Materials Science and Engineering A277 (2000) 102 – 113

www.elsevier.com/locate/msea

Microstructural and mechanical characterization of a superplastic

6xxx aluminum alloy
L.P. Troeger 1, E.A. Starke Jr *
Department of Materials Science and Engineering, Uni6ersity of Virginia, Thornton Hall, Charlottes6ille, VA 22903 -2442, USA

Received 2 February 1999; received in revised form 29 July 1999

Abstract

A new thermomechanical process has been developed which produced a fine grain structure in an Al – Mg – Si – Cu alloy. The
alloy under investigation falls within the composition limits of both 6013 and 6111. The refined microstructure has an average
grain diameter of approximately 10 mm and an average aspect ratio near 1.6. Superplasticity was investigated using ambient
pressure, uniaxial tensile tests and cone tests with backpressure. The refined material exhibits superplasticity above 500°C.
Uniaxial tests indicated a strain rate sensitivity of 0.5 at 540°C, where the elongation reached 375% for a flow stress of 680 psi
(4.7 MPa). Cone tests revealed excellent overall formability, and the suppression of cavitation with backpressure. The effects of
strain on grain size and porosity were determined. © 2000 Elsevier Science S.A. All rights reserved.

Keywords: Microstructure; Mechanical characterization; Superplastic; Aluminium alloy

1. Introduction strengthened appreciably during the paint-bake cycle

used by the automotive industry.
Aluminum alloys have been the material of choice Superplastic forming (SPF) is a manufacturing pro-
for aircraft construction since the 1930s. The aerospace cess that exploits the phenomenon of superplasticity
industry relies heavily on 2xxx and 7xxx alloys, while a [3–6]. Currently, there is increased commercial interest
wide variety of aluminum alloys are used by the auto- in SPF by both the automotive and aerospace indus-
motive industry. Currently, the automotive industry is tries. During SPF, materials are formed into complex
showing an increasing interest in aluminum alloys as shapes using gas pressure with minimal energy expendi-
structural materials [1,2]. 6xxx Aluminum alloys are of ture (gas pressure B 7 MPa (1000 psi)) due to the low
particular interest to both the aerospace industry (for resistance of these materials to plastic flow. SPF re-
fuselage skins and other applications) and automotive duces tooling costs since only one single-surface tool is
industry (for body panels and bumpers) because of required. Another benefit of SPF derives from the
their attractive combinations of properties. The benefits complexity of parts that can be formed in a single
of 6xxx alloys include medium strength, formability, operation. This leads to reduced part counts and a
weldability, corrosion resistance, and low cost. 6xxx reduced need for fasteners and connectors, decreasing
alloys have corrosion resistance which is superior to the weight of the final product. Decreased product/ve-
2xxx and 7xxx alloys, traditional aerospace alloys that hicle weight improves fuel efficiency. One disadvantage
are prone to intergranular attack. The low cost of 6xxx
of SPF is that it is a slow process due to the low
alloys is especially significant to the aerospace industry,
strain-rate range in which superplasticity normally oc-
that relies heavily on the more expensive 2xxx and 7xxx
curs (near 10 − 4 s − 1). This can lead to forming times of
alloys. Since 6xxx alloys are heat treatable, they can be
0.5–2 h per part. Therefore, decreased forming times
obtained through microstructural refinement can be
* Corresponding author. Tel.: +1-804-924-6335; fax: + 1-804-924-
1353.
extremely beneficial.
1
Currently National Research Council Post-Doctoral Research Microstructures suitable for SPF can be produced
Associate, NASA Langley Research Center, USA. through thermomechanical processing (TMP). A fine,

0921-5093/00/$ - see front matter © 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 9 2 1 - 5 0 9 3 ( 9 9 ) 0 0 5 4 3 - 2
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 103

