3D thixo-printing : a novel approach for additive manufacturing of biodegradable Mg-Zn alloys

(1)

UNIVERSIDADE ESTADUAL DE CAMPINAS

SISTEMA DE BIBLIOTECAS DA UNICAMP

REPOSITÓRIO DA PRODUÇÃO CIENTIFICA E INTELECTUAL DA UNICAMP

Versão do arquivo anexado / Version of attached file:

Versão do Editor / Published Version

Mais informações no site da editora / Further information on publisher's website:

https://www.sciencedirect.com/science/article/pii/S0264127520306961

DOI: 10.1016/j.matdes.2020.109161

Direitos autorais / Publisher's copyright statement:

©2020

by Elsevier. All rights reserved.

DIRETORIA DE TRATAMENTO DA INFORMAÇÃO

Cidade Universitária Zeferino Vaz Barão Geraldo

CEP 13083-970 – Campinas SP

Fone: (19) 3521-6493

http://www.repositorio.unicamp.br

(2)

3D thixo-printing: A novel approach for additive manufacturing

of biodegradable Mg-Zn alloys

Dalton Daniel Lima, Kaio Niitsu Campo, Sergio Tonini Button, Rubens Caram

⁎

School of Mechanical Engineering, University of Campinas (UNICAMP), Campinas, SP 13083-860, Brazil

H I G H L I G H T S

• Investigation of an additive manufactur-ing method for semi-solid Mg-Zn alloys (Mg-xZn, x = 30 - 45 wt. %).

• Simulations and thermal analysis sug-gested that Mg-38Zn is a promising composition for semi-solid applications. • Successful printing of biodegradable Mg-38Zn was carried out at 420 °C, resulting in ﬁne and globular microstructures. G R A P H I C A L A B S T R A C T

a b s t r a c t

a r t i c l e i n f o

Article history: Received 14 May 2020

Received in revised form 15 September 2020 Accepted 17 September 2020

Available online 18 September 2020

Keywords: Mg-Zn alloys Biomaterial

Biodegradable materials Fused deposition modeling Semi-solid printing 3D Thixo-printing

A simple and widely used additive manufacturing technique for polymeric materials is fusedﬁlament fabrication (FFF). In the semi-solid state, metallic materials may show rheological features comparable to those of polymers. Thus, they can be processed accordingly. The use of biodegradable Mg-based materials is an interesting approach to avoid removal surgeries and release of toxic corrosion products and wear debris. Therefore, in this study, the FFF technique was applied using a biodegradable Mg-Zn alloy in the semi-solid state. Some preliminary compo-sitions were investigated through thermodynamic simulations to verify their compatibility with the process. Among them, Mg-38Zn was selected to be experimentally evaluated. Metallicﬁlaments were produced via hot extrusion, which also aided in obtaining a globular microstructure in the semi-solid state. FFF was performed at 420 °C without any obstruction at the nozzle channel, which allowed the production of sound parts with ac-ceptable welding between the deposited layers. This indicated that this technique (termed as“3D thixo-printing”) provides a promising additive manufacturing route to produce biodegradable Mg-based implants. © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://

creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Orthopedic implants made of conventional metallic biomaterials, such as stainless steels and Ti- and Cr-Co-based alloys, have limitations as they can release corrosion products and wear debris which can result in severe inﬂammatory processes [1,2]. Furthermore, their elastic mod-ulus has also been found to be much higher than the implanted bones, which can result in stress shielding phenomenon [3]. In such a case,

mechanical stress distribution can be altered by the device, thus, limit-ing the stress transfer to the livlimit-ing bone tissue. Over time, loadlimit-ing trans-fer from the bone to the implant has been found to cause bone remodeling with a consequent bone loss (resorption process), which has been observed to decrease the bone strength and, eventually, results in bone fracture. Because of these issues, once the bone is healed, it has been considered important to remove the implanted device, a proce-dure that includes painful and costly surgery [4].

Ideally, to completely avoid bone resorption and removal surgery, the orthopedic implant should be made of a biocompatible material, having proper mechanical behavior and suitable host response. Such

⁎ Corresponding author.

E-mail address:[email protected](R. Caram).

https://doi.org/10.1016/j.matdes.2020.109161

Contents lists available atScienceDirect

Materials and Design

(3)

an implant would progressively get consumed through corrosion in the body environment, without releasing any toxic products and would get totally dissolved after the bone gets properly healed [5,6]. Biodegrad-able materials, such as those based on Mg alloys, can be used to address this issue.

The use of Mg in biodegradable devices is not entirely new in medi-cine. It wasfirst attempted in the first half of the nineteenth century in England, soon after the pure metal became available. In 1878, E.C. Huse, an American surgeon, reported that using Mg wires as ligatures could prevent bleeding after surgeries, mainly because after a certain period, Mg gets oxidized and naturally absorbed by the human body [7]. It is important to note that Mg plays a crucial role in the formation of essen-tial compounds that are required in many metabolic functions of the human body [8]. Thus, Mg dissolution is not expected to be problematic [9,10]. According to Witte [11], since the experiments conducted by Huse, Mg-based materials have been applied in several clinical applica-tions, including cardiovascular, musculoskeletal, and general surgery applications. Therefore, many patients have benefited from the use of Mg implants. In the last few years, interest in Mg-based materials in or-thopedics has increased considerably, mainly considering Mg-Zn-based alloys. According to Zhang et al. [12], besides being a vital element for the human body, Zn can also enhance mechanical behavior and control the corrosion properties, such as dissolution rate of Mg-based alloys.

