SUPERPLASTIC DEFORMATION OF TITANIUM ALLOYS

A thesis submitted to the University of Surrey

for the degree of Doctor of Philosophy on the basis

ol collaboration between the University Department

of Physics and the Royal Aircraft Establishment

by

Christopher Douglas Ingelbrecht

Materials and Structures Department
Royal Aircraft Establishment
Farnborough, Hampshire, UK

November 1985
 qTl'\iIMARY

The superplastic deformation of three alpha/beta titanium alloys;

Ti-6Al-4V, IMI 550 (Ti-4Al-4Mo-2Sn-0.5Si) and Ti-8Al-1Mo-1V and a beta
alloy; Ti-15V-3Cr-3Al-3Sn was investigated. The alpha/beta alloys
exhibited high strain rate sensitivities and superplastic strain in

these alloys tended to randomise the alpha and beta phase textures and
caused grain growth. The beta alloy showed relatively low strain rate
sensitivity and formed subgrains resulting in grain refinement.

Superplastic deformation reduced room temperature strength. In the

alpha/beta alloys this was mainly the result of recrystallisation at the

forming temperature rather than grain growth associated with the

superplastic strain. The strength of the IMI 550 alloy after forming

was raised by increasing the cooling rate from the forming temperature

and by ageing after forming.

The anisotropy of superplastic deformation, measured in terms of

the strain ratio R, was caused by microstructural directionality and

was not related texture. The R values were also influenced by the test

piece shape.

Uniaxial data were used to predict the optimum gas pressure cycle
for the superplastic forming of a hemisphere from Ti-6Al-4V sheet. The

calculated pressure cycle was found to be significantly different to that

for an isotropic, non-strain hardening material.

The uniaxial data were also used in a computer model of necking in

Ti-6Al-4V during superplastic deformation. The influence of strain

hardening and strain rate hardening were considered and predictions of
limiting strains for various initial neck sizes were made.

The work described in this thesis has shown in particular that:

1 The anisotropy of superplastic deformation in titanium alloys results

from directionality in the microstructure, but that R value measurements

can be influenced by test piece shape.

2 The alpha/beta alloys investigated exhibit relatively high strain

rate sensitivities and are more suitable for superplastic forming than

the beta alloy Ti-15V-3Cr-3Al-3Sn.

 CONTENTS
Page

1 INTRODUCTION 1

3
2 REVIEW

2.1 Models of superplastic deformation 3

2.2 Superplastic deformation of two phase materials 6

2.3 Anisotropy 8

2.4 Superplastic forming of sheet 11

3 EXPERIMENTAL PROCEDURE 15
3.1 Material 15
3.2 Superplastic deformation 15
3.3 Room temperature testing 17
3.4 Microstructural examination 18
3.5 Analysis of necking of Ti-6Al-4V during superplastic 18
deformation

4 RESULTS 20
4.1 Superplastic deformation and room temperature tensile 20

testing of titanium alloy sheet

4.1.1 Ti-6Al-4V 20

4.1.2 IMI 550 (Ti-4Al-4Mo-2Sn-0.5Si) 22

4.1.3 Ti-8Al- 1,Nlo- 1V 23

4.1.4 Ti-15V-3Cr-3Al-3Sn 25

4.2 Superplastic deformation of sheet test pieces machined 26

from Ti-6Al-4V bar

4.2.1 Test piece shape after superplastic strain 26

4.2.1.1 TL orientation 26

4.2.1.2 ST orientation 26
4.2.1.3 LT orientation 26

4.2.2 Microstructure 27

4.2.3 Flow stress 28

4.2.4 Strain rate sensitivity 29

4.2.5 R values 29
4.2.6 Texture 30
4.3 R values of Ti-6AI-4V sheet after superplastic strain 30
4.3.1 Effect of sheet thickness 30
4.3.2 Effect of test piece geometry 31
4.3.3 R values of other alloys 31
4.4 Analý, sis of necking of Ti-6Al-4V during superplastic 32
defornizit ioii
5 DISCUSS10N 34

5.1 Titanium alloy sheet 34

5.1.1 Superplastic deformation and microstructure 34

5.1.2 Room temperature tensile properties after 37

superplastic deformation

5.1.3 Texture 39

5.1.4 Activation energy 40

5.2 Superplastic deformation of sheet test pieces machined 42

from Ti-6Al-4V bar

5.2.1 Microstructure and flow stress 42

5.2.2 Strain rate sensitivity 44
5.2.3 R values 44
5.2.3.1 TL orientation 44
5.2.3.2 ST orientation 46
5.2.3.3 LT orientation 47
5.2.3.4 Comparison with other work 47
5.3 R values of Ti-6Al-4V sheet 48
5.3.1 Effect of sheet thickness 48

5.3.2 Effect of test piece geometry 48

5.4 Application of uniaxial data to hemisphere forming 51

5.5 Analysis of necking of Ti-6Al-4V during superplastic 55

deformation

6 CONCLUSIONS 60

7 SUGGESTIONS FOR FURTHER WORK 62

ACKNO14LEDGEMENTS 63

REFERENCES 70

TABLES

FIGURES
1 INTRODUCTION

The term "superplastic" indicates the ability of a material to

undergo large amounts of essentially neck free strain without failure.

Superplasticity is generally restricted to temperatures greater than

0.4T (1), where TM is the absolute melting temperature, and is promoted
M
by a small grain size and slow grain coarsening at the deformation

temperature. Superplasticity can also be induced in some materials by

cycling through a phase transformation (1), although this "environmental

superplasticity" is not the subject of this thesis.

The logarithmic stress/strain curve for the superplastic

temperature usually has a sigmoidal shape and is conventionally divided
into three regions. In regions I and III, the low and high strain rate

regimes respectively, flow stress is relatively insensitive to strain

rate. However, in region II, where superplasticity is encountered, the

flow stress often increases rapidly with strain rate. Thus, any tendency

of strain localisation during deformation is counteracted by a local

increase in flow stress and necking is avoided. The potential super-

plasticity is usually measured by the strain rate sensitivity m where:

a= Kým. In this equation a is the applied stress, ý is the strain
rate and K is a constant depending on testing conditions. This expression
incorporates no strain hardening and an m value of 1.0 indicates viscous
flow (1). In practice, m values for metals up to about 0.9 can be

obtained and some strain hardening, due to grain growth, may occur.

The two phase Ti-6AI-4V alloy is by far the most widely used high

strength titanium alloy and is highly superplastic, capable of undergoing

tensile elongations of up to about 1000% (1,2). Consequently, the

experimental data on superplasticity of titanium is most extensive for

this alloy and it has been used for almost all of the superplastically
formed titanium components so far produced.

The objectives of this thesis were to examine the characteristics

of superplastic deformation of titanium alloys with emphasis on the
influences of microstructure, texture and test piece shape on the

uniformity of superplastic flow. In a review of the literature on super-

plastic deformation particular attention has been paid to those aspects

most relevant to the experimental work described elsewhere in the thesis.

An inert atmosphere testing rig was designed and uniaxial super-

plastic testing was carried out on Ti-6Al-4V in sheet form and on sheet
test pieces machined from Ti-6Al-4V bar containing a strongly directional

microstructure and a pronounced crystallographic texture. Three other

sheet alloys, for which there was very little data on superplastic

deformation were also investigated viz:

(1) IMI 550 (Ti-4Al-4Mo-2Sn-0.5Si). This is a high strength,

two phase alloy with potential for lower forming temperatures

and significantly higher post-formed room temperature strength

than Ti-6Al-4V.

(2) Ti-Ml-Mo-W. This is a creep resistant, near alpha alloy

with a higher modulus and lower density than most other titanium
alloys.

(3) Ti-15V-3Cr-3Al-3Sn. This is a large grained, metastable

beta alloy that is cold formable in the annealed condition and

can be aged to very high strengths.

Biaxial sheet forming under gas pressure is often carried out

assuming that the material is isotropic and does not strain harden. The

superplastic data derived from the uniaxial tests on Ti-6Al-4V have been

used to estimate the effects of anisotropy, strain hardening and strain

rate sensitivity on the superplastic forming of a hemisphere and to

predict the optimum pressure cycle. The uniaxial data has also been

used to model the necking characteristics of Ti-6Al-4V during superplastic

tensile deformation.

The experimental results are discussed in detail. The superplastic

deformation in each of the alloys is compared and the alloys are assessed
for superplastic formability. The causes of anisotropy of superplastic
deformation are considered. Conclusions arising from the work are
presented and several areas of possible future work are suggested.

2
2 REVIEW

2.1 Models of superplastic deformation

A variety of mechanisms have been proposed for superplastic

deformation including slip, diffusion and dislocation creep and grain
boundary sliding with various accommodation processes. There is

evidence to support many of the models. However, no single theory

comes close to explaining all of the experimental observations, which is

not surprising in view of the wide range of single and multiphase

materials exhibiting superplasticity (1-4).

Diffusion creep processes may involve lattice diffusion

(Herring-Nabarro (5) creep, grain boundary diffusion (Coble (6) creep)

or a combination of diffusion paths (7). Both Herring-Nabarro and Coble

creep predict a strain rate sensitivity m. = 1, where m. = dlna/dlný, but

a different grain size dependence of strain rate. Diffusion creep also

predicts grain elongation (not usually observed during superplastic

deformation), strain rates usually several orders of magnitude too low

and the retention of texture after superplastic strain. However, there

is some evidence that diffusion creep plays a role in the superplastic
deformation of some materials (2) and Griffiths and Hammond (8) predicted
the superplastic strain rates of two large grained beta titanium alloys

and beta brass with reasonable accuracy using the Herring-Nabarro creep
formula.

Models based on slip alone are not satisfactory (2), because they
do not predict an equiaxed grain structure after superplastic strain and

cannot explain the randomisation of texture or the low rate of work

hardening. There is evidence (see section 2.3) that slip can play a

role in the superplastic deformation of some alloys. However, in all these

cases recovery, recrystallisation or grain boundary migration are assumed

to take place. According to Naziri and Pearce (9,10) and Schmidt-Whitley
(11,12) superplasticity occurs only when the observed flow stress would

predict a dislocation cell size greater than the grain size. Thus, subgrains

are not observed during superplastic deformation. Hayden et al (13,14)

formulated a dislocation climb model based on experimental observations

of superplasticity in Ni-Cr-Fe alloys in which a transition occurred from

essentially dislocation free grains at low stress to cell formation and

dislocation tangling at higher stresses.

3
 Chaudhari (15) produced a model of superplastic flow in Zn-Al
based on the motion of dislocations in internal stress f ields and later ( 16)

suggested a "dislocation cascade" mechanism involving diffusion creep at

the heads of dislocation pile-ups. However, the extensive dislocation

pile-ups predicted by both the models of Chaudhari (15,16) are not usually
observed (1,2) although studies of the role of dislocations in superplastic
deformation are likely to be hindered by dislocation recovery and
disappearance during cooling from the forming temperatures.

The dislocation based models discussed generally seem (1) to be

more applicable to the high stress regime ie region III than to region II

where the highest strain rate sensitivities are exhibited.

It is grain boundary sliding that is usually assumed to play the

dominant role in superplastic deformation. Observations of marker-lines
scratched onto the surface have revealed grain boundary offsets after
superplastic strain in eutectoid Al-Zn (17-19), eutectic Pb-62%Sn (20),

eutectic Mg-Al (21) and Pb-T1 (22). The grain rearrangement in the Pb-Sn

eutectic has also been studied in the scanning electron microscope by

straining in situ (23,24). It has been estimated (1,2,25-27) that grain

boundary sliding contributes 60-80% of the total strain in region II and
considerably less in regions I and III.

Grain boundary sliding cannot occur without some accommodation

process to maintain coherency between grains at edges and triple points.

The model of Ashby and Verrall (28) proposed diffusional accommodation for

grain boundary sliding and introduced the "grain switching" concept,

whereby a group of four grains rearrange themselves producing an axial
strain of 0.55. The model predicts a strain rate of:

3.36D
1000 0.72y B'
D1+ -d -D
kTd 2d
Iv

where a is the tensile stress, Q the atomic volume, d the grain size,

y the grain boundary free energy, 6 the width of the boundary diffusion

path, Dv and DB are the bulk and boundary diffusion coefficients, k is

Boltzmann's constant and T the absolute temperature. The threshold

stress 0.72y/d arises from the high energy transition state in the grain

switching process. The contribution of dislocation creep, which becomes

significant at high stress levels, was considered to be additive. The

Ashby/Verrall model has been applied with varying degrees of success to

4
a Zn-Al alloy (28,29), alpha/beta titanium (30) and a Zr based alloy (31),

and the grain switching process has been observed in practice (32).

