Smart Materials and Structures Lab | - Microstructural Characterization … · 2020-01-22 · Microstructural Characterization of Bonding Interfaces in Aluminum 3003 Blocks Fabricated

Introduction

Ultrasonic additive manufacturing(UAM) is a solid-state joining process inwhich thin metallic tapes are ultrasonicallywelded on top of one another and periodi-cally machined to create a final part. Aschematic illustration of the process isshown in Fig. 1. Along with progressivebuilding of a block through seam welding(Fig. 1A), a milling process (Fig. 1B) is usedas required to create holes or channels be-fore welding the subsequent layers. Themachining operation is also used periodi-cally to produce a flat surface to ensureproper dimensions of the finished build.This process offers many benefits over tra-

ditional fusion welding processes such as al-lowing for complex shapes and designs,having a significantly lower processing tem-perature, allowing for embedded materialsand channels, and offering the capability ofjoining dissimilar materials that are other-wise difficult or impossible due to UAMbeing a solid-state process.

The majority of research on UAM is cur-rently focused on optimizing processing pa-rameters (Refs. 1–6) and characterizing thequality and microstructure (Refs. 7–9) ofthe resulting builds. The four main param-eters in UAM are sonotrode amplitude,

travel speed, normal force applied, and pre-heat temperature. Increasing the ampli-tude, normal force, and preheat tempera-ture, while decreasing the travel speed,generally increases the quality of the bonds.However, above a threshold for each pa-rameter, no further gains are realized (Ref.5). The threshold effects with respect tosonotrode amplitude and normal force aremost likely due to the machine not being ca-pable of delivering enough power to sustainthe ultrasonic vibrations at the higher am-plitudes and forces. Additional gains inbond quality may be possible with a higher-power system, allowing for higher ampli-tudes of vibration and forces. For most cur-rent UAM machines, optimum parametersare approximately 18–21 μm amplitude,25–50 mm/s travel speed, preheat of65°–150°C, and normal forces between 800and 1500 N. Peel tests (Refs. 1, 2, 4, 6), fiberpush- out testing (Ref. 3), and microhard-ness and nanohardness tests (Refs. 2, 9, 10)have been conducted to further the under-standing of this additive manufacturingprocess. These tests are often done alongwith parameter development to comparethe bond quality between different builds.Voids are often present in UAM builds andcan be quantified with linear weld density(LWD). The LWD is defined as the lengthof a particular interface that appears prop-erly bonded divided by the total interfacelength inspected. The LWD is often used asa test to determine optimum processing pa-rameters (Refs. 1, 2, 4, 5, 11). It is generallyagreed that to improve the bond quality ofUAM builds, LWD must be kept as high aspossible. In most UAM builds, LWD den-sity ranges from 40 to 95%. Ram et al. (Ref.5) and Johnson (Ref. 7) theorized voidsform in UAM builds due to the sonotrodetransferring its texture to the workpiece.This results in a situation where the top ofeach interface is smooth, but the bottom is

Microstructural Characterization of BondingInterfaces in Aluminum 3003 Blocks

Fabricated by Ultrasonic AdditiveManufacturing

A look at linking microstructure and linear weld density to the mechanicalproperties of ultrasonic additive manufacturing builds as well as analyzing their

properties with different microscopy and testing methods

BY D. E. SCHICK, R. M. HAHNLEN, R. DEHOFF, P. COLLINS, S. S. BABU, M. J. DAPINO, AND J. C. LIPPOLD

KEYWORDS

Ultrasonic AdditiveManufacturing (UAM)

Linear Weld Density (LWD)Scanning Electron Microscopy(SEM)

Shear StrengthAl 3003-H18Transmission Electron

Microscopy

D. E. SCHICK, S. S. BABU ([email protected]),and J. C. LIPPOLD are with the Department ofMaterials Science and Engineering — Welding En-gineering Program, and R. M. HAHNLEN and M.J. DAPINO are with the Department of Mechani-cal Engineering at The Ohio State University,Columbus, Ohio. R. DEHOFF, formerly at the De-partment of Materials Science and Engineering, iscurrently at Oak Ridge National Laboratory, OakRidge, Tenn. P. COLLINS, formerly at the Depart-ment of Materials Science and Engineering, is cur-rently at Quad Cities Manufacturing Laboratory,Rock Island, Ill.

Ultrasonic additive manufacturing (UAM) is a process by which hybrid and near-net-shaped products can be manufactured from thin metallic tapes. One of the main con-cerns of UAM is the development of anisotropic mechanical properties. In this work,the microstructures in the bond regions are characterized with optical and electron mi-croscopy. Recrystallization and grain growth across the interface are proposed as amechanism for the bond formation. The presence of voids or unbonded areas, which re-duce the load-bearing cross section and create a stress intensity factor, is attributed tothe transfer of the sonotrode texture to the new foil layer. This results in large peaks andvalleys that are not filled in during processing. Tensile testing revealed the weld inter-face strength was 15% of the bulk foil. Shear tests of the weld interfaces showed almost50% of the bulk shear strength of the material. Finally, optical microscopy of the frac-ture surfaces from the tensile tests revealed 34% of the interface area was unbonded.

105-sWELDING JOURNAL

WE

LD

ING

RE

SE

AR

CH

ABSTRACT

Babu et al Supplement May 2010:Layout 1 4/12/10 1:40 PM Page 105

rough. To combat this, it has been demon-strated that milling between layer depositsto provide a smooth-to-smooth interfacecan eliminate voids, achieving 98% LWD(Ref. 5). However, no tensile, peel, or otherquantitative measurement of bond qualitywas done to verify the bond quality.

Several researchers have made prelim-inary attempts at mechanical and finite el-ement modeling (FEM) of UAM weld-ments. Doumanidis and Gao (Ref. 10)used an analytical model combined withexperimental data to produce an FEM of

UAM useful in simulating different mate-rial combinations, embedding of materi-als, and the production of complex parts.This model also proved useful in deter-mining ideal geometry for the sonotrodeand other components. Zhang et al. (Ref.12) developed a three-dimensional FEMfor ultrasonic spot welding that evaluatedthe ever-changing parameters at eachnode including normal stress, heat gener-ation, and plastic deformation. Theirmodel used thermal and mechanical con-ditions to simulate the ultrasonic welding

process and led to the theory that ultra-sonic bonds are formed due to high levelsof localized strain, high temperatures, andplastic deformation along the interface.Siddiq et al. (Ref. 13) also developed athree-dimensional model focusing on fric-tion and heat generation at the interface.Their simulation determined the effect offriction at the interface to be only useful inremoving oxides and contaminates whilethe plastic deformation of material actu-ally leads to a bond.

