Alloy design for laser powder bed fusion additive manufacturing: a critical review

Metal additive manufacturing (AM) has been extensively studied in recent decades. Despite the significant progress achieved in manufacturing complex shapes and structures, challenges such as severe cracking when using existing alloys for laser powder bed fusion (L-PBF) AM have persisted. These challenges arise because commercial alloys are primarily designed for conventional casting or forging processes, overlooking the fast cooling rates, steep temperature gradients and multiple thermal cycles of L-PBF. To address this, there is an urgent need to develop novel alloys specifically tailored for L-PBF technologies. This review provides a comprehensive summary of the strategies employed in alloy design for L-PBF. It aims to guide future research on designing novel alloys dedicated to L-PBF instead of adapting existing alloys. The review begins by discussing the features of the L-PBF processes, focusing on rapid solidification and intrinsic heat treatment. Next, the printability of the four main existing alloys (Fe-, Ni-, Al- and Ti-based alloys) is critically assessed, with a comparison of their conventional weldability. It was found that the weldability criteria are not always applicable in estimating printability. Furthermore, the review presents recent advances in alloy development and associated strategies, categorizing them into crack mitigation-oriented, microstructure manipulation-oriented and machine learning-assisted approaches. Lastly, an outlook and suggestions are given to highlight the issues that need to be addressed in future work.

Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.

Introduction
Additive manufacturing (AM), building objects layer by layer directly from three-dimensional (3D) data [1], has been regarded as a revolutionary manufacturing technology since it was invented in the 1980s [2].The early studies primarily focused on polymers and AM was used for rapid prototyping applications.Later on, with the introduction of selective laser melting (also termed laser powder bed fusion, L-PBF) in the 1990s [3], AM technology has been widely used for the production of complex metal parts.Nowadays, AM technology has gained recognition for end-use part production, particularly in aerospace, automotive and medical industries [4][5][6].
In contrast to traditional 'subtractive' manufacturing, AM offers notable advantages.AM allows for complex geometries, including internal cavities, undercuts, and intricate lattice structures, without the constraints of traditional manufacturing methods.AM also enables quick iterations in the design process, reducing time and costs associated with tooling.Additionally, parts can be designed with specific material properties and optimized for their intended function, leading to lightweight and efficient designs [7][8][9][10].
Despite notable achievements in manufacturing complex shapes and structures using AM [11][12][13][14], the availability of printable metals that are crack-free in the as-built state is still limited.The presence of defects such as cracks and pores significantly compromises the mechanical properties, as the alloys were initially designed for wrought or casting without considering the unique characteristics of the AM process.L-PBF AM involves high cooling rates, steep temperature gradients, and numerous thermal cycles [15,16].The compositions of many existing alloys are not well-suited for these characteristics, leading to cracking when directly applied to L-PBF.For instance, while boron is added to nickel-based superalloys in casting to enhance grain boundary cohesion and improve creep properties [17], its addition in L-PBF is detrimental, causing hot cracking due to reduced solidus temperature and the formation of a stabilized liquid film at lower temperatures [18].Therefore, a novel alloy design approach is necessary, taking into account the processability during L-PBF.
Metal AM has been extensively reviewed by various research groups from different perspectives, for instance the microstructure [19,20], processing [21,22], numerical modeling [23][24][25], mechanical properties [21,26,27], and post-treatments [28][29][30].The alloy development has also been reviewed, including Ti-based [31][32][33][34][35], Al-based [20,36], Nibased [37,38], Fe-based [2,39], and Mg-based alloys [40,41].However, most of the work focuses on adapting existing alloys to the L-PBF process, rather than developing specific alloys dedicated to L-PBF.Furthermore, there is a lack of analysis on the correlation between elements and the printability of an alloy system in existing reviews.Additionally, less attention has been paid to comparing the weldability and printability of the existing alloys.Therefore, in this work, we first analyze the differences between conventional casting/welding processes and L-PBF process.Subsequently, the weldability and printability of commonly used alloy systems are analyzed, i.e.Fe-based, Ni-based, Al-based, and Ti-based, assessing the validity of using convention weldability to predict printability and identifying problems when using existing alloys in L-PBF.Subsequently, strategies used in alloy design for L-PBF are summarized and categorized into crack mitigation-oriented, microstructure manipulation-oriented, and machine learning (ML)-assisted approaches.At last, the current development of alloy design for L-PBF is summarized, and suggestions for future work are provided.The purpose of this review work is to provide guidance and new insights for designing novel alloys specifically tailored for L-PBF.[44].Reproduced from [44].CC BY 4.0.

L-PBF
The prevalent metal AM technologies primarily include powder bed fusion (PBF) and directed energy deposition (DED), as shown in figure 1.The metal powder in PBF is spread uniformly over a build platform, and each layer is selectively melted according to the 3D model.In DED, both metal powders and wires can be used to precisely deposit on a substrate or previously deposited layers through a nozzle or feedstock mechanism.In general, PBF processes typically offer higher resolution and accuracy, resulting in parts with fine details and intricate geometries, comparing to DED.However, DED technologies offer greater deposition freedom than PBF, as they can be used to fabricate and repair parts in a more flexible manner, enabling large-scale production.
The PBF can be further categorized into L-PBF and electron PBF (E-PBF) based on the types of heat sources employed, as shown in figure 1.The laser beam used in L-PBF is a coherent beam of light with specific wavelengths, typically in the infrared range.The electron beam used in E-PBF consists of accelerated electrons.It has a larger interaction volume and can penetrate deeper into the material, thus creating larger melt pools than laser.Furthermore, the build chamber during E-PBF can be heated to a much higher temperature, e.g.1000 • C, while most of the commercial L-PBF equipment can only preheat the baseplate to 200 • C. L-PBF typically exhibits high cooling rates due to the localized melting and rapid solidification of the metal powder.The localized melting and solidification result in steep thermal gradients between the molten and solidified regions.Therefore, L-PBF can produce fine-grained microstructures.In this review, we primarily focus on alloy design for the L-PBF technology, but will also cite multiple works on E-PBF or DED as a comparison or reference to inspire future alloy design work for L-PBF.
As shown in figure 1, during L-PBF, a thin layer of metal powder, typically around 20-100 µm thick, is evenly spread across a metal plate.A laser beam is precisely guided across the powder bed in a pattern defined by the 3D model.As the laser beam scans the powder bed, metallic powders absorb laser light and convert it into heat energy.It should be noted that part of the laser energy can reflect off the powder's surface due to its high reflectivity.When the temperature reaches or exceeds the powder's melting point, the particles start to melt, forming a localized molten pool and then cooling down rapidly.After each layer is melted and solidified, the build platform is lowered by a layer thickness.A new layer of metal powder is then spread across the previously printed layer.The process is repeated, building the object layer by layer until it is fully formed.Metal powders used in L-PBF should have a controlled particle size and shape, as well as good flowability and homogeneity.These properties are crucial for uniform powder spreading and influence the laser-powder interaction and energy absorption, impacting the final microstructure and properties of the printed object [42].
Comparing to conventional casting, metallic parts undergo much higher cooling rates (10 3 • C • s −1 to 10 8 • C • s −1 ) [43] and heating rates during L-PBF.Due to the layer-wise printing nature, the solidified section experiences multiple thermal cycles, which is termed as intrinsic heat treatment (IHT).The IHT generates substantial internal thermal stresses and dislocations within the solidified part, playing a significant role in the part properties.This section provides an overview of the key characteristics of L-PBF, highlighting the associated advantages and challenges.

Features
In L-PBF, metallic parts undergo complex thermal cycles [45,46].Initially, the powders interact with the laser beam and are melted at a fast heating rate.As the heat source moves away, the melt pool cools down rapidly and is solidified.The as-solidified parts then experience additional thermal cycles arising from the melting of subsequent tracks and neighboring layers.

Rapid solidification
During L-PBF, the layer-by-layer scanning of high-power laser is essentially a non-equilibrium process [47].A large amount of heat input will be applied to the metal powder, and the powder temperature can exceed  [56].Reprinted from [56], Copyright (2021), with permission from Elsevier.(b) L-PBF processed over-eutectic Al-50Si alloy.Complex thermal physical processes generate non-equilibrium phase distribution [20].Reprinted from [20], Copyright (2021), with permission from Elsevier.(c) Longitudinal sections of mass flow velocity and temperature distribution of depositing single track at different times [48].Reprinted from [48], Copyright (2020), with permission from Elsevier.(d) Anisotropic stress field in each layer under the influence of multiple anisotropic beam paths [55].Reprinted from [55], Copyright (2019), with permission from Elsevier.2000 • C (figure 2(c)) [48].This results in a steep temperature gradient from the center of the melt pool to its boundary.Upon laser removal, the material rapidly solidifies at a cooling rate of roughly 10 5 • C • s −1 [49].As a comparison, the cooling rate in conventional casting is approximately 0.5 • C • s −1 [43].Consequently, the additively manufactured (AM'ed) microstructure is one or two orders of magnitude smaller than the as-cast materials [50].The repeated ultrafast heating and cooling cycles during L-PBF also gives rise to a higher dislocation density comparing to their cast counterparts [12].Moreover, the AM'ed microstructure exhibits supersaturated solid solution with micro-segregation along dendrite arms [51] as a result of the high cooling rate.Furthermore, the steep temperature gradient leads to an intense Marangoni flow in the melts [52], affecting the chemical uniformity within the melt pool [53].