statically recrystallized grain structure can be achieved report the application of the approach of Waldman et
via a discontinuous recrystallization process which ex- al. and Paton and Hamilton to commercial aluminum
ploits large (\1 mm diameter) and small (B0.5 mm alloys 6061, 6063, and 6066 [19,20], and only the latter
diameter) particles present in the microstructure [7–11]. of these reported any degree of superplasticity. Major
Waldman et al. [12] and Paton and Hamilton [13] modifications to the general approach are required to
patented general approaches for the grain refinement of produce a fine, equiaxed grain structure in aluminum
aluminum alloys through thermomechanical processing. alloy 6013/6111 sufficient to induce superplasticity.
The general theory embodied by the work of Waldman
et al. and Paton and Hamilton can be described as
follows. First, the material is given a solution heat 2. Related investigation
treatment, usually followed by a water quench. The
material is then over-aged to give a uniform distribu- A previous investigation of the grain refinement of
tion of precipitates. Cold work can be used prior to the Baseline 6013 was made by Chung et al. [21] (the
over-aging step to improve the distribution of over- composition for Baseline 6013 will be discussed later).
aged precipitates, by providing a uniform network of A thermomechanical process was developed, consisting
dislocations to serve as heterogeneous nucleation sites of solution heat treatment, 10% cold rolling, over-aging
for them [7,10]. However, dislocation interactions, at 380°C, 90% warm rolling at 190°C, and static recrys-
rather than the dislocations themselves, may provide tallization. The process produced a marginally super-
nucleation sites for precipitates [14]. Once the over-aged plastic microstructure with an approximately 12–13 mm
precipitates are present, large subsequent rolling reduc- grain size, which produced a maximum elongation of
tions result in the formation of deformation zones 230% at 520°C for a strain rate of 3× 10 − 4 s − 1. The
around the precipitates. A recrystallization treatment flow stress under these conditions was 6.7 MPa (970
follows, during which the highly strained regions of the psi). The maximum strain rate sensitivity was 0.38 for a
deformation zones serve as nucleation sites for recrys- strain rate range of 3×10 − 4 to 6× 10 − 4 s − 1 at 563°C.
tallized grains, a process known as particle-stimulated Closer examination of the precipitate structure pro-
nucleation (PSN) [15 – 18]. duced by the process of Chung et al. revealed that it
Although extensive research has been devoted to the was not homogeneous, likely leading to a non-uniform
development of fine-grained, superplastic 2xxx, 5xxx, grain structure.
7xxx, and 8xxx aluminum alloys by static or dynamic
recrystallization, few studies have been made of the
grain refinement of 6xxx alloys for superplasticity. The 3. Experimental procedure
objective of our research was to develop a microstruc-
ture for a 6013-type aluminum alloy that is conducive Aluminum alloy 6013 is a medium strength, heat-
to superplastic forming for use by the aerospace and treatable automotive/aerospace alloy developed by AL-
automotive industries. The goals of the project were: (1) COA in the early 1980s. It can be processed to develop
to produce a microstructure that exhibits superplasticity properties superior to other 6xxx alloys.[22] The pri-
through the development of a fine ( 10 mm) grain size mary alloying elements are magnesium, silicon, and
and weak or random texture by exploiting particles in copper. The alloy has corrosion resistance superior to
conjunction with thermomechanical processing and 2xxx and 7xxx aluminum alloys, which are heavily used
compositional variation; (2) to determine the high-tem- for aerospace applications. The yield strength of 6013-
perature deformation characteristics of that microstruc- T6 is 12% higher than that of 2024-T3, and it is nearly
ture; and (3) to determine the extent of cavitation and immune to exfoliation corrosion and stress-corrosion
grain growth caused by superplastic deformation. cracking. It is 25% stronger than AA 6061-T6, and
It should be noted here that although a general 6013-T4 has better stretch-forming characteristics than
approach for the grain refinement of precipitation- other aerospace aluminum alloys.[23,24]
hardening aluminum alloys existed prior to the current
investigation, it has not been applied as successfully to 3.1. Alloy de6elopment and ingot processing
the 6xxx aluminum alloys as to, for instance, the 7xxx
alloys. Waldman et al. [12], as well as Paton and The majority of the research focused on a slight
Hamilton [13], successfully applied this process to alu- variation of the composition of the Baseline 6013 stud-
minum alloy 7075. This alloy was then shown to exhibit ied by Chung et al. [21]. The iron content was reduced
superplasticity. Although Paton and Hamilton [13] slightly to reduce the volume fraction of constituent
cited several 6xxx alloys as examples for the effective- particles, which are often cracked during processing
ness of their process, limited experimental evidence and bind with manganese, making it unavailable for the
exists in the literature for the successful application of formation of dispersoid particles to control grain struc-
this method to commercial 6xxx alloys. Two studies ture. The composition of the new alloy, 6013 Variant 1,
104 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113

is lower in iron content but still falls within the com- distributions were made for these three processes
position limits of 6013 (as well those of 6111). The only.
copper content was also decreased slightly from the Identical downstream rolling and recrystallization
level of Baseline 6013 for improved corrosion resis- treatments were applied to material subjected to pro-
tance. The compositions of Baseline 6013 and 6013 cesses A, B, and C (in parallel). Through comparison
Variant 1 are shown in Table 1. The compositional of the resultant recrystallized grain structures, a single
ranges for 6013 and 6111 are shown for comparison. processing scheme was identified which produced the
All values are in wt.%; the balance is aluminum [25]. best candidate microstructure for superplasticity
The new alloy 6013 Variant 1 was cast and ingot- (patent application submitted) [28]. Process A was
processed by Reynolds Metals Company. One-half of thereby identified as the best precipitation treatment.
the ingot was preheated using a standard treatment, After a brief study of the effect of post-age rolling
while the other underwent a low-temperature preheat, strain, strain rate, and reduction per pass on the re-
shown by Thanaboonsombut and Sanders [26] to re- crystallized grain structure, the optimal downstream
sult in a finer size distribution and slightly higher processing conditions were also identified (rolling and
volume fraction of dispersoids than the standard pre- static recrystallization). Recrystallization treatments
heat. Preheat schedules are shown in Table 2. Both involved 5–60 min soaks at 540°C.
ingots were then rolled to 1 inch plate. Grain statistics were derived from image analysis of
grain boundary maps. The grain boundary maps were
generated from electron backscattering patterns mi-
3.2. De6elopment and characterization of precipitate crotexture data [29–31]. The grain boundary maps
and grain structures represented boundaries with misorientations greater
than 10°. This prevents low-angle boundaries from
Prior to processing, all material was solution heat influencing the statistical grain size. Microtexture
treated for 1 h at 540°C and water quenched. Ther- scans were obtained from midwidth, midthickness LS
momechanical processing parameters were varied to sections.
optimize the size, shape, and distribution of over-aged General texture analysis was conducted using tradi-
precipitates. Aged samples were electropolished and tional macrotexture and orientation distribution func-
observed in scanning electron microscopy (SEM) at tion (ODF) data, obtained using a Philips texture
midthickness using midwidth LS sections, see Fig. 5. goniometer and associated Philips PC-Texture/ODF
Mean particle diameter and aspect-ratio statistics were software (version 2.2). Macrotexture specimens were
obtained from digitized SEM images using commer- midthickness, LT sections.
cial image analysis software. Interparticle spacings In this manner, a candidate processing route was
were expressed as the mean free distance between identified for use in high-temperature deformation
particles (an edge-to-edge distance), hand calculated tests in order to confirm and characterize superplas-
using results of the image analysis, coupled with prin- ticity. All material used in the subsequent uniaxial
ciples of quantitative stereology. [27] Three candidate and cone tests was processed according to process A,
processes (A, B, and C) produced homogeneous dis- followed by the optimized downstream processing
tributions of micron-sized particles. Quantitative com- conditions. No further work was conducted using ma-
parisons of precipitate sizes, shapes, spacings, and terial from processes B or C.