Multiple material combinations havebeen studied, including aluminum, cop-per, titanium, and nickel (Refs. 11, 13, 14),as well as many different fibers have beensuccessfully embedded including fiber op-tics, silicon carbides, shape memory alloys,and thermocouples (Refs. 3, 6, 9, 11, 15,16). The UAM process has been found toeasily accommodate these embeddedfibers as the ultrasonic energy allows forexcellent matrix material flow around thefiber.

In all the above work, a one-to-one cor-relation of tensile and shear propertieswith the underlying microstructure hasnot been documented. Therefore, in thiswork, the mechanical properties of alu-minum builds were measured and ob-served properties were correlated with thedetailed microstructure evaluation usingoptical microscopy, hardness mapping,and electron microscopy. The results willbe compared with published literature onUAM processes as well as data from ultra-sonic spot welding. The methodology anddata generated in this research are ex-pected to provide a baseline for the devel-opment of a very high power UAM (VHP-UAM) instrument (Ref. 17). Thisinstrument will be capable of joininghigher-strength alloys including titanium,copper, nickel, shape memory alloys, car-bon steels, and low-alloy steels.

Experimentation

Alloys

In this research, a non age-hardenableAl-3003 (Al-1Mn-0.7Fe-0.12Cu wt-%)

MAY 2010, VOL. 89106-s

WELDING RESEARCH

A B

Fig. 1 — Schematic illustration of the UAM process. A — Adding a new tape layer; B — periodic milling operation to form final dimensions.

Table 1 — UAM Processing Parameters Used in the Current Research

ID Force Speed Amplitude Frequency Build Temperature(N) (mm/s) (μm) (kHz) (°C)

Tack 350 59.3 12 20 150Weld 1150 42.3 17 20 150

Fig. 2 — The UAM tensile and shear sample dimen-sions. A — Reduced shear specimens (1 and 2); B —symmetric shear specimens (3 and 4); C — transversetensile specimens; D — longitudinal tensile specimens.

A

C D

B

98.3

3.18

31.75

R 3.18

ø 3.28

3.81

12.7 9.52

Tape LayerOrientation

Key

3.56

R 239

9.52R 3.18

9.52

17.14

6.3512.7

38.19.52

9.5238.1

17.14

12.7

38.7

38.1 9.52

9.52


alloy was used as both tapes (H18, 150 μmthick, 25.4 mm wide) and substrate (H14,more than 12.7 mm thick). The composi-tion of the materials used meets the stan-dard specification of the alloy (Ref. 18).

UAM Process Parameters

The ultrasonic sonotrode was madefrom Ti-6Al-4V alloy, and the surface wassubjected to electrical discharge machin-ing (EDM) to achieve the desired surfacetexture (Ra = 7 μm). This surface textureis known to provide consistent bond qual-ity (Refs. 5, 7). During the tacking andwelding passes, the substrate was pre-heated with a hot plate to 149°C (300°F)and was maintained at that temperature.The preheat was used to soften the mate-rial, which leads to better bonding. How-ever, during processing, the tape and in-terface temperatures are not necessarilymaintained at this preheat temperaturedue to complex heat transfer across themany weld interfaces, heat generated atthe interfaces, and a heat-sinking effectdue to the sonotrode.

Sequential joining of tapes to build asmall block was achieved through tackingand welding passes. The differences be-tween the tacking and welding passes are re-lated to the magnitude of the process pa-rameters, i.e., normal load, travel speed,and amplitude of ultrasonic vibration. Inthe current research, the vibration fre-quency was kept constant at 20 kHz for allpasses due to machine and sonotrode de-sign. Table 1 provides an overview of theprocessing parameters used in the currentresearch. These processing parameterswere obtained by extensive trial and errorexperiments. One method of testing processparameters involves joining of tapes by dif-ferent process parameters and manuallypeel testing the builds. The best processingparameters are qualitatively selected whenthe manual peel test fractures the tape,rather than peeling off from the interfacial


WEL

DING

RES

EARC

H

Table 2 — Shear Test Results, Base Metal: Al 3003-H18 USS Is 110 MPa

Sample Force (N) Area (mm2) USS (MPa) % of BM

1 5089 81.6 62.4 56.72 4395 80.5 54.6 49.63 8830 215 41.1 37.44 11387 216 52.6 47.8

Average 52.7 47.9Standard Deviation 8.78 7.9

Stdev/Avg 0.167

Fig. 3 — Sample image to demonstrate methodology used for image analyses to derive the linear weld density. A — Original optical microscopy image; B —processed image using ImageJ software.

BA

Fig. 4 — Schematic illustration of steps to prepare TEM samples from builds made with UAM. Sam-ples were taken along interfaces at various heights (top, middle, and bottom) of the build. First, opti-cal microscopy samples were prepared to select the regions of interest. In the next step, the sample wastransferred to FIB instrument. Then, an interface of interest was selected, and a rectangular region oneither side of the interface was coated with platinum. After this step, the focused ion beam machin-ing was made on either side of the coated region. This leads to a thin film sample that contains thebonded interface. In this schematic representation, the n and n + 1 correspond to the successive tapesduring the UAM processing.

Table 3 — Transverse Tensile Tests, Base Metal: Al 3003-H18 UTS Is 200 MPa

Sample Force (N) Area (mm2) UTTS (MPa) % of BM

1 907 31.9 28.4 14.22 979 31.4 31.1 15.63 930 32.4 28.7 14.44 1010 31.4 32.1 16.15 859 33.5 25.7 12.96 601 31.4 19.1 9.67 1080 32.4 33.3 16.7

Average 28.3 14.2Standard Deviation 4.81 2.4

Stdev/Avg 0.170


area. It is important to note that the processparameters used here may not be optimumand are considered as the starting point forthis and future research. Details of the peeltest instrument and technique have beencovered extensively by other researchers(Refs. 19–21).

