Additive manufacturing of magnesium and its alloys: process-formability-microstructure-performance relationship and underlying mechanism

Magnesium and its alloys, as a promising class of materials, is popular in lightweight application and biomedical implants due to their low density and good biocompatibility. Additive manufacturing (AM) of Mg and its alloys is of growing interest in academia and industry. The domain-by-domain localized forming characteristics of AM leads to unique microstructures and performances of AM-process Mg and its alloys, which are different from those of traditionally manufactured counterparts. However, the intrinsic mechanisms still remain unclear and need to be in-depth explored. Therefore, this work aims to discuss and analyze the possible underlying mechanisms regarding defect appearance and elimination, microstructure formation and evolution, and performance improvement, based on presenting a comprehensive and systematic review on the relationship between process parameters, forming quality, microstructure characteristics and resultant performances. Lastly, some key perspectives requiring focus for further progression are highlighted to promote development of AM-processed Mg and its alloys and accelerate their industrialization.

These authors contributed equally to this work. * Authors to whom any correspondence should be addressed.
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. Statistically stored dislocation density ρg Geometrically necessary dislocation density ∆α Thermal expansion coefficient difference between particle and matrix ∆T d Temperature difference between test and processing dp Equivalent diameter of particle KAMave Average KAM value of the selected region µ Step size of EBSD test kws Constant, kws = 16.1 for FCC materials with b along [109] and kws = 14.4 for BCC materials with b along [110] F I Interaction factor ε Lattice strain in material Ks Shape factor, which is normally set as 0.9 λ Wavelength of x-ray source D Crystallites size θ Peak position in radians 1. Introduction

Magnesium and its alloys
Mg is the lightest structural metal in the world with a density of ∼1.738 g·cm −3 , approximately achieving 25% the density of steel, 38% that of titanium, and 65% of aluminum, which leads to a high strength-to-weight ratio and stiffnessto-weight ratio. Thus, Mg and its alloys become promising materials for structural parts in the field of aerospace and automotive [1,2]. In addition, Mg and its alloys have excellent damping performance [3], superior biocompatibility, large hydrogen capacity, and high theoretical specific capacity for battery, promoting their electrochemical and biomedical applications [4,5]. The advancement and application of Mg and its alloys have attracted increasing attention. Many new standards for Mg and Mg alloys are gradually registered and developed in recent years, such as ISO/CD 4155 in 2020 [6] and ISO/DIS 4177 in 2021 [1]. In addition, some mega enterprises are increasing investment to promote the industrial application of Mg alloys. Baosteel Metal Co., Ltd plans to invest 18.2 billion Yuan and co-establishes a 'Joint Research and Development Center for Advanced Magnesium Technology' with Chongqing University to promote the industrialization of Mg and its alloys and accelerate magnesium industry development in China. Western Magnesium Corporation proposed an innovative project that over a one-billion-dollar investment to build a plant for primary Mg production, and the production plant was chosen to be in the Ohio State in 2021. Moreover, IMS issued International Magnesium Science & Technology Award [7], and IMA issued IMA Awards of Excellence [8], to encourage more scientists and engineers to engage in the R&D and application of Mg and its alloys. utilized in preparation and repair of both metallic and nonmetallic materials [9][10][11]. From a viewpoint of engineering, AM technology has a very high freedom degree of manufacturing, thus allowing fabrication of parts with extremely complex structure [12]. In addition, AM has high material utilization. For instance, the material utilization of LPBF can reach 95%, and that of WAAM can achieve nearly 100%, which are both much higher than that of conventional subtractive manufacturing process of only 5% [13]. Furthermore, AM also helps to reduce R&D cycle and accelerate product iteration. From a viewpoint of science, the process parameters and forming materials can be in-situ tailored during AM process, combined with the unique thermal history, bringing new opportunities and challenges in material science. Zhang et al reported an advanced titanium alloy with concentration modulations, which was realized by small amounts of 316 l stainless steel addition via LPBF process [14]. This kind of titanium alloy could retain high strength while substantially improving ductility. Dependent on the above advantages, Gu et al believed that AM has the potential to revolutionize how components are designed. They proposed a holistic concept of material-structure-performance integrate AM, which is defined as a one-step AM production of an integral metallic component by integrating multi-material layout and innovative structures, with an aim to proactively achieve the designed high performance and multifunctionality [15].
AM of Mg and its alloys is of increasing interest for researchers. Current AM technologies applied to fabrication of Mg and its alloys include WAAM [16], LPBF [17], LDED [18], FSAM [19,20] and binder jetting [21]. Amongst them, most studies are focused on WAAM and LPBF, which may be ascribed to the two main application fields of Mg and its alloys, i.e. as structural material and biomedical material. WAAM has advantages in the rapid manufacturing of large-scale parts, thereby the related literature on WAAM-processed Mg and its alloys mainly aims at mechanical properties. LPBF is conducive to preparing fine and complex structures (such as lattice structure), so the relevant researches are mainly focused on corrosion resistance and biocompatibility.

The purpose of this review
Nowadays, there have been several review papers summarizing and discussing the-state-of-the-art of AM of Mg and its alloys [22][23][24][25][26][27]. Most recently, Zeng et al explored the relationship between alloying elements, microstructure and properties of Mg alloys fabricated by different AM processes [28]. However, the underlying mechanisms of such kind of relationship has not yet been systematically explored and analyzed. Therefore, the motivation of this work is to discuss and reveal the possible underlying mechanisms based on a rigorous and dedicated review on the process-formability-microstructureperformance relationship of AM-processed Mg and its alloys, with the aim to highlight future scientific research directions.

Element vaporization
Mg element, as well as some alloying elements in Mg alloys, such as Al and Zn, has a low boiling point and high vapor pressure, which make it easily to vaporize under laser irradiation during AM process. Compared to Ti and Al, Zn and Mg have relatively low boiling point (Mg: 1091 • C; Zn: 917 • C; Ti: 3287 • C; Al: 2500 • C), fusion latent heat (Mg: 8.7 kJ·mol −1 ; Zn: 7 kJ·mol −1 ; Ti: 65 kJ·mol −1 ; Al: 10 kJ·mol −1 ) and vaporization latent heat (Mg: 134 kJ·mol −1 ; Zn: 115 kJ·mol −1 ; Ti: 411 kJ·mol −1 ; Al: 291 kJ·mol −1 ) [29]. Therefore, Mg and Zn normally exhibit relatively high saturated vapor pressure and vaporization flux at various temperatures, which means Mg and Zn have strong vaporization tendency amongst the four elements [29]. That is the reason why the generally concerning elements for vaporization in AM process of Al alloys are Mg and Zn [30,31].
The Langmuir model shown below is commonly utilized to evaluate the vaporization rate J i (g · cm −2 · s −1 ) of element i in materials: where γ i is the activity coefficient of element i, X i is the molar fraction of element i, P 0 i is the saturated vapor pressure of pure metal i, M i is the molecular weight of element i, and T is the absolute temperature (K). According to the Langmuir model, Wei et al indicated that J Mg is 4.2 × 10 4 -3.5 × 10 10 times, 54-160 times, and 2.3 × 10 5 -1.2 × 10 9 times of J Al , J Zn , and J Mn , respectively for Mg-8.97Al-0.25Zn-0.14Mn alloy in the temperature range of 870 K to 2000 K [32]; Deng et al found that J Mg is 8.2 × 10 6 -2.0 × 10 14 times, 7.9-20.8 times, and 5.4 × 10 12 -1.6 × 10 27 times of J Gd , J Zn , and J Zr , respectively for Mg-11Gd-1.77Zn-0.43Zr alloy in the temperature range of 923 K to 2000 K [33]. It is evident that the vaporization rate of Mg element is the highest during AM process. However, given that the weight percentage of Mg element is extremely higher than that of Zn element in Mg alloys, the vaporization of the latter should be more intense than that of the former, resulting in increased Mg/Zn ratio in AM-processed sample [33,34]. Nevertheless, the main weight loss during AM of Mg is still caused by Mg element vaporization [35,36]. Attarzadeh and Asadi established a numerical model to investigate the weight loss of Mg element (Mg loss ) during AM process of WE43 alloy, and illustrated that the increase in the laser power and layer thickness, as well as the decrease in the scanning velocity, would induce an increment in Mg loss [36].
It should be noted that the current investigations normally utilize relative mass fraction variation to characterize the element loss during AM process of metallic alloys. Zhang et al indicated that Al was the most easily vaporization element in Ti-6Al-4V alloy, and the mass loss ranged from 0.15 wt.% to 0.3 wt.% [37]. Although such index is not accurate enough, it still can reflect Mg loss to some extent in several certain types of Mg alloys. For instance, Wei et al found that the Mg loss (relative mass fraction variation) during the LPBF process of AZ91D Mg alloy was higher than 3 wt.% [32]. This is mainly due to that Mg is more easily to vaporize than Al as mentioned above, resulting in more mass loss of Mg during AM process, even considering the vaporization of Al. Thus, the Mg loss (relative mass fraction variation) was significantly reduced. However, Zn has comparable or superior vaporization tendency to Mg. Consequently, the Mg loss (relative mass fraction variation) in Zn-containing Mg alloys is not obvious, which is consistent with the observations by Deng et al (less than 1 wt.%) [33] and Wei et al (Relative mass fraction of Mg element was increased) [34]. In this case, the relative mass fraction is far more inappropriate to be used. In addition, process parameter tuning is helpful to control the element vaporization in the AM process of Mg alloys [35,36]. Attarzadeh and Asadi optimized the process parameters to achieve only 0.23 wt.% Mg loss in the LPBF process of WE43 alloy [36].
In the LPBF process, the vaporization of elements can lead to violent powder splash, which will create depressions in the laser scanning areas, thus affecting the subsequent deposition process. Specifically speaking, gas bubbles will be involved in melt pool during the process of depression recovery. If these gas bubbles cannot escape from melt pool in time (for instance, the liquid flow inside melt pool drags the gas bubbles to the bottom), then porosity in final microstructure will be increased accordingly. In addition, the splashed partially melted powders may land on the previously deposited regions, resulting in balling phenomenon. To ensure that there will be enough powders for subsequent deposition after current deposition process, Deng et al set the actual powder bed elevated height at 2.5-3 times the layer thickness [38,39].

Defects and their formation mechanism
2.2.1. Crack. Hot crack, including solidification crack and liquation crack, as well as cold crack, has been observed and investigated in additively manufactured Mg alloys [38,[40][41][42][43]. Deng et al reported that cracks were observed even on the surface of the as-built samples under inappropriate process parameters as shown in figure 1 [38]. The occurrence of hot cracking is closely related to the formation of eutectic phase or eutectic liquid film. On the one hand, Mg alloys normally have a wide solidification temperature range, resulting in a large mush zone. During the final stage of solidification process, the segregation of alloying elements leads to the formation of eutectic liquid film, reducing the strength of the mush zone. Subsequently, the eutectic liquid film ruptures under the action of solidification shrinkage stress. Thus, solidification cracks form accordingly. Wei et al indicated that the solidification cracking tendency became serious firstly and then alleviated with the increase of Zn content during the LPBF process of Mg-xZn binary alloys [44]. They interpreted this to the competitive effect of the rupture of the eutectic liquid film and the backfilling of the residual liquid. In the case of low Zn content (<2 wt.%), the eutectic liquid was not sufficient to reduce the strength of the mush zone below the solidification shrinkage stress, thereby suppressing the formation of solidification  [38]. Reprinted from [38], Copyright (2022), with permission from Elsevier.
cracks. With the increase in Zn content, the volume fraction of the eutectic liquid was increased, leading to sufficient reduction in the strength of the mush zone, which in turns induced solidification cracking. However, as the Zn content was further increased, the morphology of the eutectic liquid film was reticular, which provided a good approach to backfill the ruptured region with residual liquid, thus reducing solidification cracking tendency. Liu et al observed the healing phenomenon of solidification cracks at the top of the LPBF-processed ZK60 sample as shown in figure 2 [42], which presents a good experimental evidence for the interpretation by Wei et al.
On the other hand, the cyclic heating process of AM demonstrates a possibility to cause the temperature of HAZ to exceed the melting points of the eutectic phases without melting the matrix. If so, the melted eutectic phases in the HAZ rupture under the action of thermal stress caused by the rapid heating and cooling of AM process, inducing the formation of liquation cracks. It is evident that the liquation cracking tendency is influence by the energy input during AM process. Liang et al found that a high energy input (486.1 J·mm −3 ) during the LPBF process would induce the formation of liquation cracks, while a low energy input (291.6 J·mm −3 ) was conducive to generating near-full dense samples [45]. Zhang et al observed liquation cracking during the WAAM of AZ31 alloy, which was also attributed to the high energy input caused by the utilized low welding speed [40].
It is noteworthy that the solidification cracks and liquation cracks can merge together to appear as long and continuous cracks [45], which are unique cracking characteristics during AM process and different from that during welding process. In addition, it is also possible that these two kinds of hot cracks remain unmerged, which is as similar as cracking phenomenon observed during welding process [42]. Grain refinement and precipitation of grain-boundary strengthening phase are beneficial for the elimination of hot cracks [46].
Deng et al found cold cracks in the middle part of the LPBF-processed Mg-10Gd-3Y-1Zn-0.4Zr alloy, while no cracks formed in the top and bottom regions [38]. Similar to liquation cracks, the formation of cold cracks is also related to thermal stress. During manufacturing process, the thermal stress is increased gradually, until it reaches to a certain degree to cause the rupture of solidified microstructure. Then cold cracks form, and the thermal stress is alleviated. The subsequent deposition process will not induce cold cracking, unless the thermal stress is increased high enough again. That is the reason why cold cracks are normally found in the middle part, not in the top and bottom part of manufactured sample. Preheating and making materials more ductile are both effective methods to suppress cold cracking formation [38].