IHT
In casting, materials typically undergo a single heating-cooling cycle.In contrast, in L-PBF, the powders are melted and cooled first, and the resulting solidified part can undergo multiple cycles of reheating or even remelting as the laser scans adjacent hatches or subsequent layers.The intricate physical process and thermal history result in high thermal residual stresses, heterogeneous metastable microstructures, and non-equilibrium compositions or phase distributions within the final part [20,54,55] (figures 2(b) and (d)) [20,48].
For instance, the IHT during L-PBF enables the element partitioning at grain boundaries and potential precipitate formation.However, this unique feature of L-PBF can also bring manufacturing challenges for some specific alloys.The precipitates at grain boundaries, dendritic boundaries, and melt pool boundaries can cause boundary liquation or local stress/strain concentration due to the layer-by-layer scanning pattern during L-PBF.For instance, micro-segregation of Nb and Ti in L-PBF Inconel 718 alloy leads to the formation of brittle Laves and δ phases, increasing the crack susceptibility.Homogenization heat treatment is thus generally employed to equalize element distribution (e.g.Ti and Nb in Inconel 718) and potentially dissolve harmful phases (e.g.Laves and δ phases) present in the as-printed alloys [54].Another typical characteristic of L-PBF is the high internal thermal stress, which is accumulated during thermal cycling.The thermal stress could cause cracks and geometric distortions of the components during L-PBF (figure 2(a)) [56].Furthermore, the L-PBF-processed alloys are inherently anisotropic [57] as a result of epitaxial grain growth, which is prevalent in L-PBF.Large columnar grain is known for the capillary effect and poor strain accommodation capability [48].As a result of the microstructural non-uniformities [54], the mechanical response depends on the loading direction relative to the build direction, leading to anisotropic mechanical properties.

Defects
In L-PBF, the quality of printed parts is influenced by multiple factors such as metallic powders (flowability, particle size distribution, et al), laser beam (beam diameter, beam intensity profile, et al), and scanning strategies (hatch manner, interlayer rotational angel, et al).The defects occurred in AM'ed alloys primarily include gas pores, keyhole, lack-of-fusion (LoF) pores, and cracks, as shown in figure 3 [58].
The gas pores in the as-built microstructure typically originate from either the gas pores in the feedstock powders or the shielding gas entrapped during intensive melt pool flow in L-PBF.The gas pores primarily occur at occasions when there is insufficient time duration for the gas bubble to escape from the melt pool prior to solidification.
LoF pores, as the name implies, are due to the insufficient melting of powders and are evident to occur at low energy input.The un-completely melted powders constantly leave a triangle space between two neighboring hatches.Due to the sharp appearance, the LoF is detrimental to the mechanical behavior.
In contrast to LoF, which occurs at insufficient energy input, the keyhole defect is observed at excessive heat input.During L-PBF, the high-energy laser beam interacts with the powder beds.The powders at the surface boil and form a strong steam jet as a result of local heating.The steam jet then pushes the melt down due to the recoil pressure, forming a vapor depression (figure 3), which is known as keyhole [59].
The keyhole, gas porosity, and LoF defects (figures 4(a)-(c)) can be minimized or completely removed in the as-built parts by optimizing the scanning parameters or applying hot isostatic pressing (HIP) treatment.However, cracks, which are intolerant in most of the L-PBF-processed parts, are still intractable for the majority of the existing metals and hinder the application of L-PBF in the industry.
Depending on their origins, cracks occurred during L-PBF are categorized as hot cracking and solid-state cracking [60].Hot cracking includes solidification cracking and liquation cracking.Solid-state cracking involves strain-age cracking (SAC) and ductility-dip cracking (DDC).
Solidification cracking, also termed as hot tearing during welding, is believed to occur at the last stage of solidification [61].Due to the rapid cooling, solute atoms are enriched in-between adjacent dendrites, reducing the solidus of the residual liquid.The shrinkage strain caused by solidification and thermal contraction can pull apart the residual liquid.If the liquid is insufficient to backfill the inter-dendritic cavities, cracks form with clear dendritic characteristics [62] (figures 4(d)-(f)).It is widely accepted that the semi-solid zone is prone to solidification cracking [63].Therefore, narrowing the solidification temperature range results in a small slurry zone and a reduced hot cracking susceptibility.
Liquation cracking occurs when the low melting point phase present in a solidified part is liquefied again due to the reheating from the adjacent regions (figures 4(g)-(i)) [64].In conventional welding, the liquation cracking is observed at the heat affected zone.During L-PBF, the liquation cracking generally occurs near the melt pool boundaries, which is attributed to the segregation induced liquation under the effect of cyclic heat treatment [65].The enrichment of solute, e.g.boron [18] in superalloys, at inter-dendritic region significantly reduces the local solidus temperature, promoting the formation of liquation cracking.
during L-PBF.The γ' phase might precipitate during the layerby-layer printing, leading to local stress concentration.The concentrated stress is superposed with the large thermal stress accumulated during L-PBF, leading to SAC.
The mechanism of DDC has not yet been completely understood.The grain boundary inoperability at the intermediate temperature range was found to be one of the factors causing DDC [66][67][68].It is found that the ability of grain boundary sliding is significantly reduced at tortuous grain boundaries or at grain boundaries where precipitates like γ' and carbides are present.The suppressed grain boundary sliding is beneficial for enhancing creep resistance, yet, it causes strain concentration and generates voids.The connection of these voids significantly lowers the ductility and eventually leads to DDC (figures 4(m)-(o)) [64].
Both processing parameters and alloy chemistry are found to have a significant impact on the cracking behavior.However, no consensus on the root cause of cracking has been reached yet.In general, the alloys with high solidification shrinkage (nickel-based superalloys) and thermal expansion coefficient (aluminum alloys) are susceptible to cracking during L-PBF [69].Moreover, the refractory metals (e.g.tungsten, tantalum, and molybdenum) and metals with high laser reflectivity (e.g.pure copper) are less printable as they are difficult to melt [70].The alloys with volatile elements (magnesium, zinc, etc) are also challenging due to the server evaporation during L-PBF.
In the past, most of the work focused on the adaption of existing alloys to AM process.However, the thermal circumstance, solidification process, the melt pool kinetics, etc of L-PBF are distinct from those of the conventional processes.Only an integrated consideration of the alloy design and the unique features of L-PBF could lead to the synergistic advancement of new materials and processing technologies [71].

Printability of existing alloys
Defects, such as cracks in alloys, are intolerable for most of the applications.However, as mentioned in section 2.2, the existing alloys designed for conventional processes are not always suitable for the L-PBF process.As a result, defects such as pores and cracks can occur in alloys processed with L-PBF, and these defects are difficult to eliminate during the post-heat treatment.While HIP can help to close internal cracks and pores, it is not capable of closing surface cracks.Moreover, HIP is both costly and time-consuming, which is less favorable for alloy processing.Hence, it is crucial to design alloys with higher crack resistance to prevent cracks from occurring in the first place.
Similar to weldability, which is widely used in the welding industry to characterize whether an alloy is easier to weld than others, printability is employed in the AM industry.It is the term rating the ability of an alloy to be printed without defects, particularly cracks [15].The printability of an alloy is generally determined by trial-and-error tests.In the literature, there are several mathematical and empirical models assessing the cracking vulnerability of alloys during L-PBF.Most of the models or equations were modified based on the understanding of welding or casting processes.However, it is worth mentioning that AM, particularly L-PBF, has much steeper temperature gradients and higher cooling rates than welding and casting, leading to large deviation in microstructure and cracking behavior.Therefore, it is essential to know whether the criterion for weldability is still valid for the L-PBF printability for designing easy-to-print alloys.In this section, the printability of the most commonly studied alloys, i.e. steels, nickel-based alloys, aluminum alloys, and titanium alloys, is evaluated with special attention on the comparison with weldability.The influence of each element on the crack formation is analyzed statistically, providing fundamental knowledge for alloy design.
The research emphasis of AM'ed steel was mainly on property enhancement as fully dense steel parts seem to be easily manufactured [77,78,85].The mechanical properties of these steels are comparable to their conventionally made counterparts.The printability of steels is fairly satisfactory.However, the fact is that only a handful of commercial steels can be printed without cracks, which is marginal compared with more than 3500 commercial steel grades [86] in practical applications.Furthermore, behind the decent manufacturability, numerous parametric studies have been conducted to screen out the processing window in which successful deposition is obtained.It is of great value to examine the printability of the steel alloys, aiming at exploring the key elements leading to cracks during L-PBF.
In the welding industry, the term carbon equivalent (CE) was widely accepted as an indicator to quantify the weldability of steels, as shown in equation ( 1) According to the welding industry, steels with CE above 0.35 wt.% have poor weldability [87].Based on the steel composition, the CE of each steel was calculated and displayed along with their printability in table 1. Surprisingly, the high CE value does not imply poor printability as expected.On the contrary, steels with unsatisfied printability, such as tool steels, have low CE but high carbon contents.The low CE in tool steels calculated with equation ( 1) is attributed to low Cr and Ni contents.This result demonstrates that the L-PBF printability is not merely linked to CE, but also affected by the microstructure which is associated with all the elements in steels.
In addition to CE, a Schaeffler diagram was widely used to estimate the phases of weld deposits based on the weldment   composition, as shown in figure 5.The Cr and Ni equivalents are introduced in the diagram, giving the proportions of austenite, martensite, and ferrite in the microstructure.As displayed in figure 5, it is evident that the cracked steels during L-PBF mainly have martensitic microstructure, which is consistent with that observed in the weldment.Ferrite is found to resist hot cracking due to its high solubility of P and S [74].Austenite has low solubility of P and S, thus facilitating the inter-dendritic segregation and leading to cracks [88].Nevertheless, compared to ferrite, austenite has better ductility and a lower thermal expansion coefficient, which can reduce the risk of distortion and cracking during welding.Therefore, it is suggested to obtain ferrite-austenite bimodal phases at the end of solidification [74].Compared with the planar grain boundaries in a single phase, the torturous inter-dendritic regions in ferrite-austenite solidification mode are more difficult for the liquid film to wet, suppressing crack propagation [74].Compared with ferrite and austenite, martensite has limited weldability due to the high carbon content and the body center cubic (BCC) crystal structure, which can lead to hard and brittle martensitic transformation during welding [89].Preheating and post-weld heat treatment are generally applied to reduce the cooling rate and facilitate the formation of soft phases such as austenite, improving the ductility and toughness.As shown in figure 5, the steels consisting of martensite are also more vulnerable to cracks compared to those with austenite and/or ferrite.This certifies that the effect of steel phases on weldability is consistent with that on the printability in L-PBF.
To summarize, the printability of steels in L-PBF is influenced by key elements such as carbon and the phase constitutions.However, it should be noted that the CE, which is commonly used to assess the weldability of steel alloys, is not valid in ranking the printability of the steels processed with L-PBF.The printability is also significantly affected by the steel microstructure.For instance, the presence of martensite in the as-printed microstructure can lead to higher susceptibility to cracking.This analysis suggests that conventional weldability may not always align with printability in L-PBF.The Schaeffler diagram is valuable in predicting steel printability with its chemistry.It is worth noticing that the Schaeffler diagram does not consider non-equilibrium solidification.During L-PBF, the elemental partitioning and the high thermal stress can induce unforeseen local phase transition, resulting in a more complex microstructure, deviating from that predicted in the Schaeffler diagram.

