j ourna l homepage: www.e lsev ie r .com/ locate /scr ip tamat
Viewpoint article
Non-beam-based metal additive manufacturing enabled by additivefriction stir deposition
Hang Z. Yu a,⁎, Mackenzie E. Jones a, George W. Brady a, R. Joey Griffiths a, David Garcia a, Hunter A. Rauch a,Chase D. Cox b, Nanci Hardwick b
a Department of Materials Science and Engineering, Virginia Tech, 445 Old Turner Street, Blacksburg, VA 24061, USAb Aeroprobe Corporation, 200 Technology Drive, Christiansburg, VA 24073, USA
Article history:Received 8 February 2018Received in revised form 13 March 2018Accepted 16 March 2018Available online 7 April 2018
Beam-based processes are popularly used for metal additive manufacturing, but there are significant gaps be-tween their capabilities and the demand from industry and society. Examples include solidification issues, aniso-tropic mechanical properties, and restrictions on powder attributes. Non-beam-based additive processes arepromising to bridge these gaps. In this viewpoint article, we introduce and discuss additive friction stir deposi-tion, which is a fast, scalable, solid-state process that results in refined microstructures and has flexible optionsfor feed materials. With comparisons to other additive processes, we discuss its benefits and limitations alongwith the pathways to widespread implementation of metal additive manufacturing.
Among the seven types of additive manufacturing (AM) technolo-gies classified by the American Society for Testing and Materials(ASTM) [1], the beam-based technologies, i.e. powder bed fusion and di-rected energy deposition, represent themain approaches for fabricatingmetals and alloys today. These are high-temperature and high-energyprocesses, in which a laser or electron beam is applied to selectivelymelt powders that are either pre-deposited to form a powder bed or de-livered by powder feed nozzles [2–9]. The interactions of the high-energy beam and powders cause a series of complicated physical pro-cesses, including powder melting, dynamic melt flow, and rapid solidi-fication [3,10]. Because the microstructure of the raw powders isdestroyed during melting, the microstructure of the as-manufacturedpart is mainly determined by the rapid solidification process, with acooling rate around 103–107 K/s [4]. The last few years have borne wit-ness to exciting advances in the research of beam-based technologies,including microstructure manipulation and quality control [2–18].
Even with these advances, significant gaps still exist between whatthe high-cost, high-energy, beam-based AM technologies offer andwhat society and industry need. First, beam-based metal AM processesare energy inefficient and incapable of reliably fabricating non-weldablealloys, such as 2xxx or 7xxxAl alloys [19,20]. In addition,most structuralapplications require isotropic mechanical properties; because of
ier Ltd. All rights reserved.
epitaxial solidification, however, beam-based AM processes generallylead to highly orientated, columnar grains with anisotropic mechanicalproperties [13,21]. Moreover, future critical applicationswill necessitatethe use of high performance alloys with specific compositions, such astwinning induced plasticity steels and quenching and partitioning steels[22–25]. To fabricate these alloys with beam-based AMprocesses, high-quality powders with the desired composition, shape, and size distribu-tion are required [26–28]. Making suitable powders is already time-consuming and expensive, let alone the subsequent efforts needed todetermine the processing conditions for optimal part quality. These fac-tors severely limit the viability of beam-based technologies formanufacturing large-scale, high-quality parts with consistent composi-tion and isotropic properties.
The above limitations stem from the nature of beam-based pro-cesses:melting of high-quality powders followed by rapid solidification.Are there alternativemetal AMapproaches that avoid these limitations?Scientists and engineers have developed several non-beam-based,solid-state additive processes [29–33], and some are promising forwidespread use. The metal AM research community is starting to ac-knowledge these alternative processes. For example, the MS&T 2017conference featured a symposium titled ‘Non-beam-based additivemanufacturing approaches for metallic parts’ [34].
In this viewpoint article, we introduce and discuss additive frictionstir deposition, which is an emerging low-temperature and low-cost ad-ditive process that consistently produces a ‘wrought microstructure’(i.e. result of thermomechanical processes) rather than a ‘cast micro-structure’ (i.e. result of solidification). It enables fast, scalable
123H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
manufacturing for a broad range of engineering alloys and composites[33,35,36]. While most solid-state additive processes are based onsheet lamination [30,37], this process enables near-net shaping viasite-specific deposition. The size of the site in this context is dependenton the in-plane resolution of the process. As a new technology that re-cently entered the public eye, however, its research is still in an elemen-tary state with many fundamental aspects not fully understood. Thisviewpoint article aims to provide an overview of the status of additivefriction stir deposition, to discuss its benefits and limitations throughcomparisons to other AM processes, and to stimulate discussions onthe pathways to widespread implementation of metal AM.
2. Basic physical processes
2.1. Process description
Additive friction stir deposition enables solid-state additivemanufacturing of metals andmetal matrix composites [35,36]. The cen-tral component of the system is a hollow shoulder, through which thefeed material in the form of either a solid rod or powder is delivered[38]. The shoulder rapidly rotates and generates heat through dynamiccontact friction at the shoulder-material interface and material-substrate interface (Fig. 1 (a)) [32,33]. Heated and softened, the fillermaterial is fed through the tool and bonds with the substrate throughplastic deformation at the interface. The transversemotion of the shoul-der results in deposition of a single track of material, typically several
Fig. 1. (a) An image of the rotating shoulder during additive friction stir deposition and a schconditions of heat flow: a comparison between the stirred material in friction stir welding and
hundreds of microns thick. The first layer is formed by the tool travelingacross the surface of the substrate; by selectively adding subsequentlayers upon the initial one, 3D parts are made.
