research papers
Texturebased residual stress analysis of laser powder bed fused Inconel 718 parts
^{a}Bundesanstalt für Materialforschung und prüfung, Unter den Eichen 87, 12205 Berlin, Germany, ^{b}Australian Nuclear Science and Technology Organisation, New Illawara Road, Lucas Heights, NSW 2234, Australia, ^{c}HelmholtzZentrum Hereon, MaxPlanckStrasse 1, 21502 Geesthacht, Germany, ^{d}Laboratory for Neutron Scattering and Imaging, Paul Scherrer Institut, Forschungsstrasse 111, Villigen 5232, Switzerland, and ^{e}Universität Potsdam, Institut für Physik und Astronomie, KarlLiebknechtStrasse 2425, 14476 Potsdam, Germany
^{*}Correspondence email: jakob.schroeder@bam.de, giovanni.bruno@bam.de
Although layerbased additive manufacturing methods such as laser powder bed fusion (PBFLB) offer an immense geometrical freedom in design, they are typically subject to a buildup of internal stress (i.e. thermal stress) during manufacturing. As a consequence, significant residual stress (RS) is retained in the final part as a footprint of these internal stresses. Furthermore, localized melting and solidification inherently induce columnartype grain growth accompanied by crystallographic texture. Although diffractionbased methods are commonly used to determine the RS distribution in PBFLB parts, such features pose metrological challenges in their application. In theory, preferred grain orientation invalidates the hypothesis of isotropic material behavior underlying the common methods to determine RS. In this work, more refined methods are employed to determine RS in PBFLB/M/IN718 prisms, based on crystallographic texture data. In fact, the employment of directiondependent elastic constants (i.e. stress factors) for the calculation of RS results in insignificant differences from conventional approaches based on the hypothesis of isotropic mechanical properties. It can be concluded that this result is directly linked to the fact that the {311} lattice planes typically used for RS analysis in nickelbased alloys have high multiplicity and less strong texture intensities compared with other lattice planes. It is also found that the length of the laser scan vectors determines the surface RS distribution in prisms prior to their removal from the baseplate. On removal from the baseplate the surface RS considerably relaxes and/or redistributes; a combination of the geometry and the scanning strategy dictates the subsurface RS distribution.
Keywords: additive manufacturing; electron backscattered diffraction; principal stress; residual stress.
1. Introduction
Layerwise additive manufacturing methods such as laser powder bed fusion (PBFLB) have attracted major interest from both academia and industry within the past decade; this interest is based on the immense geometrical design flexibility in the manufacturing of dense parts in a single manufacturing step (Attaran, 2017). In fact, the aerospace and gas turbine industry sectors demand complex geometries to increase the efficiency of lightweight construction in hightemperature applications. Further, the geometrical freedom enables the design of sophisticated internal cooling geometries in such parts. Owing to its excellent weldability (Lingenfelter, 1989) paired with its potential in hightemperature applications up to 650 °C (Collier et al., 1988), the alloy Inconel 718 (denoted IN718) is an established candidate for PBFLB processing (Volpato et al., 2022). IN718 is a niobium, aluminium and titaniumcontaining precipitationhardenable Ni–Cr–Fe–Mobased superalloy. Its high strength is achieved by the precipitation of γ′′ (Ni_{3}Nb, tetragonal D0_{22} crystal structure) and γ′ [Ni_{3}(Al,Ti), cubic L1_{2} crystal structure] phases during aging heat treatments (Cozar & Pineau, 1973).
However, the layerwise nature of the PBFLB process has certain drawbacks that undermine the applicability of the technique: Manufactured parts may suffer from defect formation such as porosity, caused by either gas inclusions or lack of fusion (Foster et al., 2018). Another inherent problem of the technique is the significant surface roughness (Foster et al., 2018) of the parts. Although the formation of defects can nowadays be greatly reduced by the selection of appropriate process parameters (Foster et al., 2018), the surface finish remains a critical aspect for engineering applications (Kasperovich et al., 2021). The localized melting and solidification mechanisms of the layerwise technique also inevitably induce large internal stresses during manufacturing (Kruth et al., 2004; Mercelis & Kruth, 2006). These are based on the temperature gradient mechanism in combination with the thermal contraction (i.e. shrinkage) during cooling of the previous layer, due to mechanical constraint by the substrate plate (Mercelis & Kruth, 2006; Kruth et al., 2004; Ulbricht et al., 2020). In extreme cases, the internal stresses may lead to cracking or delamination during production (Yadroitsev & Yadroitsava, 2015). In most cases, residual stress (RS) of high magnitude is retained in asbuilt parts as a footprint of these internal stresses during manufacturing (Schröder, Evans et al., 2021).
Diffractionbased methods allow the nondestructive determination of the RS distribution of full parts. In principle, lattice spacings (d^{hkl}) are measured and subsequently used to calculate a lattice strain by comparing them with a stressfree reference value (d_{0}^{hkl}). In the case of laboratory Xray diffraction (XRD) experiments, plane stress can be assumed, i.e. the normal stress component vanishes within the of the radiation, and a precise knowledge of d_{0}^{hkl} is not required (Spieß et al., 2009). However, whenever triaxiality of the stress state cannot be excluded, a precise knowledge of d_{0}^{hkl} is indispensable (Withers et al., 2007), in particular when using penetrant radiation such as neutrons (well suited to the determination of 3D stress fields).
With knowledge of the relationship between elastic lattice strains and macroscopic stress provided by the diffractionelastic constants (DECs) (GnäupelHerold et al., 2012), RS can be determined from measured strains on the basis of Hooke's law (Hauk, 1997). For anisotropic crystals the DECs depend on the hkl reflection used to measure the lattice spacing (GnäupelHerold et al., 2012). The DECs can be either determined experimentally or, more commonly, calculated from singlecrystal elastic tensor data of the material of interest (Hauk, 1997). In the past, several graininteraction models for polycrystalline aggregates have been developed to calculate such DECs from singlecrystal data. To name a few, these include the models of isostrain (Voigt, 1889) and isostress (Reuss, 1929), the average suggested by Hill (1952), or the Kröner model (Kröner, 1958) based on the solution of the Eshelby inclusion problem (Eshelby, 1957). Apart from the Kröner model, preferred grain orientation and graintograin interactions are neglected in these models (GnäupelHerold et al., 2012). However, from Eshelby's theory (Eshelby, 1961) it is known that the strain/stress response of a single grain depends on the elastic properties and shape of the surrounding grains (GnäupelHerold et al., 2012). The formulation of Hooke's law in the form by Dölle & Hauk (1978, 1979) overcomes the problem and considers the by introducing the stress factors.
If one wants to select an appropriate model for the calculation of the DECs, it is commonly accepted that the Kröner model provides a reasonable agreement to experimental data for equiaxed polycrystalline IN718 (Schröder, Mishurova et al., 2021) and IN625 (Wang et al., 2016). However, another consequence of the localized melting and solidification during the PBFLB process is the columnar grain growth as reviewed by Volpato et al. (2022). In such cases, it has been experimentally shown that the Reuss model represents the materials behavior for PBFLB/M/IN718 more accurately (Schröder et al., 2022; Schröder, Mishurova et al., 2021). In fact, the usage of DECs based on the Kröner model may lead to RS exceeding the yield strength of asbuilt PBFLB/M/IN718 (Pant et al., 2020; SerranoMunoz, Fritsch et al., 2021). Additionally, strong crystallographic textures are characteristic for PBFLB/M/IN718 (Gokcekaya et al., 2021), since the f.c.c. crystals grow along the 〈100〉 directions (Chalmers, 1964). On the one hand, this dependence of the texture on the heat flow allows the texture to be tailored by choosing appropriate scanning strategies and beam parameters (Gokcekaya et al., 2021). On the other hand, the presence of texture requires the usage of the stress factors for the determination of RS. Yet, in the open literature it is common to neglect crystallographic texture when determining RS in PBFLB/M/IN718. Beyond that, the validity of the general assumption that the directions of principal strain/stress are governed by the main geometrical axes should be additionally questioned (Mishurova, SerranoMunoz et al., 2020).
