FRICTION-STIR-WELDED AND SPIN-FORMED END DOMES FOR CRYOGENIC TANKS

S.J. Hales, W.A. Tayon and M.S. Domack

NASA Langley Research Center
Hampton, Virginia

ABSTRACT

Manufacturing of single-piece end domes for cryogenic tanks employing spin forming of tailored,
friction-stir-welded blanks of Al-Li alloy 2195 plate offers cost and reliability benefits. The
introduction of plastic deformation into a friction stir weld is a unique feature of the proposed
manufacturing route. This investigation addressed abnormal grain growth [AGG] within the
friction stir weldments during post-fabrication processing of a prototype dome. The phenomenon
of AGG was observed during the solution heat treatment [SHT] phase of T8 tempering and is a
major concern for meeting specifications. Such abrupt microstructural transitions can be
detrimental to notch-sensitive mechanical properties, such as ductility and/or fracture toughness.
If the issue of AGG cannot be resolved, then the acceptance of this approach as a viable
manufacturing route may be in jeopardy. The innovative approach adopted in this investigation
was the insertion of a stand-alone, Intermediate Annealing Treatment [IAT] between the spin
forming and T8 processing operations. A simple, recovery annealing step was deemed to be the
most readily-scalable solution when fabricating thin-walled, ellipsoidal domes. The research
effort culminated in the development of an effective IAT, which resulted in a significant decrease
in AGG following SHT. The processing philosophy adopted in designing the IAT is outlined and
the microstructural reasons for success are discussed. The analytical results presented are
consistent with promoting continuous grain growth during the prior IAT, thereby suppressing AGG
during the subsequent SHT.

INTRODUCTION

BACKGROUND
Manufacture of the forward and aft domes for the liquid hydrogen tank on the Space Shuttle
employed traditional gore panel construction, as illustrated in Figure 1(a). Each section was
stretch-formed to complex curvature, chem-milled, trimmed to shape, heat treated and then the
panels were fusion welded together (1). Recently, a 5 meter diameter prototype dome was
fabricated by combining friction stir welding [FSW] and single-piece, spin forming deformation
[SFD] technology at MT Aerospace in Augsburg, Germany, Figure 1(b) (2). Size limitations on
commercially available starting stock dictated that the SFD blank was two Al-Li alloy 2195 plates
welded together. FSW of 2195 plate in various gages is flight-ready technology, being used
during construction of the cylindrical sections on the Super Lightweight Tank [SLWT] (1).
Considerable weight and cost savings have been realized by adopting the FSW/SFD combined
process for the barrel sections. Further cost benefits are anticipated by using SFD technology to
fabricate the end domes on next-generation cryogenic tanks.

The current thermo-mechanical processing profile for the proposed FSW/SFD manufacturing
route is outlined in Figure 2. Alloy 2195 has the nominal composition of Al-4.0Cu-1.0Li-0.4Mg-
0.4Ag-0.1Zr [wt. %] and the starting material is 19 mm thick plate in the fully soft, O temper
condition. Fabrication commences with FSW of two plates parallel to the prior rolling direction,
followed by a stress relief annealing treatment. Subsequent machining operations produce a 5 m
diameter, tailored blank with a varying radial thickness of 5 to 8 mm. Progressive SFD at warm
temperatures gradually creates an ellipsoidal dome of uniform thickness [~ 2.5 mm]. The final
stage of fabrication involves processing of the spin-formed article to the high-strength, T8 temper
condition for service properties. The occurrence of AGG in weld nuggets has been a concern
since FSW processing of alloy 2195 was first explored for reducing the cost of the SLWT (3).
 (a) Current: Multi-piece assembly (b) Prototype: Single-piece construction

Potential Benefits:
Friction Stir Welding; Spin Forming;
• No molten material • Reduced part count
• Less welding defects • Reduced scrap rate
• Better mech. props. • Lower Fab. costs
Figure 1. Fabrication of the end domes for cryogenic propellant tanks: (a), multi-piece
assembly used on the Space Shuttle (1); (b), single-piece construction approach proposed
for future launch vehicles, combining Al-Li alloys with friction stir welding and spin
forming technologies.

Friction Stir Solution

Welding JOINING Heat
Stress Treatment
// Relief FABRICATION
T8
Anneal
TEMPER
Temperature

Spin-Forming Aging
// //

Contour Cold
Machining Stretching

Time (Deformation Strain)

Figure 2. Thermo-mechanical processing profile for the manufacturing of Al-Li 2195
prototype end domes (2): FSW of 2 plates [plus machining] produces a large, circular
blank; progressive spin forming operations create the ellipsoidal shape; T8 processing
generates service properties.

A preliminary evaluation on 2195 material produced by FSW/SFD was conducted in earlier work
(4). The effect of standard T8 tempering on the microstructure of a deformed weld nugget is
illustrated in Figure 3. The orientation of the full-penetration, butt weld with respect to a
fabricated dome is indicated beside the micrographs. It is worth outlining the anatomy of a typical
friction stir weld at this juncture. The weldment consists of the weld nugget, bordered on both
sides by thermo-mechanically affected zones [TMAZ's]. These represent microstructural
transitions from the characteristics of the weld nugget back to those of the parent metal. The
advancing side [at left] exhibits an abrupt change in microstructure at the nugget/TMAZ interface.
In contrast, the retreating side [at right] shows a more gradual shift within the TMAZ region. The
weld crown side [at top], and the weld root side [at bottom], correspond to the outer and inner
mold lines of a fully-fabricated article, respectively. Note that the actual crown and root locations
of the weld nugget were removed during the contour machining stage, prior to spin-forming
operations.

(a) As Spin-Formed condition (b) Solution Heat Treated condition

Weld Nugget Weld Crown Side Outer Mold Line

Add tool width


1.0 mm 1.0 mm
Advancing Side Retreating Side Weld Root Side Inner Mold Line

(c) Partial weld nugget cross-section

Welding Direction WELD CROWN Parent
S Metal

RETREATING SIDE
parallel to plate
ADVANCING SIDE

Axes
Rolling Direction
L T
Original thickness
of material during Material thickness
Friction Stir Welding WELD following contour
NUGGET machining and
Spin Forming

WELD ROOT
Figure 3. AGG in a deformed weld nugget during standard T8 processing of Al-Li 2195:
(a), as spin-formed condition; (b), solution heat treated condition; (c), schematic of friction
stir weld cross-section remaining after contour machining and spin forming operations.

The undesirable AGG during the SHT phase of conventional processing has transformed a
heavily deformed microstructure with indistinct features [Fig. 3(a)] into a very coarse-grained
structure [Fig. 3(b)]. It is evident that some grain boundaries extend through the entire section
thickness [primarily on the advancing side]. In both micrographs, the tortuous line weaving
through the center of the weld nugget is a remnant of the faying surface from the original butt joint
between the two plates. The accompanying schematic [Fig. 3(c)] shows the size and
approximate location of the remaining weld nugget with respect to the dimensions and orientation
of the starting plate. The thickness of the spin-formed material being evaluated comprises only
25% of the original plate gage, due to prior contour machining and SFD operations. Note that a
fully-formed dome would have a uniform thickness of ~ 2.5 mm from the 'pole' to the' rim'
sections. The material evaluated in this study was from a partially formed dome where SFD
operations were interrupted at a gage of ~ 5.0 mm in the intermediate 'membrane' section.