The first commercialized technology of this type was developed andpatented by Aeroprobe Corporation, which has a proprietary version ofthe process [36]. To highlight its deposition nature and distinguish itfromother friction stir processes,we refer to the process as additive fric-tion stir deposition throughout the paper.
2.2. Underlying materials science
Additive friction stir deposition is a thermomechanical process. Im-portant processing parameters include the shoulder rotation frequencyΩ, shoulder normal force P, filler material feed rate R, layer height h, andtransverse speed VTra [35,38]. These parameters control the heat flowand material flow processes, which are fully coupled. For the depositedmaterial, the heat generation, dissipation, and transfer mechanisms aresimilar to the stirred material in friction stir welding or processing. Inboth cases, heat is generated bydynamic contact friction betweenmate-rial and tool, dissipated by severe plastic deformation of the material,and transferred inside the material by thermal conduction and thermalconvection via material flow. The governing equation of heat transfer isexpected to be similar for the two cases [39]:
∂∂t
ρCPT þ ∇ � ρ u! CPTð Þ þ ∇ � κ∇T ¼ q�: ð1Þ
ematic to highlight the basic physical processes in the deposition region. (b) Boundarythe deposited material in additive friction stir deposition.
124 H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
Here ρ is thematerial density, CP is the specific heat capacity, T is thetemperature, u! is the material velocity, and κ is the thermal conductiv-ity. q � is the volumetric heat generation or dissipation rate. There are,however, significant differences in tool geometry and processing config-uration between the two cases, which lead to different boundary condi-tions. This point is addressed in detail in Section 2.3, where we willprovide a schematic to compare the boundary conditions in thesecases. Moreover, material addition (i.e. deposition) is involved in addi-tive friction stir deposition. For transient thermal analysis, this can besimulated using finite element modeling that incorporates techniqueslike quiet or inactive elements [40].
The maximum homologous temperature, which is defined as thepeak material temperature divided by the melting temperature in Kel-vin (Thmax= Tmax/Tm), is expected to lie between0.6 and0.9 [41]. The de-posited material is thus severely, plastically deformed by high strain-rate shear at elevated temperatures, which can result in dynamic recov-ery or dynamic recrystallization, depending on the Zener-Hollomon pa-rameter and the stacking fault energy of the material [42]. Dynamicrecrystallization can lead to refined, equiaxed grains and high-anglegrain boundaries. Similar to friction stir welding, the thermal historyplot (temperature vs. time) has a narrow peak in additive friction stirdeposition [39,43–45]. The width of this peak reflects how long the de-posited material is in contact with the shoulder, which is estimated tobe tens of seconds for typical tool geometries and processing conditions.Despite the grain growth following deposition, the resultant grain size isstill considerably smaller than that of the feed material in most cases.
Overall, the microstructure evolution is critically dependent on theconvolution of the intrinsic mechanical properties of the material withthe history of temperature and strain in additive friction stir deposition.Prediction of the final microstructure for a given processing conditionneeds two steps. First, a quantitative relationship must be establishedfrom the processing conditions to the temperature and strain historyof the deposited material. This can be achieved by combining existingfriction stir welding models with a model for the material depositionprocess. Multi-scale analysis will be necessary to account for both thematerial and heat flow in the local deposition zone and the global ther-mal field evolution on the track, layer, and part levels. Second, the effectof temperature and strain history on the final microstructure can besimulated using techniques like the cellular automatonmethod [46,47].
2.3. Relations to friction stir welding and processing
Additive friction stir deposition shares many of the same benefits asfriction stir welding and friction stir processing, especially in terms ofmicrostructure formation and applicability to non-weldable alloys[41,48–61]. Nevertheless, there are major differences. In friction stirwelding and processing, a shoulder with a protruding pin is spun at ahigh rate and inserted into the softened metal surface. The shoulderthen traverses the material to either join two interfaces in friction stirwelding or treat the surface in friction stir processing. However, in addi-tive friction stir deposition, there is no pin structure penetrating thebase material [35,36,41]. The different geometric configurations leadto different heat flow boundary conditions, and therefore, differentcooling rates. In friction stir welding and processing, the stirredmaterialis surrounded and constrained by the material in thethermomechanically affected zone, into which heat is transferred bythermal conduction. In additive friction stir deposition, the depositedmaterial of the first track of each layer is in direct contact with air on ei-ther side, and heat transfer perpendicular to the tool traveling directionoccurs by convection and radiation rather than conduction. Regardingheat generation, in friction stir welding or processing it is generated atthe shoulder-material and pin-material interfaces, whereas in additivefriction stir deposition it is generated at the shoulder-material andmaterial-substrate (or new layer-previous layer) interfaces. Fig. 1(b) compares the different boundary conditions in these cases.
In additive friction stir deposition, all the material undergoes thecharacteristic stir-zone thermomechanical process involving dynamicrestorationwhen it is being deposited, and themicrostructure is normallyuniform.When building 3D parts, additive friction stir deposition of thenth layer leads to re-deformation and re-heating in the (n-1)th layer.Therefore, internal regions in a part experience additional deformationand heating as compared to the free surface, and the grain size can differbetween the internal and surface layers in additive friction stir deposi-tion. In contrast, friction stir welding results in multiple zones with dif-ferent characteristic microstructures, such as the stir zone,thermomechanically affected zone, and heat affected zone [39]. Onlythe material in the stir zone undergoes significant dynamic restorationwith refined equiaxed grains, while the microstructure in other zonescan be highly dependent on the parent material. This can lead to signif-icant microstructure inhomogeneity across the workpieces.