It becomes clear that several metrological challenges of the RS determination in PBFLB/M/IN718 need to be tackled. In this article the strain and RS distribution in asbuilt PBFLB/M/IN718 prisms (manufactured with two different scan strategies) will be determined using a combined approach of laboratory Xray, highenergy synchrotron and neutron diffraction. These investigations are carried out on material identical to that used in the in situ loading studies reported by Schröder et al. (2022). Hence for the isotropic case, the DECs are known to be well predicted by Reuss for the 311 reflection, which mitigates one of the key unknowns for the accurate RS determination. The distribution of subsurface principal strain and stress is evaluated by strain pole figures and a subsequent eigenvalue decomposition considering texturebased stress factors. Finally, the RS calculations encompassing the crystallographic texture of the two scan strategies are compared with approaches neglecting the presence of texture. Some metrological consequences for RS determination in PBFLB/M/IN718 prisms are discussed.
2. Material and methods
2.1. Sample manufacturing
The subjects of this study are horizontally built PBFLB/M/IN718 prisms (110 × 13 × 13 mm^{3}) manufactured using an SLM 280 (SLM Solutions Group AG, Lübeck, Germany). The specimens were manufactured with their longest direction within the build plane but tilted by 12° with respect to the build plate edges (Fig. 1). The baseplate was preheated to 200°C and the processing parameters suggested by SLM Solutions were applied: laser power P = 350 W, scanning velocity v = 800 mm s^{−1}, spot size diameter of 0.08 mm defocused by 4 mm and hatch spacing h = 0.15 mm. Two different scanning strategies with an interlayer rotation of 90° were applied to produce the specimens (Fig. 1): in the first variant, the scanning tracks were aligned parallel to the specimen edges (H_{0°}), whereas the scanning pattern was rotated by 45° relative to the prism edges for the second variant (H_{45°}). The specimens were all used in the asbuilt state (i.e. no heat treatments were applied).
2.2. Microstructural analysis
2.2.1. Electron backscattered diffraction
As depicted in Fig. 1, BD–T (build–transverse directions) cross sections were extracted from sister specimens for microstructural analysis. These cross sections were ground to 1200 grit with SiC abrasive paper followed by subsequent 9, 3 and 1 µm polishing steps. The final polishing step was performed using a 0.04 µm active oxide polishing suspension (OPS, Struers GmbH, Crinitz, Germany). The samples were then mounted in an LEO 1530VP (Carl Zeiss Microscopy GmbH, Oberkochen, Germany) scanning electron microscope, equipped with an electron backscatter Bruker Nano e^{−}Flash HD 5030 detector (Bruker Corporation, Billerica, USA). For the electron backscattered diffraction (EBSD) analysis, the sample was tilted by 70° and kept at a working distance of approximately 18 mm. The acceleration voltage of the electron beam was 20 kV. In essence, for the bulk microstructure an area of 4 × 3 mm^{2} was probed over an 800 × 600 pixel map, i.e. with a pixel size of 5 µm. In contrast, the near surface maps were acquired at a higher magnification (250×) with a pixel size of 1.5 µm, i.e. over an approximate probed area of 1.2 × 0.9 mm^{2}. For data acquisition and indexing the ESPRIT (version 1.94) package from Bruker Nano was used. For data postprocessing, the opensource MTEX toolbox (Bachmann et al., 2011) installed within MATLAB (The MathWorks Inc., Natick, USA) was utilized. A misorientation angle of 10° was used as the threshold to define highangle grain boundaries, whereby grains containing fewer than ten pixels were excluded from the analysis.
The grain boundaries were then smoothed using the default kernel (25 iterations). In addition, nonindexed pixels were filled by their nearest neighbor and denoising was performed using a variational spline filter.
2.2.2. Neutron diffraction texture measurements
The bulk texture measurements were performed at the KOWARI strain scanner located at the Australian Nuclear Science and Technology Organisation (ANSTO) in Lucas Heights. For the measurements, cylinders with a diameter and a height of 8 mm were extracted from the center of the threaded region of the H_{0°} and H_{45°} tensile specimens (see Schröder et al., 2022). The neutron wavelength of 1.4 Å was selected from the 400 reflection of the Si monochromator. With the cylinders fully immersed in the beam, measurements were run with an approximate 3 × 3 (°) mesh (in φ–χ space) over the intervals χ [0, 90]° and φ [0, 360]°. Three detector positions 2θ = 43°, 2θ = 67° and 2θ = 82° corresponding to the 111, 200, 220 and 311 reflections were selected with an acquisition time of ∼2 s. Data postprocessing and analysis were performed using MTEX. First, the pole figures were normalized; subsequently, orientation distribution functions (ODFs) were calculated using a halfwidth of 5°. These ODFs were exported into ISODEC (GnäupelHerold, 2012). From these ODFs, the {200}, {220}, {111} and {311} pole figures were calculated. Furthermore, the strength of these textures can be quantified by the texture index J_{ODF} [equation (1)] as implemented in MTEX (Mainprice et al., 2015). The orientation distribution function can be described as the function f(g). In this context, the texture index J_{ODF} can be defined as the integral of f(g)^{2} over the rotationally invariant volume element dg:
This definition involves the square of f, which type of functional is called an L^{2}norm (Mainprice et al., 2015). For a uniform distribution, J_{ODF} returns a value of 1. For a single orientation, it becomes an infinitely large value (Mainprice et al., 2015).
2.3. Texturebased RS analysis
RS analysis by diffractionbased methods rests on Bragg's law (Bragg & Bragg, 1913). The lattice spacing d^{hkl} can be effectively used as a strain gauge. From the comparison between the measured d^{hkl} and a reference lattice spacing d_{0}^{hkl}, the strain can be calculated as the relative difference (Withers et al., 2007). In this regard, Hooke's law can be written in the special form of Dölle & Hauk (1978, 1979) to determine the macroscopic RS 〈σ_{ij}〉 from lattice spacings d^{hkl} [equation (2)]:
where subscript 33 denotes the laboratory direction . F_{33ij}(φ, ψ, hkl) are the stress factors introduced by Dölle & Hauk (1978, 1979), who assigned them the term F_{ij}, thereby formally missing their fourth rank tensor character. Deriving from the definition, Mishurova, Bruno et al. (2020) showed that the notation is somewhat imprecise, as in fact in the literature ɛ often replaces ɛ_{33}, and F_{ij} is effectively taken as a second rank tensor. For a material without a the stress factors are independent of the measurement directions ψ, φ; thus they become linear combinations of the DECs s_{1} and 1/2s_{2} (Hauk, 1997). However, in the presence of a preferred crystallographic orientation, the F_{33ij}(φ, ψ, hkl) depend on the measurement directions. Similar to the DECs, the stress factors can be either directly measured by in situ tests or calculated from singlecrystal elastic properties using graininteraction models – note that GnäupelHerold et al. (2012) also used the misleading notation F_{ij}, while properly defining the stress factors. In the latter case, the ODF is required to account for the crystallographic texture of the material studied (Behnken & Hauk, 1991).
In this study, the texturedependent F_{33ij}(φ, ψ, hkl) were calculated with the software ISODEC (version 3.0; GnäupelHerold, 2012) on the basis of the Reuss model (Reuss, 1929) using the singlecrystal elastic constants of IN718 (c_{11} = 242.35 GPa, c_{12} = 139.73 GPa, c_{44} = 104.44 GPa) reported by Haldipur et al. (2004). For the orientation relationships used in this study, see Fig. 2(a).
