Solvent-cast 3D printing of magnesium scaffolds

(1)

Solvent-cast 3D printing of magnesium scaffolds

Dong, J.; Li, Y.; Lin, P.; Leeflang, M. A.; van Asperen, S.; Yu, K.; Tümer, N.; Norder, B.; Zadpoor, A. A.;

Zhou, J.

DOI

10.1016/j.actbio.2020.08.002

Publication date

2020

Document Version

Final published version

Published in

Acta Biomaterialia

Citation (APA)

Dong, J., Li, Y., Lin, P., Leeflang, M. A., van Asperen, S., Yu, K., Tümer, N., Norder, B., Zadpoor, A. A., &

Zhou, J. (2020). Solvent-cast 3D printing of magnesium scaffolds. Acta Biomaterialia, 114, 497-514.

https://doi.org/10.1016/j.actbio.2020.08.002

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

(2)

Acta

Biomaterialia

journalhomepage:www.elsevier.com/locate/actbio

Full

length

article

Solvent-cast

3D

printing

of

magnesium

scaffolds

J. Dong

a ,∗

_,

_Y.

_Li

a

_,

_P.

_Lin

b

_,

_M.A.

_Leeﬂang

a

_,

_S.

_van

_Asperen

c

_,

_K.

_Yu

d

_,

_N.

_Tümer

a

_,

_B.

_Norder

e

_,

A

.A

.

Zadpoor

a

_,

_J.

_Zhou

a

a Department of Biomechanical Engineering, Delft University of Technology, Delft 2628 CD, the Netherlands b Department of Engineering Structures, Delft University of Technology, Delft 2628 CN, the Netherlands c Department of Materials Science and Engineering, Delft University of Technology, Delft 2628 CD, the Netherlands

d Department of Bionanoscience & Kavli Institute of Nanoscience, Delft University of Technology, Delft 2629 HZ, the Netherlands e Department of Chemical Engineering, Delft University of Technology, Delft 2629 HZ, the Netherlands

a

r

t

i

c

l

e

i

n

f

o

Article history: Received 10 March 2020 Revised 13 July 2020 Accepted 3 August 2020 Available online 7 August 2020

Keywords: Additive manufacturing Magnesium Scaffold Solvent-cast Sintering

a

b

s

t

r

a

c

t

Biodegradableporous magnesium (Mg) scaffolds arepromising for application inthe regeneration of critical-sizedbonedefects.Althoughadditivemanufacturing(AM)carriesthepromiseofofferingunique opportunitiestofabricateporousMgscaffolds,currentattemptstoapplytheAMapproachtofabricating Mgscaffoldshaveencounteredsomecrucialissues,suchasthoserelatedtosafety inoperationandto thedifficultiesincompositioncontrol.Inthispaper,wepresentaroom-temperatureextrusion-basedAM methodforthefabricationoftopologicallyorderedporousMgscaffolds.Itiscomposedofthreesteps, namely(i)preparing aMgpowderloaded inkwithdesiredrheologicalproperties,(ii) solvent-cast3D printing(SC-3DP)ofthe inktoform scaffoldswith 0°/ 90°/ 0 ° layers,and (iii)debindingand sin-teringtoremovethebinderintheinkandthengetMgpowderparticlesbondedbyapplyinga liquid-phasesinteringstrategy.Arheologicalanalysisofthepreparedinkswith54,58and62 vol%Mgpowder loadingwasperformedtorevealtheirviscoelasticproperties.Thermal-gravimetricanalysis(TGA),Fourier transforminfraredspectroscopy(FTIR),carbon/sulfuranalysis and scanningelectronmicroscopy(SEM) indicatedthepossibilitiesofdebindingandsinteringatonesinglestepforfabricatingpureMgscaffolds withhighfidelityanddensification.Theresultingscaffoldswithhighporositycontainedhierarchicaland interconnectedpores.Thisstudy,forthefirsttime,demonstratedthattheSC-3DPtechniquepresents un-precedentedpossibilitiestofabricateMg-basedporousscaffoldsthathavethepotentialtobeusedasa bone-substitutingmaterial.

StatementofSigniﬁcance

Biodegradableporousmagnesiumscaffoldsarepromisingforapplicationintheregenerationof critical-sizedbonedefects.Althoughadditivemanufacturing(AM)carriesthepromiseofofferingunique opportu-nitiestofabricateporousmagnesiumscaffolds,currentattemptstoapplytheAMapproachtofabricating magnesiumscaffoldsstill havesomecruciallimitations. Thisstudydemonstrated thatthesolvent-cast 3DprintingtechniquepresentsunprecedentedpossibilitiestofabricateMg-basedporousscaffolds.The judiciouschosenofformulatedbindersystemallowedforthenegligiblebinderresidueafterdebinding and theshort-timeliquid-phasesintering strategyledtoagreat successinsinteringpuremagnesium scaffolds.Theresultingscaffoldswith hierarchicaland interconnectedpores havegreatpotential tobe usedasabone-substitutingmaterial.

∗ _{Corresponding author.}

E-mail address: J.Dong-5@tudelft.nl (J. Dong).

1. Introduction

Theregeneration of critical-sizedbone defects remains a clin-ical challenge and usually necessitates bone grafting materials. However, the currentclinically available grafts andmostexisting synthetic biomaterials do not meet all clinical requirements [1] .

https://doi.org/10.1016/j.actbio.2020.08.002

(3)

Therefore, developing a new generation of suitable bone substi-tutesisdesired.Anidealbonesubstituteorimplantshould:(i)be abletodegrade invivowithin an appropriatetime periodtill the newlygrownbonetissuehasreplacedthesubstitute’sfunction[2] , (ii)exhibit osteogenic,osteoconductiveandosteoinductive proper-tiesfornewboneformation[3] ,(iii)bebiomechanicallystableand matchthe mechanicalpropertiesofthehostbone toavoidstress shielding[2 ,3] ,priortolosingitsmechanicalfunctionality,and(iv) presenta highly porous structure with fullyinterconnected pore networksto allow for bone ingrowthand facilitatethe transport ofnutrientsandmetabolicwaste[4] .

In recent years, Mg-based biodegradable metals have been consideredtobeanewclassofpromisingbiomaterialsfor ortho-pedicapplications,sincetheymaybetunedtopossessmechanical and biological performances needed for bone defect regenera-tion [5] . One of the main advantagesof Mg over currently used metallic biomaterials is its biodegradability [5] . Mgions play an importantfunctional role in physiological systems [6] . Moreover, it has been reported that Mg ions released from Mg implants stimulate bone formation [7] . Additionally, Mg possesses appro-priatemechanicalpropertiesthatareclosetothoseofthenatural bone[8] .Inother words,Mg-basedmaterials cansatisfy theﬁrst three requirements mentioned above for biodegradable implants aimed to regenerate bone. Incorporating porosity into Mg-based implants is of critical importance to meet the last requirement of an ideal bone substitute for orthopedic and traumatological applications.

Several techniques to fabricate porous Mg scaffolds have al-ready been developed including melt foaming [9 –11] , preform infiltration[12 –16] ,patterncasting[17 ,18] ,andpowdermetallurgy withspace holder[19 –22] .However, thesetraditional fabrication techniques permit neither complex exterior 3D architectures nor fully interconnected interior networks. Recent advances in additive manufacturing (AM) offer great potential to achieve a much greater degree of design andmanufacturing flexibility and efficiency over the traditional techniques, making it possible to design and fabricate fully interconnected porous structures with preciselycontrolledtopologicalparameters[23 ,24] .

Employing AMtechniquesto fabricateMg-basedmaterials has juststartedina verysmallnumber ofuniversityresearch groups

[25] .Selectivelasermelting(SLM) inthecategoryofpowder bed fusionaccordingtotheASTMclassificationisacommonlychosen AM technique for Mg scaffold fabrication [25 –30] , compared to other AM techniques [31 ,32] . The success in fabricating porous Mg alloy scaffolds with SLM is yet rather limited [33 ,34] . The primary challenge encountered is associated with the safety in operation,considering thehigh flammabilityof Mg powder with a large collective surface area. The other difficulties encountered concerntheundesirable compositionalvariationinfinalpartsdue torelatively low meltingand boiling temperatures ofMg andto rapidlyincreasedvaporpressure ofMgatitsboiling temperature

[27 ,35] ,excessiveoxidationthroughoutthewholechainofscaffold fabrication due to the high affinity of fine Mg-based powder particlesto oxygen [27] , andseriously limitedavailability of pre-alloyedMgpowders withdesiredcompositionsthat canbe used toadjustthebiodegradationrateandmechanicalpropertiesofthe porous structures [25] . Furthermore, due to the high reflectivity ofMgpowder,highlaserpowerisneeded, whichmakestheSLM equipment costly [27] , increases the chance of Mg evaporation, andcreates significant thermal gradients during SLM, leading to metallurgicaldefectsandresidualstresses[36] .Recently,afew at-temptshavebeenmadetodeveloppowderbedinkjet3Dprinting

[37] and fused ﬁlament fabrication (FFF)techniques[38] ,followed byasinteringstep,asalternativestoSLM.However,utilizingthese techniquesforthefabrication oftopologically orderedporous Mg scaffoldshasnotyetbeenreported.

