crack initiation due to slip at the surface.
The material in the parallel direction has higher fatigue life than the transverse direction under the strain-controlled LCF condition, as shown in Fig. 10.Furthermore, the material in the parallel direction has better fatigue performance than the wrought material, while the material in the transverse direction has similar performance to that of the wrought material.Similar behaviour has been reported for EB-PBF Alloy 718 at elevated temperature as well [11].The anisotropic fatigue behaviour in PBF metals, under stress-controlled high cycle fatigue (HCF) condition, due to the orientation dependent as-built surface roughness and the orientation of the defects w.r.t to the loading direction and the building direction are well established [43][44][45][46].Both the sharp edges of the LoF defects and the notch-like valleys of the asbuilt surface lead to high stress concentration and act as crack initiation sites and deteriorate the HCF performance.In general, the part orientation that leads to a higher surface roughness has inferior fatigue performance.Similarly, the fatigue performance is poor when the LoF defects are oriented perpendicular to the loading direction.In the present study, however, the anisotropy in LCF behaviour is due to the process-dependent texture-induced anisotropy of the Young's modulus.The better LCF performance of the material in the parallel direction is due to the lower stress ranges required to achieve a specific strain range than the material in the transverse direction, which is a consequence of the lower Young's modulus as described in Section 3.2.1.Therefore, it can be assumed that the resolved shear stress acting on slip planes that is responsible for slip (and dislocation multiplication and their movement), for a specific strain range, is lower for the material in parallel direction than the transverse direction.To verify this assumption, a pseudo-elastic stress estimated from the strain amplitude and Young's modulus (σ pseudo-elastic = ε a .E) is plotted against life, as a S-N type graph, in a double logarithmic scale as in Fig. 12.All the data points for the parallel direction and transverse direction, in both posttreatment conditions, merge to a linear relationship in a double logarithmic plot between pseudo-elastic stress and life.In addition, the scatter in this data is clearly related to the inclusion-based crack initiations.Such a relationship between strain, anisotropic modulus due  to crystallographic orientation and fatigue life has been shown to exist for single crystal nickel superalloys in the past [40,47,48].For the wrought reference data the Young's modulus for specimens at each of the data points was unknown; therefore, an average value is assumed from literature [49] and utilized in the pseudo-elastic stress estimation.Even the wrought data, with approximated estimates, merges with the other material conditions.Therefore, it can be inferred that the slip resistance under fatigue loading conditions is similar in all the four material conditions.Furthermore, in single crystal superalloys shear stresses and shear strains estimated to act on the slip systems have shown good correlation to fatigue life [50,51], which further strengthens the argument regarding similarity in slip resistance for the different material conditions evaluated in the present study.Based on these findings, it would be worth investigating the fatigue response of the EB-PBF manufactured Alloy 718, having a columnar microstructure, under stress-controlled conditions.

Crack path investigation
Fig. 13(a)-(c) are representative fracture surfaces of specimens tested along the parallel and the transverse direction; the fractographs show similar features in the standard and the repair treatment conditions.All the samples tested along the parallel direction have a typical transgranular crack growth appearance as shown in Fig. 13(a), which are confirmed by analysis of metallographic cross-sections of the crack path (Fig. 13(d) and (g)).The cross-sections in Fig. 13(g) reveal that there is secondary cracking along some of the grain boundaries.The angular difference between the slip line on either side of the secondary crack, visible in Fig. 13(g), indicate that this could be a high-angle grain boundary.Similar crack-branching, along high angle grain boundaries, for parallel direction has been reported for dwell-fatigue crack propagation testing at 550 °C [15].
The samples tested along the transverse direction have a columnar appearance in the fracture surfaces as shown in Fig. 13(b) and (c).The crack propagation, in most cases, appears to be at an angle to the columnar grains Fig. 13(b); between being either completely perpendicular or parallel.Metallographic sections and EBSD IPF maps perpendicular to the crack plane, shown in Fig. 13(e), (f), (h) and (i), reveal that the crack growth occurs by a combination of transgranular and intergranular modes in all the cases, irrespective of the macro appearance of the fracture surface.The angular difference between slip lines on either side of the crack, at the intergranular sections of the crack path as shown in Fig. 13(i), is high.Therefore, it is possible that intergranular cracking occurs whenever the crack tip encounters a highangle grain boundary.Based on this tendency for intermittent intergranular cracking and secondary cracking along high-angle boundaries that is discussed above, further in-depth research is required to understand if there are other metallurgical reasons, than strain incompatibility, for intergranular cracking at room temperature.
An earlier study has shown that the crack propagation rate is slower along the parallel direction than the transverse direction [15]; however, the two possible orientations of the crack tip w.r.t to the columnar grains in the transverse direction were not investigated.In the present study, only one out of the 13 specimens tested in transverse direction had crack propagation completely parallel to the columnar grains as in Fig. 13(c).Since only a few specimens had crack propagation being exactly parallel or perpendicular to the columnar grains, it was difficult to draw any meaningful conclusion about the crack propagation behaviour between the different orientations of the crack front to the columnar grains, and how it affects the fatigue life.Furthermore, investigating the differences in crack propagation behaviour is outside the scope of the current study.However, based on the fracture surface appearance dedicated fatigue crack propagation tests are needed in order to study if the material exhibits anisotropy in crack propagation rates based on the orientation of the crack tip w.r.t to the columnar grains.