2.3.1. Laboratory Xray diffraction
The surface RS measurements were performed with an Xstress G3 diffractometer (StressTech, Vaajakoski, Finland) at Bundesanstalt für Materialforschung und prüfung (BAM, Berlin, Germany). The system operates in modified χmode (see standard DIN EN 15305: 2009–01: Nondestructive testing – Test method for RS analysis by XRD) using two positionsensitive detectors, which are calibrated using copper powder. For all measurements, a Ø2 mm collimator and an acquisition time of 5 s were used. For detailed information on the measurement conditions, see Table S1 of the supporting information. The measurement plane for the surface measurements of the top surface (for H_{0°} and H_{45°}) is shown in Fig. 2(e). These measurements were performed pre and postremoval of the specimens from the baseplate to determine the redistribution of the surface RS. Data analysis was performed in the software Xtronic using a Pearson VII function to fit the diffraction peaks and determine the d^{311} values. As the classic analysis of the sin^{2}ψ method data does not allow the incorporation of texture, the calculation of the RS was performed by the matrix method (also referred to as the generalized sin^{2}ψ method) reported by Ortner (2009, 2011, 2014). For all measurements, the overdetermined system of linear equations [see equation (2)] was solved using the generalized linear model as implemented in statsmodels.api within Python (Seabold & Perktold, 2010). Since the of the Mn Kα radiation is at the maximum of ∼5 µm, the outofplane stresses were disregarded for the top (σ_{iBD} = 0) and side (σ_{iT} = 0) surfaces. Furthermore, the measurements carry the assumptions that the measured directions (φ = 0°, 90°) are principal, so that the inplane shear components vanish. As for surface measurements a precise knowledge of d_{0}^{hkl} is not required, the tensor equation (1) can be written in d^{hkl} form. This treatment yields an overdetermined set of linear equations with the unknowns d_{0}^{311}, σ_{TD} and σ_{LD} (top); and d_{0}^{311}, σ_{LD} and σ_{BD} (side).
2.3.2. Electrolytic layer removal
Incremental electrolytic layer removal was performed (after removal from the baseplate) at the side surfaces of H_{0°} and H_{45°}, as indicated in Fig. 2(e). A Kristall 650 electrolytic polishing device (ATM Qness GmbH, Mammelzen, Germany) was used, operated at a voltage of 30 V and a current of 2 A with a circular polishing diameter of 9 mm. The solution used for electropolishing consisted of 550 ml of saturated saline solution, 150 ml of water, 200 ml of ethylene glycol and 100 ml of ethanol. The depth after each removal increment was measured using an IDC series 543471B dial indicator (Mitutoyo Corporation, Kawasaki, Japan) with an accuracy of ±3 µm.
2.3.3. Synchrotron Xray diffraction
The synchrotron XRD measurements were performed at the white beam engineering materials science beamline P61A at the Deutsches ElektronenSynchrotron (DESY) in Hamburg, Germany [for details see Farla et al. (2022)]. A greatly simplified illustration of the basic instrument principle is shown in Fig. 2(b). Prior to the measurements, a diffraction angle of 2θ ≃ 11.946° was calibrated using NIST silicon powder. The specimens, mounted in an Eulerian cradle, were scanned in χmode using the energydispersive detector in the horizontal diffraction plane (ψ = χ − 90° for the detector in the horizontal diffraction plane). The specimens were measured at the top and side surfaces according to Fig. 2(e). The acquisition time varied between 10 and 20 s up to ψ = 60° and was increased to 20–40 s between ψ = 60 and 80°. The incoming beam was narrowed by the vertical and horizontal slits to a 0.5 × 0.5 mm^{2} In the diffracted beam, the slits narrowed the beam to 26 × 26 µm^{2} (for further details on the measurement conditions, see Table S1). was performed in the opensource software P61A:Viewer developed at the P61A beamline, using a pseudoVoigt function. Peaks under 100 counts were excluded from the analysis. The diffraction angle used was approximately 12°, giving a of 311 of τ_{0} ≃ 30 µm. As a consequence, stress triaxiality should not be neglected, though its influence on the obtained stress values is expected to be low.
The overdetermined set of linear equations was solved using the mean value of all measured d^{311} as stressfree reference d_{0}^{311}. Afterwards an eigenvalue decomposition was performed to determine the principal directions represented by the eigenvectors v_{1}, v_{2} and v_{3}, and eigenvalues σ′_{L}, σ′_{T} and σ′_{BD}. All calculations were repeated 10000 times, selecting a random value within the 95% confidence interval of the leastsquares solution to estimate an error band for the principal stress directions and magnitudes. The procedure is described in more detail by Fritsch et al. (2021). Although the choice of d_{0}^{311} as the mean value of all measured d^{311} values remains somewhat arbitrary, it may affect the absolute values of the principal stresses but not the principal stress directions.
2.3.4. Timeofflight neutron diffraction
Bulk d_{0}^{hkl} and strain measurements were performed at the pulse overlap timeofflight (TOF) diffractometer POLDI at the Swiss Spallation Neutron Source (SINQ) at the Paul Scherrer Institut (PSI), Villigen, Switzerland. A greatly simplified sketch of the POLDI measurement principle is shown in Fig. 2(c). POLDI uses a pulsed neutron beam with a 1D ^{3}He chamber detector. The detector is TOF and angle sensitive with an angular coverage of 2θ = 75–105°. The signal is integrated over the whole angular range. This implies that the strain component is averaged around ±7.5° from the scattering vector. All measurements were made using the 1.5 × 1.5 mm^{2} full width at halfmaximum collimator to define the diffracted beam. The incident beam shape was defined by the slit optics. A d_{0}grid was extracted from a sister specimen by electrical discharge machining as depicted in Fig. 2(e). The single cubes have the dimensions 3 × 3 × 3 mm^{3} and are connected in the grid to simplify their alignment. To fully immerse the gauge volume in the cuboids of the d_{0}grid, a 2.6 × 2.6 × 1.5 mm^{3} gauge volume was defined for the measurements of d_{0} along the three orthogonal directions BD, L and T. However, to obtain sufficient sampling statistics, a 1.5 × 1.5 × 20 mm^{3} matchstickshaped gauge volume was used to measure d^{hkl} along BD and T in the prism. For optimization, depending on the path length, the acquisition time was adjusted between 30 and 45 min.
Data analysis was performed using a Gaussian peak function within Mantid (Arnold et al., 2014). Additional information on the experimental setup and the data evaluation of POLDI can be found in the literature (Stuhr, 2005; Stuhr et al., 2006, 2005).
2.3.5. Monochromatic neutron diffraction stress analysis
Bulk residual stress determination was conducted using the KOWARI strain scanner located at the Australian Nuclear Science and Technology Organisation (ANSTO) in Lucas Heights. The principle of the technique is depicted in Fig. 2(d). In contrast to the pulsed white beam at POLDI, a specific wavelength (in our case 1.53 Å) is selected by a silicon monochromator. Using a diffraction angle of 2θ ≃ 90°, a 1.5 × 1.5 × 1.5 mm^{3} gauge volume was defined by slits in the incoming and diffracted beams. The positional accuracy was better than 0.1 mm. The detailed measurement conditions are listed in Table S1. Measurements of d_{0}^{311} were performed along the L direction of the central cube in the d_{0}grids of both H_{0°} and H_{45°}. To assess the RS distribution, an equally distributed 8 × 8 point grid was defined in the BD–T of H_{0°} and H_{45°} at the specimen midlength L = 55 mm, Fig. 2(e)]. In addition to measurements of d_{0}^{311}, the stress balance conditions based on these measured d^{311} values were applied to the T and BD components, using an inhouse developed Python script.
The obtained diffraction peaks were fitted using a Gaussian profile and the texturebased analysis of the RS was performed directly in the software ISODEC (GnäupelHerold, 2012). The set of linear equations is not overdetermined since we only measured the three orthogonal strain components ɛ_{BD}, ɛ_{L} and ɛ_{T}. Thus, the error on the stress is estimated by propagating the errors in d^{311} and d_{0}^{311}. Since neutron diffraction knowledge of d_{0}^{311} is required, the linear equation system must be expressed in the form [see equation (2)].