Solvent-cast 3D printing (SC-3DP) is another interesting tech-nique alternative to SLM, in which an ink containingmetallic or anyother powder particles,together with abinder system (com-posed of polymer, volatile solvent, and/or additive), is extruded through a nozzle and printed into a designed structure. The 3D printed structure is then subjected to debinding and sintering. SC-3DP has great potential to overcome the material, technical, and structural limitations of the other AM techniques currently applied to fabricate porous Mgscaffolds, asmentioned earlier. It offers multiple advantages including (i) 3D printing at ambient conditions,(ii)easyadjustmentofthecomponentsoftheink,(iii) low investment in equipment, and(iv) the potential to fabricate complex structures with hierarchical pores and desired alloy compositions.Inaddition,inSC-3DP,nonaturalsupportispresent foroverhungparts,whichisincontrasttothepresenceofnatural supportinpowder bedbasedAM processes.Precisematchofthe evaporation rate of solvent and printing parameters allows for printingstackedlatticestructureswithoverhangpartsandenables the fabrication of helically freeform structures without any syn-thetic support [39] . The SC-3DPmethod has been demonstrated tobe feasibleinthe caseof theAMofsteel[40 ,41] ,iron[42 ,43] , titanium-based[44] ,andnickel-based[43 ,45 ,46] micro-trusses.

Tothe best ofour knowledge, there havebeen no reportson SC-3DP of Mg scaffolds. It is worth notingthat unlike relatively less reactive metals, fabricating Mg-based materials through the SC-SDP approach could present a different set of challenges in-cluding (i) a limited choice of binder components due to the high reactivity of Mg powder, leading to interactions during the printing and debinding processes and the exclusion of the use of water-based and PLGA-containing binder systems suitable for SC-3DP ofother metals,(ii) the poor sinterability of Mg powder in the absence of external pressure, because of the inevitable presence of a stable oxide ﬁlm on Mg powder particle surface, whichactsasabarriertodiffusion.

In the present study, we attempted to overcome these chal-lenges and apply the SC-3DP technique to build topologically ordered biodegradablepure Mgscaffolds.PureMgwaschosen in this first attemptto make the material system simple andavoid unexpectedinteractionsbetweenthealloyingelementsandbinder. Thisworkwasaimedtodemonstratetheviabilityofusingthe SC-3DPtechniquetoproduceMgscaffoldsandtofindapplicableand optimized process conditions for successfully fabricating porous Mgscaffoldswithhighfidelity.First,thepreparedinks,consisting ofajudiciouslyformulatedbinderandMgpowder particles,were evaluated in their rheological behavior. Rheological characteriza-tion was intended to lay the basis for initial screening of inks prior totime-consumingtrials of3D printingandalso toprovide quantitative support for the evaluation of printability and for the descriptionofobserved behavior duringthe printingprocess. Second,theinkswithdifferentpercentagesofMgpowderloading wereprintedbyapplyingtheSC-3DPapproachandtheapplicable printabilitywindowswere defined.Finally,theeffectsofsintering temperatureandholdingtimeonthedensityofstrutswere inves-tigatedinlightofmaintainingtheshapefidelityofthescaffolds. 2. Materialsandmethods

2.1. Fabricationstrategy

TheSC-3DPapproachtothefabrication ofporousMgscaffolds consisted of three steps (Fig. 1 a): ink preparation, 3D printing of Mg scaffolds with a designed structure, and debinding and sintering.First,inks containinga bindersystem(polymer,volatile solvent, and additive), and different percentages of Mg powder loadingwereprepared.Second,thepreparedinkwiththerequired rheologicalpropertieswasextrudedthroughamicro-nozzleunder

(4)

Fig. 1. A schematic diagram of the fabrication steps and the designed structure of Mg scaffolds: (a) the fabrication process consisting of ink preparation, 3D printing, and post-processing, (b) the CAD model of the scaffold, and (c) the lay-down pattern of the scaffold.

an applied pressure and at an appropriate printing speed. The volatile solvents in the extruded struts rapidly evaporated upon extrusion, enabling the condensation of the remaining polymer and increasing the rigidity of the deposited Mg powder loaded struts.Thisallowedthestrutstoholdtheirshapeandsupportthe subsequentlayers.Thepresenceoftheadditivepreventedthe sol-ventfromexcessivelyfastdrying,duetoitslowervaporpressure, thereby permitting adjacent layers to merge during deposition. Green scaffold samples with Mg powder particles, polymer, and additives were created. Finally, the green scaffold samples were transformed intopure Mgscaffolds througha thermaltreatment, duringwhich polymer andadditiveswere thermallydecomposed andMgpowderparticlesweresinteredtogether.

2.2. Inkpreparationandcharacterization 2.2.1. Powdercharacterization

Atomizedpure Mgpowderparticleswithapurityof99.8wt% (Zn:0.0081wt%,Fe: 0.0085wt% andothers: 0.17wt%)(Tangshan Weihao MagnesiumPowder Co.,China) had a particlesize range of 25 to 80 μm (D10 = 31.11 μm, D50 = 44.96 μm and D90 =

67 μm). The morphology of the powder was examined with a scanning electron microscope (SEM, JSM-IT100, JEOL, Japan). The microstructure of the Mg powder was observed with an optical microscope(OM,VH-Z250R,KEYENCE,USA)aftergrinding, polish-ing,andetchinginasolutioncomposed ofnitricacid,aceticacid,

water,andethanolatavolumeratioof1:3:4:12.Theaveragegrain sizewasdeterminedbyusingthelineinterceptmethod.

2.2.2. Inkpreparation

Mg powder loaded inks were synthesized through mixing of (i) Mg powder particles and (ii) a binder system composed of polystyrene (Sigma-Aldrich), chloroform (Sigma-Aldrich) and dibutyl phthalate (Sigma-Aldrich). Mg powder particles were addedintothealreadymadebinderinagloveboxtopreventMg powderparticlesfrombeingoxidizedduringinkpreparation. Inks with54,58,and 62vol% powder loadingwere prepared, respec-tively.ThebinderandMgpowdermixtureswerefurthermixedby applyingmagneticstirringforimprovementsinhomogeneity.The synthesizedinkswerethencentrifugedat102rad/sfor1minand at25°Ctoremoveairbubbles.

2.2.3. Rheologicalcharacterization

Therheologicalpropertiesofthe54,58,and62vol%Mg pow-der loaded inks were determined by using a rheometer(Physica MCR 301, Anton Paar, Germany) with a plate-plate geometry (25 mm diameter, 1 mm distance). A shear stress ramp, ranging from1to10,000Paat1Hz,wasappliedtotheinkandthelinear viscoelastic range (LVR) was determined. Rotational shear rate-viscositymeasurementswereperformedintheﬂowmodeatshear rates varying between 0.1 and 220 s−1 . The linear trend when plottingtheresultsonadouble-logarithmicscalewasﬁttedintoa

(5)

powerlawequationtoquantifythedegreeofshearthinning:

η

=K

γ

˙n−1 ₍₁₎

where

η

istheviscosity,

γ

˙ theshearrate,Ktheconsistencyindex andntheﬂowbehaviorindex.

The storage (G’) and loss (G’’) moduli were obtained from a frequency sweep test over a frequency range of 0.1 to 10 Hz at a constant stress selected from the LVR (

τ

= 10 Pa). In the creepexperiment,a constantshearstressselectedwithin theLVR (

τ

₌ 10 Pa) was applied for 60 s and the resulting strain was measured.Then,intherecoverytest,thestress wasremovedand thestrain wasrecordedforanother180s.Thecompliance(J)was obtained from the ratio of the measured strain to the applied stress. The degree of recovery (R) of the ink (i.e., the ability to recoverfromdeformation)wascalculatedas:

R=Jmax− J∞

Jmax

(2)

whereJmaxandJ∞ arethecomplianceattheendofthecreepand

therecoverytests,respectively.