AGG MITIGATION
Although a concise description of the AGG phenomenon does not exist, it is widely recognized
that it falls under the realm of secondary recrystallization [RX]. It may be categorized as an
extreme example of a discontinuous grain growth [DGG] process, involving the very rapid growth
of a very limited number of pre-existing grains (5). There are three principal differences between
AGG and the mechanisms commonly associated with primary RX (6). First, there is no
nucleation of new grains involved. Second, the process occurs within a fully recrystallized
microstructure. Third, stored energy is primarily decreased by a reduction in the total grain
boundary area. It is accepted that the mechanism is promoted by microstructural
heterogeneities, such as localized variations in deformation strain, crystallographic texture, grain
size, grain orientation, and boundary misorientation. Microstructural conditions which favor the
growth of isolated grains at the expense of adjacent grains can include any combination of these
factors. Consequently, there exists a multitude of possibilities for stabilizing a grain structure with
respect to suppression of AGG (7).
In this case study, the weld nugget dynamically recrystallized during the FSW process and then
was further plastically deformed. The characteristics of the grain structure following dynamic RX
alone will be more complicated than those resulting from static RX, i.e. concurrent temperature
and deformation effects (5). The intention of the current work was to use thermal processing to
replace the inhomogeneous, coarse-grained microstructure with a uniform, refined grain
structure. As shown in Figure 4, the mechanism of DGG generally involves extensive migration
of a few, high-angle grain boundaries [HAGB's]. In contrast, the mechanism of continuous grain
growth [CGG] involves uniform microstructural coarsening via development and migration of
many HAGB's (6). The CGG process also includes recovery-type mechanisms and substructure
development, i.e. the formation, coalescence and migration of low-angle grain boundaries
[LAGB's] (8). Therefore, the approach adopted was to promote CGG during a prior IAT, in order
to reduce the stored energy for AGG during subsequent SHT.

Continuous Grain Growth Factors governing the stability

of a uniform grain assembly

Continuous
Grain Growth

Grain Size Ratio (RI /RA)

100

 Energy
Ratio 10
10 (γ I /γ A) 1 5

Discontinuous
Grain Growth
1
1 10 100
Discontinuous Grain Growth Mobility Ratio (MI /MA)

Rate of differential grain growth, dR/dt = MA.(γA/ RA) - MI.(γI / RI)

If MA.(γA /RA) = MI.(γI /RI)  dR/dt = 0  CGG occurs

After Humphreys (1997); Reproduced with permission
Figure 4. Microstructural stability depends on the grain radius [Ri ], boundary energy [γi ],
and boundary mobility [Mi ] of an Isolated grain, relative to the whole Array of grains. The
primary objective was to suppress AGG by decreasing the MA.(γA/ RA) term. Illustration
adapted from the Humphreys model (9).

A unified model proposed by Humphreys is readily applicable to AGG suppression in a high

solute, multi-phase Al alloy which has undergone complex, plastic deformation. The concepts
developed blend existing theories of recovery, recrystallization and grain growth, and explain
relative contributions to microstructural development. Specifically, the effect on DGG of factors
such as strain energy, grain size, preferred texture, microstructural heterogeneities and
precipitates have been considered (9)(10)(11). The basic model, addressing the stability of
cellular microstructures, is summarized in Figure 4. Control of microstructural conditions such
that CGG is favored over DGG is the main principle. The model indicates that DGG can be
suppressed if the growth advantage of a few, unstable grains can be eliminated. The controlling
factors are presented in terms of the ratio in grain size, boundary energy and boundary mobility
between isolated grains and the surrounding grain array. The details reveal that if (M.γ/R)A =
(M.γ/R)I, then the rate of DGG is zero, i.e. CGG occurs [where: M ≡ boundary mobility; γ ≡
boundary energy; R ≡ grain radius, and A ≡ array of grains; I ≡ isolated grain]. From a practical
perspective, this goal can be accomplished most readily by increasing R A, thereby decreasing the
value of the (M.γ/R)A term. Hence, the thermal processing philosophy adopted in this study was
to reduce the grain size differential by promoting CGG, thereby eliminating the growth advantage
and suppressing DGG.

PROCESSING PHILOSOPHY
From a practical perspective, AGG tends to occur rapidly because it is a preferential growth
process only. Consequently, the phenomenon tends to be more sensitive to absolute
temperature than heating and cooling rates. It is generally accepted that there is a threshold
temperature [TAGG], below which AGG does not occur (8). TAGG has been shown to be dependent
on many factors, including alloy composition, strain rate, deformation strain and the nature of the
stored energy. In this case, TAGG will be affected by both the FSW and SFD parameters
employed, and the varying levels of forming strain. Therefore, the approach adopted in this
investigation combined microstructural stability concepts with traditional processing protocols and
the influence of systematic microstructural gradients. The novel aspect of this work is that it
addresses suppression of AGG within a deformed FSW weld nugget. This feature separates the
current study from most of the open literature and existing patents, which tend to cover heat
treatment of as-welded material only, and include analysis of whole nugget cross-sections.

There have been two important pieces of prior thermal processing research which are helpful to
the present study. The first, shown in Figure 5, was the work performed by Alcoa in an attempt
to improve the mechanical isotropy of Al-Li alloy rolled products (12). The thermo-mechanical
processing scheme developed [Fig. 5(a)] is compared with the partial Al-Cu phase diagram [Fig.
5(b)]. The binary diagram can be employed as a convenient temperature guide because the
dominant solute in alloy 2195 is 4 wt.% Cu (7). The unusual step of correlating these data
permits identification of the temperature regimes where microstructural changes are typically
implemented using traditional processing protocols. SHT [~ 510°C] is conducted in the quasi-
single phase region [non Cu-containing precipitates, such as Al3Zr, are not included].
Recrystallization heat treatments [~ 490°C] are conducted close to the solidus temperature and
hot working operations in the temperature regime immediately below [400-475°C]. Performing
conventional recovery annealing treatments within the 250-400°C temperature range is of
particular significance to this investigation.
(a) Traditional processing protocols for (b) Partial Al-Cu binary phase diagram and
isotropic Al-Li alloy rolled product; typical processing temperature ranges;
C
After Rioja (1998) 2195 (Al - 4.0Cu+ Li+ Mg+ Ag+ Zr)
Solution
Heat
Treatment 600 Al + L
Recrystallization
Breakdown Anneal Solution Treating ~ 510 C
(Hot Roll) Solution Heat
Preheat Finish Recrystallizing ~ 490 C Al(ss) Treatments
Cross Roll/ Hot Roll
Lengthen Hot Working ~ 410 C
Recovery
Anneal
400
} Conventional
Annealing
Temperature

Annealing ~ 300 C
Treatments
Artificial 200
Age Aging Treatments
Scalp
Al + Al2 Cu
Stretch After Hatch (1984)
0
0 2 4 6 8 10 12
Time Copper Content (wt.%)
Figure 5. Establishment of typical thermal treatment practices for alloy 2195 materials:
(a), thermo-mechanical processing profile for Al-Li alloy wrought product (12); (b), partial
Al-Cu phase diagram, highlighting temperature regimes for different heat treatments (7).