Table 1
Alloy compositions

Alloy Mg Si Cu Mn Fe Other

Baseline 6013 0.89 0.74 0.90 0.33 0.26 0.03 Zr

6013 Variant 1 0.80 0.70 0.80 0.30 0.10 –
6013 range 0.8–1.2 0.6–1.0 0.6–1.1 0.2–0.8 0.5 0.1 Cr+0.25 Zn+0.1 Ti
6111 range 0.5–1.0 0.7–1.1 0.5–0.9 0.15–0.45 0.4 0.1 Cr+0.1 Zn+0.1 Ti

Table 2
Ingot preheat schedules

Preheat treatment Ramp heat-up rate (°C/h) Soak temperature (°C) Soak time (h)

Standard 50 560 4
Low-temperature 50 500 8
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 105

3.3. Formability testing imens. The recrystallized material was still allowed to
heat up for 5 min prior to forming, as in previous tests.
Uniaxial tensile tests were performed at atmospheric This series of tests was conducted primarily for condi-
pressure using a computer-driven Instron test rig fitted tions near those that produced the tallest cone in the
with a five zone, ATS clamshell-type furnace. All uniax- first series of tests. The cones in the second group of
ial experiments were conducted at constant crosshead tests were allowed to form to failure even if it took
velocity using dogbone-type samples with a 0.5-inch longer than 1 h.
gage length cut from statically recrystallized material. A third group of tests was performed on the low-tem-
Strain rate sensitivities were established at 500 and perature preheat 6013 Variant 1, grain refined by pro-
540°C using step strain rate tests [9]. The jump in cess A and the optimized rolling procedure developed
crosshead velocity was made after 20% strain had for the standard preheat material. These tests were
accumulated. performed only for those conditions which produced
The minimum strain rate investigated by this method the tallest cones from the first two series of tests.
was approximately 4× 10 − 5 s − 1, while the maximum Each cone was measured to determine its height and
was near 1×10 − 3 s − 1. The strain rate range was thickness (at the pole). The strain for the cone is
chosen to bracket those where superplasticity had pre- calculated through the thickness ratio,
viously been observed in Baseline 6013 [21]. The tem-
o= ln(t0/t)
perature and crosshead velocity ranges resulting in the
maximum strain rate sensitivity were thereby where t0 is the starting thickness of the sheet and t is
established. the thickness at the pole [8,32,33]. Strain rates were
Samples were pulled to failure under conditions calculated by dividing the true thickness strain at the
which bracketed the optimum strain rate sensitivity. A pole by the total forming time.
single temperature/strain rate combination was thus
identified which produced maximal tensile elongation. 3.5. Characterization of post-forming microstructure
For comparison, samples cut from commercially avail-
able Baseline 6013-T4 sheet were tested at the same In discontinuously recrystallized alloys, microstruc-
temperature and strain rate. tural changes during superplastic deformation include
grain growth and cavitation. These changes can affect
3.4. Cone tests subsequent formability and elongation to failure. The
grain size was therefore measured again after the tensile
While uniaxial tensile tests are commonly used to or cone tests to determine the extent of grain growth
evaluate general superplastic deformation characteris- due to exposure to stress at high temperature. This
tics, biaxial cone tests offer the advantage of more provides evidence as to the dynamic stability of the
closely simulating a real SPF environment. During the microstructure, i.e. the effectiveness of the manganese-
cone test, a sheet sample is superplastically formed into bearing dispersoids in suppressing grain growth.
a conical die cavity using a net forward gas pressure The tallest smooth cone produced in less than 1 h
[32,33]. (cone C) was sectioned and analyzed in terms of poros-
Cone tests were performed at NASA Langley Re- ity and grain size as functions of strain. Porosity data
search Center for a 59° cone with a base radius of 1 were obtained through image analysis of backscattered
inch. The first group of tests involved constant-pressure electron images of an as-polished centerline cone sec-
forming experiments carried out at 520, 540, and 560°C tion from SEM. Backscattered images were required in
for backpressures of 400, 425, and 450 psi (2.76, 2.93, order to distinguish between particles and pores. The
and 3.10 MPa, respectively). The forward pressure re- entire cone thickness was analyzed for each level of
mained constant at 500 psi (3.45 MPa), i.e. the forming strain. Grain sizes were determined for the same cone
pressure (forward pressure minus backward pressure) section using grain boundary maps generated from
was varied from 50 to 100 psi (0.35 – 0.69 MPa). For the microtexture scans, as described previously.
first and largest group of tests, the material was allowed
to recrystallize in the cone press during the 5 min
heat-up time allotted prior to testing. Each cone was 4. Results and discussion
allowed to form until rupture or until 1 h had elapsed,
whichever occurred first. The three aging conditions mentioned previously (A,
A second, brief series of tests was conducted on B, and C) all resulted in fairly uniform distributions of
material which received a 5 min static recrystallization precipitates, 1–3 mm in diameter, spaced 9–14 mm
treatment prior to insertion into the cone press, in order apart. The precipitate structures resulting from process
to fix the starting grain size at its minimum and reduce A are shown in Fig. 1. Based on the recrystallized grain
the variation of starting grain sizes between sheet spec- structures after the application of identical rolling and
106 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113

statistics were calculated from grain boundary maps, as

described previously. The grain structures resulting
from processes B and C are not as fine as that produced
by process A, probably due to the decreased volume
fraction of particles and increased interparticle spacing
from processes B and C compared with process A.
It is apparent from quantitative comparison of the
average grain diameters shown in Table 3 that process
A yielded the finest, most uniform and equiaxed grain
structure. An electron channelling contrast (ECC) im-
age of the recrystallized grain structure resulting from
process A, followed by rolling and static recrystalliza-
tion is shown in Fig. 2. For comparison, the grain
structure of the unprocessed material (following solu-
tion heat treatment) is shown in Fig. 3. Material from
process A was therefore chosen for a study of the effect
of roll speed, reduction-per-pass, and total rolling re-
duction on the recrystallized grain size and shape. For
six combinations of these parameters, average grain
sizes (on LS sections at midthickness) ranged from

Fig. 1. (a) Precipitate structure produced by process A. SEM, 500× .