Mechanical Property Testing

Previous mechanical strength studies on

UAM samples focused on peel tests (Refs.1, 2, 4, 6, 22–24). Peel tests, while useful forcomparison between parameter sets andother UAM samples, are primarily used formeasuring adhesive strength of tape, glue,or other bonded surfaces and do not pro-vide strength values useful for the design ofbulk UAM parts. In order to be utilized asan additive manufacturing process, bulkmechanical strengths such as ultimateshear and tensile strengths must be known

for design of UAM samples. To date, therehas been no reported research on such bulkstrength properties. In order to obtain bulkstrength properties of the UAM matrix,three types of samples were made: lapshear, transverse tensile, and longitudinaltensile. The geometries of these test speci-mens are presented — Fig. 2. The shearspecimens were built such that the tape in-terfaces were along the shear plane. Sheartests were conducted using a specializedshear jig and a compressive load with an av-erage displacement rate of 0.28 mm/s.

Initial shear test specimens had a re-duced interface area to ensure failurebelow the 5000-lb machine capability. Ini-tial estimates for strength assumed theshear strength for UAM specimens wouldbe approximately 75% of the bulk mate-rial. As testing revealed, the shearstrength was much lower than anticipated,and later samples were not prepared witha reduced interface area. The transversetensile specimens were built such that the

MAY 2010, VOL. 89108-s

WELDING RESEARCH

Table 4 — Longitudinal Tensile Test Results, Base Metal: Al 3003-H18 UTS Is 200 MPa

Sample Force (N) Area (mm2) ULTS (MPa) % of BM

1 2630 11.71 225 112.52 2900 12.21 238 119.03 2880 12.03 240 120.04 2870 12.13 237 118.55 2790 11.95 233 116.5

Average 234 117.0Standard Deviation 5.89 2.9

Stdev/Avg 0.025

A B

Fig. 5 — Fracture surface image analyses of transverse tensile test samples. A — Low-magnification optical image; B — high-magnification optical image show-ing regions I and III after processing. Region I is well bonded material with recrystallization across the interface. Region II is deep valleys carved by the sonotrodeduring the previous pass (not shown). Region III material is directly opposite Region II and is unaffected during the UAM process; C — SEM image of the topsurface of the tape; D — SEM image of the bottom surface of the tape; and the featureless gray regions are Region III material.

500 µm

Top Surface Bottom Surface

500 µmC D


tape interfaces were perpendicular to theapplied axial force. Transverse tensile testswere conducted using specialized speci-men shoulder grips with an average dis-placement rate of 0.32 mm/s. Longitudinaltensile specimens were built with tape in-terfaces parallel to the applied axial forceand were tested using pinned grips with anaverage displacement rate of 0.52 mm/s.

For all tests, samples were placed in auniversal tension/compression testingframe and were stressed until failure. Theapplied force was recorded using a ten-sion/compression load cell and frame ac-tuator displacement with an integratedlinear variable differential transformer(LVDT). Maximum loads were used to ob-tain ultimate stresses, and the shape of theforce-displacement plots was used to helpcharacterize specimen failures. Becausethe integrated LVDT measures the testingframe actuator displacement, all displace-ment data includes displacement gener-ated within the load train as well as thespecimen. For this reason, the shape of theforce-displacement plots can only be usedto determine if a given sample failed in abrittle or ductile mode through qualitativeanalysis. However, this cannot be used tocalculate specimen strain or related prop-erties such as the elastic modulus. Afterthe mechanical testing, the fracture sur-faces of the shear and transverse tensilesamples were examined with optical andscanning electron microscopy (SEM).

Optical Microscopy and HardnessMapping

Optical metallographic samples wereprepared using standard metallographictechniques. The samples were preparedfrom cross sections perpendicular to thetravel direction. Five optical images at10× magnification were taken from dif-ferent locations within the build. Eachimage corresponded to 1111 by 833 μm,containing five interfaces. These mi-croscopy images were analyzed with thepublic domain ImageJ software program

(Ref. 25). With linearintercept analyses, theLWD was measured asa function of distancein a direction perpen-dicular to the metallictape layers. Grayscaleimage threshold val-ues (0 to 60) were keptconstant to delineatethe void areas in allthese images. A typicaloptical image beforeand after thresholdprocessing demon-strates the effective-ness of delineating thevoids between layers— Fig. 3A, B.

For the microhard-ness testing, a LecoAMH-43 machine wasused to create a 200 × 20 map of hardnessindents with a diamond indenter. Themeasurements were made with 25-g loadand a 13-s dwell time, and spacing be-tween the hardness indents was 150 μm inboth directions. The coordinates of theindents were designed to sample the solidmatrix regions away from interfaces.Hardness measurements were made ontapes that were not ultrasonically consol-idated in the same orientation, as areference point.

Analytical ElectronMicroscopy

In order to examine the grain structureand morphology in specific locations (bot-tom, middle, and top regions of the build)through transmission electron microscopy(TEM), the samples were prepared usinga FEI Helios dual-beam focused ion beam(FIB) microscope. The samples were pre-pared from cross sections perpendicularto the travel direction along interfaceswith apparent good bonding. The FIB


WEL

DING

RES

EARC

H

Fig. 6 — Measured load vs. displacement curves. A — Shear tests with sym-metric and reduced cross sections; B — transverse tensile tests; C — longitu-dinal tensile tests.

A B

C

Table 5 — Linear Weld Density (%) from Optical Micrographs Taken from Random Locationswithin the Build

Image Interface StandardNumber 1 2 3 4 5 Average Deviation

1 49.5 86.2 74.7 91.1 48.2 69.9 20.22 70.4 74.1 61.9 88.1 51 69.1 13.83 69.5 59.6 67.5 65.5 5.24 53.1 65.7 52.2 55.2 46.7 54.6 7.05 34.2 65.6 68.1 91.6 75.2 66.9 20.9

Overall 65.2 15.3


contains both an electron beam as well asan ion beam that can be used for imaging.The electron beam is a standard second-ary electron beam, which can be used toimage topological difference but does notreveal grain structure in unetched alu-minum alloys. The ion beam was used toimage the grain structure of the materialwith contrast differences arising from gal-lium ion channeling contrast. To createTEM foils, platinum is deposited over theregion of interest to protect the foil sur-face from Ga+ implantation duringmilling. Trenches are then milled on bothsides of the platinum to create the foil.The sample is then bonded to an om-niprobe needle also using platinum. Thesides of the foil are then milled to create afree-standing foil. Once the sample is cutfree, it is lifted out using the omniprobeneedle and welded to a copper grid usingplatinum. Once the sample is welded tothe grid, it is thinned using ion milling anda series of various apertures at 30 kV. Thesteps used in making the samples areschematically illustrated in Fig. 4. Finally,the samples were then examined using aFEI Tecnai F20 operated in STEM mode.