Lack of fusion (LoF).
It should be noted that some researchers refer to LoF defect as gas pore as well [47]. However, the morphology of LoF defect is different from that of gas pore [36,48]. In this work, the pores with irregular morphology are recognized as LoF defects, while spherical pores as gas pores. Åhman et al pointed out that some irregular pores were formed due to LoF defects during the LPBF process of WE43 alloy [48]. It is generally accepted that the formation of LoF defects is ascribed to low energy input that is insufficient to completely melt alloy powders during AM process [33,39,42,43,[48][49][50]. Wu et al believed that high scanning velocity (corresponding to low energy input) led to capillary instability in melt pool during the LPBF process, and then resulted in balling phenomenon. The poor morphology of the currently deposited track would also affect the quality of the subsequently deposited tracks by suppressing sufficient liquid metal spreading. Consequently, insufficient overlapping occurred between adjacent tracks or adjacent layers, and LoF defects formed [47]. Similarly, Deng et al found that high hatch spacing would cause thicker powder layer in the overlap region than other non-overlap regions, also inducing a large amount of LoF defects since the laser energy input could not fully melt all the powders in the overlap region (insufficient overlapping) [33]. Esmaily et al observed that the LoF defects with irregular morphologies, not spherical gas pores or oxides, were preferentially located around unmelted powders during the LPBF process of WE43 alloy as shown in figure 3 [51], which proofed the interpretation of LoF defect formation. Liu et al indicated that high energy input could fully erase the LoF defects in the LPBF-processed ZK60 alloy [42].

Gas pore.
Gas pores are commonly observed in both WAAM-and LPBF-processed Mg alloys [52][53][54]. Considering the differences in process characteristics, the relative densities of Mg alloys prepared by these two manufacturing processes are also different. In general, the WAAM-processed Mg alloys are more easily to achieve a high relatively density of over 99.9% than the LPBF-processed ones as shown in table 1. For instance, the minimum porosity of the GTAW WAAM-processed AZ91 can be reduced to 0.026% through parameter optimization [53], while that of the LPBF-processed AZ91D is still as high as 0.48% [55]. As mentioned, near-fully dense microstructure can be easily realized in the WAAM-processed Mg alloys by optimizing process parameters, thereby few investigations are conducted to  [51]. Reprinted from [51], Copyright (2020), with permission from Elsevier. study the formation mechanism of pores and the corresponding interpretation is superficial.
The formation of gas pores in the LPBF process is generally ascribed to four mechanisms. (i) Gas pores originate from raw powder materials [48]. GA powders are the most widely used raw materials for LPBF process. Since the preparation technology of GA Mg alloy powders is still imperfect, the appearance of hollow powders is inevitable, which provides a good gas source for pore formation. Such kind of gas pore formation mechanism is also common in laser additively manufactured steel, nickel-based superalloy and titanium alloy [13]. (ii) The gaps between powder particles on powder bed is another source for pore formation. This kind of gas pores is almost inevitable due to that the rapid cooling rate (nearly 10 6 k·s −1 ) of LPBF process is conducive to gas capture. (iii) For alloys containing low-boiling-point elements, like Al, Mg and Zn, such elements easily evaporate during LPBF process, thus promoting pore formation. (iv) Keyhole formation and collapse have a great impact on the formation of gas pores [47]. Martin et al believed that the gas formation during LPBF process was attributed to the appearance and subsequent collapse of a deep keyhole depression at laser turn points, which trapped gas inside melt pool, and pores formed accordingly [56]. They proposed a mitigation strategy for the LPBF process to avoid the formation of pores. Recently, Huang et al conducted a more detailed investigation on keyhole collapse inducing pore formation, and indicated that pores formed due to different mechanisms during the transition process of a keyhole from stable state to unstable state [57]. In the transition regime (figures 4(a) and (b)), the hump, formed at the FKW due to the dependence of laser absorption on angle of incidence, tends to reflect laser beam and vapor flow towards the RKW. Intensive evaporation and recoil pressure are formed on the RKW, and a stagnation pressure is established, which in turn makes the RKW deform and expand. However, if the reflected laser beam and vapor flow are blocked or redirected due to keyhole perturbations, the temperature of the RKW will be reduced due to no illumination, which leads to the enhancement in surface tension, overcoming the recoil pressure. Then, keyhole rear wall collapses, and bubbles form. The bubbles will be pushed to the back of melt pool by Marangoni flow, and captured to form pores. In the unstable regime (figures 4(a)-(c)), the hump tends to direct laser beam and vapor flow towards the bottom of keyhole, and a vapor cavity will be formed. Because the energy is concentrated in this small cavity, the keyhole is prone to capillary instability and sometimes collapses, pinching off the cavity to form bubbles full with vapor. Although a few bubbles are re-captured by the expanding keyhole, most of the bubbles are almost immediately captured by the solidified melt pool to form pores. It should be noted that the gas formation mechanisms mentioned above do not work in isolation. For instance, element evaporation was also observed during keyhole dominated gas formation mechanism.

Balling
Balling phenomenon is rarely observed in WAAM-processed materials, while it is commonly reported in LPBF-processed materials. Most of the investigations show that balling phenomenon occurs at a relatively low energy density caused by high scanning velocity [55,62] or low laser power [47,68] or large hatch spacing [61] or their combination  [55]. Balling occurrence under low energy density is closely related to interfacial tension. As shown in figure 6(a), the melt pool can be divided into two parts, i.e. an upper deposition part consisting of alloy powders and a lower remelting part on substrate/previously solidified layers. Zhang et al indicated that the gas-liquid interfacial tension of the upper deposition part was beneficial for inducing balling occurrence, while the lower remelting part tended to inhibit the balling process of the upper part [69]. In case of low energy density, the remelting area was very small, and the liquid melt demonstrated a poor wettability with the solidified layers or the substrate. Under the action of interfacial tension, the liquid melt tended to become a sphere, resulting in balling occurrence [45]. In addition, it is also reported that the Plateau-Rayleigh capillary instability is easily occur at low energy densities, breaking up LPBF-processed single track, and thus leading to balling [62,70]. The balling phenomenon of deposited layer induced by interfacial tension is also observed in WAAM-processed Mg alloys [71]. Nevertheless, the WAAM-processed single track is much harder to break up into smaller droplets than the LPBF-processed one, since the former normally has a large diameter of several millimeters that is nearly 20 times the diameter of the latter, making it hard to rupture according to the optimized criterion for the Plateau-Rayleigh capillary instability [62].
However, high energy density can also induce balling occurrence. Yang et al indicated that too high an energy density (222.2 J·mm −3 ) during the LPBF process of ZK60 would cause the balling phenomenon [72]. Liang et al also showed that the balling occurred at a high energy density of 486.1 J·mm −3 in the LPBF-processed ZK60 alloy [45]. It is generally accepted that the balling under high energy density is attributed to droplet/powder splashing caused by evaporation of alloying elements [45,69,70]. As shown in figure 6(b), excessive energy density leads to the rapid formation of metal vapor, creating a high recoil pressure in melt pool, which will cause some liquid melt to escape from melt pool. Subsequently, the ejected liquid melt breaks up into microdroplets, which will become spherical under the action of gasliquid interfacial tension. These spherical micro-droplets fall back to the surface of material, causing balling particles to appear. Moreover, the unmelted powders around melt pool may be splashed sideways under the effect of metal vapor and stick to the surface of material, also resulting in balling phenomenon [45,69].

Process parameters-forming quality relationship
The internal and external qualities of additively manufactured Mg and its alloys are dependent on the utilized energy density. As mentioned, too high or too low energy density both will cause quality deterioration in AM process of Mg alloys [73,74]. Many investigations have established quantitative maps of process parameters and forming quality, such as printing speed-current process map for WAAM-processed AZ91D [75], laser power-scanning speed process map for LPBF-processed pure Mg [76], and scanning speed-hatch spacing process map for LPBF-processed GZ112K Mg alloy [33].
During the LPBF process of Mg alloys, when the energy density is extremely low (under the conditions of low laser power, high scanning speed, and high hatch spacing), most powder particles cannot be melted and maintain their initial morphologies, thus there is only limited consolidation between them. High porosity, caused by LoF accompanied with a small amount of gas pores, is generated in this case (Region I in figure 7) [33,[49][50][51]. Region I is called the under-heating region [47], and strong balling phenomenon can be observed in this region [61]. As the energy density is increased, a melt pool with a small circumference-to-length ratio can be created, which is in an unstable state and tends to break up. Low porosity, insufficient dimensional accuracy caused by rough surface, and accordingly poor mechanical properties are normally obtained in this region (Region II in   [47,55,61,76,77]. If the energy density is further increased to be located within Region III, dense samples with optimal dimensional accuracy (flat surface) can be generated. Region III is also known as forming zone, which is generally observed in the middle of process map [47,51]. Region IV is an over-energy density zone, in which the evaporation of Mg and Zn (if Zn is an alloying element) occurs, making melt pool splash under the combined action of recoil pressure and Marangoni convection. Consequently, key-hole induced pores normally appear, and a poor dimensional accuracy is generated [47,51,76].
For WAAM-processed Mg alloys, it is accepted that current should be linearly related with welding speed for a good forming quality as shown in figure 8 [75]. When manufacturing is conducted in the region that deviates from the linear relationship, droplet transfer mechanisms will be changed, resulting in the occurrence of 'undercooked' line, insufficient deposition, excessive deposition and 'Burnt' line, thus deteriorating the forming quality. It is evident that the process map of LPBF-processed Mg alloys is different from that of WAAM-processed Mg alloys, which is attributed to different process characteristics. However, the current research on WAAM process of Mg alloys is insufficient, which limits our understanding on the intrinsic mechanism of the influence of process parameters on forming quality.
The establishment of the above process maps only consider the variation of porosity and dimensional accuracy with process parameters, and does not consider cracking, because cracking is determined by both process parameters and material compositions. Therefore, the relationship between process parameters and cracking cannot be plotted independently of material compositions. Nonetheless, when one of the variables Figure 7. Laser power-scanning speed process map for LPBF-processed WE43 alloy [51]. Reprinted from [51], Copyright (2020), with permission from Elsevier. is fixed, the relationship of cracking with the other variable can be plotted. For instance, Wei et al plotted a schematic diagram illustrating the effect of Zn content on solidification cracking behavior of LPBF-processed Mg-Zn binary alloys as shown in figure 9(a) [44]. They indicated that the solidification cracking became more serious firstly and then alleviated with the increase of Zn content due to the competitive effects of shrinkage stress and backfilling. In addition, Liu et al established a process map to demonstrate the liquidation cracking behavior variation of the LPBF-processed ZK60 alloy with process parameters as shown in figure 9(b) [42]. They found that the reduction in energy density was beneficial for liquidation cracking elimination.