(b) Precipitate structure produced by process A. SEM, 5000 × .

Table 3
Grain statistics for processes A–C, following rolling and static
recrystallizationa
Fig. 2. Grain structure produced by rolling and recrystallization after
Process DAVG (mm) sD (mm) AR sAR process A. SEM/ECC, 150 ×.

A 9.48 4.72 1.65 0.55

B 11.63 5.58 1.85 0.80
C 10.80 5.63 1.88 0.67

a
DAVG, average grain diameter; sD, standard deviation of grain
diameters; AR, average grain aspect ratio; sAR, standard deviation of
grain aspect ratios.

recrystallization treatments, process A (followed by the

optimized rolling and recrystallization treatment) was
chosen as the best candidate for superplasticity. The
resultant grain structure was the finest, most uniform,
and most equiaxed of those studied. (These grain char-
acteristics are represented statistically by the average
grain diameter, its standard deviation, and the average
grain aspect ratio, respectively.)
Statistics related to the grain structures ultimately Fig. 3. Grain structure of unprocessed material following solution
produced by processes A – C are shown in Table 3. All heat treatment. SEM/ECC, 150 ×.
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 107

Note that the scanned area and map are both 200 mm
wide.
Based on these results, processing parameters were
identified which produced the finest and most equiaxed
grain structure following process A. Variation in recrys-
tallization temperature was shown to have little qualita-
tive effect on the recrystallized grain structure. Grain
boundary maps taken from the LS, LT, and ST planes
following a 5 min soak at the recrystallization tempera-
ture (540°C) revealed a three-dimensional average grain
size of 10.2 mm. The average increased to 10.7 mm after
a 1 h exposure to the same temperature. The details of
the variation of grain size with orientation and soak
time are shown in Table 4. ECC images of the recrystal-
lized grain structures following the 5 min and 1 h soaks
are shown in Figs. 5 and 6, respectively.
For the low-temperature preheat material subjected
to process A, rolling, and static recrystallization, the
resultant grains were finer but less equiaxed than those
from the standard preheat, as shown in Table 5. The
low-temperature preheat schedule should have resulted
in a higher volume fraction of smaller manganese-bear-
ing dispersoid particles, due to a higher driving force
for nucleation of precipitates and lower solubility at the
reduced preheat temperature, relative to the standard
preheat. Recall that this was shown to be the case in
previous work on 6013-type alloys [26]. The Zener drag
pressure increases with increasing particle volume frac-
tion and decreasing particle radius; therefore, a disper-
sion comprised of a higher volume fraction of smaller
particles should be more effective at pinning grain
boundaries than a lower volume fraction of larger
Fig. 4. (a) Optimized grain structure resulting from process A. particles.
Scanned area is 200 mm wide ×150 mm high. SEM/ECC, 500 × . (b)
Map of grain boundaries with greater than 10° misorientation, corre-
sponding to the optimized grain structure from process A. Map is 200
mm wide.

Table 4
Grain statistics for process A followed by rolling and static
recrystallizationa

Plane Soak time (min) DAVG (mm) sD (mm)

LS 5 9.48 4.47
LT 5 10.35 5.50
ST 5 10.90 4.97
LS 60 10.97 4.72
LT 60 10.47 4.54
ST 60 10.72 4.87

a
DAVG, average grain diameter; sD, standard deviation of grain
diameters.

 9.5 to 11.5 mm, with standard deviations increasing

with grain size from 4.4 to 5.7 mm. The finest grain size
corresponded to the slower roll speed and higher total
rolling reduction, and is shown in Fig. 4a; its corre- Fig. 5. Three-dimensional grain structure produced by rolling and 5
sponding grain boundary map is shown in Fig. 4b. min recrystallization treatment after process A. SEM/ECC, 150 ×.
108 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113

