The Mechanism of Grain Coarsening in Friction-Stir Welded …li.mit.edu/Stuff/RHW/Upload/Chen_MetallTrans_FSW_Ri... · 2010. 2. 11. · 1 1. INTRODUCTION Friction stir welding (FSW)

The Mechanism of Grain Coarsening in

Friction-Stir Welded AA5083 after Heat Treatment

K. Chen+, W. Gan**, C. Kim#, K. Okamoto*, K. Chung++, R. H. Wagoner+

+Department of Materials Science and Engineering, The Ohio State University 2041 College Road Columbus, OH 43210, U.S.A.

** Edison Welding Institute 1250 Arthur E. Adams Drive Columbus, Ohio 43221, U.S.A.

#Materials and Processes Laboratories General Motors R&D Center 30500 Mound Road, MC 480106224, Warren, MI 48090, U.S.A.

*R&D Div., Hitachi America, Ltd. 34500 Grand River Ave. Farmington Hills, MI 48335, U.S.A.

++School of Materials Science and Engineering Seoul National University 56-1 Shinlim-dong, Kwanak-gu Seoul, 151-742, Korea

Submitted to Metall. Mat. Trans. A

Version Revision V44; September 2Dec. 7, 2009

ii

ABSTRACT

Friction stir welding (FSW) takes place in the solid state, thus providing potential

advantages of welds of high strength and ductility because of fine microstructures.

However, post-FSW heat treatment can create very coarse grains, potentially reducing

mechanical properties. AA5083-H18 sheets were friction-stir butt-welded using three

sets of welding parameters representing a wide range of heat input. They were then heat

treated for 5 minutes at 738 K (465 °C), producing grain sizes exceeding 100 μm near the

top weld surfaces, with the coarse grains extending toward the bottom surface to various

degrees depending on the welding parameters. Electron backscatter diffraction (EBSD),

transmission electron microscopy (TEM), scanning electron microscopy (SEM), optical

metallography, inductively coupled plasma mass spectrometry (ICP-MS) and Vickers

hardness testing were used to characterize the regions within welds. Particle pinning was

determined quantitatively and used with Humphreys’ model of grain growth to interpret

the behavior. The mechanism responsible for forming the large grains was identified as

abnormal grain growth (AGG), with AGG occurring only for regions with pre-heat-

treatment grain sizes smaller than 3 μm. Second-phase particle volume fractions and

sizes, textures, dislocation contents, and solute concentrations were not significantly

different in AGG and non-AGG regions. Ultra-fine grain layers with grain diameters of

0.3mm were characterized and had high densities of pinning particles of MgSi2, Al2O3

and Mg5Al8. Strategies to eliminate AGG by alloy and weld process design were

proposed.

1

1. INTRODUCTION Friction stir welding (FSW) is a relatively new welding technique, invented in

1991.[1][1] It has broad promise for joining similar and dissimilar materials which are

difficult or impossible to weld through conventional welding techniques. FSW can

produce a fine microstructure with fewer defects, lower residual stresses, less distortion,

better retained mechanical properties, and better dimensional stability as compared with

conventional welding.[2-5][2-5] FSW is also a candidate process for joining sheets of

dissimilar thickness or composition to create tailor-welded blanks[6-8][6-8] that retain the

capacity for enhanced ductility and strength via fine grain structure. Recent research

focuses on FSW in joining materials such as magnesium alloy and steel.[9-12][9-12]

Commercialization of FSW, mainly for aluminum alloys, has occurred in the

transportation industry, for applications such as automobiles, railway vehicles, ships, and

rockets.[13-17][13-17]

1.1. FSW Background

FSW is achieved by severe plastic deformation and heating induced by friction

and plastic work created by the rotation of a non-consumable tool embedded in the

workpiece, Fig. 1Fig. 1.[4][4] Severe microstructural changes can occur in the welded

region.[4, 18-25][4, 18-25] Peak temperatures, reportedly as high as 0.6-0.8 times the melting

temperature,[26-28][26-28] have been correlated to combinations of welding parameters ν

(weld feed rates) and ω (tool rotation speeds) through a “heat input parameter” or a “heat

index”, ω2/ν[26, 29][26, 29] or ω/ν[30][30], respectively.

Three distinct microstructural zones have been identified in the welded region –

the weld nugget zone (WN), the thermo-mechanically affected zone (TMAZ), and the

heat-affected zone (HAZ).[4][4] The weld nugget zone (delineated approximately by the

dashed line in b), or “stirred zone”, refers to a bowl-shaped[31, 32][31, 32] or elliptical[5, 33,

34][5, 33, 34] region that is highly strained and dynamically recrystallized at elevated

temperature. Dynamic recrystallization and texture development occur in the weld

nugget zone (delineated approximately by the dashed line in Fig. 1Fig. 1b) during FSW,[5,

28, 35-39][5, 28, 31-35] while precipitate dissolution and coarsening occur in all three zones[35, 39,

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Formatted: Font: Not Bold, (Asian)Korean

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Formatted: Font: Not Bold

Field Code Changed

Field Code Changed

2

40][31, 35, 36]. A fine and equiaxed grain structure is obtained after FSW with grain size

ranging from 1 μm to 10 μm.[18, 35, 37, 39-41][18, 31, 33, 35-37] Ultrafine-grained microstructures

with average grain size smaller than 1 μm have also been obtained using special cooling

methods.[18, 42][18, 38]The thermo-mechanically affected zone surrounds the weld nugget

zone; it experiences more moderate temperatures and strains, and consequently no

recrystallization.[43-45][35-37] The heat affected zone surrounding the thermo-mechanically

affected zone experiences only thermal disturbance during FSW, with possible

coarsening of the microstructure.[37, 41][38, 39] Particles may be coarsened,[46][40] or fully or

partially dissolved during FSW[25, 39, 47][25, 41, 42].

1.2. Recrystallization during and after FSW

Dynamic recrystallization and texture development occur in the nugget zone,[5, 28,

35-39][5, 28, 38, 41, 43-45] while precipitate dissolution and coarsening occur in all three zones[35,

39, 40][41, 43, 46]. A fine and equiaxed grain structure is obtained after FSW with grain size

ranging from 1 μm to 10 μm.[18, 35, 37, 39-41][18, 38, 39, 41, 43, 46] Ultrafine-grained

microstructures with average grain size smaller than 1 μm have also been obtained using

special cooling methods.[18, 42][18, 47]

Post-FSW heat treatment may be carried out for annealing or aging purposes, it

may be necessary for further forming (i.e. using FSW tailor-welded blanks for GM’s

Quick Plastic Forming process at elevated temperature[48][39][48]), or it may be an

incidental effect of, for example, the automotive paint-bake hardening cycle. Grain

growth to sizes as large as several millimeters can occur during heat treatment in FSW

AA7075,[49][40][49] AA6061,[31][41][31] AA7010,[33][42][33] AA7475,[5][5] and AA2095[32][43][32].

Large grains are known to degrade the mechanical properties of welds.[50-53][44-47][50-53]

The formation of coarse grain structures in FSW appears to be discontinuous[5, 33,

34, 54, 55][5, 42, 48-50][5, 33, 34, 54, 55] because an obvious initiation step was observed[34][48][34]. In

such “initiation” step, several much larger grains formed in a surrounding of stable

smaller grains. The subsurface region, which directly contacts the welding tool shoulder,

is the one most often reported to initiate coarse grain structure formation[34, 54, 56][48, 49,

51][34, 54, 56] Coarse grains may also form from the bottom of the nugget[33, 34, 56][42, 48, 51][33,

34, 56] and on the edges of nugget close to the WN-TMAZ boundariesthermo-mechanically

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Formatted: Do not check spellingor grammar

Field Code Changed

3

affected zone[34, 54][48, 49][34, 54]. The regions most susceptible to discontinuous grain

coarsening are those with smaller grains and higher fraction of low angle boundaries

(%LAB) in the nugget.[33, 54, 57][42, 49, 52].[33, 54, 57, 58][33, 54, 57, 58]

Primary recrystallization and secondary recrystallization are two possible

discontinuous processes leading to the coarse grain structure formation. Similar to a

discontinuous phase transformation process, there are also two characteristic steps in

primary recrystallization and secondary recrystallization: initiation and growth.[59-63][53-57]

Borrowing the nomenclature from discontinuous phase transformation, the “initiation” is

also called “nucleation”. In this paper, “nucleation” and “nuclei” are used in order to be

more consistent with previous literature, but it should be noted that the term “nucleation”

as used here does not occur by thermal fluctuation or atom by atom construction. The

nuclei in primary recrystallization are either recovered cells or subgrains pre-existing in a

deformed microstructure;[62, 63][56, 57] those in secondary recrystallization are actually

grains already present in initial grain structure[64, 65][58, 59].