Grain features and formation mechanism
3.1.1. Grain morphology, texture and size.
Unlike additively manufactured titanium alloys and nickel-based superalloys that mainly exhibiting epitaxially growing columnar grains [13], both equiaxed grain and columnar grain have been observed in additively manufactured Mg and its alloys [48,51,53,54,58,[78][79][80]. Li et al found a fully equiaxed-grain structure in the WAAM-processed AZ31 alloy as shown in figures 10(a) and (b) [81]. However, Wang et al indicated that columnar grains normally formed at the bottom of a melt pool while equiaxed grains formed in the middle region during the WAAM process of AZ31 alloy as shown in figures 10(c) Figure 9. Process map for Mg-Zn alloys: (a) Zn content-solidification cracking process map for LPBF-processed Mg-Zn binary alloys [44]. Reprinted from [44], Copyright (2019), with permission from Elsevier. (b) Parameters-liquidation cracking process map for LPBF-processed ZK60 alloy [42]. Reprinted from [42], Copyright (2021), with permission from Elsevier.  [81]. Reprinted from [81], Copyright (2021), with permission from Elsevier. (c) YOZ plane and (d) XOZ plane in the LPBF-processed AZ31 alloy [71]. Reprinted from [71], Copyright (2021), with permission from Elsevier. and (d) [71]. The similar phenomenon has also been reported in the LPBF-processed Mg alloys. For instance, Esmaily et al presented a fully columnar-grain structure in the LPBFprocessed WE43 alloy as shown in figure 11(a) [51]. On the contrary, Bär et al found that elongated columnar grains nucleated at the edge of a melt pool and grew towards the center during the LPBF process of WE43 alloy, while equiaxed grains nucleated ahead of the columnar grain zone due to CET occurrence as demonstrated in figure 11(b) [82]. The formation mechanism of different grain morphologies in additively manufactured Mg and its alloys will be discussed in detail in section 3.1.2.
Most of the investigations demonstrate that the additively manufactured Mg and its alloys with equiaxed-grain structure exhibits weak texture [78,81,[83][84][85]. However, Wang et al fabricated an equiaxed-grain structure in Mg-3.26Y-3.51Sm-2.42Zn-0.62Zr alloy via LPBF, and the EBSD results illustrated a strong texture as shown in figure 12 [86]. Unfortunately, they did not clarify the reasons for this abnormal phenomenon in their paper. The recent investigation Figure 11. Grain morphologies of LPBF-processed WE43 alloy: (a) columnar-grain structure in reference [51]. Reprinted from [51], Copyright (2020), with permission from Elsevier. (b) EBSD map presenting the grain morphologies of the last melt pool and its surroundings [82]. Reprinted from [82], Copyright (2019), with permission from Elsevier. by Bär et al provided a possible explanation for the above observation [82]. As shown in figure 11(b), the laser movement was along the x-direction, and the columnar grains of the last track in the last layer grew along the direction perpendicular to the melt pool boundary. Consequently, columnar grains elongated in the y/z direction could be identified (Region I). As for the second-last layer, the laser movement was along the ydirection, and the columnar grains should be elongated in the x/z direction accordingly, which means these grains grew perpendicular to the observation plane. This is to say, the columnar grains in the second-last layer were equiaxed-like in the yz plane (Region II). It is reasonable that the equiaxed-like grains in Region II exhibited a strong texture. Therefore, the grains manufactured by Wang et al may be columnar, and just equiaxed-like in their observation plane. There is also another explanation on the strong-textured equiaxed grains. It is well known that hot extrusion process induces dynamic recrystallization in Mg alloys, and the newly formed recrystallized grains normally have strong texture. In the process of AM, on the one hand, cyclically rapid heating and cooling process generally leads to large thermal stress formation; on the other hand, layer-by-layer deposition feature inevitably enhances thermal accumulation, the effect of which acts as post heat treatment process for the previously deposited layers. Consequently, dynamic recrystallization may occur during AM process. Hu et al found that the thermal stress could exceed 600 MPa and the overall temperature of deposit mainly ranged from 800 • C to 1050 • C in the process of laser additively manufactured Inconel 625 superalloy [87]. When the thermal stress was higher than the yield strength of Inconel 625 in the temperature range, the plastic deformation occurred as well as dynamic recrystallization. The recrystallization temperature and yield strength of Mg alloys are much lower than those of Inconel 625. Theoretically speaking, plastic deformation and dynamic recrystallization may also occur in the process of additively manufactured Mg alloys although there is no relevant report. In this case, equiaxed grains with strong texture may be formed under the comprehensive action of thermal stress and thermal accumulation during AM process of Mg alloys. This kind of strong-textured equiaxed-grain structure has been reported in a solid-state additively manufactured magnesium alloy [88]. Table 2 summaries the grain sizes of additively manufactured Mg and its alloys in different references. It is evident that the additively manufactured Mg and its alloys have much finer grains than the corresponding as-cast materials [59,63,81]. For instance, Zumdick et al indicated that the LPBF-processed WE43 magnesium alloy exhibited extremely fine grains with the size of approx. 1 µm, which is much smaller than the ascast WE43 with an average grain size of approx. 44.3 µm [59]. In addition, it is also observed from table 2 that the grain sizes of LPBF-processed Mg and its alloys (ranging from ∼0.8 µm to ∼21.6 µm) are finer than that of WAAM-processed ones (ranging from ∼17 µm to ∼55 µm). Cooling rate is one of the key factors influencing grain size in additively manufactured materials. For the material with same chemical compositions, the higher the cooling rate, the finer the grain size. It is generally accepted that the cooling rate of these three processes have the following relationship: LPBF > WAAM > cast. As a result, the LPBF-processed Mg and its alloys normally have the finest grains, while the as-cast materials exhibit the coarsest grains.
Chemical composition of material is another key factor affecting grain size. Up to now, the effects of Al [43,79], Mn [93][94][95], Zn [44], Cu [92,96], Ce [97], Dy [98] on the grain size of additively manufactured magnesium alloys have been investigated. The results demonstrated that the addition of Al, Cu, Zn, Ce and Dy can significantly refine the grains, and the grain size decreases monotonically with the content increase of these elements. For instance, Shuai et al found that a 1 wt.% of Al additive remarkably reduced the grain size of LPBF-processed Mg-3Zn alloy from (21.6 ± 2.6) µm to (10.5 ± 1.9) µm, and with increasing the Al content to 7 wt.%, the grain size was further decreased to (6.8 ± 1.5) µm [79]. However, it is reported that the Mn addition had little impact on grain refinement [94]. The above phenomenon is ascribed to the undercooling degree variation with chemical composition addition. Specifically speaking, the elements of Al, Zn, Cu, Ce and Dy can enhance the undercooling degree during AM of magnesium alloys while Mn contributes little to the undercooling degree.
Other additives, such as La 2 O 3 [99], SiCnps [64] and GO [100], were also utilized to refine the grain size in additively manufactured magnesium alloys. Niu et al indicated that the average grain size of LPBF-processed AZ91D alloy was decreased from 3.3 µm to 1.2 µm as the content of SiCnp additive was increased from 0 to 7.5 wt.% [64]. The reasons for grain refinement induced by SiCnp are as follow: on the one hand, the SiCnp at the S-L interface remarkably inhibited the diffusion of Al atom, resulting in solute enrichment and undercooling enhancement at the interface; on the other hand, the undissolved SiCnp in melt pool may act as extra heterogeneous nucleation sites, thus increasing potential nucleus.
The unique 'layer-by-layer deposition' feature of AM process significantly influences the final grain characteristics of magnesium alloys [101,102]. First of all, the deposition of current layer has a thermal impact on the previously deposited layers, leading to grain characteristic variation. Wang et al observed that a HAZ existed between two adjacent deposited layers during the WAAM process of AZ31 alloy, and the HAZ interlayers had coarse equiaxed grains while the deposited layers had fine columnar dendrite grains [71]. The similar phenomenon was also reported in a LPBF-processed magnesium alloy. Wang et al found that refined equiaxed grains emerged in the center of a scanning track during the LPBF of Mg-3.4Y-3.6Sn-2.6Zn-0.8Zr, while columnar grains formed in the overlapping region between two tracks [103]. Secondly, the 'layer-by-layer deposition' feature induces the appearance of cyclic thermal history, promoting the formation of a periodic microstructure [71]. Thirdly, thermal accumulation can be caused by the 'layer-by-layer deposition' feature, and grain characteristics will be changed from the bottom to the top of built sample [86,104]. The investigation by Takigi et al demonstrated that the bottom region of the WAAM-processed AZ31B alloy was composed of the finest equiaxed grains, and the grain size was increased gradually along the building direction [54]. Bär et al indicated that the grain morphologies and sizes changed remarkably during the LPBF process of WE43 alloy, and the texture of the current layer was smaller than that of the previously deposited layers [82].
Apart from process parameters, post treatment process is another common way to adjust grain characteristics [39,65,84,89,99]. Cheng et al indicated that arc oscillation (including spiral oscillation and asymmetric trapezoid oscillation) was helpful to refine the microstructure of the WAAMprocessed AZ31 alloy [89]. Deng et al investigated the effects of heat treatment temperature and duration on the grain size of the LPBF-processed GZ112K (Mg-11Gd-2Zn-0.4Zr) alloy in detail [39]. They found that grain coarsening was inevitable during the post heat treatment process. With the increase of heat treatment duration, the grain coarsening rate was decreased. In addition, it is noteworthy that the change in grain size was not obvious in the temperature range of 350 • C to 450 • C, while the grains significantly grew at 480 • C-520 • C, which means that the heat treatment temperature contributes more to grain coarsening than duration. Similar grain coarsening phenomenon in the post heat treatment process is also observed in the WAAM-processed AZ91 alloy by Zhang et al [84]. They indicated that the texture was obviously enhanced after the heat treatment process of 415 • C for 8 h, which was ascribed to the growth of grains with preferential orientation leading to the disappearance of other orientated grains. The current investigations are mainly focused on grain growth during post heat treatment process. However, recrystallization behavior of AM-processed magnesium alloys and its influence on grain characteristics have not yet been studied. In addition, grain growth during high temperature process inevitably results in strength reduction according to Hall-Petch relation. In order to address this issue, Deng et al introduced FSP to alter the grain characteristics of the LPBFprocessed G10K (Mg-10Gd-0.2Zr) alloy [65]. It is evident from figure 13 that the original columnar grains in the asbuilt state (figure 13(a)) were changed to the equiaxed grains in the FSP state (figure 13(c)) and the grain size was significantly decreased from 26.8 µm (figure 13(b)) to 5.8 µm (figure 13(d)). The results proved that the FSP is a promising post treatment process for grain refinement in additively manufactured magnesium alloys.

Formation mechanism.
For AM-processed singletrack material, its final grain morphology is closely related to the fact that whether new grains are formed at the front of S-L interface during solidification process. If the solidification conditions at the front of S-L interface cannot promote adequate nucleation sites to become new grains, then columnar grains dominate; otherwise, CET occurs and equiaxed grains predominate. Furthermore, it is noteworthy that the dissolution or growth of potential nucleation sites is determined by constitutional undercooling degree at the front of S-L interface. High constitutional undercooling degree is beneficial for the formation of new grains.
Equation (2) is the criterion for constitutional undercooling occurrence, where, G represents the temperature gradient, R represents the solidification velocity, m is the liquidus slope, C 0 is the initial compositions of alloying elements, k 0 is the partition coefficient, D L is the diffusion coefficient of solutes in liquid, Q = mC 0 (1 − k 0 ) represents the growth restriction factor and ∆T = mC 0 (1 − k 0 ) /k 0 represents the solidification temperature range.
On the one hand, it is evident from the viewpoint of the left side of equation (2) that small temperature gradient G and high solidification velocity R are conducive to generating constitutional undercooling when the chemical compositions of materials are fixed. In other words, large G combined with low R will induce the formation of columnar grains due to lack of enough constitutional undercooling. During AM process, the bottom of melt pool has the largest temperature gradient and the smallest solidification velocity of close to 0. Therefore, columnar grains are normally observed in the bottom region of melt pool [34,45,71,82]. As melt pool further solidifies, G becomes smaller and R becomes higher, which contributes to CET occurrence and equiaxed grain formation [45,82]. In the process of manufacturing block materials, thermal accumulation caused by 'layer-by-layer deposition' feature normally reduces temperature gradient G. That is to say, constitutional undercooling will be much easier to appear as deposition process progresses, leading to equiaxed-grain structure formation [105]. That is one of the main reasons why most of the additively manufactured Mg and its alloys are characterized by equiaxed grains.
On the other hand, chemical compositions also play a vital role in generating constitutional undercooling as shown in the right side of equation (2). Generally speaking, any alloying elements that can enhance the growth restriction factor Q or the solidification temperature range ∆T are conducive to the occurrence of constitutional undercooling [106][107][108]. Table 3 presents the liquidus slope m, equilibrium partition coefficient k 0 and m (1 − k 0 ) of various alloying elements in magnesium [108]. It is evident that the addition of Al, Zn, Cu and Ce can enlarge the value of m (1 − k 0 ) in magnesium, and higher growth restriction factor Q will be obtained with more concentration of such alloying elements. As a result, an equiaxed-grain structure is likely to be prepared in additively manufactured magnesium alloys, which is consistent with the experimental observation in references [43,44,79,96,97]. This is another main reason for the formation of equiaxed-grain structure in additively manufactured magnesium alloys. However, as mentioned in section 2.2, excessive grain-refinement elements (such as Zn) may result in cracking. Therefore, the concentration of alloying elements in magnesium alloys should be carefully designed to achieve fine grains with maintaining good formability during AM process.
The variation of chemical compositions inevitably induces precipitation of second phases. It is reported that some kinds of second phases (such as Mg-Zn-Dy phase in Mg-3Zn-xDy alloy [98]) can precipitate before matrix solidifies, and thus provide extra heterogeneous nucleation sites, which promotes grain refinement [43,98]. This case is similar to the situation that adds high-melting-point particles into Mg or Mg alloys to  Table 3. Q values of alloying elements in Mg alloys at C 0 = 1 wt.% for binary systems [109,110]. prepare Mg-based composites. Although undercooling degree is not enhanced, the number of heterogeneous nucleation sites is remarkably increased, resulting in more new grains forming ahead of S-L interface and final microstructure is refined accordingly. The utilization of frequency pulsed arc in WAAM of Mg alloys has a similar effect [52,111]. Guo et al indicated that the pulsed arc induced melt pool oscillation, leading to dendrite fragmentation at the front of S-L interface and providing extra nucleation sites. Consequently, the grain size was decreased compared with the case using continuous arc [111].
Since there are no alloying elements to provide constitutional undercooling, the undercooling degree during solidification process of pure Mg is small so that CET should not occur. Theoretically speaking based on the above analysis, AM-processed pure Mg is expected to be mainly consist of columnar grains. However, it is interesting to note that the microstructure of additively manufactured pure Mg is also characterized by equiaxed grains [67,76]. One possible explanation is that the equiaxed-grain structure is attributed to dynamic recrystallization, which is caused by the combination action of low recrystallization temperature of pure Mg, thermal accumulation, and high thermal stress during AM process.
In a short summary, the CET and dynamic recrystallization are two main methods to generate equiaxed grains in additively manufactured Mg and its alloys. Current research usually considers CET to be the main cause of forming equiaxed-grain structure, while ignoring the effect of dynamic recrystallization. This means that more effort needs to be made to clearly elaborate the detailed underlying mechanism of grain formation in AM of Mg and its alloys.