3. Results
3.1. Microstructure and texture
The orientation maps viewed along L acquired by EBSD of the specimens H_{0°} and H_{45°} are shown in Figs. 3(a)–3(c) and Figs. 3(e)–3(g). In addition, the calculated {200} pole figures (for the maps acquired on the cross section) are shown in Figs. 3(c1) and 3(g1). The nearsurface maps qualitatively reveal that no texture gradient towards the surface exists. However, they also show that the lateral and top surfaces of both H_{0°} [Figs. 3(a) and 3(b)] and H_{45°} [Figs. 3(e) and 3(f)] exhibit a degree of surface roughness, as no contouring was performed during manufacturing. The highest peaktovalley measure of the surface roughness based on the localized region (i.e. statistically very limited) of the EBSD maps in Fig. 3 is of the order of 70 µm. The neutron [Figs. 3(d) and 3(h)] and EBSD texture measurements [Figs. 3(c1) and 3(g1)] of the bulk yield similar {200} pole figures. In essence, a cubetype texture can be observed in both H_{0°} [Fig. 3(d)] and H_{45°} [Fig. 3(h)] specimens. Since the scanning vectors are aligned with the geometry in H_{0°}, the 〈100〉 directions are aligned with the L, T and BD directions. The texture strength of H_{0°} is characterized by the texture index J_{ODF}(H_{0°}) ≃ 1.8. While the texture intensities in the {200} pole figure are equal along L and T, the {220} pole figure shows that a mixed 〈100〉/〈110〉type texture is present along BD. Even though the 〈100〉/〈110〉type texture is preserved along BD in H_{45°}, the change of the scan pattern causes a 45° rotation of the cubetype texture around BD (i.e. 〈110〉/〈111〉type texture along L and T). This texture is characterized by a texture index J_{ODF}(H_{45°}) ≃ 2.1. EBSD as a surfacespecific technique provides spatial resolution to characterize grain morphology and texture. However, the calculation of a representative ODF is limited by the sampling statistics. In this context, the neutron diffraction texture measurements probed the entire volume of the cylinders (∼402 mm^{3}), rather than the 4 × 3 mm^{2} area probed by EBSD (the of the electron beam is only a few nanometres). Thus, although the textures determined by EBSD and neutron diffraction are in agreement, all subsequent texturebased RS determinations (i.e. bulk and surface) use calculated ODFs from neutron texture measurements, because the probed volume is millions of times larger. Such data show the strongest texture and thereby represent the worst case scenario of the influence of crystal orientation on the RS determination.
3.2. Stress factors
Taking into account the calculated {311} pole figures shown in Figs. 3(d) and 3(h), the stress factors F_{33ij}(φ, ψ, 311) of H_{0°} and H_{45°} are shown in Figs. 4(a) and 4(b) as a function of ψ and φ, respectively. As previously mentioned, the calculations based on the hypothesis of isotropic elasticity are linear combinations of the DECs s_{1} and 1/2s_{2} and show a linear dependence of F_{33ij} on sin^{2}ψ in the plane containing the load axis. The calculated F_{33ij} according to the texturebased Reuss model are very different for the two specimens H_{0°} and H_{45°}. As an effect of the difference in texture (Fig. 3), F_{33ij} is larger for H_{0°} up to sin^{2}ψ ≃ 0.5 but smaller above sin^{2}ψ ≃ 0.5 [Fig. 4(a)]. Note that the of the stress factors in Fig. 4(a) arises from the cubetype texture (see Fig. 3): the intensity in the {311} pole figures at φ = 0° and ψ = 45° is identical for H_{0°} [Fig. 3(d)] and H_{45°} [Fig. 3(h)]. In principle, the textures of H_{0°} and H_{45°} are akin, just rotated by 45° around the build axis. Therefore, the stress factors are also offset by 45° as they are weighted according to their orientation distribution function. In the plane perpendicular to the applied load, F_{33ij} is independent of φ for an isotropic material (= s_{1}), while becoming dependent on φ in the presence of crystallographic texture [Fig. 4(b)].
3.3. Xray diffraction: surface and subsurface RS
3.3.1. RS before removal from the baseplate
The surface RS maps (L and T directions) for a quarter of the sample surface of the prisms H_{0°} and H_{45°} are depicted in Fig. 5. The drop in RS close to the specimen edges (width = 6 mm, length = 54 mm) is associated with misalignment. If we ignore these points near the edges, an average maximum stress of 383 ± 28 MPa is present in H_{0°} along L prior to removal from the baseplate. In contrast, a minimum average stress of 255 ± 25 MPa is present in the T direction. For H_{45°} the appears broadly isotropic, as the average stresses have similar magnitude when considering the error: 355 ± 33 MPa along L and 305 ± 34 MPa along T.
3.3.2. RS after removal from the baseplate
Once the specimens are removed from the baseplate, stress redistribution and relaxation occur, due to distortion in the L direction: the surface longitudinal stress relaxes (from the edge up to L = 42 mm) to an average magnitude of 121 ± 17 MPa (≃ 68% relaxation) and 88 ± 54 MPa (≃ 75% relaxation) for H_{0°} and H_{45°}, respectively. However, close to the edges, a highermagnitude tensile RS of about 240 MPa is present, which introduces a comparable bending moment in the two specimens. Along the T directions, stress redistribution is negligible and only small relaxations of about 55 MPa (≃ 21%) in H_{0°} and 40 MPa (≃ 13%) in H_{45°} are observable.
3.3.3. Determination of subsurface principal stress
The strain pole figures acquired at the synchrotron beamline P61A are shown in Figs. 6 and 7 for the top (points 3–5) and side surfaces (points 8 and 9), respectively. From these subsurface strain pole figures, the inplane principal strain can be directly determined in a qualitative fashion. In all strain pole figures acquired close to the center (i.e. points 3, 4, 8, 9), a strain plateau at ±30° in φ is observable around the direction of maximum and minimum strain. This plateau begins to transform into a uniform `ring' of large strain at about 10 mm from the edges (i.e. stress state becomes transversely isotropic) of the top surface point 5. This observation is in line with the postremoval XRD measurements (Fig. 5). Further, the strain pole figures show that the direction of largest subsurface strain in H_{0°} coincides with the transverse direction T for measurements in the L–T plane (Fig. 6); the smallest strain (i.e. average slope of the ɛ versus sin^{2}ψ curve) is found along the longitudinal direction L. In contrast, the strain pole figures of H_{45°} in the L–T plane reveal a rotation of the inplane subsurface principal axes around BD towards the geometrical axes (Fig. 6). Such qualitative observations are confirmed quantitatively by the eigenvalue decomposition results as shown in Fig. 6. The smaller magnitude of the subsurface principal stress at measurement point 3 in H_{45°} corresponds to local stress relaxation induced by the layer removal performed on the side surface. In the case of the side surface measurements (7–9), the strain pole figures (Fig. 7) reveal the alignment of the maximum subsurface principal strain with BD irrespective of the scanning strategy used. The eigenvalue decomposition reveals an ∼120 MPa larger subsurface deviatoric principal stress difference σ′_{BD} − σ′_{T} in the H_{0°} specimen than in H_{45°}. Also in this case, the stress state is transversely isotropic with respect to BD (i.e. the stress difference σ′_{L} − σ′_{T} ≃ 0). The resulting subsurface principal stress values of all measured points 1–9 can be found in Table S1.
3.3.4. Layer removal
Although correction formulae for the determination of RS upon layer removal are available (Moore & Evans, 1958), for relatively shallow removal depths it is known that the differences of residual stress between the measured and corrected values are negligible. Therefore, Fig. 8 shows the uncorrected results of the layerremoval method measurements up to a depth of 700 µm (∼5% of the total thickness). For both specimens H_{0°} [Fig. 8(a)] and H_{45°} [Fig. 8(b)], the RS state at the surface is characterized by tensile stresses of small magnitude along BD and around 0 MPa along L. At shallow depths (first 100 µm), an increase of the stress is observed, until a stress plateau of σ_{BD} = 350 MPa and σ_{L} = 100 MPa is reached. This behavior is believed to be connected to the inherent surface roughness of the specimens, as the of Mn Kα radiation in IN718 is small. Even though the scanning strategy was different, the average stress at the plateau appears to be similar in the two specimens. Yet at shallower depths (e.g. 125 µm) the maximum stress is larger in H_{0°} (≃ 410 MPa) than in H_{45°} (≃ 330 MPa).