The driving force of primary recrystallization is the stored strain energy in the

form of dislocation content within grains.[62][56] It is associated with the formation of new

strain-free grains in a deformed matrix and the subsequent growth of these new set of

grains.[62, 63][56, 57] The formation of coarse grain structure by primary recrystallization is

a result of the competing of nucleation rate (.

N ) and nuclei growth rate (.

G ). A smaller

ratio of ..

/ GN leads to fewer nuclei and therefore a coarser final grain structure after these

nuclei grow to impingement.[61, 66, 67][55, 60, 61]

Secondary recrystallization is driven by the interfacial energy of grain

boundaries.[64, 68][58, 62] It takes place after primary recrystallization when normal grain

growth is inhibited, except for a few favored grains.[64, 68][58, 62] In order to distinguish it

from normal grain growth (NGG), it is also commonly called abnormal grain growth

(AGG) or discontinuous grain growth.[64, 65, 69][58, 59, 63] For simplification and clear

presentation, the terminology “AGG” is used here instead of secondary recrystallization.

1.3. Abnormal grain growth (AGG) background

AGG can produce grain sizes of hundreds of microns to several millimeters.

Large grains grow more slowly than smaller ones in an ‘ideal grain assembly’ where the

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

4

energies and mobilities of all boundaries are equal.[64][58] Only when such normal grain

growth is inhibited is AGG possible. Therefore, second phase particles, texture, surface

effects, and solute segregation which lead to diverse driving forces and mobilities of local

boundaries are important factors leading to AGG.[50-53, 64][44-47, 58] Various models of

AGG have been proposed, emphasizing the effects of the second phase particles on

pinning the grain boundary movement.[50-53, 70][44-47, 64]

The Humphreys’ a model for recrystallization and grain growth,[71][65] is based on

a cellular microstructure of a mean grain diameter gd , uniform intrinsic grain boundary

energy and mobility, and a volume fraction ( pf ) of spherical particles of uniform

diameter pd .[71][65] The behaviors of “particular grains” are considered, where the size

ratio of a particular grain, X, is defined as gg ddX /= , where gd is its diameter and gd

is the average grain diameter. AGG occurs if such a “particular grain” grows faster than

the matrix, i.e.:

0>⎟⎟⎠

⎞⎜⎜⎝

⎛

g

g

dd

dtd or 0>−

dtdd

ddt

ddd gg

gg (1)

Based on Hillert’s formulation[72][66] for grain growth driving force and Zener’s

particle pinning force[73][67], Humphreys derived the criterion for AGG as follows:

( ) ( ) 044414 2 >−Ψ−+−Ψ XX (2)

where Ψ is a “particle pinning parameter”, equal to p

pg

dfd

23

.

The roots of the equation based on Inequality 2 (X1 and X2, where X2 > X1),

represent critical particular grain sizes for transitions between no grain growth and AGG.

For Ψ < 0.25, normal grain growth (NGG) is possible and AGG is possible for grains of

size ratio X in the following range: X2 (=Xmax) > X > X1 (=Xmin). For Ψ > 0.25 (the case

applicable for conditions in the current study), NGG is not possible but AGG can occur

for a grain with size ratio X > X2 (=Xmin), where Xmin has the following value:

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

5

142222)25.0(

2

2min −Ψ−Ψ+Ψ+Ψ

=>Ψ= XX (3)

The branches of Inequality 2 are shown graphically in Fig. 2Fig. 2, where the

regions corresponding to NGG, AGG, and no grain growth are shown. For Ψ > 0.25,

with the increase of the pinning parameter Ψ, the required minimum size ratio (Xmin) for

AGG increases. For a normal distribution of grain sizes, the largest grain is 2.5 gd

(dashed horizontal line in Fig. 2Fig. 2). For Ψ ≥ 1, no grain can grow because the

particle pinning is so strong that Equation 2 cannot be satisfied for any particular grain

size ratio X.

1.4. Second-phase particles in AA5083

Because of the crucial roles of particles in pinning grain boundary, knowledge of

the important second-phase particles in FSW AA5083, is necessary. Based on the likely

effectiveness as grain boundary pins during the post weld heat treatment (465 °C),

particles in AA5083 may be segregated into four categories: In particular, the following

particles have been identified as present in AA5083:[46, 74-81] (Fe,Mn)Al6, Al2O3, Mg2Si,

(Fe,Cr)Al7, Mg5Al8, Al12(Fe,Mn)3Si, and Al11Cr2. These particles may be segregated into

four categories based on their likely effectiveness as grain boundary pins at the heat-

treatment temperature of 465 °C:

A. Fine and stable particles: dispersoid (Fe,Mn)Al6 and Al2O3. Dispersoid

(Fe,Mn)Al6 particles (< 0.4 μm) are known to be semi-coherent and effective in

the pinning of grain boundaries[77, 79, 82][68-70] and are susceptible to coarsening

above 500 °C.[46, 77][68, 71] They have a plate-like appearance that has been

associated with the development of pancake-shaped grains.[77][68] Al2O3 particles

have been identified in FSW weld nugget zones,[80, 81][72, 73] likely arising from the

breakup and entrainment of the surface oxide. The melting point of Al2O3 is 2327

K (2054 °C),[83][74] thus these particles are very stable during the post weld heat

treatment (465 °C).

B. Coarse, incoherent, constituent particles: Mg2Si, large (Fe,Mn)Al6, (Fe,Cr)Al7.

These particles are formed during solidification, with non-uniform



Field Code Changed

Formatted: Numbered + Level: 1 +Numbering Style: A, B, C, … +Start at: 1 + Alignment: Left +Aligned at: 0.63 cm + Tab after: 1.27 cm + Indent at: 1.27 cm

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

6

distribution.[74][75] Mg2Si is thermally stable, with a melting point of 1360 K

(1087 °C).[84][76] The melting temperature of CrAl7 is 790 °C (1063 K),[85][77]

higher than the melting temperature of AA5083 (574 °C, or 847 K[86][78]). Other

than (Fe,Mn)Al6 dispersoids, (Fe,Mn)Al6 can be particles of several

microns,.[46][71][40] These particles have been reported to be unimportant to grain

boundary migration,[75][79] although they may inhibit the migration of grain

boundaries by creating holes in the grain boundaries.[87][80]

C. Thermally unstable particles: Mg5Al8. Mg5Al8 is not heat resistant:[84][76] an aging

temperature of 623 K (350 °C) eliminated Mg5Al8[88][81] while it precipitated out

during deformation at 608 K (335 °C)[89][82]. The solvus temperature of Mg is 573

K (300 °C). Therefore, it is expected that Mg5Al8 is unstable during typical post-

weld heat-treatments at 738 K (465 °C), and thus should have little effect on grain

boundary pinning, above approximately 300 °C – 350 °C.

D. Rarely reported particles: Al12(Fe,Mn)3Si and Al11Cr2. These particles have

occasionally but rarely been reported in AA5083, with Al12(Fe,Mn)3Si identified

as constituents and Al11Cr2 identified as dispersoids.[75][79] Because

crystallographic data needed to index these phases through OIM are not available

in literature and they cannot be identified unequivocally, they have been ignored

in the current EBSD study. However, with their larger Z number than matrix Al,

they will be imaged via SEM in backscatter (BSE) mode.

1.5. Grain coarsening in heat-treated AA5083 FSW

Fig. 3Fig. 3 and Fig. 4Fig. 4∗ illustrate the role of FSW conditions and heat

treatment time on the occurrence of coarse grain structure following heat treatment.[90][83]

Coarse grains are favored by lower FSW temperatures as measured or as predicted by

lower heat input or lower heat index,[26, 29, 30][26, 29, 30] Fig. 3Fig. 3. (As described later, the

“heat input” and peak temperatures encountered during FSW increase with higher

rotational speed, ω, and lower translational speed, v.) For the extreme low-heat-input

case shown, i.e. ω = 1000 rpm and v = 300 mm/min, coarse grains are observed

∗ Fig. 3Fig. 3 and Fig. 4Fig. 4 are taken from reference 90 and used with permission of the Journal of Engineering Materials and Technology - Transactions of the ASME.

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed



Field Code Changed


Formatted: Do not check spellingor grammar, Superscript

Field Code Changed


Formatted: Font: 10 pt, Not Bold

Formatted: Font: 10 pt, Not Bold

7

throughout the weld nugget region. For the extreme high-heat case, ω = 1500 rpm and v

= 150 mm/min, coarse grains appear only at the top surface of the weld nugget. Coarse

grains are favored by longer heat treatment times, Fig. 4Fig. 4, starting from the top

region and proceeding downward as the heat-treatment time (at 465 °C) increases. Fig.