However, the grain size dependence given by the model is not consistent

with many of the experimental data (33-35), it predicts a stress exponent

(n = 1/m) tending to unity and it is based on a two dimensional array of
grains, which can deform without any increase in surface area of the

specimen.

A three dimensional grain boundary sliding model with diffusional

accommodation at grain corners has been proposed by Geckinli (36). The

predicted strain rate is:

1.25D 60d 3.
B 22.7y)
CF- d
K2 kT
v

where Kv is the volume of matter diffused for each grain for sliding to
d3
occur and the other variables are as defined above. Since Ka
v
this
-3
predicts ad grain size dependence of strain rate, which agrees with
results on Pb-Sn eutectic (23) and Sn-5%Bi (37). However, the stress
dependence strain ýa (a -a0), is
of rate which gives m1 not

commonly observed. '

Gifkins (38) suggested a "core and mantle" model, whereby each grain

consists of a core region surrounded by a deformable mantle of variable

width. At low stress (region I) diffusion creep takes place with diffusion

confined to a relatively narrow mantle. In region II the mantle deforms by

dislocation climb and glide and at higher stress extensive slip occurs in

the core and subgrains may form. This model was subsequently extended (22)

to three dimensions to allow grains from one layer to slide between two

others in an adjacent layer. Good agreement with experiment was

claimed (22).

Beere (39) proposed a deforming mantle model for cubic grains based

on diffusion creep. It was shown (39) that if one set of interfaces has a

greater sliding resistance than another then grain rotation is inevitable.

Ball and Hutchison (19) working on eutectoid Zn-Al suggested that

groups of grains slide as units until they are blocked by a grain of

unfavourable orientation. Dislocations generated in this grain pile up at

the opposite grain boundary and eventually climb into the boundary itself.

5
This led to the rate equation:

AD
BG2b2 (
E= -1ý-T ý -D -j

where A is a constant.

According to Mukherjee (40) dislocations are generated by grains

sliding individually rather than in groups and encountering ledges or

other protrusions in adjacent grains. The rate of sliding is then
controlled by the climb rate of dislocations into the boundary. This
leads to a rate equation similar to that of Ball and Hutchison given
above and gives a reasonable correlation with data on Zn-Al eutectoid
(40).

Gittus (41) proposed a theory of superplastic flow involving

grain switching specifically for two phase materials. According to the

model sliding occurs by dislocations gliding in the interphase boundaries

and piling up at triple edges before climbing away into disordered

segments of the interphase boundary.

2.2 Superplastic deformation of two phase materials

Models of superpl-istic deformation in two phase alloys are likely

to be complicated by quite different properties in the two phases and the

presence of interphase as well as intergranular boundaries. Titanium

alloys are likely to represent an extreme case in view of the relatively

high lattice diffusion rates in the beta phase; D /D = 100-1000 in the
CL
superplastic temperature range (42). Hamilton et al (30) produced, by

hydrogenation, phase proportions from 40-100% beta in Ti-6Al-4V at 870'C

and found that flow stress at 870'C increased with decreasing beta phase
content, although this comparison was complicated by rapid grain growth
in the near beta alloys at low strain rates, which tended to increase the
flow stress. The highest flow stresses were recorded for an alpha alloy
Ti-6A1. Similarly, Sastry et al (43) measured flow stresses at 900'C in

an alpha titanium alloy (Ti-5Al-2.5Sn), a beta alloy (Ti-15V-3Cr-3Al-3Sn)

and an alpha/beta alloy (Ti-6Al-4V). The alpha phase alloy was found to
have the highest flow stress and the alpha/beta alloy the lowest, although
the results were again influenced by excessive grain growth in the single

phase alloys.

6
 The superplastic properties of each phase can be studied

separately using single phase alpha or beta alloys as indicated above.

However, the problem remains of combining the results to predict the
behaviour in two phase materials. Hamilton et al (30) stated that the
Ashby/Verrall model (28) of the diffusion accommodated grain boundarv

sliding could be used to calculate the a/ý relationship for Ti-6Al-4V

from the data for the single phase materials using a rule of mixtures
approach and by assuming the same strain rate in each phase. According

to this "isostrain rate" model the strain rate in the alloy is essentially
determined by the deformation characteristics of the harder (alpha) phase.
However, this is not consistent with other work on titanium alloys (44-46),

alpha/beta brass (47,48) and the zirconium based "Zircaloy" (31), where
it was suggested that the deformation is largely restricted to the beta

phase, which tends to form a continuous matrix and behaves like a deforming

mantle around the alph4, grains. This type of behaviour is better

described by the term "isostress", with the two phases deforming at

different rates. Similarly, Springarn and Nix (49) proposed a model in

which the faster diffusing beta phase collects at alpha/alpha boundaries

and pinches off the alpha grains as observed by Naziri et al (50) in

Zn-22%Al and by Hidalgo-Prada and Mukherjee (46) in Ti-6Al-4V-2Ni.

Boundary sliding measurements on a Pb-62%Sn alloy (20) and on

Zn-22%Al (17) both revealed that the sliding rate for each type of
interface depended on the value of 6D where 6 is the boundary width
9,
and D9 is the coefficient of boundary diffusion. This was the basis for

a deformation model (51) for anisotropic superplasticity in two phase

alloys, which accounted for the break-up of banded microstructures and

predicted a flow softening by the progressive and irreversible conversion

of alpha/alpha interfaces into beta/beta interfaces.

Dunlop et al (52) observed that cavitation in an aluminium bronze

(Cu-9.5Al-4Fe) occurred mainly at alpha/beta boundaries and suggested
that sliding was concentrated at these boundaries as a result of the

strong texture, which indicated a relatively low misorientation across

alpha/alpha and beta/beta boundaries with a correspondingly high sliding

resistance. Similarly, Chandra et al (53) found that the sliding rate in

alpha/beta brass decreased in the order and that cavitation

occurred predominantly at alpha/beta interfaces. It was concluded that

7
cavitation was inhibited in the beta/beta boundaries by plastic flow in

the relatively soft beta phase and Patterson and Ridley (54) showed that
void content decreased with increasing beta phase proportion at 600% in

alpha/beta brasses with slightly differing compositions.

Attempts have been made (44-46,55,56) to improve the superplastic

properties of Ti-6Al-4V by small additions of Ni, Co or Fe to increase

the beta phase proportion and to increase effective diffusion rates in

the beta phase. Both Ni and Co have tracer diffusivities about two

orders of magnitude higher than that of titanium in beta titanium and

it is reasoned (57) that, because gross solute segregation can be avoided
by migration of the phase interfaces, the diffusivity can be controlled
by the faster diffusing species even at concentrations of only a few
percent, the effective diffusion rate in a two component system obeying:
D ýCBDA+CAD where CA and CB are the atomic fractions of
eff B'
components A and B and DA and DB are the tracer diffusivities. The

modified alloys were shown to have similar properties to the base alloy,
but at temperatures typically 100*C lower. Some improvement in room
temperature strength due to solid solution hardening was also recorded (55).

It was shown (44) that the fast diffusing species ie Co or Ni tended to

segregate to the boundaries perpendicular to the tensile axis during

superplastic strain and that longitudinal boundaries were depleted.

According to Ma et al (58) the reverse is true for the relatively slow

diffusing element Mo in IMI 550.

2.3 Superplastic anisotropy

Anisotropic superplastic properties have been reported for a wide
range of alloys (1,2,51,59-79) and have generally been ascribed either to
the effects of pre-existing crystallographic texture, inferring slip
controlled deformation, or to mechanical fibreing or grain elongation,
in which case grain boundary sliding is considered to be dominant. There
is substantial evidence for both, although in many cases either texture

or microstructure are discussed in isolation and the results are not

completely unambiguous.

In general (1,2,25,32), both microstructural directionality and

texture intensity are reduced by superplastic strain, although when

substantial amounts of slip occur the texture change may involve a systematic

8
rotation (69) or the stabilisation of certain texture components (76,80-82)

and in two phase alloys the effect of superplastic strain on texture may
be different in each phase (32,75,80-83). However, even when superplastic
anisotropy appears to be texture related, it is not suggested that slip
is the only deformation mode in operation. Superplastic deformation does

not result in grain elongation normally associated with slip at relatively

low temperatures and high strain rates. To take account of this, additional
processes such as recrystallisation, recovery or grain boundary migration
are assumed to take place. It is usually accepted (75,76,78,80-82) that
texture randomisation is associated with grain boundary sliding, whereas
the stabilisation or even intensification of some texture components may
be due to slip (76,81,82,84-86) or anisotropic diffusion (80).

Packer et al (66) working on eutectic Zn-Al noticed that round

tensile specimens developed elliptical cross-sections during superplastic
strain. This was accompanied by both a reduction in texture intensity

and the microstructure becoming more equiaxed. However, the elliptical

specimens were re-machined to a round cross-section and were still found

to behave anisotropically after the microstructural directionality had

been removed. It was concluded that the strain anisotropy was the result
of texture in the zinc rich phase and that the deformation was occurring
largely by slip, although concurrent grain boundary migration or
recrystallisation was invoked to account for the fact that grain elongation
did not occur. Similarly Johnson et al (67) produced spheroidised
grains in both eutectic and eutectoid Zn-Al by hot rolling, which also
resulted in a strong texture. This material behaved anisotropically.
However, in material quenched to produce an equiaxed grain structure,
but with a random texture, the superplastic deformation was isotropic.

Nuttall also reported (72) isotropic superplastic properties in quenched

Zn-Al eutectoid alloy.

The anisotropy in Zn-0.4%Al sheet was recorded (68,69) in terms of

the plastic strain ratio R, defined as the ratio of width to thickness

strain. The R values were found to increase towards 1.0 with superplastic
strain and strain rate sensitivity became more isotropic (71). The

deformation at superplastic strain rates was apparently slip controlled

with simultaneous recovery occurring, although the microstructure was not

studied. Heubner et al(70), working on eutectoid Zn-Al, also noticed R

9
values increasing with superplastic strain and related this to decreasing

texture intensity. Kaibyshev et al (87) reported texture intensities

decreasing with superplastic strain in a range of Zn-Al alloys and also

noticed a transition from single to multiple slip in the Zn rich phase with
increasing strain rate in region II. Investigations of superplastic
deformation in Al-Li alloys (61,79), Supral 220 (61) and the aluminium
alloy 7475 (61) also revealed anisotropy decreasing with strain.

According to Kaibyshev et al (83) the presence of a strong rather

than a random texture in Zn-22% Al with equiaxed grains resulted in a
decrease in superplastic flow stress, an increase in elongation and in the

strain rate corresponding to maximum m, although no anisotropy was observed

in the flow stress at the superplastic temperature. It was suggested that
the strong texture increased the rate of grain boundary sliding and
facilitated the climb or glide of dislocations into the boundaries. Thus

there are inconsistencies between this work and other results on Al alloys,
which generally indicate some degree of texture controlled anisotropy.

Superplastic anisotropy has also been investigated in Pb-Sn

eutectoid alloy (73-75). The evidence consistently suggests that, for

this material, microstructural banding or grain elongation are more

important than texture effects. In the material used by Melton et al (73)

the initial texture was weak and was practically unchanged by superplastic

strain. Break-up of the banded microstructure with strain was said to

explain the increasingly isotropic behaviour and compression testing

confirmed that the flow stress was highest parallel to the rolling
direction. These results are consistent with those of Kashyap and Murty
(74) who investigated Pb-Sn both with grains elongated in one direction by

extrusion and swaging and grains elongated in two directions by rolling.

Texture was not discussed in this work. However, Cutler and Edington (75)

found that textures in both the Pb and Sn rich phases were slowly

randomised by superplastic strain, consistent with grain boundary sliding

rather than slip. Similar observations of texture randomisation were made

on aluminium bronze by Dunlop et al (78).

Bricknell and Edington used detailed texture measurements in the

form of crystallite orientation distribution functions (77) to predict
(76) the effect of superplastic strain on R value of Al-6Cu-0.3Zr based on

10
 [111) <110> The predictions in
single and multiple slip. were good
agreement with experimental data for high strain rates, but at lower

rates the predictions were increasingly inaccurate suggesting a gradual

transition to grain boundary sliding. The increase of R value with

strain was ascribed to the development of a <111> fibre texture at high

strain rate and to the break-up of microstructural banding at lower rates.