2.3.3Dprinting

2.3.1. Flowratemeasurement

The 54,58, and 62 vol% Mg powder loaded inks were trans-ported into plastic syringes (EFD, Nordson, Germany) with a 410 μm tapered nozzle (EFD, Vieweg, Germany), which were mountedona 3D Bioscaffolderprinter (BS3.2, GeSim, Germany). Todeterminetheflowratesoftheinkswithdifferentpercentages ofMgpowderloading,theinkswereextrudedunderapplied pres-suresof60,80,100, and500kPa for5, 10,and15min. Because the 62 vol% ink could not be extruded at the applied pressures of60,80or100kPa, anadditionalextrusion pressureof600kPa was applied for evaluating 62 vol% ink flow rate. The extruded Mgpowder containingstrutswere collectedandweighedusinga high-precisionbalanceafterdryinginafumehood.Then,themass flowratewasdeterminedbydividingthetotalweight(theweight oftheevaporatedsolventswasalsoincludedincalculation)ofthe extrudedstruts dispensed atvarious time points along a certain dispensing time (5, 10, or 15 min) to obtain an average mass flowrate.The averagevolumetricflowratewasthenobtainedby convertingthemassflowrateusingtherespectiveinkdensity.The induced shear rate, as the ink flew through the nozzle tip, was correlatedwiththeflowrateandwasestimatedby

˙

γ

=3n+1 4n ·

4Q˙

π

R3 (3)

where

γ

˙ is the shear rate, R the radius of the nozzle, n the shear-thinningcoeﬃcient,andQthevolumetricﬂowrate.

2.3.2. Scaffoldstructuredesign

Acuboidalarchitecture (dimensions:7mm× 7mm × 8mm) withastrut size of410 μm,a spacingdistanceof 390μmanda lay-down pattern of 0°/90°/0°(Fig. 1 b), and a relative density of 53.7%(the designvalue, calculated by Solidworks), wasdesigned usingthe GeSimcustom software.A linearinﬁll pattern(Fig. 1 c) was adopted for the designed structure, since it is the most reliable pattern to create uniform porous structures and allows for a fast and facile fabrication process. This designed structure was meant to easily and quickly determine the printability of the created Mg powder loaded inks and the formability of the scaffolds,aswellastheoptimizedprintingprocessparameters.

2.3.3. 3Dprintingandprintabilitywindow

The54,58,and62vol%Mgpowderloadedinkswereextruded under various applied pressures at room temperature. The ex-truded struts were deposited on a glass slide as the substrate.

The distancebetweenthe glass slideandthe nozzletip, andthe layerheightwerebothsettobe320μminthesystem,whichwas almost 80% ofthe inner diameter of the nozzle.Considering the capacityoftheprinter used,appliedpressuresrangingbetween1 and600kPa (the maximumpressureoftheprinter)andprinting speedsvarying between 1and18 mm/swere used to determine the printability windows ofthe three Mgpowder loaded inks to realizethedesiredarchitecture.First,theoperatingprocess param-eters(appliedpressureandprintingspeed)thatenabledthewidth ofextrudedstrutsina rangeof300–500μm werechosen asthe preliminary parameters. Consideringthe diﬃculties to obtain the strutshavingexactlythesamediameterasthenozzlesize,a rela-tivelylargerangeof300–500μmwasempiricallychosentomake theresultantprintingwindowsuitablefordifferentprintingcases ofMgscaffolds,sinceinevitableexperimentalerrorscould greatly affect the size of extruded struts during the printing process. Then,theseparameterswerefurtherscreenedthroughbuilding3D structuresandevaluatingthem.Finally,theprintabilitywindowsof theinks tofabricate thedesignedMg scaffoldswere determined. The criteria for the printability of the scaffolds were as follows (i) thedeﬂectiondegree(

δ

z) ofspannedstrutsmust belessthan 5% of the extruded strutsdiameter[47] and (ii) there should be no apparent defects ordisplacements of the printedstruts. After the printability windows were constructed,a number ofprinting processparametersets wereselectedforthesezonesandmarked as A, B, C, D, E and F in the windows, and then the resultant scaffoldswereimagedunderSEMandtakenastheexamples.

2.4. Debindingandsintering

The samples with 58 vol% Mg powder loading, struts of 350 ± 10 μm, and an interspace of 450 ± 9 μm, printed with the parameters selected from the deﬁned printability window, were subjectedtodebindingandsintering.Thedebindingandsintering processes were conducted in a tube furnace with a controlled atmosphereofhigh-purityargon(purity≥ 99.9999%).Theprinted materials were heated at 5 °C/min from room temperature to 650°Cwithvarious dwellingtimes(5, 10,and35min),followed by furnacecoolingto room temperature.To investigatetheeffect of sintering temperature, part of the 3D printed samples was sinteredat660and670°C.

2.5. Materialcharacterization

Todeterminetheexacttemperatureofbinderremoval, thermal-gravimetricanalyses (TGA8000,PerkinElmer,USA)ofpurebinder samplesandgreensampleswith58vol%Mgpowderloadingwere conductedbyheatingthesamplesto650°Catarateof5°C/min andatan Arflowrateof40mL/min.Theas-receivedMgpowder and the as-sintered Mg samples were also subjected to TGA, following the same procedure as above. The analyses were per-formed intriplicate.The adsorption spectraofthe greensamples andthose heatedto 220, 450,and650 °C were generated using a Fourier transform-infrared spectroscope (FTIR, Spectrum 100, PerkinElmer, USA).A portionof the greensample wasplaced on thediamond attenuatedtotalreflection(ATR) crystal.The spectra overarangeof4000to600cm−1 wereobtainedby30scanswith a spectral resolution of 4 cm−1 . Phase constituents in pure Mg powder, printed and sintered scaffolds were identified by using an X-ray diffractometer (XRD, Bruker D8 Advance diffractometer Bragg–Brentanogeometry).The diffractometerwasequippedwith a Lynxeye position sensitive detector andoperated at45 kVand 40 mA with a scan range of 10–110° and a step size of 0.030° using Cu Ka radiation. Quantitative analysisof magnesium oxide (MgO) was performed using Bruker software (Diffrac.Suite EVA,

(6)

Fig. 2. The characteristics of the Mg powder: (a) a SEM image showing spherical powder particles and (b) the microstructure of the Mg powder.

5.2). Chemical analyses of the as-received Mg powder and sin-tered parts in terms of carbon content were performed using a carbon/sulfur analyzer(CS744, LECO, USA). Three repeat samples foreachgroupweremeasured.

The printed 2D pattern and the cross section of the sintered samples were imaged under the optical microscope. The green samples and sintered samples were also examined using SEM. The relative density of the sintered struts was measured from theopticalimagesofthepolishedcrosssectionsusingtheImageJ software (National InstitutesofHealth, US).The shrinkage ofthe sinteredsampleswasdeterminedbymeasuringthedimensionsof the samplesbeforeandaftersinteringusinga digitalcaliper.The sintered scaffolds were imaged by using X-ray micro-computed tomography(μCT,Nanotom180NF,Phoenix)atacurrentof140μA and a voltage of 140kV, andresolutions of5 μm and 2 μm. To calculate the porosity, struts size, and pore size of the sintered samples, the scanned images were exported to Fiji (National InstitutesofHealth, US). Theregions ofinterest(ROIs)were then deﬁned. Subsequently, the BoneJ [48] plugin was used for the calculations.

2.6. Statisticalanalysis

Statistical analysis was performed using ANOVA (one way or two way) or the Scheirer-Ray-Hare test with post-hoc Tukey’s multiple comparison test. p _< 0.05 was considered statistically signiﬁcant.

3. Results

3.1. CharacteristicsoftheMgpowder

Atomized pure Mg powder particles appeared to be spherical (Fig. 2 a) under SEM, typical of the characteristics of an argon-protectedcentrifugally atomized powder.Onthe cross-sectionsof powderparticles,theaveragegrainsizewas8.6±1.7μm(Fig. 2 b).

3.2. Rheologicalcharacteristicsoftheinks

The viscosity-shearstress curve of each ofthe threeprepared inks (with 54, 58 and 62 vol% Mg powder loading) showed a steepdropinviscosityatacertainshearstress(Fig. 3 a).Arelative steady region was observed before reaching the critical shear stress, knownastheyield stress(

τ

y). Theyield stresswas

deter-mined at the intersectionof the two tangents:one in the stable region of viscosity where the material was deformed elastically and another in the region where the viscosity dropped and the material startedto ﬂow.The yieldstress oftheinkwitha higher percentageofMgpowderloadingwashigherthanthatoftheink witha lower percentageof Mgpowder loading(Table 1 ). Fig. 3 b

Table 1

Rheological parameters of the inks with different percentages of Mg powder loading, in which n is the shear-thinning coeﬃcient and K the consistency index.

