Ultrafine Fe-Fe2Ti eutectics by directed energy deposition: Insights into microstructure formation based on experimental techniques and phase field modelling

Journal Pre-proof Ultrafine Fe-Fe2Ti eutectics by directed energy deposition: insights into microstructure formation based on experimental techniques and phase field modelling G. Requena, K. Bugelnig, F. Sket, S. Milenkovic, G. R¨odler, A. Weisheit, J. Gussone, J. Haubrich, P. Barriobero-Vila, T. Pusztai, L. Gr´ an´ asy, A. Theofilatos, J.C. da Silva, U. Hecht PII: S2214-8604(19)31466-6 DOI: https://doi.org/10.1016/j.addma.2020.101133 Reference: ADDMA 101133 To appear in: Additive Manufacturing Received Date: 31 August 2019 Revised Date: 13 February 2020 Accepted Date: 13 February 2020 Please cite this article as: Requena G, Bugelnig K, Sket F, Milenkovic S, R¨odler G, Weisheit A, Gussone J, Haubrich J, Barriobero-Vila P, Pusztai T, Gr´ an´ asy L, Theofilatos A, da Silva JC, Hecht U, Ultrafine Fe-Fe2Ti eutectics by directed energy deposition: insights into microstructure formation based on experimental techniques and phase field modelling, Additive Manufacturing (2020), doi: https://doi.org/10.1016/j.addma.2020.101133 This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2020 Published by Elsevier. 1 Ultrafine Fe-Fe2Ti eutectics by directed energy deposition: insights into microstructure formation based on experimental techniques and phase field modelling G. Requena1,7, K. Bugelnig1, F. Sket2, S. Milenkovic2, G. Rödler3, A. Weisheit3, J. Gussone1, J. Haubrich1, P. Barriobero-Vila1, T. Pusztai4, L. Gránásy4, A. Theofilatos5, J. C. da Silva6, U. Hecht5 1 German Aerospace Center DLR, Institute of Materials Research, Linder Höhe D-51147 Cologne, Germany 2 IMDEA Materiales, Eric Kandel 2, Tecnogetafe, 28906 Getafe, Madrid, Spain 3 Fraunhofer Institute for Laser Technology ILT, Steinbachstraße 15, 52074 Aachen, Germany 4 Wigner Research Centre for Physics, Institute for Solid State Physics and Optics, KonkolyThege M. út 29-33, 1121 Budapest, Hungary 5 Access e.V., Intzestr. 5, 52072 Aachen, Germany 6 European Synchrotron Radiation Facility, Grenoble, France 7 Metallic Structures and Materials Systems for Aerospace Engineering, RWTH Aachen University, 52062 Aachen, Germany., Abstract We investigated the Fe-Fe2Ti eutectic microstructure obtained by Directed Energy Deposition (DED) with a hypereutectic composition of Fe-17.6 at.% Ti. Ultrafine lamellar spacings as low as 200 nm were achieved, features which otherwise can only be obtained in thin specimens, e.g. by suction casting. However, at interlayer boundaries (ILBs) a globular morphology of the primary Fe2Ti phase is observed with halos of the Fe phase. For the given DED conditions the crystalline structure is thus discontinuous across the ILBs. Both 2D and 3D analysis methods were used to quantify the microstructure, including high resolution synchrotron holographic X-ray computed tomography (HXCT). The generic behaviour of eutectic systems under conditions that qualitatively correspond to those of laser additive manufacturing was explored by phase-field modelling for selected nucleation scenarios and alloy compositions spanning from eutectic to hyper-eutectic. While providing valuable insights into microstructure formation, the simulations point out the need to further deepen our understanding about melting under additive manufacturing conditions in order to implement suitable nucleation and / or free growth models. The simulations also show that globular ILBs can be prevented when using exactly eutectic alloy compositions. 1. Introduction Eutectic alloys have been in the focus of materials science and engineering over many decades because they are a prime example for spontaneous pattern formation and selforganization during solidification [1] while also bearing promise as in-situ composites with Journal Pre-proof 2 unique functional and structural properties. Their microstructure commonly consists of grains with a lamellar, fibrous, or more complex periodic arrangement of the solid phases. The characteristic length scales of the structure are the grain size (GS) and the eutectic spacing (), both of which depend on the alloy composition and the processing conditions. A scaling law was proposed by Jackson and Hunt in 1966 [2] to link the average spacing𝜆 to the local growth velocity v, e.g. 