Fundamentally, our focus here is not a welding or surface treatmentprocess. Instead, it is a deposition process formanufacturing, which com-bines the friction stir concept with a material feeding process. Note thatthere is another friction stir-based AM technology called friction stir ad-ditive manufacturing, which is a sheet lamination process. We will com-pare it with additive friction stir deposition in Section 5.2.
3. Features and capabilities
There are several key features and capabilities, including the solid-state nature, resultant wrought microstructure, feed material options,applicability to engineering alloys, and scalability, that make additivefriction stir deposition promising as an alternative to beam-basedmetal AM.
3.1. Solid-state nature
The peak temperature of an additive friction stir deposition processnormally ranges from 60 to 90% of the melting temperature [41]. Be-cause of the absence of melting, parts do not suffer from the issues asso-ciated with rapid solidification, e.g. porosity, hot cracking, elementalsegregation and dilution, aggregation of finely dispersed oxide particles,and high residual stresses [2–20]. The lowporosity and residual stress inthe as-deposited part render post-processing heat treatment unneces-sary, although post-processing for surface finish is often required.
As a solid-state process, additive friction stir deposition also providesa natural route for 3D printing of non-weldable alloys. That being said, avery high shoulder rotation frequency or a low shoulder transversespeed may lead to adiabatic heating and local melting during deposi-tion. Such phenomena have been observed in friction stir weldingwhere the melting and solidification process yields an increased num-ber of defects [62,63]. Understanding the processing-temperature rela-tionship in additive friction stir deposition is therefore critical to avoidsuch potential problems.
3.2. Characteristic microstructure
As introduced in Section 2, the thermomechanical process in addi-tive friction stir deposition is similar to the process in the stir zone offriction stir welding. The resultant microstructure in additive frictionstir deposition features the refined, equiaxed grains formed during dy-namic recrystallization and isotropic mechanical properties that arecharacteristic of the stir zone. Microstructural inhomogeneity can arisebetween the internal and surface regions of a part, because the formerexperience re-deformation and re-heating processes when the layerabove is being deposited.
Fig. 2 (a) and (b) show opticalmicrographs of themicrostructures ofInconel 625, where the grains are substantially refined from the as-received to the as-deposited state (10.3 μm to 2 μm) [41]. In contrast,beam-based processes result in columnar grains due to epitaxial solidi-fication; the resultant high mechanical anisotropy can be devastating
Fig. 2. Characteristic microstructure. Optical micrographs of (a) the as-received Inconel625 filler material and (b) the as-deposited Inconel 625 showing grain refinement.Electron backscatter diffraction maps showing (c) equiaxed grains of as-depositedInconel 625 and (d) columnar grains in the build direction of as-printed Inconel 718 bypowder bed fusion. Images reproduced with permission from references [41,64].
Table 1Representative material systems deposited by additive friction stir deposition.
Deposited Material Al 6061 Mg-WE 43 Ti-6Al-4V Nb Ta Inconel 625Substrate Material Al 6061 Mg-WE 43 Ti-6Al-4V Cu Cu HY 80
Fig. 3. Wide applicability. (a) Images of a bend test for tantalum and niobium coatingsdeposited on a copper substrate, which show strong mechanical bonding and nodelamination under bending. (b) Flow curves of the deposited material and substrate atdifferent strain rates: Cu at 0.82/s and 81/s [70], Nb at 1/s and 100/s [71], and Ta at 1/sand 100/s [71,72]. (c) A scanning electron microscopy image of an as-deposited Al-Mocomposite, which shows a relatively uniform distribution of the reinforcement particles(~30% volume fraction) in the aluminum matrix. Inset: a transmission electronmicroscopy image showing the formation of a buffer layer between the matrix andreinforcement.
125H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
for structural applications. The electron backscatter diffraction results inFig. 2 (c) and (d) compare the equiaxed grains from additive friction stirdeposition [41] to the columnar grains from selective laser melting [64].
3.3. Flexibility in feed material
Beam-based processes have restrictions on feed material in terms ofpowder composition, shape, and size. This imposes a great barrier forwidespread implementation of metal AM. In contrast, additive frictionstir deposition can use a solid rod made from cast or wrought alloys asthe feed material. Traditional atomized powder feedstock can also beused without stringent requirements on powder attributes. High qual-ity, reproducible Al 6061 and Inconel 625 parts have beenmanufacturedusing solid rod feed material, and Al-Mo and Al-SiC composites havebeen manufactured from powder feedstock [36]. To qualify as viablefeed materials for additive friction stir deposition, consistent as-deposited microstructures and properties are necessary. The compati-bility with inexpensive and mature technologies like casting providesa convenient and economic route to test the printability of feedmaterials.
3.4. Wide applicability: alloys, dissimilar materials, and composites
In addition to non-weldable materials, additive friction stir deposi-tion has shown encouraging success in fabricating a broad range of al-loys. For example, it has enabled fabrication of Al-based (1xxx, 2xxx,5xxx, 6xxx, 7xxx series) and Mg-based (AZ 31, WE 43) alloys withhigh part quality [65,66]. It has also been used to fabricate high strengthalloys such as nickel-based superalloys and titanium alloys [36,41].Table 1 summarizes the representative material systems (depositedmaterial and substrate) that have been successfully deposited by thisprocess.