The most thorough made so far of superplastic

analysis anisotropy
in titanium alloys is by McDarmaid et al (51,59-62,64,65) working on
strongly banded Ti-6AI-4V with a pronounced texture. A gradual randomisa-
tion of alpha phase texture was found with anisotropy apparently arising
from contiguous alpha grains aligned in the rolling direction. However,
at relatively high strain or low temperatures
rates a transition to slip
controlled deformation deformation occurred (61,64,65). At a temperature
close to the beta transus when the alpha phase content was small the
deformation was nearly isotropic and was not influenced by strain rate.
Paton and Hamilton investigated (88) Ti-6Al-4V sheet containing similar
microstructural banding and found anisotropic flow stress and strain
rate sensitivity. Russian data (63) on the VT6 alloy Ti-6.5Al-5. lV are
apparently contradictory in that superplastic anisotropy was reported in

material with nominally equiaxed grains. This was said to result from the
strong texture present. However, the sense of the anisotropy and its

reversal at lower temperatures were the same as reported for the banded
Ti-6Al-4V (60-62,64,65) suggesting that microstructure rather than
texture may have been responsible.

2.4 Superplastic forming of sheet

Despite the wide range of materials in which superplasticity has

been investigated (1-4), the commercial exploitation of the phenomenon has
concentrated on the sheet forming of aluminium. (89-91) and titanium (89-95)

alloys with microduplex stainless steel (96) and eutectoid Zn-22%Al

(91,92,97) occasionally used.

Alloys based on Al-Ca and Al-Cu-Zr (Supral) have been specially

developed for superplastic forming and a wide range of Supral components
from ejector seat head boxes to machine covers, cladding for internal and

external walls and roofs and even car body parts have been produced. For

higher strength applications Si, Mg and Ge additions have been made to

the Supral material and thermomechanical grain refining treatments have

been developed (98-101) for the high strength7000 series alloys, which
improve the superplastic properties. All the aluminium alloys suffer
11
from cavitation during superplastic strain (79,102,103) which degrades

the service properties, although forming with a superimposed hydrostatic

pressure can alleviate the problem (104).

The titanium alloy Ti-6Al-4V is highly superplastic and unlike

aluminium alloys does not cavitate significantly during superplastic
forming. It also readily diffusion bonds at temperatures in the

superplastic range so that forming and bonding can often be carried out
in the same operation and complicated, internally stiffened components

can be produced from a lay-up of several sheets. Superplastic forming

or the combined superplastic forming/diffusion bonding (SPF/DB)

fabrication processes have already been successfully applied to a number

of demonstration and production aerospace components with cost savings

up to 50% and weight savings up to 30% (95) compared with conventionally

produced titanium parts. These savings are achieved because fewer

fasteners are required, much machining is eliminated and more structurally

efficient designs are possible. Titanium SPF/DB components can also

economically replace aluminium or steel parts in some cases.

For superplastic forming of the aluminium. Supral alloys aluminium

tools can be used, whereas steel tools are required for the 7000 series
r-a-M.
alloys, which, have higher superplastic flow stresses. Titanium alloy
forming requires much more expensive cast or machined steel or nickel
based dies capable of withstanding temperatures up to about 950'C.

A variety of superplastic forming techniques for aluminium. alloys

have been investigated (89,91) in order to minimise the thickness variation

of the forming and to reduce material wastage. Simple gas pressure blowing
into a female mould results in a decrease in thickness from the edge of the
forming to the corners, which contact the die wall last. Drape forming

over a male die sitting in a cavity improves the thickness distribution,

but leads to some material wastage. Reverse billowing is another technique

used to reduce thickness variation. In this process the sheet is initially

free-blown away from the mould by gas pressure to a height greater than

the mould depth. The pressure is then reversed and the sheet blown into

the mould so that the peak of the dome is the first part of the forming to

touch the mould surface. For relatively deep formings it is usual to use

male tools moving into the sheet. In plug assisted forming a prestretch
is carried out using a moving tool and then pressure is used to complete

the forming into a female mould.

12
 Titanium sheet formings are almost always carried out using argon
gas pressure and a female mould. When diffusion bonding is also required
then a lay-up or pack of several sheets is assembled beforehand with a
stop-off compound, usually yttria, applied where bonding is not required.
The sheets may also be profiled by chemical milling before forming or be

stamped out, depending on the reinforcement or the geometry of the

internal stiffening required. Diffusion bonding may be by platen pressure
before or during the forming operation or by gas pressure. Joining by
TIG welding before inflation of the pack is also occasionally carried
out.

Resistance to neck growth during plastic deformation at room

temperature is largely determined by the rate of work hardening with the

strain rate sensitivity m becoming important in the post uniform

stage (105). The normal anisotropy, measured by the R value can also
significantly affect sheet formability. Sheet drawability (106), for

example, is enhanced by high R values as the potential failure site at

the bottom of the cup wall, where plane strain deformation occurs, is

effectively strengthened, whereas the flange area (pure shear), near the

cup rim, is effectively weakened. The influence of R on limiting strain

(the strain at which localised necking begins) also depends on the stress

state. In uniaxial tension increasing the R value delays the

development of the plane strain deformation that is required for

localised necking (105) and thus increases the limiting strain. This is

true for all cases where the minor strain e2 is negative (107,108).

However, under plane strain conditions (T05,109) the limiting strain is

independent of R and for positive minor strains (E >0) the limiting
2
strain decreases with R value. Strongly basal textured titanium alloys

can have R values up to 12 (105,107).

The deformation limits for cold forming are determined by minimum

bend radii and by strain and strain ratio, which are considered
together in the forming limit diagram. Failure by localised necking or
fracture usually occurs if the limits are exceeded. Each of the

processes of drawing, bending, or stretching over a punch have different

constraints and friction considerations associated with them and

different mechanical properties may be critical in each case.

Superplastic forming differs from cold sheet forming in a number

The is usually assumed to be the most important factor
of ways. m value

13
in promoting large elongations and the material may be processed
beforehand to give a fine, stable grain size for this reason. The R
value has the same influence on superplastic formings as on cold sheet
formings. However, superplastic deformation is not usually influenced by
texture and the normal anisotropy under superplastic conditions is

unlikely to be as pronounced as that observed during room temperature

deformation.

Superplastic sheet forming is normally carried out by gas pressure

blowing such that the stress state is biaxial tension. This mostly
precludes localised necking (106) and even where localised necking could
occur such as in corners or in long trough-shaped formings, the material
is usually sufficiently strain rate sensitive to avoid localised

necking and diffuse necking only will occur. Thus, failure of non-
cavitating material during superplastic forming is rarely a problem.

14
 EXPERIMENTAL PROCEDURE

3.1 Material

The Ti-6Al-4V sheet was supplied by Reactive Metals Inc. in four

thicknesses: 0.9,1.8,2.0 (batch A) and 3.3mm. A further two batches

(B and C) of 2. Omm Ti-6Al-4V sheet and 2. Omm IMI 550 sheet
(Ti-4Al-4Mo---2Sn-0.5Si) were obtained from IMI Titanium. All of the
Ti-6Al-4V and IMI 550 sheet had been cross rolled and annealed at 700'C

for 2 hr. The Ti-6Al-4V bar 160 x 55mm was also supplied by IMI and was
annealed for 4hr at 700'C before machining. The 2. Omm Ti-8Al-1Mo-1V
(duplex annealed: 790*C 8hr, furnace cooled, 790*C 1/4 hr, air cooled)
and the 2. Omm Ti-15V-3Cr-3Al-3Sn (annealed 790*C, 2hr) were supplied by

Timet.

3.2 Superplastic deformation

Sheet test pieces were machined with the final rolling direction

parallel to the tension axis W and perpendicular to the tension axis (T).

The test piece heads were reinforced by platos of 2mm thick Ti-6Al-4V

sheet spot welded onto each side in order to minimise head distortion

during superplastic testing. A diagram of the test piece is given in

Fig 1. For the determination of superplastic properties of sheet

material (section 3.1) a gauge length (1 of 25mm and a gauge width

0)
(w of 16mm was used, except for the Ti-8Al-lMo-lV tests (w 12mm).
0) 0=
For the series of tests investigating the effect of test piece shape on
R value L orientation test pieces of 3.3mm thick Ti-6Al-4V sheet were

used either (a) with a gauge length of 10mm and gauge widths of 4,8 16 or

19.5mm or (b) with a gauge width of 16mm and gauge lengths of 2,5,10 or
25mm as shown in Fig 1.

Sheet test pieces were machined from the 55mm thick Ti-6Al-4V

bar in three orientations; TL, ST and LT where the first letter

indicates the tensile axis of the test piece and the second letter

refers to the orthogonal direction in the plane of the test piece and
L, T and S are the longitudinal, transverse and short transverse
directions of the original bar. A cutting diagram is shown in Fig 2.
The bar thickness was insufficient for the extraction of the ST

orientation test pieces. Therefore, 4mm thick blanks for the gauge
length were machined from the bar in the ST orientation and extended by

electron beam welding pieces of 4mm thick Ti-6Al-4V sheet to each end.

15
Final machining of the test pieces to 2mm thickness was then carried

out with a gauge length of 25mm and agauge width of 12mm.

C2
Uniaxial superplastic straining was carried out on a screw
driven tensile testing machine with the test piece enclosed in a quartz
tube sealed at each end by water cooled brass caps. A gas tight seal was
maintained by heat resistant O-ring seals around the pull rods and an
inert atmosphere was provided by a slow flow of argon, which was passed
over titanium swarf at 900*C in a separate furnace in order to getter
the oxygen before entering the testing chamber. The pull rods, test

piece grips and loading pins were made from nimonic 75 and were assembled
before each test with graphite flakes on the threads to prevent sticking.
A vertical three zone furnace provided a hot zone to ± 2'C over a length
of 100mm. A series of circular stainless steel heat shields push-fitted
onto the pull rods inside the quartz tube both above and below the test

piece helped to maintain the hot zone and the furnace was lagged top and
bottom to reduce heat loss. Temperature was monitored by three thermo-

couples which passed through the top cap and upper heat shields and were
wired onto the test piece. Cross-head displacement was measured by an
integral digital extensometer and load, extension and temperature were
recorded on a moving chart. The testing apparatus is shown in Fig 3.

The test temperatures used for the rolled sheet test pieces were:
Ti-6Al-4V: 925'C, IMI 550: 900*C, Ti-8AI-lMo-IV: 910,940,970 and
1010'C, Ti-15V-3Cr-3AI-3Sn: 810,860 and 910'C. A strain rate of
3x 10-4s-l was used for the majority of the tests-with some tests on the
3.3mm Ti-6AI-4V sheet carried out at 10-3s-1 and 2.5 x 10-3s-1. For the
Ti-6Al-4V test pieces machined from bar four test temperatures of 800,
875,925 and 975'C and two strain rates of 3x 10-4s-1 and 1.5 x 10-3s-1

were employed.

The strain rate was maintained nominally constant by increasing

the cross-head speed as appropriate after each 50% strain increment.

Flow stress and strain rate sensitivity were determined for the 3.3mm

Ti-6Al-4V sheet for up to 400% superplastic strain by interrupting the

test after 200% strain and remachining the gauge length parallel. Flow

stress and m value data for the other Ti-6Al-4V sheet thicknesses and
for the other alloys were determined for superplastic strains nominally

up to 200% using single, rather than interrupted tests.

16
 The strain rate sensitivity index m was measured over the

strain rate range 2x 10-5s-1 -4x 10-3s-1 for each alloy by

repeatedly doubling the cross-head speed and recording the maximum

load after each speed increment. The m value was calculated from

the approximation (1):

ln(P2 /P
dlncY 1
d1ni ln(V /V
2 1

where P1 and P and V1 and V2 are the load and cross-head speed before
2'
and after the strain rate increase. This "step-strain rate" procedure

was carried out after a 4% prestrain at 2x 10-4s-1. In order to

minimise the effects of strain it was desirable to cover the entire
strain rate range in the smallest strain increment possible, whilst
allowing steady state flow to be achieved after each increase in

cross-head speed. It was found that, for the lowest stresses, a constant
or maximum load was not reached and, in these casesl the strain rate
steps were carried out after 1% strain increments. In nearly every case
the entire strain rate range was covered within a total strain increment

of 25%.

Strain rate sensitivity was also determined as a function of

strain by temporarily increasing the cross-head speed by 25% after each
50% strain increment.

Calculations of flow stress were made from estimates of cross-

sectional area at various strains during the test based on micrometer

measurements of width and thickness strain made on the gauge length

centre after the test. These measurements were also used to calculate

the R values: R=6 /E ln(w/wo)/ln(t/to)

t _ý
The cooling rate from the test temperature was 25*C min-1,

except for some of the IMI 550 tests, for which faster cooling rate of
150* min-' was used. The appropriate cooling rate was maintained down

to 700% and air was not admitted to the test chamber until the temperature

of the test piece was 300*C or lower.