Ni-based alloys
Nickel-based superalloys generally have high strength, excellent creep resistance, and fatigue life at high temperatures [97].They are widely used in severe high-temperature environments such as the marine and aerospace industries [98].Conventional manufacturing of nickel superalloys includes casting, directional casting, forging, and rolling.However, the more and more complex part structure challenges conventional manufacturing methods.A more geometry-adapted fabrication method is an urgent need.In recent years, L-PBF of nickel-based superalloy has been extensively studied and great progress has been achieved.For instance, more than 30 000 fuel nozzle tips have been successfully printed with Nibased alloy by GE Aviation and used for LEAP engines [99].However, the AM of nickel-based superalloys is still facing severe cracking problems.To date, only a limited amount of nickel superalloys can be AM'ed within a large crack-free window (the crack-free processing window is defined as the laser scanning parameter gap in which alloys can be printed without cracks and the scanning parameter can be input energy density), such as Inconel 718 and Inconel 625.
Ni-based superalloys can be strengthened by solid solution (Hastelloy X (HX)), γ' precipitates (CM247LC et al), and γ" precipitates (Inconel 718 et al).The alloys containing high fractions of γ' phase are known for their non-weldable behavior due to the high Al + Ti contents (>6.5 wt.%).Based on the welding industry, the nickel-based alloys containing Al + Ti > 4.5 wt.% are typically considered as nonweldable alloys [100], as shown in figure 6. Haafkens and Matthey proposed a more generalized weldability assessment chart involving Cr and Co content, as shown in figure 7 [101].The composition and printability of some nickel-based hightemperature alloys are shown in table 2.
Considering that L-PBF is essentially similar to the welding process, it is interesting to know whether the weldability criterion is still valid for assessing printability.As shown in figures 6 and 7, the cracked nickel alloys are scattered in both charts, particularly in figure 7 involving Cr and Co, manifesting that the printability of nickel-based superalloys is not always consistent with their weldability.For instance, the HX and Haynes 230 were readily weldable, according to figure 7.However, both alloys encounter severe cracking during L-PBF.Through efforts over the years, the crack-free HX by L-PBF is eventually achievable (this is also the reason that HX is labeled as a printable alloy in figures 6 and 7), but with strict control of minor elements.Tomus et al [102] found that the  HX powders with low Mn and Si contents could significantly reduce cracks during L-PBF.Later, they found that the Si and C contents played a major role in cracking mitigation as the two elements facilitated the formation of σ phase and carbides (MC, M 6 C) due to the inter-dendritic segregation.Though the cracking mechanism of HX during L-PBF is still debatable, crack-free HX powders are commercially available nowadays.Unlike HX, the cracking of Haynes 230 during L-PBF has not been completely solved without composition modification [103].ABD900 and ABD850 are two newly developed nickelbased superalloys specifically for L-PBF [72].Instead of having poor printability, as depicted in figure 7, the two novel alloys are fairly processable during L-PBF without cracks.In the design of these two alloys, researchers considered the solidification temperature range, low melting point eutectic phases, and solid-state cracking.
The discrepancy between the printability and weldability of nickel-based superalloys is attributed to the unique processing characteristics associated with L-PBF, e.g.fast cooling rates, steep temperature gradients, cyclic heat treatment, and high thermal stresses.In conventional welding, higher Al + Ti content leads to an increased amount of γ' phase, while excessive Ti + Al can result in the precipitation of the σ phase.The σ phase acts as nuclei for the formation of harmful topologically close-packed phases, which can lead to welding cracks.However, in L-PBF, the rapid cooling rate significantly suppresses the precipitation, resulting in lower fractions of precipitates compared to what is expected based on the Ti + Al criterion.Therefore, the traditional weldability assessment criteria, i.e.Ti + Al content, may not directly apply to printability prediction of nickel-based superalloys in L-PBF.
In summary, the conventional metal manufacturing process considers the bulk composition and relatively low cooling rates, which enables substantial elemental distribution.However, during L-PBF, the extremely fast cooling rates combined with the IHT lead to a complicated near-atomic-scale composition variation, deviating significantly from the bulk composition [18].Therefore, the bulk criterion for Ni-based alloy's weldability, e.g.(Ti + Al)%, is not always valid for the atomic-scale occasions during L-PBF.An in-depth understanding of the solute behavior at the atomic scale allows us to derive strategies to avoid cracking during L-PBF.

Al-based alloys
Aluminum alloys are known for their excellent strengthdensity compromise and are widely used in aerospace and automobile industries, in which both lightweight and mechanical properties are needed [15].L-PBF of aluminum alloys enables the design of geometrically complex parts with refined microstructure and enhanced mechanical performance [118].Most Al-based alloys currently investigated in AM are directly adopted from those developed for the conventional cast and post thermo-mechanical processes but are not optimized for AM processing.Therefore, the intrinsic rapid solidification characteristics during beam-based AM processes raise significant challenges for the manufacturing of Al-based alloys, such as the high reflectivity, hot cracking, excessive loss of volatile alloying elements due to evaporation, porosity formation, and inadequate mechanical performance for structural applications that require high strength [5,30,34].
High reflectivity of Al and its alloys at a laser wavelength (around 1064 nm) necessitates a high laser energy density during L-PBF and L-DED processing to achieve a dense part [119].The high energy input during beam-based AM processing leads to the evaporation of versatile elements.Significant vaporization losses of alloying elements, such as Zn, Mg, Li, and Mn, have been widely reported in Al-Mn (3xxx), Al-Mg (5xxx), Al-Zn-Mg (7xxx), and Al-Li-based (8xxx) alloys during laser-based AM processing due to their low boiling points (907 • C for Zn, 1090 • C for Mg, 1342 • C for Li, and 2061 • C for Mn) and high equilibrium vapor pressure (Zn > Mg > Li > Mn > Al > Fe > Ni > Ti) [120,121].In addition, Al-based alloys suffer from porosity issues during AM processing.The formation of excessive pores is mainly related to impropriate process conditions (LoF or keyhole pores), gas pores (vaporization of alloying elements, entrapment of shielding gases or air), and hydrogen pores (moisture, grease, or hydrocarbon contaminants) due to the significant solubility difference of hydrogen in liquid and solid aluminum [122].Hence, optimizing the process parameter and enhancing the protective environment help in minimizing the porosity defects.

The composition, printability and weldability of
Similar to that in welding, hot cracking is one of the most critical problems in AM processing of Al-based alloys.Majority of Al-based alloys required for structural applications in aerospace and automotive industries are prone to hot cracking during beam-based AM processing.The nonweldable 2xxx series Al-Cu-Mg-based and 7xxx series Al-Zn-Mg-based, particularly Al2024 and Al7075, are not printable during AM due to their high hot cracking susceptibility [120,123].During conventional welding, the 5xxx series Al-Mg-based and 6xxx series Al-Mg-Si-based alloys are susceptible to hot cracking or stress corrosion cracking depending on the Mg content [124].Fortunately, the cracking can be readily prevented by using filler metals, such as Al-Mg-Mn (Cr, Ti) (Al5356), AlMg4.5Mn0.7 (Al5183), or Al4043 (AlSi5) [65].However, the 5xxx and 6xxx series aluminum alloys suffer from hot cracking during AM processing where filler metals are not available [125][126][127].It should be noted that manufacturing crack-free 2xxx series and 6xxx series aluminum alloys using L-PBF has proven to be possible by tuning the process conditions.For instance, hot cracking of Al-Cu-Mg-Mn can be avoided during L-PBF processing using a combination of low laser power (200 W) and low scan speed (83.3 mm s −1 ) due to the enhanced liquid backfilling [128].Crack-free AA6061 parts can be manufactured using high temperature baseplate preheating at 500 • C during L-PBF due to significantly reduced thermal stresses [129,130].However, these strategies are not feasible solutions for widely industrial applications because the former sacrifices the production rate and the latter can cause substantial powder sintering and part oxidation.The current printable commercial Al-based alloys are limited to the 4xxx series, Al-Si-based alloys with a near eutectic composition, or 3xxx series Al-Mn-based alloys, for instance, AlSi7Mg, AlSi10Mg, and AlMn3 alloys [129,130].However, the low-to-medium mechanical properties of Al-Si-based and Al-Mn-based alloys could not meet the property requirements for structural applications in the aerospace and automotive industries.Therefore, there is a strong need to develop new high strength Al-based alloys tailored specifically to AM.
To design new Al-based alloys that can be processed crackfree using fusion-based AM technologies, it is crucial to identify the key influencing factors and their coupling effect on the cracking susceptibility.Table 3 lists the chemical composition and printability during fusion-based AM processing of commonly used aluminum alloys.A Pearson correlation map was constructed to screen out the key factors responsible for the cracking of aluminum alloys, as shown in figure 8.The Pearson correlation coefficient is a statistical measure used to assess the strength and direction of the linear relationship between two variables.It is calculated using the following formula: Here, ρ represents the Pearson coefficient, x1 and x2 are the two variables of interest, COV denotes the covariance between x1 and x2, and σ(x1) and σ(x2) represent the standard deviations of x1 and x2, respectively.The Pearson coefficient ranges from −1 to 1, where 1 represents a perfect positive correlation, −1 indicates a perfect negative correlation, and 0 signifies no linear correlation.By calculating the Pearson correlation coefficient, one can gain insights into the interrelationship between variables and assess the strength and direction of their association.
In this study, the Pearson correlation diagram was generated using the SciPy module in Python, incorporating data from table 3. Table 3 presents the calculated values for three features: solidus temperature (ST), brittleness temperature range (∆T), and maximum solidification cracking index (SCI) at the brittleness temperature range.The SCI was calculated based on Kou's criterion [131], which was determined by the maximum slope of the T-(f s ) 1/2 curve, where f s is the solid fraction.Brittleness temperature range (∆T) represents the temperature range when the solid fraction is between   0.8 and 0.99 [132].The three features were calculated using Scheil model in Thermo-Calc program with the TCAL6 and MOBAL5 database.The default phase composition, a step size of 5 • C, an ambient temperature of 25 • C, and a pressure of 10 6 Pa were used during the calculation.Once the Scheil solidification data was obtained, a Python program was written to process the exported data and calculate the three features.By examining the correlation between each element and these features, it is possible to identify the element's influence on the solidification process and its association with hot cracking.Note that this batch of alloys was not treated with the addition of secondary particles to eliminate hot cracking, as their crack elimination mechanism mainly involves grain refinement, which has little impact on the Scheil solidification curve.Therefore, they will be discussed separately in section 4.2.3.
As depicted in figure 8, Si and Mg are the two elements that have strong correlations with the three features.Si positively correlates with solidification temperature and negatively correlates with the maximum SCI (SCI max) and brittleness temperature range (∆T).It means that increasing Si content leads to a higher solidus and a lower SCI and ∆T, thus reducing cracks.In contrast to Si, increasing Mg content results in a decreased solidus and increased SCI and ∆T.This result implies that Mg should be lowered to reduce cracks during L-PBF.Furthermore, based on the Pearson correlation, Mg has the strongest linear correlation with ST, followed by Zn and Cu.
The specific correlations between key alloying elements on the three features were plotted in figure 9. Increasing Si content raises the ST but reduces the SCI and ∆T.Higher Mg content results in a lower ST, and higher ∆T and SCI, increasing crack vulnerability, which is consistent with the Pearson correlation.Based on the statistical analysis, increasing Si and decreasing Mg are beneficial for crack mitigation for L-PBF-processed aluminum alloys.This is also in line with the experimental observations.Hyer et al [133] investigated the effect of Si concentration on the cracking of Al-Si binary alloys processed via L-PBF.The results demonstrated that higher concentrations of Si in the binary Al-Si alloys exhibited lower cracking susceptibility.Moreover, Si is present as fine Si particles at low Si concentration and a web-like network of Al-Si eutectic at high Si concentration.An increase in Si concentration results in a eutectic network structure, which can act as a barrier to dislocation motion, enhancing the alloy strength.Li et al [134] studied the Al-xMg-0.2Sc-0.1Zr(x = 1.5, 3.0, 6.0 wt%) alloys and found that the crack density during L-PBF increased with the Mg content.However, when 1.3 wt.% Si was added, the hot cracking disappeared due to the significantly refined microstructure and the formation of interdendritic Al-Mg 2 Si.
To summarize, the aluminum alloys typically exhibited poor printability and weldability.This is primarily attributed to the large solidification temperature range, which increases the hot crack sensitivity.Furthermore, the coefficient of linear expansion in aluminum is higher compared to the other three alloy systems [135], which could induce large thermal stress and increase crack vulnerability.Pearson correlation map helps screen out the decisive factors for cracking during L-PBF of aluminum alloy among a large number of elements.Big data and statistical analysis provide a potential methodology to discover the hidden correlation between alloy chemistry and cracking susceptibility.