In additive friction stir deposition, the substratematerial and the de-posited material do not need to be the same. When fabricating dissimi-lar materials, the thermomechanical process can lead to metallurgicalbonding and good adhesion at the interface despite the differences inchemistry, mechanical properties, and thermal properties. Examples
include the deposition of Nb and Ta coatings on Cu substrates, shownin Fig. 3 (a). No interfacial delamination is observed when applying ahigh shear stress at the interface through bending. The strong bondingbetween dissimilar materials will also allow for multi-material AM,which enables design and fabrication of functionally graded materials[67,68].We note that to ensure strong solid-state bonding, the substrateand deposited material cannot have drastically different mechanicalproperties at the strain rates typical in friction stir processes, whichcan be as high as ~100–102/s [69]. A significant difference of the flowcurves may lead to milling of the substrate by the deposited material
Fig. 4. Examples of manufactured parts. (a) A large window frame made of AA 6061.(b) (Left) A large AA 6061 pressure vessel printed in less than 2 h and machined to finalsurface finish. (Right) The internal structure of the pressure vessel shows an overhangangle as high as 54°.
126 H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
or inadequate plastic deformation of the surface layers of the substrate,which results in weak bonding. Based on the data in references [70–72],characteristic flow stress curves of Nb, Ta, and Cu are shown in Fig. 3(b) for a range of strain rates. At the similar strain rate, although the ini-tial yield strength differs considerably between the deposited material(Nb and Ta) and substrate (Cu), the difference in flow stress signifi-cantly reduces at higher strain levels. This could be an important reasonof successful fabrication of thesematerial systemsusing additive frictionstir deposition.
For all material systems investigated so far, solid-state bonding is aresult of the plastic deformation at the interface, without residentpreheating. For future research, the effects of preheating are interestingto explore. As demonstrated in friction stirwelding, it could enhance theinterface bonding, lead to less substrate deformation, and potentiallyimprove the shoulder lifetime [73]. This is especially important for de-positing high melting temperature materials or materials that are par-ticularly destructive to the shoulder.
By pre-mixing different powders and consolidating them as the feedmaterial, metal matrix composites (MMCs) can be deposited. Materialsystems like Al-Mo, Al-W, and Al-SiC have been demonstrated with re-inforcement volume fractions up to 30%. Our preliminary investigationreveals a relatively uniform distribution of Mo particles within the Almatrix in Al-Mo composites (Fig. 3(c)). Because of severe plastic forma-tion, elevated temperatures, and intimate contact, mechanically-induced mixing can occur in normally immiscible materials [74,75]. Inthe Al-Mo system, for example, a layer of Al-Mo intermetallic phase isformed at the matrix-reinforcement interface (the inset of Fig. 3(c)),which is expected to enhance the adhesion and prevent interface poros-ity. Micro X-ray computed tomography confirms that the composite isfully dense without observed voids [76].
3.5. Speed and scalability
The build rate in additive friction stir deposition is controlled by theshoulder size and filler diameter, and is comparable to that of otherlarge-scale AM processes such as wire arc AM. For example, the currenttooling of additive friction stir deposition is capable of build rates of1000 cm3/h for Al 6061 alloys with a 1/4″ diameter feed rod. To com-pare, wire-based electron-beam freeform fabrication (EBF) can have abuild rate of up to 2500 cm3/hr with a wire diameter of about 1/16″[77,78]. These build rates are much higher than beam-based processesinvolving powder melting.
Additive friction stir deposition is a free space manufacturing pro-cess like directed energy deposition, so the part size is not constrainedby a vacuumchamber or powder bed. This allows for large-scale fabrica-tion – parts with dimensions of ~1.5 m have been fabricated. As exam-ples, Fig. 4 (a) shows a large window frame, and Fig. 4 (b) shows analuminum pressure vessel printed in less than 2 h and machined tofinal surface finish. Since there is no need for atmospheric control, im-plementation of rapid field repair is also possible.
The most salient benefits of additive friction stir deposition areshown in Fig. 5.
4. Limitations and open questions
4.1. In-plane resolution
The development of additive friction stir deposition-based technol-ogy has focused on increasing the part size and build rate. As a result,the shoulder and filler dimensions are large, leading to a large in-plane feature size. While powder bed fusion can make parts with asmall feature size on the order of 0.5 mm, the in-plane resolution of ad-ditive friction stir deposition is limited to ~10 mm or above given thecurrent tooling capability. It is thus classified as a near-net shaping pro-cess rather than a net shaping process. Intuitively, the in-plane
resolutionmay be improved by reducing the size of the tooling andfillermaterial.
4.2. Overhang angle
One important metric of AM processes is the maximum self-supporting angle for making overhangs [26,79]. Compared to beam-based processes, there is less concern about material flowing off duringadditive friction stir deposition as it is a solid-state process. In fact, it hasbeen shown to successfully produce a large overhang angle of 54° with-out support material (as seen in Fig. 4(b)). However, we note that thenormal and shear forces imposed by the shoulder possibly lead to me-chanical instability and therefore a limit of the maximum self-supporting angle; experimental investigation of the overhang anglelimit can be an important direction for future research.
4.3. From computer-aided design (CAD) to part
In additive friction stir deposition, the deposition path is guided by aG-code-based hatch pattern. Since algorithms are available to automat-ically convert the CAD file to G-code, it is straightforward to go fromCADfile to part as long as the geometry is printable. Theprintable geom-etry is limited by the in-plane resolution andmaximum overhang anglein additive friction stir deposition. Given the layer-by-layer nature of ad-ditive friction stir deposition, parts may require additional post-processing to compensate for the stepwise transition at the edge ofeach layer. This is highly dependent on the feature size and partgeometry.
4.4. Buckling
A potential concern related to mechanical instability is the printingof high-aspect-ratio components, because the normal force from theshoulder may cause buckling. For fabrication of a column structure,
Euler Buckling Theory gives the critical height-to-diameter ratio Ld ¼ π
4KffiffiffiffiEσ
q[80]. Here σ is the average normal stress from the shoulder, E is
Fig. 5. A chart outlining the benefits of additive friction stir deposition.