Conclusions
In this work, LCF properties of Alloy 718 manufactured by EB-PBF process and subjected to two different post-treatments have been investigated.The tests were conducted at room temperature under straincontrolled conditions with a tensile bias in the applied strain range such that R ε = 0. Alloy 718 manufactured by EB-PBF process has fatigue properties that is comparable to, or exceeding that of, wrought material.
• The cyclic properties exhibit anisotropy in stress evolution and cyclic plasticity (hysteresis loops) between the parallel and transverse directions, corresponding to the respective Young's modulus and yield strength.
• The standard and repair treatment lead to different size and volume fraction of δ and γ" precipitates.Accordingly, the CSS behaviour is different between the two treatmentsstandard treatment leads to cyclic softening while the repair treatment leads to neither hardening nor softening.
• In strain-controlled fatigue conditions, the parallel direction out- performs the transverse direction i.e. has longer life.The difference in fatigue life between the standard and repair treatment is not significant.
• The pseudo-elastic stress vs. fatigue life approach indicates that the anisotropy in life is related primarily to the anisotropy in Young's modulus.Such an approach can, potentially, be used to handle anisotropy in fatigue life estimation of additively manufactured engineering components.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
5(b) and 6(b)).The carbide particles were similar in size as the standard treatment (compare Fig. 5(b) and Fig. 6(b)) and were vertically aligned.

Fig. 4 .
Fig. 4. (a) LOM image of gas pores.(b) SE image of a gas pore at high magnification.(c) SE image of a spherical inclusion.(d) SE image of high aspect ratio oxide inclusion.

Fig. 5 .
Fig. 5. Microstructure in standard treatment condition.(a) SE image showing grain morphology.(b) SE image of area marked in (a) showing intra and inter-granular δ phase.(c) BSE image showing vertically aligned carbides.(d) SE image showing strengthening precipitates.

Fig. 6 .
Fig. 6.Microstructure in repair treatment condition.(a) SE image showing grain morphology and δ phase distribution.(b) SE image of area marked in (a) showing smaller inter-granular δ phase and NbC.(c) SE image showing strengthening precipitates.

Fig. 10 .
Fig. 10.Strain amplitude vs. fatigue life plots.Note: Open symbols indicate inclusion-based crack initiation, and solid symbols indicate crack initiation due to slip at the specimen surface.For specimens tested at ε a = 0.875%, fractography was not performed as specimens were preserved for crack path investigation.

Fig. 11 .
Fig. 11.(a) SE image of fracture surface of sample with crack initiation at an oxide inclusion at the surface.(b) High magnification SE images of the area marked in (a) showing crack initiation site.(c) SE image of fracture surface of sample with crack initiation at surface.(d) High magnification SE images of the area marked in (c) showing crack initiation site.
study through the SUMAN Next project (20160281).The support from Sandvik Machining Solutions AB, Quintus Technologies AB and GKN Aerospace Engine Systems AB with Alloy 718 powder, HIP, and heat treatment procedures, respectively, are acknowledged.The authors are grateful for the contributions of Jonas Olsson for helping with manufacturing EB-PBF specimens.The authors would also like to thank Mats Högström and Håkan Backström for their help with setting up and performing the LCF tests.The authors thank Prajina Bhattacharya and Peter Karlsson for their inputs towards presentation of the fractographic information and strain-life relationship calculations, respectively.

Fig. 13 .
Fig. 13.Stereomicroscope fractographs of specimen in (a) parallel direction (b) transverse direction (crack front perpendicular to columnar grains) (c) transverse direction (crack front parallel to columnar grains).(d)-(f) EBSD IPF maps from crack initiation surface corresponding to images (a)-(c).LOM images at crack tip in a (g) parallel specimen (h) transverse specimen.(i) Area in (h) at higher magnification.Note: Specimens were tested at different strain ranges.IPF maps are w.r.t crack plane (not building direction).SC -Secondary Cracking, TG -Transgranular, IG -Intergranular.Readers are referred to high resolution images in the web version of the article for details.

Table 1
Nominal chemical composition of the Alloy 718 powder used in this work (in weight percent).

Table 2
Post-treatment details.

Table 3
First cycle properties.

Table 6
Strain-life relationship constants.