Recently, development of additive manufacturing (AM) technology has been transforming the manufacturing of orthopedic implants. Cus-tomized devices with complex geometry, which were impracticable be-fore the advent of this technology, have become feasible. Nonetheless, AM techniques for metallic materials usually depend on the use of ex-pensive lasers or electron beam devices, which require sophisticated and expensive technological apparatus. In addition, the feedstock is typ-ically based on metallic powders with very speciﬁc characteristics, which can only be obtained through complex processing techniques, making theﬁnal product not entirely accessible [13,14].

A simpler and widely used AM technique is fusedfilament fabrica-tion (FFF). In this printing method, a continuousfilament is heated and extruded through a moving print head and then deposited layer-by-layer based on a computer design to produce 3D parts [15]. This technique has been primarily limited to polymers, but an innovative ap-proach has recently emerged for efficient processing of metallic mate-rials. In the semi-solid state, if the remaining solid phase exhibits a globular morphology, metallic materials can exhibit thixotropy, which refers to rheological features comparable to those of thermoplastic poly-mers. Thus, metallic materials can be used as polymers in FFF. Thixot-ropy is a time-dependent phenomenon in which the viscosity of a semi-solid material decreases when it is subjected to shear stress [16]. On one hand, even if this material is partially liquid, its viscosity is rela-tively high while resting, and it can sustain its shape if the amount of liq-uid is not excessive. On the other hand, under moderate shear stress, the apparent viscosity decreases, thus allowing controllable extrusion through a heated nozzle and layer-by-layer deposition, which are the main processing steps in FFF.

The application of semi-solid alloys in AM has been discussed in some studies recently. One of theﬁrst studies regarding was conducted by Rice et al. [17]. They developed a rheocaster, an apparatus where the metallic alloy is completely melted, then cooled to a partially solidiﬁed state, and subsequently deposited on a moving substrate. In 2002, Finke and Feenstra investigated the feasibility of a new process based on the FFF technique [18]. They discussed the relationship between the microstructure and processing parameters and showed that the ex-trusion and deposition processes strongly depend on the rheological be-havior of the semi-solid alloy. In another study, Mireles et al. evaluated the FFF technique applied to low melting temperature metallic alloys, mostly of the Bi-Sn system, to produce electronic interconnects [19]. They presented a successful deposition of single-layer solder lines with varied thickness, including sharp 90° angles and smooth curved lines. In 2017, Chen et al. also described another technique to process

semi-solid alloys [20]. To demonstrate the feasibility of their approach, they processed binary Bi-Sn alloy. Recently, Jabbari and Abrinia also re-ported the processing of low melting temperature Sn-Pb alloys [21,22]. A review of relevant literature revealed that the application of FFF technique to metallic materials has always been restricted to very low melting temperature alloys, mostly Pb- and Sn-based, with processing temperatures of approximately 220 °C. It was also found that the pro-cessing of Mg-based materials using FFF has not been reported, al-though it could provide a simple and efﬁcient method to produce biodegradable implants with complex geometries. Therefore, this study aimed to demonstrate the application of the FFF technique to bio-degradable Mg-Zn alloy. Initially, a suitable composition that could ad-dress the process features was selected. Thereafter,ﬁlaments with proper microstructures were prepared by hot extrusion, which were subsequently printed at a processing temperature of 420 °C.

2. Experimental procedures

The composition of the Mg-Zn system to be used in the FFF tech-nique was selected based on the criteria typically applied for thixoforming processes, which primarily deal with the aspects of the evolution of the liquid fraction as a function of temperature [23–28]. These criteria were considered and PANDAT™ software was employed to simulate the liquid-solid transformation under equilibrium and non-equilibrium (Scheil) conditions using a thermodynamic database collected/optimized by Povoden-Karadeniz [29].

Thermodynamic simulation results suggested that biodegradable Mg-38Zn alloy could be a suitable candidate (all compositions in this study were given in wt%). Therefore, cylindrical ingots of 20 mm diam-eter of this composition were prepared by melting and mixing pure el-ements for 2 h at 750 °C in an induction heating furnace having a sealed steel crucible, under a positive pressure atmosphere of high-purity argon. Chemical homogeneity was improved by mechanical stirring every 30 min in the molten alloy. The chemical composition was con-ﬁrmed using energy dispersive X-ray ﬂuorescence spectroscopy in a Shimadzu EDX7000 analyzer.

Samples weighing 50 mg were extracted from the produced ingots for differential scanning calorimetry (DSC) tests, which were performed in a Netzsch STA 409 thermoanalyzer. The DSC experiments were car-ried out in an Al2O3crucible under a dynamic argon atmosphere and

at heating/cooling rates of 10 °C/min. The simulated solid-to-liquid transition was compared with the DSC test results, and it was found that some processing temperatures could be selected and exploited.