3.4. Neutron diffraction: bulk RS
3.4.1. The stressfree reference d_{0}^{311}
Spatially resolved measurements of d_{0}^{311} were performed on the d_{0}grid (see above) at the POLDI beamline. The results are shown in Fig. 9(a). No clear variation with respect to the build height or transverse direction can be observed for H_{0°} [Fig. 9(a)]. Even though the 2.6 × 2.6 × 1.5 mm^{3} gauge volume used for the POLDI measurements is close to full immersion, different d_{0}^{311} values were measured in the three directions [Fig. 9(a)]. However, a pointwise average for the three directions corresponds well to the L direction of the measured 3 × 3 × 3 mm^{3} cuboids. The overall average [dashed line in Fig. 9(a)] corresponds to the L direction at positional index 5. This average was used for the calculation of lattice strain from POLDI data for H_{0°}. In fact, such strain agrees with the strain determined by KOWARI using the d_{0}^{311} measured along L at positional index 5 [Figs. 9(b) and 9(c)]. As opposed to H_{0°}, the directional spread of d_{0}^{311} is much smaller in H_{45°}, yet the overall average has a slightly worse correlation (although still within the error bar) to the L direction in the center of the d_{0}grid (i.e. at positional index 5, see Fig. 10). For the determination of all subsequent bulk RS values (for H_{0°} and H_{45°}) from measurements at KOWARI, d_{0}^{311} along L at positional index 5 is used.
3.4.2. Stress mapping
The RS maps acquired at the strain scanner KOWARI in the cross sections displayed in Fig. 2(e) are shown in Fig. 11. It is evident from these measurements that the tensile RS close to the surface is balanced by compressive stress in the bulk. Furthermore, a slight asymmetry in the stress maps from left to right can be observed. The stress relaxation on removal from the baseplate results in a low stress (about 50 MPa) along L close to the sample surface (center of gauge volume 1.25 mm below the surface) in both specimens. Overall, the RS distributions look alike, except for a larger compressive stress preserved in the H_{0°} specimen.
4. Discussion
4.1. Influence of preferred grain orientation
In theory, the presence of crystallographic textures invalidates the use of methods based on the hypothesis of isotropic elastic behavior. Yet, in most cases the hypothesis of isotropic elastic constants is used. This holds true even though it is known that crystallographic texture is present in PBFLB/M/IN718 manufactured specimens (Volpato et al., 2022). Fig. 12 shows the d^{311}–sin^{2}ψ curves and their relative intensities at φ = 90° in the BD–L plane for H_{0°} and H_{45°}. In a case without such a distribution should exhibit linearity (Vanhoutte & Debuyser, 1993). In addition, the relative intensity should be nearly independent of ψ (Spieß et al., 2009), yet gradually decrease at higher ψ angles due to the grazing incidence. Instead, both d^{311}–sin^{2}ψ curves are nonlinear (especially for H_{45°}), i.e. show clear evidence of crystallographic texture. Such a nonlinearity has been recently observed for the {311} lattice planes in PBFLB/M/IN718 (Mishurova et al., 2018; SerranoMunoz, Fritsch et al., 2021), although the {311} lattice planes are supposed to behave in an isotropic manner. In fact, Mishurova et al. (2018) and SerranoMunoz, Fritsch et al. (2021) determined the RS using a linear fit. Although Mishurova et al. (2018) and SerranoMunoz, Fritsch et al. (2021) proved this to be a fair approximation, this approach may lead to significant errors in the calculation of RS, when compared with approaches fitting nonlinear functions (i.e. those considering texture) to d–sin^{2}ψ curves (Vanhoutte & Debuyser, 1993).
To quantify the difference between texturebased and isotropic calculations, we used both isotropic and texturebased calculations within ISODEC; the results are outlined in Table 1. The differences in the obtained RS are small and well within the error bar of the measurements. This corroborates the assumption made by Mishurova et al. (2018), SerranoMunoz, Fritsch et al. (2021) and Thiede et al. (2018). Most probably, the mild texture of the 311 reflection (maximum 1.6 m.r.d.) has a rather minor effect on the calculated RS and one could still use the hypothesis of isotropic elastic constants. In fact, the isotropic and texturebased calculations of the neutron diffraction and energydispersive RS data are comparable for H_{0°}. However, when stronger cubetype textures are modeled in MTEX and accounted for in the texturebased analysis of H_{0°}, the absolute stress difference between isotropic and texturebased calculations increases up to 80 MPa (Fig. 13). This difference is well beyond the error bar of the determination and is above 15% of the actual stress value. Especially since in highpower PBFLB (1000 W) strong cubetype textures (t ≃ 20) are realized (Zhong et al., 2023), texturebased methods should be employed in such a case. However, for texture indices J_{ODF} < 3, the effect of texture on the RS values seems to rapidly decrease (Fig. 13). It must be emphasized that this observation is based on modeled textures applied to experimental data possessing much lower crystallographic texture. In reality, it is practically impossible to produce material with different textures yet the exact same residual stress field using PBFLB.

Therefore, the general assumptions used for a diffractionbased analysis of RS must be checked on a casebycase basis (low texture factor, columnar grain shape). Whenever the ODF is known, the use of texturebased methods for the determination of RS is recommended. Once different reflections are used for RS analysis (e.g. energydispersive methods), texturebased approaches become unavoidable.
4.2. The scanning strategy determines the RS distribution
Several studies reporting surface RS distributions have shown that longer scan vectors lead to higher tensile RS in Ti6Al4V (Kruth et al., 2012; Ali et al., 2018) and IN718 (SerranoMunoz, Ulbricht et al., 2021). Furthermore, it is known that the larger principal residual stress is always parallel to the track of the scan direction in the final deposited layer while the specimen is attached to the baseplate for PBFLB/M/Ti64 (Levkulich et al., 2019). In contrast, Bayerlein et al. (2018) showed (for an unspecified scanning strategy) that the principal directions are approximately aligned in the direction of the sample edges for asbuilt PBFLB/M/IN718 cuboids. Similar observations have been made for PBFLB/M/IN625 structures, where the principal direction coincides with the main geometrical axis of the structure (Fritsch et al., 2021). However, Fritsch et al. (2021) showed that the determination of the principal stress is only independent of the choice of the measurement directions if one uses nine directions.
These observations seem to be transferrable to PBFLB/M/IN718. In H_{0°} the scanning vector was oriented along the length (110 mm) and width (13 mm) of the rectangular prism for alternate layers. The largest stress along the L direction in H_{0°} (Fig. 5) can thus be explained by the larger thermal gradient when scanning along this direction: in fact, the aspect ratio between the scan length of alternate layers is about 7. If the scan vectors become of equal length, as in the case of a 45° rotation to the geometrical axis in H_{45°}, the surface RS magnitudes along T and L become similar. Further, the scanning strategy influences the orientation of the surface principal stress axes relative to the geometrical axis. It is hypothesized that, prior to removal from the baseplate, the scanning direction dictates the subsurface principal stress direction (∼45° to L and T in H_{45°}). This would explain the equivalent surface RS values along the L and T directions prior to removal: both L and T lie at 45° from the principal axis.
4.3. RS redistribution on removal from the baseplate
In agreement with the present work, Thiede et al. (2018) found a similar relaxation pattern of the surface RS in horizontally manufactured IN718 prisms (with a rounded tip). Prior to removal from the baseplate, the surface RS had high tensile magnitude with insignificant changes across the specimen surface. On removal from the baseplate, an overall relaxation with a steep increase of the surface RS in the longitudinal direction towards the tip was found, irrespective of the scanning strategy applied [see also SerranoMunoz, Ulbricht et al. (2021)]. In contrast to our work, Thiede et al. (2018) observed the surface RS in the transverse direction to be the largest prior to removal and it additionally showed significant relaxation. However, both the specimen (20 × 20 mm^{2}) and the stripewise scanning strategy were substantially different compared with this study. Therefore, the disagreement with the present study outlines the influence of such aspects on the surface RS distribution.
Additionally, the surface RS values reported by Thiede et al. (2018) were significantly higher than those observed in our study. On the one hand this is connected to the choice of a Krönertype graininteraction model [see also Pant et al. (2020)]. In fact, SerranoMunoz, Ulbricht et al. (2021) showed that the use of the Reuss model for similar specimens yields a more sensible magnitude of surface and subsurface RS. On the other hand, this is – to a degree – also dependent on the geometry and the process parameters [i.e. the scanning strategy (Nadammal et al., 2021)]. Distortion measurements of this kind of sample geometry show that the sample tends to deform towards the tip (Thiede et al., 2018; Mishurova et al., 2018; SerranoMunoz, Fritsch et al., 2021). In addition, the distortion tends to be somewhat dependent on the scanning strategy (SerranoMunoz, Ulbricht et al., 2021).