127H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
the Young's modulus, and K is the effective height factor. We note thatthis buckling model cannot be directly applied. When a new layer isbeing deposited, the temperature of previously-deposited materialmay remain elevated. This thermal gradient can yield a compliance gra-dient and a decrease in the critical height-to-diameter ratio. As seen inultrasonic additive manufacturing, a more compliant material belowthe tool can influence the differential motion between the depositedlayer and the previous layer, therefore lowering the quality of the inter-facial bond [81,82]. Experimental investigationswill be needed to quan-titatively understand the buckling limit while taking into account theseeffects. In situ monitoring of the temperature gradient along the build-ing direction can further help to correct the bucklingmodel. For fabrica-tion of high-aspect-ratio components by additive friction stirdeposition, the authors recommend effective cooling through deposi-tion interruption between layers to avoid the compliance gradient.
4.5. Tooling development
Shoulder lifetime is of utmost concern for additive friction stir depo-sition, as the machine's performance is tied to the tool's ability to reli-ably stir material. A balance must exist between the economic viabilityand degradation rate of the shoulder. The target shoulder properties in-clude high wear resistance, high fracture toughness, high compressiveyield strength at elevated temperatures, high thermal fatigue strength,creep resistance, chemical stability, and low cost.
For low strengthmaterials such as Al andMg alloys, tool steel shoul-ders have been observed to have almost no wear degradation. For hightemperature or high strength alloys, such as Inconel or steel, tungstencarbide shoulders are used in the current equipment, with a low wearrate and a reasonable cost. For future research involving deposition ofmaterials with even higher strength or higher melting temperatures,shoulders made of advanced ceramics or composites with better prop-erties will be needed, such as hot isostatically pressed polycrystallinecubic boron nitride (PCBN). PCBN has excellent fracture toughness butis expensive and difficult to process to full density due to the covalentbonds and reversible phase transformation at high temperatures [83].
We envision a great amount of work to be done on design and process-ing of tooling materials for depositing high strength and high tempera-ture alloys.
Despite the similarities of tool requirements between additive fric-tion stir deposition and friction stir welding, the absence of pin in theformer may lead to selection preferences towards particularly toughmaterials for the shoulder. When passing the feed material throughthe hollow shoulder, the compressive normal force exerted on thefeed material leads to in-plane expansion, imposing an outward forceon the internal surface of the shoulder. Additional stress arises from dif-ferential thermal expansion between the feed material and the shoul-der. These effects have led to catastrophic failure of ceramic shouldersin previous investigations by the authors. Finally, we note that the ab-sence of pin in additive friction stir deposition allows for sharp transi-tions in the deposition path, which would lead to pin fracture infriction stir welding due to the high shear stress.
5. Comparison to other solid-state AM processes
In this section, two alternative solid-state AM processes are intro-duced and compared to additive friction stir deposition: ultrasonic addi-tive manufacturing and friction stir additive manufacturing. Both ofthem belong to the sheet lamination classification defined by ASTM,which requires additional subtractive processes to make 3D parts withcomplex geometries [30,37]. In contrast, additive friction stir depositionadds the material in a site-specific fashion and enables near-netshaping.
5.1. Comparison to ultrasonic additive manufacturing
Ultrasonic additive manufacturing is a sheet lamination process thatuses high frequency sound waves for interfacial cleaning and bondingfollowed by selective CNC (computer numerical control) milling. By vi-brating a textured wheel above 20,000 Hz on the top surface of a metalfoil, the oxide layers are removed from the interface between this foiland the previously deposited one, resulting in atomically clean surfaces
Table 2A comparison of different solid-state AM processes.
ASTM classification Sheet lamination Sheet lamination N/AHybrid process Yes Yes NoResolution limiting factor Subtractive process Subtractive process Tool geometryTemperature Relatively low Relatively high Relatively highMicrostructure Similar to pre-processed Refined, equiaxed in the stir zone only Refined, equiaxed
128 H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
in intimate contact [29,37]. Adiabatic heating and severe plastic defor-mation lead to recrystallization and formation of a metallurgical bondat the interface, resulting in 100% dense parts [84–87]. Ultrasonic addi-tive manufacturing allows for low operating temperatures and severalunique advantages. For example, the microstructure of the metal foilsis largely preserved throughout the sheet lamination process, post pro-cessing is generally unnecessary, and temperature-sensitive functionaldevices including electronics can be embedded into the printed parts.The final in-plane resolution in ultrasonic additive manufacturing is de-termined by the additional subtractive machining processes.
Additive friction stir deposition is fundamentally different in thatit is a material deposition process rather than a sheet lamination pro-cess. In addition, the microstructure resulting from the ‘intensive’friction stir process is dependent on dynamic recovery or dynamicrecrystallization, whereas the ‘gentle’ ultrasonic process gives a sim-ilar microstructure before and after fabrication. Having said that, ul-trasonic additive manufacturing can also lead to severe plasticdeformation and recrystallization at the interface resulting in finergrains. Finally, the base material in additive friction stir depositionis a metal rod or powders, while ultrasonic additive manufacturingrequires using thin metal foils.
5.2. Comparison to friction stir additive manufacturing
Friction stir additivemanufacturing is another sheet lamination pro-cess, in which multiple metal layers are stacked and then bound to-gether through friction stir welding. The welding is performed on thetop layer of sheets with a custom designed pin penetrating them verti-cally, rotating, and traversing to create a joining line throughout theoverlapping sheets [30]. Repeating this stacking andwelding process al-lows for fabrication of large components. A key difference between thisprocess and ultrasonic additive manufacturing is the layer thickness.While the latter usually uses metal foils with a thickness of ~0.1 mm,the former can weldmultiple thicker metal sheets. For example, Mishraand coworkers successfullywelded fourWE43 sheets eachwith a thick-ness of 1.7 mm [30].