3.3 Room temperature tensile testing

The test pieces used for the determination of room temperature

tensile of the Ti-6Al-4V sheet were standard "D-size" test

properties
(gauge length 15.9mm) machined from the (larger) gauge lengths
pieces

17
of the superplastically deformed specimens. It was subsequently found
to be more convenient simply to remachine the edges of the superplastic
tensile test pieces and retest at room temperature and this approach
was used for the other alloys. The tensile test pieces used for the
as-received conditions and for material annealed at the superplastic
temperatures were standard "B-size" (gauge length 32mm). The annealing

was carried out in a vacuum furnace at a pressure of < 10-4 torr with
the appropriate coolingrate achieved by slowly sliding the furnace off
the vacuum tube. The room temperature testing was carried out at a
strain rate of 5x 10-5s-1.

3.4 Microstructural examination

The etchants used for the microstructural examinations were the

stain etch (40% methanol, 40% glycerol, 15% bezalkonium chloride, 5%

hydrofluoric acid) for the quenched microstructures and Kroll's reagent
(1% hydrofluoric acid, 12% nitric acid, 87% water) in all other cases.
Phase proportions in the quenched Ti-6Al-4V sheets were determined by
point counting using a 10 x8 grid over 10 different areas ie 800 points
for each microstructure. The mean linear intercept (mli) of the alpha

phase was measured ie alpha/alpha grain boundaries were ignored and then
the contiguous alpha grain size g was calculated in the two principal
directions on the LS and TS sections from:

g=Px mli

where P is the alpha phase proportion. The alpha phase aspect ratios
were then determined from the grain size measurements. Alpha phase
aspect ratios were also determined for the 3.3mm Ti-6Al-4V sheet after
superplastic strain and after reheating to the forming temperature and
quenching into water.

3.5 Analysis of necking Ti-6Al-4V during superplastic deformation

An attempt has been made to analyse neck development during

superplastic deformation using uniaxial flow stress data measured on L

orientation Ti-6Al-4V sheet test pieces of initial thickness 3.3mm at a
testing temperature of 925*C. A computer programme was written to

predict the effect of a small strain increment in the uniform region

of an imaginary test piece on the strain, strain rate, strain rate
sensitivity and strain hardening in a pre-existing neck or inhomogeneity.

18
The initial area of the inhomogeneity was varied from 50% to 99% of

that of the uniform region. For the purposes of the calculation the
neck and uniform regions were treated as two separate test pieces
pulled in series such that the strain rate in the uniform region was
constant at 3x 10-4s-1 (approximately that of maximum m at low

strain). The neck strain rate was assumed to be constant during each
strain increment, but both the uniform and the necked regions were
allowed to strain harden during each increment according to the true

stress-true strain curves for the strain rate in each region. After

each increment a new value of neck strain rate was calculated.

Interpolations of the experimental data were made by a curve

fitting subroutine and the values of strain rate sensitivity dlna/dlný

and strain hardening exponent (n = dlna/dlnc) obtained by

differentiation.

The experimental data used is shown in Fig 4. Step-strain rate

tests were used to determine the lna/lný relationships at true strains
of approximately 0.25 and 0.8, true stress-true strain curves were
determined for three different strain rates of 3x 10 -4 s,-1 10 -3 s -1 and
1
2.5 x 10-3 S_ and the gradient (dlna/dlný) was determined as a function
3x 10 -4 -1 Flow
of strain at a strain rate of s stresses at zero strain

were estimated by extrapolation of the true stress-true strain curves

back to zero strain. The experimental data, shown by circles in Fig 4,
1
were extrapolated up to a strain rate of 5x 10-2 s- and down to
4.5 x 10-5 s -1 However, data for strain rates <3x 10-4 s -1 were not

required in the calculation.

19
4 RESULTS
4.1 Superplastic deformation and subsequent tensile properties
4.1.1 Ti-6Al-4V

Plots of lk.,ý cy against ici,, ý are given in Fig 5 for L and T

orientations for sheets of initial thickness 0.9-3.3mm. The 1n o/lný

relationships for the three batches of 2mm sheet are shown in Fig 6.

For each of the sheets there was apparently little effect of test piece
orientation on flow stress. The flow stresses were similar for each of
the sheets tested except that the 2mm thick batch A material showed
slightly lower flow stress at low strain rates, although this is

exaggerated by the logarithmic axes. The effect of strain rate on

strain rate sensitivity m for each of the Ti-6Al-4V sheets is shown
in Figs 7 and 8. In each case a peak in m value occurred at a strain

rate of about 2x 10-4 s -1 At a given strain rat4 the variation in m

value between the sheets was typically ± 0.05 and repeat tests on

several of the sheets resulted in variation in m values of about ± 0.02

at a given strain rate.

The effect of superplastic strain on flow stress at 925*C and

3x 10 -4 -1 in Fig 9. A solid line has been drawn through
s is recorded
the data points for the sheet of initial thickness 3.3mm. The flow stress
increased with strain up to a true strain of about 1.4 (300%). There

was little effect of strain on flow stress for higher strains.

Strain rate sensitivity m decreased with strain at 925'C as

shown in Fig 10, which gives data for the 3.3am Ti-6Al-4V sheet.

The effects of superplastic strain on room temperature 0.2% proof

stress (0.2PS) and tensile strength (TS) are shown in Figs 11 and 12. The

full tensile results for the 3.3mm sheet are given in Table 1 and the data

for the 2. Omm batch A material can be found in ref 110. No room

temperature tensile tests were performed on the 2mm batch C material.

The effect of annealing at the superplastic temperature, corresponding to

zero strain in Figs 11 and 12, was to reduce the 0.2PS by-8-t2% and the TS

by 1-10% compared with the as-received (mill annealed) material. Superplastic

true strain of 0.9 reduced the 0.2PS and TS typically by a further 2%,

although for the T orientation there was apparently no effect of

superplastic strain on strength for several of the sheets when compared

with the heat cycled condition. Uniform and total elongations and moduli

were generally unaffected by superplastic strain.

20
 The textures ranged from a basal "edge" texture (basal
starting
poles parallel to the T direction) in the thickest sheet (Fig 13) to

a "sheet" texture (basal poles at about 20* to the sheet normal inclined

towards the final rolling direction) in the thinnest sheet. The 2.0mm

thick sheets showed intermediate textures. Despite the cross rolling

process only the 2. Omm batch B material showed four-fold symmetry in

the pole figure. The starting textures for each of the sheets, except
for the 1.8mm and 2. Omm batch C sheets, are given elsewhere (111).

Only the texture of the 3.3mm sheet is considered further here.

The basal pole figure for the sheet surface (Fig 13a) showed
concentrations of poles parallel to the transverse direction and normal
to the sheet plane. The intensities of peaks close to the edge of the

pole figure cannot be accurately determined or the peaks may remain

undetected. Hence, the intensities of poles close to the transverse

direction are likely to be higher than indicated. The prism planes at
the sheet surface (Fig 13b) lay in the plane of the sheet or inclined

towards the L direction. The (110 ) texture pattern (Fig 13c) was similar
to that of the (0002),. The pole figures for the sheet centre were

generally more distinct and with more intense peaks than those of the

sheet surface. The (0002) pole figure (Fig 13d) showed an edge texture
a
component and an annular concentration around the sheet normal. The
(1010) pole figure for the sheet centre (Fig 13e) is different from that

of the sheet surface (Fig 13c) and indicates a rotation of 30' about the

c axis such that for the sheet centre <11ý0> was parallel to the sheet

normal, whereas for the sheet surface <11h> was parallel to the L direction.
The (100k <110> texture (Fig 13f) is a typical beta rolling texture
.j
(112-114) filo) to the T direction and the L direction
with poles parallel
(not detected) and at 45' to the sheet normal.

Annealing at 925* (Fig 14) caused a sharpening of the texture

(112,115) with significantly higher peak intensities in the sheet centre.
The (0002) and (110) pole figures for the sheet surface showed the
a
development of four new intensity peaks (Fig 14a and c).

Superplastic strain caused a substantial reduction in texture

intensities (Fig 15). Neither the alpha nor the beta phase textures

removed, although for the sheet surface only the (0002)

were completely a
(110) close to the sheet normal were detected (Figs 15a and c).
and peaks

21
 The microstructure of the LS section of the 2mm batch A

material, shown in Fig 16a consisted of alpha grains elongated in the

final rolling direction with particles of beta phase between the alpha
grains. The microstructures of the other Ti-6Al-4V sheets were similar
(111). Heating to the forming temperature (925*C 1hr furnace cooled
,
1)
at 25"C min- resulted in more equiaxed and larger grains (Fig 16b).
1
Superplastic strain to 150% followed by cooling at 25'C min- (Fig 16c)

caused a further increase in grain size compared with the heat cycled
material and largely removed directionality in the microstructure.

The microstructures in Fig 16 are not representative of the

structure at the forming temperature because of the transformation which

occurred during cooling. Fig 17, which shows the microstructure of the

mill annealed 3.3mm sheet after quenching into water from 9259C provides

a better indication of the high temperature structure. The LS section

microstructure (Fig 17a) was much more strongly directional than that

of the TS section (Fig 17b). However, this difference was much less
marked in the other batches of Ti-6Al-4V sheet (111). Superplastic strain
in the L direction (Fig 18) increased the grain size and reduced the
directionality in the microstructure. Each of the microstructures in

Fig 18 have been reheated to the forming temperature and water quenched.

The alpha aspect ratio ie ignoring alpha/alpha boundaries was

measured for the quenched LS microstructure and is shown as a function

of superplastic strain in Fig 19. The alpha phase appeared to be

completely equiaxed after a strain of about 1.5 in the L direction.

4.1.2 IMI 550 (Ti-4Al-4Mo-2Sn-0.5Si)

The flow stress of IMI 550 at 900'C is given as a function of

strain rate in Fig 20. There was almost no effect of test piece

orientation on flow stress and the m values for the L and T orientations

were also very similar (Fig 21) with the peak value corresponding to a

strain rate of about 10-4sýl. The flow stress at 900*C increased gradually

with superplastic strain (Fig 22) and the strain rate sensitivity

apparently showed a slight increase up to a strain of about 0.8 (Fig 10).

The effect of superplastic strain at 900*C at a strain rate of

1 is in 23 and
3x 10-4 s- on the 0.2PS and TS of IMI 550 shown Fig the

full tensile results are given in Tables 2-4. Annealing at the forming

22
temperature caused a slight reduction in 0.2PS and TS, compared with
the as-received material. Superplastic strain up to 190% apparently had
little further effect on room temperature strength, although there was
some scatter in the L orientation results (Fig 23). Total elongations

were significantly lower after superplastic strain than in the

as-received condition.

The effect of post-forming heat treatment on the room tempature

tensile properties of the L orientation was investigated. Ageing at

500*C for 24 hrs after forming increased the 0.2PS and TS by about 14%.
1.
The standard post-forming cooling rate was 25'C min- However, a faster
i
cooling rate of 150'C min- produced, in the as-formed condition, 0.2PS

and TS increases of about 11% compared with the slow cooled samples.
Ageing of the fast cooled material further increased the 0.2PS by 10% and

the TS by 8%.

Tensile tests were also carried out on material annealed at the

forming temperature for 1/2 hr to simulate the heat cY'cle associated with
1
superplastic strain. Cooling rates of 25 and 150*C min- were used and
both aged and unaged specimens were tested. The proof, tensile strengths

and total elongations for these conditions (Table 2) were found to be

slightly higher than the correspQnding values for the superplastically

formed and heat treated conditions (Table 4).

Table 2 also includes tensile properties for

material annealed at
1.
90D'C for 31hrs and cooled at either 25 or 150*C min- This relatively
long anneal reduced proof and tensile strengths, although the material

remained sensitive to cooling rate.

The as-received (annealed) LS section microstructure of IMI 550

(Fig 24a) contained highly deformed alpha grains elongated in the

longitudinal direction. The TS section was similar with alpha grains

elongated in the transverse direction. The alpha phase was much more
19
equiaxed after heat cycling (900'C 1hr, cooled at 25'C min- Fig 24b)

strain up to 150% further reduced the directionality in

and superplastic
the slightly increased the grain size (Fig 24c).
microstructure and

4.1.3 Ti-8Al-lMo-lV

The flow stress of Ti-8Al-1Mo-1V increased with strain rate and

decreasing temperature as shown in Fig 25. The strain rate corresponding

to value increased with temperature (Fig 26) and a maximum m

peak m

23
value of 0.82 was found for both orientations for a test temperature of
1010% for both orientations. The m value was highly strain rate

sensitive at this temperature (Fig 26). Superplastic strain caused

some increase in flow stress at the higher test temperatures (Fig 27),

but at 910'C strain softening apparently occurred. Strain rate

sensitivity at 1010% decreased rapidly with strain (Fig 10).