Ti-based alloys
Titanium alloys, possessing desired strength, biocompatibility, and excellent corrosion resistance, have been widely used in biomedical, aerospace, and defense industries.Depending on alloying elements, titanium alloys can be classified as α, near α, α-β, near β, β-titanium, and metastable β titanium alloys [71].Conventional titanium producing processes include forging, casting, and powder metallurgy.Machining is generally needed to achieve the final geometry of the parts.However, titanium alloys are known as hard-to-machine materials due to their low Young's modulus and thermal conductivity, hindering their wide application in diverse industries [157].As L-PBF is a near-net shaping technology, manufacturing hard-to-machine materials such as titanium alloys has drawn extensive attentions.
The easy-to-print titanium alloys include CP-Ti, which is pure titanium, and Ti6Al4V.The AM of CP-Ti initially faces the impurity problem, particularly in controlling the oxygen level as titanium can form strong bonding with oxygen.Recent studies demonstrated that the oxygen content can be wellcontrolled, and the density of CP-Ti can achieve 99% [158].Another titanium alloy that has been extensively studied in the AM community is Ti6Al4V, which is a typical α-β alloy.The as-fabricated Ti6Al4V has predominantly refined α' martensite through the β to α' martensitic transition, leading to a higher yield strength (YS) (>1100 MPa), but a lower elongation to failure (5%-10%), compared to the wrought alloy [71].Lu et al [159] found that the tensile strength is determined by the content of fine α' martensite and lath thickness, while plasticity is influenced by prior β grain size and lath length.Increasing α' fraction and refining prior β grains and α laths simultaneously improves strength and plasticity.Wang et al [160] improved the ductility and mechanical isotropy by reducing the Al content in Ti6Al4V from 6 wt.% to 4 wt.%.This is due to the activation of multiple slip modes and twins during solidification and deformation.Chen et al [161] utilized blended powders of Ti6Al4V and 3 wt.%Fe particles during L-PBF.The as-built microstructures transit from α' dominated microstructure to a nearly complete β-dominated microstructure, exhibiting high strength and enhanced ductility without post-heat treatments.Oh et al [162] obtained martensitic α' and α + α' + β microstructures by utilizing two different laser scanning speeds.The critical cooling rate of fully martensitic transformation is in the range between 2900 • C s −1 and 6500 • C s −1 .Although Ti6Al4V is widely used due to its inherent resistance to in-vivo corrosion, favorable osteointegration properties, and excellent strength-to-weight ratio, the release of toxic vanadium (V) ions in the human body poses health risks [163].Issariyapat et al [164] investigated the Ti-Zr alloy as a substitute for Ti6Al4V.They successfully prepared high-density Ti-Zr materials with homogeneous Zr distribution through L-PBF using pre-mixed blends of Ti 0-10 wt% ZrH 2 powders.The presence of Zr promoted grain refinement, resulting in a fine acicular grain structure.Increasing Zr content significantly enhanced the tensile YS, while maintaining acceptable ductility.Ti-6Al-7Nb alloy, which replaces V with Nb in its composition, exhibits a martensitic α' phase hardened by dispersed precipitates.Chlebus et al [165] studied the effect of L-PBF manufacturing strategy on the mechanical properties and microstructure of Ti-6Al-7Nb alloy and found higher tensile and compressive strength but lower ductility compared to conventional methods.Ti-24Nb-4Zr-8Sn (Ti2448) is a near β-Ti alloy with a low Young's modulus (42-50 GPa) [166].Introducing porosity further reduces the Young's modulus [167], making it suitable for porous titanium structures that match the human skeleton (4-30 GPa), which can be produced using L-PBF [168].
As discussed above, different from nickel-based and aluminum alloys, studies on the crack formation and mitigation of titanium alloys during L-PBF are rarely reported in literature, suggesting a good printability of titanium alloys.Compared to nickel-based superalloys, titanium alloys have lower Young's modulus [169].The low Young's modulus enables titanium alloys to accommodate larger strains during L-PBF.Such large deformation of titanium releases the high residual stress accumulated during IHT of L-PBF process, alleviating the stress concentration and thus cracking behavior.Therefore, fully dense titanium alloys are achievable by optimizing scanning parameters, as shown in table 4. The L-PBF of titanium alloys mainly focuses on suppressing epitaxial grain growth of long columnar grains, enhancing mechanical properties, and designing lattice structures for energy absorption and biomedical applications.A molybdenum equivalent MoE is also widely used to analyze the phase constitutions of titanium alloys, as shown in equation ( 3).If 0 ⩽ MoE < 5, the titanium alloy is rich in β phase.If 5 ⩽ MoE < 10, the alloy is near β phase.The alloy is metastable in the case of 10 ⩽ MoE < 30, whereas for MoE > 30, alloys are considered stabilized.MoE = 1.0 (wt.%Mo) + 0.67 (wt.%V) + 0.44 (wt.%W) + 0.28 (wt.%Nb) + 0.22 (wt.%Ta) + 2.9 (wt.%Fe) + 0.28 (wt.%Nb) + 0.22 (wt.%Ta) + 2.9 (wt.%Fe) + 0.28 (wt.%Nb) + 0.22 (wt.%Ta) + 2.9 (wt.%Fe) As discussed above, titanium alloys generally have fewer cracking problems during L-PBF comparing to Ni-based and Al-based alloys.This can be attributed to the smaller number of elements, which results in fewer low-melting-point precipitates at grain boundaries.Moreover, Ti-based alloys generally have a narrower solidification temperature range and a lower Young's modulus, which further reduces the hot crack sensitivity.Similarly, the weldability of titanium alloys is also quite satisfactory comparing to the other three alloy systems investigated in this work.
To summarize, while L-PBF offers numerous advantages, such as the ability to create complex geometries and produce functional metal parts, several specific issues can arise when using existing alloys in L-PBF.
(1) Defects: during the L-PBF process, rapid and localized heating and cooling cycles induce significant thermal stress in the printed part.This thermal stress can cause  cracking and distortion, as existing alloys are not specifically designed to resist such stress during the forming process.Alloys with narrower solidification ranges tend to exhibit better printability, while those with large solidification ranges like CM247LC and AA7075 are more prone to cracking during L-PBF.Alloys with higher thermal conductivity can dissipate heat more efficiently, reducing the risk of thermal stress build-up and cracking.Mismatched thermal conductivities can lead to thermal gradients, warpage, and residual stresses, affecting part quality.Additionally, L-PBF requires high-quality powders as feedstock.Powder characteristics, such as normally distributed particle size, spherical morphology, and excellent flowability, are required to achieve desired print quality.However, not every conventional alloy can be processed into powders that meet these requirements.Powders with entrapped gas, undesirable morphology, or poor flowability can lead to defects such as porosity and LoF in the printed part.(2) Compositions: the alloy impurity in conventional processes can be reduced to extremely low levels using specific equipment like vacuum degassing furnace.However, in the powder-making process, the lack of specialized equipment or contamination from the refractory materials used during powder-making could increase the impurity levels in the powders.Additionally, the use of certain alloying elements with low melting point in conventional alloys, such as Mg and Al, can lead to issues like excessive powder spattering or vaporization during laser exposure.Furthermore, despite flushing the L-PBF equipment with inert gases before printing, there might still be residual oxygen atoms present.This can result in oxidation of alloys that are prone to oxidation, which can degrade their mechanical properties.It is important to note that alloys highly reactive with oxygen, moisture, or other environmental gas, such as Mg, should be printed with specialized equipment or in controlled environments.(3) Microstructure: due to layer-wise build-up manner during L-PBF, the epitaxial grain growth is facilitated, leading to columnar grains along the building direction, which is undesirable for conventionally designed equiaxed grain structures.The rapid solidification during L-PBF can result in finer microstructures, but it can also lead to some metastable phases.The IHT occurring during L-PBF can further contribute to the formation of undesirable phases, necessitating additional post-processing like heat treatment.(4) Post-processing: L-PBF parts often require postprocessing steps such as heat treatment, HIP, or surface finishing to achieve the desired properties.However, these steps may differ from the conventional post processing optimized for casted or wrought alloys.Therefore, additional research and process optimization are required to develop post-processing techniques suitable for L-PBF fabricated parts using existing alloys.(5) Mechanical properties: L-PBF fabricated parts exhibit anisotropic mechanical properties due to the layer-by-layer building process.The microstructure anisotropy might limit their performance in certain applications.Furthermore, the LPBF-processed traditional alloys might have reduced ductility and lower fatigue resistance due to the formation of fine-grained or non-equilibrium microstructures.