Metallicfilaments were prepared using a two-step extrusion process carried out in a hydraulic press at 330 °C. First, a cylindrical ingot of 20 mm diameter was extruded to 5 mm diameter bars (area reduction of 16:1) using a maximum load of 220 kN. Afterwards, the bars were further deformed to obtain 1.75 mm diameter wires (area reduction of 8:1) using a maximum load of 20 kN. Besides providing the proper way to obtain thefilaments for the FFF, the hot extrusion process was also intended to prepare a suitable thixotropic material because the de-formation process was able to refine the dendritic microstructure, which resulted from the solidification process, thus, allowing to obtain solid particles with globular morphology in the semi-solid state [30]. To analyze the globularization behavior of the prepared feedstock, sam-ples were extracted from thefilaments and isothermally heat-treated in an induction furnace under an inert atmosphere at 420 °C for 10, 300, and 600 s, followed by water quenching.

To enable the use of the metallic material in the FFF technique, an in-novative print head was designed to print median-melting-temperature alloys to be able to work at up to 700 °C. 3D thixo-printing was performed in a machine developed in-house, which was similar to a Prusa FFF printer, as can be seen in theFig. 1a. A comparison between the original components and those developed for this work can be seen inFig. 1b and c. The Mg-based material was processed at 420 °C, with a printing speed of 5.0 mm/s, nozzle diameter of 1.0 mm,

D.D. Lima, K.N. Campo, S.T. Button et al. Materials and Design 196 (2020) 109161

(4)

layer thickness of 1.0 mm, nozzle to substrate distance of 1.0 mm, and deposited on a 1200 grit sandpaper (substrate).

Microstructural characterization was performed during all steps of this study. Metallography preparation for this included sanding with 200, 400, 800, and 1200 grit sandpapers and polishing with 9, 6, 3, 1, and 0.25μm diamond cloths. Microstructural features of non-etched samples were observed through scanning electron microscopy (SEM) with a Zeiss EVO MA15 microscope using backscattered electron imag-ing. Quantitative image analyses were carried out using the ImageJ soft-ware [31], which was applied to measure the liquid fraction, globule size, and shape factor of the solid particles in the semi-solid condition. For this, Eq.(1)was applied to determine the average diameter (equiv-alent diameter): D¼ 2 ffiffiffi A π r ð1Þ where A is the measured area. Furthermore, Eq.(2)was employed to determine the shape factor (circularity):

C¼ 4πA

P2 ð2Þ

where P is the perimeter. 3. Results and discussion

For semi-solid FFF processing, the liquid fraction should ensure con-tinuous extrusionﬂow to avoid nozzle obstructions and discontinuities. While working at low liquid fractions requires higher extrusion forces, making the process complex or eventually, unfeasible, the use of exces-sive liquid fractions may decrease the viscosity to a point where forma-tion of drops in the nozzle exit channel can be observed. Drops were found to be undesirable because their size and deposition could not be appropriately controlled, which decreased the process resolution. Thus, it is crucial to select a suitable liquid fraction range to simulta-neously address both issues. Considering these limitations and based on the results of preliminary experiments and available literature about the topic [18,21,22], it was found that sound printed parts can

usually be obtained by employing liquid fractions between 0.5 and 0.8. Furthermore, it is also desirable that this range could be achieved at lower temperatures to reduce oxidation, which can affect the adhesion between the deposited layers.

Because of the crucial role of liquid fraction and the existence of tem-peraturefluctuations or variability during the process, it is also impor-tant that the liquid fraction does not significantly change with temperature. This implies that the alloy must exhibit a low liquid frac-tion sensitivity at the target temperature. Here, the maximum value for this parameter was assumed to be 0.015 °C−1, i.e., at a temperature variation of 10 °C, the liquid fraction would have changed by 15%. This specific value was chosen because it has been commonly employed when dealing with conventional semi-solid processing [26–28]. Fur-thermore, to decrease the possibility of hot tear formation, solidi_fication temperature range was restricted to less than 150 °C [27]. Therefore, the main criteria that guided the selection of a promising candidate for printing was based on liquid fraction sensitivity at the desired liquid fraction range and the size of the solidification interval.

Mg-Zn phase diagram, shown inFig. 2[29], indicates that a mixture of liquid and solid phases is found above 340 °C in hypoeutectic compo-sitions ranging between 7.5 and 51.7% of Zn. Significantly low Zn con-tents were not considered appropriate here because the intended liquid fraction range is only achieved at temperatures significantly higher than the eutectic temperature (340 °C). Furthermore, all alloys with compositions between 7.5 and 30% of Zn present a solidification temperature range higher than 150 °C, which is the highest value allowed to avoid the possibility of hot tearing. Also, it is not possible to employ near-eutectic compositions, specifically above 45% of Zn, be-cause they exhibit a liquid fraction higher than the established upper limit of 0.80 just after eutectic melting. Therefore, compositions ranging between 30 and 45% Zn were found to be good preliminary candidates and were further evaluated.