3Fig. 3 and Fig. 4Fig. 4 suggest that the top part of the weld nugget has a microstructure

which promotes coarse grain structure relative to other areas. No coarse grains are found

outside of the weld region, even for extended heat treatment times, consistent with the

literature for AA7075.[34][34] This is made apparent by comparing the weld nugget zone

in b with the extended heat-treatment micrographs, , Views (7) and (8).

Dynamic recrystallization dominates the initial fine grain formation in the weld

nugget during the FSW process, producing very fine grains[28, 42, 91, 92], ranging from 0.1

μm[42][47] to 0.8 μm[28][28]. These grains are typically smaller than the original subgrains

(about 1 μm[25, 44][25, 36]), tending to confirm the importance of dynamic recrystallization

near the tool pin as established by high energy storage, and large strain and strain

gradients, complex strain paths.

In order to assess the mechanisms for coarse grain structure formation during

heat-treatment of aluminum alloys, aluminum alloy 5083-H18 was friction-stir butt-

welded using three sets of weld parameters representing a wide range of heat input.

Aluminum alloy 5083 is of interest for superplastic forming in transportation industry[93-

100][84-91], where rapid grain growth in FSW regions could be detrimental to final service

properties. Following analysis of the as-welded microstructures, samples were heat

treated for 5 minutes at 738 K (465 °C). The material and heat-treat cycle represent those

used in elevated temperature forming in the GM Quick Plastic Forming process,[48][39][48]

a target application. Microstructures between weld regions of various heat-input samples

and between different locations within a single weld were compared to reveal the

controlling microstructural factors for the undesirable grain coarsening. As-welded

samples were also heat-treated at other temperatures for various length of time in order to

study the effect of recovery annealing on the coarse grain formation.




Field Code Changed

Field Code Changed


Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed



Field Code Changed

8

2. EXPERIMENTAL PROCEDURES Commercial AA5083 is a non-age hardenable aluminum alloy with additions of

Mg for solid solution strengthening. Its chemical composition (in weight percent) is

0.40 % max Si, 0.40 % max Fe, 0.10 % max Cu, 0.40 % to 1.0 % Mn, 4.0 % to 4.9 % Mg,

0.05 % to 0.25 % Cr, 0.25 % max Zn, 0.15 % max Ti, and remaining Al.[86][78] It was

obtained in the fully strain-hardened temper (H18), with thickness of 1.6 ± 0.03 mm.

2.1. FSW conditions

AA5083-H18 sheets were welded by Hitachi using a 3D-FSW machine with

welding tool made of matrix high speed tool steel.[101][92] The diameters of the shoulder

and the pin are 10 mm and 4 mm respectively. The pin is threaded for increasing the stir

effect and lowering the welding force. The pin length is 0.3 mm less than the sheet

thickness (1.6 ± 0.03 mm) and the tool shoulder penetrated approximately 0.2 mm into

the workpieces during welding. All the welding directions were aligned with the rolling

direction (RD). The as-welded plate with locations and orientation labels is

schematically shown in Fig. 1Fig. 1a.

As shown in Table 1Table 1, the three weld conditions are labeled H1, H2, and

H3 in order of progressively higher temperatures measured during their welding (and

corresponding to the progressively higher indices as shown).[26, 29, 102][26, 29, 93] Two such

indexes have been proposed, as used in Table 1Table 1:

Heat Index 1[26, 29][26, 29]: v/2ω (4)

Heat Index 2 [30][30]: v/ω (5)

where ω is the tool rotation speed (rev/s) and ν is the tool translational rate (mm/s).

These heat indexes have been correlated phenomenologically with observed

temperatures and times-at-temperature during FSW. For a given translational speed, a

higher tool rotation speed produces more stirring in a spatial region, thus producing

higher temperatures. Conversely, for a given rotational speed, a slower travel will

produce more heat at a given location.

Table 1Table 1 compares measured peak temperatures and the residence times

above 300 °C and 400 °C 2 mm from the weld centerline (i.e. at the pin-workpiece

interface) and 0.6 mm below the initial top surface of the 1.6 mm sheet. The

Field Code Changed

Field Code Changed





Field Code Changed




Field Code Changed


Field Code Changed



9

methodology by which these measurements were made has been described

elsewhere.[90][83][89] Note that measured times-above-temperature are longer for H2 than

for H3 although the peak temperatures are slightly lower. This is reflected in Heat Index

2, which is nearly equal for the two cases.

Table 1. Welding parameters and associated thermal characterization

Label H1 H2 H3 Tool rotation speed, ω – rpm (rps) 1000 (17) 1000 (17) 1500 (25) Translational rate, v – mm/min (mm/s) 300 (5.0) 100 (1.7) 150 (2.5) Heat Index 1, ω2/ν[26, 29][26, 29] – (rps)2/(mm/s) 56 163 250 Heat Index 2, ω/ν[30][30] – rps/(mm/s) 3.3 9.8 10.0 Peak Temp, AS, RS*[90][83] – °C 459, 400 522, 476 526, 500 Avg. Peak Temp*[90][83] – °C 430 499 513 Time Above 300 °C (Avg. AS, RS)[103][94] – s 2.6 10.8 6.5 Time Above 400 °C (Avg. AS, RS)[103][94] – s 0.7 5.6 4.1

Key: AS, RS = Advancing side, Retreating side

2.2. Sample preparation, heat treatment, hardness testing

Test coupons (10 mm x 25 mm) were sectioned from the welded sheets with weld

region at the longitudinal center, then heat treated for 5 minutes at 738 K (465 °C). This

heat treatment is denoted here as QPF because it represents that used before the GM

Quick Plastic Forming (QPF) process,[48][39][48] a target application. Additional test

coupons were used to study alternate heat treatments.

Full annealing of AA5083 (solidus temperature is 847 K, or 574 °C[86][78]) to

achieve the “O” condition is typically done at 618 K (345 °C) for the time required to

heat the whole sample to temperature.[104][95] Full annealing in the current work has been

done at 618 K (345 °C) for 1 hr and 380 °C. During all heat treatment, an aluminum

sheet was preheated in the furnace and coupons were placed on the sheet to promote

rapid heat transfer. The heat treat times were recorded from the time that the furnace

regained the set temperature after placement of the coupons.

Coupons were mounted using Epoxy or conductive Bakelite for hardness testing

using a Buehler Micromet II digital micro hardness tester along the centerline of the weld

Field Code Changed


Field Code Changed

Field Code Changed

Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed

Field Code Changed

Field Code Changed



10

from top surface to bottom on the LT-ST plane (see Fig. 1Fig. 1). The samples were

mechanically polished, and a load of 50 g force and dwell time of 15 seconds were used.

At least five Vickers microhardness test values were recorded at each distance from the

top surface near the centerline and were averaged after excluding the maximum and

minimum values in order to minimize the effect of locally varying conditions, such as

locating the indentation on a cluster of particles. Along the thickness, ten positions were

measured for Vickers hardness. Therefore, a minimum of 50 hardness measurements

were made on each welded sample. For unwelded specimens, the measurements were

averaged to obtaina single hardness value.

Macro- and micro-structural characterization was done on the cross sectioned

plane (LT-ST) unless otherwise noted. All samples were successively ground using

silicon carbide grinding paper and then polished using diamond compounds (final grit: 1

μm) for optical microscopy (OM). Samples were etched with a solution of 12 mL HCl

(conc), 16 mL HNO3 (conc), 17 mL H2O, 1 mL HF (48 %), and 4.8 g chromic acid on

sample surface for 1 to 4 min [105][96][105].

2.3. EBSD procedures

Samples for electron backscattered diffraction (EBSD) analysis were polished as

for optical microscopy and then using 0.05 μm colloidal silica in a vibratory polisher.

EBSD scans were conducted using TSL OIM Data Collection Program on a Philips XL-

30 Environmental SEM (25kV) with field emission gun. Low angle boundary

measurement was found to be dependent on the settings of EBSD collection, especially

on step size, the binning of Kikuchi band images, and the minimum cutoff misorientation

value. The step size of the scans was chosen to be 0.25 μm, the binning for the Kikuchi

pattern image to be 4 x 4, and the minimum cutoff value wais 3°. Low-angle boundaries

(LAB) in this paper are defined as having misorientation angles from 3° to 15°. For most

scans the scan area wais 50 by 50 μm2. Hough peaks of scanned points were collected

first. Using the known crystallographic structures of Al, (Fe,Mn)Al6, Mg2Si, Mg5Al8,

and (Fe,Cr)Al7, and Al2O3 phases, the initially collected Hough-peak data were used to

generate orientation data for identifying these phases. , , and pole

figures for the top and middle regions of H1, H2, and H3 samples were recorded by OIM.