Alloy design strategies for L-PBF
As discussed above, using existing alloys developed for conventional casting and forging processes in L-PBF can be challenging due to differences in solidification behavior, microstructure, properties, and post-processing requirements.To fully leverage the capabilities of L-PBF, it is crucial to develop alloys specifically designed for this technique.
Commercial alloys are designed for specific applications.The mechanical properties of commercial alloys have been certified and well-known.It is undesirable to modify the alloy composition as this would lead to deterioration of the mechanical performance.However, as discussed in section 3, the existing alloys are designed for traditional alloy-making processes without AM in mind.Consequently, server cracking is constantly observed in most of the commercial alloys processed with AM, particularly the L-PBF.Multiple works attempt to modify the commercial alloy composition within the specification to improve printability.This is conservative as it is challenging to completely remove all the cracks in this way although it does not need to undergo an additional certification process.Some newly designed alloys achieve great success in crack mitigation, such as ABD900-AM [72], a novel nickelbased superalloy tailored for the L-PBF process.However, a knowledge-based alloy design guideline specifically for L-PBF is still absent.Three primary strategies are found in the literature.With mechanical properties in mind, one focuses on crack mitigation by studying the cracking mechanism, and the second tends to refine the grain structure.The third strategy employs the recently popular ML process without an in-depth understanding of the cracking or grain refining mechanism.

Crack mitigation-oriented strategies
Even in conventional casting, the mechanism of cracking is still under debate, although this topic has been studied extensively for years [179].With the advance of AM, the attention on the cracking mechanism has been raised again.A lot of efforts have been spared on designing crack-free alloys for the L-PBF process.In this section, the crack criteria used during alloy design was presented at first.Subsequently, the grain and cell boundary manipulation work were reviewed as most of the crack-mitigation work focuses on the grain/cell boundary manipulations.

Cracking criteria
As discussed in section 2.2, cracks in LPBF-processed metals are classified as hot cracking, i.e., solidification cracking and liquation cracking, and solid-state cracking, which includes SAC and DDC.
RDG criterion for cracking is widely applied in L-PBF [180].According to this criterion, hot tearing occurs if the pressure within the semi-solid mushy zone drops beyond a critical value.Besides, a viscous liquid is more difficult to flow and backfill the existing porosity or crack at the end of solidification.In the RDG model, the grain boundary can be categorized into repulsive (stabilization of liquid films) and attractive boundaries (dendrite coalescence).The coalescence undercooling is found to be higher at high angle grain boundaries (HAGBs) than at low angle grain boundaries (LAGBs), indicating that the liquid film at HAGBs is stable at lower temperatures.This is consistent with the experimental observation that cracks are found to propagate along HAGBs, which is similar to welding [61].
Clyne and Davis [181] considered the time duration of the solidification intervals at high temperatures (between 10 and 60 vol.% liquid, t r , indicating the time available for stress relaxation) and low temperatures (between 1 and 10 vol.% liquid, t v , indicating the vulnerable time when cracks can propagate easily).The ratio of t v /t r is defined as the hotcracking sensitivity (HCS).This model has succeeded in predicting the cracking susceptibility in aluminum alloys [180].
Xu et al [63] proposed a heat resistance and deformation resistance (HR-DR) model for nickel-based superalloys processed via L-PBF.In this model, cracks are attributed to low HR-DR at grain boundaries.HR is linked to the ST difference between the interdendritic core (ID) and the dendritic core (DC) region.DR is associated with the difference in the YS between DC and ID.According to the HR-DR model, alloying elements such as Zr, B, and Hf dramatically reduce the cracking resistance due to their strong segregation to inter-dendritic regions.Moreover, the γ' fraction has limited influence on the L-PBF as their precipitations are suppressed during the rapid cooling in L-PBF.The HR-DR model has been successfully validated using a novel nickel alloy MAD542 [182] and the commercial Rene 104 (also termed as ME3).
Another widely used cracking criterion was proposed by Kou [131], who considered the correlation among (1) separation of grains from each other due to solidification contraction, (2) the growth of two neighboring grains toward each other, and (3) the liquid backfilling between the intergranular channel.Based on these understandings, Kou derived an SCI to evaluate the hot cracking vulnerability of an alloy during solidification, as shown in equation ( 4) where T is the temperature and f s is the fraction of solid.The index was successfully verified with experimental data in casting and welding of Al alloys [131].Additionally, Tang et al [72] calculated the SCI of nickel-based superalloys, and found that the SCI of CM247LC, Inconel 939 and Inconel 738LC was significantly higher than those easy-to-print alloys such as Inconel 718, which was consistent with the experimental observation.
Solidification temperature range/freezing temperature range has been widely employed to qualitatively assess the printability of alloys during L-PBF.Griffiths et al [183] found that decreasing Hf significantly reduces the solidification temperature range of the nickel-based superalloy CM247LC.An Hf-free version of CM247LC was developed with crack density reduced by 36%-75%.Tomus et al [102] used two batches of pre-alloyed HX powders with different Si, Mn, and C contents for L-PBF.It was found that low Si and C contents reduced the solidification interval, helping to avoid crack formation.Therefore, by lowering Si and C content, crackfree HX was fabricated via L-PBF.In terms of aluminum alloys, it is found that introducing silicon to aluminum alloys such as 2024, 6061, and 7075 can eliminate cracks due to the reduced solidification range, thermal expansion, and solidification shrinkage [184].However, the solidification temperature range is not always valid in interpreting cracks.For instance, Zhao et al [103] found that the addition of Zr (1 wt.%) in Haynes 320 resulted in a higher solidification temperature range.Instead of increasing HCS, the cracking was completely suppressed.Similarly, Sun et al [185] introduced Al element into a high-entropy alloy CoCrFeNi, which was crack susceptible.Although the solidification temperature range was enlarged, adding 0.5 wt.% Al significantly decreased the crack density.This was attributed to the grain boundary phase formation that lowered the tensile stress or even altered the stress state at the grain boundaries.

Grain\cell boundary manipulation
The in-depth characterization of cracks demonstrates that grain boundaries and cell boundaries are the most vulnerable regions for cracking in L-PBF-processed microstructure.The rapid solidification during L-PBF leads to a supersaturated solid solution matrix.On top of that, the solutes that partition into the inter-dendritic liquid are insufficient to equilibrate via diffusion.This results in elemental enrichment at the solidification front based on the classical constitutional undercooling theory.The elemental enrichment assists the formation of intermetallics, carbides, or oxide inclusions, as shown in figure 10.In casting, the intermetallic is beneficial for preventing hot cracking [186].However, in the L-PBF process, the intermetallic at grain boundaries might be detrimental to the grain boundary integrities [187].The occurrence of cracking during L-PBF is tightly associated with their boundary characteristics.Therefore, the concept of grain boundary segregation engineering was employed to manipulate the grain boundary features to counteract the cracking problems in metal AM [185].
Sun et al [185] found that the elemental segregation lead to a liquid film at the grain boundary at the end of solidification.Consequently, discontinuous precipitates formed along the grain boundary.The precipitates could accommodate the tensile residual stress through grain deformation and molar volume expansion.The tensile stress could thus switch into compressive stress, suppressing the formation of cracks, as displayed in figure 11(d).However, when the brittle precipitates form a continuous film, they promote crack initiation instead of acting as strain absorbers.Later, they developed a thermodynamics-guided approach for L-PBF alloy design using the concept of grain boundary segregation engineering [188].The authors solved the hot cracking problem of Inconel 738LC alloy through lowering the Si content from 0.11 wt.% to 0.03 wt.%.They concluded that the occurrence of hot cracking during L-PBF required (a) pores to act as crack nucleation sources and (b) liquid films with a low ST to facilitate crack propagation, i.e.B in their work.Zhou et al [107] found that with the carbon content increase from 0.11 wt.% (Inconel 738LC) to 0.3 wt.% and 0.6 wt.%, crack-free Inconel 738 was produced in a wide range of L-PBF parameters (76-126 J mm −3 ).With the increase in carbon content, the γ/γ' eutectic decreased, and the cracking sensitivity decreased correspondingly (figure 11(a)).
Kontis et al [18] found that a segregated film at an HAGB can switch into a liquid film, as depicted in figure 11(b).The local thermal stresses and segregation-induced grain boundary liquation jointly cause hot cracking.Fewer grain boundaries lead to more concentrated and continuous liquid films extending over several millimeters, which is detrimental to mechanical performance.Therefore, they proposed to obtain an equiaxed or a columnar microstructure with a grain width smaller than 100 µm to avoid cracking, despite strong grain boundary segregation.
Thapliyal et al [189] proposed a theoretical framework applying the concept of segregation engineering of grain boundaries during L-PBF.In order to use the proposed framework, the authors address the premises, which are: (a) certain elements must segregate at grain boundaries due to the low solid solubility, (b) solute segregation at the grain boundaries leads to reduced energy and mobility of the boundaries, and (c) segregation of the solute may lead to either an enhanced or deteriorated boundary cohesion depending on whether the solute is of strengthening or embrittling type (figure 11(c)).They explored the concept to facilitate microstructural heterogeneity and hierarchy, suppressing cracking-susceptible columnar growth and leading to good printability at a wide range of process parameters.
Unlike most studies, in which researchers attempted to reduce the liquid film at grain boundaries, Zhao et al [103] introduced liquid backfilling and a network of segregation phases at grain boundaries to alleviate thermal stress.Following this strategy, they added Zr in Haynes 230 and obtained crack-free specimens with an extraordinary combination of strength and plasticity.
To summarize, the rejection of solutes during rapid solidification in L-PBF leads to cell/dendrite segregation.The enrichment of solutes in the residual liquid reduces the grain boundary energy and mobility and affects the grain boundary cohesiveness.Depending on whether the solute is of strengthening type or embrittling type, the grain boundary is then strengthened or easily cracked.Moreover, the segregationinduced grain boundary characterization can lead to the lowmelting-point phase formation, which is not desired considering the hot cracking.Nevertheless, the precipitates can also reduce the detrimental effect of residual thermal strain and facilitate liquid backfilling, which are beneficial for cracking mitigation.Therefore, the grain boundary segregation should be carefully manipulated in order to design novel crack-free alloys for L-PBF.