The second significant piece of work was performed by Alcan to eliminate AGG during SHT
following FSW 7xxx series Al alloys (13). Particle pinning concepts were adopted for
microstructural stabilization; the aim was to modify the size and distribution of Al3Zr dispersoids.
The approach involved the application of prolonged anneals [≤ 72 hrs] at temperatures very close
to the SHT temperature. The ‘annealing’ temperature was necessarily high because Al3Zr is a
high melting point phase and the very long annealing times accounted for the thermal stability of
the dispersoids. In this case study, the Alcan approach was considered applicable to post-FSW
thermal processing only, as the exposure of deformed weldments to highly elevated temperatures
would cause AGG. This implies that the IAT temperature selected needs to be higher than
conventional recovery temperatures, but lower than the TAGG. Accordingly, results from the
combined studies were employed to define the upper and lower limits for the annealing
temperature, and to bracket the scope of the annealing times evaluated.

EXPERIMENTAL

The material used in this study was extracted from the same location in the prototype dome for all
of the processing and microstructural evaluations. Thermal treatment procedures conformed to
current, commercial guidelines, e.g. temperature control margins, heating rates, soak times,
cooling rates and furnace atmospheres (14). The longitudinal direction [L] is parallel to both the
rolling direction [RD] and the welding direction [WD]. The size of the heat treat coupons was
such that the smallest dimension corresponded to the short-transverse [S], or through-thickness,
direction in the weld nugget, i.e. width [// FSW; // L] ≥ 2X thickness [ FSW; // S]. As a
consequence, any 'free-surface' effects on AGG behavior, and the resultant microstructure, would
be comparable with processing of full-scale domes (6). The initial objective of the IAT studies
was to establish TAGG. The first series of heat treatments performed were of short duration [1 hr],
because AGG occurs very rapidly. Subsequent heat treatment studies were conducted at a
temperature suitably below TAGG in order to establish the annealing time. The second series of
heat treatments performed were of practical duration for an effective IAT, considered to be
between 12 and 72 hours.

Specimens for metallurgical analyses were sectioned perpendicular to the weldment at the mid-
plane of heat treated coupons, i.e. plane as remote from the free surfaces as possible.
Qualitative optical metallography [OM] of the entire weld nugget cross-section was conducted
during the IAT design studies. Two standard specimen preparation/ imaging mode combinations
were employed; Keller's chemical etching/ bright field [K/BF] images and Barker's electro-
chemical etching/ cross-polarized [B/XP] images. Once the best annealing temperature and time
for the IAT had been identified, electron back-scattered diffraction [EBSD] techniques were
employed along a diagonal traverse of a weld nugget. Quantitative metallurgical analyses
concentrated on the differences in [sub]grain structure, microtexture and [sub]grain boundaries
between 'the start’ and 'the end’ of the most effective IAT. Note that the term [sub]grain is used
throughout to denote the crystallites within an aggregate when the distinction between 'subgrains'
and 'grains' is not drawn. First, OM images were collected contiguously across a nugget and
then EBSD analyses were conducted on the same samples. This was to ensure that data
gathered from limited area was representative of the bulk microstructure. In both cases,
quantitative metallography practices were in conformance with ASTM specifications (15)(16).
Measurements were conducted along a diagonal traverse of the weld nugget to capture the
contribution of both horizontal and vertical microstructural gradients.

RESULTS

THERMAL PROCESSING STUDIES

A selection of the extensive K/BF micrographs collected in order to arrive at an effective IAT
design is shown in Figure 6. The annealing temperature [Figs. 6(a)-(d)] and the annealing time
[Figs. 6(e)-(h)] were derived empirically in separate studies. Note that the material is in the
intermediate annealed condition only in Figs. 6(a)-(d), and has not been exposed to the SHT
temperature. Progressively increasing the temperature for the 1 hour soak from T1 to T4 by equal
increments revealed that TAGG= T4. The highest annealing temperature for the shortest duration
was the original goal. A suitable buffer to allow for furnace variations and scale-up to large parts
was also desirable. Therefore, a recovery annealing temperature of T3 was selected for the IAT.
This may be considered a ‘non-conventional’ annealing temperature in the sense that it is in the
regime that is normally reserved for RX heat treatments and/or hot working operations [Figure 5].
The second study was used to identify the appropriate annealing time, and the number of
iterations was reduced by down selecting the T3 annealing temperature only. Note that the
material is in the ‘IAT + SHT’ condition in Figs. 6(e)-(h). The series of images reveal that the
substantial areal fraction of oversized grains after exposure for Ƭ1 hours was essentially reduced
to zero following Ƭ4 hours. Progressively increasing the duration of the recovery annealing
treatment gradually reduced AGG until the reaction was effectively suppressed. Consequently, a
recovery annealing time of Ƭ4 hours was selected as the most promising for the IAT.
(a) (e)
1 hr @ T1 Ƭ1 hrs @ T3
+ SHT

(b) (f)
1 hr @ T2 Ƭ2 hrs @ T3
+ SHT

(c) (g)

1 hr @ T3 Ƭ3 hrs @ T3
+ SHT

(d) (h)

1 hr @ T4 Ƭ4 hrs @ T3
+ SHT

1.0 mm

Figure 6. Definition of parameters for an effective IAT [Ƭ4 hrs at T3 ] : Figs.(a)-(d), the
annealing temperature [T1< T2< T3< T4 ; T4= TAGG], material in ‘IAT only’ condition; Figs.(e)-
(h), the annealing time [Ƭ1< Ƭ2< Ƭ3< Ƭ4], material in ‘IAT + SHT’ condition.

The microstructural changes occurring during the best IAT are represented on a broad scale by
the B/XP images shown in Figure 7. The characteristics of the nugget cross-section after 1 hour
at T3, the 'Start of the IAT' [Fig. 7(a)], and following Ƭ4 hours at T3, the 'End of the IAT' [Fig. 7(b)],
can be compared. The images exhibit very similar characteristics with few visible differences.
The IAT has resulted in a general sharpening of the contrast and the emergence of a scattering of
small grains along the weld root side. In both micrographs, a common characteristic of FSW weld
nugget microstructures, i.e. the onion rings, is revealed by the different preparation technique
(17). The rings are so well-defined because B/XP imaging relies on crystallographic orientation
contrast. The banding of grains with differing orientation has been associated with the varying
intensity and directionality of shear deformation created during the FSW process (18). As
documented in the earlier work, these features are primarily a product of spatial variations in
texture and grain morphology, rather than grain size per se (4).