The room temperature transverse tensile properties of

Ti-Ml-lMo-W before and after superplastic strain are given in Table 5.

The effect of annealing at each of the forming temperatures of 940,

970 and 1010% was to reduce the 0.2PS and TS by about 5% compared with
the as-received condition. Superplastic strain of 200% further reduced
the proof and tensile strengths by another 5%. There was apparently
no effect of forming temperature in the range 940-1010'C. The room

temperature properties following forming at 9100C could not be evaluated

because each of the test pieces pulled at this temperature failed during

superplastic testing after about 200% strain.

The alpha and beta phase textures (not included) of the

Ti-8Al-1Mo-1V sheet after 120% strain at 1010*C were similar to those

of the as-received sheet. The pole figures showed (0002) and (110)
U
peaks close to the sheet normal tilted slightly towards the rolling
direction with indications of another set of peaks close to the

transverse direction. Pole figures for material annealed at the forming

temperature were not determined. However, the texture results on this

alloy were consistent with those on Ti-6Al-4V (Figs 13-15), which

indicated a sharpening of texture associated with recrystallisation
during annealing and then a gradual reduction of texture intensity with

superplastic strain.

The microstructures of the LS sections of the Ti-Ml-lMo-W in

the as-received (duplex annealed) condition, after annealing 1010*C for

1
1hr and cooling at 25'C min- and after superplastic strain of 150% in

the L direction at 1010*C are shown in Fig 28. Annealing at the forming

temperature caused a marked increase in grain size and superplastic

strain produced a further slight increase. The superplastically formed

Ti-Ml-lMo-W consisted of about 50% transformed beta in the form of

Widmanstatten alpha plates (Fig 28c), whereas the heat cycled material

contained no such secondary alpha plates.

24
4.1.4 Ti-15V-3Cr-3Al-3Sn

Flow stress and m values for Ti-15V-3Cr-3Al-3Sn are plotted

against strain rate in Figs 29 and 30. The m values decreased slowly
with strain rate and there was little effect of temperature in the
range 810-910'C. The flow stress remained roughly constant up to
superplastic true strains of 1.0, but some strain softening occurred
thereafter (Fig 31). Strain rate sensitivity at 910'C increased from
0.35 to about 0.4 after a strain of 1.4 (Fig 10).

The effect of superplastic strain on room temperature tensile

properties of Ti-15V-3Cr-3Al-3Sn are shown in Table 6. Both strength

and ductility were reduced by 300% superplastic strain, although the

post-forming properties appeared to be independent of forming temperature

in the range 810-910*C.

The (110) pole figures for the Ti-15V-3Cr-3Al-3Sn before and

after superplastic strain are given in Fig 32. The X-ray analysis
suggested that the presence of a small amount of alpha phase may have
slightly increased the apparent texture intensities in Fig 32a. However,
it is clear that the texture in this alloy is not randomised by

superplastic strain, unlike that of the Ti-6Al-4V (Figs 13-15).

Superplastic strain of Ti-15V-3Cr-3AI-3Sn at each of the

temperatures investigated produced surface rumpling as shown in Fig 33.

This extended intothe test piece heads in the regions close to the ends

of the gauge length.

The microstructures of the material in the as-received (annealed)

condition and after various amounts of superplastic strain at 810*C

are shown in Fig 34. AiAnealing at the forming temperature caused a large
increase in grain size (Fig 34a), but superplastic strain resulted in

grain refinement (Figs 34b-g). The effect of superplastic strain at 810

and 860% on grain length in the L direction is shown in Fig 35. Both

the heat cycled and the superplastically formed material contained alpha

precipitates in two forms: relatively large intragranular precipitates

and much finer particles decorating the grain and subgrain boundaries

(Fig 34h). This precipitation probably occurred during cooling from the

forming temperature.

25
4.2 Superplastic deformation of sheet test pieces machined from

Ti-6Al-4V bar

4.2.1 Test piece shape after superplastic strain

The appearance of the test pieces after superplastic deformation

under each of the conditions of temperature and strain rate is shown in

Fig 36.

TL orientation

The TL orientation test pieces after testing were characterised

by a very irregular strain distribution along the gauge length with
bands of relatively undeformed material running across the test piece.
This behaviour was most marked for the 875*C and 925*C test temperatures

and the banding is shown at A in Fig 36. The banding was more apparent
in the low strain rate test pieces than in the corresponding high strain

rate test pieces for test temperatures 800-925'C. The test pieces pulled

at 975' failed at a relatively low strain with no evidence of banding

across the gauge width.

4.2.1.2 ST orientation

For the ST orientation the banding direction was through the

thickness of the test pieces and small blocks of relatively undeformed

material were left protruding out of the surface eg Fig 37a. This effect

was more clearly visible near the ends of the test piece gauge lengths

ie in the low strain regions than in the centre of the gauge lengths or
in the necked regions, where the test piece surfaces were relatively

smooth. The test pieces pulled at 975*C exhibited relatively low total

elongations and, for both strain rates, showed shallow trenches in the

test piece surface running roughly perpendicular to the tensile axis

(Fig 38). Each of these features was an incipient neck as shown

schematically in Fig 38b. This behaviour was peculiar to the ST test

piece orientation. Final necking at 975*C occurred along a line at an

angle to the tensile axis. The 8000C ST test pieces also showed this
"slant" type local necking (Fig 36).

4.2.1.3 LT orientation
By contrast, the LT orientation test pieces pulled at temperatures

up to 925'C showed more uniform deformation with relatively smooth

surfaces and parallel gauge lengths, except that shallow troughs and

to the test piece axes were visible (Fig 37b).

ridges, running parallel
The height and depth of these features was much less than that of the

irregularities occurring in the test pieces of the other

surface
26
orientations. These surface grooves were also observed on round
test pieces pulled in the L direction.

4.2.2 Microstructure
The microstructure of the as-received material contained bands

of heavily deformed material with strongly directional microstructure

eg at A in Fig 39 and more homogeneous, non-directional regions such as
at B in Fig 39. The microstructures of material quenched into water
from each of the forming temperatures are shown in Fig 40. The alpha

phase volume fractions, estimated from the data in ref 64, were, at
800,875,925 and 975'C respectively, 81%, 68%, 50% and 10%. As discussed
in refs 59,64,65 and 111, the banded regions, such as at A in Fig 39,

contain equiaxed, contiguous alpha grains aligned in the rolling

direction. These aligned alpha grains can be seen on the LS and LT

sections in Fig 40c. There is also evidence from Figs 39 and 40 of alpha
phase alignment, or possibly grain elongation in the T direction, although
to a lesser extent than in the L direction. The distribution of aligned
microstructure is shown schematically in Fig 41, reproduced from ref 51.
The nature of the surface protrusions in Fig 37a lend some support to
the shape suggested in Fig 41 for the bands of aligned alpha grain. In
Fig 40d the etchant has attacked the martensite matrix and martensite
plates are visible between the alpha grains. Each of the microstructures
in Fig 40 shows the alpha phase alignment, although this is less marked
for the higher forming temperatures as a result of the lower alpha
phase content. The material quenched from 975*C (Fig 40d) showed some
variation in primary alpha phase proportion associated with the regions
of aligned and non-aligned alpha grains, although it was not clearly
determined that the alpha rich regions corresponded either to the

aligned or to the non-aligned regions. The TS section of Fig 40d shows

an alpha rich region at A and an alpha lean region at B.

A section of 925*C TL test piece, through a region such as at

A in Fig 36b, is shown in Fig 42. Two distinct areas of different

microstructure can be seen; aligned alpha grains such as at A and

equiaxed microstructure at B. Other banded regions on the TL test pieces

were all found to contain the aligned microstructure. Similarly, the

protrusions on the surfaces of the ST orientation test pieces (Fig 37a)

were found to correspond to areas of aligned microstructure, such as at

A in Fig 43, and depressions in the surface were associated with more

equiaxed microstructure (at B in Fig 43).

27
 The effect of superplastic strain was to increase the average
grain size, particularly during the first 100% strain increment, and to
break up the banded microstructure. The LS section is shown in Fig 44

after various amounts of strain in the S direction at 875'C and at a

strain rate of 3x 10 -4 s -1 After testing, the material was reheated to
875'C and quenched into water. The microsections for Figs 44a-d were
taken from parts of the test piece where the irregular surface
suggested the microstructure was banded. The micrographs in Figs 44e

and f corresponded to regions near the centre of the gauge length where
the test piece surface was relatively smooth. The values for axial strain
given in Fig 44 were calculated from measurements of local width and
thickness strain and by assuming constant volume.

A section through an ST test piece pulled at 975'C revealed that

the microstructure in the necked regions contained practically no primary
alpha phase and showed relatively large alpha plate colonies indicative

of a large prior beta grain size (Fig 45). The neck which propagated
to failure contained only a few isolated primary alpha grains indicating

that the composition at the forming temperature was almost completely

be.ta (Fig 46a). By contrast, the microstructure of a 975*C TL test

piece showed a uniform distribution of primary alpha phase eg Fig 46b

which shows the microstructure close to the point of failure (compare with
Fig 46a).

The ST test pieces B2 and B3 and the TL test piece A4, which had

almost necked to failure (Figs 36c, d, f), showed cavitation in the

necked region eg Fig 47. The cavitation seemed to originate either in

the beta phase or at the alpha/beta phase boundaries (Fig 47) and the

cavities appeared to elongate or coalesce parallel, rather than

perpendicular to the tensile axis.

A scanning electron micrograph of the edge of a test piece after

superplastic strain is shown in Fig 48. The surfaces, which had a

machined finish prior to testing, have been roughened by grain boundary

sliding and the emergence of new surface grains.

4.2.3 Flow stress

The superplastic flow stresses for each orientation and test

temperature are given in Fig 49. These values were calculated from the

load, which occurred in each test at 15% strain or less, and the
maximum
initial cross-section area. At such low strains the test piece gauge

28
lengths would still be approximately of uniform cross-section and

parallel sided. Flow stresses were higher for the high strain rate
and decreased with increasing temperature. For each strain rate the
flow stresses were highest for the LT orientation and lowest for the ST

orientation.

4.2.4 Strain rate sensitivit

The m values, measured at the point of maximum load by

temporarily increasing the crosshead speed by 25%, are plotted against

temperatui: e in Fig 50. For each temperature the m values were higher
for the low strain rate tests and for each orientation and strain rate

reached a maximum at 925*C. With the exceptions of the high strain rate
tests at 8000C and the low strain rate tests at 975'C, the strain rate
sensitivity values were in the order mST ý'mTLý'mLT This was the reverse
of the order of flow stresses (Fig 49).

4.2.5 R values

R values were measured at various points along the gauge length

of each of the test pieces and have been plotted against the local

axial strain c1 in Fig 51. The two strain rates given in Fig 51 were

only nominal values as the irregular strain distributions in the TL and

ST test pieces implied a wide variation in local strain rate.

The most significant feature of the R values for the ST test

pieces was that for test temperatures of 800-925*C all the R values

measured were greater than 1.0, although a large amount of scatter was

recorded. For the 975'C test pieces R values ranged from 0.77 inside the

necked regions to about 1.0 in the adjacent unnecked area (see Fig 38).

For the LT orientation there was much less variability in R than

for the other orientations. All the R values measured for this

orientation were less than 0.9.

The R values recorded for the TL orientation were measured both

on the banded regions of the test pieces, such as at A in Fig 36, and in

the regions between the bands and have been plotted separately as closed

and open symbols respectively. No trend of R value with strain was

apparent, although for the low strain rate tests (Fig 51a) the R values

corresponding to the banded regions tended to be higher than those

measured in the non-banded areas, particularly for the 875*C test

temperature. This is shown in Fig 52 where the true width and thickness

29
strains E:-,. and E have been plotted against 6 for the 8750C low
wt1
strain rate test and the measurement locations on the test pieces
have been identified.

4.2.6 Texture
The pole figures for the material before and after superplastic
strain are given elsewhere (116). They revealed that, unlike the case
of the rolled sheet (section 4.1.1), annealing at the forming

temperature did not appreciably sharpen the texture and superplastic

strain at 925*C caused a gradual reduction in texture intensity. However,

even after 374% strain neither the alpha nor the beta phase texture
intensities were completely randomised.