Ink Yield stress (Pa) n K Recovery degree 54 vol% ink 2677 ± 87 a _{0.18 ± 0.03}a 240 ± 36 a _{26% ± 1%}a 58 vol% ink 4753 ± 6 b _{0.06 ± 0.01}a 1313 ± 337 a 29% ± 3% a 62 vol% ink 5917 ± 76 c _{0.05 ± 0.01}a 6782 ± 419 b 27% ± 3% a a–c: different lowercase letters represent statistically signiﬁcant differences between the groups of the different percentages of Mg powder loading at p < 0.05 (comparison in column).

The same letter indicates that the values are not signiﬁcantly different.

showsthedecreases inviscositywithincreasingshearrateforall thethreesamples,whichisknownastheshear-thinningbehavior, beingofgreatimportancefortheinkstobeextrudedthroughfine nozzleandatthe sametimetoretain theshape afterdeposition. Overthefullshearraterange,theviscosityvalueoftheinkwitha higherpercentageofMgpowderloading washigherthanthat of theink with alower percentage ofMgpowder loading (Fig. 3 b). Theshear-thinningdegreesofthethreeinksweresimilar(Fig. 3 b), whichcould beobservedfromthefittedresults(i.e.,thevaluesof theshear-thinningcoefficient,n)(Table 1 ).

Thestorage(G’)andloss(G’’)moduliweredeterminedthrough a frequency sweep test to examine the viscoelastic properties of the prepared inks, which were considered crucial for successful printing.Inthefrequencysweeptesting,boththestorageandloss moduli increased with rising Mg powder loading and frequency (Fig. 3 c).ForalltheinkswiththethreepercentagesofMgpowder loading,thestoragemoduli were allhigherthan theloss moduli, indicating that the inks showed elastic or solid-like behavior, whichwasnecessaryforretainingtheshapeofextrudedpartand supporting its own weight and the layers on top duringSC-3DP. Thecreep-recoverytestingwasalsoperformedtocharacterizethe viscoelastic properties of the inks, expressed by the compliance

J(t)asafunctionoftime(Fig. 3 d).Theresultsshowedanonlinear growth trend in the creep test and an exponential decay in the recoveryregion.TheinkwithdifferentMgpowderloadingshowed similardegreeofrecovery(Table 1 ).

3.3.3Dprinting

Fig. 4 a showsthattheflowratesoftheinkssteadilyincreased withrisingappliedpressure.Whentheflowratesoftheinkswith different percentages of Mg powder loading at a given applied pressurewerecompared,theinkwith54vol%Mgpowderloading exhibited the highest flow rate, while the ink with 62 vol% Mg powderloadinghadthelowest flowrate. Itshould benotedthat the ink with 62vol% Mg powder loading could not be extruded outofthenozzlewhenapressurelessthan350kPawasapplied.

(7)

0.1

1

10

100

1

10

100 1000

10000

100000

54 vol.% 58 vol.% 62 vol.% Fitted 54 vol.% Fitted 58 vol.% Fitted 62 vol.%

V

iscosity (Pa

. s)

Shear rate (s

-1

₎

1

10

100 1000

10000

1

100 1000

10000

100000

1E6

1E7

V

iscosity (Pa

. s)

Shear stress (Pa)

54 vol.% 58 vol.% 62 vol.%

0

2

4

6

8

10 2000

6000

10000

14000

18000

22000

Modulus (Pa)

Frequency (Hz)

G', 54 vol.% G', 58 vol.% G', 62 vol.% G'', 54 vol.% G'', 58 vol.% G'', 62 vol.%

0

50

100

150

200

250

0

0.4

0.8

1.2

1.6 Compliance,J (10

-3

Pa

-1

)

Time (s)

54 vol.% 58 vol.% 62 vol.%

10 1E8

Fig. 3. The rheological characteristics of the Mg powder loaded inks: (a) the results obtained from the shear stress sweep tests at 1 Hz, (b) the results obtained from the shear rate sweep tests at 10 kPa, (c) the G’ and G’’ values determined from the frequency-sweep tests, and (d) the compliance values determined from the creep-recovery tests.

Weusedthemeasured ﬂowratestocalculatethepredictedshear ratesandsubsequentlythe correspondingviscosity values(atthe point when the inks were extruded through the nozzle) using

Eq. 3 (Table 2 ). Clearly, the shear rateis dependent on the ﬂow rateatagivenexternalpressure.Anincreaseintheapplied pres-sureresultsinasigniﬁcantincreaseintheshearrate,accompanied byadecreaseinviscosity(Table 2 ).

Three separate operating windows (Fig. 4 b) were deﬁned as guidelinesfor selecting the printing process parameters (applied pressureand printing speed)for the designedMg scaffolds with different percentages of Mg powder loading. In each plot, four regions are presented, in which scaffolds with differentqualities canbeobtained.ZoneIindicatesthatonedimensional(1D)struts with widths in the range of 300 to 500 μm and 2D array can be successfully fabricated. As part of zone I, zone II represents therealprintable region for3D printedMgscaffolds, whilezone III and zone IV represent the unprintable regions, in which the widthsofstrutswereout ofthedeﬁnedrangeof300–500μmas aresultofthemismatchofprintingparameters.

Severalcombinations ofthe process parameters selected from the different zones were adopted for 3D printing trials and the resultingspecimenswereconsideredastherepresentativesofthe regions (Fig. 4 b). The struts printedwith theprocess parameters at point A in zone IV were much thinner than the expected struts(Fig. 5 a)andthelowerappliedpressure (orhigherprinting speed) resulted in the displacement of the deposited struts. On theother hand,point Cselectedinzone IIIled tothe fabrication of much thicker struts (Fig. 5 c) and the closure and fusion be-tween adjacent struts. Point B in zone I enabled the successful fabrication of1D strutsor2D arrayswithadesiredgapbetween two struts, without any displacements or fusions (Fig. 5 b). The green scaffold samples fabricated with the process parameters at point D in the zone I of the 54 and 58 vol% windows had a collapsedsurfacewithalargedegreeofdeﬂection(Fig. 5 d),while thescaffoldsamplesprintedwiththeprocessparametersatpoint F in Zone I ofthe 58 and 62 vol% windows possessed excessive defects(Fig. 5 f).TheSEMimageofthescaffoldfabricatedwiththe processparametersatpointEinzoneIIofthethreeinkwindows

(8)

0 100 200 300 400 500 600 Applied pressure (kPa)

Flow rate (µL/s) 54 vol.% 58 vol.% 62 vol.% 16 0 14 12 2 10 8 6 4

54 vol.% 58 vol.% 62 vol.%

2 6 10 14 18 10 70 130 190 250 2 6 10 14 18 10 70 130 190 250 1 1.5 2 2.5 200 300 400 500 600

Applied pressure (kPa)

C D A A C E B C A F B

IV

III

II

I

III

IV

II

I

Printing speed (mm/s) % . l o v 2 6 % . l o v 8 5 % . l o v 4 5

(a)

(b)

B E E D

II

F

Fig. 4. The characteristic printing behavior of the Mg powder loaded inks: (a) ﬂow rate as a function of the applied pressure ∗ _{and (b) the printability windows (II: 3D} printable zone; I: 2D printable zone; III and IV: unprintable zones).

(9)

Table 2

Predicted shear rate values and corresponding viscosity values during the extrusion of the inks at different applied pressures.

Ink 80 kPa 120 kPa 500 kPa

54 vol% Shear rate (s −1 ) 419 ± 17 a1 _{589 ± 12}b1 _{4598 ± 10}c1 Viscosity (Pa s) 1.70 ± 0.05 a1 _{1.29 ± 0.02}b1 _{0.24 ± 0.00}c1 58 vol% Shear rate (s −1_{) 358 ± 2}a2 _{559 ± 84}b1 _{8459 ± 98}c2

Viscosity (Pa s) 5.23 ± 0.02 a2 _{3.46 ± 0.50}b2 _{0.26 ± 0.00}c1 62 vol% Shear rate (s −1 ) 0 a3 ₀a2 _{159 ± 2}b3

Viscosity (Pa s) ∞ a3 _∞a3 _{54.94 ± 0.50}b2 a–c: different lowercase letters represent statistically signiﬁcant differences between the groups of different applied pressures at p < 0.05 (comparison in raw);

1–3: different Arabic numbers represent statistically signiﬁcant differences between the groups of different percentages of Mg powder loading at p < 0.05 (comparison in column); The same letter or number indicates that the values are not signiﬁcantly different.