Fig. 3a presents the results of the non-equilibrium thermodynamic simulation (Scheil model) of liquid fraction versus temperature for some compositions in the selected range. The non-equilibrium condi-tion was used because the samples used in this study were not homog-enized after casting. Nonetheless, as shown later, the differences related to the equilibrium condition were found to be minimal. As shown in

Fig. 3a, all the highlighted alloys exhibited the same melting/

(5)

solidiﬁcation behavior (i.e., curve proﬁle). The curves showing solid-to-liquid transition exhibit a knee that is associated with the eutectic reac-tion. Furthermore, as the amount of Zn increases, the liquid fraction at which the knee is formed also increases. It should be noted that during melting, after reverse eutectic transformation, the primary phase melts very slowly with increasing temperature, which is considered to be a

good feature since it promotes very low values for liquid fraction sensi-tivity. This behavior was more effectively described through the deriva-tive of the curves presented inFig. 3a, which is considered ef_{ﬁcient to} provide information about the sensitivity parameter. The results of this procedure, as presented inFig. 3b, indicates that all evaluated com-positions present sensitivity values between 0.001 and 0.008 °C−1for primary phase melting, which implies that even under critical condi-tions, for higher sensitivity values, a variation of 1 °C results in a minimal liquid fraction variation of approximately 0.8%, which is lower than the established limit of 0.015 °C−1. According to the thermodynamic simu-lations, all investigated compositions were considered to be good candi-dates for processing using the FFF technique. Therefore, the intermediary candidate (Mg-38Zn) was selected for experimental examination.

Fig. 4shows the DSC results of the heating and cooling cycles for Mg-38Zn alloy. Evaluation of the heating curve revealed the occurrence of three thermal events due to phase transformations. Event 1, which is highlighted in theﬁgure inset ofFig. 4, corresponds to a weak peak caused by a low energy solid-solid phase transformation that occurs near the melting onset, as seen inFig. 2. Furthermore, event 2 is related to the melting of the eutectic formed by the Mg-rich and Mg7Zn3phases

and event 3 corresponds to the primary phase (Mg-rich phase) melting. On cooling, these same reactions were observed to reversibly occur. It should be noted that a peak associated with event 1 was not detected during the cooling scan, however, this was not considered relevant due to the scope of this study.Table 1summarizes the solidus and liquidus temperatures obtained from the DSC measurements, which were also compared with the simulated results. The onset temperatures (solidus on heating and liquidus on cooling) are mostly similar to the simulated temperatures, whereas a larger difference is observed in the offset temperatures (liquidus on heating and solidus on cooling). This was due to presence of thermal inertia effect during the DSC measure-ments, which induced the reactions to occur in a wider temperature

Fig. 3. (a) Liquid fraction (Lf) versus temperature and (b) liquid fraction sensitivity versus temperature curves for different Mg-Zn alloys.

Fig. 4. DSC curves of Mg-38Zn alloy obtained for heating and cooling cycles. Fig. 2. Mg-Zn phase diagram obtained from the thermodynamic simulation using the

database of Povoden-Karadeniz [29].

Table 1

Solidus and liquidus temperatures determined from simulation and DSC measurements for Mg-38Zn alloy.

Type of analysis Temperature (°C)

Solidus Liquidus

Simulation 340 478

DSC– heating 347 500

DSC– cooling 317 473

(6)

range. Thus, the solidiﬁcation range appeared to be wider in the exper-imental results compared to the simulated results. It was determined to be 138, 153, and 156 °C for simulation, DCS on the heating cycle, and DSC on the cooling cycle, respectively.

On the heating cycle, as the liquid was formed, energy was absorbed. Consequently, endothermic peaks correspond to liquid formation, and the evolution of the experimental liquid fraction as a function of the temperature can be obtained by integration of the DSC curve [23], as shown inFig. 5, which also shows two simulated curves. Theﬁrst simu-lated curve corresponds to a process in equilibrium and the second is as-sociated with non-equilibrium (Scheil model). The use of heating cycles is more convenient because no superheating is expected, whereas cooling cycles usually exhibit some degree of supercooling. Analysis of the liquid fraction curve obtained by DSC reveals that eutectic melting occurred in a small temperature range. As pointed out earlier, this is caused by thermal inertia because temperature acquisition was re-stricted to conduction through the ceramic crucible in the absence of di-rect contact between the thermocouple and the sample. This delayed the acquisition of energy variation by the equipment. Thus, the knee re-gion of the experimental curve is not as sharp as it is in the simulated ones. Regarding the simulated curves, no signiﬁcant difference is ob-served between the equilibrium and non-equilibrium conditions. The

difference corresponds to approximately 0.01 at 340 °C and decreases as the temperature increases. A comparison between simulated and DSC results reveals some differences within the intended liquid frac-tions (between 0.5 and 0.8). In the theoretical analyses, the working temperature range is approximately 90 °C, whereas the experimental range is approximately 75 °C. Despite this discrepancy, both results are considered suitable for semi-solid processing as the average liquid

Fig. 5. Liquid fraction (Lf) versus temperature determined by the integration of DSC heating curve compared to the simulated ones.