Both friction stir additivemanufacturing and additive friction stir de-position are ‘intensive’ processes that involve severe plastic deforma-tion at high temperatures. However, there are significant differencesbetween them. One is a hybrid (both additive and subtractive) sheetlamination process involving welding that results in microstructureswith cyclic variations [30,48,60], whereas the other is a purely additiveprocess that deposits a new track with each pass. With each individuallayer undergoing the same type of severe plastic deformation during de-position and reheating afterwards, additive friction stir deposition can
Table 3A summary of the unique advantages of various metal AM processes.
Process Powder bed fusion Directed energy deposition U
Solid-state No No YAdvantages High resolution Scalable; capable of mixing powders on the
flyS
Nicheapplications
Complexgeometries
Large parts; multi-materials; repair Ls
lead to a much more homogeneous microstructure. Table 2 comparesthe three solid-state metal additive processes in detail.
6. Pathways to widespread implementation of metal AM
For widespread implementation, metal AM needs to possess a seriesof attributes, such as (1) affordability to customers, (2) convenience andease of use, (3) energy efficiency, (4) environmental friendliness,(5) wide applicability to engineering materials, (6) compatibility withmature technologies, and (7) reliability of part properties.
Unfortunately, not a single metal AM process fully meets the abovestandard at this moment. The question is, for current technologies,what aspects need to be improved for widespread implementation? Re-garding the prominent beam-based processes, the capability of produc-ing cheap, accessible powders of high quality and arbitrarycompositions is a pressing demand. Strategies are also needed to signif-icantly reduce the energy and cost of production and to enable the fab-rication of non-weldable alloys. Regarding the non-beam-basedprocesses, strategies for controlling in-plane resolution should be fur-ther developed, because these processes either have a low in-plane res-olution or require an additional subtractive process. In the case ofadditive friction stir deposition, the low in-plane resolution is not a re-sult of its physical nature but rather a result of design choices thatfavor scalability. Scaling down the tool size and using a small-diameter filler may improve the in-plane resolution without interferingwith the physical processes.
For all of these metal AM processes, quality control is a key techno-logical challenge. Substantial efforts have been made towards in-process qualification, where in situ and non-destructive process moni-toring techniques are employed to increase reliability in as-manufactured parts [88]. For example, infrared and high-speed camerashave been used to monitor the thermal evolution in the melting pooland during powder consolidation, 3D laser scanners have been used tomonitor surface topography, and in situ ultrasonic testing has beenused to monitor the defect formation and evolution [89–93]. With sen-sitivity analysis of the part attributes with respect to process parame-ters, closed loop control can be implemented; once a process anomalyis detected, the process parameters will be adjusted for online defectmitigation. In-process qualification is thus necessary for improvingpart consistency to enable widespread use of metal AM. The first at-tempt to instrument the basic additive friction stir deposition facilitywith multiple sensors is underway, which should enable in situ moni-toring of thermal fields, surface topography, and defect levels.
Despite their imperfections, it is important to recognize the uniqueadvantages of existing technologies, as shown in Table 3. Additive
es Yescalable; low operation temperature Scalable; flexible feed materials
arge parts; embedding electronics orensors
Large parts; non-weldable alloys;repair
129H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
friction stir deposition is well suited for the rapid fabrication of large-scale components, investigation of non-weldable alloys, and printingof alloys that are difficult or expensive to be made in powdered form.Powder bed fusion is the best for making high-resolution complex ge-ometries. Directed energy deposition enables mixing multiple powderson thefly to formalloys in situ [94]. Both directed energy deposition andadditive friction stir deposition can be good for field repair. Ultrasonicadditive manufacturing has its own advantages in low operation tem-perature and incorporation of electronics. Given that these technologiesall have their niches, a synergistic effort from various processes provid-ing complementary capabilities will likely make metal AM viable for awide variety of materials, applications, and fields. In that sense, a com-prehensive understanding of the benefits and limitations of each tech-nology will be essential for widespread implementation of metal AM.
Acknowledgement
HZYwould like to acknowledge the support from the Department ofMaterials Science and Engineering and College of Engineering at Vir-ginia Tech. MEJwould like to acknowledge the support from the CharlesBlankenship Engineering Scholarship. The authors would also like tothank Dr. William Reynolds for helpful comments.
Declaration of interest
The authors claim no conflicts of interest.
References
[1] ASTM F2792, Standard Terminology for Additive Manufacturing Technologies,ASTM International, 2015.
A. Wilson-Heid, A. De, W. Zhang, Prog. Mater. Sci. 92 (2018) 112.[5] W.J. Sames, F.A. List, S. Pannala, R.R. Dehoff, S.S. Babu, Int. Mater. Rev. 61 (2016) 315.[6] W.E. Frazier, J. Mater. Eng. Perform. 23 (2014) 1917.[7] X. He, J. Mazumder, J. Appl. Phys. 101 (2007), 053113. .[8] A. Mukesh, D. Bourell, J. Beaman, H. Marcus, J. Barlow, Rapid Prototyp. J. 1 (1995) 26.[9] J.-P. Kruth, P. Mercelis, J. Van Vaerenhergh, L. Froyen, M. Rombouts, Rapid Prototyp.