Fig. 5. The results obtained from the printing trials of Mg scaffolds: (a)–(c) printed single-layer struts with the printing process parameters corresponding to points A–C in the printability windows ( Fig. 4 ), respectively, (d)–(f) the representative samples showing samples with deﬂected surface, accurate alignment and excessive defects, printed with the process parameters corresponding to points D-F in the printability windows ( Fig. 4 ), respectively, and (g),(h) the microstructures of the printed struts initially with 54, 58, and 62 vol% Mg powder loading, respectively.

(10)

Sintering temperature ( °C)/dwelling time (min) 650/35 650/10 650/5 660/5 670/5 Relative density of struts 70% ± 6% a _{61% ± 4%}b _{54% ± 6%}c1 _{58% ± 7%}12 _{63% ± 7%}2 a–c: different lowercase letters represent statistically signiﬁcant differences between the groups of different sintering times at p

< 0.001.

1–2: different Arabic numbers represent statistically signiﬁcant differences between the groups of different sintering temperatures at p < 0.001;

The same letter or number indicates that the values are not signiﬁcantly different.

showed that the adjacent layers cohered well together and the strutsalignedaccurately(Fig. 5 e).

The differences in macro-morphology between the struts printedfromtheinkswith54,58,and62vol%Mgpowderloading couldbe observedthroughtheclose-upviewsofthestrutsinthe scaffolds(Fig. 5 g-i).Ascomparedto58vol%struts(Fig. 5 h),inthe 54vol%struts,Mgpowderparticleswerewellcoveredandbonded by excessive binder (Fig. 5 g), whilethe surface of62 vol% struts wasroughduetoinsuﬃcientcoveragebythebinder(Fig. 5 i).

The macrographs of the printable scaffold samples (Fig. 5 e) showed that the top layer of struts perpendicularly stacked on the layerunderneath accurately.Furthermore,there wasan over-hung partoutsidethe outermostwallofthescaffolds, whichwas printed without any additional supporting structure. It did not bend or deform during3D printing orsubsequent handling. The shrinkage of the printed scaffold samples was less than 5%, as comparedwiththedesignedstructure.

3.4. Debindingandsintering 3.4.1. Debinding

TGA of pure binder without Mg powder added revealed two-step weight losses (Fig. 6 a). The binder was almost totally decomposed after being heated up to 410 °C and thus above 410 °C, the curve remained steady. FTIR analysis showed that at room temperature, there was no new peak appearing in the spectrumoftheinkwith58vol%MgpowderloadingaftertheMg powder was added to the binder solution (Fig. 6 c), ascompared with the spectrum of pure binder. With increasing temperature, TGA of the as-printed sample with 58 vol% Mg powder loading (Fig. 6 b) showed a similar trendto that of pure binder (Fig. 6 a). Theﬁrststepofthermaldecompositionstartedtooccurataround 115 °C and progressively continued to 210 °C, indicating the removal of plasticizers. Indeed, the peaks stemming from the plasticizersdisappearedintheFTIRspectrumofthematerialafter the green samples had been heated to 220 °C, corresponding to the ﬁrststep debinding, ascompared to thespectrum ofthe ink at room temperature. The leftover peaks at1492, 1452,743, and 679 cm−1 corresponded to the characteristic absorption frequen-cies of backbone polymer (Fig. 6 c). The second step of binder decompositionoccurredintherangeof320to440 °C,wherethe pyrolysis of polymer occurred with the evidence of the absence of the absorption bands that displayed after the material was heatedto220°C(Fig. 6 c).Thetotalweightlossofthe58vol%Mg powder loaded samplebetween115and440°C wasfoundto be around 20 wt%. As compared with the spectrum of the sintered Mgpowderwithoutbinder,therewerenodifferencesbetweenthe scaffolds andpure Mgpowder samplesheatedto 650°C. Dueto the absence ofvibrations orrotations of organicgroups in these samples,bothofthemshowedanearlysteadyspectrum(Fig. 6 c).

UnliketheTGAcurveoftheas-printedsample,theTGAcurves of the as-received Mg powder and as-sintered samples did not show any pronounced two-step weight losses, but maintained the weight percentage at around 100% before being heated up to 400 °C (Fig. 6 b), indicating the absence of the binder in the

sintered scaffolds. Above the debinding temperature, the TGA curves of all the three samples showed slight increases prior to decreasing.TGAoftheas-receivedMgpowder andtheas-printed samplerevealedaweightgainof6.5_{± 1.3}wt%overatemperature rangeof400 to630°C andaweight gain of2.0± 0.3wt% (p<

0.01,comparedtotheas-receivedMgpowdergroup),respectively, whiletheweight gain oftheas-sintered Mgsamplestartedfrom 500°Candendedat625°Cwithaweightgainof2.8± 0.4wt%(p < 0.01,comparedto the as-receivedMgpowder group)(Fig. 6 b). XRD showed that only the pure Mg phase was present in the as-receivedMgpowderandintheas-printedMgsample,whilein additiontothepure Mgphase,1–2wt%MgOphasewasdetected in the as-sintered sample (Fig. 6 d). From the elemental carbon measurement,therewas0.056 ± 0.004wt%carbonresidueinthe sinteredMgsamplesaftersintering at650°C,compared to0.011 ± 0.006wt%carbonmeasuredfromtheas-receivedMgpowder(p <0.001).

3.4.2. Sintering

The effects of the sintering temperature and time on the ﬁdelity of sintered 58 vol% Mgpowder loaded scaffolds and the sinteringneckevolution within strutswereexamined underSEM andopticalmicroscope. Duringsinteringat 650°C,necksformed within struts in scaffolds (Fig. 7 a2–e2). The sintering results of the Mg scaffolds affected by the sintering temperature and time could be seen from the cross-section close-up micrographs of the struts (Fig. 7 a4–e4). Near-net struts were created with necks formed between Mg powder particles and the relative density of struts (excluding the nodules to be described below) increasedwithincreasing holdingtime andsinteringtemperature (Table 3 ).

Interparticleneckformationproceededgreatlyaftersinteringat 650°Cfor35minandsomepowderparticleswerefoundtohave merged together to form a larger body (Fig. 7 a2). Reducing the holdingtime enabledthealleviationofthemergingphenomenon of Mg powder particles, but interparticle necks were still well maintained. The growth of necks, the reduction of micro-pores, and powder particle merging tended to progress further as the sinteringtemperatureincreased(Fig. 7 d2ande2).

Due to theexcessive flow of molten Mg andpoor wettability betweenmoltenMgandsolidMgpowderparticles,somenodules appeared in the fabricated Mg scaffolds (Fig. 7 a1–e1) sintered undercertain sinteringconditions. Largenodules could be found onthesurfaceofthestrutsandbetweenthestrutsofthescaffolds sintered at 650 °C for 35 min (Fig. 7 a1 ). The nodules could be large enough to fill the space or macropores between adjacent struts, which could be observed from the cross-section view of thesinteredsamples(Fig. 7 a3).Thesizeofthenodulesdecreased asthe holding time decreased, while no nodules were observed when the holding time was only 5 min (Fig. 7 c1–c4). However, they appeared whenthe temperaturewas increasedfrom650 to 660 °Candeven to670 °C,although theholding time wasonly 5 min(Fig. 7 d1 andFig. 7 e3). Many of the macroporesbetween thestrutswerefilledbysphericalnodulesaftersinteringat670°C, adverselyaffectingthescaffoldfidelity(Fig. 7 e3).

(11)

Fig. 6. The characteristics of the decomposition and interaction in the 58 vol% Mg powder loaded scaffold during fabrication process: (a) TGA results of pure binder, (b) TGA results of as-received Mg powders, as-printed and as-sintered scaffolds, (c) FTIR analysis results after the ink was subjected to heating at different temperatures and (d) XRD results of as-received Mg powders, as-printed and as-sintered scaffolds.

(12)

Fig. 7. The sintering behavior and ﬁdelity of the scaffolds initially with 58 vol% Mg powder loading in relation to sintering temperature and holding time.

3.4.3. Hierarchicalstructureofscaffolds

The macrograph of the sample after sintering at 650 °C for 5 min (Fig. 8 b) showed the retention of the shape with high ﬁdelityascomparedtothe 3Dprintedsample(Fig. 8 a). Thestrut size aftersintering was 323 ± 15 μm.The designedmacropores remained open after sintering, as conﬁrmed by the 3D recon-structionofμCTimages(Fig. 8 c).Theporesizewas476± 11 μm. Themicro-channelsandsurroundingnecks withinthestrutswere clearly visible from the reconstructed models (Fig. 8 c) and the microporesinthestrutswereinarangeof19to100μmaccording to the calculation of ROI from the μCT images. The porosity of thesinteredsampleswas78.4%.Bothmacroporesandmicropores

aremostlyinterconnected (Fig. 8 c). Theshrinkageofthe sintered sample,relativetothe3Dprintedsample,was2.0± 0.8%.