𝜆≈ 𝐾 ∙ 𝑣−0.5. The constant K depends solely on material properties. The Jackson-Hunt (JH) scaling law holds for a large range of growth velocities, spanning over several orders of magnitude from about 10-7 to 100 m/s. Correction terms are necessary if growth occurs at very low velocity in a high temperature gradient. For high growth velocities the diffusion length ld = 2D/v may be as small as the spacing and under these conditions of rapid solidification the JH-scaling law must be amended [3] or even abandoned in favour of models that predict banded structure formation [4]. The growth velocities to be expected in Directed Energy Deposition (DED) commonly reach values in the order of 10-4 m/s corresponding to the velocity of isotherms inside the melt pool underneath the travelling laser beam. For these conditions ultrafine eutectic spacings may be obtained, well below 500 nm. This possibility seems to be unique: conventional casting cannot achieve the high cooling rates required to drive isotherms at such high velocity except potentially in small diameter capillaries. This has indeed been explored in so called suction casting experiments using arc melting followed by suction casting into cold metallic moulds with cylindrical cavities of diameter 1 to 4 mm. An impressive body of literature is published on ultrafine eutectics prepared by suction casting [5, 6], however without means to bridge towards applications. Laser additive manufacturing (L-AM) is the first technology to offer exploitation means for all the knowledge developed so far by the suction casting experiments, provided that structural inhomogeneities at interlayer boundaries can be avoided or tolerated. Here we present the results from ongoing research work dedicated to this subject. We selected the Fe-Fe2Ti eutectic for experiments and a generic A-B eutectic for modelling the structure evolution during DED. The ferrite-Laves phase eutectic Fe-Fe2Ti has been cast and described [7] as a potential material for high temperature applications and we attempted to investigate its manufacturing, microstructure and properties using DED. This paper focusses on microstructure formation, characterization and modelling as follows: Section 2 describes sample manufacturing and the characteristic features observed in the microstructure, as well as a detailed analysis of the 3D microstructure obtained by high resolution synchrotron holographic X-ray computed tomography. Section 3 presents and discusses the results from phase field modelling and simulations which were set up to include AM-specific thermal boundary conditions for melting/remelting and solidification. Conclusions and an outlook to future work are given in Section 4. 2. Experimental methods and microstructure characterization 2.1 Sample manufacturing and characteristic microstructure features The alloy Fe-17.6 at.% Ti with a slightly hypereutectic composition (eutectic composition cE=15.8 at.% Ti) was selected for the experimental work. For a very similar alloy Journal Pre-proof 3 composition, e.g. Fe-18.1 at.%Ti corresponding to 16 wt.%Ti, Tokoro et al. [8] reported the dependency of the average eutectic spacing𝜆 on growth velocity based on unidirectional solidification experiments in a Bridgman furnace. From these data the Jackson-Hunt constant K was evaluated to be K=271 µm³/s. Other alloy properties were calculated with the ThermoCalc software [9] and the database TCBIN, specifically the liquidus temperature Tliq, the eutectic temperature TE, the solidification interval T = Tliq-TE and the phase fractions at T=TE-ΔTTable 1 lists these alloy properties. DED was conducted using a fiber coupled diode laser system with a maximum laser output of 