J. 11 (2005) 26.[10] M.J. Matthews, G. Guss, S.A. Khairallah, A.M. Rubenchik, P.J. Depond, W.E. King, Acta
Mater. 114 (2016) 33.[11] W. Xu, M. Brandt, S. Sun, J. Elambasseril, Q. Liu, K. Latham, K. Xia, M. Qian, Acta
Mater. 85 (2015) 74.[12] Z. Wang, T.A. Palmer, A.M. Beese, Acta Mater. 110 (2016) 226.[13] A. Basak, S. Das, Annu. Rev. Mater. Res. 46 (2016) 125.[14] M. Yampolskiy, W. King, G. Pope, S. Belikovetsky, Y. Elovici, in: M. Rice, S. Shenoi
(Eds.), Revised Selected Papers, Springer International Publishing, Cham 2017,pp. 23–44.
[15] D.D. Gu, W. Meiners, K. Wissenbach, R. Poprawe, Int. Mater. Rev. 57 (2012) 133.[16] E.D. Herderick, JOM 68 (2016) 721.[17] R. Cunningham, S.P. Narra, T. Ozturk, J. Beuth, A.D. Rollett, JOM 68 (2016) 765.[18] D.L. Bourell, J.J. Beaman, M.C. Leu, D.W. Rosen, Proceedings of RapidTech 2009: US-
549 (2017) 365.[20] L.E. Murr, S.M. Gaytan, D.A. Ramirez, E. Martine, J. Hernandez, K.N. Amato, P.M.
Shindo, F.R. Medina, R.B. Wicker, J. Mater. Sci. Technol. 28 (2012) 1.[21] J.J. Lewandowski, M. Seifi, Annu. Rev. Mater. Res. 46 (2016) 151.[22] G. Frommeyer, U. Brux, P. Neumann, ISIJ Int. 43 (2003) 438.[23] O. Grassel, L. Kruger, G. Frommeyer, L.W. Meyer, Int. J. Plast. 16 (2000) 1391.[24] B.C. De Cooman, Y. Estrin, S.K. Kim, Acta Mater. 142 (2018) 283.[25] D.V. Edmonds, K. He, F.C. Rizzo, B.C. De Cooman, D.K. Matlock, J.G. Speer, Mater. Sci.
Eng. A 25 (2004) 438–440.[26] I. Gibson, D. Rosen, B. Stucker, Additive Manufacturing Technologies: 3D Printing,
Rapid Prototyping, and Direct Digital Manufacturing, Springer, New York, 2015.[27] J. Slotwinski, E. Garboczi, P. Stutzman, C. Ferraris, S. Watson, M. Peltz, J. Res. Natl.
Inst. Stand. Technol. 119 (2014) 460.[28] V. Bhavar, P. Kattire, V. Patil, S. Khot, K. Gujar, R. Singh, Conference: 4th International
Conference and Exhibition on Additive Manufacturing Technologies-AM-2014, Sep-tember 1 & 2, 2014 (Banglore, India).
[29] C.Y. Kong, R.C. Soar, P.M. Dickens, Mater. Sci. Eng. A 363 (2003) 99.[30] S. Palanivel, P. Nelaturu, B. Glass, R.S. Mishra, Materials & Design, 65, 2015 934.[31] D. White, Object Consolidation Employing Friction Joining, 2002 (US 6457629 B1).[32] H.Z. Yu, N. Hardwick, MS&T Conference, 2017 (Pittsburgh, PA).
Whittington, D. Francis, J. Su, R.L. Martens, N. Hardwick, Mater. Sci. Eng. A 694(2017) 1.
[42] F.J. Humphreys, G.S. Rohrer, A. Rollett, Recrystallization and Related Annealing Phe-nomena, 3rd ed. Elsevier, Amsterdam; Oxford; Cambridge, 2017.
[43] V. Manvatkar, A. De, L.E. Svensson, T. DebRoy, Scr. Mater. 94 (2015) 36–39.[44] P. Edwards, M. Ramulu, Sci. Technol. Weld. Join. 15 (6) (2010) 468–472.[45] S. Swaminathan, K. Oh-Ishi, A.P. Zhilyaev, C.B. Fuller, B. London, M.W. Mahoney, T.R.
McNelley, Metall. Mater. Trans. A 41 (3) (2010) 631–640.[46] P. Asadi, M.K. Besharati Givi, M. Akbari, Int. J. Adv. Manuf. Technol. 83 (1) (2016)
301–311.[47] D. Raabe, Annu. Rev. Mater. Res. 32 (1) (2002) 53–76.[48] R.S. Mishra, Z.Y. Ma, Mater. Sci. Eng. R. Rep. 50 (2005) 1.[49] T.W. Nelson, C.D. Sorensen, S. Packer, P.A. Felter, Friction Stir Welding of Metal Ma-
trix Composites, Ferrous Alloys, Non-ferous Alloys, and Superalloys Using aSuperabrasive Tool, 2004 (US 7124929 B2).
[50] F.C. Liu, Y. Hovanski, M.P. Mile, C.D. Sorensen, T.W. Nelson, J. Mater. Sci. Technol. 34(2018) 39.
[51] M.W. Mahoney, C.G. Rhodes, J.G. Flintoff, W.H. Bingel, R.A. Spurling, Metall. Mater.Trans. A 29 (1998) 1955.