Fig. 6. Backscattered electron image (SEM) of the typical microstructure of Mg-38Zn alloy after extrusion at 330 °C.

Fig. 7. Backscattered electron images (SEM) of the typical microstructure of the Mg-38Zn alloy after treatment in the semi-solid state at 420 °C for (a) 10, (b) 300, and (c) 600 s, followed by water quenching.

(7)

fraction sensitivity in this range (considering a straight line) is approx-imately 0.003 and 0.004 °C−1, respectively.

To ensure that the liquid fraction should range between 0.5 and 0.8, preliminary FFF tests were carried out at temperatures where both simulated and experimental results indicated the liquid

fraction values to be within the intended range. Accordingly, the tests were performed at 380, 400, and 420 °C (Fig. 5). However, it was observed that printing could be done without the obstruction at the nozzle channel only at the highest temperature, thus, produc-ing samples with apparently reasonable weldproduc-ing between the

Fig. 8. (a) Deposited multilayer of the Mg-38Zn alloy. (b) Typical microstructure of a Mg-38Znﬁlament printed by semi-solid FFF, (c), (d), (e) and (f) examples of printed shapes by the semi-solid FFF, (c) and (e) are the top views of (d) and (f), respectively.

(8)

printed layers. This implied that at 420 °C, the semi-solid material exhibited the best rheological behavior for the printing process. It has been well-established that the semi-solid rheological character-istics depend on temperature, shear rate, time, and solid phase mor-phology [32]. Regarding the morphology, a globular solid phase is considered fundamentally important and it must be obtained. Dur-ing theﬁlament preparation step, the dendritic microstructure was broken-up by the deformation process. A typical microstructure of the extrudedﬁlament is shown inFig. 6. This backscattered electron image (SEM) clearly shows an oriented microstructure with de-formed grains aligned parallel to the extrusion direction. Further-more, microstructural defects that may negatively affect the printing process, such as pores and cracks, are not observed.

For transforming the hot-worked microstructure into a globular mi-crostructure, samples of thesefilaments were reheated to the semi-solid state and isothermally heat-treated at 420 °C for 10, 300, and 600 s, followed by water quenching. The period of 10 s was selected to simu-late the extrusion process through the heated nozzle during the real printing procedure, while treatments with longer exposures were car-ried out to detect microstructural coarsening. Fig. 7 presents the resulting microstructures in the semi-solid state after the treatments. The images show dark gray globules surrounded by a lighter matrix cor-responding to the transformed solid and liquid phases, respectively. As expected, the globular solid phase coarsened when exposed to the high temperature. Furthermore, they became seemingly more globular, although the shortest treatment already provided a suitable microstruc-ture. This feature is very important because the need of longer isother-mal treatments for obtaining a globular solid phase would have impaired the continuity of the printing procedure. The recrystallization process that takes place during the reheating stage formed near-equiaxial grains, which are expected to rapidly globularize. It should also be noted that there are many small particles dispersed within the matrix. They are associated with the primary phase, but were liquid in the semi-solid state, i.e., they originated from a process similar to the secondary solidification process that takes place in rheo-casting. Such secondary particles are commonly observed in samples treated at tem-peratures significantly higher than the knee of the liquid fraction versus the temperature curve. This implies that a significant amount of primary phase also melted during the partial melting process. During solidi fica-tion, they tend to nucleate at the existing primary phase, thus, causing their growth. However, because of rapid quenching procedure, suf fi-cient time was not available for extensive occurrence of epitaxial nucle-ation, thus, many particles were observed to nucleate within the liquid, forming these small particles. These secondary particles are relatively small because of the extensive nucleation rate and short time for growth.

Fig. 8depicts the macro- and microscopic features of the 3D-printed samples processed at 420 °C.Fig. 8a shows the image of the resulting macrostructure, demonstrating the deposited layers with the expected geometric characteristics and adhesion. Fig. 8b shows the typical image of the microstructure of the printed material. It resembles the globular microstructures shown inFig. 7; however, it should be noted that no secondary particles were found because of the slower solidi fica-tion rate, which allowed effective diffusion, and thus, epitaxial nucle-ation in the existing solid particles. Furthermore, it should be noted that the solid globules are significantly smaller than the globules ob-tained from isothermal heat treatments (Fig. 7). This was due to the shearing of the material during printing, which promoted microstruc-tural refinement, thus facilitating the printing process.Fig. 8c to f show the examples of geometries produced by 3D printing, demonstrat-ing the versatility of applications and freedom to develop intricate shapes.

Results of quantitative evaluation of liquid fraction, mean particle size, and circularity are shown inFig. 9. The data obtained were used to compare the results from static heat treatments and semi-solid print-ing. For the following evaluations, secondary particles were evidently

taken as part of the transformed liquid. For this, they were deleted dur-ing image processdur-ing.Fig. 9a shows the results of the mean particle di-ameter. During isothermal heat treatment, the average diameter increased from 10.4 ± 0.4 (10 s) to 24.2 ± 0.4μm (600 s), which was observed to decrease the total solid-liquid interfacial area. On the other hand, the average particle size of the printed samples was approx-imately 5μm.Fig. 9b presents the shape factor results. The highest circu-larity value of approximately 0.91 was found in the printed condition. This was due to the smaller sizes of the remaining solid particles and en-hanced diffusion caused by the applied shear rate, where both

Fig. 9. Quantitative results obtained from the image analysis of the isothermally treated samples in the semi-solid state and after semi-solid printing: (a) mean equivalent diameter, (b) mean circularity, and (c) liquid fraction.