2 kW. The first samples with the dimensions of 4 x 10 x 3 mm3 were produced from Fe (99.8%) and Ti (99.7%) elemental powders blended in a tumble mixer. The particle size distribution of the elemental Fe-powder was 20-90µm and 45-90µm for the Ti-powder, respectively. The powder blend was fed through a coaxial nozzle by an inert gas stream of Ar, which also shields the melt pool from the surrounding atmosphere. DED was carried out with a bidirectional hatching strategy using a 1.4301 steel substrate at room temperature and the parameters listed in Table 2. The main microstructure characteristics are displayed in Fig. 1, from optical microscopy (Fig.1a) and backscatter electron microscopy (Fig. 1.b). (a) (b) Fig. 1: As-built microstructure in a section plane perpendicular to the deposited tracks showing elongated eutectic grains in the bulk layers/tracks and interlayer boundaries of thickness “” with a distinct, globular morphology; (a) optical microscopy, (b) backscatter electron microscopy. The two-phase eutectic microstructure is composed of the hexagonal Laves phase Fe2Ti (space group no. 194) and the body centered (Fe) (space group no. 229) with elongated lamellar grains inside the bulk volume of the layers / tracks. At interlayer boundaries however, the microstructure is globular, being composed of rounded primary Fe2Ti particles which are enveloped by (Fe) halos. The thickness of the interlayer boundaries (ILBs) ranges around =12±5 µm, while the eutectic spacing in the lamellar grains ranges around =190±25 nm (compare Fig. 5 in section 2.2.). More details are revealed by electron backscatter diffraction (EBSD) as shown in Fig. 2. The EBSD maps were acquired in an SEM type Zeiss Gemini 1550 equipped with a HKL Nordlys detector and using the software Oxford INCA Crystal. The maps in Figs. 2(a) for Fe2Ti and (Fe) respectively, encompass one ILB running from the upper left to the lower right side of the map. The crystal orientations are provided as inverse pole figure colour maps relative to one direction only (here the transverse direction TD was chosen). It is noteworthy to observe that for the given = 12 µm Journal Pre-proof 4 DED process parameters the crystalline orientation of both phases is discontinuous across the ILB. A magnified view of the crystal structure inside the ILB is presented in Figs. 2(b), showing the globular Fe2Ti phase particles being enveloped by halos of (Fe). In each (Fe) halo several (Fe) grains can be identified. From the top-most globular particles new grains develop into the following layer. (a) Transverse direction TD  (b) Transverse direction TD  (c) IPF triangle Fig.2: EBSD maps encompassing an interlayer boundary (a) reveal that the crystalline orientation of the two phases is discontinuous across the ILB. A magnified view of the structure inside the ILB (b) shows that each Fe2Ti-particle is enveloped by a halo of (Fe). In each halo several (Fe) grains can be identified. For simplicity the orientation maps are shown for one direction only (here TD) using the inverse pole figure colouring of the IPF-triangles (c). The phase fraction of the Fe-phase is calculated to be 49.4% and for the Fe2Ti phase 50.5%. The origin of the globular morphology at interlayer boundaries is proposed to follow the sequence of events (i) through (iii) upon melting of the previously deposited layer as follows: (i) Initially the alloy displays coupled eutectic growth, even if the composition is hypereutectic on the primary Fe2Ti-side of the phase diagram. Therefore, the bulk structure is a lamellar eutectic. (ii) Upon melting, some part of the previously deposited layer will melt completely. However, at the bottom of the melt pool the system will evolve towards thermodynamic equilibrium and display a mushy zone composed of primary Fe2Ti+Liquid over a distance  that roughly corresponds to the ratio between the solidification interval T=Tliq-Te. and the temperature gradient G. In this region, primary Fe2Ti lamellae remain unmolten protruding into the surrounding liquid