distinct rotated cubeND {001}Ž310 component. This

component has been reported by Engler and Lücke [34]
to be indicative of the nucleation of recrystallization at
particles.
While the overall texture is weak, there is a marked
retention of a portion of the b fiber. This normally
would indicate an unrecrystallized structure, but the
fact that the ODFs are nearly identical following the 5
min and 1 h treatments indicates that the orientations
present are stable. The retention of a rolling texture in
materials experiencing PSN has been observed previ-
ously. Ørsund and Nes [35] developed a preliminary
model based on experimental evidence for the retention
of rolling texture components in a 3xxx aluminum
alloy. They linked the recrystallization texture to the
location of the nucleation event inside the deformation
zone. The size of the deformation zone is on the order
of the particle diameter, and contains a gradient of
subgrain orientations and dislocation densities which
vary with distance from the particle. Assuming that
recrystallization occurs by the growth of subgrains
within the deformation zone, the location of the sub-
grain that ultimately grows to consume the deformation
Fig. 6. Three-dimensional grain structure produced by rolling and 1 h zone will determine the relative orientation of the re-
recrystallization treatment after process A. SEM/ECC, 150 × . crystallized grain that results.
Ørsund and Nes [35] theorized that if the nucleus
Table 5 originated at the edge of the deformation zone, where
Grain statistics for low-temperature preheat materiala the local degree of misorientation is small, and grows
towards the particle, the resultant grain will have an
Soak time DAVG (mm) sD (mm) AR sAR
orientation closely related to that of the matrix defor-
5 min. 9.23 4.36 1.96 0.71 mation texture. If the nucleus originated at the matrix/
1 hour 9.75 4.73 2.05 0.83 particle interface where the local misorientation is most
intense, the nucleus and resultant grain have an orienta-
a
DAVG, average grain diameter; sD, standard deviation of grain
tion distinctly different from that of the deformed
diameters; AR, average grain aspect ratio; sAR, standard deviation of
grain aspect ratios. matrix. In this case, the recrystallized texture in nearly
random, given a uniform particle distribution. How-
Although a dispersion of small particles can be ever, there is currently little microstructural evidence
beneficial in preventing grain growth, they may also that nucleation of recrystallization actually occurs in
the periphery of the deformation zone [18].
retard PSN by exerting a drag force on recrystallization
nuclei. Recrystallization nuclei with enough driving
4.1. Uniaxial tensile tests
force to overcome the pinning effect of the dispersoids
may then consume the deformation zones of adjacent
Fig. 8 shows the variation of strain rate sensitivity
nuclei that cannot overcome the dispersoids. Mechani-
with strain rate, for the step strain rate tests of opti-
cal fibering of the dispersoids prevents advancement of mally processed, statically recrystallized 6013 Variant 1
the boundaries of the recrystallization nuclei in the at 500 and 540°C. The material showed a maximum
direction normal to the rolling plane, so the final recrys- strain rate sensitivity of 0.5, which occurred at 540°C
tallized grain structure is elongated along the rolling for a strain rate range of 2×10 − 4 to 5 ×10 − 4 s − 1
direction. (based on initial gage length). Fig. 9 shows the elonga-
ODF and microtexture analyses indicated a very tion as a function of strain rate. The elongation to
weak texture on both LT and LS planes following the 5 fracture reached 375% for a strain rate of 5× 10 − 4 s − 1,
min and 1 h recrystallization treatments. An ODF from with a corresponding maximum stress of  680 psi (4.7
a midthickness, LT plane for material subjected to MPa). Failure was due to cavitation.
process A and the optimized downstream rolling and 1 Fig. 10 shows an undeformed sample alongside sam-
h recrystallization treatment is shown in Fig. 7. Exami- ples which elongated 350 and 375% prior to fracture.
nation of the 82 = 90° section reveals the presence of a (A Baseline 6013-T4 sheet tested under the same condi-
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 109

Fig. 7. ODF after process A, followed by rolling and a 1 h recrystallization soak at 540°C. LT plane.

tions fractured after about 120% elongation with a Very near the site of fracture, the average grain size was
maximum stress of 860 psi (6.0 MPa).) The best less than 17.5 mm. In theory, the presence of man-
uniaxial results produced by the current TMP are com- ganese-bearing dispersoid particles in the alloy should
pared with those from a previous work [21] in Table 6. exert a drag force on grain boundaries, retarding grain
Analysis of microtexture data taken from an area
located at about one-third the gage length (from a
sample which fractured at midgage after 350% elonga-
tion) indicated an average grain size of 14.4 mm, with a
standard deviation of approximately 6.7 mm. The aver-
age aspect ratio for this group of grains was approxi-
mately 2.0. The volume fraction of pores between
quarter-gage and midgage length for the same sample
was 14–20%.

4.2. Cone tests

Figs. 11 and 12 show that the dependence of cone Fig. 8. Strain rate sensitivity as a function of strain rate for material
height-to-radius ratio on strain rate and forming time, subjected to process A followed by rolling and static recrystallization.
respectively, is weak for tests with backpressure over
the strain rate range investigated. The tallest smooth
cone from the in-situ-recrystallized series was consid-
ered for analysis (cone C). Cone C, the tallest smooth
cone produced in less than 1 h, had a height-to-radius
ratio (h/r) of 1.15. This cone was formed at 540°C with
75 psi (0.52 MPa) forming pressure (500 psi (3.45 MPa)
forward pressure minus 425 psi (2.93 MPa) backpres-
sure). Cone C ruptured near the crown after 26 min of
forming, corresponding to a strain rate in the order of
1×10 − 3 s − 1.
Figs. 13 and 14 show the variation of grain size and
mean grain aspect ratio with strain, respectively, for
cone C. The mean grain aspect ratio remained fairly
constant during forming, characteristic of superplastic- Fig. 9. Elongation to failure as a function of strain rate for tests at
ity. For the deformed areas of the cone, the grain size 540°C for material subjected to process A followed by rolling and
remained steady at around 15 mm up to strains near 1.2. static recrystallization.
110 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113

strain rate were observed to follow the same trends as

the uniaxial length strain measurements [36]. A similar
comparison is not feasible for the current investigation,
since the uniaxial measurements of Agrawal et al. [36]
were performed under constant true strain rate, while
those in the current research were conducted at con-
stant crosshead speed. Constant crosshead speed tests
lead to a decreasing strain rate as deformation pro-
gresses. However, the strain rate for a cone test is also
probably not constant since experimental observations
by Guo et al. [37] showed that bulge-formed domes of
aluminum alloy 7475 underwent a sigmoidal increase in
cone height with time, but the thickness gradient within
the dome decreased nearly linearly with the height of
Fig. 10. Superplastically deformed tensile specimens which fractured
some reference point measured from the dome periph-
after 375% (top) and 350% (middle) at 540°C. An undeformed ery. Guo et al. [37] also observed that the domes often
specimen is shown for comparison (bottom). deviated from the theoretically spherical shape.

Table 6
Comparison of current and previous TMPs in terms of induced
superplastic performance

Parameter Current process Chung et al.

(1994) [21]

mmax (temp.) 0.50 (540°C) 0.38 (563°C)

o; for mmax (s−1) 2×10−4 to 3×10−4 to
5×10−4 6×10−4
Maximum elongation 375% 230%
o; for maximum elongation 5×10−4 3×10−4
(s−1)
s for maximum elongation 680/4.7 970/6.7
(psi/MPa) Fig. 12. Variation of cone height-to-radius ratio with time to failure
for all cones formed to failure.

Fig. 13. Average grain diameter as a function of strain for cone C.