[53] Y.S. Sato, H. Yamanoi, H. Kokawa, T. Furuhara, Scr. Mater. 57 (2007) 557.[54] Y.S. Sato, S.H.C. Park, M. Michiuchi, H. Kokawa, Scr. Mater. 50 (2004) 1233.[55] D.P. Field, T.W. Nelson, Y. Hovanski, K.V. Jata, Metall. Mater. Trans. A 32 (2001)
2869.[56] M.A. García-Bernal, R.S. Mishra, R. Verma, D. Hernandez-Silva, Mater. Sci. Eng. A 670
(2016) 9.[57] S. Nimer, J. Wolk, M. Zupan, Acta Mater. 61 (2013) 3050.[58] W.M. Thomas, E.D. Nicholas, Mater. Des. 18 (1997) 269.[59] A.P. Reynolds, Scr. Mater. 58 (2008) 338.[60] R.S. Mishra, Z.Y. Ma, I. Charit, Mater. Sci. Eng. A 341 (2003) 307.[61] R. Nandan, T. DebRoy, H.K.D.H. Bhadeshia, Prog. Mater. Sci. 53 (2008) 980.[62] A. Gerlich, M. Yamamoto, T.H. North, J. Mater. Sci. 43 (1) (2008) 2–11.[63] A.P. Gerlich, T. Shibayanagi, Sci. Technol. Weld. Join. 16 (4) (2011) 295–299.[64] E. Cakmak, M.M. Kirka, T.R. Watkins, R.C. Cooper, K. An, H. Choo,W. Wu, R.R. Dehoff,
S.S. Babu, Acta Mater. 108 (2016) 161.[65] D. Gary, Modeling and Characterization of Friction Stir Fabricated Coatings on
Al6061 and Al5083 Substrates(Ph.D. thesis) Virginia Tech, 2009.[66] J.R. Calvert, Microstructure and Mechanical Properties of WE43 Alloy Produced Via
Additive Friction Stir Technology(Master thesis) Virginia Tech, 2015.[67] D. Garcia, M.E. Jones, Y. Zhu, H.Z. Yu, J. Mater. Res. 33 (2018) 58.[68] H.Z. Yu, S.R. Cross, C.A. Schuh, J. Mater. Sci. 52 (2017) 4288.[69] Y. Morisada, T. Imaizumi, H. Fujii, Scr. Mater. 106 (2015) 57–60.[70] P.S. Follansbee, U.F. Kocks, Acta Metall. 36 (1) (1988) 81–93.[71] G.Z. Voyiadjis, F.H. Abed, Int. J. Plast. 22 (8) (2006) 1398–1431.[72] R. Kapoor, S. Nemat-Nasser, Metall. Mater. Trans. A 31 (13) (2000) 815–823.[73] A.K. Shukla, H. Chelladurai, S. Tiwari, Materials Today: Proceedings 4 (2, Part A),
2017 3618–3626.[74] Z.C. Cordero, C.A. Schuh, Acta Mater. 82 (2015) 123.[75] A.O. Aning, Z. Wang, T.H. Courtney, Acta Metall. Mater. 41 (1993) 165.[76] J.M. Sietins, Private communications.[77] K.M. Taminger, R.A. Hafley, Electron beam freeform fabrication for cost effective
near-net shape manufacturing, NATO/RTOAVT-139 Specialists' Meeting on Cost Ef-fective Manufacture Via Net Shape ProcessingNATO, Amsterdam (The Netherlands),2006.
[78] D. Ding, Z. Pan, D. Cuiuri, H. Li, Int. J. Adv. Manuf. Technol. 81 (1) (2015) 465–481.[79] M. Van Elsen, P. Corbeels, H. Van de Zande, J.-P. Kruth, 7th Annual International Con-
ference on Rapid Prototyping Technologies for Competitive Tooling, Cape Town,South Africa, November 1–3, 2006.
[81] J.M. Gibert, E.M. Austin, G. Fadel, Rapid Prototyp. J. 16 (4) (2010) 284–294.[82] C. Robinson, C. Zhang, G.D. Janaki Ram, E. Siggard, B.E. Stucker, L. LiMaximum, Proc.
of 17th Solid Freeform Fabrication Symp., Aug. 14–16, 2006, Austin, TX, 2006.[83] M. Collier, J.M. Torralba, T. Sakai, Mater. Sci. Forum 3011 (2003) 426–432.[84] R.R. Dehoff, S.S. Babu, Acta Mater. 58 (2010) 4305.[85] M.R. Sriraman, S.S. Babu, M. Short, Scr. Mater. 62 (2010) 560.[86] H.T. Fujii, S. Shimizu, Y.S. Sato, H. Kokawa, Scr. Mater. 135 (2017) 125–129.[87] S. Shimizu, H.T. Fujii, Y.S. Sato, H. Kokawa, M.R. Sriraman, S.S. Babu, Acta Mater. 74
(2014) 234–243.[88] S.K. Everton, M. Hirsch, P. Stravroulakis, R.K. Leach, A.T. Clare, Materials & Design,
vol. 95, 2016 431–445.[89] H. Krauss, C. Eschey, M.F. Zaeh, 23rd International Solid Freeform Fabrication Sym-
130 H.Z. Yu et al. / Scripta Materialia 153 (2018) 122–131
[90] T. Craeghs, S. Clijsters, E. Yasa, J.-P. Kruth, 22nd International Solid Freeform Fabrica-tion Symposium; Austin, TX, 2011.
[91] D. Hu, R. Kovacevic, Int. J. Mach. Tools Manuf. 43 (1) (2003) 51–60.[92] S. Liu, W. Liu, M. Harooni, J. Ma, R. Kovacevic, Opt. Lasers Technol. 62 (10) (2014)
124–134.
[93] N. Karthik, H. Gu, D. Pal, T. Starr, B. Stucker, 24th International Solid Freeform Fabri-cation Symposium, UT Austin, Austin, TX, 2013.
[94] D.C. Hofmann, S. Roberts, R. Otis, J. Kolodziejska, P.R. Dillon, J. Suh, A.A. Shapiro, Z.Liu, J. Borgonia, Sci. Rep. 4 (2014) 5357.