like thin plates. The thin plates in contact with the liquid will undergo a fragmentation process. It is not fully clear, whether this process is a Plateau-Rayleigh capillary instability accomplished through an interface perturbation [10, 11] or simply a fast melting and rupture process at random locations, e.g. at subgrain boundaries, dislocations etc. Recent literature based on MD simulations [12, 13] indeed Journal Pre-proof 5 reports on pathways to very fast melting. Wu et al. [13] report rupture times of about 12 milliseconds for a 10 µm diameter rod. To our best knowledge, similar results for plates have not been reported. Investigating the details of the fragmentation process remains a challenging research task for the future. (iii) The fragmentation process is necessarily accompanied by partial dissolution of Fe2Ti fragments, more pronounced in the upper part of the mushy zone. This is because the phase fraction of Fe2Ti in the mushy zone depends on the local temperature in the temperature gradient. The fragments are free to spheroidize, move and rotate along with the flow in the melt pool. At the transition from melting to solidification these fragments will experience free growth and the top-most ones will further re-initiate coupled growth. For laser-based AM-conditions, however, more research work is required in order to fully understand the process of melting, fragmentation and the interlayer boundary solidification structures as well as their impact on mechanical properties. Irrespective of the open questions, the microstructure analysis results may be summarized along the following lines: (i) From the thickness of the interlayer boundaries =12±5 µm one may estimate the local temperature gradient G at the bottom of the melt pool 𝑮 = 𝚫𝑻 𝜹 ⁄. Here T is the solidification interval of the slightly hypereutectic alloy, as listed in Table 1. For this sample the estimated value is 𝑮~𝟐. 𝟔 × 𝟏𝟎𝟔 𝑲/𝒎. As highlighted before, more research including melt pool simulations are required to refine this analysis. (ii) From the average lamellar spacing=190±25 nm in the bulk eutectic grains one may estimate the local growth velocity 𝒗 corresponding to the average isotherm velocity in the melt pool using the Jackson-Hunt scaling law 𝒗 = 𝑲𝑱𝑯 𝝀𝟐 ⁄. For this sample the estimated value is 𝒗~𝟕. 𝟓 × 𝟏𝟎−𝟑𝒎/𝒔. The order of magnitude seems to be in reasonable agreement with the applied laser scanning velocity 𝑽𝑳 of 800 mm/min which amounts to 𝑽𝑳=𝟏𝟑 × 𝟏𝟎−𝟑𝒎/𝒔. Since the melt pool is solidified from virtually two sides (radially), the necessary growth velocity expected in a steady state process would correspond to roughly half the laser scanning velocity. Melt pool simualtions are currently ongoing to refine this analysis. For comparison, the lamellar spacing obtained in the same alloy by suction casting in a rod with a 3x3 mm cross sectional area amounts to =720±180 nm, as will be presented elsewhere. 2.2 Microstructure characterization by high resolution synchrotron holographic X-ray computed tomography (HXCT) The 3D microstructure of the Fe-17.6 at.%Ti DED sample was investigated using synchrotron holographic X-ray computed tomography (HXCT) at the nano-imaging beamline ID16A of the European Synchrotron Research Facility (ESRF), Grenoble, France [14]. To this purpose, a cylindrical sample with a diameter of ~ 15 µm and ~ 20 µm height was prepared by focused ion beam (FIB) in a FEI Helios Nanolab 600i dual beam microscope from a representative region containing the interface between the globular and the lamellar region (see Fig. 1b). In fact, the volume was extracted from the bottom part of an ILB encompassing a lamellar region just below the ILB (at the bottom of a melt pool). Journal Pre-proof 6 Table 3 displays the parameters used for HXCT. The sample was illuminated with a magnifying X-ray cone beam of 33.6 keV focused by Kirkpatrick–Baez mirrors using the zoom HXCT approach [15]. 