Our synchrotron experiments reveal that the subsurface principal axes are aligned with the geometry if the scanning vectors are alternatingly parallel to L and T. However, the principal directions in the L–T plane are rotated by ∼13° from the main geometrical axes if the scanning vectors are oriented 45° to the geometry. This indicates that a `back rotation' of the subsurface principal components around BD occurs, due to the distortion on removal from the baseplate. This last finding would explain the similarity of the surface RS for H_{0°} and H_{45°} after removal from the baseplate (Fig. 5). The inherent distortion causes the geometry to influence the subsurface principal direction. As a consequence, the slight rotation, in conjunction with the strain plateau of ±30°, results in similar RS values along the geometrical axes. This is emphasized by the negligible difference between the subsurface deviatoric stress along T (σ_{T} − σ_{BD} = 373 ± 13 MPa) and the maximum principal components for H_{45°} (σ′_{T} − σ′_{BD} = 381 ± 18 MPa) measured at point 2. In neutron diffraction measurements, the detectors typically average over a range of ±15° (±7.5° from the diffraction vector). This average implies that any small difference between geometrical and stress axes would not influence the RS values.
On the other hand, one of the subsurface principal stress axes always remains aligned with BD irrespective of the scanning strategy. In fact, the laser beam parameters predominantly determine the RS distribution along BD, rather than the scanning strategy. This results in similar distortion along BD for different scanning strategies.
4.4. On the choice of the stressfree reference
A et al., 2007). It has thus been proposed to utilize different methods to crosscheck d_{0}^{hkl} values (Withers et al., 2007). In fact, the crosscheck between mechanically relaxed cubes and the application of theoretical boundary conditions such as the stress balance yields a suitable sanity check for the measured d_{0}^{hkl} values. However, for the applicability of the stress balance method it must be ensured that no spatial variation of d_{0}^{hkl} exists within the of interest (Withers et al., 2007). In the case of PBFLB it has been shown that no large variations of d_{0}^{hkl} across the specimen occur (SerranoMunoz et al., 2022; Bayerlein et al., 2018), at least when significant heat concentrations are avoided (Capek et al., 2022). This has also been observed for the specimens in this work (Fig. 9). The d_{0}^{311} values calculated from the application of the stress balance condition to bulk data are listed in Table 2. The use of the stressbalancebased d_{0}^{311} (instead of the one based on measurements of the coupons) would shift the calculated stress by about 70 MPa for H_{0°} and 30 MPa for H_{45°}. This may be because the surface RS was not included in the stress balance. In fact, if the surface RS (accounting for the surface roughness) is not included in the stress balance, a deviation between experimentally measured and theoretical d_{0}^{311} occurs (SerranoMunoz et al., 2022). The use of the postremoval RS values of the relaxed surface (Fig. 5) should shift the stressbalancebased d_{0}^{311} to smaller values. Although no spatial gradient of the experimentally determined d_{0}^{311} exists, a directional dependence is evident (Fig. 9). Such a directional dependence has been reported by other researchers for PBFLB/M/IN718 (Bayerlein et al., 2018; Thiede et al., 2018) and PBFLB/M/316L (Ulbricht et al., 2020). This direction dependence might arise from possible retention of macroscopic (if the gauge volume is not fully immersed in the cuboid) or intergranular stress (Withers et al., 2007). As we lack evidence of the cause of this directional dependence, we considered the global average of all d_{0}^{311} as an appropriate value. Yet, the fact that the gauge volume was close to full immersion implies the prevalence of intergranular over macro stress. If one accounts for the directional dependence of d_{0}^{311} [Fig. 9(a)], the RS values would shift in H_{0°} but not in H_{45°} (the directional variation is much smaller, see Fig. 10). Finally, the similarity between the XRDbased (where no precise d_{0}^{311} is required) and neutrondiffractionbased RS strongly indicates that the directionindependent d_{0}^{311} of the L component (being similar to the overall average) is appropriate in this special case.
of uncertainty for the determination of the bulk RS by neutron diffraction techniques may arise from inaccuracy of the stress (or strain)free reference (Withers

4.5. Throughthickness stress distribution
A et al., 2022). In fact, the mean roughness of PBFLB specimens manufactured without a contouring parameter set is reported to be in the range 10–25 µm (Fritsch et al., 2022; Mishurova et al., 2019; Sprengel et al., 2022). It is further known that high tensile stresses are usually present in the subsurface region (Bayerlein et al., 2018; SerranoMunoz et al., 2022; SerranoMunoz, Fritsch et al., 2021; SerranoMunoz, Ulbricht et al., 2021; Busi et al., 2021). The layer removal plus the XRD experiments we performed revealed a subsurface stress plateau rather than a peak stress. Interestingly, such behavior has also been found by SerranoMunoz et al. (2022) using neutron diffraction. Therefore, additional sample preparation (e.g. electro polishing) or use of highenergy XRD techniques is recommended to overcome such surface roughness effects (Mishurova et al., 2019).
of the stress profile within PBFLB manufactured alloys is the distribution close to the surface. Overall, the increase of the RS magnitudes in the subsurface region can be linked to the surface roughness of the parts, since the roughness contributes to a stress relaxation in the vicinity of the surface (SerranoMunozFig. 14 shows the throughthickness stress profiles for the BD and L components of the specimens H_{0°} and H_{45°}, combining surface XRD (layer removal) and bulk neutron data. The full profiles are drawn assuming symmetry of the surface and subsurface RS with respect to the sample center point. It becomes apparent that a strong RS gradient must be present at depths of 0.7–2.75 mm. In this context, SerranoMunoz et al. (2022) recently showed that the RS decreased at 1.4 mm depth from the lateral surfaces in a 20 × 20 mm^{2} prism produced with a 67°rotation scan strategy. However, SerranoMunoz et al. (2022) showed that the plateau below 1.4 mm displayed higher RS compared with our study. First and foremost, the buildup of RS in PBFLB/M/IN718 is known to depend on the build height [much larger for SerranoMunoz et al. (2022) than in the present study]: the addition of new layers produces tensile stress in the material directly below (Bayerlein et al., 2018). In fact, Pant et al. (2020) reported that the magnitude of RS depends on the build orientation of Lshaped specimens produced with a 13° interlayer rotation. The horizontally built specimen (10 mm build height) showed the lowest magnitudes of residual stress, while the largest magnitudes were found for the vertical build orientation (build height 55 mm). Secondly, the use of upskin (also referred to as contouring) processing is known to cause higher RS magnitudes in Ti6Al4V (Artzt et al., 2020). While the rotation scanning strategy used by SerranoMunoz, Ulbricht et al. (2021) would lead to lower RS values compared with other scanning strategies, the effect of the contour and the addition of layers prevails in the present case. Interestingly, the RS profiles observed by Pant et al. (2020) show a similar distribution in their horizontally built specimen: tensile stress is present at the side surfaces along BD, while it is observed along the short direction at the top surface of the structure. Note that the RS values reported by Pant et al. (2020) are not metrologically comparable to our study, because the diffraction elastic constants were calculated using the Kröner model.
5. Conclusions
This work discusses the texturebased determination of residual stress in asbuilt PBFLB Inconel 718 prisms. Different crystallographic textures were obtained by employing different scanning strategies. Scan vectors aligned with the specimen geometrical axes resulted in 〈100〉 inplane texture. In contrast, those rotated by 45° to these axes, while maintaining the 90° interlayer rotation, resulted in 〈111〉/〈110〉 inplane texture. Residual stress determination was performed by utilizing laboratory XRD methods and employing stress factors to account for the specimen texture. Additional laboratory Xray (layer removal) and neutron diffraction measurements provided further insight into the residual stress distribution after removal from the baseplate. Furthermore, subsurface principal stress was assessed by energydispersive synchrotron diffraction. The consequences of the presence of crystallographic texture on the residual stress determination were studied for both surface and bulkrelated measurements. The following conclusions can be drawn:
(1) Under the conditions used in this study (texture indices < 3), the preferred grain orientation (i.e. the crystallographic texture) has a negligible influence on the determined residual stress values. We identified that the high multiplicity of the 311 reflection, its propensity to exhibit mild texture intensities when compared with other reflections (e.g. 200) and its quasiisotropic elastic behavior produce such a result.