Microstructure manipulation-oriented strategies
In section 4.1, we summarized the work on designing alloys for L-PBF with the intention of removing cracks.However, crack mitigation is not the sole purpose of alloy design for L-PBF.In addition to cracks, the microstructure of LPBFprocessed alloys also significantly impacts the mechanical Schematic of the microstructural and compositional evolution during AM of the superalloys with hot cracks.The progressive enrichment in Cr, Mo and B at grain boundaries over the building process causes grain boundary segregation induced liquation [18].Reprinted from [18], Copyright (2019), with permission from Elsevier.(c) Schematic of solute segregation effect on interfacial energy and solidification behavior of metastable Fe-Mn-Co-Cr-Si high entropy alloy processed via L-PBF [189].Reprinted from [189], Copyright (2021), with permission from Elsevier.(d) Schematics of residual stress minimization for Al0.5CoCrFeNi alloy during L-PBF.The image indicates the Al segregation toward the interdendritic/grain boundaries, discontinuous BCC/B2 grains formed along the grain boundaries, the tensile residual stress accommodated by the BCC/B2 grains, and the switch from a large tensile residual strain a small compressive residual strain [185].Reprinted from [185], Copyright (2021), with permission from Elsevier.
properties, corrosion resistance and other functional properties, thus affecting the alloy's application.Generally, the layerby-layer printing nature of the L-PBF process is favorable for the columnar grain growth.Columnar microstructures, ideally single crystalline, are desirable for applications such as gas turbine blades to enhance creep and fatigue properties.However, in most cases, microstructure anisotropy is undesirable, and a refined and equiaxed microstructure is preferred.Furthermore, due to the fast cooling rate, the as-built alloys are constantly a supersaturated solid solution.While this may be favorable for solid solution alloys, precipitation hardening alloys typically require post-heat treatment to achieve optimal properties.Therefore, when designing alloys for the L-PBF process, it is crucial to consider the microstructure preferences of the specific alloy.By accounting for the desired microstructural characteristics during alloy design, extensive parameter optimization work and post-heat treatment can potentially be avoided, simplifying the overall manufacturing process.
The high cooling rate and high thermal gradient of L-PBF almost exclusively lead to columnar grains, which are undesirable considering the anisotropic mechanical properties in most of the applications [190].According to the wellestablished grain refinement theory in casting, nuclei provide a site for growth, and solute leads to constitutional supercooling, which activates adjacent nuclei [191].A growth restriction factor, Q, defined as the development rate of a constitutionally undercooled zone, is employed to assess the effect of an element on grain refinement [192].Tan et al [193] reconsidered the grain refinement theory and found that the classic theory (i.e.Q value) could not be directly applied to metal L-PBF.The large thermal gradient during L-PBF could significantly inhibit the formation of a compositional supercooling zone ahead of the solid/liquid (S/L) interface.In addition, the solute rejection resulted in a lag in the actual growth of the S/L interface compared to the theoretical solidification rate.The growth restriction factor (Q value) cannot characterize the role of solutes in L-PBF.Despite the lack of exact grain refinement theories, a considerable amount of work has been reported in the literature, achieving refined microstructure in newly designed or tailored commercial alloys for L-PBF.Three different strategies were found in the literature to promote the columnar-to-equiaxed (CTE) transition (CET) during L-PBF, i.e. optimizing process parameters such as in steels [194,195] and nickel-based alloys [196], the addition of solute elements with a high supercooling capacity like in titanium alloys [190,[197][198][199][200], and exploring grain refiners for aluminum alloys [201][202][203][204][205][206].
In this section, the microstructure manipulation-oriented strategies were discussed, including phase transition, CET, and grain refinement.Phase transition involves making use of the rapid cooling and IHT in L-PBF to induce specific phase transitions, such as the precipitation of secondary phases.CET is employed to promote the formation of equiaxed grains through increasing the supercooling capacity.Grain refinement is realized by adding nucleating agents to increase the nucleation sites and promote the equiaxed grain formation.

Phase transition
Due to the layer-by-layer printing nature and the complex thermal cycles during L-PBF, it is feasible to control the phase transition guided by thermodynamic calculation beforehand.Therefore, during alloy design for L-PBF, it is essential to take into account the potential phase transition that may occur during laser processing.
Guo et al [194] proposed a phase-transformation-guided approach for alloy development in the L-PBF of PH17-4 steel.Through controlling scanning parameters, they minimized the formation of δ-ferrite but facilitated the austenitization.Next, they removed the minor alloying elements C, Mn, and Si, which reduced the Ms temperature, to promote the austenite-to-martensite transformation.In the end, UW_17-4 steel was developed with a fully martensitic microstructure at various cooling rates (figure 12).The authors address the importance of understanding the phase transformation dynamics prior to alloy design for AM.Wang et al [195] fabricated 316L with L-PBF and obtained a super combination of YS and ductility, breaking the strength-ductility dilemma.It was interpreted that the unique cellular structures, LAGBs, and dislocations formed during L-PBF resulted in the high strength, and the work-hardening mechanism contributed to the high elongation.
The LPBF-processed alloy is constantly a supersaturated solid solution with a low YS and thus requires a post heat treatment.In order to avoid the post processing, in-situ precipitation by optimizing scanning parameters is proposed to induce nano precipitations under the IHT during AM.For example, Fe19Ni5Ti steel was tailored for laser AM by local control of both nanoprecipitation and martensitic transformation [196].The as-built steel was hardened in-situ by NiTi nanoprecipitation and it consists of alternating soft (without precipitates) and hard layers (with a high volume fraction of nanoscale precipitates) with a tensile strength of 1.3 GPa and 10% elongation.

CET
During L-PBF, if the easy growth direction of crystals is parallel with the thermal gradient, epitaxial growth is facilitated.As a result, columnar grains spanning multiple melt pools are constantly present in the as-built microstructure.Variation in processing parameters during L-PBF determines the temperature gradient and cooling rates, thus affecting the crystal morphology and phase transformation.Although a lot of attempts have been made to promote the CET during L-PBF by optimizing the scanning parameters as shown in figure 13 [65], it is still challenging to obtain fully equiaxed grains, particularly for titanium alloys [207].Therefore, it is significant to design alloys with high CET tendency in order to achieve fully equiaxed grain structures.
Mereddy et al [197] added different amounts of carbon to Ti6Al4V during AM, and they found that the addition of 0.03-0.41wt.% carbon reduced the prior β grain size and α lath by factors of 5-6.The carbon segregated at grain boundaries during rapid cooling decreased the solidification temperature, The microstructures for all conditions are fully martensitic.All IPFs share the same color code, which is shown in the inset of (a) [194].Reprinted from [194], Copyright (2022), with permission from Elsevier.leading to constitutional supercooling.Higher constitutional supercooling capacity thus suppresses the epitaxial growth, leading to a refined microstructure [192].Gou et al [198] proved that the addition of Nb in Ti6Al4V increased the supercooling capacity, and the planar grain growth mode switched to the cellular grain growth, leading to a refined microstructure.Zhang et al [190] developed Ti-Cu alloys that have a high constitutional supercooling capacity.The as-printed Ti-Cu alloys have a fully equiaxed microstructure without any complex processing parameter control.They demonstrated a pathway to additively manufacture Ti-Cu alloys with fine equiaxed prior-β grains and an ultrafine eutectoid lamellar structure.Similarly, Narayana et al [199] and Simonelli et al [200] investigated the addition of Fe in Ti6Al4V alloy during DED and L-PBF, respectively.They found that the addition of 3%-4% Fe resulted in a significant reduction of prior β grains from ∼190 µm to ∼70 µm [199] and the further increase of Fe (>4%) content showed only a limited further benefit [200].Zhang et al [208] found that adding boron to Ti6Al4V resulted in a reduced grain size and refinement of α lath during AM.
Moreover, grain refinement weakens the texture, reducing the tensile property anisotropy.