Field Code Changed


11

Coordinate axes are ST and LT as shown in Fig. 1a. , , and pole

figures for the top and middle regions of H1, H2, and H3 samples were recorded by OIM

2.4. Grain size conversion

The orientation data were analyzed using TSL OIM Analysis Program, with a 5-

degree grain tolerance angle, that is, 5 degrees of misorientation is the minimum for

defining a grain boundary. The grain size statistics were generated directly by the TSL

OIM Analysis Program. The accuracy and reproducibility of grain sizes were established

by performing 3 scans at adjacent regions in specimens of H1-Mid and H3-Mid, and

calculating standard deviations of the measured values. The uncertainty was found to be

within 4% of the measurement. In grain growth model, mean grain diameter is used.

Therefore, grain size in this paper is represented by grain diameter. With the assumption

that all grains are spherical, the mean 2D equivalent circle diameter (ECD) for

characterization locations were calculated from EBSD measured average grain area ( A )

as follows:[106][97] 2/1)/(2 πAECD ×= (6)

The 3D grain diameter ( gd ) was converted from the ECD based on a published

stereological relationship:[106][97]

ECDECDd g 225.1)3/2(2/1 == − (7)

2.5. SEM procedures

Samples for scanning electron microscope (SEM) investigation were prepared as

for EBSD. A Philips XL-30 Environmental SEM (with field emission gun) was operated

at 15 kV for imaging and an energy dispersive X-ray spectroscopy (EDS) was used for

chemical analysis. A Sirion high-resolution FEG-SEM was operated at 5 kV under

backscattered electron (BSE) mode for analyzing second phase particle larger than 0.2

μm. All photomicrographs were taken under a constant magnification of 1500X, and

approximately a total of 3000 particles were analyzed. The resolution of these

photomicrographs was such adjusted to make a single pixel size corresponding to 0.016

μm. A low electron voltage of 5 kV was used to reduce the information depth.

Compared with an information depth of about 400 nm at 10 kV, it is only about 120 nm


Formatted: Font color: Blue

Field Code Changed



Field Code Changed



12

for a 5 kV beam.[107][98] A large information depth will bias the SEM analysis away from

the 2-D condition, producing biased results of particle sizes and volume fractions which

are calculated based on stereologic theories relating exact 2D and 3D information.[108][99]

2.6. TEM procedures

Second phase particles smaller than 0.2 μm were imaged using TEM. Thin-foil

disk specimens from various weld locations were mechanically thinned to about 120 μm

and then electro-polished using a solution of 70% methanol and 30% nitric acid at about

12 °C. A Philips CM-12 transmission electron microscope (TEM) was used, operating at

120 kV. Convergent beam electron diffraction was used for sample thickness estimation.

Because of the projection nature of TEM imaging and following the general assumption

that all particles are spherical, the measured sizes can be taken as 3D sizes. Particles with

diameter smaller than 200 nm were quantified by TEM. Volume fractions of particles

were obtained by the total volume of particles divided by the measured sample volume

(imaging area multiplying sample thickness).

2.7. Mass spectrometry

A Perkin-Elmer Sciex ELAN 6000 Inductively Coupled Plasma Mass

Spectrometer (ICP-MS)[109][100] with laser ablation was used to measure the spatial

distribution of magnesium content (top to bottom of weld). Samples were cross-

sectioned, and measurements were made on ST-LT plane, Fig. 1Fig. 1b. The laser beam

was scanned from top to bottom along the centerline of the H2 weld, and top-to-bottom

for the base H18.

Field Code Changed



Field Code Changed



Field Code Changed




13

3. EXPERIMENTAL RESULTS A typical optical FSW macrostructure, Fig. 1Fig. 1b, shows a basin-shaped

dynamically recrystallized weld nugget zone. After heat treatment at 465 °C for 5 min

(QPF), etching reveals the extent of coarse grained area, depending on weld parameters,

Fig. 5Fig. 5. The lower the heat input is, the more extensive the coarse grained area,

spreading from the upper surface. Microstructure analyses focus on the top and middle

regions of these three samples to differentiate the conditions favoring the production of

coarse grains. The centers of all the 50 by 50 μm2 scans for top regions are

approximately 50 μm below the sample top surface to avoid near-surface effects.

The H2 condition was chosen to further explore the annealing behavior of FSW

welds because 1) only part of the weld experienced coarse grain formation during QPF, 2)

this welding condition was determined to be optimal for mechanical properties,[90][83] and

3) this condition was used to generate Fig. 4Fig. 4.[90][83]

As shown in Fig. 6Fig. 6, coarse grains begin to appear at the top surface at

420 °C for 5-minute annealing. For higher annealing temperatures, the coarse grained

area extends downward progressively. For 510 °C and above, the whole weld region

becomes coarse grained. A closer examination of the 510 °C, 540 °C, and 560 °C

annealing cases shows that the final grain size in the center of the sheet decreases as the

annealing temperature increases. Also, elongated coarse grains begin to extend into the

base material at these temperatures. This observation differs from an earlier report that

coarse grains cannot extend outside of FSW welds into the base metal.[34][48][34]

In order to see the grain coarsening behavior and its relationship to the FSW

process, a plan-view of the H2 weld after QPF is shown in Fig. 7Fig. 7a. A convolution

pattern is clearly seen corresponding to the rotation of the welding tool, with the pattern

concave in the welding direction. A previous study reported a similar pattern.[45][101][37]

Particle-rich (dark band) and particle-poor (bright band) regions follow a similar

convolution pattern, Fig. 7Fig. 7b. This similarity suggests the importance of particle

pinning in the coarse grain formation.

3.1. Hardness measurements



Field Code Changed




Field Code Changed




Field Code Changed


Field Code Changed


14

The hardness of the H18 base material was 132 ± 2 Hv. After annealing at 345 °C

and 380 °C for 1 hr, hardnesses of 69 ± 2 Hv and 67 ± 3 Hv were achieved respectively.

(The error for each Hv measurement is the standard deviation of at least three

measurements in closely adjacent regions.) Published hardness of the O condition is

approximately 67 Hv.[46][71][40]

Plots of hardness values along the weld centerlines of as-welded conditions and

after QPF are compared with base H18 and O materials in Fig. 8Fig. 8. There is little

variation of hardness laterally (standard deviation of 1 Hv) in the weld zone within 4 mm

of the weld centerlines.[90][83] For all as-welded conditions, the top region is harder than

the middle region by 2-3 Hv. This result contrasts with those for a heat-treatable alloy

(AA7050-T7451) where the top region is softer.[44][102][36]

After QPF the whole thickness of FSW-H1 and the upper region of FSW-H2 and

FSW-H3 exhibit coarse grains and hardness reduced by 6-8 Hv while the lower regions

of FSW-H2 and FSW-H3 have unchanged hardness, within 2 Hv. For the H2 case, the

hardness readings in the grain coarsening regions were reduced from 75 ± 2 Hv to 69 ± 3

Hv while the non-coarsening regions were 72 ± 1 Hv and 71 ± 1 Hv before and after QPF.

As shown in Table 2Table 2, QPF has little or no effect on hardness except in the regions

where the grain coarsening occurs.

Table 2. Hardness (Hv) upper (where coarse grains form after QPF) and lower (where fine grains remain after QPF) of FSW-H2 at various heat treatment stages

FSW-H2 as welded 350°C/24hr 380°C/24hrQPF (465°C/5min)

350°C/1hr+QPF

380°C/24hr +QPF

Upper 75 ± 2 72 ± 2 71 ± 1 69 ± 3 67 ± 2 68 ± 2 Lower 72 ± 1 71 ± 1 70 ± 1 71 ± 2 72 ± 2 71 ± 2

3.2. Recovery annealing tests

A series of two-stage heat treatments was conducted to probe the nature of the

driving force for the formation of the coarse grains. Recovery anneals of FSW-H2

material were performed over a range of times and temperatures as follows: 80°C/1hr,

Field Code Changed


Field Code Changed



Field Code Changed


15

150°C/1hr, 200°C/1hr, 250°C/1hr, 300°C/1hr, 350°C/1hr, 350°C/24hr, 370°C/24hr, and

380°C/24hr. (Temperatures above 380 °C were not used because at 380 °C large grains

were formed at the boundary of weld nugget, Fig. 9Fig. 9, as marked in Fig. 9Fig. 9a

View (3).) Also shown in Fig. 9Fig. 9, even the most aggressive recovery anneal has no

apparent effect on the formation of large grains during the subsequent QPF heat treatment.