(2) Significant redistribution and relaxation of the residual stress (both bulk and surface) occur after the removal from the baseplate. Prior to removal, the longitudinal residual stress is the highest if the scan vectors are aligned with the sample geometrical axes, but longitudinal and transverse stress components become similar when the scan vectors are rotated by 45°. After removal, the residual stress redistributes in such a way that the longitudinal stress relaxes and a bending moment is induced in the specimens. On the other hand, the transverse component barely shows any signs of relaxation or redistribution.
(3) Postremoval synchrotron XRD measurements in the plane perpendicular to the build direction revealed an alignment of the subsurface stress tensor principal axes with the geometrical axes if the scan vectors are aligned with the geometrical axes. In contrast, the subsurface stress tensor principal axes rotate around the build direction when the scan vectors are aligned by 45° to the geometry. This rotation seems to be influenced by the residual stress redistribution on removal from the baseplate. Furthermore, the subsurface strain does not vary as a function of angle around the principal axes; therefore, determining bulk residual stress using measurements along the geometrical axes does not induce large errors.
(4) The combination of laboratory Xray and neutron diffraction allowed further insight into the residual stress formation and spatial distribution: irrespective of the scanning strategy, similar residual stress distributions after removal from the baseplate were found. By combination of Xray electrolytic layer removal and neutron diffraction data, the throughthickness stress profile was successfully determined, revealing a subsurface tensile plateau balanced by compressive stress in the bulk.
6. Data availability
Datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.
Supporting information
Supporting tables. DOI: https://doi.org/10.1107/S1600576723004855/xx5022sup1.pdf
Acknowledgements
The conceptualization was carried out by JS. VL prepared and performed the neutron diffraction measurements with the associated data curation at ANSTO. JS and AK performed the laboratory XRD measurements. JS, AE, EP and JC performed the neutron diffraction experiments at POLDI. JS, AE, GAF, SD and GD performed the energydispersive synchrotron diffraction measurements. JS evaluated, interpreted and visualized the data. The main manuscript was written by JS and GB, and all authors reviewed the manuscript. Project administration was done by GB and AE. The authors express their gratitude to Gunther Mohr (BAM) for specimen production. In addition, we thank Romeo SaliwanNeumann (BAM) and Benjamin Piesker (BAM) for their help in preparation and conduction of electron backscatter analysis. Alexander Ulbricht (BAM) is acknowledged for the provision and help with the stress balance script. We must also thank Thomas GnäupelHerold (NIST) for personal support in working with ISODEC. For the provision of neutron beam time at the instrument KOWARI, we acknowledge the Australian Nuclear Science and Technology Organisation. The Paul Scherrer Institut is acknowledged for provision of neutron beam time at the POLDI beamline (proposal No. 20211090). JC gratefully acknowledges financial support from the Strategic Focus Area Advanced Manufacturing (SFAAM) initiative of the ETH Board. Finally, the Deutsches ElektronenSynchrotron DESY is acknowledged for the provision of beam time at the white beam engineering materials beamline P61A (proposal No. 20200890). Open access funding enabled and organized by Projekt DEAL.
References
Ali, H., Ghadbeigi, H. & Mumtaz, K. (2018). Mater. Sci. Eng. A, 712, 175–187. Google Scholar
Arnold, O., Bilheux, J. C., Borreguero, J. M., Buts, A., Campbell, S. I., Chapon, L., Doucet, M., Draper, N., Ferraz Leal, R., Gigg, M. A., Lynch, V. E., Markvardsen, A., Mikkelson, D. J., Mikkelson, R. L., Miller, R., Palmen, K., Parker, P., Passos, G., Perring, T. G., Peterson, P. F., Ren, S., Reuter, M. A., Savici, A. T., Taylor, J. W., Taylor, R. J., Tolchenov, R., Zhou, W. & Zikovsky, J. (2014). Nucl. Instrum. Methods Phys. Res. A, 764, 156–166. Web of Science CrossRef CAS Google Scholar
Artzt, K., Mishurova, T., Bauer, P. P., Gussone, J., BarrioberoVila, P., Evsevleev, S., Bruno, G., Requena, G. & Haubrich, J. (2020). Materials, 13, 3348. Google Scholar
Attaran, M. (2017). Bus. Horiz. 60, 677–688. Google Scholar
Bachmann, F., Hielscher, R. & Schaeben, H. (2011). Ultramicroscopy, 111, 1720–1733. Web of Science CrossRef CAS PubMed Google Scholar
Bayerlein, F., Bodensteiner, F., Zeller, C., Hofmann, M. & Zaeh, M. F. (2018). Addit. Manuf. 24, 587–594. Google Scholar
Behnken, H. & Hauk, V. (1991). Z. Metallkd. 82, 151–158. CAS Google Scholar
Bragg, W. H. & Bragg, W. L. (1913). Proc. R. Soc. London Ser. A, 88, 428–438. CrossRef ICSD CAS Google Scholar
Busi, M., Kalentics, N., Morgano, M., Griffiths, S., Tremsin, A. S., Shinohara, T., Loge, R., Leinenbach, C. & Strobl, M. (2021). Addit. Manuf. 39, 101848. Google Scholar
Capek, J., Polatidis, E., Casati, N., Pederson, R., Lyphout, C. & Strobl, M. (2022). Mater. Des. 221, 110983. Google Scholar
Chalmers, B. (1964). Principles of Solidification. New York: Wiley. Google Scholar
Collier, J. P., Wong, S. H., Tien, J. K. & Phillips, J. C. (1988). Metall. Trans. A, 19, 1657–1666. Google Scholar
Cozar, R. & Pineau, A. (1973). Met. Trans. 4, 47–59. Google Scholar
Dölle, H. & Hauk, V. (1978). Z. Metallkd. 69, 410–417. Google Scholar
Dölle, H. & Hauk, V. (1979). Z. Metallkd. 70, 682–685. Google Scholar
Eshelby, J. D. (1957). Proc. R. Soc. London Ser. A, 241, 376–396. CrossRef Web of Science Google Scholar
Eshelby, J. D. (1961). Elastic Inclusions and Inhomogeneities. Progress in Solid Mechanics, Vol. II, pp. 88–140. Amsterdam: NorthHolland. Google Scholar
Farla, R., Bhat, S., Sonntag, S., Chanyshev, A., Ma, S., Ishii, T., Liu, Z., Néri, A., Nishiyama, N., Faria, G. A., Wroblewski, T., SchulteSchrepping, H., Drube, W., Seeck, O. & Katsura, T. (2022). J. Synchrotron Rad. 29, 409–423. CrossRef CAS IUCr Journals Google Scholar
Foster, S. J., Carver, K., Dinwiddie, R. B., List, F., Unocic, K. A., Chaudhary, A. & Babu, S. S. (2018). Metall. Mater. Trans. A, 49, 5775–5798. Google Scholar
Fritsch, T., FarahbodSternahl, L., SerranoMunoz, I., Leonard, F., Haberland, C. & Bruno, G. (2022). Adv. Eng. Mater. 24, 2100689. Google Scholar