4. Discussion

Inthisstudy,fortheﬁrsttime,anextrusion-based3Dprinting technique wasutilized to manufacture porous Mg scaffolds. The major challengesinvolvedin applyingthistechnique, namely the limited number of binder choices and poor sinterability in the absenceofexternalpressure,wereovercome.Theseachievements openupmanypossibilitiestoexploitthisapproachfurther.

(13)

Fig. 8. The macrographs of the fabricated samples: (a) 3D printed sample, (b) sample sintered at 650 °C for 5 min, and (c) reconstructed sintered sample from μCT images.

4.1.Inkpreparation

ThecoreoffabricatingMgscaffoldsthroughtheSC-3DPprocess isthedesignofMgpowderloadedinksthatshouldbechemically stableandexhibit requiredrheological propertiesto allowthe3D printingofdefect-freeMgscaffoldsunderambientconditions.The selectionof binder,the rheologicalproperties oftheink, andthe relationship between the ink and printing process are discussed below.

4.1.1. Selectionofbinder

The selectionofaproperbinderisa challengingtaskandalso acritical stepprior tothe fabrication ofMg-basedscaffolds with theSC-3DPtechnique.First,consideringtheintrinsicpropertiesof Mgpowder, any water-basedbinder should be excluded. That is becauseMgpowderreactswithanaqueousbinderandproducesa greatamountofhydrogenintheinkfeedstock.Bubbles were eas-ilyobservedwhenMgpowderwasmixedwithawater-basedPVA binder (Fig. S1, see the Supplementary material), which resulted indiscontinuousextrusionandseveredefectsinthemanufactured struts.Second, anypossible interactions betweenthe Mgpowder andtheby-products ofpolymeric binderduringdebindingshould becarefullyconsidered.Forinstance,commerciallyavailable PLGA (polylactide-co-glycolide)has been widely utilizedas a binder in combinationwithvolatilesolventsforproducingvariousother ma-terials,suchasiron[43] ,nickel-basedalloys[45 ,46] ,tungsten[49] , graphene[50] ,andceramic[51] usingtheSC-3DPtechnique.Inthe caseoftheMgpowderloaded inkwithPLGAasthebinder, how-ever,wefoundthat duringthedebindingprocess,eithertheester group(-COOH-)ofPLGAorthermally-evolvedPLGAfragments ad-heredtothesurfacesofMgpowderparticlesinevitablycontaining the hydroxyl (-OH) group, probably leading to the formation of an intermediate compound, magnesium carboxylate (Fig. S2, see theSupplementary material).Similarreactions werefoundonthe surfacesofthealumina[52] andleadtitanate[53] sampleswhere poly(methyl methacrylate)(also containingthe ester group)was involved.ThisreactionproducedorganicresiduesanchoredonMg powderparticlesurfaces(Fig.S3,seetheSupplementarymaterial), resultingintheretentionofnon-volatilecharonMgpowder par-ticlesurfacesandhinderingthesinteringofMgpowder,although metalcarboxylatewasnotstableathightemperatures.

Finally,apartfromtheconcernsaboutthepossibleinteractions betweenthe binder and the Mgpowder, the temperatureof the pyrolysisprocessandtheamountofbinderresidueshouldalsobe considered,becausethe applicablesinteringtemperatures ofpure Mgwithameltingpointof650°Carerelativelylowascompared tothoseapplicabletoothermetals,suchasironandnickel.

Based on the above considerations and the lessons learned from many experimental trials, we designed a graded volatile binder system that consisted of polymer, volatile solvents, and additives for fabricating Mg scaffolds. From the results obtained fromtheFTIRanalysisandTGA,itwasconﬁrmedthatthisbinder systemdid not react withthe Mgpowder during 3D printingor debinding. This is because that unlike water-based PVA binder, the solvent we used is a kindof water-free organic solvent,and unlikePLGAcontainingbinder,thebackbonepolymerusedinthis studydoesnotincludethefunctionalgroup thatmightreactwith Mg powder. After debinding, almost no binder residue retained, whichcouldotherwisehavehinderedthesinteringofMg(Fig. 6 ). Moreover, the carbon residue results showed that there is very little carbon residue(0.056%) in the sintered samples. Therefore, the chosen binder system would not affect further sintering process.

4.1.2. Rheologicalpropertiesoftheinks

Theviscoelasticpropertiesofaviableinkareofcritical impor-tance for the success of the SC-3DP approach. A high degree of shear-thinning behavior is particularly important, which enables theink material tobe extrudedthrough finenozzles, onthe one hand,andtopossessasufficientlyhighrigidityforshaperetention on the other hand [54] . Meeting these criteriarequires a proper control of ink formulation andrheological propertiesto generate a stable suspension that promotes the “fluid-to-gel” transition, therebyensuringtheretentionoftheprintedshapeandthefusion with the previously deposited layer. The fluid-to-gel transition can be achieved by applying different approaches to different 3D printingtechniques, such astemperature-controlledtransition

[55 ,56] ,gellingofprecursorsolution[57] ,andsolventevaporation, whichwasappliedtothepresentprintingtechnique(i.e.,SC-3DP). Solvent evaporation is the most commonly used approach for increasing the rigidity of materials right after extrusion, and offersa numberofbeneﬁts,suchassimplicity,room-temperature

(14)

tomaximizeitsprintability.

Solidsloadingiscommonlyusedtoquantifytheamountof sus-pendedsolidsinasubstance.Inthepresentstudy,thisisreferred to as the amount ofMg powder loaded intothe binder to form feedstockorink forSC-3DP. Thevolumefractionofsolidsloading (Ø)iscloselyrelatedtotheviscosityofasphericalparticleloaded ink [58] .When Ø increasestowards a maximumpackingfraction (Øm),therelativeviscositytendstobeinﬁniteandthesuspension

exhibits yield stress behavior. Usually, spherical particles with unimodal sizes, no matter what their sizes may be, can achieve a typical value of Øm= 0.64 for a random close-packing [59] . This was conﬁrmed experimentally by printing the 64 vol% Mg powder loadedink andasexpected, itsﬂow wastotallyhindered even though themaximum pressureof theprinter,600 kPa, was applied. Therefore,54, 58,and62 vol% Mgpowder loading were usedinthepresentstudyfortheoptimizationofthesolidloading intheinks.

Inthe presentcase, therheological behaviorofthe inkscould be tailoredby varying thepercentageof Mgpowder loading(54, 58,and62vol%)(Fig. 3 ).Theresultsobtainedfromtheshearstress sweep, frequencysweep,andcreep-recoverytestsall showedthe viscoelasticpropertiesoftheinks.First,intheshearstresssweep test, the obtained yield stress revealed an “at-rest” viscoelastic characteristic, below which the material behaved as a solid. It is reported that the “yield behavior” really exists in highlyfilled viscous polymers[59] .Asuspension composed ofa low-viscosity fluidorequivalentlyahighconcentrationofsolids(Ø>0.5)usually exhibitsa highyield stress that mustto be exceededin orderto initiate the flow, which is attributed to the particle interaction forces within materials [59] . However, it has been observed that suspensions with particles larger than 10 μm in diameter (non-colloidalsuspensions) orvolumefractionlessthan 0.3usuallydo not showasignofyieldstressduetothelesssignificant interac-tion forcesamonglarger particles. AlthoughMgpowder particles withdiametersinarangeof25to80μmwereusedinthisstudy, the “yield behavior” wasobvious fortheMg powderloaded inks (Fig. 3 ),which wasmostlikelydueto theadditionofplasticizers to the inks. The plasticizers could activate the surface of solid powderparticlesand,therefore,increasetheinterfacialinteraction between powder particles [60] , which promoted the viscoelastic propertiesof theMg powderloaded inks. Theyield stress ofthe inkwithahigherpercentageofMgpowderloadingishigherthan that of the ink with a lower percentage of Mg powder loading, indicatingthat ahigherprintingpressureisrequiredforinitiating theflowoftheinkwithahigherpercentageofMgpowderloading (Fig. 4 ).