(9)

conditions enhanced the globularization process. The sample which was isothermally treated for 10 s was observed to exhibit a circularity value of 0.79 ± 0.02, which evolved to 0.86 ± 0.01 after 300 s. Nonetheless, it was not observed to improve after a more prolonged treatment (600 s), indicating that the obtained value was already elevated and the driven force, therefore, was reduced. These results indicate that the shear ap-plied in the nozzle channel is beneﬁcial in improving the thixotropic material.

Results related to liquid fraction are shown inFig. 9c. It was observed that the results for isothermally treated samples in the semi-solid state showed no signiﬁcant difference. The measured liquid fraction was ap-proximately 0.51, a value smaller than values predicted by simulation and obtained by the DSC measurement (Fig. 5). Such discrepancies may be considered to originate due to the artifacts resulting from the quenching procedure. It is very difﬁcult to completely avoid some phase growth; thus, the room temperature microstructure does not per-fectly correspond to the semi-solid one. Limodin et al. [33] observed the discrepancies after performing in situ analysis, demonstrating that the quenched microstructure should be analyzed with caution. Another point to consider is that the liquid fraction obtained from the thermody-namic simulations was related to the mass fraction, whereas it was as-sociated with the volume fraction when determined by the image analysis. This implies that some differences can be observed based on the densities of the solid and liquid phases. It should be noted that the liquid fraction of the printed sample showed a higher value (around 0.57). However, it should be considered that the cooling rate was signif-icantly lower, which combined with the reheating caused by the subse-quent layer deposition, may have contributed to the development of the phases closer to the path expected according to the phase diagram (Fig. 2). In such a case, the primary phase tends to decrease as the tem-perature decreases, i.e., the measured liquid phase may be overestimated. Furthermore, liquid segregation that occurred in the nozzle exit channel was also considered to explain this higher liquid fraction, as shown later.

The obtained results for mean particle size (Fig. 9a), combined with high values for mean circularity (Fig. 9b), indicated that the microstruc-tural behavior of Mg-38Zn alloy was excellent for application in the semi-solid processing, which was remarkably favored by the small par-ticles with high circularity (values closer to 1) [34–36]. Such character-istics can reduce the segregation of the liquid phase and also decrease

the force required to promote theﬂow of the semi-solid material [,35–39].

For better understanding the microstructure evolution along the printing process, aﬁlament was naturally solidiﬁed inside the nozzle channel and subsequently evaluated using SEM, resulting inFig. 10.

Fig. 10a shows the typical microstructure of the Mg-38Zn_{ﬁlament in} the cold zone of the nozzle, which is basically the extrudedﬁlament (Fig. 6). Furthermore,Fig. 10b shows the microstructure in the cold/ hot transition zone, where recrystallization, partial melting, and globularization mechanisms occur, transforming the aligned micro-structure into a globular one within a region shorter than 200μm. Fi-nally,Fig. 10c reveals the coarsened globular microstructure of the material just before it leaves the hot nozzle channel. It should be noted that the remaining solid phase is much coarser than the printed material (Fig. 8b), however, this was due to longer exposure at high temperatures as cooling occurred naturally. This implies that the mate-rial cannot be exposed to high temperatures for long periods to avoid excessive coarsening, which can clog the nozzle. Another point to con-sider is the evident lower liquid fraction. This could be due to liquid seg-regation that can occur during deformation [39,40], causing the material that leaves the nozzle to exhibit a higher amount of liquid, and the ma-terial at the end of the nozzle to exhibit a smaller liquid fraction. Reduc-tion in the liquid fracReduc-tion due to segregaReduc-tion can also explain the coarse globules observed in this region. As the liquid fraction decreased, the diffusion distances also decreased, contributing to an increased coarsen-ing rate.

4. Summary and conclusions

In this study, Mg-Zn alloys were investigated to select a suitable composition to be processed by the FFF technique. The solid-to-liquid transition was investigated using thermodynamic simulation, which re-sulted in some promising compositions. Among them, Mg-38Zn was se-lected for experimental evaluation. DSC analysis was carried out with this composition, and the results were found to be reasonably consistent with the simulated results. The microstructure evolution in the semi-solid state at 420 °C suggested a good globularization behavior. The iso-thermal treatment for 10 s already provided a proper microstructure with aﬁne and globular solid phase. Furthermore, longer treatments were observed to slightly improve the shape factor (more globular

Fig. 10. Schematic of temperature (TL- liquidus temperature, TE- extrusion temperature and TS- solidus temperature) and microstructural evolution of Mg-38Zn alloy within the nozzle

channel: (a) cold zone, (b) cold/hot transition zone, and (c) hot zone.