4.3 R values of Ti-6AI-4V sheet after superplastic strain

4.3.1 Effect of sheet thickness

The effect of uniaxial superplastic strain on R for the L and T

orientations of each of the sheets is shown in Fig 53. All the R values
determined were less than 1.0 ie strain through the thickness exceeded
strain across the width. The R value increased with superplastic strain
for both orientations of each sheet, with the R values generally higher
for the L orientation. It also appeared that the R value decreased with
initial sheet thickness.

The microstructure of the 3.3mm Ti-6AI-4V as-received sheet

(Fig 17 and ref. 111) showed some elongation and alignment of the alpha

grains along two orthogonal directions in the plane of the sheet (the

rolling directions) though this was much less severe than that observed
in highly textured Ti-6Al-4V bar as discussed in section 5.2 where the

anisotropy of superplastic flow was clearly related to the effects of

microstructure. In Fig 54 alpha phase aspect ratios of the material
quenched from the forming temperature are plotted against the R value

corresponding to a true axial strain of 0.4. For each point the aspect
ratio plotted was measured on the cross-section of the corresponding

specimen ie for the L orientation specimens the alpha phase aspect ratio

on the TS section is recorded. This figure shows that the exceptionally

low R value measured for the T orientation of the 3.3mm sheet was

accompanied by a relatively high alpha phase aspect ratio and when the R

value was high (0.9mm thick sheet, L orientation) a correspondingly low

alpha phase aspect ratio was observed. The other orientations of the

0.9 and 3.3mm sheet and the other sheet thicknesses showed intermediate

values of R and aspect ratio.

30
4.3.2 Effect of_test piece geometry

The effect of test piece geometry was investigated using test

pieces of 3.3mm Ti-6Al-4V sheet. The true width and thickness strains,
ewt et, are plotted against axial strain e1 in Fig 55a for test pieces
with the same initial gauge length, but different gauge widths. The
axial strain at the centre of each specimen was calculated from the width
and thickness measurements assuming constant volume : cl =- (E +Et
The strain was more anisotropic for larger gauge widths, but points for
the two narrowest specimens (4mm and 8mm gauge lay
widths) on the same
curve. This indicated that there was no effect of gauge width below 8mm.
Similarly, varying the gauge length (Fig 55b) resulted in more
anisotropic behaviour at shorter gauge lengths.

Test pieces of various initial widths are shown after straining

in Fig 56. In each case constraint by the test piece head near the ends
of the gauge length has reduced the width strain there, resulting in

tapering along the gauge length. However, the taper was less severe for

the narrower specimens and for the smallest gauge width (4mm) the gauge
length was approximately parallel and free from the influence of the
head over most of its length (Fig 56d).

Plotted in Fig 57 are width and thickness strain measurements

taken from different points along the gauge length of a specimen pulled
to failure (test piece 9 in Fig 56). These points are compared with
values for the 4mm gauge width specimen (taken from Fig 55) extrapolated
to E:1 = 4.0 which approximately represent the width and thickness strain
that would be measured with no constraint from the test piece head. The

two sets of curves are coincident for a value of axial strain of about
2.0, but diverge at lower and higher strains indicating increasingly

anisotropic behaviour towards the failure site and towards the test piece
head in the test piece pulled to failure.

4.3.3 R values of other alloys

The R values of the 3.3mm Ti-6Al-4V sheet after superplastic
strain are compared with those of the other alloys investigated (all of
initial thickness 2mm) in Fig 58.

The R values for the alpha/beta alloys in this comparison were

measured after superplastic strain at the temperatures corresponding to

equal phase proportions in each alloy.

31
The data for the Ti-6Al-4V are interpolated data calculated from

the width and thickness strain curves in Fig 55a for the 4mm gauge

width test piece, whereas the other points have been calculated directly
from width and thickness measurements. The IMI 550 behaved in a very
similar way to the Ti-6Al-4V, although with apparently little effect of
test piece orientation. Less data were available for the other two alloys.
However, the R values seemed to be relatively high and similar for the
L and T orientations in each case.

4.4 Analysis of necking of Ti-6Al-4V during superplastic deformation

The computer programme predicted the development of a neck after

each of a series of small strain increments in the uniform region of

the gauge length. The analysis assumed that the neck strain rate was

constant during each increment (see section 3), whereas, in practice, the

neck strain rate would change continuously. Thus, the accuracy of the

prediction of neck development depends on the size of the strain increment

chosen and is reduced by increasing the strain increment size. The effect

of different uniform strain increment size on predicted neck strain is

shown in Fig 59 for an initial neck area of 0.95 (normalised with respect
to the starting area of the uniform section of the gauge length). The

true strain increment size was varied from 0.025 to 0.1. The difference

only became significant at high total strains and for all subsequent

analyses a strain increment of 0.05 was used.

Neck strain has been plotted against uniform strain in Fig 60

for initial (normalised) neck cross-section areas ranging from 0.5 to

0.99. In each case the uniform strain at the point of failure is

indicated by the point at which the gradient becomes infinite. For neck

areas of 0.5,0.8,0.9,0.95 and 0.99 this occurred at uniform strains

about 60,200,300,450 and 850% respectively. The solid line in Fig 60

of
indicates equal strain in the "neck" and uniform regions ie the
inhomogeneity size is zero. The effect of strain on strain rate in the

for an inhomogeneity of normalised area 0.8 is shown in Fig 61.

neck
The neck strain rate initially decreased slightly and reached a minimum

of about 10%. The (constant) strain rate in the uniform region

at a strain
is shown by a solid line in Fig 61. The neck strain rate at zero strain
is indicated by the point on the vertical axis. The load, normalised with

respect to the initial load, is also shown in Fig 61.

32
 The flow stresses in the uniform and necked regions are shown

as a function of strain in Fig 62. Strain hardening and strain rate

hardening caused the neck flow stress to rise more rapidly than the
flow stress in the uniform region where the strain rate was constant.

Strain rate sensitivity m and strain hardening exponent n are

given in Fig 63. The m. values decreased with increasing strain, with
the m value in the neck decreasing sharply as the failure point was
approached. The n values in both the uniform and necked regions
initially increased and passed through maxima. A small degree of strain
softening at high strain is indicated in Fig 63 by negative n values.
This was due to a slight misfit of the polynomial to the true stress-true
strain data points. There is no experimental evidence of strain softening
and the dashed line in Fig 63 indicates the more likely effect of strain
on n.

The total strain hardening rate can be written as the sum of

strain hardening and strain rate-hardening terms. -

da ýcy (3aý dý acy öm

E + dE

where a is the flow stress and e is the strain. This ignores any effects

of increasing surface area and applies to isothermal conditions only. The

of each of thtt two terms on the right hand side of equation (1) were
values
determined as a function of strain in the neck and plotted in Fig 64 along

with the total strain hardening rate do/de. This figure also includes

(3a/3F-) for the uniform deformation. The term incorporating dý/de is

zero for this region. The value of 3a/3e. 'for the neck and uniform regions
decreased with strain and would tend--towards zero in the absence of strain
(dý)
cym for increased from to
softening. The value of the neck close zero
de
and became dominant above about 85% uniform strain. Consequently the total

strain hardening rate for a neck of initial (normalised) area decreased

strain, reached a minimum at about 70% strain and then increased rapidly.
with

33
 DTSCITqqTON

5.1 Titanium alloy sheet

5.1.1 Superplastic properties and microstructures

All of the Ti-6Al-4V sheets tested were nominally of superplastic

quality. However, Figs 7 and 8 show that for a given strain rate the m
values varied significantly (by up to about 0.18). It is well known
(117-119) that high strain rate sensitivity is favoured by fine grain
size, but m values are also sensitive to grain size distribution, grain
aspect ratio and small variations in composition (88,118,120,121). It
is likely that all of these factors contributed to the variability of m
value and flow stress between the different Ti-6Al-4V sheets examined
here. The data in Figs 5-8 indicate the range of m values and f low

stresses likely to be found in commercial superplastic quality Ti-6Al-4V

sheet and suggest that a very detailed microstructural analysis would be

required to predict the relative superplastic forming potential of

various batches.

The superplastic properties of each of the alpha/beta alloys

were qualitatively similar with peaks in m value in the strain rate
1-3x 1 (Figs
range 10-4 s- 10-4 s- 7,8,21,26) and an increase in grain

size with superplastic strain (Figs 16,18,24,28). By contrast, the

Ti-15V-3Cr-3Al-3Sn showed relatively low m values decreasing with
increasing strain rate (Fig 10) and grain size decreasing with strain
(Figs 34 and 35).
1
It is interesting to note that at a strain rate of 1.5 x 10-4 s-
corresponding roughly to that of maximum m value the flow stress of

each of the alpha/beta alloys at the appropriate temperature for equal

phase proportions was practically identical, whereas the flow stress of

the Ti-15V-3Cr-3AI-2Sn at this strain rate was much higher (Fig 65).

However, it is unlikely that the similarity of flow stress in the alpha/

beta alloys is of particular significance in view of the difference in

grain size between the alloys (Figs 16,24,28). At the strain rate

corresponding to maximum m the deformation is likely to result largely

from grain boundary sliding (1,2,18,21,32,37,122), but the average grain

boundary shearing rate will be lower for the finer grained IMI 550 than

for the other two alpha/beta alloys. This would tend to reduce the flow

stress. However, the IMI 550 contains a relatively high proportion of

molybdenum, which compared to other alloying elements, has a low

34
diffusivity in beta titanium reducing the effective lattice and grain
boundary diffusion rates (44,45,55,56). Paton and Hall (56) reported
that Mo additions to Ti-6AI-4V increased the superplastic flow stress.
The flow stress is also likely to depend on shear modulus (123), which
in turn varies with the composition. Thus, the flow stress at any
particular strain rate or phase proportion is determined by a combination

of factors and it should not be inferred from Fig 65 that there is a

unique value of flow stress for alpha/beta titanium alloys at the 50/50

phase temperature and at the strain rate corresponding to maximum m.

It has been established (88,124,125) that the increase in flow

stress of Ti-6Al-4V during superplastic strain (Fig 9) is almost entirely

due to grain coarsening. It can be seen from Fig 9 that the flow stress
did not increase above about 1.7 true strain, consistent with the stable
grain size at this strain (Fig 18). Grain growth during superplastic
strain also occurred in the other alpha/beta alloys (Figs 24 and 28) and
the strain hardening in these cases can presumably be explained in the
same way. The small amount of strain softening that occurred in the
Ti-Ml-lMo-W was possibly associated with the b reak-up of the directional
microstructure, as reported for extruded 60/40 brass (126) and Cu-P
alloys (127), or may have been due to inaccurate estimation of the
instantaneous cross-section area towards the end of the superplastic test
(the Ti-Ml-lMo-W test pieces pulled at 9100C either failed or necked
substantially).

The strain hardening rate (Fig 22) and, therefore the rate of
grain growth, was lower for the IMI 550 at the 50/50 phase proportion
temperature (9000C) than for the Ti-6Al-4V or the Ti-Ml-lMo-W at 9250C

and 1010% respectively. This was partly due to the relatively low

temperature and partly as a result of the low diffusivity of the Mo in

the IMI 550. The grain size stability of the IMI 550 was also reflected
in the increase in m value with superplastic strain up to 130% (Fig1O).

This increase in m was probably related to the microstructure becoming

more equiaxed, which appeared to override the tendency of the simultaneous

grain coarsening to reduce m. Higher superplastic strains (128) would

presumably have decreased the m value as the effect of grain coarsening

became dominant.

Superplastic strain produced subgrains and refined the grain size

of Ti-15V-3Cr-3Al-3Sn, whereas static annealing, which occurred in the

test piece head, caused rapid grain growth. It has been reported (129,130)

35
that sliding can occur in subgrain boundaries in aluminium, with the

subgrain sliding preceded by a gradual increase in misorientation between

adjacent subgrains, leading to the formation of high angle boundaries.

This is analogous to the method proposed by Hayden et al (122) for the
break-up of fibrous microstructures during superplastic strain, whereby
high aspect ratio grains are divided along their lengths by subgrain
structures. Alternatively, Griffiths and Hammond (8), working on beta
titanium alloys and beta brass, suggested that subgrain walls form

continuously by dislocation climb with equiaxed grains and grain refine-

ment resulting from the migration of subgrain boundaries and the pinching-
off of grains. Therefore, it seems likely that the reduction in grain
size in Ti-15V-3Cr-3Al-3Sn occurred by the development of a subgrain
dislocation structure with sliding eventually taking place on the subgrain
boundaries. The surface rumpling (Fig 33) indicated that grain boundary

sliding was occurring. The decreasing grain size resulted in an increase

in strain rate sensitivity at 910'C (Fig 10) up to 300% strain and a

small degree of strain softening.