Fritsch, T., Sprengel, M., Evans, A., FarahbodSternahl, L., SaliwanNeumann, R., Hofmann, M. & Bruno, G. (2021). J. Appl. Cryst. 54, 228–236. Google Scholar
GnäupelHerold, T. (2012). J. Appl. Cryst. 45, 573–574. Google Scholar
GnäupelHerold, T., Creuziger, A. A. & Iadicola, M. (2012). J. Appl. Cryst. 45, 197–206. Web of Science CrossRef IUCr Journals Google Scholar
Gokcekaya, O., Ishimoto, T., Hibino, S., Yasutomi, J., Narushima, T. & Nakano, T. (2021). Acta Mater. 212, 116876. Google Scholar
Haldipur, P., Margetan, F. J. & Thompson, R. B. (2004). AIP Conf. Proc. 700, 1061–1068. Google Scholar
Hauk, V. (1997). Structural and Residual Stress Analysis by Nondestructive Methods. Amsterdam: Elsevier Science. Google Scholar
Hill, R. (1952). Proc. Phys. Soc. A, 65, 349–354. CrossRef Web of Science Google Scholar
Kasperovich, G., Becker, R., Artzt, K., BarrioberoVila, P., Requena, G. & Haubrich, J. (2021). Mater. Des. 207, 109858. Google Scholar
Kröner, E. (1958). Z. Phys. 151, 504–518. Google Scholar
Kruth, J. P., Deckers, J., Yasa, E. & Wauthle, R. (2012). Proc. Inst. Mech. Eng. Part B J. Eng. Manuf. 226, 980–991. Google Scholar
Kruth, J. P., Froyen, L., Van Vaerenbergh, J., Mercelis, P., Rombouts, M. & Lauwers, B. (2004). J. Mater. Process. Technol. 149, 616–622. Google Scholar
Levkulich, N. C., Semiatin, S. L., Gockel, J. E., Middendorf, J. R., DeWald, A. T. & Klingbeil, N. W. (2019). Addit. Manuf. 28, 475–484. Google Scholar
Lingenfelter, A. (1989). Superalloy 718 Metallurgy and Applications, pp. 673–683. The Minerals, Metals and Materials Society. Google Scholar
Mainprice, D., Bachmann, F., Hielscher, R. & Schaeben, H. (2015). Geol. Soc. London Spec. Publ. 409, 251–271. CrossRef Google Scholar
Mercelis, P. & Kruth, J. P. (2006). Rapid Prototyping J. 12, 254–265. CrossRef Google Scholar
Mishurova, T., Artzt, K., Haubrich, J., Requena, G. & Bruno, G. (2019). Metals, 9, 261. Google Scholar
Mishurova, T., Bruno, G., Evsevleev, S. & Sevostianov, I. (2020). J. Appl. Phys. 128, 025103. Google Scholar
Mishurova, T., Cabeza, S., Thiede, T., Nadammal, N., Kromm, A., Klaus, M., Genzel, C., Haberland, C. & Bruno, G. (2018). Metall. Mater. Trans. A, 49, 3038–3046. Google Scholar
Mishurova, T., SerranoMunoz, I., Fritsch, T., Ulbricht, A., Sprengel, M., Evans, A., Kromm, A., Madia, M. & Bruno, G. (2020). Structural Integrity of Additive Manufactured Materials and Parts, edited by N. Shamsaei & M. Seifi, pp. 122–138. West Conshohocken: ASTM International. Google Scholar
Moore, M. G. & Evans, W. P. (1958). SAE Trans. 66, 340–345. Google Scholar
Nadammal, N., Mishurova, T., Fritsch, T., SerranoMunoz, I., Kromm, A., Haberland, C., Portella, P. D. & Bruno, G. (2021). Addit. Manuf. 38, 101792. Google Scholar
Ortner, B. (2009). Powder Diffr. 24, S16–S21. Web of Science CrossRef CAS Google Scholar
Ortner, B. (2011). Mater. Sci. Forum, 681, 7–12. Google Scholar
Ortner, B. (2014). Adv. Mater. Res. 996, 215–220. Google Scholar
Pant, P., Proper, S., Luzin, V., Sjostrom, S., Simonsson, K., Moverare, J., Hosseini, S., Pacheco, V. & Peng, R. L. (2020). Addit. Manuf. 36, 101501. Google Scholar
Reuss, A. (1929). Z. Angew. Math. Mech. 9, 49–58. CrossRef CAS Google Scholar
Schröder, J., Evans, A., Mishurova, T., Ulbricht, A., Sprengel, M., SerranoMunoz, I., Fritsch, T., Kromm, A., Kannengießer, T. & Bruno, G. (2021). Metals, 11, 1830. Google Scholar
Schröder, J., Evans, A., Polatidis, E., Čapek, J., Mohr, G., SerranoMunoz, I. & Bruno, G. (2022). J. Mater. Sci. 57, 15036–15058. Google Scholar
Schröder, J., Mishurova, T., Fritsch, T., SerranoMunoz, I., Evans, A., Sprengel, M., Klaus, M., Genzel, C., Schneider, J. & Bruno, G. (2021). Mater. Sci. Eng. A, 805, 140555. Google Scholar
Seabold, S. & Perktold, J. (2010). Proceedings of the 9th Python in Science Conference, pp. 57–61. SciPy Conferences. Google Scholar
SerranoMunoz, I., Evans, A., Mishurova, T., Sprengel, M., Pirling, T., Kromm, A. & Bruno, G. (2022). Adv. Eng. Mater. 24, 2100895. Google Scholar
SerranoMunoz, I., Fritsch, T., Mishurova, T., Trofimov, A., Apel, D., Ulbricht, A., Kromm, A., Hesse, R., Evans, A. & Bruno, G. (2021). J. Mater. Sci. 56, 5845–5867. Google Scholar
SerranoMunoz, I., Ulbricht, A., Fritsch, T., Mishurova, T., Kromm, A., Hofmann, M., Wimpory, R. C., Evans, A. & Bruno, G. (2021). Adv. Eng. Mater. 23, 2100158. Google Scholar
Spieß, L., Teichert, G., Schwarzer, R., Behnken, H. & Genzel, C. (2009). Moderne Röntgenbeugung, 2nd ed. Wiesbaden: Vieweg + Teubner Verlag. Google Scholar
Sprengel, M., Mohr, G., Altenburg, S. J., Evans, A., SerranoMunoz, I., Kromm, A., Pirling, T., Bruno, G. & Kannengiesser, T. (2022). Adv. Eng. Mater. 24, 2101330. Google Scholar
Stuhr, U. (2005). Nucl. Instrum. Methods Phys. Res. A, 545, 319–329. Google Scholar
Stuhr, U., Grosse, M. & Wagner, W. (2006). Mater. Sci. Eng. A, 437, 134–138. Google Scholar
Stuhr, U., Spitzer, H., Egger, J., Hofer, A., Rasmussen, P., Graf, D., Bollhalder, A., Schild, M., Bauer, G. & Wagner, W. (2005). Nucl. Instrum. Methods Phys. Res. A, 545, 330–338. Google Scholar
Thiede, T., Cabeza, S., Mishurova, T., Nadammal, N., Kromm, A., Bode, J., Haberland, C. & Bruno, G. (2018). Mater. Perform. Charact. 7, 717–735. Google Scholar
Ulbricht, A., Altenburg, S. J., Sprengel, M., Sommer, K., Mohr, G., Fritsch, T., Mishurova, T., SerranoMunoz, I., Evans, A., Hofmann, M. & Bruno, G. (2020). Metals, 10, 1234. Google Scholar
Van Houtte, P. & De Buyser, L. (1993). Acta Metall. Mater. 41, 323–336. CrossRef CAS Web of Science Google Scholar
Voigt, W. (1889). Ann. Phys. 274, 573–587. Google Scholar
Volpato, G. M., Tetzlaff, U. & Fredel, M. C. (2022). Addit. Manuf. 55, 102871. Google Scholar
Wang, Z. Q., Stoica, A. D., Ma, D. & Beese, A. M. (2016). Mater. Sci. Eng. A, 674, 406–412. Google Scholar
Withers, P. J., Preuss, M., Steuwer, A. & Pang, J. W. L. (2007). J. Appl. Cryst. 40, 891–904. Web of Science CrossRef CAS IUCr Journals Google Scholar
Yadroitsev, I. & Yadroitsava, I. (2015). Virtual Phys. Prototyping, 10, 67–76. CrossRef Google Scholar
Zhong, Q., Wei, K. W., Lu, Z., Yue, X. Z., Ouyang, T. Y. & Zeng, X. Y. (2023). J. Mater. Process. Technol. 311, 117824. Google Scholar
This is an openaccess article distributed under the terms of the Creative Commons Attribution (CCBY) Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original authors and source are cited.