Relationship between chemical composition, process parameters and grain size.
A analytical model that takes into account both chemical compositions and nucleation particles was developed by St John et al to describe the relationship between growth restriction factor and grain size at fixed process parameters [108,112]. The mathematical expression for this model is shown below, where d is the grain size, ρ is the density of nucleation particles, f is the fraction of the activated nucleation particles, ∆T n is the undercooling degree required for nucleation, v is the growth rate of grain, a and b are fitting factors. It has been proved that equation (3) is applicable for cast Mg alloys  [115]. Reprinted from [115], Copyright (1995), with permission from Elsevier. [108,113,114]. Figure 14 summaries the grain sizes of LPBFprocessed Mg alloys and their corresponding Q values [43,79,92,97,98]. It is evident that the linear relationship between grain size d and 1/Q (R 2 value is higher than 0.8) is suitable for most of the LPBF-processed Mg alloys. However, there still exists poor linear fits (R 2 value is lower than 0.4). For instance, Wang et al found that the grain size was decreased first and then increased with increasing Cu content (red ball in figure 14) [92]. They thought that the thermodynamic driving force should be considered when describing the relationship between d and 1/Q. Specifically speaking, if Cu addition was excessive, although the number of heterogeneous nucleation sites was increased, the thermodynamic driving force during solidification was reduced as well, and the heterogeneous nucleation process was inhibited accordingly, leading to coarse grain formation. The same phenomenon has also been reported in cast Mg alloys [110] and Al alloys [115], which is believed to be ascribed to the variation of undercooling degree with solute concentration. Figure 15 demonstrates how the average grain size varies over the Q −1 value. It is well known that the undercooling degree ∆T = ∆T c + ∆T t + ∆T r , where ∆T c represents the constitutional undercooling degree, ∆T t represents the thermal undercooling degree and ∆T r is the curvature undercooling degree. When the concentration of alloying elements is low, the grain growth is controlled by thermal conditions, indicating that ∆T t controls the grain size (Area I in figure 15). Increasing alloying element concentration will cause a transition to diffusion-controlled equiaxed grain growth. And in Area II in figure 15, ∆T c dominates the grain size. Then with further raising the alloying element concentration, ∆T r works (Area III in figure 15). It is evident that this explanation believed that the dominating driving force for grain growth at high alloying element concentration is the curvature undercooling degree [110,115], which is contradiction with the explanation of Wang et al. Different researchers present different explanation on the same phenomenon, suggesting a lack of in-depth and comprehensive understanding on the formation and growth of grains in Mg alloys. Energy density is widely used to evaluate the effect of process parameters on microstructure. In this study, VED and AED are utilized for block samples and single track, respectively. The expressions for VED and AED are shown below, where P is the laser power, V is the scanning velocity, h is the hatch spacing and t is the layer thickness. Similar to the relationship between grain size d and 1/Q, linear relationship can be also observed between d and energy density as shown in figure 16 (the data is from [33,43,61,117]). It is generally accepted that grain size is increased with enhancing energy density [118], which is attributed to the accordingly reduced cooling rate. The decrease in cooling rate not only reduces potential nucleation sites ahead of S-L interface, but also provides more time for grain growth, thus leading to grain coarsening. It is reasonable that grain size will not demonstrate limitless linear increase with energy density. Ng et al found that when the AED was enhanced to 9.8 × 10 6 J·mm −2 , the increasing rate of grain size was reduced as shown in figure 16(b) (red balls) [117]. In addition, an opposite linear relationship that grain size is decreased with energy density has been reported by Niu et al as shown in figure 16(a) (red, orange and green balls) [43]. They ascribed this abnormal phenomenon to the Zenner pinning effect of Mg 17 Al 12 particles on grain boundary migration. In cast Mg-Al alloy, Cao et al elaborated a native grain refinement mechanism in Mg alloys [116], which may present another explanation on Niu's observation. As shown in tables 3-2, the Q values of Zn and Al are very close, which suggests that the grain size of Mg-Zn alloy should be equivalent to that of Mg-Al alloy. However, Cao et al found that the Mg-3Zn alloy owned a much coarser native grain than their Mg-3Al alloy under the same processing conditions [116]. They believed that the Al 4 C 3 phase with high melting point was the underlying reason for native grain refinement as it increases the amount of valid potential nucleation sites during solidification [116,119]. Therefore, another possible reason for Niu's observation is that the Al 4 C 3 phase particles may also form during the LPBF of Mg-Al alloys, and higher VED lead to more Al 4 C 3 particles, thus realizing finer grains. Recently, Xu et al observed that a low laser energy density led to coarser grain formation than high laser energy density, and lamellar structure formed in the large grains [118]. However, they did not present an interpretation on this abnormal but interesting phenomenon. A more accurate and in-depth understanding of the underlying grain refinement mechanism is essential to help us establish a more effective relationship between chemical composition, process parameters and grain size.

Phase constituent
The phase constituents of Mg alloys are highly dependent on their chemical compositions. Furthermore, process parameters and post heat treatment processes also have a vital role in altering the phase constituents. Nowadays, additively manufactured Mg alloys can be mainly classified into three categories, including Mg-Al series, Mg-RE series, and Mg-Zn series, which will be described below in detail.  [123]. Reprinted from [123], Copyright (2021), with permission from Elsevier. (b) LPBF-processed AZ61 [124,125]. Reprinted from [124], Copyright (2020), with permission from Elsevier.

Mg-Al series.
Mg 17 Al 12 is the most common secondary phase in WAAM-and LPBF-processed Mg-Al series magnesium alloys, such as AZ31 [52,89,120], AZ61 [53,121], AZ91 [64,91], AZ80M [122], AEX11 [90]. It is observed that the Mg 17 Al 12 phase is normally precipitated in the inter-dendritic region with an irregular shape as shown in figure 17 [90,120]. According to Mg-Al phase diagram, the formation of Mg 17 Al 12 phase is due to the eutectic reaction caused by Al segregation during the end of non-equilibrium solidification. Therefore, the morphology and size of the Mg 17 Al 12 phase particles are closely related to previously formed dendrite framework. On the one hand, high cooling rate leads to small primary dendrite arm spacing, resulting in fine Mg 17 Al 12 phase particles [55]. On the other hand, various combination of temperature gradient and scanning velocity induces different dendrite morphology, which brings about changes in Mg 17 Al 12 phase morphology. That is the main reason causing differences in size and morphology of Mg 17 Al 12 phase particles between WAAM-and LPBF-processed Mg-Al alloys. The above analysis also enlightens us that the additives that can refine the dendritic structure have a potential to refine the Mg 17 Al 12 phase particles, such as La 2 O 3 [99] and SiCnp [64]. In addition, the Al content also has an inevitable impact on the Mg 17 Al 12 phase. Niu et al illustrated that the volume fraction of the Mg 17 Al 12 phase particles was increased, and the morphology gradually became reticulate, with the increase of Al content [43].
Post heat treatment processes can be utilized to optimize the size and morphology of Mg 17 Al 12 phase. It is generally accepted that the solution heat treatment at a temperature higher than 400 • C with a duration of several hours can induce the Mg 17 Al 12 phase particles dissolution into the matrix. During this process, the globularization of the Mg 17 Al 12 phase occurs, and its morphology is changed from irregular shape to granular shape, accompanied with a more dispersed distribution [84,90,99,122]. Subsequently, an aging heat treatment at 180 • C-220 • C induces the formation of secondary Mg 17 Al 12 phase particles [92,99,122]. Compared with the primary Mg 17 Al 12 phase particles, which is formed during solidification process, the secondary ones formed during aging heat treatment process exhibit a much smaller size and a more uniform distribution.
The thermal accumulation caused by 'layer-by-layer deposition' feature during AM process normally leads to inhomogeneous characteristics of the Mg 17 Al 12 phase along the building direction [105,123,125]. Guo et al indicated that the stable temperature during WAAM process was always higher than 150 • C, and the peak temperature could reach 400 • C-450 • C, which approached or even exceeded the dissolution temperature of the Mg 17 Al 12 phase (437 • C) as shown in figure 18(a) [123]. Consequently, the volume fraction of the Mg 17 Al 12 phase was decreased from 6.72% in the 49th layer ( figure 18(b)) to 3.75% in the 46th layer (figure 18(c)) due to its partially dissolution caused by the subsequent several depositions. As the deposition process continued, the average temperature of the deposit was decreased, and the thermal effect would act like aging heat treatment instead of solution heat treatment, resulting in the precipitation of secondary Mg 17 Al 12 phase and thus increasing its content as demonstrated in figure 18(d). Furthermore, the thermal accumulation gradually changes the solidification conditions of deposit, which influences the formation of the Mg 17 Al 12 phase according to the analysis mentioned in the first paragraph in section 3.2.1 [125]. However, the variation of Mg 17 Al 12 phase along the building direction is not observed in LPBF-processed Mg-Al alloys. When heat input is low, the thermal accumulation may not be adequate to dissolve primary Mg 17 Al 12 phase and induce the formation of secondary Mg 17 Al 12 phase. As a result, a relatively homogeneous phase constituents along the building direction will be obtained.
There are also other phases in AM-processed Mg-Al alloys, especially when adding other elements. For instance, Al 2 Y phase was observed in WAAM-processed Y-containing Mg-Al alloy (AZ80M) [122,123]; Al 2 Cu phase was formed in LPBF-processed Cu-containing Mg-Al alloy [92]; needle-like Al 4 Ce phase was precipitated in LPBF-processed Ce-adding AZ61 alloy [97]. Mn is one of the most conventional alloying elements in Mg alloys. In Mn-containing Mg-Al series magnesium alloys, Al 8 Mn 5 is the second common phase in addition to Mg 17 Al 12 [52,55,60,125,126]. It is reported that  [125,127]. The morphology of the Al 8 Mn 5 phase particle is normally spherical or granular [52] with distributing both inside the grains and along the grain boundary [53]. Furthermore, Yang et al figured out that the Al 8 Mn 5 phase particles were effective nucleation sites for the Mg 17 Al 12 phase formation [125]. However, there are still some contradictions over the formed phase type in Mn-containing Mg-Al series magnesium alloys. Kim et al indicated that the main phase in the WAAM-processed AZ91D is orthorhombic Al 5 Mg 11 Zn 4 via XRD testing [75,128]. The investigation by Li et al showed that the Al 86 Mn 14 phase rather than Al 8 Mn 5 phase formed in the WAAM-processed AZ31 alloy [81]. Nevertheless, these different conclusions are drawn based on XRD only, which is lack of more accurate proof.