With or without the recovery anneal, the QPF heat treatment produces coarse grains only

in the upper fourth of the H2 weld.

EBSD characterization was conducted on the marked region in Fig. 9Fig. 9a View

(3), where the grain coarsening initiation appears. As shown in Fig. 9Fig. 9b, those large

grains (~ 100 μm) were found to be surrounded by fine grains with a size ratio of about

40. (Fig. 9Fig. 9b is an EBSD grain structure map. Color code indicates the crystal

orientation. Therefore an area belonging to a single grain shows similar color.)

Typical reduction of hardness by recovery is shown as a function of annealing

time at 350 °C in Fig. 10Fig. 10 and as a function of annealing temperature for 24-hour

anneals in Fig. 11Fig. 11. In these figures, the average hardness readings are shown

separately for the regions that show the formation of coarse grains (upper fourth of the

thickness) and those regions that do not (lower three-fourths of the thickness∗) eventually

after QPF heat treatment.

The hardness measurements for the most aggressive recovery anneals at various

stages are summarized in Table 2. Recovery of upper regions (where coarse grains

eventually form) reduces the hardness from 75 to 71-72 but leaves the hardness of lower

regions (fine grains) unchanged. Within the typical measurement scatter of 2-3 Hv, the

upper (coarse grain) regions are reduced from 75 Hv to 68 Hv (similar to O condition)

after QPF, independent of whether a pre-anneal is performed. (68 Hv is the hardness

expected for a full anneal heat treatment to the O condition of rolled sheet.) The lower

(fine grain) regions have an unchanged hardness of 71 Hv as welded, before or after

recovery annealing, before or after QPF. The lower regions remain slightly harder than

full O-annealed sheet.

∗ The hardness measurements closest to the bottom of the sheet were excluded because they differ sharply from the remainder of the region.









16

The recovery anneal results show that recovery annealing a) has no effect on the

subsequent growth of coarse grains during QPF heat treatment, and b) the hardness of the

lower (fine grain) regions are unaffected by either the recovery anneal or subsequent QPF.

Thus, it may be concluded that the driving force for the formation of coarse grains is not

the elimination of dislocation content, i.e. that recrystallization does not occur throughout

the FSW-H2 weld zones, even for heat treatments at higher temperatures and for longer

times than that required to obtain the O condition for strain-hardened AA5083.

3.3. EBSD imaging: as-welded

The microstructures of the as-welded specimens are presented in this section.

Those after QPF (465 °C for 5 min), and after various recovery anneals are presented in

the next section.

An EBSD phase-contrast map, Fig. 12Fig. 12, reveals the grain structure and

known second phase particles of as-received H18 base material: pancake-shaped of

widely varying grain size. Intercepts give dg,1 = 3.0 μm (LT direction) and dg,2 = 1.3 μm

(ST direction). Given that grains in rolled sheets are usually larger in the RDL direction

than in LT, the average grains intercept may be approximated as 3 μm (i.e. average of 3,

1, and 5 μm).

The effects of particle shape and distribution are important factors in

recrystallization and grain growth via changing the boundary mobility,[73][67] with the

largest influence from high densities of particles of size smaller than 1.0 μm.[64, 76, 88][58, 81,

103] EBSD is useful for identifying second-phase particles, but is semi-quantitative at best

for particles of size less than approximately 200nm. In view of many particles in the

range of ~ 30 nm, quantification of sizes and volume fractions was accomplished using

SEM and TEM, as described later.

Fig. 13Fig. 13 is similar to Fig. 12Fig. 12, but compares weld locations shown in

Fig. 5Fig. 5. All locations are recrystallized, with smaller grains near the tops of the

welds. The average grain size, gd d, determined by the mean equivalent 3D diameter,

appear on each figure with uncertainty of ± (0.02-0.04)×d gd , that is, within 4% of the

Formatted: Outline numbered +Level: 2 + Numbering Style: 1, 2,3, … + Start at: 3 + Alignment:Left + Aligned at: 0 cm + Tabafter: 0.63 cm + Indent at: 0.63cm


Field Code Changed

Field Code Changed




17

measurement∗. Each figure labeled by “AGG” refers to a region that exhibits grain

coarsening after QPF. All such locations (Fig. 13Fig. 13a-d) had gd < 3 μm while all

fine-grain locations after QPF (Fig. 13Fig. 13e and f) had gd > 3 μm.

Larger grains correlate to regions of higher heat input (i.e. higher peak

temperatures or longer times-at-temperature during welding). This can be seen by

comparing H1 - H3 (increasing order of heat index) for either upper or middle locations,

Fig. 13Fig. 13. The middle regions (lower strains, lower strain rates[103][94]) have larger

grain sizes than the corresponding upper locations (higher strains, higher strain

rates[103][94]). This is contrary to the previous observations, i.e. for AA2519[110][104] and

AA7010[33][42][33], where grain sizes decreased from top to bottom of the FSW region.

The through-thickness temperature gradients in the current weld are expected much

smaller than in those previous cases because the sheet is thinner (1.65 mm here vs. 25.4

mm[110][104] and 6.4 mm[33][42][33]) and higher heat inputs were used (56-250 rpm2 mm/s

here vs. 5-35 rpm2 mm/s[33, 110][42, 104]) This expectation is consistent with FSW thermo-

mechanical simulation results for the conditions used here showing no difference of peak

temperature between the top and middle regions during FSW.[103][94] Therefore, while

large strain rate gradients are still expected from top to bottom (because of the dominant

role of shoulder region of the tool), the thermal gradients are nearly absent. Thus, the

upper region is expected to experience higher flow stress for dynamic recrystallization

during FSW and thus smaller grains.[111][105]

The composition of the second phase particles is shown by the color codes on Fig.

12Fig. 12 and Fig. 13Fig. 13. Large particles (in purple) are (Fe,Mn)Al6, consistent with

reports in the literature.[46, 77][68, 71] Because sparse, large particles have little affect on

AGG,[64, 73, 76][58, 67, 103] these scan fields were chosen to avoid these large particles, and

thus do not show representative numbers of (Fe,Mn)Al6 particles. Analysis of second

phase particles appears in a subsequent section.

3.4. EBSD imaging: FSW + heat treatment

∗ The uncertainty is determined via three adjacent scans on the H1 middle region and the H3 middle. The standard deviation is calculated for each location.

Formatted: Font: Not Bold, Italic

Formatted: Font: Not Bold, Italic


Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed

Field Code Changed



Field Code Changed

Field Code Changed



Field Code Changed



Field Code Changed





Field Code Changed

Field Code Changed

18

After QPF, areas of the weld that experienced grain coarsening have grains as

large as 100 μm (Fig. 3Fig. 3), while elongated grains after extended heat treatment can

grow to 500 μm,[90][83] (Fig. 4Fig. 4, View (8)). The disparity of grain size across a

boundary separating AGG and non-AGG regions* is striking, Fig. 14Fig. 14.

A region 4-5 μm thick at the very top of the weld nugget zone has a high density

of second phase particles and irregular, ultra-fine grains, gd ~ 0.3 μm, Fig. 15Fig. 15.

The difference of grain size across the boundary of this region is approximately 300, or

2.5 orders of magnitude (0.3 μm vs. 100 μm). The high fraction of second phase particles,

such as entrained weld debris and surface oxides, presumably inhibits grain growth. The

existence of this region rules out a surface-nucleation of the coarse grains (but not

necessarily an interface nucleation). The equiaxed nature of the coarse grains near the

top of the welds also suggests a more homogeneous activation (or nucleation). The ultra-

fine grain layer was observed in as-welded samples as well as after QPF.

Details of the microstructure near the top surface are revealed by Fig. 16Fig. 16

for another condition: FSW-H1+QPF (having coarse grains in the top, middle, and

bottom regions). Starting from the top, there are four distinct layers delineated by grain

size and appearance: Layer 1, thickness ~ 5 μm, gd ~ 0.3 μm (Fig. 15Fig. 15); Layer 2,

thickness ~ 25 μm, gd ~ 20 μm (Fig. 16Fig. 16); Layer 3, thickness ~ 70 μm, gd ~ 60 μm

(Fig. 16Fig. 16); and Layer 4 (remainder), thickness ~ 1.5 mm, gd ~ 500 μm (Fig. 16Fig.

16). Such a layering macrostructure was confirmed by optical microscopy. The number

and thickness of the layers are larger for welds with lower heat input (H1>H2>H3) and

vary depending on lateral position in a single weld. (See also Fig. 1Fig. 1a and

accompanying text later in this paper for SEM imaging of these layers.