Results and Discussions

Mechanical Properties

The original mechanical properties ofAl 3003-H18 alloys are as follows: The ul-timate tensile strength (UTS) is 200 MPa,the yield strength (YS) is 186 MPa, andthe ultimate shear strength (USS) is 110MPa (Ref. 18). The mechanical propertydata from this research program are sum-marized in Tables 2–4. All shear tests re-sulted in a linear force-displacement re-lationship, indicating samples failed in amacro-level brittle fracture mode. Asshown in Table 2, an average USS of 52.7MPa with a standard deviation of 8.78MPa was found. The average USS was ap-proximately 48% of that of the solid basematerial. The results from the transverse

tensile tests are shown inTable 3. The average ulti-mate transverse tensilestrength (UTTS) was 28MPa, approximately 15%of the tensile strength ofsolid base material. Stan-dard deviation for UTTSwas 4.57 MPa.

To understand the re-duced strength of thetransverse tensile testsamples, the fracture sur-faces were characterizedusing optical and scan-ning electron microscopy.The fracture surfacesfrom the samples from 1to 5, as well as samples 7and 8, share a similar frac-ture surface — Fig. 5.These images indicatethat the interface regionshave many small speckled-like features dispersedthroughout the bond area. This feature isdue to small areas of bonded material(marked as I in Fig. 5A) mixed with smallareas of unaffected material (marked asIII in Fig. 5B). In order to make sure theRegion I (Fig. 5B) is a true bond, the frac-ture surfaces from either sides of the frac-ture were characterized with scanningelectron microscopy (Fig. 5C, D). Scan-ning electron microscopy showed that thefracture surfaces do show localized duc-tile failure with typical microvoid coales-cence features. This is irrespective of thefact that the load-displacement curves donot show appreciable macro level ductil-ity. These observations proved that theultrasonic additive manufacturing didnot reduce the inherent ductility of thematerial; however, on a macro scale, thematerial behaved in a brittle fashion dueto the voids. The lack of a yield point andhardening region is related to prematurebrittle fracture caused by the voids.

Transverse tensile samples behaved sim-ilar to the shear samples, as indicated by the

linear force-displacement plots —Fig. 6A,B. From Fig. 6B, sample 6 is considered tobe an outlier, as seen by its much lower fail-ure force when compared to the other sam-ples. Upon examination, the fracture sur-face of the sample 6 showed interestingfeatures that were different from the othersamples inspected — Fig. 7A. Optical mi-croscopy showed trenches and ridges, whichare typical of surfaces created by a millingoperation (Ref. 7). Furthermore, scanningelectron microscopy showed that the bondshave formed along these ridges and havefailed again by ductile mode — Fig. 7B, C.Cursory evaluation of the above fracturemorphology may be puzzling; however, thisphenomenon may be explained. During theUAM process, at frequent intervals amilling operation is performed to achieve aflat surface to ensure dimensional accuracyof the finished part. In sample 6, the failureoccurred at one such milled interface. Thefracture surface showed that the area frac-tion of the bonded region was small com-pared to unbonded regions, likely the rea-

MAY 2010, VOL. 89110-s

WELDING RESEARCH

A

Fig. 7 — Fracture surface analyses of transverse tensile test 6. A — Opti-cal image showing crosshatching from the milling pass; B — SEM imageof the top surface of the tape containing fractured regions. It is apparentall bonding occurred in the light-colored regions, less than half of theavailable surface area; C — SEM image of the bottom surface of the tapecontaining fractured regions. It is apparent all bonding occurred in thelight-colored regions, less than half of the available surface area.

B

C


son for the premature failure of this sample.All of the samples tested had this flat passwithin them; however, only sample 6 failedin this manner. Because it is not known whythe other samples did not fail at this loca-tion, it is believed that sample 6 does notgive a true representation of UAM bondtensile strength and is also excluded in sta-tistical analysis. The sensitivity of surfaceroughness on the bond quality has been ad-dressed by previous researchers (Refs. 7,22) by relating the surface roughness of thesonotrode to changes in linear weld density.

Unlike the other UAM samples, thelongitudinal tensile samples (Table 4) ex-hibited a substantial plastic yielding re-gion after the linear elastic region — Fig.6C. This is more typical of aluminum al-loys and indicates that failure occurred ina ductile mode. All tested samples exhib-ited a higher than expected tensilestrength. The average ultimate longitudi-nal tensile strength (ULTS) was 234 MPa,17% more than the original Al 3003-H18tape based on published properties (Ref.18). This is a departure from both thetransverse tensile samples and shear sam-ples previously tested in which the failurestresses were significantly lower than theAl 3003-H18 tape based on publishedproperties(Ref. 18). In this orientation, nodrop in tensile strength was expected asthe load was transmitted along the solidtapes as opposed to across the interfacesbetween them. However, the increase instrength above the base material was notexpected and further explanation of thisphenomenon is required.

Microstructure and MechanicalHeterogeneity

To rationalize the reduction in me-chanical properties in the transverse load-ing condition, the LWD of the builds indifferent regions were analyzed. A typicaldata set of linear void density (inverse ofLWD) is shown in Fig. 8. The image analy-ses show the LWD can vary from 35 to99%, depending on the interface. The av-erage LWD of all the images analyzed wasfound to be 65.2±15.3%, (Table 5). Imageanalyses of optical micrographs of thefracture surface of the transverse tensilesamples (Fig. 5) yielded 66±2% bondedarea. In Fig. 5D, it is clear the voided re-gions are random in nature. In stereologi-cal terms, randomly placed line segmentsin cross-sectional images are proportionalto an objects area in a 2-D plane (Ref. 26).However, in the current study, only onecross-sectional plane was used. This pre-vents the conclusion that LWD is directlyrelated to area density of properly bondedmaterial in UAM builds, despite the aver-ages being comparable. With additionalangular cross sections and more samples,it may be possible to confirm a possible

one-to-one relationship. Regions withlower amounts of bonded area within thebuild are expected to reduce the trans-verse and shear strength significantly. Thishypothesis is consistent with the conclu-sions made by previous researchers thatlinear weld density is a good measure ofUAM bond quality.