Mg-RE series.
In AM-processed Mg-RE series magnesium alloys, oxides formation is inevitable even if in an atmosphere with extreme low oxygen content [39,48,49,51,59,65]. Esmaily et al conducted a real-time monitor on the oxygen level during LPBF process of WE43 alloy [51]. They found that the remaining oxygen in the chamber reached near 0 ppm and stayed at that extreme low level throughout manufacturing as shown in figure 19(a). However, strong evidence was still provided to prove the formation of various oxygen-rich phases, such as Y 2 O 3 ( figure 19(c)), Nd-Y-O ( figure 19(d)) and Y-Nd-Zr-O ( figure 19(e)). They ascribed the appearance of oxides to the original powders, which have oxides formed during powder preparation due to the remarkably high affinities of the main alloying elements (such as Y and Zr) in WE43 alloy for oxygen. Lovašiová et al also gave a similar explanation on the oxides formation in LPBF-processed WE43 alloys [129]. Zumdick et al compared the XRD result of WE43 original powders with that of LPBF-processed WE43 block [59]. The results proved that the WE43 original powders did have oxides as demonstrated in figure 20(a), and the LPBF-processed material had more oxides than the original powders as shown in figure 19(b) [59], suggesting that the formation of oxides might also be attributed to the residual low oxygen content in the chamber [49,65]. The oxides in AM-processed Mg-RE alloys, such as Y 2 O 3 and Gd 2 O 3 , always have a flaky shape as shown in figure 19(b). Apart from Y 2 O 3 and Gd 2 O 3 , MgO phase is also found in AM-processed Mg-RE material. However, it is difficult to proof this since MgO phase particles have no prominent features that can be discerned via SEM or OM and the diffraction peaks of MgO phase coincide with those of Mg 3 Gd phase as shown in figure 20. The investigation by Wei et al indicated that MgO phase could be detected by XRD in the LPBFprocessed ZK60 alloys [34], which provides a circumstantial evidence for the present of MgO phase in AM-processed Mg-RE alloys. Nevertheless, the content of MgO should be low in additively manufactured Mg alloys, since AM process has a high heating rate which suppress the oxidation of Mg [130]. Tong et al calculated the Gibbs free energy of some  [51]. Reprinted from [51], Copyright (2020), with permission from Elsevier. possible oxidation reactions in Mg-RE alloys, and indicated an oxygen affinity order of Y > Gd > Nd > Mg > Zr [131], confirming that Y 2 O 3 was the preferred oxide phase in AMprocessed Mg-RE alloy, which is consistent with the experimental observations.
It is worth noting that the oxide inclusions or oxidations are rarely observed in WAAM-processed Mg-RE alloys. Wang et al indicated that no obvious oxide inclusions was observed in the WAAM-processed WE43 alloy [101]. Tong et al also found that the volume fraction of oxide inclusions in the WAAM-processed WE43 was less than 0.3%, significantly lower than that in the LPBF-processed counterparts (at least 7.3%), and further illustrated that the number density of the oxide inclusions did not vary a lot along the building direction [131]. They ascribed this to the low content of oxide in the feedstock materials, low exposure time for the melt pool to the atmosphere and the positive effect of the arc on removing oxide film. Such observation is quite inspired, since the chamber is no longer necessary for the WAAM of Mg alloys, which means that the size of Mg alloy part is not restricted. More experiments are needed to confirm this.
The formation of oxides will influence solidified microstructure as well as secondary phase precipitation during heat treatment process. Hyer et al found that the formation of Y 2 O 3 suppress the precipitation of β ′ -(Mg 12 NdY)/(Mg 24 Y 2 Nd 3 ) and β ′′ -(Mg 3 (Y, Nd)) since the oxide occupied the main forming element Y of these two kinds of precipitates [49]. In addition, the Y 2 O 3 phase did not dissolve or coalesce during the heat treatment process at a temperature of 536 • C for a duration of 24 h, which further inhabited the appearance of β ′ and β ′′ phases. The phenomenon that oxides do not dissolve during solution heat treatment process has also been reported by other researcher [39,51,66,132].
The phase constituents of AM-processed Mg-RE alloys are more complex than those of AM-processed Mg-Al alloys, since different combinations of RE elements result in various phase constituents. Table 4 summaries the phase constituents in the as-built and heat treated Mg-RE alloys fabricated by LPBF process [33, 35, 38, 39, 48-51, 59, 63, 65, 82, 85, 86, 103, 129, 133]. It is evident that the Mg 5 Gd phase and LPSO structure are only observed in LPBF-processed Mg-Gd-Zn-Zr alloys [39,65,85]. Yuan et al indicated that the  [59]. Reprinted from [59], Copyright (2019), with permission from Elsevier. morphology of Mg 5 Gd changed from lamellar to networklike with the increase of Gd content in the LPBF-processed Mg-Gd binary alloys [134]. Zheng et al found that metastable Mg 3 Gd to stable Mg 5 Gd transition occurred during the LDED of GA151K (Mg-15Gd-1Al-0.4Zr) alloy due to multiple thermal cycles [102]. In addition, the multiple thermal cycles also lead to in-situ precipitation of Mg 7 Gd with a high area number density of 1.56 × 10 −3 nm −2 . When Gd element is changed into Nd, the main precipitate in LPBFprocessed Mg-Nd-Zn-Zr alloys is Mg 12 Nd [133]. In addition, it is reported that Mg 3 RE (Mg 3 Gd, Mg 3 Nd), Mg 41 Nd 5 and Mg 24 Y 5 are all observed in LPBF-processed Mg-Y-N-Zr alloys [48,51,129]. Furthermore, the Mg 3 RE phase is also the main precipitate in LPBF-processed Mg-Y-Sm-Zn-Zr alloys [35,86,103].
It can be seen from table 4 that AM-processed Mg-RE series magnesium alloys exhibit changeable phase constituents, even if the same alloy is considered. Esmaily et al found that Mg 41 Nd 5 and Mg 24 Y 5 phases formed in the LPBFprocessed WE43 alloy [51]. However, Hyer et al did not observe and detect these two phases by SEM, TEM and XRD during LPBF of WE43 [49]. Interestingly, they indicated that Al 3 Zr might be precipitated in the as-built microstructure, while nobody else reported a similar phenomenon. In addition, hcp-Zr particles and intergranular Mg 14 (Nd, Gd) 2 Y phase are only observed in the LPBF-processed WE43 alloy by Nilsson Åhman et al [48] and Liu et al [50], respectively. Similar controversy also exists in WAAM-processed Mg-RE alloy. Wang et al found Mg 41 Nd 5 and Mg 3 Nd phases in the WAAMprocessed WE43 alloy [101]. However, in addition to Mg 3 RE phase, Tong et al also observed Mg 12 RE phase in the WAAMprocessed WE43 alloy that was not reported in other relevant investigations [131]. In fact, it is easy to understand that the above phenomena have a close relation with the characteristics of AM process. On the one hand, the AM of materials is a non-equilibrium solidification process, in which metastable phases (such as Mg 3 RE, Mg 24 Y 5 , Mg 12 YNd in table 5) formed at elevated temperature are easily retained till room temperature. On the other hand, thermal accumulation occurs during AM process as mention in section 3.2.1. When the thermal accumulation is high, the deposition of current layers will have a heat treatment effect on previously deposited layers, resulting in the appearance of phase transitions or phase dissolution [102]. For instance, according to the precipitation sequences in Mg-RE alloys as shown in table 5, stable phases, such as Mg 41 Nd 5 , Mg 5 Gd, Mg 14 Nd 2 Y, may be precipitated in the previously deposited layers during AM process. Zheng et al confirmed the precipitation of stable phase Mg 5 Gd and the dissolution of metastable stable phase Mg 7 Gd due to thermal accumulation caused by thermal cycles during the LDED process of GA151K (Mg-15Gd-1Al-0.4Zr) alloy [102]. However, the analysis is only a rough prediction based on the as-cast microstructure and AM process characteristics. Although the observed phenomena in the investigations can be Mg-Nd-Zn-Zr Mg 12 Nd Dot particle Inside grain [133]  well explained though the analysis, it is still insufficient to precisely tune the phase constituents to meet specific performance requirements.
Mg-RE alloys are expected to have high strength and excellent creep and corrosion resistance. The LPSO structure is essential to remarkably enhance this expected performance of Mg-RE alloys. It is well known that the LPSO structure in Mg alloys is generally composed of regularly alternating of HCP α-Mg layers and RE-and TM-rich FCC layers [137]. There are mainly five types of LPSO structure, including 6H, 10H, 14H, 18R and 24R, which is classified depending on the number of Mg layers that intersperse FCC layers [137][138][139]. Different kinds of LPSO structures can be interconverted. In additively manufactured Mg-RE alloys, only 14H-LPSO structure is reported as shown in figures 21(a) and (b) [38,39] due to limited investigations. In addition, it is noteworthy that the 14H-LPSO structure is barely observed in as-built Mg-RE alloys fabricated by LPBF process [33,38,39]. Theoretically speaking, the formation of LPSO structure is closely related to the SFE of Mg alloys. Higher SFE leads to easier formation of LPSO structure. Shuai et al found that Mn addition could lower the SFE of ZK30-10Gd alloy and promote the precipitation of LPSO structure in as-built LPBF microstructure [95]. However, the precipitation behavior of LPSO structure was still not adequate during the LPBF process. Nevertheless, their investigation still provides an effective approach for in situ preparation of LPSO structures in LPBF-process Mg alloys without post heat treatment.
Basal γ ′ precipitate, which is the basic unit of 14H-LPSO structure, can be formed during AM process as shown in figures 21(c) and (d). There are orientation relationships between basal γ ′ precipitate and Mg matrix, which are   [39]. Reprinted from [39], Copyright (2022), with permission from Elsevier. (c) and (d) Basal γ ′ precipitate [38]. Reprinted from [38], Copyright (2022), with permission from Elsevier. range of 400 • C-480 • C was appropriate for the full transformation from eutectic phase to lamellar 14H-LPSO structure in the LPBF-processed Mg-11Gd-2Zn-0.4Zr alloy [39]. Higher temperature would cause the disappearance of LPSO structure. Furthermore, the area fraction of lamellar 14H-LPSO structure was increased first and then decreased with prolonging duration at 450 • C [39]. The decrease in area fraction of 14H-LPSO structure was ascribed to the growth of X phase at grain boundaries, since their formation both needed to consume Gd and Zn elements. Deng et al further illustrated that the crystal structure, lattice parameters, and chemical composition of lamellar LPSO structure and X phase were identical [38]. However, the 14H-LPSO structure was normally precipitated from supersaturated Mg matrix and distributed inside grains, while the X phase was mainly transformed from the eutectic phase and located at grain boundaries.

Mg-Zn series.
It is interesting to note that there is no commonly accepted conclusion on the phase constituents of LPBF-processed Mg-Zn alloys. Some researchers believe that MgZn 2 is the only intermetallic phase in Mg-Zn-Zr system and is easily precipitated at grain boundaries [93,96,100,140]. For instance, Tao et al indicated that the precipitates formed at the grain boundaries in the LPBF-processed ZK30-xGO (GO stands for GO) were MgZn 2 phase particles [100]. However, the conclusions about the phase constituents in LPBF-processed Mg-Zn alloys given in these investigations are not very reliable, because only EDS and XRD are used  [42]. Reprinted from [42], Copyright (2021), with permission from Elsevier. to identify second phase particles. On the one hand, the EDS analysis results in the investigations cannot coincide well with the standard chemical formula of MgZn 2 [96,100,140]. On the other hand, the diffraction peaks of MgZn 2 and Mg 7 Zn 3 are relatively close, which makes it difficult to distinguish them via XRD. When other atoms enter these two types of second phases in the form of solid solution or substitution, lattice distortion appears, resulting in shift of diffraction peaks. This will further reduce the accuracy of phase constituent identification. On the contrary, the other researchers indicate that the grainboundary phase is Mg 7 Zn 3 [42,44,46] while MgZn 2 phase is distributed inside grains [45,141]. As shown in figure 22, Liu et al utilized TEM and SAED to present strong evidences that Mg 7 Zn 3 phase formed at grain boundaries in the LPBFprocessed ZK60 alloy [42]. Furthermore, Liang et al identified that the nano-scale particles inside grains were Zn 2 Zr 3 core with MgZn 2 shell in the LPBF-processed ZK60 alloy as demonstrated in figure 23 [45,142]. According to the above comparison and analysis, it is reasonable to believe that Mg 7 Zn 3 phase forms at grain boundaries and MgZn 2 forms in the grains during LPBF process of Mg-Zn alloy.
The change in Zn content, as well as the introduction of other alloying elements, alters the phase constituents of LPBFprocessed Mg-Zn alloys. Wei et al found that the volume fraction of Mg 7 Zn 3 was increased with increasing Zn content, and the morphology of Mg 7 Zn 3 transformed gradually from granular shape to nearly reticular structure [44]. In addition, it is reported that Ag, Cu, Mn + RE, Al, Ca and Dy additives can induce the formation of Mg 54 Ag 17 [143], MgZnCu [144], LPSO structure [95], Mg 17 Al 12 [79], amorphous phase [17] and Mg-Zn-Dy phase [98] in LPBF-processed Mg-Zn alloys, respectively. The emergence of these phases caused by alloying elements addition in Mg-Zn alloys will undoubtedly change the morphology, size, and volume fraction of the eutectic Mg 7 Zn 3 phase at grain boundaries, thereby reducing crack sensitivity and improving performances.