Grain refinement
Epitaxial grain growth, prominently prevalent in L-PBF, is known for the capillary effect and poor strain accommodation capability.Therefore, a lot of efforts have been made to refine grains to prevent hot cracking and enhance the mechanical property isotropy.Adding solute atoms as nuclei to promote equiaxed grain formation has been widely explored in aluminum alloys and titanium alloys [14,[209][210][211].However, in nickel-based alloys and steels, the inoculation treatment is not a common practice due to the lack of knowledge of the inoculants [103].
Researchers have attempted to add elements to the standard alloy as inoculants for grain refinement.A successful example is the Scalmalloy, which is developed and patented by the Airbus Group [201].The introduction of Sc or Zr element to aluminum alloys promotes the formation of a supersaturated solid solution under rapid cooling during L-PBF.The alloy was then strengthened by direct aging to form Al 3 Sc or Al 3 Zr, which is an L1 2 structure and is strongly coherent with the Al matrix [202].The Al 3 X precipitates act as nucleating agents, leading to significant microstructure refinement for the aluminum alloys.The Scalmalloy has both good printability and excellent mechanical properties, i.e.YS = 520 MPa, ultimate tensile strength (UTS) = 530 MPa, and elongation = 14% [201].Due to the success of Scalmalloy with regard to both printability and mechanical properties, similar strategies were explored by other researchers.
Jia et al [203] explored the potential of replacing Sc with Er in the Al-Sc-Zr alloy.They found that the microstructure of the laser remelted Al-Sc-Zr region was mainly composed of fine columnar grains, while the remelted region of Al-Er-Zr alloy consisted of a large-grained structure (figure 14(a)).The Al-Sc-Zr alloy has higher hardness and thermal ability compared with the Al-Er-Zr alloy.Manca et al [204] printed the novel Al-3Ce-7Cu alloy via L-PBF with a low porosity of 0.03%.The mechanical performance is satisfactory with YS = 274 MPa and UTS = 459 MPa at room temperature.The high strength was attributed to the fine Al6.5CeCu6.5 and Al11Ce3 eutectic phase precipitated in the as-built Al-3Ce-7Cu alloy, demonstrating good promise as novel heatresistant alloys.They also investigated the addition of Cu in the Al-Si-Ni-Fe alloy [205].The yield compression stress is 355 MPa at 200 • C due to the refined microstructure as a result of Si, Al 5 Fe(Ni,Cu) and Al 3 (Ni,Cu) phase precipitation (figure 14(b)).Croteau et al [206] designed an Sc-free, agehardenable aluminum alloy Al-Mg-Zr specifically tailored for L-PBF.The developed Al-Mg-Zr alloy was successfully fabricated via L-PBF without cracks and had an excellent combination of high YS (∼354 MPa) and high tensile ductility (∼20%) after aging due to solid strengthening by Mg in the matrix and the Hall-Petch effect of both sub-micrometer and nano-sized Al 3 Zr precipitates.Li et al [123] studied the effect of Zr content on the printability, microstructure evolution, and mechanical properties of Al-Cu-Mg-Mn alloy.With the increase of Zr content from 0 to 4 wt%, the grain size decreased and (figure 14(c)) the YS and ultimate tensile stress increased.Nie et al [212] also studied the effect of Zr content on the Al-Cu-Mg-Mn alloy and found that with the increase of Zr content, the grain size decreased first and then tended to be stable (figure 14(c)) while the YS and ultimate tensile stress increased first and then decreased.
In summary, microstructure-oriented strategies are employed to develop alloys tailored for L-PBF in three main ways: (1) Phase transition-guided approach: this involves leveraging the IHT during L-PBF to induce specific phase transitions.By utilizing thermodynamic and kinetic software to calculate phase diagrams, the desired phase constitutions are achievable.
(2) Columnar or equiaxed grain structure design: the design of the microstructure focuses on either columnar or equiaxed grain structures, depending on the targeted application.
The CET theory is utilized to guide the design by optimizing scanning parameters and controlling the thermal gradient.
(3) Addition of nucleating agents: nucleating agents are added to the alloy to refine the microstructure.This strategy considers the interaction of the nucleating agents with the matrix and their influence on solid solubility and precipitation behavior of other elements in the alloy.

ML-assisted strategies
In sections 4.1 and 4.2, we discussed two primary strategies for alloy design in L-PBF.These strategies rely on a deep understanding of physical mechanisms such as cracking mechanism and phase transitions.In order to accurately design the composition of an alloy with specific properties, validated models or theories are required that can link the composition with defects, microstructure, and properties.Unfortunately, at present, either the cracking mechanism or the phase transition during the complex thermal cycles in L-PBF has not been completely understood.Consequently, a validated correlation between alloy composition and final properties has not been well-established, restricting the alloy development for L-PBF.
In recent years, ML, a method to create logistic models from large datasets following specific algorithms, has attracted a lot of attentions in diverse industries.Since ML does not solve complex physical equations based on mechanistic models, the computation is fast and is believed to be capable of making more accurate predictions than human beings [213].Through learning from the literature data, ML enables the prediction of material compositions and properties.
Significant advancements have been achieved in the development of ML-driven metals [88,[214][215][216][217][218][219][220][221].For instance, Zhang et al [222] successfully employed ML to develop four novel Cu-In alloys that possess both high UTS and Reprinted from [205], Copyright (2019), with permission from Elsevier.(c) EBSD maps of Al-4.24Cu-1.97Mg-0.56Mn alloy with different Zr contents, (i) 0 wt.%, (ii) 0.6 wt.%, (iii) 2 wt.%, (iv) 2.5 wt.% [212].Reprinted from [212], Copyright (2018), with permission from Elsevier.electrical conductivity (EC).These alloys have the potential to replace costly Cu-Ag alloys currently used in railway wiring.Furthermore, Zhang et al [223] combined feature screening strategy with Bayesian optimization to iteratively design alloys.This approach led to the development of a copper alloy, Cu-1.3Ni-1.4Co-0.56Si-0.03Mg,which exhibits exceptional mechanical and electrical properties.Wang et al [224] proposed an ML design system (MLDS) that enables performance-oriented compositional design for complex alloys.The MLDS incorporates ML modeling, composition design, and property prediction.Through this system, several high-performance copper alloys were designed, offering an UTS range of 600-950 MPa and EC of 50.0%International Annealed Copper Standard (IACS).To further enhance performance-oriented design, Jiang et al [225] extended the MLDS and successfully designed a new type of aluminum alloy with remarkable UTS (707-736 MPa) and elongation (7.8%-9.5%).Additionally, Xue et al [226] optimized the composition of Ti-Ni-based shape memory alloy using an active learning strategy and discovered a new alloy, Ti-Ni-Cu-Fe-Pd, with reduced transformation thermal hysteresis (1.84 K).Successful applications of ML in alloy design can also be found in the development of high entropy alloys [227], aluminum alloys [228], and Co-based superalloys [214] that undergo conventional casting or forging processes.
However, the utilization of ML in metal AM remains limited.Existing literature on ML in the AM field primarily concentrates on optimizing processing parameters [229][230][231], controlling part geometry [232][233][234], and detecting defects [234][235][236][237].The incorporation of ML for alloy design specifically for AM is notably scarce in current research.
Du et al [218] reduced the defects during L-PBF by combining physical-informed ML, mechanistic modeling, and experimental data (figure 15(a)).In order to predict the spheroidizing defect during L-PBF, parameters such as bulk energy density (E), surface tension (F), Marangoni number (M), Richardson number (R), the aspect ratio of the molten pool (ε), and solidification time (T) were taken into account.The spheroidizing sensitivity index was deduced by a genetic algorithm.Two key factors affecting spheroidization were thus found in this way, i.e.Marangoni number and the solidification time of the molten pool.They formulated an equation relating the spheroidizing sensitivity index with the aforementioned six parameters.With this equation, they successfully predicted the spheroidizing defects with an accuracy of 90%.Mondal et al [221] also applied ML in crack prediction, as shown in figure 15(a).Four parameters were used in this ML model, including cooling rate (T), the ratio of the temperature gradient to the solidification growth rate (ε), solidification stress (σ), and the ratio of the vulnerable and relaxation times (β) [221].The four variables were selected by Pearson correlation analysis.They are independent of each other and can represent the physical characteristics of the cracks [238,239].Ultimately, a linear correlation was proposed to calculate the crack sensitivity index.
Computational alloy design is being widely utilized in various applications.The integration of tools like Thermo-Calc can significantly enhance the efficiency of alloy development.
In cases where direct physical predictions are impractical, ML methods can be employed to forecast properties based on existing literature data.Multi-objective genetic algorithms are capable of achieving this objective, as shown in figure 15(b).The genetic algorithm is an optimization algorithm that emulates biological evolution processes to iteratively improve solutions to a given problem.The genetic algorithm evolves the chromosomes through genetic operations such as selection, crossover, and mutation.These operations simulate the natural selection process, where individuals with better fitness (i.e.solutions that perform well on the given problem) have a higher probability of being selected for reproduction (figure 16(b)-genetic algorithm illustration).Dreano et al [217] employed qualitative constraints in their study, aiming to control the fraction of face-centered cubic (FCC) phase within the range of 0.7-0.94 and ensure that the solidification process starts with the formation of FCC phase.They utilized a genetic algorithm that aimed to minimize the solidification temperature range and maximize the strength index.Alloys meeting all the constraints were evaluated based on their fitness scores.In this way, a crack-free aluminum alloy during L-PBF was designed (figure 15(b)-Al-alloy composition design).Similar methods have also been applied in the design of crack-free austenitic stainless steels for L-PBF (figure 15(b)-austenitic stainless steel composition design).Sabzi et al [88] defined constraints and goals for the genetic algorithm, including the solidification temperature range, performance index, and the ratio of Cr equivalent to Ni equivalent.They optimized the composition of a 316L alloy and validated its printability.The results demonstrated a significant improvement in the formability of crack-free L-PBF 316L alloy compared to 316L alloys with pores and deformation.
To summarize, the use of ML in alloy design for AM is still in its early stages.This is attributed to the lack of an accurate and statistically representative database, which is essential for the prediction accuracy of ML.Furthermore, many mechanistic variables of alloys might not be readily available and thus should be calculated based on known data.For instance, temperature gradient, cooling rate, and stresses, which influence the solidification microstructure and cracking, should be calculated by combining calculation of phase diagram (CALPHAD)-based computational thermodynamic and finite-element-based thermal models.These mechanistic variables are important inputs in the ML model construction.