The mechanical properties measuredalong longitudinal sections showed a 17%

increase in ULTS compared to that of theoriginal Al 3003-H18 tape materials. Inorder to rationalize this increase instrength, hardness mapping was per-formed on the UAM builds. The mapshows a soft substrate and UAM build re-gions with large variations in hardness —Fig. 9A. The hardness data were analyzedin terms of frequency distribution — Fig.9B. This graph also shows the hardness


WEL

DING

RES

EARC

H

A

Fig. 8 — A — Plot of linear void density (LVD) vs. five interfaces; B — the corresponding optical imagenumber 5. The LVD point for each interface was taken as the high point and is shown. Linear void den-sity is the inverse of linear weld density.

Fig. 9 — Microhardness plot of a UAM build. A — The map was 200 indents tall by 20 indents wide, with thesofter substrate at the bottom. No gradient in hardness (either from bottom to top of build or left to right ofbuild) was observed. This indicates later passes have minimal effects on the hardness of previously depositedlayers; B — histogram showing bimodal hardness distribution with the UAM build foils significantly harderthan the substrate. The hardness of unconsolidated foils is also overlaid on same plot and is below the peakhardness of the UAM build; C — optical image showing a high hardness foil indent on left (90 HV) next to aweaker interface indent on the right (30 HV). Interface areas with voids or defects caused by insufficient ma-terial flow to fill in grooves cut by sonotrode during previous pass had lower hardness.

A B

B

Linear Void Density (%)0 10 20 30 40 50 60 70 80

0

200

400

600

800

1000

1200

1400

24.7

8.5

31.9

33.8

65.8200 μm

Imag

e H

eigh

(μ

m)

C


distribution from the original Al 3003-H18 foils, which had an average hardnessof 64.5±2.7 HV. The UAM build had anaverage hardness of 73.7±1.9 HV. Thedata show that the UAM builds are indeedharder than the stock foils and provide aqualitative explanation of the increase inULTS. Careful analyses of hardness in-dents in certain regions also showed inter-esting features — Fig. 9C. In one region, asmall indent showing high hardness wasright next to a large indent showing lowhardness. This low hardness was associ-ated with a large planar defect (marked byarrows). Although the weakened regionsmay be explained with the presence of un-bonded areas, it is necessary to evaluatethe hardened regions through detailed mi-crostructure characterization.

Transmission electron microscopesamples from the interfaces from the bot-

tom (near substrate), middle, and top re-gions of the build were extracted throughFIB machining. The electron microscopyimages are presented — Fig. 10A–E. Themicrostructure of the original foil is alsoprovided for comparison — Fig. 10F. In-terestingly, the microscopy images fromthe bottom (Fig. 10A) and middle (Fig.10B) regions failed to show any sharp in-terface region indicating the formation ofa metallurgical solid-state bonding. Thegrains were equiaxed in nature, quite dif-ferent from that of elongated grains of theoriginal Al 3003-H18 tapes — Fig. 10F.This suggests that the bond formation maybe associated with recrystallization. In ad-dition to the equiaxed grains, fine Al-Mn-Fe-based intermetallics were observed inthe samples along the grain boundariesand within the matrix grains. These inter-metallics are found in the original Al 3003-

H18 tapes and do not appear to be af-fected by the UAM process. The interfacemicrostructure (Fig. 10C) from the top re-gion showed interesting features. Theoriginal interface location can be inferredfrom the sudden change in the grain struc-ture. The microstructure in the (n + 1)th

tape shows the original pancake structure,which transitions sharply to a coarse andrecrystallized grain structure close to theoriginal interface location. The mi-crostructure from the nth layer does notshow any pancake structure, rather morerecrystallized structure. Moreover, a re-gion of grain boundary decohesion wasalso observed. This grain boundary deco-hesion was also confirmed with high-mag-nification analyses — Fig. 10D. A surveyof many samples from different regionsalso showed the interface regions con-tained fine recrystallized grains (< 500

MAY 2010, VOL. 89112-s

WELDING RESEARCH

A

C

D

B

Fig. 10 — Six TEM images. A — Bright field TEM image taken from an interface location with apparent good bonding. The interface cannot be determined easily,indicating potential recrystallization across the interface. Small, white Al-Mn-Fe intermetallics can be seen here; B — another interface location again showing the dif-ficulty in discerning the bond line; C — a third interface location where the bond line can be determined, as pointed out by the red arrows. The blue arrow points to asmall void that appears to have migrated from the interface into the bulk of the material; D — high-magnification bright field image of the void in C; E — dark fieldimage showing the high levels of dislocations within the grains and the Al-Mn-Fe intermetallic particles; F — bright field image of the original foil before consolida-tion. The as-rolled structure is pancake-like grains with some dislocations present. Dislocation content significantly lower than that observed after UAM processing.

E F


nm) with relatively low dislocation densityand coarse grains (500 nm to 2 μm) withrelatively high dislocation density asshown by dark field microscopy — Fig.10E. The original foil, before consolida-tion, shows (Fig. 10F) pancake-like grains,as expected from as-rolled material. Dis-locations were present, though a muchlower concentration than in the grainsalong interfaces after UAM processing.

Discussion on Process-Structure-Property Correlations

In order to understand the interfacemicrostructure, it is important to reviewthe steps involved in the UAM process,shown schematically in Fig. 11. In Fig.11A, a first layer has been bonded to thesubstrate, with the top of this layer left ina rough condition after the sonotroderolled over it. When the next layer is ap-plied, the bottom of the new layer is rela-tively flat, creating an interface between asmooth surface and a rough surface — Fig.11B. When the sonotrode comes directlyon top of the interface during the tackingpass, the relative motion between the twolayers creates frictional and deformationalheating and partially collapses asperities

— Fig. 11C. This results in a weak bondbetween the layers, with many voids asshown in Fig. 11D. During the weldingpass, more ultrasonic energy (higherforces and amplitude of vibration) is usedto finish the bond — Fig. 11E. Some resid-ual voids remain, as shown in Fig. 11F.