Current issues.
Through the above analysis, it is found that there are still some issues for the phase constituents in AM-processed Mg alloys. A sufficient understanding of the above information can help to develop novel Mg alloys and customized heat treatment for AM process, thereby accelerating the process of industrial application of additively manufactured Mg alloys.

Room temperature tensile property, strengthening and ductilization mechanism
4.1.1. Room temperature tensile properties. Figure 24 summarizes the room temperature tensile properties of additively manufactured Mg alloys in both as-built and heat treated states in recent published investigations [33, 38, 39, 44, 45, 49, 52-55, 59, 61, 64, 65, 71, 78, 90, 91, 105, 111, 122-125, 127, 131, 141, 145-148]. The following two main points of information can be obtained from figure 24. Firstly, the strength and ductility of additively manufactured Mg alloys can vary over a wide range, including UTS of 150-450 MPa with El of 0%-27%. And from an overall viewpoint, the WAAM-processed Mg alloys have lower strength and higher El compared to the LPBF-processed Mg alloys. Secondly, the additively manufactured Mg alloys demonstrate obvious strength-ductility tradeoff, which is also a common issue in Figure 23. TEM micrographs of the LPBF-processed ZK60 alloy: (a) HAADF image of nano-scale particles inside grains; (b) details of one nano-scale particle; (c) high resolution TEM micrograph and FFT filtered micrograph in the nano-scale precipitate [45]. Reprinted from [45], Copyright (2022), with permission from Elsevier.  engineering materials, such as Ti alloys, steel, and nickelbased superalloys. Anisotropy in mechanical properties is also observed in additively manufactured Mg alloys as shown in table 6. It is worth noting that the changing tendencies of tensile strength and El in vertical direction and horizontal direction are different in different investigations. There is still controversy in mechanical property variation with directions.
(1) Guo et al found that the average strength and El of WAAM-processed AZ80M in horizontal direction (normal to building direction) were superior to those in vertical direction (parallel to building direction) [105], which is consistent with the investigations by Cao et al [52], Guo et al [122], Tong et al [131] and Wang et al [71]. Guo et al believed that this phenomenon was ascribed to the inhomogeneous microstructure along building direction [105]. Ni et al [149] and Tong et al [131] proved that the mechanical properties of WAAM-processed Mg alloy varied along building direction, and further indicated that the vertical sample exhibited inferior mechanical properties to the horizontal sample extracted from the middle region of deposited block. (2) However, Yang et al showed that the average tensile strength and El of WAAM-processed AZ31 Mg alloy in horizontal direction were far inferior to those in vertical direction [125]. They ascribed this to the direction of the dendrites and the volume fraction of the interdendritic eutectic. Specifically speaking, they observed that the cracks easily appeared at the interfaces between dendrite arm and interdendritic eutectic. When the loading direction is parallel to the direction of the primary dendrite arms (corresponding to the case of vertical sample), the propagation of cracks between different dendrite arms was impeded because of the continuity of the dendrite arms, thereby resulting in good properties. When the loading direction is normal to the direction of the primary dendrite arms (corresponding to the case of horizontal sample), cracks could easily propagate along the primary dendrite arms, leading to poor properties. Such anisotropy in mechanical properties caused by columnar-grain structure has also been observed in AM-processed nickel-based superalloys and titanium alloys [13]. In addition, inhomogeneous microstructure along the building direction can also lead to poor mechanical properties in horizontal direction [149]. (3) Furthermore, some investigations reported that the vertical sample demonstrated comparable mechanical properties to the horizontal sample [78,90,122,127,145,148]. The difference between the mechanical properties in two directions was within the range of error. They interpreted the isotropy in mechanical properties to the formation of equiaxed grains and the homogeneous microstructure along the building direction.
Post heat treatment process is generally utilized to optimize the mechanical properties of additively manufactured Mg alloys [38,39,90,122,124,141]. For instance, Gneiger et al indicated that the heat treatment process of (415 • C/6 h + 430 • C/6 h)/water quenching + 260 • C/30 min/water quenching remarkably increased the strength and ductility of the WAAM-processed AEX11 alloys from 243 MPa and 4.5% to 299 MPa and 12.6% via dissolving hard and brittle Mg 17 Al 12 phase particles [90]. Liang et al designed a heat treatment process including solution stage (410 • C/24 h/water quenching) and double aging stage (90 • C/24 h/air cooling + 180 • C/24 h/air cooling) to significantly enhance the strength of LPBF-processed ZK60 alloy with maintaining the ductility. In addition, it is noteworthy that post heat treatment process can reduce, however cannot completely eliminate, the anisotropy in room temperature tensile properties [90,122] as shown in table 6. FSP is also an effective post treatment approach to ameliorate the mechanical properties of AM-processed Mg alloys. Deng et al demonstrated that the FSP sample exhibited superior combination of UTS and El (271 MPa and 7.5%) to the as-built sample (228 MPa and 2.2%) for LPBF-processed G10K (Mg-10Gd-0.2Zr) alloy [65]. Nevertheless, the applicability of FSP for additively manufactured Mg alloys is not as good as that of post heat treatment process due to difficult process integration of FSP and AM.

Contribution of different strengthening mechanisms to
yield strength. In general, the strengthening mechanisms of AM-processed Mg alloys are almost the same as those of traditionally fabricated counterparts. However, due to the distinct microstructure characteristics of additively manufactured Mg alloys, the contribution of various strengthening mechanisms to yield strength should be accordingly different compared with traditionally fabricated Mg alloys. Therefore, this section is mainly focused on the differences in predominant strengthening mechanisms among WAAM-processed, LPBF-process and traditionally fabricated Mg alloys. The detailed description on strengthening mechanisms and their related models is put into supplementary materials.
Grain boundary strengthening is critical in both AMprocessed and traditionally fabricated Mg alloys. According to the data provided by Deng et al [33], the contribution of grain boundary strengthening could reach 183-226 MPa, which provided 53.8%-69.7% of the overall yield strength in the LPBF-processed Mg-Gd-Zn-Zr alloy. However, the contribution of grain boundary strengthening in the as-cast counterpart was only 45.5 MPa, which was 27.9% of the overall yield strength. The result is easy to understand. The grain size in the as-cast microstructure is much higher than that in the LPBF-processed microstructure. According to the Hall-Petch relation, the larger the grain size, the smaller the strengthening effect it provides. It should also be noted that the Hall-Petch coefficients k H utilized in different papers are different, making it difficult to compare the computation results among different investigations. For instance, Chang et al utilized k H = 164 MPa·µm 1/2 to evaluate the grain boundary strengthening effect and indicated that the contribution of grain boundary strengthening in LPBF-processed Mg-9Al-1Zn-0.5Mn alloy could reach 38.1% of the overall yield strength [150], which is smaller than the results shown by Deng et al. If the utilized k H value was increased to 280 MPa·µm 1/2 (which is the same as the k H value used by Deng et al), the contribution proportion of grain boundary strengthening would be increased to 65.2%, which is consistent with the investigation by Deng et al. The similar comparison is also found in the WAAM-processed Mg alloys. Li et al [151] and Cao et al [148] used k H = 158 MPa·µm 1/2 and 250 MPa·µm 1/2 , respectively, to evaluate the grain boundary strengthening effects in the WAAM-processed Mg alloys. Li et al found that the contribution of grain boundary strengthening to yield strength in the WAAM-processed AZ31 alloy only occupied ∼20% of the overall yield strength, while Cao et al indicated that the proportion of grain boundary strengthening effect could reach 36.4% in the WAAM-processed Mg-Gd-Y-Zr alloy. Similarly, if they both used the same k H = 250 MPa·µm 1/2 , then the grain boundary strengthening contribution would be similar. Although the k H value affects the evaluation of grain boundary strengthening, it can still be concluded from the above comparison that the grain boundary strengthening effect in WAAM-processed Mg alloys is weaker than that in LPBFprocessed Mg alloys, since the former has much larger grain size than the latter as shown in table 2.
Solid solution strengthening is also an important strengthening mechanism in AM-process Mg alloys. For instance, Chang et al indicated the solid solution strengthening contributed 52.9 MPa to yield strength of LPBF-processed Mg-9Al-1Zn-0.5Mn alloy with the proportion of 17.7% [150]. Theoretically speaking, the solid solution strengthening effect has little to do with the forming process, which means the AMprocessed microstructure should exhibit similar solid solution strengthening effect to the as-cast microstructure as for a certain Mg alloy. However, the AM process has extremely high cooling rate that exceeds the casting process, leading to the formation of a far-from-equilibrium solidified microstructure. In this case, supersaturated solid solution will be generated, enhancing the solid solution strengthening effect in AMprocessed Mg alloys.
Dislocation strengthening contribution in WAAMprocessed Mg alloys is significantly different from that in LPBF-processed Mg alloys. Li et al indicated that the dislocation strengthening could only contribute 3.2-8.6 MPa to the overall yield strength (the proportion was 5%-8%) [151]. However, Chang et al found that the dislocation strengthening contribution in the LPBF-processed Mg alloy could reach 55.9 MPa, occupying 18.7% of the overall yield strength [150]. This observation is ascribed to the heat accumulation and cooling rate during AM process. Heat accumulation can provide in-situ annealing effect to reduce dislocation density, and high cooling rate normally results in high dislocation density. Compared with the LPBF process, the WAAM process has more remarkable heat accumulation and lower cooling rate, thus exhibiting weaker dislocation strengthening effect.
There are also some other strengthening mechanisms beneficial for AM-processed Mg alloys, such as HDI strengthening, LPSO strengthening, precipitate strengthening. However, the current literature in this field is very limited, making it difficult to conduct an in-depth discussion. For instance, Chang et al [150] and Li et al [151] discussed the Orowan strengthening effect caused by large nanoparticles in LPBF-and WAAMprocessed Mg alloys, respectively. However, the variation of precipitate strengthening effect (Orowan mechanism to cutting mechanism) with nanoparticle size is still unclear. Moreover, the precipitates in AM-processed Mg alloys are not all spherical. The combined strengthening effects of precipitates with various morphologies are also unclear. Much more effort is necessary in this field for generating high-strength Mg alloy for structural application. Table 7 summaries the corrosion resistance and degradation performances of additively manufactured Mg alloys in SBF. Firstly, it is apparent that the corrosion rate calculated by corrosion current density is lower than that computed by weight loss. The reason for this phenomenon is the so-called NDE [152]. Cathode of samples is activated under anodic polarization, resulting in the formation of surface films with a persistent cathodic effect. Consequently, the net anodic current density will be reduced by a cathodic value. In comparison, the corrosion rate determined by weight loss measurement or hydrogen volume detection is much more accurate. Nevertheless, the corrosion rate based on corrosion current density always shows the same trend as that evaluated from weight loss [100,144]. However, for Mg alloy scaffolds, the weight loss based corrosion rate is not accurate [135] because the corroded Mg may induce breakoff of uncorroded Mg and accordingly increase the weight loss [132]. Secondly, the corrosion rate usually exhibits a trend of first decreasing and then increasing with the increase of additive content. For instance, Shuai et al indicated that the minimum corrosion rate of LPBF-processed AZ61-xTi alloys appeared at x = 0.75 wt.% [153]. The increase or decrease of Ti additive content would increase the corrosion rate. Thirdly, solid solution treatment normally deteriorates the corrosion resistance of additively manufactured Mg alloys. Xie et al found that the corrosion rate of LPBF-processed ZK30 alloy in the as-built state (1.06 mm·yr −1 ) was much smaller than that in the SST state (1.38 mm·yr −1 ) [140]. Lastly, the relationships of the corrosion rates between as-cast and AM-processed Mg alloys are uncertain, which is ascribed to the possible appearance of defects such as pores and LoF. Wu et al illustrated that the LPBF-processed ZK60 owned a better corrosion resistance than the as-cast ZK60 as indicated by a lower I corr of 8.89 µA·cm −2 as compared with 18.5 µA·cm −2 for as-cast ZK60 with a nobler E corr of −1.52 V vs. SCE, and also a 30% decrease in hydrogen evolution rate [47]. However, Lovašiová et al demonstrated that the LPBF-processed WE43 alloy had a higher corrosion rate in SBF than the as-cast WE43 alloy [129]. Similar to the findings by Lovašiová et al, an inferior corrosion resistance of WAAM-processed Mg alloy to as-cast counterpart was also reported by Li et al [81].