After superplastic deformation and furnace cooling the

Ti-Ml-lMo-W contained approximately 50% transformed beta (Fig 28c),

whereas the annealed material taken from the test piece head contained

only intergranular beta and no secondary alpha. There was also evidence

of this behaviour in the Ti-6Al-4V (Figs 16b and c). The likely

explanation for this is simply that the cooling rate in the test piece
heads, which were covered by the grips, would have been lower than that
in the gauge lengths. Consequently, the beta-alpha phase transformation
in the test piece heads occurred with migration of the alpha/beta
boundaries and enlargement of the existing alpha grains. In the gauge
lengths the greater undercooling and slightly larger grain size favoured

the nucleation and growth of Widmanstatten alpha plates (131) inside

the beta grains. This explanation is consistent with the grain size
decreasing in the order Ti-8Al-1Mo-1V>Ti-6Al-4V>IMI 550 after superplastic

strain. Higher superplastic strains and, hence larger grain sizes, can

result in the formation of Widmanstatten alpha plates in IMI 550 also

(132).

The highest m values measured were for the Ti-8Al-lMo-lV at

1010'C (Fig 26). However, for commercial forming operations other
factors must be taken into consideration when selecting an alloy. A low

forming temperature is desirable so that cheap die materials can be used,

36
die life prolonged and contamination minimised. An m value which is

roughly constant over a wide range of strain rates is also an asset as

some variation in strain rate is to be expected in a complicated
forming. The variation in m value for the IMI 550 (Fig 21) over the

strain rate range was relatively small, although the peak values were
slightly lower than those of the Ti-6Al-4V or Ti-8Al-lMo-lV. By

comparison, the m value of the Ti-8Al-lMo-lV was highly strain rate

sensitive at 1010'C (Fig 26). A further disadvantage of the Ti-8Al-lMo-lV
is that for this alloy the phase proportions are relatively sensitive to
temperature (133). Thus, close temperature and strain rate control
would be required in order to maintain optimum superplastic properties
in this material.

The advantages of the high lattice diffusivity of the beta alloy

Ti-15V-3Cr-3Al-3Sn in promoting high strain rate sensitivity were out-
weighed by the large initial grain size and the rapid grain coarsening
that occurred during annealing before superplastic deformation.

Consequently, the Ti-15V-3Cr-3Al-3Sn appeared to be the least attractive

of the alloys for superplastic forming showing relatively low strain rate
sensitivity and an irregular surface after forming. Low total elongations
are reported (8,134) for beta titanium alloys during superplastic
deformation.

5.1.2 Room temperature tensile properties after superplastic deformation

The most marked effect of superplastic strain on the room

temperature tensile properties of Ti-6Al-4V was the reduction of 0.2PS

and TS by about 10% associated purely with the thermal cycle. This was

caused by recrystallisation (134,135) that removed the worked structure

resulting in larger, more equiaxed grains (136). By comparison the

effect of superplastic strain on room temperature strength, which was

the result of grain coarsening (137), was small. These observations on
the effects of annealing at the forming temperature and superplastic

strain agree with other uniaxial data (62,134,137,138) on Ti-6Al-4V. The

results for biaxially formed material (139-141) show similar trends,

but more scatter, probably reflecting slightly different forming and

testing techniques. The British standard (142) for Ti-6Al-4V sheet

specifies minimum 0.2PS and TS levels of 900 and 960 MPa respectively.

The strength loss inherent in superplastic forming often means that formed

components do not reach these minimum values.

37
 The anisotropy in room temperature properties was due to

crystallographic texture. The anisotropy was particularly pronounced

for the edge textured 3.3 mm Ti-6Al-4V sheet (Figs 11 and 12 and Table 1)
for which the modulus and strength were greater in the transverse
direction than in the longitudinal direction.

The IMI 550 alloy also showed some loss in room temperature

strength after annealing at the forming temperature (Table 2). The

additional effect of superplastic strain on strength was slight (Fig 23).

The standard heat treatment for IMI 550 involves an air cool from the

solution temperature (900'C) and ageing at 500'C to produce a fine alpha

precipitation. This ageing response is enhanced by increasing the cooling

rate as more metastable beta phase in retained and this phase is richer
in alpha stabilising elements. The post-forming cooling rate of 25'C
1
min- was sufficient to give a 0.2PS and TS increase of about 10% on

ageing. Resolution treatment at 900*C, air cooling and ageing can

increase room temperature strength at least to that of the as-received
(mill annealed) material and usually beyond (110,128,132). Such a heat

treatment applied to a formed sheet component may, in practice, cause

contamination and distortion problems. However, since the optimum forming

temperature is the same as, or close to, the solution temperature the

same result can be achieved simply by increasing the cooling rate after
the forming operation and subsequently ageing. It is likely that cooling
1
rates higher than the 150'1C min- used in this work could easily be

achieved in practice.

The as-formed 0.2PS and TS advantage of the IMI 550 over the
1.
Ti-6Al-4V was about 5% for a cooling rate of 250C min- Ageing of the

IMI 550 at 500*C after forming further increased the 0.2PS and TS by
5%. The highest strengths achieved in the IMI 550, by cooling at 1500C
1
min- and ageing, gave 0.2PS and TS levels typically 18% greater than

those of the as-formed Ti-6Al-4V.

Tables 2-4 indicate that superplastic strain reduced total

elongations of IMI 550 at room temperature. This contradicts other

results (110) on IMI 550 sheet, although Duffy (128) also reported some
loss of ductility. It is possible that slight surface contamination by

oxygen occurred during superplastic testing accounting for the reduction

in elongation.

38
 The as-formed strength of the Ti-Ml-Mo-W was low by comparison

with the Ti-6Al-4V and the IMI 550, although modulus and elongation were
both higher.

The strength and ductility of the Ti-15V-3Cr-3Al-3Sn alloy were

both reduced by superplastic deformation (Table 6). Ageing to precipitate
alpha would presumably strengthen the alloy considerably and increase

the modulus after superplastic strain. There appears to be little

advantage in superplastically forming this alloy for commercial

applications in view of its cold formability (143) and competition from

the highly superplastic alpha/beta alloys.

5.1.3 Texture

The texture variation through the thickness of the 3.3 mm

Ti-6Al-4V sheet (Fig 13) was indicative of different deformation

conditions at the sheet centre and the sheet surface during rolling,
although this was not reflected in the microstructure . The (0002) and
a
(110) pole figures were similar suggesting that the Burgers orientation
relationship (0002) //{110, ja9 <1120> //<111> was generally obeyed.
a a a
However, examination of the (10TO) pole figures 13b, e and 14b, e shows
Ot
that the Burgers variantwith (0002) parallel to the T direction was
a
preferred. This occurred because the transverse basal texture component
develops during sheet rolling (112), whereas the remaining variants can
only form by transformation during cooling.

Superplastic strain of Ti-6Al-4V caused a reduction in texture

intensity at both the sheet surface and the sheet centre (Fig 15)

consistent with grain boundary sliding throughout the material. The

peak intensities in the alpha and beta pole figures were similar both

before and after superplastic strain suggesting that the grain rotation

was not restricted to one phase. These observations of the effect of

strain on the alpha texture are in agreement with other work (59,138) on

uniaxial superplastic deformation of Ti-6Al-4V and similar results were

obtained for IMI 550 (132). Biaxial forming of Ti-6A1-4V (139-141) appears

to reduce the alpha phase texture intensity in a similar way. Much less

attention has previously been paid to the effect of superplastic strain

on beta textures. McDarmaid et al (60) reported that in highly textured

Ti-6Al-4V bar beta texture was almost completely removed by 300% super-

plastic strain, whereas the alpha texture was retained, although with a
intensity. It was suggested (61) that this was caused by a
reduced

39
preference for in the beta/beta boundaries (51). Ma and Hammond
sliding
also found evidence (44) by high temperature electron microscopy that

superplastic deformation in Ti-6Al-4V is associated with much larger

distortions of the beta phase than of the alpha phase. Therefore, these

results (44,60) differ frQm the work reported here and the precise role
of the two phases remain unclear.

The Ti-6Al-4V sheets investigated had a wide range of initial

textures, but there was no evidence of any influence of texture on

superplastic properties. This agreed with other work on Ti-6Al-4V (59,88)

and was consistent with the assumption that the superplastic strain was
largely accommodated by grain boundary sliding.

The texture of the Ti-15V-3Cr-3Al-3Sn sheet Fig 32 was either

{1 101<1 10> U 1101 planes parallel to the sheet plane and <1 10> parallel
to the rolling direction) or J110}<100>. Neither of these textures is a
common bcc rolling texture (113,114). Superplastic strain apparently
did not reduce the texture intensity. This was consi. stent with the

occurrence of slip deformation, which led to the formation of subgrains.

5.1.4 Activation energy

Estimates of the activation energy Q for superplastic flow in

Ti-6Al-4V at 850-925*C were made by Arieli et al (124) using the Dorn
creep equation (123):

AGb bPnQ
ý= ý--TG (-,j) (ý) D. exp -- RT

where A and p and constants, G is the shear modulus, b is the Burgers

vector, k is Boltzmann's constant, T is the absolute temperature, R is

the gas constant, d is the grain size, D0 is a frequency factor and n is

the stress exponent (n = 1/m). The estimated values of Q (124) ranged
1
from 224-299 kJmol- and increased with superplastic strain.

An attempt was made to calculate the activation energy for

superplastic deformation for the Ti-8Al-lMo-lV alloy based on the data

in Fig 25 and using a similar method to that of Arieli et al (124). The

flow stress was determined for each test temperature for a strain rate

of 10-4 s,-1 which was assumed to lie within region II in each case. A
-n Gn-1 I/T
plot of RInTa against was made using the appropriate values of
n and G for each temperature. However, the points did not fall on, or

even near a straight line and a meaningful estimation of Q could not be

40
made. Instead, it is useful to consider the sources of error in the

determination of
Very little information is available for the shear modulus of
titanium alloys at high temperature. The estimations of Q for the
Ti-8Al-lMo-lV alloy and that for Ti-6Al-4V (124) used extrapolated shear
modulus data (144) for alpha phase titanium, which is likely to be a

source of significant error.

The calculated value of Q is particularly sensitive to the strain

rate sensitivity and its temperature dependence. There are several ways
(1,145) in which the m value can be estimated from load displacement

curves and these can yield significantly different results (145).

Arieli and Rosen (145) claim that the m value for Ti-6Al-4V at 900'C
is really constant at 0.5 throughout region II. However, the over-
whelming weight of evidence (1) indicates that this is not the case and
that m phases through a maximum in region II. The m values given
in this thesis were calculated using one technique only, as described
in section 3.2, in order that the results should be comparable. In the
determination of Q for Ti-6Al-4V Arieli et al (124) assumed that the m

value did not vary (for a particular strain rate) in the temperature

range 850-925*C. Clearly, a similar assumption is not justified for the

Ti-8Al-lMo-lV alloy (Fig 26). Other workers (8,128,146) have chosen

simpler rate equations to describe superplastic flow and have ignored

either the influence of shear modulus or the temperature variation in m

when calculating the activation energy.

Further sources of inaccuracy are the variability in test piece

machining and microstructure. The small variations in m, recorded

during repeat tests on material of the same batch could not be entirely

accounted for by the random inaccuracies in load measurement.

A number of different models have been formulated for superplastic

flow (reviewed in section 2.1 and in refs 1,35,40,147) with activation

energies corresponding to grain boundary diffusion, lattice diffusion

or phase boundary diffusion. Measured values of Q (reviewed in ref 1)

are generally of the order of that for grain boundary diffusion in some
and of the order the activation energy for lattice diffusion in
systems
For titanium alloys the reported activation energies (55,124,
others.
128,141,148,149) seem to be greater than the activation energies for

diffusion in either the alpha (42) or the beta (150) phases and
self

41
certainly greater than the activation energies for grain or phase boundary
diffusion. This is not consistent with the models of superplastic
deformation and the scatter in the results highlights the difficulties

associated with the experimental determination of Q. It has been

suggested (43) that dynamic recrystallisation in two phase titanium alloys

might explain the anomolously high activation energies, but no experimental

evidence has been put forward in support of this.

5.2 Superplastic deformation of sheet test pieces machined from

Ti-6, Al-4V bar

5.2.1 Microstructure and flow stress

Most models of superplastic flow eg refs 22,28,36,38 invoke grain boundary

sliding with accommodation at the grain corners by diffusion or dislocation

motion. The microstructural observations in section 4.2 indicate that

a model of this type applies to Ti-6Al-4V, although the precise role of

each phase is not clear. In Fig 48 whole grains have moved relative to
their neighbours and emerged from the test piece surface. There was no
evidence of any intergranular voids, apart from in the necked regions
where strain rates were locally high and the deformation non-superplastic.