(10)

morphology), but the mean globule diameter was observed to signi fi-cantly increase. The globularization behavior was considered to be ben-e_{ficial, as a proper microstructure should be rapidly obtained to not} impair the continuity of the printing process. Thixo-printing performed at 420 °C was conducted without any obstruction at the nozzle channel, thus, allowing the production of sound parts with acceptable welding between the deposited layers. The microstructural investigation re-vealed that the printed parts presented a more refined microstructure compared to those which were statically heat-treated in the semi-solid state, because of the shear applied in the nozzle channel. Further-more, liquid segregation was also observed, which increased the liquid fraction in the deposited layer. In conclusion, the good results obtained with the biodegradable Mg-38Zn alloy demonstrate the feasibility of the semi-solid FFF process (also termed as“3D thixo-printing”), opening new perspectives for the AM technology.

CRediT authorship contribution statement

Dalton Daniel de Lima: Conceptualization, Investigation, Formal analysis, Writing - original draft. Kaio Niitsu Campo: Conceptualization, Formal analysis, Writing - review & editing. Sergio Tonini Button: Writing - review &editing. Rubens Caram: Funding acquisition, Super-vision, Writing - original draft.

Declaration of Competing Interest

We declare that no conﬂict of interest exists. Acknowledgments

The authors are grateful for theﬁnancial support provided by the Brazilian research funding agencies CNPq (National Council for Scien-tiﬁc and Technological Development), Grant # 405054/2016-5 and FAPESP (São Paulo Research Foundation) Grant # 2018/10190-5. References

[1] N.J. Hallab, J.J. Jacobs, Biologic effects of implant debris, Bull. NYU HJD. 67 (2009) 182–188.

[2]J.J. Jacobs, J.L. Gilbert, R.M. Urban, Current concepts review - corrosion of metal or-thopaedic implants, J. Bone Joint Surg. Am. 80 (1998) 1554.

[3] P.J. Denard, P. Raiss, R. Gobezie, T.B. Edwards, E. Lederman, Stress shielding of the humerus in press-ﬁt anatomic shoulder arthroplasty: review and recommendations for evaluation, J. Shoulder Elb. Surg. 27 (6) (2018) 1139–1147.

[4]S. Kamrani, C. Fleck, Biodegradable magnesium alloys as temporary orthopaedic im-plants: a review, BioMetals 32 (2) (2019) 185–193.

[5]D. Zhao, F. Witte, F. Lu, J. Wang, J. Li, L. Qin, Current status on clinical applications of magnesium-based orthopaedic implants: a review from clinical translational per-spective, Biomaterials 112 (2017) 287–302.

[6] D. Vojtěch, J. Kubásek, J. Šerák, P. Novák, Mechanical and corrosion properties of newly developed biodegradable Zn-based alloys for boneﬁxation, Acta Biomater. 7 (2011) 3515–3522.

[7]E.C. Huse, A new ligature, Chicago Med. J. Exam. (1878) 171–172.

[8]R.A. Reinhart, Magnesium metabolism: a review with special reference to the rela-tionship between intracellular content and serum levels, Arch. Intern. Med. 148 (1988) 2415–2420.

[9]X. Zhang, Z. Wang, G. Yuan, Y. Xue, Improvement of mechanical properties and cor-rosion resistance of biodegradable Mg-Nd-Zn-Zr alloys by double extrusion, Mater. Sci. Eng. B 177 (2012) 1113–1119.

[10] Y. Chen, Z. Xu, C. Smith, J. Sankar, Recent advances on the development of magne-sium alloys for biodegradable implants, Acta Biomater. 10 (2014) 4561–4573.

[11] F. Witte, The history of biodegradable magnesium implants: a review, Acta Biomater. 6 (2010) 1680–1692.

[12]S. Zhang, X. Zhang, C. Zhao, J. Li, Y. Song, C. Xie, H. Tao, Y. Zhang, Y. He, Y. Jiang, Y. Bian, Research on an Mg-Zn alloy as a degradable biomaterial, Acta Biomater. 6 (2010) 626–640.

[13] L.E. Murr, E. Martinez, K.N. Amato, S.M. Gaytan, J. Hernandez, D.A. Ramirez, P.W. Shindo, F. Medina, R.B. Wicker, Fabrication of metal and alloy components by addi-tive manufacturing: examples of 3D materials science, J. Mater. Res. Technol. 1 (2012) 42–54.

[14]R. Galante, C.G. Figueiredo-Pina, A.P. Serro, Additive manufacturing of ceramics for dental applications: a review, Dent. Mater. 35 (6) (2019) 825–846.

[15] S. Singh, G. Singh, C. Prakash, S. Ramakrishna, Current status and future directions of fusedﬁlament fabrication, J. Manuf. Process. 55 (2020) 288–306.

[16] Z. Ma, H. Zhang, X. Zhang, X. Wu, H. Fu, L. Jia, H. Zhang, Rheological behaviour of par-tially solidiﬁed A356 alloy: experimental study and constitutive modelling, J. Alloys Compd. 803 (2019) 1141–1154.

[17] C.S. Rice, P.F. Mendez, S.B. Brown, Metal solid freeform fabrication using semi-solid slurries, JOM 52 (2000) 31–33.