Fig. 11. Variation of cone height-to-radius ratio with strain rate for
all cones formed to failure.

growth. Although the grain structure has been shown

to be statically stable, some dynamic grain growth
occurred at high strain levels.
Previous researchers have derived strain rate sensitiv-
ity data from plots of log stress versus log strain rate
for cone tests and correlated them with those from
uniaxial tests with reasonable success. For their experi-
ments, variations in engineering thickness strains with Fig. 14. Grain aspect ratio as a function of strain for cone C.
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 111

initial sample cross-sectional area for the uniaxial tests.

In general, the flow stresses associated with the cone
tests were higher than those of the uniaxial tests. How-
ever, the cone tests did overlap the stresses of the
uniaxial tests near the regime of ‘‘optimal’’ superplastic-
ity as indicated by the step strain rate tests. The stresses
for the cone samples at 540°C ranged from 631 to 1040
psi (4.35–7.17 MPa), with stress increasing with strain
rate. The stresses for the uniaxial samples ranged from
482 to 1215 psi (3.32–8.38 MPa), also increasing with
strain rate. Thus, the conditions of the cone tests were
likely a subset of the conditions of the uniaxial tests to
failure as far as flow stresses were concerned.
The range of strain rates investigated by the cone
Fig. 15. Variation of porosity with strain for cone C. tests which most closely paralleled the conditions of the
uniaxial tests to failure (at 540°C for standard preheat
It was neither intended nor attempted in the current material only) was 3.2× 10 − 4 to 3.8 ×10 − 3 s − 1. The
work to compare uniaxial and cone test results in terms results of the cone tests showed a slight increase in cone
of strain rate sensitivity, strain, or strain rates for height with decreasing strain rate. The lack of a pro-
several reasons. First, the uniaxial tests were performed nounced variation in cone height with strain rate is
to obtain strain rate sensitivity values (through the likely due to: (1) the narrow strain rate range over
accepted method of jump strain rate tests) and engi- which the tests were conducted (about one order of
neering strains to failure for comparison with similar magnitude difference in strain rate); (2) differences in
studies of other alloys and/or processes. Second, while the starting microstructure due to the fact that most of
in theory there may be a correspondence between strain the cone samples were recrystallized in-situ in the cone
due to length increase in a uniaxial test, and strain due press, known to cause slight deviations from the ideal
to thickness decrease for a cone sample deformed under grain structure (non-constant initial microstructures
identical loading and strain rate conditions, experimen- might prevent the identification of clear trends in the
tal variation in cone test results (due to method of formability data); and (3) slight variation in the temper-
recrystallization, heat-up rate, work hardening or lack ature of the cone tests, due to experimental difficulty in
thereof, and temperature gradients) makes one-to-one maintaining a constant temperature, which would ob-
comparison of uniaxial and cone test results unjustified scure data trends.
for the current study. Microstructural analysis of cone C indicated that this
The cone tests involved in the current study were test (and, one can assume, any at slower cone strain
conducted near the temperature shown by the uniaxial rates) did in fact involve superplasticity. This conclu-
tests to produce superplasticity. The cone tests were sion is based upon two observations. First, there was
intended to preliminarily screen overall formability due not a significant increase in the grain aspect ratio for
to different recrystallization treatments, forming tem- this cone from the undeformed region to the failed
peratures, forming pressures, and ingot preheat treat- region at the pole. The maintenance of a relatively
ments. Formability assessments were made in terms of equiaxed microstructure is inherent to superplasticity,
cone height, with the understanding that thickness uni- while some grain elongation may occur in mixed mode
formity is also indicative of formability. While calcula- deformation processes. Second, the voids observed in
tions of flow stresses from the cone tests (for conditions the more highly strained portion of cone C exhibited
most nearly matching those of the uniaxial tests to angular morphologies, encompassing groups of grains.
failure) showed an overlap with those of the uniaxial Such cavity morphologies have been previously associ-
tests, the strain rates at which the tests were conducted ated with superplastic deformation (region II) from
were not compared due to uncertainty in determining tests of 7475 [9].
the actual strain rate for a cone tests to failure, and the Fig. 15 shows the variation of porosity with strain
fact that the strain rate was definitely not constant for for cone C. The scatter of the data reflects the variation
the uniaxial tests. of porosity through the thickness of the material for a
For the cone tests most closely related to the uniaxial given level of strain. Multiple overlapping micrographs
tests, flow stresses for the failed cones at 540°C (stan- were required to span the cone thickness. Each data
dard preheat material only) were calculated using the point represents the total areal fraction of voids in one
equations of Mackay et al. [32] and Goforth et al. [38]. micrograph. (There was no clear relationship between
The flow stresses for the cone tests were compared with the level of porosity at a given location and its depth
the stresses calculated from the maximum load and below the surface of the cone.) The average volume
112 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113