The final microstructure at the inter-face can be summarized to consist of threeregions as shown in Fig. 11G. During thewelding of the previous layer, the top sur-face of each foil interacts with thesonotrode and becomes rough. This roughsurface becomes the bottom of the follow-ing interface. Where peaks occurred alongthe rough surface, contact was made withthe next foil and a bond resulted (RegionI). This region constitutes recrystallizedmicrostructure (500 nm to 2 μm) acrossthe interface and has good metallurgicalbonding. It is believed when these peaksare brought into contact with the new foillayer sufficient strain energy, tempera-ture, and forces exist to force dynamic re-crystallization. However, where valleys oc-curred due to the sonotrode texture,Region II, they were often too deep tomake contact with the next layer beingadded. This resulted in voids along the in-terfaces and created the Region III mate-

rial on the foil directly above it. Region IIIis the unaffected original foil surface thathas not been touched by either thesonotrode or the foil layer beneath it. Re-gion III was only found on the top surfaceof the interface. Region II material was di-rectly opposite and the cause of Region IIImaterial. In this study, focus was given tounderstand the mechanism of the grainstructure evolution in Region I. Based onthe microstructure from Fig. 10A–C andE, we can conclude that the original pan-cake grain structure was modified to formsub grains with sizes ranging from 500 nmto 2 μm with different levels of dislocationdensity. To understand this reduction ingrain size, we assume this process is simi-lar to that of hot working of aluminum al-loys. The subgrain size (dsub in μm) duringhot working can be related to Zener-Hol-lomon (Zh) parameter and peak tempera-ture (TP) achieved during hot working(Refs. 27, 28).

Equation 1 has been used to estimatethe grain size in both friction and frictionstir welding. The Zener and Hollomon(Zh) parameter has been estimated for

d log Zsub h = − +[ ]−0 60 0 08 1

1. . ( ) ( )


WEL

DING

RES

EARC

H

A

D E F

G

B C

Fig. 11 — Schematic representation of the UAM process highlighting the vari-ous stages. A — Beginning of new layer, top of previous layer textured bysonotrode during previous layer bonding. B — New layer, 2, placed by feedingmechanism in front of sonotrode; C — sonotrode tacks the new layer down, gen-erating frictional heat and forming a weak bond; D — new layer tacked down,many residual voids present; E — sonotrode passes again for welding pass, de-forming the top surface as it passes; F — layer 2 attached, some voids are stillpresent between layers 1 and 2; G — third layer ready to be added. Enlargementof bond interface showing the three regions. Region I is well-bonded material,Region II is valleys carved by sonotrode, and Region III is untouched material.

Fig. 12 — A contour plot estimating the final grain size using the Zener-Hol-loman parameter. The numbered, curved lines represent the possible combi-nations of temperature and strain rate to achieve the same grain size. Dashedlines represent the range of grain sizes found along UAM interfaces. The ver-tical line is the calculated strain rate derived from Equations 3–6.


aluminum alloys as a function of strainrate (ε̊ in s–1) and peak temperature (TP)(Ref. 29).

In order to understand the subgrainstructure in UAM process, Equation 2 wasused. The strain rate during ultrasonic ad-ditive manufacturing is calculated usingthe following approximation. The totaldisplacement due to the plastic deforma-tion, a thin slab of material under thehorn, can be taken as the horn amplitude,i.e., (Δ d = 26 × 10–6 m). This assumesthere is no slippage of the interface mate-rial. The asperity height is estimated as thepeak-to-peak height of the tape surface.This surface is assumed to be a negativeimage of the sonotrode texture, which hasa value of 7 × 10–6 m, as reported by John-son (Ref. 7). Thus, the peak-to-peakheight of the average asperity is 14 × 10–6

m. Furthermore, the height of the asperi-ties is assumed to have negligible changewith respect to time. With these assump-tions, displacement of the bonded regionswith respect to time can be given by theexpression:

Asperity velocity is calculated as the de-rivative of displacement with respect totime:

The shear strain of an asperity is given bythe equation:

Shear strain rate is then found by takingthe derivative of strain with respect to timeas follows:

Over one ultrasonic cycle, an asperitywill have a strain rate varying between±2.3 × 105 rad/s with an RMS value of

±1.1 × 105 rad/s. Because we do not knowthe peak temperatures experienced by in-terface regions, the subgrain sizes wereevaluated as a function of peak tempera-ture and strain rate. This is shown as aform of contour plot — Fig. 12. The cal-culated micro strain rate from Equation 6results in a peak temperature of around300 K for the 500-nm grains and a peaktemperature of 900 K for the 2-μm grains.This range of temperatures is larger thanexpected, but this may have been causedby the approximations in Equations 1 to 6.In Equation 6, perfect transfer of strainwas assumed, no slipping between thesonotrode and the foil was accounted for.Account for slipping, the resulting strain,and therefore peak temperature requiredto achieve a certain grain size, would havedecreased. Meanwhile, Equations 1 and 2were developed for simple monotonoushot working conditions and not reversiblestrains that are experienced during UAM.Gunduz et al. (Ref. 30) estimated a localstrain rate of 1 × 104 s–1 at a temperatureof 513 K based on vacancy calculationsand the diffusion profile observed in alu-minum-zinc ultrasonic welds. This workwas based on finding the vacancy concen-tration required to reduce the meltingtemperature so the observed small meltregion was possible at ultrasonic weldingtemperatures. Their result is within therange of grain size, temperatures, andstrain rates studied here. Conversely,macro strain rates were studied by Gaoand Doumanidis (Ref. 31) by placing astrain gauge near, but not directly be-neath, the welding sonotrode. They foundmaximum strains of 90 × 10–6 over 0.5 s or1.8 × 10–4 s–1. This low strain rate is ex-pected as Gao and Doumanidis measuredmacro strains with a strain gauge of amuch larger size scale than the asperitiesused in Equation 6.