Second, in the frequency sweep test, all the three inks con-formed to the G’> G’’ relationship in the linearregion (Fig. 3 c). Thephase angle,

δ

(tan

δ

=G’’/G’),isanindicator oftheelasticor viscousbehaviorofanink.When

δ

fallsinarangeof0°<

δ

<90°, the materialpresents viscoelasticbehavior, inwhich thematerial with

δ

>45° ismoreviscousandthatwith

δ

<45° ismoreelastic

[61] . Therefore, the inkswith various percentages ofMg powder loading showed dominantly elastic behavior featured by a phase angle smaller than 45°. In the presentstudy, a minimum elastic modulus G’value that is required to have a midpoint deﬂection ofnogreaterthan5%ofthestrutdiameterwasconsidered,based onasimpleelasticbeammodel[47] .Thisassessmentwaswidely adopted by other researchers toevaluate therigidity ofthe inks to support the entire structure [61-63] . A minimum G’ value of 7.05–7.28 Pawasrequired,based on thecalculation withthe ink densityvaluesof1.50–1.55g/mlfor54–62vol% Mgpowder load-ing,410μminstrutdiameter,and390μminspacingdistance.The

G’valuesofall theinksobtainedfromthefrequencysweep tests

Finally,the viscoelastic propertiesof the inks were conﬁrmed bythecreep-recoverytests,whereaquickviscosityrecoverywith no shear stress applied was required (Fig. 3 d). All the prepared inks showed viscoelastic characteristics with about 25% recov-ery degree. The shear thinning of the inks is a prerequisite for extrusion-based 3D printing. All the preparedinks exhibited the shear-thinningbehavior(Fig. 3 c),andthedegreeofshearthinning did not show large differences between the inks with different percentagesofMgpowderloading(Table 1 ).Thisagreeswiththe ﬁndings ofother researchers, showingthat n remains unchanged upon the additionof various volume fractions ofspherical solids duetothepresenceofparticulates[64] .

Theviscosityofthesuitableinksforextrusion-based3D print-ing variescase by case. In the presentstudy, the shearrates for extrudingdifferentinksatvariousapplied pressureswerederived fromthemeasuredﬂow rates.Thecorresponding viscosityvalues oftheinksexperiencingextrusionforsuitable 3 Dprinting(Fig. 4 ) were between 1 and 60 Pa_s (Table 2 ), which is in the same viscosityrangeasreportedbefore[61] .

Although the rheological properties are of help in gaining a fundamentalunderstandingoftheprintabilityoftheinks,thereal printingprocess issigniﬁcantly morecomplex than what can be captured from the shear-viscosity measurements. Nevertheless, rheologicalevaluationcanprovideaguidelineforinitialscreening oroptimizing the suitable inks prior to time-consuming trials of 3Dprinting. Indeed,inthepresentresearch,therheological char-acterizationofthepreparedinks withthree differentpercentages ofMgpowder loadingprovidedtheoretical supportfordescribing theprintability andprintingbehavior, whichwillbe discussed in thenextsubsection.

4.2.3Dprintingprocess

The dimensional accuracy of material deposition during the 3Dprintingprocessisstronglydependent ontheflowrateofthe ink (i.e., the extrusion rateof the ink from the nozzle underan external pressure).The flow rateis influenced primarily by three factors,namelythephysicalandrheologicalpropertiesoftheink, the geometry of the nozzle, and user-imposed conditions, such asthetemperature, whichwasnotincludedinthepresentstudy. First, under the same applied pressure, the flow rate of the ink witha lower percentage ofMgpowder loadingwas found tobe largerthanthatoftheinkwithahigherpercentageofMgpowder loading(54 vol% > 58vol% > 62vol%). Thatis becausethe flow rate is usually inversely proportional to viscosity. The viscosity valueoftheinkattheshearrateinducedbyextrudingwasinthe ascendingorder:54vol%<58vol%<62vol%(Table 2 ).Combined withthe printability window (Fig. 4 b), the appropriate flow rate forprintingstrutswitha fidelitytolerance below25%(the width ofstrutsinarangeof300to500μm,whena410μmnozzlewas used) wasat a level of 10−4 to 10−3 ml/s with printing speeds ina rangeof 1to 18 mm/s(Fig. 4 a). A widerange offlow rates (i.e., 0.07 nL/s - 4 mL/s) were reported for various inks suitable forextrusion-based3Dprinting[65] .Theflowratesinthepresent case were, therefore, considered reasonable. Second, the nozzle sizeandgeometrycouldaffecttheflowrate.Ataperednozzlewas usedfor3D printinginthepresentstudy,sinceitisbettersuited forachievinghigh-pressuregradients or higherflow rates thana cylindricalnozzle[66] .

Theapplicableprintingparametersaredeterminedby theﬂow rateofthe ink. Fora givenink, asﬂow rateincreasesata given printingspeed, nozzlesize,andgeometry,thewidthofthe resul-tantstrands increases.Inthiscase,an increasedprintingspeed is neededtoensurethestrutswidthremainswithinadesiredrange.

(15)

Thiscorrelationcangiveareasonableexplanationfortheinclined region of the printability window, especially zone I for 2D strut printing.Theselectionofsuitableprocessparametersthatcouldbe usedforsuccessfulfabricationof3Dscaffoldsismuchstricterthan that for2D printing, since maintainingstructural integrity would haveto be considered in the 3D printing process (Fig. 4 b). Two criteriawouldhaveto be metfor structuralintegrity: (i) smooth transitionsbetweenadjacentlayers,whichrequiresenough liquid-ityfordepositedlayerstoenableseamlessfusionwiththe subse-quentlayers,and(ii)enoughrigidity offirstly depositedlayersto supportthe wholestructure, which requiresrapid evaporationof solventsrightafterextrusion.Itwasobservedthattheevaporation oftheinksplayedanimportantroleinmeetingthesetwocriteria forbuildingthedesignedstructurewithoutdistortions ordefects, although the presence of additive (i.e., the remaining low vapor pressuresolvent)alreadyimpartedtheliquiditytoacertainextent. Atahighprintingspeed,pointDinthe54and58vol%windows led to insufficient evaporation of solvents from the deposited layers,resulting inthe slumpsofthe structure (Fig. 5 d). Alower printingspeed (forexample,point Finthe58vol%window) ora thinstrutresulting froman insufficientapplied pressure(pointF inthe62vol%window)resultedintoomuchdryingofthelayers and,thus,the unmergedordiscontinuouslayers withpoor adhe-sion(Fig. 5 f).Intheprintabilitywindowsofthe54and62vol%Mg powderloadedinks,theprintablezoneswerelocatedatlow print-ingspeeds(below5mm/s)forthedesigned3Dscaffolds(Fig. 4 b). The54vol%Mgpowderloaded inkneededmoretimefor evapo-rationthantheinkwithahigherpercentageofpowderloadingto reachthesamelevelofrigidity,becausetheformerhadmore sol-vents.Alowerprintingspeed meansalongertime forthenozzle toreach thesame displacementand, therefore,a longer time for the evaporation of the deposited struts before the deposition of nextlayers. Bycontrast,theapplicable printingspeed fortheink with62vol%Mgpowderloadingwaslimitedtoaverylowspeed duetoaverylow flowrate.Incomparisonwiththeinkswith54 and62vol%Mgpowderloading,theinkwith58vol%Mgpowder loadingcould beprintedoverthewidestrangeofprintingspeeds (i.e.,4–12mm/s).

In summary, the ﬂow rate (affected by the applied pressure), the evaporation rate (a physical property of the ink) and the depositionrate(determined byprinting speed)mustbe matched witheachother forprintingthedesired3Dstructure successfully. The established operating window can be used as a quick guid-anceforfabricating 3Dporous scaffoldswithstructuresizesona millimeterscale.Itisimportanttonotethattheoperatingwindow depends on the size and geometry of the designed structure in theevaporation-controlled printingprocess. In thepresent study, theprinted structure showeda highﬁdelity without defects and the side pores remained open even over a long spanninglength (Fig. 8 a). The main limitationof the printing process is that the inksmustbeprintedwithin1.5–2hafterhavingbeenloadedinto the cartridge due to the continuous evaporation of the inks in thecartridge and, thus, an alteredviscosity that could affectthe printingprocess.

Apartfromtheconsiderations regardingtheprintability ofMg powderloadedinks,ahigherpercentageofMgpowderloadingis beneﬁcial for the subsequent step (i.e., debinding andsintering), asahighrelativedensityofstrutscanbeobtainedandthe reten-tionof strut/scaffold shape can be ensured after binder removal. However,theinkwith62vol%Mgpowderloadingisnotthemost promising one, because of its narrow printability window and low printingeﬃciency(i.e., the low permissibleprinting speeds). Therefore,theprinted scaffoldswith58 vol%Mg powderloading were chosen for further study on the debinding and sintering behaviorofthescaffolds.