2000 projections were acquired between 0° -180° with acquisition times of ~ 1 s/projection employing a CCD camera at four sample-to-focal-point distances (4.19, 4.37, 5.09 and 6.59 mm) for efficient phase retrieval of the holo-tomographic reconstruction. The experiment was carried out under ultra-high vacuum atmosphere. The following steps were performed to analyse the 3D data: (i) Reconstruction of the tomographic volume was performed by using a filtered back projection algorithm implemented in the PySHT software of the ESRF with a resulting voxel size of (10 nm³). Prior to segmentation, the quality of the reconstructed volume was improved using a band pass filter and 2D or 3D anisotropic diffusion filters available in Fiji [16] and Avizo Fire 9.5 in order to minimize ring artefacts and smooth the images. After filtering, the volume was converted from 16 to 8 bit format and the grey value histogram was inverted so that the appearance of the microstructural constituents is analogous to SEM-BSE images. (ii) The segmentation of the microstructural constituents was carried out by global grey value thresholding applying three different grey value thresholds for each segmented constituent (best threshold determined by eye ± 2 grey values) to ensure better representativity. 3D visualizations were produced using Avizo Fire 9.5. (iii) The quantification of the 3D microstructure of the material was carried out for the entire volume of ~ 12.1 × 10.8 × 12.6 µm³ as well as for individual volumes containing separately the eutectic lamellar region (7.4 × 10.8 × 12.6 µm³) and the globular interlayer boundary (4.7 × 10.8 × 12.6 µm³). The global interconnectivity of the Fe and Fe2Ti phases was calculated as the volume of the largest particle divided by the total volume of the phase considered [17,18]. The 3D thickness distribution was determined individually for the Fe and Fe2Ti phases using the software Avizo Fire 9.5 to evaluate the coarseness of the microstructure. This parameter is calculated for each voxel of the considered phase after skeletonization and is defined as the diameter of the largest sphere that fits within the considered structure [18]. Fig. 3 depicts the as-reconstructed HXCT volume and a portion of a reconstructed HXCT slice. This volume / slice contains few coarse (Fe) particles just below the interlayer boundary. These particles are not predominant for the whole sample volume (compare Fig. 1 and Fig. 2). The volume also contains a small amount of oxide particles and fine pores (black). We were not able to make a clear distinction between oxides and pores. Journal Pre-proof 7 (a) (b) Fig. 3: 3D visualization of (a) the as-reconstructed HXCT volume (a) and portion of a reconstructed HXCT slice (b) at the interface region containing lamellar and globular structures; bright grey = Fe, dark grey = Fe2Ti, black = oxides. ROI: ~ 12.1 × 10.8 × 12.6 µm³, voxel size = 10x10x10 nm³. Fig. 4 shows 3D visualizations of the as-reconstructed volume of 12.1 × 10.8 × 12.6 µm³ with the Fe (a) and Fe2Ti (b) phases segmented separately. The blue dashed line indicates the boundary between the lamellar and globular regions. The Fe (blue) and Fe2Ti- (red) phases form highly interconnected 3D networks which are coarser in the globular region. The volumes are also shown in color-coded views corresponding to the 3D thickness of the phases (c), (d). The coarse spherical (Fe) particles emerge with orange colour (c-d) being heterogeneously distributed in the lamellar region just below the globular ILB region. Inclusions, likely oxide particles, not shown here, were found in the center of the Fe spheres, but never in the centers of the globular Fe2Ti particles. The (Fe) and Fe2Ti phases both present a global interconnectivity ~ 99 % and volume fractions of 46.44 ± 2.2 vol. % and 39.95 ± 2 vol. % for the Fe-network and 47.02 ± 2.3 vol. % and 66.58 ± 2.6 vol. % for the Fe2Ti-network in the lamellar and globular regions, respectively. Spherical Fe has a volume fraction of 1.1 ± 0.03 vol. % within the entire volume, while oxides/pores have a sphericity between 0.6 – 1, an aspect ratio between 1.2 – 2.5 