The series of micrographs showing a progressive reduction in AGG with annealing time, Figs.
6(e)-(h), reveal a persistent trend. There is a preponderance of oversized grains adjacent to the
intersection between the advancing side interface and the root side of the partial weld nugget. As
a result, Figure 8 serves to illustrate the incorporation of previously documented microstructural
gradients in friction stir welds into the quantitative EBSD analysis of the reduced weld cross-
section. Earlier reports on whole nuggets document systematic strain gradients from both the
advancing side [AS] to the retreating side [RS] (17), and from the root [R] to the crown [C] sides
of weldments (19). Non-symmetrical variations in grain size/morphology and crystallographic
texture have also been noted, the nature of which tend to depend on FSW parameters (20). The
schematic in Figure 8(a) combines these findings and outlines the working hypothesis for the
analytical approach adopted in this case study. Quantitative metallurgical analyses were
conducted using a diagonal line connecting the 'AS -Root' [AR] and 'RS-Crown' [RC] intersections,
i.e. along 'AR to RC' traverses of the deformed weld nugget.

(a) Start of IAT: 1 hr @ T3 CROWN SIDE

RETREATING SIDE
ADVANCING SIDE

1.0 mm

(b) End of IAT: Ƭ4 hrs @ T3

1.0 mm

ROOT SIDE
Figure 7. Microstructural characteristics revealed by combination of Barker’s etching and
cross-polarized light imaging of spin-formed, partial weld nuggets: (a) at the beginning of
the IAT; (b) at the conclusion of the IAT.

(a) (b)
T Crown Side

Premise;
FSW Most of strain energy
is stored in (sub)grain
Adv. Ret. boundaries
AGG
S Side Side
1.0 mm
Root Side

(c) RC

Ɛ d Ɛ d _
d

1.0 mm
Adv. T Ret. Root S Crown
Side Side AR
Figure 8. Inclusion of gradients from FSW into EBSD analysis of a spin-formed weld
nuggets: (a), schematic of working hypothesis developed; (b), ‘SHT only’ condition –
typical location of AGG; (c) ‘IAT only’ condition – measurements along diagonal traverse.

The grain size following primary [static] RX tends to be inversely proportional to the level of stored
strain energy (6). Accordingly, it may be inferred that a similar scenario exists for the complicated
case of dynamic RX during the FSW process. The location-dependent balance between work
and heat input within the weld nugget will be manifested in variations in the final grain size of the
fully-recrystallized microstructure (20). Combining this notion with the premise that most stored
energy is contained in grain boundaries constitutes the hypothesis developed. The prevalence of
AGG at the AR location is illustrated most clearly in Figure 8(b), where large grains predominate
on the root side and extend further into the weld nugget on the advancing side. Application of
these concepts is summarized in Figure 8(c), showing material in which AGG has been
suppressed. The location of a typical 'AR-RC traverse', along which quantitative EBSD analysis
was performed, is shown.

QUANTITATIVE MICROSTRUCTURAL ANALYSES

The various types of data generated via EBSD analysis along A R-RC traverses are illustrated in
Figure 9, using the AR location as an example. Figure 9(a) shows the use of inverse pole figures
[IPF's] and IPF maps to determine the frequency and spatial distribution of microtextural
components, respectively. The data can be interpreted as [sub]grain orientation information and
can be transformed such that RD/WD [// L] becomes the reference axis, rather than the traditional
ND [// S]. Furthermore, the IPF data can be reduced to permit comparison with the texture
characteristics common in Al-Li alloy wrought product. The large quantity of data produced can
also be exploited to quantify the statistical distribution of the sizes and boundary misorientations
of [sub]grains. Figure 9(b) shows the [sub]grain area distribution information extracted. It is
evident that the distribution is approximately log-normal and is a consistent observation at all of
the locations analyzed. Similarly, Figure 9(c) shows the that the [sub]grain boundary
misorientation distribution differs from that of a random polycrystalline aggregate (21). The
distributions intersect at a boundary misorientation, θ, of ~ 30°. The frequencies for θ>30°
boundaries are lower [HAGB's], and that for θ<30° boundaries are higher. It should be noted that
this latter category contains all the LAGB's associated with subgrain structure, i.e. θ<15°. This
observation proved to be common in the microstructure at all locations evaluated, and is
indicative of the level of substructure present.
(a) Microtexture Frequency & Spatial Distribution (b) (Sub)Grain Area Distribution
[100] 0.12
Start of IAT
// to RD; 111 0.1
Frequency

<uvw>’s 0.08
112
112 0.06

0.04

0.02
001 101
0
Texture Fibers: 1 10 100 1000
<100>; RX [Sub]Grain Area (μm2)
<110>; Shear
@ AR
<111>; Extrusion
<112>; Rolling (c) Boundary Misorientation Distribution
0.08 J.K. MacKenzie,
0.07 MacKenzie Biometrika, Vol.
0.06 Start of IAT 45, p.229, 1964.
Frequency

0.05
0.04
0.03
0.02
0.01
0
Θ = 30
0 20 40 60 80

@ AR Misorientation (degrees)
25 μm
Figure 9. Types of quantitative data generated via EBSD analyses along AR-RC traverses
of deformed weld nuggets [using AR location as an example]: (a) preferred [sub]grain
orientations; (b) average [sub]grain dimensions; (c) [sub]grain boundary misorientations.
The frequency of preferred [sub]grain orientations is represented by the variation in areal fraction
of common texture fibers along the AR-RC traverse in Figure 10. The information allows the
textural composition of the material at the start [1 hour @ T 3] and end of the IAT [Ƭ4 hours @ T3]
to be compared and contrasted. Note that texture fibers are generally families of {hkl}<uvw>
crystallographic orientations with a common <uvw> or {hkl} component (22). Specific to FSW of
Al alloys torsional deformation results in the emergence of so-called A-{111}<uvw> and B-
{hkl}<110> fibers and the C-{001}<110> component. It is noteworthy that such dynamically
recrystallized microstructures have also been shown to exhibit a higher fraction of LAGB's (23).
In Al-Li wrought product, the <100>-fiber is typically associated with static recrystallization [RX],
even though the microstructures tend to be predominantly unrecrystallized. The presence of
<110>-, <111>-, and <112>-fibers is common, and formation can be related to the type of
deformation encountered (24). The <110>- and <111>-fibers are related to simple shear, and
mixed shear deformation conditions, respectively. The <110>-fiber is also observed in many
wrought materials (22), and the <111> fiber tends to dominate in Al-Li extrusions (25). The
<112> fiber is often referred to as a 'skeleton line' of ideal crystallographic orientations, but it is in
close proximity to the β-fiber frequently formed in Al alloy rolled products (26).