In order to reveal the nature of the ultra-fine-grained regions near the top surface,

SEM imaging and analysis were performed on a FSW-H3-QPF specimen, Fig. 17Fig. 17.

The gradient of grain size across the 20 μm-thick layer can be seen in Fig. 17Fig. 17a, (<

1 μm near the top to 3 μm at the interface with the AGG region, where the grain size is * The possibility that significant normal grain growth occurs during QPF was eliminated by measuring before-and-after grain sizes for H2-M. These were found to be 3.5 and 3.7 μm, respectively, identical within the 4% scatter of the measurement.


Field Code Changed














19

hundreds of microns). At the interface, a band with a high density of second-phase

particles can be seen, with darker appearance (lower atomic number) than the

predominantly Al matrix.

The three numbered particles labeled in Fig. 17Fig. 17a were analyzed using EDS,

Fig. 17Fig. 17b and c. The results for Particle 1 are consistent with MgSi2, Particles 2

and 3 are identified as Al2O3 and Mg5Al8. Similar compositions were determined for

particles found elsewhere, but nowhere else does the volume fraction of such particles

approach that in the band.

3.5. EBSD imaging: grain boundary misorientations

The kinetics of recrystallization and grain growth has been associated with the

character of the grain boundaries involved.[63, 112][57, 106] Deformation leads to the

formation of cells and subgrains and therefore produces higher fraction of LAB and

increase %LAB.[113, 114][107, 108] Grain-to-adjacent-grain misorientation angle (θ)

distributions over a range of 5° to 65° are shown in Fig. 18Fig. 18 as compared with a

random texture.[115][109] The top region (red/solid lines) has more low angle boundaries

(LAB) (definition: θ = 5° to 15°) than the middle region (blue/dashed lines), Fig. 18Fig.

18a. Both regions have more LAB than a random distribution,[115][109] as would be

expected in view of the higher cold work.[113, 114][107, 108] AA5083-O misorientations

correspond to a random texture, while AA5083-H18 has predominantly low-angle

boundaries (again, as correlated to cold work), Fig. 18Fig. 18b. Fig. 18Fig. 18c

summarizes selected information in Fig. 18Fig. 18a and b, by plotting the number

fraction less that for a random distribution. Annealed AA5083-O is nearly random, while

the middle FSW regions, top FSW regions, and heavily rolled AA5083-H18 exhibit

progressively higher proportions of lower-angle boundaries (5° - 20°) and lower

proportions of higher-angle boundaries (30° - 50°). The order is consistent with the

corresponding hardnesses, Fig. 8Fig. 8.

The as-welded grain sizes and fractions of LAB for all six characterization

locations are plotted in Fig. 19Fig. 19. LAB here is defined to be boundaries with

misorientation from 3° to 15°. This plot shows a location of smaller as-welded grain size



Field Code Changed



Field Code Changed




Field Code Changed




Field Code Changed



Field Code Changed








20

generally has a higher fraction of LAB. Regions (in red) of smaller grain size and higher

LAB% are more susceptible to coarse grain structure formation.

3.6. Particle and pinning parameter quantification

Table 3 summarizes the microstructural measurements and corresponding particle

pinning parameters. The particles were characterized using TEM for diameters (dp) less

than 200 nm and SEM for dp > 200 nm, the latter converted to an adjusted spherical

diameter by dividing the raw measurement by 0.816.[108][99] The presented standard

deviations were obtained by characterizing at least three adjacent fields of view for each

case.

Note that the particle area fraction ( pf ) and mean particle diameters are identical

within scatter for each location and whether as-welded or after QPF (SEM only). In view

of this result and in order to improve the uncertainty of the combined particle pinning

parameters, all particles analyzed (8330 for pd > 200 nm; 246 for pd < 200 nm) were

lumped into a single population. The effective particle dispersion Eff

p

p

df⎟⎟⎠

⎞⎜⎜⎝

⎛was obtained

using the following equation*:

np

np

p

p

p

p

Eff

p

p

df

df

df

df

,

,

2,

2,

1,

1, ...++=⎟⎟⎠

⎞⎜⎜⎝

⎛ (8)

where n is 8576, the total number of particles measured, and the volume fraction for an

SEM-measured particle is the area fraction determined from adjusted particle and the

following equation,

( )Tot

TEM

ipip V

df

6

3,

,

π= (9)

is used for TEM-measured particles, where TotTEMV is the total TEM –interrogated volume

containing the particles.

* To check whether Eq. 8 can be reasonably applied particle-by-particle, an alternate analysis was carried out by assuming a normal distribution of particle sizes with average and standard deviation corresponding to the measured population. The dispersion was identical, 0.11 μm-1. A simple number average of particle sizes, as sometimes seen in the literature, yielded a value of 0.15 μm-1.

Field Code Changed



21

The particle dispersion obtained by this approach is 0.11 μm-1 ± 0.01 μm-1, where

the scatter represents the range of individually computed particle populations for H2-T,

H2-M, H3-T, and H3-M. This values was also used to determine the other particle

pinning parameters presented in Table 3, under the assumption that the particle

dispersions are the same for all cases. Further analysis showed that particles larger than

0.6 μm had no significant contribution to Eff

p

p

df⎟⎟⎠

⎞⎜⎜⎝

⎛, or thus to Ψ.

Table 3. Volume fractions, mean spherical diameters, and computed pinning parameters for various locations

H1-T H1-M H2-T H2-M H3-T H3-M H2-T +QPF Average

gd (μm) 1.9±0.1 2.7±0.1 2.0±0.1 3.5±0.1 2.3±0.1 4.0±0.2

maxX 2.2 2.4 2.4 2.1 2.4 2.2

SEM fp

0.015 ±0.004

0.017 ±0.006

0.017 ±0.002

0.020 ±0.008

0.020 ±0.006

0.019 ±0.004

0.021 ±0.009

0.018 ±0.003

pd (μm)

0.42 ±0.03

0.38 ±0.03

0.45 ±0.04

0.40 ±0.02

0.43 ±0.05

0.39 ±0.03

0.43 ±0.03

0.41 ±0.03

TEM fp

0.012 ±0.004

0.012 ±0.007

0.010 ±0.007

0.011 ±0.003

0.011 ±0.01

pd (μm)

0.10 ±0.01

0.09 ±0.02

0.10 ±0.01

0.09 ±0.03

0.10 ±0.01

Ψ 0.31* ±0.03 0.45* ±0.04

0.33 ±0.03

0.58 ±0.05

0.38 ±0.04

0.67 ±0.06

Key: gd - average grain size; maxX - maximum size ratio; fp - particle volume fraction;

pd - average particle diameter; Ψ - particle pinning parameter.

* Based on the assumption that Eff

p

p

df⎟⎟⎠

⎞⎜⎜⎝

⎛is equivalent at all locations.

The maximum grain size ratio in Table 3, maxX , is determined by dividing the

diameter of the largest observed grain (2D, unadjusted) by the adjusted average grain

22

diameter. That is, the largest measured diameter is assumed to have come from the

equatorial cut of the grain (consistent with it being the largest cross-section observed).

The grain size ratios are similar to those computed for ideal random microstructures,

2.5[71][65]. The NGG limiting grain size[71][65] for the computed particle dispersion 1.5 μm,

smaller than all the as-welded cases, Fig. 13Fig. 13. Therefore, normal grain growth is

not expected unless the temperature is high enough to alter the particle population.

Field Code Changed

Field Code Changed


23

4. DISCUSSION AND ANALYSIS

4.1. Establishment of grain structure during FSW

The grain size and grain misorientation distributions established during welding

are consistent with dynamic recrystallization and recovery, with possible minor grain

growth. The most severe deformation (ε ~ 300) and the highest strain rate (2000 s-1 –

3000 s-1) occur near the top surface while stirring, while the temperatures (typically up to

0.8 of the melting temperature[26-28][26-28]) fall in a narrow range throughout the thickness

for 1.6 mm AA5083 sheets.[103][94]

Aluminum alloys are known to dynamically recover during hot working at

temperatures below 450 °C[116, 117][110, 111] at strains up to 5[116, 117][110, 111]. In the presence

of large second-phase particles (> 0.6 μm), dynamic recrystallization by nucleation and

grain growth may also occur.[117, 118][111, 112] Nucleation of nano-scale grains in regions of

high strain during FSW was observed,[28, 42, 91, 92][28, 38, 113, 114] and such nuclei have been

observed to grow to sizes of several microns after heating 1-4 min at 350-450 °C,[92][114]

the grain sizes being similar to those in the current study. More nuclei are formed in

regions of high strain (and thus high flow stress), thus producing smaller final

grains[111][105] toward the top of the weld zone.