Corrosion resistance and degradation.
The best corrosion rate, which is computed by weight loss, reported in each investigation is plotted in figure 25 [93-95, 97, 100, 121, 129, 132, 153, 154]. It is evident that almost all the corrosion rates of additively manufactured Mg alloys exceed the maximum corrosion rate (0.5 mm·yr −1 ) required for biodegradation implants, which means they can hardly be applied without any post treatment currently. Only AZ61-1.2Ce alloy demonstrated a desirable corrosion rate of 0.21 mm·yr −1 in SBF [97]. The authors interpreted that a moderate amount of Ce addition facilitated the formation of an appropriate amount of Al 4 Ce phase. The appearance of acicular Al 4 Ce phase consumed Al content in the matrix and thus reduced island Mg 17 Al 12 phase. On the one hand, the electrical resistance of acicular Al 4 Ce phase was greater than that of island Mg 17 Al 12 phase. On the other hand, the potential difference between Al 4 Ce phase and Mg matrix was much smaller than that between Mg 17 Al 12 phase and Mg matrix. Therefore, 1.2 wt.% Ce addition leads to minimum corrosion rate. Considering the poor corrosion resistance of AMprocessed Mg alloys without protective layer, surface modification is normally utilized. Li et al indicated the AM-processed WE43 scaffold without surface modification had a corrosion rate of ∼1.5 mm·yr −1 , while the utilization of plasma electrolytic oxidation (PEO) significantly reduced the corrosion rate to <0.1 mm·yr −1 [132].

Corrosion mechanism of Mg scaffold.
Mg and Mg alloy are usually made into scaffold for better biological applications. The corrosion mechanism of the periphery of scaffold is the same as that of bulk mentioned in section S3 in supplementary materials. However, the center of Mg scaffold demonstrates a unique corrosion behavior as shown   in figure 26. Specifically, the periphery of scaffold experiences a uniform corrosion while the center undergoes a localized corrosion [135]. At the initial stage, uniform corrosion predominates. As the corrosion process progresses, the contribution of localized corrosion gradually exceeds that of uniform corrosion. Consequently, the degradation products accumulate between the struts, creating a relatively narrow space, which may lead to limited diffusion of Mg 2+ ions in the center where crevice-like corrosion may occur. Then a build-up of Mg 2+ ions with a concentration gradient is set up between the entrance and the end of the space. Subsequently, the negatively charged Cl − ions migrate into the narrow space under the attraction of the positively charged Mg 2+ ions. Hydrolysis of chlorides lowers the pH and locally destroys the passive layer in the narrow spaces. Meanwhile, the periphery of the scaffold with the passive layer may act as a cathode, establishing a corrosion cell with the Mg material inside the scaffold, further accelerating the corrosion in the center. In addition to the crevice-like corrosion mechanism, the difference in surface roughness of struts between the periphery and the center of scaffold may lead to different corrosion behaviors as well.

Approaches to enhance corrosion resistance and the
underlying mechanisms. Currently, many approaches have been conducted to enhance the corrosion resistances of additively manufactured Mg alloys, and the underlying mechanisms are elaborated, which are summarized as follows:

Grain refinement.
Most investigations point out that grain refinement is one of the main mechanisms for corrosion resistance enhancement. Gu et al [155] and Shuai et al [156] found that the ZK60 alloy with fine grains exhibited a better corrosion resistance than the counterpart with coarse grains. It is well known that grain boundary is one of the major crystallographic defects in Mg and Mg alloys, which acts as a corrosion barrier to inhibit the corrosion propagation. Moreover, high density of grain boundaries can also reduce the internal stress by providing vacancy, resulting in the reduction of cracks in the surface oxide film to protect the remaining Mg substrate from the invasion of soaking fluid. Grain refinement introduces more grain boundaries, and thus is beneficial for corrosion resistance enhancement. Most of the additives currently utilized in additively manufactured Mg alloys have an effect of grain refinement, such as Al [79], Ce [97], GO [100,154], Ti [153], Mn [93,94], Ag [143]. Solid solution treatment normally increases the grain size and leads to inferior corrosion resistance [140]. However, it should also be noted that several researchers believe the deterioration in corrosion resistance of Mg alloys is related to grain size reduction [81,157]. They indicate that the grain boundaries, as crystallographic defects, are more susceptible to be corroded than the grains and thus facilitate corrosion process of Mg alloys. Li et al found that the WAAM-processed AZ31 alloy with finer grains exhibited an inferior corrosion resistance in comparison with the as-cast one with coarser grains [81], although no defects were detected in both samples, which was different from the interpretation by Lovašiová et al [129]. Li et al believed that the variation of corrosion mechanism with grain size is the main reason causing this phenomenon [81]. The corrosion mechanism of the as-cast AZ31 alloy with large grains was dominated by the micro-galvanic coupling between second phases and Mg matrix, while that of the WAAM-processed AZ31 alloy was controlled by intergranular corrosion.

Phase constitute optimization.
It is generally accepted that the second phases in Mg alloys typically deteriorate corrosion resistance, since they usually act as a cathode to accelerate corrosion. For instance, excessive Mn addition can lead to Mn particle formation in additively manufactured Mg alloys, thus leading to high corrosion rate although Mn can refine grain structure [93]. This means that a balance between grain refinement and second phase formation leads to the content of additives not being too high or too low in order to obtain a best corrosion resistance.
In general, there are mainly four methods to optimize phase constitute: (a) reducing second phase content. Tao et al indicated that the GO addition remarkably decreased the MgZn 2 phase content in the LPBF-processed ZK30 alloy, and thereby lowered down the corrosion rate [100]. (b) Promoting the substitution of low-potential second phases for highpotential second phase. Shuai et al found that the Ti addition in the LPBF-processed AZ61 alloy promoted the formation of divorced eutectic α phase that exhibited lower potential difference with the Mg matrix than the Mg 17 Al 12 phase [153]. The substitution of Mg 17 Al 12 phase with eutectic α phase decreased the susceptibility to galvanic corrosion, and was conducive to enhancing corrosion resistance. It is reported that the addition of Al or Nd can also bring about the similar effect, by promoting the formation of low-potential Al 4 Ce [97] or Mg 12 Nd [132] phases, respectively. (c) Induce the formation of dense LPSO structure. Shuai et al indicated that dense LPSO structure lead to the formation of a homogeneous and compact degradation product film by providing substantial sites for nucleation, which isolated the Mg matrix from the corrosive solution and protected the samples from further degradation [95]. (d) Forming a continuous net-like structure. Zhang et al illustrated that the continuously distributed Mg 17 Al 12 phase formed a net-like structure in the WAAMprocessed AZ91 alloy, thereby exerting an inhibitory effect on corrosion process [84]. 4.2.3.3. Generating protective layers. Surface modification, such as micro-arc oxidation and PEO [132], can provide a protective layer before corrosion and thus decrease degradation rate of additively manufactured Mg alloys. Gao et al treated AZ61 powders with 40% hydrofluoric acid solution for 24 h and then conducted LPBF process [121]. A continuous insoluble MgF 2 network structure was generated during LPBF process when using the treated AZ61 powders and the corrosion resistance was improved. In addition, several additives can induce the formation of protective layers during corrosion. Yang et al introduced BG into LPBF-processed ZK60 alloy, and found that a bone-like apatite formed on the surface during corrosion process [72,158]. The deposited apatite possessed a stable and compact structure compared with Mg(OH) 2 , thereby effectively retarding the degradation of Mg matrix.

Smoothening surface.
Compared to machined materials, the surfaces of additively manufactured materials are often rough. Benn et al investigated the effect of surface condition on the corrosion behavior of LPBF-processed WE43 alloy [159]. They found that the surface smoothening was a possible approach to reduce hydrogen release and Figure 26. Schematic illustration of corrosion mechanism at the center of the WE43 scaffold [135]. Reprinted from [135], Copyright (2018), with permission from Elsevier. decrease initial large pit formation, thus resulting in slow degradation process. In addition, phosphoric etching was confirmed to be more effective than machining on obtaining low degradation rate, since phosphoric etching was conducive to generating various protective apatites on the P enriched spots.

Design of open porous structure.
Design of open porous structure is of great importance to improve the corrosion resistance of Mg alloy scaffold. Firstly, process optimization is essential for scaffold fabrication to minimize internal porosity of struts and generate high dimensional accuracy, since these two factors greatly influence the corrosion resistance [160]. The relationship between process parameters and forming quality has been elucidated in detail in section 2. As for Mg scaffold, Liu et al indicated that with the customized energy input and scanning velocity could lead to high relative density of 99.5% in struts, and low geometrical error of <10% between the designed and the fabricated. Secondly, the effects of pore size [161], strut diameter [132] and architecture (biomimetic, diamond, and gyroid) [162] on the corrosion behavior of Mg scaffold have been investigated currently. The results demonstrate that small pores, large strut diameters, and gyroid architecture are conducive to promising corrosion resistance.
It is noteworthy that additives often enhance corrosion resistance through more than one mechanism. For instance, Ti addition can promote grain refinement, reduce undesirable Mg 17 Al 12 phase, and induce the formation of net-like protective layer composed of low-potential eutectic α phase [153]. Mn addition can refine grain structure [93], induce the formation of dense LPSO structure [95], and produce a relatively protective manganese oxide film [94], to decrease the degradation rate.

Biocompatibility.
An undeniable fact is that the released OHand Mg 2+ ions during Mg degradation increase the PH value and osmotic pressure in culture medium, thus inhibiting cell growth [72]. However, Xie et al indicated that the relatively fast Mg 2+ release during the early implantation stage is a potential antibacterial mechanism of Mg-based implant [133], which means the release of Mg 2+ should be controlled to an extent rather than completely suppressed. In addition, excessive hydrogen gas release can interfere with bone healing process, leading to callus formation and cortical defects [50]. It is believed that enhanced corrosion resistance normally leads to low release rates of metallic ions and hydrogen [121], thus alleviating the cytotoxicity [153] and resulting in improved biocompatibility. Nevertheless, not all the additives that can improve corrosion resistance are suitable for biological application. Ag + ion has cytotoxicity since it can destroy active proteins. So although additively manufactured Ag-modified Mg alloy is reported to possess favorable biodegradability and good antibacterial ability [143], it is still inappropriate for medical implant currently. In addition, Ho et al found that the incremental addition of HA (Ca 10 (PO 4 ) 6 OH 2 ) deteriorated the corrosion resistance of FSAM-processed AZ31B alloy [163], while improved the hemocompatibility, biocompatibility and cell adhesion [164].

Micro-wear resistance.
There is only limited literature focused on the micro-wear resistance of additively manufactured Mg alloys. Wang et al investigated the microwear performance of LPBF-processed Cu-introduced [92] and La 2 O 3 -introduced [99] AZ61alloy in both as-built and heattreated states. They believed that hard second phase formation (such as Mg 17 Al 12 , Al 2 Cu, Al 11 La 3 ), grain refinement and increased solid solution atoms resulted in hardness enhancement and accordingly micro-wear resistance improvement. Based on the above mechanisms, the best Cu addition was confirmed to be 2 wt.% since higher Cu content led to grain coarsen and deteriorated micro-wear resistance [92]. In addition, aging heat treatment, which could promote the formation of second phases, was more effective on improving micro-wear performance than solid solution treatment. La 2 O 3 addition was also conducive to enhancing micro-wear resistance, however, its best content was not determined in current investigation [99].

Corrosion fatigue behavior.
Mg alloys are usually subjected to both fatigue and corrosion when used as biomedical implants in human body. Wegner et al [152] and Li et al [83] investigated the corrosion fatigue behaviors of LPBF-processed WE43 bulk and scaffold in SBF, respectively. Wegner et al indicated that the general corrosion influence on the fatigue properties seemed to be dominant here, and the process-induced surface roughness remarkably decreased fatigue strength. As for scaffolds, Li et al found that cyclic loading during fatigue test accelerated the biodegradation process. Localized corrosion is the main degradation mechanism, and the center is suffered from more localized corrosion than the periphery. Nevertheless, most of the fatigue cracks initiated in the struts positioned on the periphery.