RC
0.30
0.25 (e)

Frequency
_ 0.20
f(g) 0.15
0.30 0.10
0.25 (d) 0.05
1.0 mm
Frequency
0.20 0.00
0.15 <001> <011> <111> <112>
AR 0.30 0.10
@ RC Texture Fiber

0.25 (c) 0.05

Start of IAT 0.00
Frequency

0.20
End of IAT <001> <011> <111> <112>
0.15
@ 3t/4 Texture Fiber
0.30 0.10
0.25 (b) 0.05
0.00 Similar Microtexture
Frequency

0.20
<001> <011> <111> <112> Composition @ AR, t/2 & RC
0.15 @ t/2 Texture Fiber
0.10
0.30
0.25 (a) 0.05
Deformation Texture Components, %
0.00
Frequency

0.20
<001> <011> <111> <112> Fiber AR t/4 t/2 3t/4 RC
0.15 @ t/4 Texture Fiber
0.10 <111> 14.8 6.1 12.9 5.1 7.2
0.05
<112> 22.5 14.7 20.0 12.5 16.5
0.00 Similar Microtexture
<001> <011> <111> <112> <111> 22.2 5.3 6.2 4.9 11.9
@ AR Texture Fiber Composition @ t/4 & 3t/4
<112> 29.2 15.5 22.0 10.5 19.2

Figure 10. Effect of IAT on microtexture distribution along AR-RC traverse of deformed
weld nugget: Figs.(a)-(e) show relative contribution of selected texture fibers. Two distinct
microtexture compositions, and IAT causes subtle changes only.

The microtexture data indicate that the texture along the AR-RC traverse may be divided into two
distinct compositions. Although varying in intensity across the weld nugget, the textural
composition at the AR, t/2 and RC locations is similar, i.e. the <112>-fiber makes the largest
contribution [Figs. 10(a), (c), &(e)]. Likewise, the texture characteristics at the t/4 and 3t/4
locations are similar, but of a different composition, i.e. the <110>-fiber contributes the most
[Figs. 10(b) &(d)]. The IAT does seem to have a subtle effect on texture composition within the
two distinct categories. However, caution is exercised in interpretation of this data because of the
fluctuations introduced by the onion ring features [Fig. 7]. The tabulated data [inset] highlights
both the cross-sectional variation in, and the effect of the IAT on, two of the more prominent fibers
noted. These may be significant because <111> and <112> provide the highest contribution at
the AR location and it appears to increase during the IAT. It may be surmised that subtle shifts in
(sub)grain orientation reduce the growth advantage of certain grains residing in the region where
AGG is the most prevalent.

The EBSD data was also employed to generate the statistical distribution of [sub]grain areas
along the AR-RC traverse presented in Figure 11. The effects of variability in [sub]grain size and
morphology noted in earlier work (4), by determining L, T, and S dimensions via mli methodology,
can be reduced in this manner. The series of plots also show that the grain area distribution is
approximately log-normal throughout the cross-section. Of significance, the distribution appears
to be narrower at the AR and t/4 locations and the IAT has resulted in further sharpening [Figs.
11(a) &(b)]. The distribution does appear to be broader [peak height marginally lower] at other
locations [Figs. 11(d) &(e)] and particularly at t/2 [Fig. 11(c)]. Again, caution should be exercised
in interpretation of this data because of fluctuations potentially introduced by the onion rings.

RC
0.12 (e)
0.10

Frequency
_ 0.08
A
0.06
0.04
0.12 (d) 0.02
1.0 mm 0.10
0.00

Frequency
0.08
1 10 100 1000
AR 0.06 @ RC [Sub]Grain Area (µm 2)
0.04
0.12 (c) 0.02
Start of IAT 0.10
0.00
Frequency

0.08
End of IAT 1 10 100 1000
0.06 @ 3t/4 [Sub]Grain Area (µm2)
0.04
0.12 (b) 0.02
0.10
0.00 [Sub]Grain area distribution is
Frequency

0.08
1 10 100 1000
0.06
@ t/2 [Sub]Grain Area (µm 2)
approximately log-normal
0.04
0.12 (a) 0.02
0.10
0.00 Average (Sub)Grain Area, µm2
Frequency

0.08 1 10 100 1000

0.06 @ t/4 [Sub]Grain Area (µm2) IAT AR t/4 t/2 3t/4 RC
0.04
0.02 Start 65 75 76 93 76
0.00 Avg. [Sub]Grain area
1 10 100 1000
@ AR [Sub]Grain Area (µm 2) increases by ~30% End 84 88 103 116 103

Figure 11. Effect of IAT on [sub]grain area distribution along A R-RC traverse of deformed
weld nugget: Figs.(a)-(e) show approximately log-normal distribution throughout and an
approximately 13% increase in average [sub]grain diameter during the IAT.

When examined as a whole, the distribution plots in Figure 11 also reveal that the IAT has
resulted in a consistent movement in the peak position. This is indicative of an increase in the
average [sub]grain area throughout the weld nugget cross-section. The tabulated data [inset]
effectively highlights both the cross-sectional variation in, and the effect of the IAT on, the
[sub]grain size. The dimensions are lowest at the AR location, gradually increase toward 3t/4 and
then decrease toward the RC location. The most important finding is that the average [sub]grain
2
area increases by approximately 30% across the entire weld nugget. Based on the π.r
relationship, this translates into a uniform increase in [sub]grain diameter of approximately 13%.
The location of the finest (sub)grains at the AR location may be significant, because it
corresponds with the region which consistently contained the most AGG.

The uniform microstructural coarsening means that the slight grain size gradient documented is
not modified by the IAT. This raises the question as to whether there is a change in the
associated (sub)grain boundaries which could make a supplemental contribution to AGG
suppression. Further analysis of the EBSD data produced the statistical distribution of [sub]grain
boundary misorientations along the AR-RC traverse shown in Figure 12. Throughout the cross-
section, the misorientation distribution differs from that of a randomly-oriented grain structure
(21). As shown earlier for the AR location [Fig. 9(c)], the distributions coincide at a boundary
misorientation, θ ~ 30°. Above this value there are less HAGB's than random, and below it there
are more boundaries, including LAGB's [θ<15°]. Dynamic RX, the presence of substructure, and
retention of deformation-type texture components, are most likely responsible for this
characteristic (20). Further, variations in the relative contributions of heat and work input during
the FSW process may account for the different LAGB/HAGB ratio along the traverse (17).

RC
0.07
0.06 (e)
0.05

Frequency
_
ϴ 0.04
0.03
0.07 0.02
0.06 (d) 0.01
1.0 mm 0.05 0

Frequency
0.04 0 20 40 60 80
AR 0.03 @ RC Misorientation (deg)
0.07 0.02
0.06 (c) 0.01
Start of IAT 0.05 0
Frequency

End of IAT 0.04 0 20 40 60 80

0.03 @ 3t/4 Misorientation (deg)
0.07 0.02
0.06 (b) 0.01
0.05 0 More subgrain
Frequency

0.04 0 20 40 60 80 structure
0.03 @ t/2 Misorientation (deg)
0.07 0.02
0.06 (a) 0.01
0.05 0
Boundary Misorientation (ϴ <15 ), %
Frequency

0.04 0 20 40 60 80
0.03 @ t/4 Misorientation (deg) IAT AR t/4 t/2 3t/4 RC
0.02
0.01 Start 13 14 23 22 23
0
0 20 40 60 80 Less subgrain
@ AR Misorientation (deg)
structure End 16 13 30 23 27

Figure 12. Effect of IAT on [sub]grain boundary misorientation distribution along AR-RC
traverse of deformed weld nugget: Figs.(a)-(e) show more LAGB’s, and consequently less
HAGB’s, at AR and t/4 locations. IAT causes subtle changes only.