Recently, Johnson has proposed thatthe materials under reversible strainingconditions may exhibit an UltrasonicBauschinger effect (Ref. 7). However,the interaction of these effects with heat-ing and subgrain formation is not clear.In addition, the estimated strain rateshave to be validated based on detailed fi-nite element deformation models (Ref.12), which considers the spatial variationsas well as dynamic strain hardening orsoftening. The localized temperaturealong interfaces may be affected by thefriction and rapid deformation condi-tions. In the current UAM process, thesubstrate temperature is maintained at149°C (422 K). This heat will diffuse fromthe substrate to the entire build. As a re-sult, with the progress of UAM builds, thepreviously welded interfaces will be sub-jected to an isothermal hold close to thistemperature throughout the processingof the build. This isothermal hold is also

expected to induce some of the recrystal-lization and grain growth observed. Thissuggests the need for measuring the spa-tial and temporal variations of the tem-perature during the UAM process. Thiswill be the focus of the future work (Ref.32). The next step is to provide some di-rections to rationalize the measured me-chanical properties. From the above dis-cussions, it is apparent that all UAMsamples will have large voids along the in-terfaces as well as localized hard and softregions. The voids can be treated as em-bedded cracks, which cause stress con-centrations that resulted in brittle frac-ture of the shear and transverse tensilespecimens. The loading of the transversetensile samples results in a mode I frac-ture, while the shear sample loading in-duces mode II fracture. For a given crackand load magnitude, mode I fractureloading typically exhibits the largeststress intensity factor (SIF) followed bymode II fracture (Ref. 33). Because load-ing is parallel to the embedded cracks inthe longitudinal tensile samples, there isno SIF and the strength in this orienta-tion was not reduced. This fracture me-chanics perspective further explains whythe shear and transverse tensile sampleshave lower than normal strengths andbrittle fracture characteristics, while thelongitudinal samples were not weakenedby the presence of voids and insteadfailed with ductile characteristics. Again,the above discussion is simplistic, doesnot provide a predictive capability, anddoes not account for all transients thathave been observed, such as the varia-tions in tensile testing shown — Fig. 6.Further work is necessary to develop de-tailed computational models that incor-porate the spatial variation of mi-crostructure and voids and constitutiveresponse of the bulk and interface loca-tion. To facilitate the development ofconstitutive properties of the interfacelocations, the grain size distributionalong the interface has to be character-ized close to the voids and away from thevoids using orientation-imaging mi-croscopy. This grain orientation and sizedistribution will allow us to develop mul-tiscale models similar to the ones beingdeveloped by Ghosh (Ref. 34).

Finally, in order to overcome the deepchannels carved by the sonotrode, a veryhigh power UAM system is being developedby EWI (Ref. 17). It is believed that higherultrasonic power input, higher amplitudes,and normal forces will increase the plasticflow at the interfaces. This should enablegreater LWD, reducing the inherent stressconcentrations and improving the tensileand shear strength of UAM builds. Higherplastic flows should also improve the metal-lurgical bonds by ensuring all of the oxidesare removed from the interface.

Z expTh

p = ∈ ×

⎡

⎣⎢⎢

⎤

⎦⎥⎥

° 18 7722

,( )

γ�

�

( )( )

td t

h

d

1

=⎛⎝⎜

⎞⎠⎟

+

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥⎥⎥

×1

2(( )

( )

t

h

d th

1

=⎛⎝⎜

⎞⎠⎟

+

⎡

⎣

⎢⎢⎢⎢

⎤

⎦

⎥⎥1

2 ⎥⎥⎥

× v t

h

( )( )6

γ = ⎛⎝⎜

⎞⎠⎟

−tand t

h1 5

( )( )

d t v t cos t°

= = × × ×( ) ( ) . ( 20,000 )

3 3 2π

(4)

d t sin t( ) ( 20,000 )

= × × × ×−26 10 26 π (3)

MAY 2010, VOL. 89114-s

WE

LD

ING

RE

SE

AR

CH


Conclusions

The present study focused on linkingmicrostructure and LWD to mechanicalproperties of ultrasonic additive manufac-turing builds. Using TEM, SEM, and op-tical microscopy along with microhardnessand tensile and shear testing, the micro-scopic and macroscopic properties ofUAM builds were analyzed. The followingwas found:

1. The average shear strength of thetested UAM samples was approximately48% of the expected 110 MPa ultimateshear strength of Al 3003-H18. The aver-age transverse tensile strength was ap-proximately 14% of the expected 200 MPatensile strength of Al 3003-H18. Trans-verse tensile and shear testing results areindicative of bond quality alone; failureoccurs before microstructure becomes significant.

2. Without optimized parameters,UAM weldments result in voids scatteredthroughout all interfaces. This ultimatelycaused the samples to fail in a low ductil-ity manner with low strength values.

3. Image analysis of cross-sectionedsamples found an average linear weld den-sity of 67.4±16.1%. Image analysis oftransverse tensile fracture surfaces foundan average area weld density of 66±2%. Adirect comparison between LWD and areaweld density was not possible based on thesample size.

4. The average longitudinal tensilestrength was approximately 117% of theexpected tensile strength of Al 3003-H18.This indicates the foils were strengthenedduring processing and was confirmed bymicrohardness testing. Microhardnesstesting found the average hardness of theUAM foils increased almost 15%, from64.5±2.7 HV in the original foils to73.7±1.9 HV, during processing.

5. A hypothesis relating grain refine-ment to strain and temperature using theZener-Hollomon parameter was devel-oped. Microstrain rates were estimatedbased on operating conditions to bearound 1 × 105 s–1. From this and an ob-served grain size of 500 nm to 2 μm, an es-timated peak temperature range for theUAM process of 300 to 900 K was calculated.

Acknowledgments

The authors would like to thank the Co-operative Research Program of EdisonWelding Institute for supporting this re-search. In addition, we thank Dr. K. Graff(EWI), Dr. M. Sriram (OSU), and MattShort (EWI) for suggestions and fruitfuldiscussions during preparation of themanuscript. R. M. Hahnlen and M. J.

Dapino are grateful to the member organi-zations of the Smart Vehicle Concepts Cen-ter (www.SmartVehicleCenter.org), the Na-tional Science FoundationIndustry/University Cooperative ResearchProgram (I/UCRC), and the Smart VehicleConcepts Graduate Fellowship Program.

References

1. Kong, C., Soar, R., and Dickens, P. 2004.Optimum process parameters for ultrasonicconsolidation of 3003 aluminium. Journal ofMaterials Processing Technology 146: 181–187.

2. Kong, C., Soar, R., and Dickens, P. 2004.A model for weld strength in ultrasonically con-solidated components. Journal of MechanicalEngineering Science 219: 83–91.

3. Yang, Y., Janaki Ram, G. D., and Stucker,B. E. 2007. An experimental determination ofoptimum processing parameters for Al/SiCmetal matrix composites made using ultrasonicconsolidation. American Society of MechanicalEngineers 129: 538–549.