and represent a volume fraction of 0.34 ± 0.02 vol. % within the entire volume. Fig. 5 shows the evaluated 3D thickness distributions determined for the Fe-network (a) (b) and the Fe2Ti network (c), (d), respectively. The relative frequency distributions are displayed separately for the lamellar region (a), (c), and for the globular region inside the ILB (b), (d) being displayed in blue and red color, respectively. A clear difference in coarseness between the lamellar and globular regions can be seen. The thickness of the phases in the lamellar region peaks around 100 nm, which corresponds to an average lamellar spacing  of about 200 nm. For the globular region, comparatively coarser structures with 3D thicknesses up to ~ 0.36 µm for the Fe-network and up to ~ 0.5 µm for the Fe2Ti can be observed. The few coarse Feparticles have a diameter in the range from 0.28 - 1.04 µm. Journal Pre-proof 14 Next, we changed the initial composition of the melt (representing the average composition of the powder mixture in the experiments). We carried out simulations at the eutectic and 3 offeutectic compositions, c0 = 0.50, 0.55, 0.60 and 0.65 (Fig. 11). At the eutectic composition, as the phase diagram of our model is also symmetric with respect to the components A and B, a fully symmetric system was obtained: there should be no preference of the nucleation of the B rich phase over the A rich phase. Instead of the simultaneous nucleation of roughly equal numbers of globular particles rich in A or rich in B, which then develop to a eutectic structure, we observed the appearance of globular particles that had a disordered eutectic structure already in their core. This is in contrast with the simulation results at the off-eutectic compositions (c0>0.5) where the nucleation of B-rich particles is always preferred. This results in a visible difference between the two types of simulations. The one corresponding to the eutectic composition does not, while the ones corresponding to the off-eutectic compositions do have a clearly visible globular structure in the fully solidified state. Finally, we tuned the amplitude of the Gaussian white noise added to the phase-field equation of motion to represent thermal fluctuations present in the system. We made simulations with 4 different values, 0.5, 0.67, 0.75, and 1 in relative units (Fig. 12). Although in nature the strength of the fluctuations cannot be easily tuned independently from temperature, these simulations demonstrate the importance of nucleation in forming the microstructure. The larger the noise amplitude, the higher the nucleation rate, which results in wider globular layers of smaller particle size. Fig. 11: The effect of alloy concentration on the final microstructure as predicted by the phase-field simulations. From top to bottom, c0 = 0.50, 0.55, 0.60 and 0.65, respectively. Fig. 12: The effect of the amplitude of the Gaussian white noise used in the equation of motion of the phase-field. From top to bottom, the noise amplitudes are 0.5, 0.67, 0.75 and 1 in relative units. Note that c0=0.6. Journal Pre-proof 15 4. Summary and outlook Directed Energy Deposition (DED) was used as a method to produce an ultrafine eutectic microstructure composed of BCC-Fe and the hexagonal Laves phase Fe2Ti in a binary alloy with a slightly hypereutectic composition of Fe-17.6 at.% Ti. Lamellar spacings as low as 200 nm were obtained, however the microstructure is discontinuous across interlayer boundaries. In these regions, i.e. at the bottom of the melt pool, conditions prevail which lead to a globular morphology of the primary Fe2Ti phase. 