Although the variations in the misorientation distribution along the traverse are not systematic, the
data indicate the presence of more HAGB's at AR and t/4 [Figs. 12(a) &(b)], than at the t/2, 3t/4
and RC locations [Figs. 12(c), (d) &(e)]. This may be interpreted as evidence that there is less
substructure within grains on the advancing side and more on the retreating side of the weld
nugget. Accordingly, the highest fraction of substructure appears to be at the t/2 location, where
the frequency of LAGB’s appears to be much larger [Fig. 12(c)]. This mid-weld region is
coincident with the axis of the FSW tool and mass flow in this vicinity is likely to be the most
complicated (18). As observed in Figure 7, the bands of grains at this location exhibit a broad
spacing and are horizontally inclined, unlike the t/4 location. Again, the onion rings may cause
some data distortion due to sampling location and interpretation should be guarded (4). The
tabulated data [inset] highlights both the cross-sectional variation in, and the effect of the IAT on,
the frequency of subgrain boundaries [θ<15°]. A lower proportion of LAGB’s at the AR and t/4
locations with little change in frequency during the IAT is readily apparent. The existence of a
higher proportion of HAGB’s at the AR to t/4 locations may be of significance. As mentioned
earlier, this is the region of the weld nugget where AGG was observed to be the most persistent.
 DISCUSSION

The insertion of an effective IAT into the dome processing schedule culminated in a significant
reduction in the area fraction of oversized grains within the deformed weld nugget. The
combination of FSW and SFD dictates that the microstructural condition in and around the nugget
is complex. The abrupt transition between the nugget and the TMAZ at the advancing side, and
the intersection of this interface with the root side of the weld result in very complicated
characteristics. Current observations of AGG in the presence of a microstructural gradient are
consistent with initiation within this region of highest stored energy, and the smallest grain size.
In the 'SHT only' material condition, the oversized grains persisted at the A R location during IAT
development, Figs. 6(e)-(h). Therefore, variations in the primary driving force [reduction of stored
energy] may be correlated with the changes in grain size across the weld nugget. The A R
location represents the region best-suited for the initiation of AGG, with propagation along the
resident gradient[s]. The use of diagonal traverses of a weld nugget during quantitative analyses
offers an innovative perspective on how the effects of overall material flow during FSW are
manifested in the microstructure of the whole cross-section (20).

A synthesis of the microstructural changes resulting from the IAT is presented in Figure 13,
which combines the tabulated data contained in Figures 10, 11, and 12 in to two plots. Figure
13(a) shows the initial cross-sectional variations in microstructure, and a comparison with Figure
13(b) reveals the effect of the IAT on these microstructural characteristics. The [sub]grain
diameter data suggests that the microstructure is most refined at the AR location. Comparing the
two figures shows the ~13% increase in grain size across the whole weld nugget during the IAT.
Although there are perturbations in grain diameter along the AR - RC traverse, a subtle upward
gradient toward the RC location is evident. This effect may be attributed to gradients in the mass
flow, and varying heat/work input, characteristic of the FSW process, e.g. higher temperature
correlates with larger grain size (20). The gradient persists following the IAT, which indicates that
the microstructural coarsening was uniform throughout the cross-section. The combined data are
consistent with the growth advantage of isolated grains being mitigated by a reduction in the grain
size differential with the surrounding array of grains (9). Whether a uniform increase in grain size
of ~13% across the weld nugget is sufficient (in itself) to eliminate AGG is uncertain.

(a) Start of IAT (b) End of IAT

35 Diam. (um) 35
<111>-fiber
30 <112>-fiber
30
LAGB's (%)
25 25
Frequency

Frequency

20 20

15 15

10 10

5 5

0 0
AR
A R t/4 t/2 3t/4 R
RCC AR
AR t/4 t/2 3t/4 RC
RC
Location in Weld Nugget Location in Weld Nugget

Figure 13. Graphical summary of tabulated data contained in Figures 10, 11, and 12, which
can be correlated with [sub]grain size, orientation, and boundary misorientation variations
across a deformed weld nugget: (a) beginning of the IAT: (b) conclusion of the IAT.

It is evident that the prime objective of promoting CGG at the expense of DGG has been
achieved, and it may be inferred that the net driving force for AGG has been neutralized. The
differential between driving and retarding forces was expected to be heterogeneous as a result of
the microstructural gradients observed through the cross-section. As mentioned earlier, the Alcan
approach of increasing the particle-pinning effect of insoluble dispersoids on migrating grain
boundaries was inapplicable (13). It is reasonable to assume that the primary source of retarding
force, namely the highly stable Al3Zr precipitates, would be unaffected by exposure to such low
homologous temperatures (7). Any other retarding forces, such as those associated with soluble
precipitates, would be expected to be homogenized across the nugget as a result of the IAT. The
findings are inconsistent with an increase in retarding force contributing to AGG suppression at
the AR location. A pressing question remains; "Is there anything else that is different in the
vicinity of the AR location that could contribute to a reduction in the driving force for AGG ?", i.e.
the shaded area on the plots in Figure 13.

It is clear that the combination of FSW and SFD results in very complicated microstructural
characteristics within the deformed weld nugget. The complex [sub]grain structure stems from
the occurrence of dynamic RX and the concurrent development of a strong deformation texture
(6). During the IAT, the microstructure is further transformed by the prolonged exposure at a
relatively high, recovery annealing temperature (8). As a consequence of uniform grain
coarsening and no new grains emerging, it is not surprising that the overall textural composition
across the weld nugget is relatively unaffected by the IAT (26). It also explains why the intensity
of the <100>-fiber [RX] is weak at all locations, i.e. dynamic RX + recovery only. Of significance,
is the apparent increased susceptibility to AGG at the AR location, where the <111>- and <112>-
fibers dominate the texture composition. The microtexture data shows that the contribution of
these fibers increases during the IAT, which indicates that annealing causes these deformation-
type components to intensify. This is not unexpected because texture sharpening is commonly
associated with recovery mechanisms, and the development of better-defined (sub)grain
structures (6).