While peak temperatures and times-at-temperature are not greatly different

through the thickness of the thin sheets used in this study, the role of thermal history is

revealed by comparing similar locations within welds made using alternate welding

parameters. For example, the time above 300 °C or 400 °C during welding is 4-6 times

longer for H2 than for H1 (Table 1Table 1). Grains in H2 weld zones are larger than

those in corresponding H1 locations, but there is little difference in grain size between H2

and H3, which have similar heat inputs.

Larger fractions of low angle boundaries (%LAB) are found in regions of higher

strain and lower heat input, for example top regions as compared with middle regions,

Fig. 19Fig. 19. For either top or middle regions, H2 and H3 have smaller %LAB than H1,

which has the lowest heat input.

4.2. Effect of grain size and recovery on hardness

Field Code Changed

Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed



Field Code Changed





24

The relative contribution of recovery and grain growth on the hardness of grain

coarsening regions (where the grain size changes from several microns to hundreds

microns during QPF) in the as-welded FSW weld zone can be estimated as follows. The

hardness of upper region of H2 (as-welded gd = 2 μm) is reduced from 75 Hv to 71 Hv

( gd = 4 μm) by the recovery anneal of 380 °C for 24 hours, and can be further decreased

to 68 Hv ( gd = 80 μm) by QPF. Starting from the recovered and grain-coarsened

hardnesses and the grain sizes (i.e. gd = 4 μm, 71 Hv; and gd = 80 μm, 68 Hv,

respectively), the Hall-Petch slope (kH) can be computed as 7.7 Hv μm1/2. This is half of

a reported value of 14 Hv μm1/2,[119][115] which seems to have been determined from FSW

regions without a post-FSW recovery treatment and thus includes the effect of both grain

size differences and dislocation density differences. Following a similar procedure for

the current data yielded an apparent kH of 12 Hv μm1/2, in good agreement. Using the 7.7

Hv μm1/2 Hall-Petch slope, it is concluded that about two thirds (Δ = 4.5 Hv) of the

hardness reduction during QPF where coarsening occurs is attributable to the increase of

grain size (from 2 μm to 80 μm) and one third (Δ = 2.5 Hv) is attributable to the reduction

of dislocation density.

4.3. The mechanism of grain coarsening

Extreme grain coarsening during post-weld annealing has been widely observed

in various friction-stir welded aluminum alloys.[4, 32, 33, 54, 57][4, 42, 43, 49, 52] Although the

phenomenon has been attributedable to AGG in the literature, evidence has not been

provided,[32, 33, 57][42, 43, 52] and in fact a recrystallization mechanism has alternately been

claimed[33, 54][42, 49][33, 54]. Our recovery anneal tests show conclusively that dislocation

content is not the driving force for the grain size changes and thus recrystallization is not

responsible. The correspondence of AGG regions with high %LAB is consistent with the

easier occurrence of discontinuous grain coarsening in a less homogeneous distribution of

boundary mobilities.[59][53]

Between the two grain growth mechanisms, NGG and AGG, the distinguishing

characteristic is whether grain structure coarsens uniformly or not. As shown in Fig. 9Fig.

Field Code Changed



Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed


25

9, the “nucleation” of a large grain among small grains occurs, a key characteristic of the

discontinuous AGG process. It is therefore concluded that the mechanism for coarse

grain formation in FSW 5083 is AGG.

4.3.1. Factors leading to the occurrence of AGG

For the cases studied here, the principal factor determining whether a region of

FSW undergoes AGG during the QPF heat treatment is the prior grain size. There is a

one-to-one correspondence of AGG regions with as-welded grain sizes smaller than 3 μm

and no observation of AGG for larger grain sizes, Fig. 5Fig. 5. This is consistent with the

role of particle pinning in AGG,[64, 65, 71][58, 59, 65] where a fixed number and volume

fraction of particles are less effective pins when distributed over a larger grain boundary

area because of a smaller mean grain size.

It should also be noted that there is a correspondence between as-welded grain

size and %LAB, Fig. 19Fig. 19, and thus a correspondence between as-welded %LAB

and the occurrence of AGG. The linkage that emerges from the evidence presented in

this paper is that higher strain or lower-heat input FSW regions favor both smaller grains

(more nuclei for dynamic recrystallization) and higher %LAB.

Other factors generally cited as contributing to the occurrence of AGG via grain

boundary mobility distributions are solute segregation[65][59] or texture differences[64][58].

We find no compelling evidence for the roles of these factors in the cases studied here.

The only solute of importance in AA5083, Mg, was found to have constant concentration

throughout the thickness of the FSW zone, Fig. 21Fig. 21, and unchanged from the

unwelded base material. No correlation between texture and AGG was revealed other

than that two dominant components were found for top regions, Fig. 22Fig. 22. Texture

with more than one dominant component has been reported unlikely to promote

AGG.[64][58]

4.3.2. Application of grain growth model

Fig. 23Fig. 23 shows the application of the data in Table 3Table 3 to Humphreys’

model.[71][65] The horizontal bars represent the uncertainty of pinning parameters. (The

cases here meet one of the key assumptions of the model: that the grain size dispersions

represent that of a random grain size dispersion of 2.5, Table 3.)


Field Code Changed


Field Code Changed

Field Code Changed



Field Code Changed



Field Code Changed

26

The red labels on Figure 23 connote conditions for which AGG was observed,

while blue-labeled conditions exhibited no grain growth. There is a complete agreement

of the predictions and the results, although one condition, H2-M, lies at the critical

condition separating AGG from no grain growth. Fig. 4Fig. 4 confirms that H2-M is very

close to the critical condition: AGG does not occur for QPF, but does occur if the 5

minute heat treatment interval is extended to 1 hour. The 465 °C / 1 hour condition is

also shown on Fig. 23Fig. 23 taking into account the measured size ratio of 50

representing the large AGG grains from the upper region adjacent to the middle regions.

The theory predicts that these large grains can grow into elongated grain shapes,

consistent with their appearance at later times, Fig. 4Fig. 4.

The base material is also shown on Fig. 23Fig. 23, with the pinning parameter

computed using the same particle population as for other cases and the measure average

grain size of 6.6 μm. No AGG is observed for the base material in the O condition, as

predicted.

It should be noted that at higher temperatures, AGG can occur when not predicted

as shown on Fig. 23Fig. 23. Two examples were observed in the current work: AGG can

occur when H2-M or base O material is heated to temperatures of 510 °C and greater.

This transition is attributable to the known coarsening of fine dispersoid (Fe,Mn)Al6

particles at temperatures above 500 °C.[46, 77][68, 71]

4.4. Possible methods to avoid AGG

Fig. 23Fig. 23 provides a conceptual guide for avoiding AGG in FSW joints of AA 5083

subjected to QPF heat treatment, and thus the deleterious effects of very large grains.

Two approaches are suggested, the first being to increase the as-welded grain size to

move the pinning parameter to higher values (more particles per grain boundary area).

For the alloy and cases shown in this study, an average grain size greater than 4 μm

should prevent AGG throughout the weld zone. The grain size after FSW depends on the

relative rates of dynamic recrystallization (favored by large strains) and grain growth

(favored by high temperatures for longer times). The changes to the process to obtain > 4

μm grains (from 2-3 μm) should not be overwhelming; the result is likely attainable by

several means or a combination therein, e.g.: 1) adjusting welding parameters in the






Field Code Changed


27

normal FSW process to obtain a higher heat input and thus increasing grain growth

relative to dynamic recrystallization rates (consistent with literature[33][42]), 2) adding an

auxiliary heat source during the welding for the same effect, or 3) by changing the shape

of the tool, penetration, angle, or surface condition to reduce the peak strains in the weld

zone. The second method involves changing the alloy content to obtain higher pinning

parameters by changing the particle size and volume fraction (similar to previous

suggestions in literature[5, 33][5, 42]). For the minimum grain size of the current welds (2

μm), particle dispersion greater than 0.2 μm-1 (double the current level) must be attained

to prevent AGG throughout the weld zones. While this could be attained by doubling the

Mn content to double the number of small dispersoid (Mn)Al6 particles (say by raising

the Mn content by 0.8 wt. %,[88][81] this may be disadvantageous in terms of promoting

more large eutectic particles of the same type. Addtions of zirconium may be a better

choice because it forms finer coherent ZrAl3 particles in the size range 0.025 to 0.1 μm,[88,

120][81, 116] with twice the pinning power because of their coherency[73, 88][67, 81]. For an

assumed average ZrAl3 particle size of 0.05 μm, the required Zr addition to prevent AGG

represents a volume fraction of 0.0027 corresponding to a weight percentage of 0.2%.