fraction of voids remained at 2 – 3% for true strains of recrystallization treatment are applied to the low-
up to 1.4. (Recall that cone C was formed at 540°C, temperature preheat material, the resulting grain
with a 425 psi (2.93 MPa) backpressure.) Examination structure is finer but less equiaxed than the refined
of a section of the cone near the crown showed that standard preheat material. Grain diameters for the
failure was probably due to the interlinkage of cavities. low-temperature preheat material average  9.5
The level of backpressure used was apparently enough mm, while grain aspect ratios average 2.0. The
to suppress but not eliminate cavitation as compared grain size is statically stable. The increase in grain
with the uniaxial results. An increase in the level of aspect ratio is attributed to the increased effective-
backpressure above 425 psi (2.93 MPa) might improve ness of dispersoids in the low-temperature preheat
forming performance in terms of cavitation, strain, and material, allowing recrystallization nuclei which can
cone height. overcome the drag of the dispersoids to consume
For the second group of cone tests, which were nuclei which cannot.
statically recrystallized prior to insertion in the cone 4. The finest grained material produced by process A,
press, the conditions of forming were chosen to bracket rolling and static recrystallization exhibits super-
those of cone C. In addition, tests which were previ- plasticity above 500°C. The strain rate sensitivity
ously interrupted after 1 h were allowed to form until reached a maximum of 0.5 for strain rates of 2×
failure occurred. Of these, the tallest cones were N and 10 − 4 to 5×10 − 4 s − 1 at 540°C. The corresponding
O. Cones N and O were formed using 50 psi (0.34 maximum uniaxial elongation, occurring for a strain
MPa) differential pressure (500 psi (3.45 MPa) forward rate of 5× 10 − 4 s − 1, was 375% with a flow stress of
pressure minus 450 psi (3.10 MPa) backpressure). Al- 680 psi (4.7 MPa). Failure of the sample under
though the h/r ratios for cones N and O were less than uniaxial tension was due to cavitation damage.
8.7% higher than that of cone C, cones N and O took Porosity reached 14–20% near midgage length.
four times longer to form than cone C. While the grain 5. The strain rate sensitivity, flow stress, and maximum
size was not actually measured for cones N and O, an
elongation achieved by applying the current ther-
ECC image of cone N at a location where the true
momechanical process to 6013 Variant 1 represent a
strain was near 1.7 (very near the pole) showed that
substantial improvement over that of Baseline 6013
large, elongated grains existed near this region. Grain
refined using a previous processing method.
growth in this sample due to extended exposure to the
6. (a) During cone forming tests with backpressure,
forming temperature may have lead to an increase in
material from process A, rolled and recrystallized
flow stress and a decrease in formability as testing
in-situ in the cone press and deformed at a strain
progressed.
rate near 1× 10 − 3 s − 1, produced a cone with h/r=
1.15 and true strain (at the crown) of 1.7. (This cone
5. Conclusions was formed with 500 psi (3.44 MPa) forward pres-
sure and 425 psi (2.93 MPa) backpressure at 540°C.
The following conclusions can be drawn with respect It was the tallest produced in less than 1 h (cone
to the thermomechanical processing of Baseline 6013 C).) Failure was due to cavitation damage. (b) For
and 6013 Variant 1 for grain refinement by particle statically recrystallized material, deformation at a
stimulated nucleation of recrystallization. Additional strain rate of 3.2× 10 − 4 s − 1 produced a cone
conclusions relate to the superplastic behavior of fine- with h/r=1.25 and a true strain of 2.0. Failure was
grained 6013 Variant 1. due to cavitation damage.
1. Three processes were developed (A, B, and C), each 7. The dependence of cone height on strain rate and
of which produced a uniform distribution of mi- forming time was weak for strain rates between
cron-sized particles. The optimal over-aging treat- 2.5× 10 − 4 and 3.7× 10 − 3 s − 1, and forming times
ment was chosen based on the characteristics of the between 10 and 110 min. This is probably due to the
recrystallized grain structure it ultimately produced. narrow strain rate range investigated. Dynamic
2. When the over-aged structure produced by process grain growth may also have occurred at the slower
A is subsequently rolled and statically recrystallized, strain rates/longer forming times. Dynamic grain
the resulting grain structure consists of equiaxed growth leads to increased flow stress and decreased
grains whose diameters average  10 mm and whose formability. Optimal formability thus arises from a
aspect ratios average 1.6. The grain size is statically balance between strain rate and length of exposure
stable. The texture is weak, although there is a to the forming temperature.
retention of some components of the deformation 8. For the tallest cone produced in less than 1 h (cone
texture, even after prolonged exposure to high C), the volume fraction of voids increased with
temperature. strain up to approximately 3% porosity for a true
3. When process A and the optimized rolling and strain of 1.4. The application of backpressure re-
 L.P. Troeger, E.A. Starke / Materials Science and Engineering A277 (2000) 102–113 113