Consistent with previous work, the presence of the <111>-fiber suggests that the deformation
behavior during FSW is similar to that which occurs during extrusion (17). It may also be inferred
that there is a significant shear component oriented in the through-thickness direction [// S], which
has been associated with helical mass flow (18). The presence of such a preferred texture
accounts for some of the differences between the observed misorientation distributions and the
traditional MacKenzie plot (21). The misorientation distribution data shows that the frequency of
subgrain boundaries [θ<15°] is lower at the AR and t/4 locations, and also reveals only subtle
changes during the IAT. In general, the microstructure comprises ~80% HAGB’s and ~20%
LAGB's, and the ratio is affected little by the IAT. A higher frequency of LAGB's than random
polycrystalline aggregate throughout the cross-section is not to be unexpected. Al-Li alloys are
highly resistant to static recrystallization, and dynamic recrystallization tends to produce more
subgrain structure (20). The observation of a change in the distribution of boundary
misorientations within each of the two categories may be significant.

The strong <111>-fiber is probably related to the directionality of shear deformation around a
rotating and translating FSW tool (17). The strong <112>-fiber texture at the AR location is not
related to deformation by rolling, as is the case for the β-fiber (26). However, such a texture
characteristic may result from the rigid body rotation of the idealized texture fibers associated with
multi-directional shear deformation (20). That this should occur adjacent to the advancing
interface is reasonable, based on the turbulent material flow envisioned in this vicinity during the
FSW process (18). This may have provided additional impetus for the reaction to occur at the A R
location, but only in the as-formed condition. The microtexture data also suggests that some
strengthening of the <111>- and <112>- fibers occurred during the IAT. It is possible that small
changes in grain orientation can reduce the driving force for the AGG process by eliminating the
growth advantage of selected grains at the AR location (9).

Crystallographic texture with a strong single component has been recognized as a driving force
for AGG (6), but it may also act as a retarding force under the certain circumstances. For
example, the effect of systematic texture gradients on DGG per se has yet to be determined.
Furthermore, any influence of the periodic texture fluctuations associated with the onion ring
features is likely to be more complicated. The onion rings can be associated with the 'advance
per revolution' characteristic of FSW processing and localized directionality of material flow (18).
The banding results from grain orientation variations created by the transient shear deformation
(17). The varying spacing between bands likely coincides with the localized levels of strain
across the nugget. The rings are most evident between t/4 and 3t/4 in the cross-section because
that is where the rotational contribution to material flow is considered to be highest (20). As the
translational contribution increases toward the extremities of the nugget, the nature of the shear
deformation involved becomes more complex. It is likely to be more severe on the advancing
side, and reach a maximum adjacent to the TMAZ interface.

However, the extreme inclination of the onion rings at t/4 and 3t/4 may explain why the <110>-
fiber is so well-developed at these locations, i.e. mass flow dominated by multi-directional shear
deformation [torsion] (22). Toward the lateral extremes of the FSW nugget, the translational
contribution increases, with similarities to an extrusion (17). This is particularly true on the
advancing side where the passage of the FSW tool results in severe deformation and creates an
abrupt microstructural interface (18). This explains why the shear deformation textures are
strongest in this region, i.e. conflicting rotational and translational contributions. It is reasonable
to assume that the area of the microstructure that is the most heterogeneous would also be most
susceptible to AGG. In this case, isolated grains with the largest growth advantage over the
surrounding array are likely present in the vicinity of the AR location.

The effects of the IAT on boundary misorientation distribution at the AR location are subtle, and
are presented in more detail in Figure 14. As mentioned, the frequency of LAGB's is higher, and
that of HAGB's lower than for a randomly-oriented array of grains (21). It was observed that the
division between θ<30° and θ>30° boundaries was relatively unaffected by the IAT. However,
the frequency of LAGB's [θ<15°] does seem to increase, which is consistent with the formation of
substructure via recovery-type mechanisms during the IAT. Amongst the HAGB’s [θ>15°], the
frequency of 25-50° misorientations decreases and the frequency of 50-60° misorientations
increases. It is widely accepted that certain grain boundary types can migrate faster during
recrystallization (6). Specifically, it has been documented that boundary misorientations close to
the axis-angle pair relationship 40°<111> exhibit enhanced mobility in Al alloys (27). Therefore,
these transitions may be evidence of an overall decrease in grain boundary mobility at the A R
location, and may have contributed to AGG suppression.

0.08
MacKenzie
0.07 Start of IAT
End of IAT
0.06
Θ= 15
Frequency

0.05
LAGB’s HAGB’s
0.04

0.03

0.02

0.01

0
0 10 20 30 40 50 60 70

@ AR Misorientation (degrees)
Figure 14. Effect of the IAT on [sub]grain boundary misorientation distribution at the AR
location. The overall distribution is different from a random distribution (21), but the two
are coincident at θ ~30° The division between θ>30° and θ<30° boundaries is unaffected
by the IAT, but the frequencies within the two categories undergo subtle changes.
 SUMMARY

The fact that the material examined in this case study had been deformed after FSW, and only a
partial weld nugget remained after SFD, presented a unique opportunity. Consequently,
metallurgical analyses were conducted along a diagonal traverse of the much reduced cross-
section, in order to effectively capture any gradient effects. Useful insight on the reduction in the
net driving force for AGG was gained by quantifying microstructural differences between the start
and end of the IAT, particularly at the AR location. The current findings are consistent with
modification of a dynamically recrystallized microstructure via CGG and recovery-type
mechanisms.

The results presented suggest that the growth advantage of isolated grains can be negated by
employing a simple, in-process IAT. Uniform microstructural coarsening within the deformed
weld nugget, plus subtle changes in grain orientation and boundary misorientation, is deemed to
be responsible. The proposed mechanism for the reduction in net driving force for AGG is a
combination of factors, particularly as documented at the A R location. This was the region of the
deformed weld nugget where AGG was the most persistent during the thermal processing
studies.

It is suggested that the first reduction in stored energy is by elimination of total boundary area,
resulting from a measurable (~30%) increase in (sub)grain area. The second involves re-
alignment of the grain structure (to a lower energy configuration), as evidenced by subtle shifts in
microtextural composition. The third may involve a reduction in grain boundary mobility, as
evidenced by the shift in the misorientation distribution of HAGB's. Individually, these
contributions may not be sufficient to suppress AGG, but in combination they may exert a
substantial influence. Returning to the concepts adopted for this case study (Figs. 4, 5 & 8), it
may be assumed that the combined contribution will have a significant effect on the '(M.γ/R)A
term’ in the Humphreys model (9).

CONCLUSIONS

The application of a simple IAT to a friction-stir-welded and spin-formed Al-Li alloy 2195 material
proved effective in suppressing AGG. The analyses suggest that uniform coarsening of a
dynamically recrystallized microstructure was primarily responsible.

The approach combined microstructural stability concepts developed by Humphreys with gradient
microstructure effects and traditional processing protocols. The objective was to promote CGG
(during the IAT) to eliminate DGG (during SHT), i.e. suppress AGG.

First, the AGG threshold temperature was established, and second, the duration of a recovery
annealing treatment below that temperature was determined. The processing philosophy
adopted is applicable to many Al alloys scheduled for post-FSW deformation.