4.3. Debindingandsintering

DebindingandsinteringofporousMgscaffolds areevenmore challengingthanthe3Dprintingstep.Themainchallengeinvolved in debinding is mostly related to the binder selection that has already been discussed earlier. TGA and FTIR analysis (Fig. 6 ) conﬁrmed the proper choice of the binder, as the vast majority of the binder had been degraded before reaching the sintering temperature and no interaction was found between the reactive Mg powder and the pyrolysis products of the binder during the debindingprocess.

Apartfromtheadvantagesoftheseamlessattachmentoflayers during 3D printing, a binary-component binder system is also beneficialfordebinding.Theplasticizerwasremovedfirsttoopen the paths for further diffusionof the degraded gaseous products (Fig. 6 ). The major polymer, which served as backbone, retained thestrut shape untilbeingalmost totallydecomposedbefore the sinteringtemperature, 650°C,wasreached, withlowashresidue (Fig. 6 ).Around20wt% ofthebinder wasthermallydecomposed duringthe2-stepdegradation, whichwasnearly allthebinderin theinitialfeedstock(Fig. 6 b).Asthebackbonepolymerwasmostly removed before 450 °C was reached, the ash or carbon residue contentwasofparamountimportanceforretainingtheshapeand preventingthe scaffolds fromslumpingand distortingbefore the formation of the inter-particle necks occurred during sintering. Such a strategy (i.e., a multi-component binder system) is often used in the debinding of metal injection molding (MIM) [67] . A heatingrateof5 °C/minwaschosen fordebindingandsintering, whichwasrelativelyhighascomparedwiththattypicallyapplied in MIM. In general, a high heatingrate is not preferred in MIM because integrity lose may occur during the debinding process whensuddenevaporationofabinderfractionoccurwithnoclear exitpath.Insolvent-cast3Dprintingappliedinthepresentstudy, however,the backbone binder accountedonly for37 vol% of the binder andthe restwere solvent andplasticizer. As a result, the evaporation of the solvent and the decomposition of plasticizer provided sufficient path for the evaporation of polymer. Sudden internalstress thatmaycausethecollapseofthestructurewould unlike occur. In addition, the carbon residue on the surfaces of Mg powder particles played a vital role in increasing the fiction betweenMgpowderparticles.Inotherstudiesonsolvent-cast3D printingofmetals,scaffolds couldbe fabricated successfullyeven ataheatingrateof10°C/minduringthedebindingprocess[45 ,46] . ConsideringarelativelylongdebindingtimebeingtypicalinMIM, asingle-stepprotocolwasemployedinthepresentstudyby com-bining debinding and sintering, during which no dwelling time fordebindingwasallocated.The relativelyquickdebindingmight be attributed to the pressure-free thin struts that enabled fast diffusionpaths for pyrolysisproducts to move tothe green body surfaceswherethedecomposedbinderunderargongasflowcould quicklyescape[69] .Thisone-steptreatmentgreatlyincreasedthe fabrication efficiencyand reduced the operational complexity, as comparedwiththeusualpracticeinthecaseofMIM.

The changes of the Mg powder throughout the fabrication process were also studied. XRD revealed that the Mg powder in the as-printed samples remained the same as the as-received powderinphase constitution(Fig. 6 d), whichwasalsoconﬁrmed by theFTIRresults(Fig. 6 c).Inthesintered sample,however,the MgOphasewasfound.InTGA,above400°C,weightgainoccurred to all the tested samples (Fig. 6 b) due to the oxidation of Mg. In another study, it also observed that the oxidation of Mg was accelerated above a critical temperature (i.e., 400 °C) at various heatingrates evenin an ultra-high-purity Argas atmosphere(Ar purity: 99.9999%) [68] . After being heatedabove 400 °C inTGA, theas-printedsampleappearedtobelessoxidized(withaweight

(16)

Fig. 9. A schematic illustration of liquid-phase sintering of the Mg 3D scaffolds: (a) proper liquid-phase sintering and (b) excessive liquid-phase sintering.

gain of 2.0 wt%) when compared with pure Mg powder (with a weight gain of 6.5 wt%). This was probably because a small amountofcarbonresiduefromthebindercovered somesurfaces of Mg powder particles in the as-printed sample, while the as-received Mg powder particles possessed more exposed surfaces. Theweightgainoftheas-sinteredMgsamplewasalsolowerthan that oftheas-receivedMgpowder overthetemperaturerangeof 400 to 650 °C.This wasbecause the as-sintered Mgsample had already been oxidizedonce during the sintering process prior to XRDanalysis, whilepureMgpowderparticles stillhad“relatively clean” surfaces to react with oxygen during XRD test. Therefore, compared withthe as-receivedpure Mgpowder, the Mgpowder in the as-printed Mg scaffolds did not change chemically after being mixed with the binder, while that in the as-sintered Mg scaffoldspresentedonlysmallincreasesinMgOcontent(1–2wt%) andcarboncontent(0.045wt%).

Sintering of the porous Mg scaffolds was challenging, as the presence ofoxidelayers on Mgpowderparticle surfacesseverely retarded the sintering process. Unlike other metals, such as iron and copper [70] , the diffusion coeﬃcient of Mg atoms (i.e., 5.25× 10−24 _at₆₅₀_°C)_through_the_oxide_layer_is_about₁₂_orders

of magnitude lower than their self-diffusivity (3.01 × 10−12 _at

650 °C) [71] . Furthermore, the magnesium oxide (MgO) layer on Mg powder particle surfaces is rather stable even beyond the evaporation temperature of Mg. In addition, there is no real possibility of reducing MgO into Mg during sintering. Therefore, solid-state sintering of loose Mg powder is not feasible and breaking uptheoxide layeristheonlywaytoallow foreffective sintering kinetics, which will provide paths for the diffusion of Mgatoms. Withrespectto this,the compactionofMgpowderis usually used inorder to mechanically rupturethe oxide layer in theconventionalpowdermetallurgytechniques.However,the sin-teringofthedesignedporousscaffoldsexcludesthepossibilityof applyingexternalmechanicalpressureonthe3Dprintedsamples. Nevertheless, liquid hot isostatic pressing (HIP) which has been once tried to consolidate a magnesium alloy solid part [72] may bestillapplicable.

Liquid-phase sintering offers a possibility for the sintering of loose Mgpowder, inwhich thepresence ofa smallvolume

frac-tionoftheliquidphasefacilitatesthedisruptionoftheoxidelayer and enhances the diffusion. This strategy has been also utilized inthesinteringofZK60scaffoldsfabricated bypowder-bedinkjet 3D printing [71] . The control of the presence of a solid-liquid coexistencezone enabledthe liquidfraction inthe ZK60 alloyto varyfrom14to43vol%overatemperaturerangeof535to610°C, whichgaveawiderangefortheselectionofsinteringtemperature. IncomparisonwiththatMgalloy,itismuchmorediﬃculttotake advantage ofthe formed liquidphase forthe sintering ofwith a liquidmonophasezone. Thatis becausesinteringatorabove the meltingpointofMgisnecessarytogeneratetheliquidphase,but itinvolves therisk ofdisrupting theshape ofthe specimensand evenatotalcollapseofthescaffolds.

Withthe above considerations, in the present study, we pro-posedashortsinteringprocesstoovercomethechallengeinvolved inpure Mgsintering.Itwasfoundthat a5minholdingtimewas enoughtoforminter-particlenecksat650°C.Generally,650°Cis themeltingpointofpureMg. Althoughthesinteringtemperature reachedthemelting point ofMg, theshape of scaffoldswasstill wellretainedwithoutanycollapse(Fig. 8 ), owingtothepresence ofaMgO filmonthesurfacesofMgpowderparticles.Adepicted model(Fig. 9 ) displayshowpure Mgcanbe successfullysintered withinashortholdingtime.Beforetheoccurrenceofsintering,an oxidelayer coveredMgpowder surfaces,whichseverely impedes diffusion(Fig. 9 a).Duringheatingtowards650°C,theliquidphase appearedalonggrain boundariesorsome defectsinsideMg pow-derparticles.AstemperatureincreasedtothemeltingpointofMg, agreater volumefractionoftheliquid phaseformed inside pow-der particles. The volumetric expansion of Mg powder particles causeda gradual increase ofinternal pressure inside Mgpowder particlesuntiltheoxidelayerruptured.The cracksonthesurface ofpowderparticles (Fig. 7 c2 ) providedshort-circuitpaths forthe interiorliquid Mgto seep onto the surfaces ofpowder particles. Thereleasedliquidwasthendrawnintotheinter-particleneck re-gionsduetothecapillaryforce,leadingtotheformationofnecks. Insummary,theenvelopingMgOfilmenabledtheretentionofMg powderparticlesandscaffoldshape,whilethecracks oftheMgO filmgavethemeltachancetoflowoutandconsequentlytoform necks.