Potential application of additively manufactured Mg alloys
Magnesium alloys were originally used in aerospace and defense since World War II. At that time, the flammability, poor fatigue, and creep resistance limit the wide application of Mg alloys. With the development of Mg-RE series alloys and the gradually increasing in-deep understanding on the relationship between processing, microstructure and performances, the application of Mg alloys is further expanded, and more and more researchers show their increasing interest in Mg alloys, especially under the global trend of energy conservation, emission reduction and green manufacturing. To date, Mg alloy components are currently used in many fields including aerospace, weapons, national defense, automobile and biomedical [165][166][167][168][169].
Currently, there is no report about the practical application of additively manufactured Mg alloys in industry, which may be ascribed to the following three reasons. From the viewpoint of science, there is still a lack in the systematic understanding on process-microstructure-performance relation. From the viewpoint of engineering, current AM equipment/system is not sufficient for the fabrication of Mg alloy. Moreover, the intrinsic processing characteristics of AM technology makes it difficult to fulfill the requirements of mass production in several fields, such as automobile and 3C products, although some research institutes have printed related parts via AM. For instance, Fraunhofer Institute for Laser Technology (Fraunhofer ILT) prepared a topology-optimized motorcycle triple clamp in 2016 [170] and TCT Korea manufactured earphone housing made of AZ91 Mg alloy [171] in 2018 as shown in figures 27(a) and (b).
According to the current development trend, we believe that the most likely potential application of additively manufactured Mg alloys in future is in bio-medical and aerospace, since the products in these fields normally have high added value, demanding less on cost control requirement. In the bio-medical field, LPBF technique has outstanding advantages on preparing components with extremely complex structures over other AM techniques because of its small spot and high forming accuracy. Fraunhofer ILT used LPBF to fabricate degradable porous Mg scaffold with the dimension of 10 mm × 10 mm × 7.5 mm and a structure thickness of only 0.4 mm as shown in figure 27(c). Recently, several other institutes or universities, such as Shanghai Jiao Tong University [172], Carlos III University of Madrid [173], Delft University of Technology [174] and University of Shanghai for Science and Technology [175], also reported that Mg alloy based implants were successfully fabricated via LPBF in the laboratory. It is believed that additively manufactured Mg alloy implants will usher in the dawn of clinical application in the visible future with its unique advantages. Compared with LPBF technique, WAAM, wirebased LDED and FSAM have advantages in deposition rate (or deposition efficiency) and are suitable for the preparation of large-scale components in the field of aerospace, such as cockpit instrument panel (made of AZ31B), service door inner panel (made of AZ31B or ZK10) and rudder pedal (made of AZ80A) [167]. Figure 27(d) is a WAAM-processed large-scale Mg alloy component by our group. Although this is not a practical component, its successful fabrication still verifies the feasibility of WAAM process on the manufacturing of large-scale special-shaped structural components. In addition, with the emergence and continuous progress of multi-laser PBF technique, its deposition rate may reach or exceed other AM process [176], accelerating the application of LPBF in the manufacturing of large-scale Mg alloy components. Reproduced with permission from [170]. © Fraunhofer ILT, Aachen. (b) LPBF-processed earphone housing by TCT Korea. Reproduced with permission from [171]. (c) A LPBF-processed degradable porous Mg scaffold by Fraunhofer ILT. Reproduced with permission from [170]. © Fraunhofer ILT, Aachen. (d) A WAAM-processed large-scale Mg alloy component by our group.

Summary
This work reviews the current state of additively manufactured Mg and its alloys in regard of defect formation, microstructural characteristics and evolution, and performances. The process-formability-microstructure-performance relationship is discussed, and the underlying mechanisms are analyzed. The main conclusions are drawn in this review as follow: (1) Mg and Zn (if possible) are two main vaporization elements in the AM process of Mg alloys. Considering that the weight percentage of Mg element is extremely higher than that of Zn element in Zn-containing Mg alloys, the vaporization of the latter should be more intense than that of the former, resulting in an increased Mg/Zn ratio. The element vaporization during AM process is unavoidable, but can be greatly reduced through optimizing process parameters. (2) Gas pore, LoF, crack and surface balling are normally observed in AM-processed Mg alloys. The formation of gas pores is related to powder characteristics (such as hollow powders), gaps between adjacent powders, element vaporization and keyhole collapse. The appearance of LoF is ascribed to insufficient energy input and inadequate overlapping. Both hot cracks and cold cracks are attributed to high residual stress caused by rapid heating and cooling during AM process. In addition, the appearance of surface balling is normally induced by droplet/powder splashing at high energy input and interfacial tension at low energy input. Appropriate process parameters can eliminate LoF and cracks. It is interesting to note that WAAM process is more conducive to generating dense Mg alloys than LPBF process. (3) Equiaxed grain-structure is the most observed microstructure in AM-processed Mg alloys, although columnar grains are also reported. However, the underlying mechanisms for the formation of equiaxed grains are still unclear. The phase constituents of AM-processed Mg alloys are very complex, and vary with alloying elements, heat treatment processes, building height and process parameters. According to the currently published investigations, Mg 17 Al 12 is the most common secondary phase in Mg-Al series alloys. As for Mg-Zn series alloys, Mg 7 Zn 3 and MgZn 2 phases are widely reported, although the dominate secondary phase is still controversial in different investigations. However, the phase constituents of Mg-RE series alloys still remain unclear. For instance, nine different phases have been reported in the AM-processed Mg-Y-Nd-Zr alloys. (4) WAAM-processed Mg alloys generally exhibit higher strength and lower ductility than LPBF-processed Mg alloys. As can be seen from the statistical image of tensile results, strength-ductility tradeoff exists. In addition, anisotropy in mechanical properties is observed, possibly due to the inhomogeneous microstructure along the building direction and the possible formation of columnar grains. The strengthening and ductilization mechanisms for AM-processed Mg alloys are almost the same as those for traditional Mg alloys. The minor differences in strengthening and ductilization mechanisms between AMprocessed Mg alloys and traditional Mg alloys are reflected by the smaller grain size and higher dislocation content in the former, resulting in more significant grain boundary strengthening and strain strengthening effect. (5) The relations of the corrosion rates of as-cast and AMprocessed Mg alloys are still uncertain, which is ascribed to the possible appearance of defects such as pores and LoF. The corrosion mechanism of AM-processed Mg alloys is similar to that of traditional Mg alloys. However, the effect of grain size on the corrosion resistance is still controversial. The corrosion resistance of AM-processed Mg alloys (including bulk and scaffolds) is currently difficult to fulfill the requirements of practical application. Phase constituents optimization, and surface modification are essential to improve the corrosion resistance and biocompatibility.

Perspectives
AM of Mg alloys attracts increasing attentions currently. To achieve the industrial applications of additively manufactured Mg alloys, a full chain development from raw materials to equipment and process should be emphasized in the future as shown in figure 28, which requires a joint effect of scientists and engineers from both the academia and industry.
(1) Preparation of powders and wires It is well known that the preparation of Mg alloy powders is highly dangerous due to their low melting point and boiling point. Currently, the enterprises (such as Tangshan Weihao Magnesium Powder Co. Ltd) normally adopt rotating-disk centrifuge atomization technology, rather than the mainstream gas atomization technology (which is one of the main preparation methods of Ti alloys, Ni-based alloys, Fe-based alloys, etc), to prepare Mg alloy powders. The oxygen content can be controlled by adding inert gas during the rotating-disk centrifuge atomization process, so the manufacturing process is relatively safe. However, the subsequent powder sifting and equipment cleaning are often carried out in the atmosphere, presenting high safety risks. Therefore, it is important to realize the whole preparation process (including manufacturing, sifting, cleaning, and storage) of Mg alloy powders in a protective atmosphere. In addition, it is also essential to ensure the stability and consistency of chemical composition, sphericity, particle size distribution, fluidity, apparent density and other indicators of different batches of powders. Compared to powders, the preparation of Mg alloy wires is relatively safe, which is normally achieved by drawing or extrusion process. However, the hexagonal close-packed structure of Mg alloy makes it difficult to undergo plastic deformation. The preparation of wire products is commonly associated with issues such as rough surfaces, cracking susceptibility, and blockage of extrusion channel. Our group has developed reciprocating extrusion process for Mg alloy wire preparation. Through reciprocating hot extrusion of cast ingot, Mg alloy (for instance, AZ31, ZK60, Mg-Gd-Y-Zn) and Mg matrix composites (MMCs) (carbon nanotubes, SiC, etc) wires are successfully prepared with a diameter of 0.8-1.5 mm and more than 5 kg per plate. In the future, the preparation of Mg alloy wires with a wider diameter range, greater single-wire length and more types is an important research direction. In addition, it is also necessary to ensure the uniformity of chemical composition and size for different batches of wire.
(2) AM equipment Mg and its alloys are easily vaporized and oxidized during AM process. It is crucial to ensure a low oxygen level in the chamber in the AM process of Mg alloys, which requires the design and optimization of nozzles, the development of in-situ monitoring systems, and the construction of a fully enclosed chamber with controllable oxygen content. Currently, most research on AM of Mg and its alloys is concentrated on LPBF and WAAM processes. In the future, as other AM processes (such as binder jetting, LDED) for Mg alloys receive increasing attention, related equipment will undergo improvements accordingly. Specifically, FSAM process, as a solid-state fabrication technique, has significant advantages in preparing large-sized and dense Mg alloy parts due to high deposition rate and small grain size that is comparable to LPBF-processed counterparts. FSAM of Mg alloys may gradually become a popular direction. However, the development of FSAM equipment is still in the preliminary stage. In addition, the  [45,64,122,[182][183][184][185]. Reprinted from [45], Copyright (2022), with permission from Elsevier. Reprinted from [64], Copyright (2021), with permission from Elsevier. Reprinted from [122], Copyright (2022), with permission from Elsevier. Reprinted from [182], Copyright (2021), with permission from Elsevier. Reprinted from [183], Copyright (2023), with permission from Elsevier. Reprinted from [184], Copyright (2023), with permission from Elsevier. Reprinted from [185], Copyright (2021), with permission from Elsevier. design and development of integrated equipment for additive, equivalent and subtractive manufacturing, multi-energy field assisted AM equipment, and multi-laser AM equipment are also expected to be hotspots.
(3) Underlying mechanisms of process-formabilitymicrostructure-performance relation The morphology, size and orientation of grains play a vital role in performance of AM-processed Mg alloys. Unfortunately, the current understanding of grain formation mechanism is insufficient as mentioned in section 3.1.2. So, in the future investigations, in-situ monitoring of melt pool, simulation of temperature field and flow field, as well as solidification theory should be combined to reveal the in-depth mechanisms.
The phase constituents of additively manufactured Mg alloys are still controversial, especially for Mg-RE series, as mention in section 3.2. In addition, the utilization of post heat treatment processes makes the phase constituents of additively manufactured Mg alloys more diverse and less predictable. Therefore, the thermodynamic and kinetic calculation database of Mg alloys should be established first in the future. Subsequently, advanced microstructure characterization methods, such as three-dimensional atom probe tomography, should be used to identify the phase constituents, and further analyze the effect of alloying elements, especially RE element. We believe that AM of Mg-RE series alloys containing LPSO structure will be one of the future research hotspots.
AM process has advantages in manufacturing porous scaffolds that is important for human implants. Therefore, more effort should be made to explore various Mg alloy porous scaffolds with different architecture (such as octettruss and auxetic [177]), different pore sizes, different strut diameters and different volumes [178], even functionally graded lattice structure [177], to achieve optimization and better performance. In addition, the optimization of scaffolds may bring about new corrosion mechanisms, thus deepening the understanding of the corrosion behavior of Mg alloys. Meanwhile, more biologically appropriate environment should be established [179], and performances in a situation more consistent with actual application environment (such as corrosion fatigue) should be assessed, to better evaluate Mg alloy scaffold degradation in vivo. Moreover, the effects of RE elements on corrosion behavior and biocompatibility of Mg alloys remain unclear.
(4) AM of heterostructured Mg alloys Recently, the design and fabrication of heterogeneous structures gradually become popular in the field of materials science and engineering [180]. One the one hand, heterostructured materials can achieve excellent properties that are difficult for traditional homogeneous materials to realize. On the other hand, the introduction of heterogeneous structures can realize linear or spatial change of material composition or microstructure, then bringing gradual variation of mechanical and functional properties to meet practical application requirements that different regions of a particular structural part may work in different conditions. The complex microstructure characteristics of Mg alloys, especially Mg-RE series alloys, combined with the in-situ controllable thermal history during AM process, are conducive to the design and preparation of heterogeneous structures, and bring more possibilities for the improvement of performances. However, the current research on AM of heterostructured Mg alloys is still limited.
(5) AM of MMCs Adding reinforcing particles into Mg alloys to form MMCs is an effective approach to improve the properties. The current utilized reinforcing particles are usually ceramic particles. The interface between ceramic particle and Mg matrix is normally considered as the weak region, resulting in premature failure. Thus, the composition of MMCs should be carefully designed. Multi-principal element alloy (MPEA, also known as high-entropy or medium-entropy alloy) particle may be a better choice than ceramic particle, since the interface between MPEA particle and Mg matrix may be another kind of MPEA rather than intermetallic. In addition, the influences of reinforcing particles on the solidification process, microstructure evolution and performances of MMCs are still not deeply understood. Zhang et al conducted a detailed and comprehensive review on AM of MMCs. Therefore, it will not be discussed here. Please refer to [168] for more information.
(6) Customized Mg alloys for AM The currently studied Mg alloys are usually developed based on the understanding on the features of casting or plastic deformation process. Such kind of Mg alloys may not suitable for AM process, since they cannot take full advantage of the process characteristics of AM, such as the in-situ heat treatment effect caused by reciprocating heat cycle and heat accumulation. Therefore, customized Mg alloys specifically for AM process should be highlighted. A machine learning assisted high-throughput alloy screening process can be utilized to achieve novel Mg alloy design [181,182].