[18] S. Finke, F.K. Feenstra, Solid freeform fabrication by extrusion and deposition of semi-solid alloys, J. Mater. Sci. 37 (2002) 3101–3106.

[19] J. Mireles, H.C. Kim, I.H. Lee, D. Espalin, F. Medina, E. Macdonald, R. Wicker, Develop-ment of a fused deposition modeling system for low melting temperature metal al-loys, J. Electron. Packag. Trans. ASME. 135 (2013) 1–6.

[20] Wen Chen, Luke Thornley, Hannah G. Coe, Samuel J. Tonneslan, John J. Vericella, Cheng Zhu, Eric B. Duoss, Ryan M. Hunt, Michael J. Wight, Diran Apelian, Andrew J. Pascall, Joshua D. Kuntz, Christopher M. Spadaccini, Direct metal writing: Control-ling the rheology through microstructure, Applied Physics Letters 110 (9) (2017).

[21] A. Jabbari, K. Abrinia, A metal additive manufacturing method: semi-solid metal ex-trusion and deposition, Int. J. Adv. Manuf. Technol. 94 (2018) 3819–3828.

[22]A. Jabbari, K. Abrinia, Developing thixo-extrusion process for additive manufactur-ing of metals in semi-solid state, J. Manuf. Process. 35 (2018) 664–671.

[23] D. Liu, H.V. Atkinson, H. Jones, Thermodynamic prediction of thixoformability in al-loys based on the Al-Si-Cu and Al-Si-Cu-Mg systems, 53, Acta Materialia, 2005 3807–3819.

[24] A. Maciel Camacho, H.V. Atkinson, P. Kapranos, B.B. Argent, Thermodynamic predic-tions of wrought alloy composipredic-tions amenable to semi-solid processing, Acta Mater. 51 (2003) 2319–2330.

[25] Q. Han, S. Viswanathan, The use of thermodynamic simulation for the selection of hypoeutectic aluminum-silicon alloys for semi-solid metal processing, Mater. Sci. Eng. A 364 (2004) 48–54.

[26] Y.D. Li, D. Apelian, B. Xing, Y. Ma, Y. Hao, Commercial AM60 alloy for semisolid pro-cessing: alloy optimization and thermodynamic analysis, Trans. Nonferrous Metals Soc. China 20 (2010) 1572–1578 (English Ed.).

[27] Y.Q. Liu, A. Das, Z. Fan, Thermodynamic predictions of Mg-Al-M (M=Zn, Mn, Si) alloy compositions amenable to semisolid metal processing, Mater. Sci. Technol. 20 (2004) 35–41.

[28] K.N. Campo, D.D. de Lima, É.S.N. Lopes, R. Caram, On the selection of Ti–Cu alloys for thixoforming processes: phase diagram and microstructural evaluation, J. Mater. Sci. 50 (2015) 8007–8017.

[29] E. Povoden-Karadeniz, Thermodynamic Data from MatCalc Database“mc_sample_ mg” for simulations in AZ Mg-base Alloys,https://www.matcalc.at/2013. [30] F. Czerwinski, Critical assessment 32: controling the melting behaviour of cast

struc-tures through solid-state deformation, Mater. Sci. Technol. 35 (2019) 999–1006.

[31] C.A. Schneider, W.S. Rasband, K.W. Eliceiri, NIH image to ImageJ: 25 years of image analysis, Nat. Methods 9 (2012) 671–675.

[32] M.C. Flemings, Behavior of metal alloys in the semisolid state, Metall. Trans. A. 22 (1991) 957–981.

[33] N. Limodin, L. Salvo, M. Suéry, M. DiMichiel, In situ investigation by X-ray tomogra-phy of the overall and local microstructural changes occurring during partial remelting of an Al-15.8 wt.% Cu alloy, Acta Mater. 55 (2007) 3177–3191.

[34] D.J.B.D. Brabazon, D.J. Browne, A.J. Carr, Mechanical stir casting of aluminium alloys from the mushy state: process, microstructure and mechanical properties, Mater. Sci. Eng. A 326 (2002) 370–381.

[35] O. Lashkari, R. Ghomashchi, The implication of rheology in semi-solid metal pro-cesses: an overview, J. Mater. Process. Technol. 182 (2007) 229–240.

[36] K.N. Campo, C.C. de Freitas, L. Fanton, R. Caram, Melting behavior and globular mi-crostructure formation in semi-solid CoCrCuxFeNi high-entropy alloys, J. Mater. Sci. Technol. 52 (2020) 207–217.

[37] D.H. Kirkwood, Semisolid metal processing, Int. Mater. Rev. 39 (1994) 173–189.

[38] H.V. Atkinson, Modelling the semisolid processing of metallic alloys, Prog. Mater. Sci. 50 (2005) 341–412.

[39] G. Hirt, R. Kopp, Thixoforming: Semi-solid Metal Processing, John Wiley & Sons, 2009.

[40] M. Qi, J. Li, Y. Kang, Correlation between segregation behavior and wall thickness in a rheological high pressure die-casting AC46000 aluminum alloy, J. Mater. Res. Technol. 8 (4) (2019) 3565–3579.