3D microstructure analysis using high resolution synchrotron holographic X-ray computed tomography (HXCT) was successfully performed with a voxel size of 10x10x10 nm³. HXCT was used to characterize the microstructure and to verify if oxide particles or other potential inclusions can be detected in the center of the globular primary Fe2Ti particles. This is not the case and hence we conclude that the nucleation of the globular Fe2Ti particles is not caused by heterogeneous nucleation. With our phase-field model we also investigated the possibility of nucleation on partially remelted and spherodized lamellae as seed particles. Though the simulations should capture this automatically, with the simple symmetric model system and process parameters used we could not observe this directly. But we have shown that if we assume that these seeds are present in the system, the nucleation and growth of spherical particles of the primary phase is reproduced and the resulting microstructure is very similar to those of obtained by homogeneous nucleation and to the experiments. Then homogeneous nucleation scenarios were further explored in detail by phase field modelling for a generic eutectic phase diagram. Simulation results indeed showed that a layered microstructure similar to the one observed in the experiments can be reproduced using a cyclic temperature history that mimics the melting by the laser beam and the subsequent solidification. Experimental observations suggest yet another mechanism to explain the globular structure of primary phase particles at interlayer boundaries: Most likely these particles grow from partially remelted and spheroidized parts of eutectic lamellae from the previous layers. A dedicated study focusing on the remelting of a solid eutectic structure is required to clarify this issue. Furthermore, the experimental results were used to estimate the local growth conditions in DED using simple scaling laws: from the measured eutectic spacing  the isotherm velocity v in the melt pool was estimated to reach~7.5 × 10−3𝑚/𝑠 following the Jackson-Hunt scaling law 𝒗 = 𝑲𝑱𝑯 𝝀𝟐 ⁄. The local temperature gradient G at the bottom of the melt pool was estimated from the width of the interlayer boundaries 𝜹 and the equilibrium melting /solidification interval of the alloy 𝚫𝑻. Using 𝑮 = 𝚫𝑻 𝜹 ⁄ the temperature gradient was estimated to reach ~2.6 × 106 K/m. We wish to highlight that binary eutectic alloys could be used to validate melt pool simulations in the future. The only condition that needs to be checked beforehand is the verification that the coupled zone of the eutectic at case allows achieving coupled growth at high growth velocitites. Journal Pre-proof 16 Acknowledgments The authors gratefully acknowledge funding through the European M-Era.Net Project “ELAM” funded through national agencies under grants number BMBF 03XP0121B, 03XP0121C, the National Research, Development, and Innovation Office (NKFIH), Hungary under contract NN-125832 and Agencia Estatal de Investigación, Spain, under grant number PCIN-2017-11. The ESRF is acknowledged for the provision of synchrotron facilities at beamline ID16-A in the frame of proposal MA-4351. References [1] S. Akamatsu, M. Plapp, Eutectic and peritectic solidification patterns. Current Opinion in Solid State and Materials Science, 20(1), (2016), 46-54. [2] K. A. Jackson, J. D. Hunt, Lamellar and rod eutectic growth, Transactions of the Metallurgical Society of AIME, Vol. 236, (1966), 1129-1142. [3] N. Wang, R. Trivedi, Limit of steady-state lamellar eutectic growth, Scripta Materialia 64, (2011), 848–851. [4] S. C. Gill and W. Kurz, Rapidly solidified Al-cu alloys-IICalculation of the microstructure selection map, Acta metall mater. Vol. 43, No. 1, (1995), 139-151. [5] J. M. Park, K. B. Kim, W. T. Kim, M. H. Lee, J. Eckert, D. H. Kim, High strength ultrafine eutectic Fe–Nb–Al composites with enhanced plasticity. Intermetallics, 16(5), (2008), 642-650. [6] L. Sheng, J. T. Guo, L. Z. Zhou, H. Q. Ye, Microstructure and compressive properties of NiAl–Cr (Mo)–Dy near eutectic alloy prepared by suction casting. Materials Science and Technology, 26(2), (2010), 164-168. [7] D. Barbier, M.X. Huang, O. Bouaziz, A novel eutectic Fe-15 wt.