duced the level of porosity significantly below that [15] U. Köster, Recrystallization involving a second phase, Met. Sci.
8 (1974) 151 – 160.
of the uniaxial samples.
[16] F.J. Humphreys, The nucleation of recrystallization at second
9. For the tallest cone produced in less than 1 h (cone phase particles in deformed aluminium, Acta Metall. 25 (1977)
C), the grain size increased with strain for a net 1323 – 1344.
increase of about 4 mm from the undeformed condi- [17] F.J. Humphreys, Particle Stimulated Nucleation of Recrystallisa-
tion up to true strains near 1.7. The absence of tion, in: T. Chandra (Ed.), Recrystallization ‘90, 1990, pp.
113 – 122.
significant grain growth is likely due to the presence
[18] F.J. Humphreys, M. Hatherly, Recrystallization and Related
of dispersoid particles and the short forming time, Annealing Phenomena, Pergamon, Oxford, 1995.
i.e. limited exposure to the high-temperature [19] J.S. Washfold, I.R. Dover, I.J. Polmear, Thermomechanical
environment. processing of an Al – Mg – Si alloy, Met. Forum 8 (1) (1985)
56 – 59.
[20] E. Kovács-Csetenyi, T. Torma, T. Turmezey, N.Q. Chinh, A.
Acknowledgements Juhász, I. Kovács, Superplasticity of AlMgSi alloys, J. Mater. Sci.
27 (1992) 6141 – 6145.
[21] Y.H. Chung, L.P. Troeger, E.A. Starke, Jr., Grain refining and
This project was supported by the NASA Graduate superplastic forming of aluminum alloy 6013, Proceedings of the
Student Researchers Program and the Zonta Interna- Fourth International Conference on Aluminum Alloys, I, 1994,
tional Amelia Earhart Fellowship Program. pp. 434 – 442.
[22] B. Thanaboonsombut, T.H. Sanders, Jr., A review of the physical
metallurgy of 6013, Proceedings of the 4th International Confer-
References ence on Aluminum Alloys, Vol. III, 1994, pp. 197 –201.
[23] ALCOA Aerospace Technical Fact Sheet: Alloy 6013 Sheet,
Bettendorf, Iowa.
[1] A.K. Vasudevan, R.D. Doherty (Eds.), Preface to Aluminum
[24] ALCOA, Aluminum Alloy 6013, Alloy Digest, Orange, NJ,
Alloys: Contemporary Research and Applications, Treatise on
September 1987.
Materials Science and Technology, Vol. 31, Academic Press,
[25] J.R. Davis (ed.), Aluminum and Aluminum Alloys, ASM Interna-
New York, 1989.
tional, Materials Park, OH, 1993.
[2] J.E. Hatch, Aluminum: Properties and Physical Metallurgy,
[26] B. Thanaboonsombut, T.H. Sanders, Jr., The effects of modifica-
American Society for Metals, 1984.
tions to composition and processing on the recrystallization
[3] J.W. Edington, K.N. Melton, C.P. Cutler, Superplasticity, Progr.
resistance of 6013, Proceedings of the Fourth International
Mat. Sci. 21 (2) (1976) 61–158.
Conference on Aluminum Alloys, Vol. III, 1994, pp. 202–215.
[4] T.G. Langdon, Experimental observations in superplasticity, Su-
[27] E.E. Underwood, Quantitative Stereology, Addison-Wesley Pub-
perplastic Forming of Structural Alloys, The Metallurgical Soci- lishing Co., Reading, MA, 1970, pp. 82 – 83.
ety of AIME, Warrendale, PA, 1982, pp. 27–40. [28] L.P.Troeger, E.A. Starke, Jr., Mater. Sci. Eng. (to be submitted).
[5] J. Pilling, N. Ridley, Cavitation in aluminium alloys during [29] D.J. Dingley, V. Randle, Microtexture determination by electron
superplastic flow, in: H.C. Heikkenen, T.R. NcNelley (Eds.), back-scatter diffraction, J. Mater. Sci. 27 (1992) 4545–4566.
Superplasticity in Aerospace, The Minerals, Metals and Materi- [30] B.L. Adams, S.I. Wright, K. Kunze, Orientation imaging: the
als Society, Warrendale, PA, 1988, pp. 183–198. emergence of a new microscopy, Met. Trans. A. 24A (1993)
[6] C.C. Bampton, J. Wadsworth, A.K. Ghosh, Superplastic alu- 819 – 831.
minum alloys, in: A.K. Vasudevan, R.D. Doherty Jr (Eds.), [31] G. Gottstein, O. Engler, Local texture measurements with the
Aluminum Alloys—Contemporary Research and Applications. scanning electron microscope, J. Phys. IV 3 (1993) 2137–2142.
In: Treatise on Materials Science and Technology, vol. 31, [32] T.L. Mackay, S.M.L. Sastry, C.F. Yolton, Metallurgical charac-
Academic Press, Inc, New York, 1989, pp. 189–216. terization of superplastic forming, AFWAL-TR-80-4038, Septem-
[7] J.C. Williams, E.A. Starke Jr, The role of thermomechanical ber 1980.
processing in tailoring the properties of aluminum and titanium [33] R.J. Lederich, S.M.L. Sastry, M. Hayase, T.L. Mackay, Super-
alloys, in: Deformation, Processing, and Structure, ASM, Metals plastic formability testing, J. Met. 34 (1982) 16 – 20.
Park, OH, 1984, pp. 279–354. [34] O. Engler, K. Lücke, Mechanism of recrystallization texture
[8] C.H. Hamilton, A.K. Ghosh, J.A. Wert, Superplasticity in engi- formation in aluminum alloys, Scr. Metall. 27 (11) (1992) 1527–
neering: a review, Met. Forum 8 (4) (1985) 172–190. 1532.
[9] J. Pilling, N. Ridley, Superplasticity in Crystalline Solids, The [35] R. Ørsund, E. Nes, A model for the nucleation of recrystallization
Institute of Metals, London, England, 1989. from particles: the texture aspect, Scr. Metall. 22 (1988) 671–676.
[10] M.A. Zaidi, J.A. Wert, Thermomechanical processing of alu- [36] S.P. Agrawal, M.L. Rommel, M.T. Hahn, Design, development,
minum alloys, in: A.K. Vasudevan, R.D. Doherty Jr (Eds.), and characterization of advanced aluminum structural alloys for
Aluminum Alloys—Contemporary Research and Applications. enhanced superplasticity, Final Technical Report, Contract No.
In: Treatise on Materials Science and Technology, vol. 31, NAS1-17623, NASA Langley Research Center, Hampton, VA,
Academic Press, Inc, New York, 1989, pp. 137–167. 1991.
[11] T.G. Nieh, J. Wadsworth, O.D. Sherby, Superplasticity in [37] Z.X. Guo, J. Pilling, N. Ridley, Bulge-forming of domes: a
Metals and Ceramics, Cambridge Solid State Science Series, comparison of theoretical prediction and experiment, in: C.H.
Cambridge University Press, Cambridge, 1997. Hamilton, N.E. Paton (Eds.), Superplasticity and Superplastic
[12] J. Waldman, H.V. Sulinski, H. Markus, US Patent No. Forming, The Minerals, Metals and Materials Society, Warren-
3,847,681, 12 November 1974. dale, PA, 1988, pp. 303 – 308.
[13] N.E. Paton, C.H. Hamilton, US Patent No. 4,092,181, 30 May [38] R.E. Goforth, N.A. Chandra, D. George, Analysis of the cone
1978. test to evaluate superplastic forming characteristics of sheet
[14] W.A. Cassada, G.J. Shiflet, E.A. Starke Jr, Mechanism of metal, in: H.C. Heikkenen, T.R. McNelley (Eds.), Superplasticity
Al2CuLi (T1) nucleation and growth, Met. Trans. A. 22A (1991) in Aerospace, The Minerals, Metals and Materials Society, War-
287 – 297. rendale, PA, 1988, pp. 149 – 166.