Future outlook and conclusions
Since the emergence of AM in the 1980s, a considerable amount of work on metal AM has been published.Despite significant progress in AM, some scientific and technological challenges still need to be addressed.For instance, cracking, anisotropy in building direction, low ductility, and low fatigue resistance constantly occur in additively manufactured alloys, particularly in the LPBF-processed alloys.The previous work has proposed various remedies to tackle these issues during alloy design for L-PBF, as shown in figure 16.However, printable metals with satisfactory mechanical properties represent only a small fraction of commercially available alloys.The following aspects should be emphasized in future research.
(1) The defect formation mechanism in L-PBF requires clarification.Although our understanding of the L-PBF process has significantly advanced (e.g.RGD theory, Kou's criterion), the root cause of cracking during L-PBF has not been completely understood.Due to the complex metallurgical reactions (laser-powder-gas-bulk interactions) during L-PBF, a practical in-situ crack observation technology that can visualize the crack initiation and propagation during L-PBF is in urgent need.(2) Solidification theories tailored for the L-PBF process need to be developed.The layer-by-layer powder deposition in L-PBF facilitates epitaxial grain growth, which is detrimental to the property isotropy in the final component.Considering the L-PBF process features, a solidification theory accounting for fast cooling, steep temperature gradients, and huge internal stress is required.Additionally, tailored thermodynamic packages that consider the phase transitions during thermal cycles in L-PBF are necessary.(3) Further exploration of grain boundary segregation engineering is needed during alloy design for L-PBF.Grain boundary manipulation is crucial for the development of crack-resistant alloys, as cracks primarily initiate from grain boundaries.It is important to note that the subgrain boundary (cell/dendrite) segregation also plays a significant role in crack formation.The enrichment of solute atoms can lead to precipitations, which reduce grain boundary energy and hinder boundary mobility through pinning mechanisms.This pinning effect promotes the formation of equiaxed grains but also introduces stress concentration.Care must be taken when assessing the influence of precipitates on crack sensitivity due to their embrittling nature.(4) Further research is needed to apply ML in alloy design for AM, accompanied by the high-quality database.The application of ML in alloy design for AM is still in its nascent stages.This is primarily due to the limited volume of data available for ML in the AM field.Additionally, data consistency in AM is unsatisfactory.For instance, the crack density of a sample can vary significantly when printed with different machines or using different batches of powders.Furthermore, more efforts should be made toward developing physics-informed ML models that aid in understanding the cracking and strengthening mechanisms of LPBF-processed alloys.(5) High-throughput L-PBF equipment or methods are in urgent need, taking into account the time-consuming and costly powder making and conventional L-PBF processes.
In-situ alloying presents a high-throughput methodology to design novel alloys for AM.By combining multiple elemental powders during printing, engineers can design and test various alloy compositions to achieve desired material properties, reducing the need for pre-alloyed powders that may have limited availability or high costs.By blending elemental powders during printing, the desired alloy composition can be achieved precisely where and when it is needed, providing greater flexibility in material selection.This flexibility in composition enables the rapid exploration of new alloy systems.For instance, blends of titanium and Cr powders have been manufactured using DED, and a desirable combination of strength and ductility was obtained [240].Grigoriev et al mixed Ti, Al, and Nb elemental powders in a turbula mixer and processed the mixture with L-PBF to produce Ti 2 AlNb parts [241].However, incompletely melted and unevenly distributed Nb particles were observed in the printed specimen due to the high melting point of Nb (2469 • C).Balancing the melting of elements with high melting points while avoiding the evaporation of volatile elements during parameter optimization remains a challenge in in-situ powder mixing for L-PBF.
This review assessed the printability of four commonly used alloy systems (Fe-, Ni-, Al-, and Ti-based alloys) mainly in L-PBF AM and compared with their weldability.The application of weldability criteria in assessing printability for each alloy was also evaluated.Various pathways for AM alloy design were reviewed and analyzed.The following conclusions were drawn: (1) The presence of martensite in the as-printed steel microstructure increases crack susceptibility.The Schaeffler diagram can be used to estimate the printability of steels based on their chemistry.(2) The weldability of nickel-based superalloys does not always align with their printability.The weldability assessment chart, which considers the effects of Ti, Al, Cr, and Co, cannot directly categorize the printability of nickel-based superalloys during L-PBF.(3) A Pearson correlation map was constructed to identify key factors contributing to cracks in aluminum alloys processed via L-PBF.The map revealed that increasing silicon content and decreasing magnesium content can enhance crack resistance in aluminum alloys during L-PBF.(4) Titanium alloys generally exhibit less severe cracking during L-PBF.Efforts have been focused on promoting the CET during the AM of titanium alloys.(5) Alloy design strategies for L-PBF can be categorized into three types: (1) crack mitigation-oriented strategies that aim to mitigate cracks through in-depth analysis of the crack formation mechanism, (2) microstructure refiningoriented strategies that focus on synergistically improving crack resistance and mechanical properties by adding alloying elements or nucleation particles, and (3) MLassisted strategies based on AM databases and data analysis.(6) To accelerate alloy design for L-PBF, there is a need to enhance fundamental understanding of defect formation, building direction anisotropy, low ductility, and low fatigue resistance in LPBF-processed alloys.Thermo-Calc calculations can be utilized to address complex phase transitions during AM by considering features such as extremely fast cooling and cyclic heat treatment.Grain boundary should be carefully manipulated considering the strengthening and embrittling effects of precipitates.Insitu alloying provides a high-throughput method for alloy design, but challenges related to composition homogeneity and melting point variation among elements need to be resolved.
In summary, the complex metallurgical reactions and phase transitions during L-PBF pose challenges to the quality of printed parts.However, they also open avenues for revolutionary alloy design, offering the potential for unprecedented metals with tailored properties in the near future.By addressing the L-PBF processability challenges and exploring innovative alloy design strategies, this review contributes to a better understanding of alloy design for L-PBF AM and guides future research.

Figure 2 .
Figure 2. Effects of complex physical processes and thermal history of L-PBF on microstructure, phase and stress distribution.(a) Schematic diagram of cracking in Al-Cu-Mg alloy and the microstructure of cracked and uncracked regions [56].Reprinted from [56], Copyright (2021), with permission from Elsevier.(b) L-PBF processed over-eutectic Al-50Si alloy.Complex thermal physical processes generate non-equilibrium phase distribution[20].Reprinted from[20], Copyright (2021), with permission from Elsevier.(c) Longitudinal sections of mass flow velocity and temperature distribution of depositing single track at different times[48].Reprinted from[48], Copyright (2020), with permission from Elsevier.(d) Anisotropic stress field in each layer under the influence of multiple anisotropic beam paths[55].Reprinted from[55], Copyright (2019), with permission from Elsevier.

Figure 6 .
Figure 6.Correlation between printability and alloying elements in nickel-based superalloys.

Figure 7 .
Figure 7. Effect of compositions on weldability of nickel-based superalloys.Modified weldability assessment chart after considering the effects of Cr and Co [104].Reproduced from [104].CC BY 4.0.

Figure 8 .
Figure 8. Pearson correlation between alloying elements and three features for aluminum alloys.ST represents solidus temperature, ∆T denotes brittleness temperature range when the solid fraction is between 0.8 and 0.99, and SCI max indicates the maximum value of SCI in ∆T.
The solidus temperature, maximum SCI, brittleness temperature range, and printability of commonly seen aluminum alloys.Composition(wt.%)
The composition, types, MOE and printability of Ti-based alloys.

Figure 10 .
Figure 10.Grain boundary features during laser powder bed fusion additive manufacturing.

Figure 11 .
Figure 11.Suppressing cracks during L-PBF using the concept of grain boundary segregation engineering.(a) Schematic of the inhibition effect of carbon on the cracking of L-PBF Inconel 738.The addition of carbon reduces the enrichment of element B at grain boundaries and decreases the formation of low-melting-point phases [107].Reprinted from [107], Copyright (2022), with permission from Elsevier.(b)Schematic of the microstructural and compositional evolution during AM of the superalloys with hot cracks.The progressive enrichment in Cr, Mo and B at grain boundaries over the building process causes grain boundary segregation induced liquation[18].Reprinted from[18], Copyright (2019), with permission from Elsevier.(c) Schematic of solute segregation effect on interfacial energy and solidification behavior of metastable Fe-Mn-Co-Cr-Si high entropy alloy processed via L-PBF[189].Reprinted from[189], Copyright (2021), with permission from Elsevier.(d) Schematics of residual stress minimization for Al0.5CoCrFeNi alloy during L-PBF.The image indicates the Al segregation toward the interdendritic/grain boundaries, discontinuous BCC/B2 grains formed along the grain boundaries, the tensile residual stress accommodated by the BCC/B2 grains, and the switch from a large tensile residual strain a small compressive residual strain[185].Reprinted from[185], Copyright (2021), with permission from Elsevier.

Figure 12 .
Figure 12.Microstructure of UM_17-4 under different cooling rates.(a) EBSD of as-cast UW_17-4 fabricated by arc-melting.The left panel is an inversed pole figure (IPF).The right panel is the corresponding image quality (IQ) map.(b) EBSD of UW_17-4 after a single-layer laser melting (transverse cross-section).The substrate is a cast, fully martensitic UW_17-4 after a solution heat treatment.(c) EBSD of UW_17-4 after laser spot welding under 156 W laser power with 1 ms laser duration (transverse cross-section).(d) EBSD of a zoom-in area from (b).The microstructures for all conditions are fully martensitic.All IPFs share the same color code, which is shown in the inset of (a)[194].Reprinted from[194], Copyright (2022), with permission from Elsevier.

Figure 13 .
Figure 13.The effect of temperature gradient G and growth rate R on grain size and morphology.S and L stand for solid and liquid, respectively.[65] John Wiley & Sons.Copyright © 2003 John Wiley & Sons, Inc.

Figure 15 .
Figure 15.The application of machine learning in printing crack-free alloys in additive manufacturing.(a) Schematic diagram of machine learning model to avoid balling and cracks[218,221].Reprinted from[218], Copyright (2021), with permission from Elsevier.Reprinted from[221], Copyright (2022), with permission from Elsevier.(b) Schematic diagram of multi-objective genetic algorithm combined with thermodynamic calculation for designing aluminum alloys and austenitic stainless steels[88,217].Reprinted from[88], Copyright (2020), with permission from Elsevier.Reprinted from[217], Copyright (2022), with permission from Elsevier.

Figure 16 .
Figure 16.An overview of various alloy design strategies for laser powder bed fusion.

Table 1 .
The printability of Fe-based alloys during additive manufacturing.