% Ti alloy with an ultrafine lamellar structure for high temperature applications, Intermetallics 35, (2013), 41-44. [8] K. Tokoro, Y. Kimura, Structural and magnetic properties of unidirectionally solidified Fe-Fe2Ti eutectic alloys, Tetsu-to-Hagane, The iron and steel institute of Japan, Volume 60, (1974), 386-396. [9] Thermo-Calc Software, available online, accessed February 2019, www.thermocalc.com, [10] C.A. Norlund and R. Trivedi, Eutectic interface configurations during melting, Metallurgical and Materials Transactions A, 31A, (2000). 1261-1269. [11] J. K. Lee and T. H. Courtney, Two-dimensional finite difference analysis of shape instabilities in plates, Metallurgical Transactions A 20.8, (1989), 1385-1394. [12] A. Samanta, M.E. Tuckerman, T.Q. Yu, E. Weinan Microscopic mechanisms of equilibrium melting of a solid, Science 346.6210 (2014), 729-732. Journal Pre-proof 17 [13] L.K. Wu, B. Xu, Q. L. Li, W. Liu, Self-instability of finite sized solid-liquid interfaces, Scientific reports 5, (2015), 18466. [14] J. C. da Silva, A. Pacureanu, Y. Yang, S. Bohic, C. Morawe, R. Barrett, P. Cloetens, Efficient concentration of high-energy x-rays for diffraction-limited imaging resolution, Optica 4(5), (2017), 492-495. [15] G. Requena, P. Cloetens, W Altendorfer, C Poletti, D. Tolnai, F. Warchomicka, H.P. Degischer, Sub-micrometer synchrotron tomography of multiphase metals using Kirkpatrick– Baez optics, Scripta materialia 61, (2009), 760-763. [16] Schindelin, J.; Arganda-Carreras, I.; Frise, E., Fiji: An open-source platform for biological-image analysis. Nat. Methods 9, (2012), 676-682. [17] G.C. Requena, P. Degischer, E.D. Marks, E. Boller, Microtomographic study of the evolution of microstructure during creep of an AlSiCuMgNi alloy reinforced with Al2O3 short fibres. Mater. Sci. Eng. A 487, (2008), 99–107. [18] T. Hildebrand, P. Rüegsegger, A new method for the model-independent assessment of thickness in three-dimensional images, Journal of Microscopy 185(1), (1997), 67-75. [19] D. Lewis, T. Pusztai, L. Gránásy, J. Warren, W. Boettinger: Phase field models for eutectic solidification, JOM - J. Min. Met. Mat. S. 56, (4), (2004), 34-39. [20] L. Gránásy, T. Börzsönyi, T. Pusztai: Nucleation and bulk crystallization in binary phase field theory. Phys. Rev. Lett. 88, (2002). 206105-1-4. [21] L. Gránásy, T. Pusztai, T. Börzsönyi, J. A. Warren, J. F. Douglas, A general mechanism of polycrystalline growth. Nature Materials 3, (2004), 645-650. [22] L. Gránásy, T. Pusztai, G. Tegze, J. A. Warren, J. F. Douglas: Growth and form of spherulites. Phys. Rev. E. 72, (2005), 011605-1-15. [23] L. Gránásy, L. Rátkai, A. Szállás, B. Korbuly, G.I. Tóth, L. Környei, T. Pusztai: Phasefield modeling of polycrystalline solidification: From needle crystals to spherulites – a review. Metall. Mater. Trans. A 45, (4), (2014), 1694-1719. [24] C. Fetzer, L. Gránásy, T. Kemény, E. Kótai, M. Tegze, I. Vincze, W. Howing, F. van der Woude: Laser-melted amorphous and crystalline Fe-B alloys. Phys. Rev. B 42, (1990), 548554. [25] L. Gránásy, G.I. Tóth, J.A. Warren, F. Podmaniczky, G. Tegze, L. Rátkai, and T. Pusztai, Phase-field modeling of crystal nucleation in undercooled liquids – A review, Prog. Mater. Sci. 106, (2019) art. no. 100569, 1-51. [26] T. E. Quested and A. L. Greer, Athermal heterogeneous nucleation of solidification, Acta Mater. 53, (2005), 2683-2692. [27] L. Wei, Y. Cao, X. Lin, W. Huang, Cellular Automaton Simulation of the Growth of Anomalous Eutectic during Laser Remelting Process, Materials 11, (2018), 1844 /1-9. Journal Pre-proof 18 Table 1: Alloy Fe-17.6 at.% Ti thermodynamic equilibrium data. Alloy, at.% Tliq, °C Eutectic temperature, °C Solidification interval, °C Phase fractions @T=TE-ΔΤmol% Fe-17.6 Ti 1323 1292 31 44.1 mol% Fe2Ti Table 2: DED process parameters for a small sample with dimensions of 4 x 10 x 3 mm3 Laser Power (W) Velocity (mm/min) Powder Mass Flow (g/min) Laser Beam Diameter (mm) Layer Height (mm) 200 800 1.2 0.6 0.2 Table 3: Parameters for HXCT experiments at the beamline ID16A/ESRF. Detector Energy (keV) Field of view (µm2) Sample-todetector distance (mm) Exposure time (s/proj) Proj. Voxel size (nm3) CCD 33.6 20.5 × 20.5 4.19 0.25 2000/distance 103 4.37 5.09 6.59 Journal Pre-proof