Field Code Changed

Field Code Changed

Field Code Changed

Field Code Changed



Field Code Changed

28

5. CONCLUSIONS The following conclusions apply to friction-stir welded AA5083 (originally in the H18

condition), post-weld heat-treated at 465 °C for 5 minutes (except as noted):

1. The mechanism for forming coarse grains (>100 μm) is abnormal grain growth

(AGG). This conclusion differs from some reports in the literature concluding

that recrystallization is responsible.[33, 54][42, 49][33, 54]

2. The presence and critical nature of AGG for the post-FSW microstructures is

consistent with a theory of grain growth appearing in the literature[71][65]. The

FSW microstructures are near the critical condition for AGG to occur.

3. The principal determinant of whether AGG occurs or not is the as-welded grain

size. AGG only occurs when the average grain diameter is less than 3 μm,

corresponding to a critical pinning parameter of less than 0.6. For this reason,

AGG is favored by higher strains and lower temperatures (i.e. lower heat input

indices) during welding. Weld conditions with high heat input show coarse grains

in only the upper region of the weld zone; low heat input welds show large grains

throughout the weld zone.

4. The percentage of low-angle boundaries (LAB) is correlated inversely with grain

size, with grains smaller than 3 μm having more than 18% LAB.

5. Within the measurement accuracy, there are no differences in second-phase

particle sizes, volume fractions, or solute content in AGG vs. non-AGG regions.

Textures vary, but likely do not favor or inhibit AGG.

6. AGG cannot be avoided by a prior recovery anneal.

7. Two strategies were recommended to avoid AGG in 5083 FSW by increasing the

particle pinning parameter above 0.6: a) increase welding heat input by

controlling welding parameters or adding additional heat sources to obtain larger

as-welded grains, and b) .increase particle content by adding ~ 0.8 w.t.% Mn or

~0.2 w.t.% Zr.

8. Particles of diameter greater than 0.6 μm have no significant effect on grain

boundary pinning in this alloy.

Formatted: Numbered + Level: 1 +Numbering Style: 1, 2, 3, … +Start at: 1 + Alignment: Left +Aligned at: 0.63 cm + Tab after: 1.27 cm + Indent at: 1.27 cm

Field Code Changed

Field Code Changed

29

9. A Hall-Petch slope reported in the literature[119][115] for 5083 FSW (14 μm1/2)

includes equal contributions of dislocation density and grain size. Experiments in

the current work with and without recovery yield Hall-Petch slopes of 12 and 7.7

μm1/2, respectively.

10. Hardness changes of 7Hv for FSW regions undergoing AGG to a final grain size

of 80 μm from a starting grain size of 4 are attributable 2/3 to the grain size

change and 1/3 to removal of dislocation content.

11. One or more distinct layers of ultra-fine equi-axed grains (including regions of

average grain sizes of 0.3 μm) were observed near the top weld surface. The

number of layers, their thickness(es), and the grain sizes depend on the weld

conditions, heat treatment, and lateral location, with the finest grains closest to the

surface. The layers contain much denser arrays of pinning particles compared

with elsewhere, three kinds of which were identified: MgSi2, Al2O3 and Mg5Al8.

Field Code Changed



30

ACKNOWLEDGEMENTS

This work was principally supported by the General Motors R&D Center to the Ohio

State University with matching support by the OSU Transportation Endowment Program

(Gan, Wagoner). Other participants' were supported by parallel GM grants and

additional funds were provided by the Korea Science and Engineering Foundation

(KOSEF) through the SRC/ERC Program of MOST/KOSEF R11-2005-065 (Chung) and

Hitachi of America (Okamoto). Fig. 3Fig. 3 and Fig. 4Fig. 4 areis reproduced from

reference[90][83] with permission of the Journal of Engineering Materials and Technology -

Transactions of the ASME. The authors are grateful for helpful discussions with James G.

Schroth (GM), Harsha Badarinarayan, Qi Yang, Frank Hunt, Isamu Takahashi, and

Seung-Hwan Park (Hitachi).



Field Code Changed



31

REFERENCES

1. C.J. Dawes: Weld. Met. Fabrication, 1995, vol. 63, pp. 13-16.

2. Y.S. Sato and H. Kokawa: Metallurgical and Materials Transactions a-Physical

Metallurgy and Materials Science, 2001, vol. 32, pp. 3023-31.

3. W.J. Arbegast and P.J. Hartley: Proc. Fifth Int. Conf. on Trends in Welding

Research, Pine Mountain, GA, USA, 1998, pp. 541-46.

4. R.S. Mishra and Z.Y. Ma: Mater. Sci. Eng., R, 2005, vol. 50, pp. 1-78.

5. I. Charit, R.S. Mishra, and M.W. Mahoney: Scr. Mater., 2002, vol. 47, pp. 631-36.

6. F.I. Saunders and R.H. Wagoner: Metall. Mater. Trans. A, 1996, vol. 27, pp.

2605-16.

7. M.C. Stasik and R.H. Wagoner: Int. J. Form. Processes, 1998, vol. 1, pp. 9-33.

8. Z.C. Yao, L. Zhang, S. Kariat, and Y.J. Zhou: SAE 2007-01-0421

9. Y.S. Sato, T.W. Nelson, C.J. Sterling, R.J. Steel, and C.O. Pettersson: Mater. Sci.

Eng., A, 2005, vol. 397, pp. 376-84.

10. S.H.C. Park, Y.S. Sato, H. Kokawa, K. Okamoto, S. Hirano, and M. Inagaki: Sci.

Technol. Weld. Joi., 2005, vol. 10, pp. 550-56.

11. Y.S. Sato, T.W. Nelson, and C.J. Sterling: Acta Mater., 2005, vol. 53, pp. 637-45.

12. S.H.C. Park, Y.S. Sato, and H. Kokawa: J. Mater. Sci., 2003, vol. 38, pp. 4379-83.

13. T. Kawasaki, T. Makina, S. Todori, H. Takai, M. Ezumi, and Y. Ina: Proc. 2nd

Int. Symp. on FSW, Gotherburg, Sweden, 2000.

14. M.R. Johnsen: Weld. J., 1999, vol. 78, pp. 35-39.

Formatted: Font: 小四, Do notcheck spelling or grammar

Formatted: Font: 小四, Italic, Donot check spelling or grammar




























32

15. E.D. Nicholas and W.M. Thomas: Int. J. Mater. Prod. Tec., 1998, vol. 13, pp. 45-

55.

16. S.W. Kallee and A. Mistry: Proc. 1st Int. Symp. on FSW, Thousand Oaks,

California, 1999.

17. O.T. Midling, J.S. Kvale, and O. Dahl: Proc. 1st Int. Symp. on FSW, Thousand

Oaks, California, USA, 1999, pp. 14-16.

18. S. Benavides, Y. Li, L.E. Murr, D. Brown, and J.C. McClure: Scripta Materialia,

1999, vol. 41, pp. 809-15.

19. O.V. Flores, C. Kennedy, L.E. Murr, D. Brown, S. Pappu, B.M. Nowak, and J.C.

McClure: Scr. Mater., 1998, vol. 38, pp. 703-08.

20. Y. Li, L.E. Murr, and J.C. McClure: Scr. Mater., 1999, vol. 40, pp. 1041-46.

21. G. Liu, L.E. Murr, C.S. Niou, J.C. McClure, and F.R. Vega: Scr. Mater., 1997,

vol. 37, pp. 355-61.

22. L.E. Murr, G. Liu, and J.C. McClure: J. Mater. Sci., 1998, vol. 33, pp. 1243-51.

23. Y.S. Sato, H. Kokawa, M. Enomoto, and S. Jogan: Metall. Mater. Trans. A, 1999,

vol. 30, pp. 2429-37.

24. Y.S. Sato, H. Kokawa, M. Enomoto, S. Jogan, and T. Hashimoto: Metall. Mater.

Trans. A, 1999, vol. 30, pp. 3125-30.

25. C.G. Rhodes, M.W. Mahoney, W.H. Bingel, R.A. Spurling, and C.C. Bampton:

Scr. Mater., 1997, vol. 36, pp. 69-75.

26. W.J. Arbegast: Hot Deformation of Aluminum Alloys III, TMS, San Diego,

California, USA, 2003, pp. 313-27.





















Documents

The Mechanism of Grain Coarsening in Friction-Stir Welded …li.mit.edu/Stuff/RHW/Upload/Chen_MetallTrans_FSW_Ri... · 2010. 2. 11. · 1 1. INTRODUCTION Friction stir welding (FSW)