AGG was most prevalent at the intersection between the advancing side [AS] and the root side of
the deformed weld nugget. EBSD analyses revealed that this 'AR' location exhibited the smallest
grain size, strongest texture and highest frequency of HAGB’s.

The most effective IAT caused an approximately 30% increase in (sub)grain area across the weld
nugget. The subtle changes in grain orientations and HAGB misorientations accompanying the
(sub)grain coarsening were consistent with CGG and recovery-type mechanisms.

Insertion of a stand-alone IAT may allow for the cost effective manufacture of large, complex-
shaped articles using single-piece forming operations. The current work implies that the size
limitations on commercially-available stock can be overcome by using tailored starting blanks
which have been friction stir welded together.
 FUTURE WORK

Objectives of ongoing investigations will include;

 applying the processing philosophy used to the heat treatment of other Al alloy materials
subjected to post-FSW deformation, including different forming techniques.

 categorizing HAGB's in terms of coincident site lattice [CSL] boundaries, to establish

whether average grain boundary mobility at the AR location is affected by the IAT.

 evaluating the influence of soluble, second phase particles on grain boundary mobility,
i.e. increased solute drag effects resulting from dissolution of precipitates during the IAT.

ACKNOWLEDGMENTS

The authors are very grateful to Joel A. Alexa and Harold D. Claytor of Lockheed-Martin Space &
Science Division, Hampton, Virginia, for performing the processing work in the Light Alloy
Laboratory at NASA Langley Research Center. The first author (SJH) dedicates this paper to his
mentor and friend, Mr. Thomas T. Bales (NASA retired), who made him a better blacksmith.

REFERENCES

1. Space Shuttle External Tank System Definition Handbook, Vols. 1 & 2. LMC-ET-SE61-1.
New Orleans, LA: Lockheed Martin Michoud Space Systems, December 1997.

2. Curreri, P.A, et al. Aluminum-Lithium, Friction Stir Welded, Spun-Formed Domes for
Light-weight Cryogenic Propellant Tanks. NASA/TP-11-216462. Huntsville, AL: Marshall
Space Flight Center, March 2011.

3. Nelson, T.W., Steel, R.J., and Arbegast, W.J. Investigation of heat treatment on the
properties of friction stir welds. AeroMat 2001. Long Beach, CA: ASM International, June
2001.

4. Hales, S.J., and Tayon, W.A. Heat treatment of a friction-stir-welded and spin-formed Al-Li
alloy. Procedia Engng. 2011, Vol. 10, pp. 2496-2501.

5. Doherty, R.D., et al. Current issues in recrystallization: a review. Mater. Sci. & Engng.
1997, Vol. A238, pp. 219-274.

6. Humphreys, F.J., and Hatherly, M. Recrystallization and Related Annealing Phenomena.

Second Edition. New York, NY: Elsevier, Ltd., 2004.

7. Hatch, J.E. [ed.]. Aluminum - Properties and Physical Metallurgy. Metals Park, OH: ASM,
1984.

8. Rollett, A.D., Brahme, A.P., and Roberts, C.G. An overview of accomplishments and
challenges in recrystallization and grain growth. Mat.Sci.Forum. 2007, Vols. 558-559, pp. 33-
42.

9. Humphreys, F.J. A unified theory of recovery, recrystallization and grain growth, based
on the stability and growth of cellular microstructures - I. The basic model. Acta Mater.
1997, Vol. 45, pp. 4231-4240.

10. Humphreys, F.J. A unified theory of recovery, recrystallization and grain growth, based
on the stability and growth of cellular microstructures - II. The effect of second phase
particles. Acta Mater. 1997, Vol. 45, pp. 5031-5039.
11. Dennis, J., Bate, P.S., and Humphreys, F.J. Abnormal grain growth in Al-3.5Cu. Acta
Mater. 2009, Vol. 57, pp. 4539-4547.

12. Rioja, R.J. Fabrication methods to manufacture isotropic Al-Li alloys and products for
space and aerospace applications. Mat.Sci.& Engng. 1998, Vol. A257, pp. 100-107.

13. Ehrstrom, J.-C., and Warner, T. Manufacturing method for friction welded aluminum alloy
parts. US Patent 7,490,752 B2, February 17, 2009.

14. Heat treatment of wrought aluminum alloy parts. AMS2770H. Aerospace Material
Specifications. Warrendale, PA: SAE International, August 2006.

15. Standard test methods for determining average grain size. ASTM E112 - 96. International
Book of Standards. West Conshohoken, PA: ASTM, 2004. Vol. 03.01.

16. Practice for determining average grain size using electron backscatter diffraction
(EBSD) in fully recrystallized polycrystalline materials. ASTM E2627 - 10. International Book
of Standards. West Conshohoken, PA: ASTM, 2010. Vol. 03.01.

17. Schnieder, J.A., and Nunes, A.C. Characterization of plastic flow and resulting
microtextures in a friction stir weld. Met.& Mat. Trans. 2004, Vol. 35B, pp. 777-783.

18. Fonda, R.W., Reynolds, A.P., et al. Material flow in aluminum friction stir welds. Mater.
Sci. Forum: Trans. Tech. Publ., Switzerland, 2012. Vols. 706-709, pp. 983-989.

19. Hassan, Kh.A.A., et al. Stability of nugget zone grain structures in high strength Al-alloy
friction stir welds during solution treatment. Acta Mater. 2003, Vol. 51, pp. 1923-1936.

20. Field, D.P., Nelson, T.W., et al. Heterogeneity of crystallographic texture in friction stir
welds of aluminum. Met. Mater. Trans. 2001. Vol. 32A, 11, pp. 2869-2877.

21. MacKenzie, J.K. Statistics associated with random disorientation of cubes. Biometrika.
1958. Vol. 45, p. 229.

22. Hosford, W.S. The Mechanics of Crystals and Textured Polycrystals. New York, NY:
Oxford University Press, 1993.

23. Fonda, R.W., and Knipling, K.E. Texture development in friction stir welds. Sci.& Tech.
Welding & Joining. 2011, Vol. 16, 4, pp. 288-294.

24. Hales, S.J., and Hafley, R.A. Texture and anisotropy in Al-Li alloy 2195 plate and near-
net-shape extrusions. Mat.Sci.& Engng. 1998, Vol. A257, pp. 153-164.

25. Hales, S.J., and Hafley, R.A. Structure-property correlations in Al-Li integrally stiffened
extrusions. NASA/TP-2001-210839. Hampton, VA: Langley Research Center, April 2001.

26. Fricke, W.G., and Przystupa, M.A. Texture. Aluminum Alloys - Contemporary Research and
Applications. A.K. Vasudevan and R.D. Doherty (eds.). New York, NY: Academic Press, 1989,
Vol. 31, pp. 563-578.

27. Huang, Y., and Humphreys, F.J. Measurements of grain boundary mobility during
recrystallization of single-phase aluminium alloys. Acta Mater. 1999, Vol. 47, 7, pp. 2259-
2268.