4. Kong, C., Soar, R., and Dickens, P. 2003.Characterisation of aluminium Alloy 6061 forthe ultrasonic consolidation process. MaterialsScience and Engineering A 363: 99–106.

5. Janaki Ram, J. D., Yang, Y., and Stucker,B. E. 2006. Effect of process parameters onbond formation during ultrasonic consolidationof aluminum Alloy 3003. Journal of Manufac-turing Systems 25: 221–238.

6. George, J., and Stucker, B. E. 2006. Fab-rication of lightweight structural panelsthrough ultrasonic consolidation. Virtual andPhysical Prototyping 1: 227–241.

7. Johnson, K. 2008. Interlaminar subgrainrefinement in ultrasonic consolidation. PhDthesis, Loughborough University, Loughbor-ough, UK.

8. Adams, B. L., Nylander, C., Aydellote, B.,Ahmadi, S., Landon, C., Stucker, B. E., andJanaki Ram, G. D. 2008. Accessing the elastic-plastic properties closure by rotation and lami-nation. Journal of Acta Materialia 56: 128–139.

9. Li, D., and Soar, R. C. 2008. Plastic flowand work hardening of Al alloy matrices duringultrasonic consolidation fiber embeddingprocess. Materials Science and Engineering A498: 421–429.

10. Doumanidis, C., and Gao, Y. 2004. Me-chanical modeling of ultrasonic welding. Weld-ing Journal 83(4): 140-s to 146-s.

11. Janaki Ram, G. D., Robinson, C., Yang,Y., and Stucker, B. E. 2007. Use of ultrasonic con-solidation for fabrication of multi-material struc-tures. Rapid Prototyping Journal 13: 226–235.

12. Zhang, C., and Li, L. 2009. A coupledthermal-mechanical analysis of ultrasonic bond-ing mechanism. The Minerals, Metals and Materi-als Society and ASM International: 196–207.

13. Siddiq, A., and Ghassemieh, E. 2008.Thermomechanical analysis of ultrasonic weld-ing process using thermal and acoustic softeningeffects. Mechanics of Materials 40: 982–1000.

14. Matsuoka, S., and Imai, H. 2009. Directwelding of different metals using ultrasonic vi-bration. Journal of Materials Processing Technol-ogy 209: 954–960.

15. Kong, C. Y., Soar, R. C., and Dickens, P.M. 2004. Ultrasonic consolidation for embed-ding SMA fibers within aluminum matrices.

Composite Structures 66: 421–427.16. Hahnlen, R. 2009. Development and

characterization of NiTi joining methods andmetal matrix composite transducers with em-bedded niti by ultrasonic consolidation. Mas-ter’s thesis, The Ohio State University, Colum-bus, Ohio.

17. Sriraman, M. R., Babu, S. S., and Short,M. 2010. Bonding characteristics during veryhigh power ultrasonic additive manufacturingof copper. Scripta Materialia 62: 560–563.

18. Aluminum Standards and Data. 1993.Aluminum Association.

19. White, D. 2003. Ultrasonic object con-solidation. U. S. patent US 6,519,500 B1.

20. White, D., and Carmein, D. E. E. 2002.Ultrasonic object consolidation systems andmethod. U. S. patent US 6,446,349 B2.

21. White, D. 2003. Ultrasonic consolida-tion of aluminum tooling. Advanced Materialsand Processes: 64, 65.

22. Li, D., and Soar, R. 2009. Influence ofsonotrode texture on the performance of ultra-sonic consolidation machine and the interfacialbond strength. Journal of Materials ProcessingTechnology 209: 1627–1634.

23. Li, D., and Soar, R. C. 2009. Character-ization of process for embedding sic fibers in Al6061 O matrix through ultrasonic consolida-tion. Journal of Engineering Materials and Tech-nology 131: 021016.

24. Hazlett, T. H., and Ambekar, S. M. 1970.Additional studies of interface temperature andbonding mechanisms of ultrasonic welds. Weld-ing Journal 49(5): 196-s to 200-s.

25. Abramoff, M. D., Magelhaes, P. J., andRam, S. J. 2004. Image processing with ImageJ.Biophotonics International 11(7): 36–42.

26. Mouton, P. R. 2002. Principles and prac-tices of unbiased stereology, an introduction forbioscientists. John Hopkins University Press.

27. Frigaard, O., Grong, O., and Midling, O.T. 2001. A process model for friction stir weld-ing of age hardenable alloys. Metallurgical andMaterials Transactions A32: 1189–1200.

28. Midling, O. T., and Grong, O. 1994. Aprocess model for the friction welding of Al-Mg-Si alloys and Al-SiC metal-matrix compos-ites: HAZ temperature and strain-rate distri-bution. Acta. Metall. Mater 42: 1595–1609.

29. Grong, O. 1997. Metallurgical Modelingof Welding, 2nd ed. The Institute of Materials.Materials Modeling Series.

30. Gunduz, I. E., Ando, T., Shattuck, E.,Wang, P. Y., and Doumandis, C. 2005. En-hanced diffusion and phase transformationsduring ultrasonic welding of zinc and alu-minum. Journal of Acta Materialia 52: 939–943.

31. Gao, Y., and Doumanidis, C. 2002. Me-chanical analysis of ultrasonic bonding for rapidprototyping. Journal of Manufacturing Scienceand Engineering 124: 426–434.

32. Schick, D. E. Submitted 2010. Transientthermal response in ultrasonic additive manu-facturing of aluminum 3003. Rapid PrototypingJournal.

33. Dieter, G. E. 1986. Mechanical Metal-lurgy, 3d ed. McGraw Hill.

34. Ghosh, S., Bai, J., and Paquet, D. 2009.Homogenization-based continuum plasticity-damage model for ductile failure of materialscontaining heterogeneities. Journal of Mechan-ical Physical Solids 57: 1017–1044.


WE

LD

ING

RE

SE

AR

CH


Copyright of Welding Journal is the property of American Welding Society and its content may not be copied

or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission.

However, users may print, download, or email articles for individual use.

Documents

Smart Materials and Structures Lab | - Microstructural Characterization … · 2020-01-22 · Microstructural Characterization of Bonding Interfaces in Aluminum 3003 Blocks Fabricated