For grain boundary sliding and grain rotation to occur in the

aligned regions of the Ti-6AI-4V microstructure the contiguous alpha

grains must slide apart as they cannot readily rotate or move in unison.
A consequence of this is the break up of the contiguous alpha phase, as

shown in Fig 44. Thus, the flow stress in the area of aligned micro-

structure where many alpha/alpha boundaries must be sheared is likely to

be higher than in the more equiaxed areas where more shear along beta/

beta boundaries can occur. The reason for this is the relatively iCW

lattice and grain boundary diffusion rates in the alpha phase (see

chapter 2). A further factor, tending to increase the flow stress in the

aligned regions, is that the aligned grains are likely to be of a similar

orientation within each string of aligned grains with relatively narrow

grain boundaries and possibly even some low angle boundaries. Evidence

for this is that the alpha phase texture is much more intense in the

aligned regions than in the non-aligned regions (59) and the boundaries

between the contiguous, aligned grains are difficult to detect optically

as they etch slowly.

Quantitative analysis of the break-up of the contiguous alpha

phase is not easy because measurements have to be taken from several

42
areas that may not have been similar before straining. However, the

measurements of alpha phase aspect ratios in ref 61 and the

contiguous
microstructures in Fig 44 show that the aligned microstructure does break

up during straining and it is assumed that those areas most strongly

aligned before deformation are those which deform at the lowest rate.

For the TL and ST orientations the areas of aligned microstructure

remained relatively undeformed during testing (Figs 42 and 43) compared
with those areas where the two phases were homogeneously distributed.
Therefore, for these two orientations the measured flow stress
corresponded roughly to that of the flow stress of the equiaxed areas.
For the LT orientation, where the tensile axis was parallel to the banding
direction, the bands of aligned microstructure were constrained to deform

at the same rate as the surrounding areas and a relatively high flow

stress was measured (Fig 49). The orientation with the lowest flow

stress was consistently the ST orientation. This can be explained by the

observation that, although the alpha phase was most strongly aligned in

the L direction, there was also some alignment in the T direction (Figs

39 and 40) resulting in flow stresses in the order aL >a T >0 S'

The multiple necking that occurred in the ST orientation test

pieces pulled at 975*C (Fig 38) appeared to be the result of strain

localisation in bands where the alpha phase proportion was low at 975'C.

Deformation of these areas would be favoured because sliding of beta/beta

interfaces only would be required. The final beta grain size in these

areas was much larger than in adjacent regions where the alpha phase
retarded the beta grain growth. Presumably, this type of necking did

not occur in test pieces of other orientations because the beta phase
bands were not suitably oriented in relation to the tensile axis. It is

shown schematically in Fig 41 that the regions of aligned and non-aligned

microstructure were roughly planar in the LT plane. Thus, at 975*C the

beta rich regions, would lie in planes perpendicular to the S direction.

The variation in phase proportion suggested small variations in

chemical composition. Electron probe microanalysis of this material (59)

has not revealed any such variations. However, analysis (151) of similar
hot rolled Ti-6Al-4V bar has detected a slightly higher vanadium

concentration in the bands of more heavily deformed microstructure than

in other areas, possibly accounting for the origin of the banding. There-

fore, it is likely that the chemical composition of the 55 mm bar used here

was not uniform, although this may not be detectable throughout.

43
 It is to be expected that the beta rich areas would also exhibit
lower flow stresses than those areas low in beta phase for forming

temperatures lower than 9750C. Thus, as well as grain alignment, a

second factor that may contribute to the irregular superplastic strain
distribution is a variation in phase proportion.

5.2.2 Strain rate sensitivity

The m values at each testtemperature were lower for the high

1)
strain rate tests U=1.5 x 10-3 s- and Fig 36 shows that, for this

strain rate, the rate of neck development was higher than for the
1
corresponding tests at the lower strain rate of 3x 10-4 s- particularly
for the ST and LT orientation test pieces. The tests at 975'C resulted
in failure at a low strain, typically 150%, consistent with the low
strain rate sensitivities at this temperature (Fig 50). The m values
increased with temperature and reached a maximum at 925'C, when the phase
proportions were roughly equal. Above this temperature rapid beta grain
growth caused a sharp reduction in m value.

For the TL and ST-orientation tests the strain tended to be

localised in the areas of non-aligned microstructure. Therefore, the m

values measured on the TL and ST orientation test pieces corresponded

approximately to the m values of the non-aligned microstructure,

whereas for the LT orientation tests each type of microstructure deformed

at the same rate and an. "average" m value was recorded. Thus it appears
that the strain rate sensitivity of the aligned microstructure was lower
than that of the non-aligned microstructure.

5.2.3 R values

5.2.3.1 TL orientation

In the TL orientation the banding direction was across the gauge

width and R values measured on the bands, such as points Al-A3 in Fig 52,

were less than 1.0. It might be anticipated that the R values measured
at other points along the gauge length eg B1-B4 in Fig 52 would be

closer to 1.0 reflecting the more isotropic microstructures in these

areas. However, this was not the case and for test temperatures in the

range 800-925'C lower R values were measured in the non-banded areas.

The results on superplastic anisotropy of rolled Ti-6Al-4V sheet (section

4.3) showed that R values measured after superplastic strain were

influenced by the initial test piece shape. If a test piece of low

44
gauge length to width ratio was used then constraint by the test piece
heads restricted the strain in the width direction and low R values were
measured. Increasing the gauge length or reducing the width minimised
this geometrical effect. In the same way, R values measured in the
non-banded areas of the Ti-6Al-4V used here were influenced by the
adjacent, less deformed areas. An illustration of this effect is given
in Fig 66, which shows a segment of the gauge length of TL orientation

test piece A9 pulled at 875*C at a strain rate of 1.5 x 10-3 s.-1 The
effect of microstructure in the banded region was to resist strain in the
width direction and an R value of less than 1.0 was measured. To the
right of this region constraint by both the test piece head and the block

of strongly banded material has resulted in very low R values(O. 4), even
though the microstructure here was probably relatively equiaxed. To the
left of the banded area the constraint becomes progressively less and
the R values rise from 0.51 to 0.67. Thus, the R value measured depends
strongly on the position in which the measurement is made, which partly
explains the wide range of R values determined for the TL orientation
for test temperatures up to 925*C. This shallow "necking" or waisting
of the test piece was most pronounced for the low strain rate, 875*C
test (Fig 36c) and the difference in R value between banded and non-banded
areas was greatest for these test conditions (Fig 51). It is shown in
Figs 36a-c that for each test temperature the final shape of the TL test
pieces was much less irregular for the high strain rate tests than for
the low strain rate tests. Therefore, the geometrical constraint effects
invoked to explain the lower R values in the non-banded areas are likely
to be less important. Fig 51b, for the high strain rate tests, does not

show the trend of low R values in non-banded areas which is exhibited

for the low strain rate tests (Fig 51a).

The TJ, test pieces pulled at 975*C showed higher R values,

reflecting the low alpha phase proportion and much less directional

microstructure at this temperature.

Any contribution of alpha phase texture to anisotropy above 800*C

is likely to be small compared to that at room temperature and no effects

of texture on R value are expected for the TL orientation because the

tensile axis is parallel to the c axis and, therefore, alpha phase prism

slip cannot easily occur and neither thickness strain nor width strain is

preferred.

45
 5.2.3.2 ST orientation

For this test orientation the alpha grain alignment acted to

reduce the thickness strain during superplastic deformation and for test
temperatures of 875'C and 925'C all the R values were greater than 1.0.
However, the test piece surfaces were very irregular, as shown in Fig 37a.
Thickness measurements were taken on the surface protrusions and in the
depressions at random and, therefore, a wide variation in R was recorded

and any trend of R with strain el- was masked.

Results (65) on round test pieces from the Ti-6Al-4V bar pulled
in the S direction show that for a test temperature of 800%
the sense
of the strain anisotropy reverses for strain rates above about 4x 10-4 -1
s
For the sheet test pieces of ST orientation the texture of the bar is
such that <11h> prism slip would increase the thickness strain at the
expense of the width strain and reduce R values ie the effects of
texture would oppose those of the banding in the microstructure. Fig 51b
1
shows that, for a stra in rate of 1.5 x 10-3 s- at 800'C, R values ranged
'
from 0.9-1.1 indicating that the effects of texture were sufficient to

cancel those of microstructure.

The test pieces pulled at 9750C showed relatively low elongation

and incipient necking at several locations. As each neck began to

develop the deformation tended towards plane strain and the thickness
strain increased relative to the width strain. This reduced the local R

value, as shown in Fig 38. However, for a neck to form across the test
piece perpendicular to the tensile axis
yield the Von Mises
criterion
predicts (152) that the axial stress required is a= 2a IV3, where af
f
is the flow stress. Therefore, a neck of this geometry cannot develop
because the applied stress necessary exceeds that required for general
deformation along the gauge length. Under these circumstances final

necking occurs along a line inclined to the tensile axis and along which
the strain rate can approach zero. For isotropic material this necking
line makes an angle of 54.7" with the tensile axis. In both test pieces
in Fig 38 the growth of necks running perpendicular to the tensile axis
has been arrested at an early stage and neck development along a line
inclined to the tensile axis has been favoured. This type of localised

necking also occurred in the 800*C ST orientation test pieces for which
the strain rate sensitivity was relatively low (Fig 50).

46
 5.2.3.3 LT orientation

The LT orientation test pieces showed a more uniform strain

distribution along the gauge length than the corresponding TL and ST
test pieces. Consequently, there was less variation in R for each test
temperature (Fig 51). All the R values measured for this orientation
were less than 1.0 as a result of the small degree of grain alignment or
grain elongation in the T direction.

The R values corresponding to the 800*C tests at a strain rate

-3 -1
of 1.5 x 10 s were lower than those for the other test temperatures

at this strain rate. This was due to the influence of texture, which
for this orientation, tended to reduce R.

5.2.3.4 Comparisons with other work

The R value results on the sheet test pieces are, for the most
part, in agreement with the diametral strain ratio measurements made
(62,64,65) on the round test pieces machined from the same bar. An

exception to this was the round test pieces pulled in the S direction;

those regions of the gauge length containing aligned microstructure were

found (64,65) to deform more anisotropically than the non-aligned areas.
It was not possible to discriminate between aligned and non-aligned areas
in R value measurements made on the ST orientation sheet test pieces.
However, R values for the different microstructures were recorded for

the TL orientation tests and showed, on average, more anisotropic

deformation in the non-aligned areas (Fig 51a) ie the opposite of the
behaviour predicted by the round test piece results. This can be explained
by the constraints present in the sheet test pieces, as discussed in

section 5.2.3 and illustrated in Fig 66. Similarly, the results for the

S orientation round test pieces showed (64,65) anisotropy at 970'C in the

opposite sense to that indicated in Fig 51 for the ST orientation sheet

test pieces at 975*C. The reason for this is the unusual necking behaviour

related to the beta rich phase bands, which influenced the R value measure-

ments sufficiently to reverse the apparent sense of anisotropy. This type

of necking, approaching plane strain, can only occur in sheet test pieces.
Thus, the use of sheet rather than round test pieces does affect the high

temperature anisotropy in this material.

Superplastic anisotropy of rolled Ti-6Al-4V sheet of various

thicknesses is discussed in section 5.3. The sheet microstructures were

much less directional than that of the bar. Nevertheless, R values as

47
low as 0.5 were measured with the lowest R values occurring in the sheet
with the highest alpha phase aspect ratio. Thus, although the rolled
sheet microstructures were homogeneous ie without bands of aligned and
non-aligned microstructure, the superplastic anisotropy was similar in

nature to that of the bar material.

5.3 R values of Ti-6Al-4V sheet

5.3.1 Effect of sheet thickness

It is apparent from Figs 53 and 54 that superplastic strain

occurred most easily in the direction in which the average alpha grain
width was smallest, and that the anisotropy was most pronounced for
high alpha phase aspect ratios. The thickness effect shown in Fig 53
(R decreasing with increasing sheet thickness) was therefore observed
because the thinner sheets exhibited the most isotropic microstructures,
presumably as a result of recrystallisation during rolling or during
annealing at the superplastic temperature before straining. The effect
of superplastic strain in the L direction on the LS section microstructure
of the sheet of initial thickness 3.3 mm is shown in Fig 18 (compare with
Fig 17a) and Fig 19 shows that the alpha phase aspect ratio decreased
with superplastic strain. This is reflected in the increase of R with
strain (Fig 53).

This behaviour is consistent with the superplastic strain

anisotropy shown by heavily banded Ti-6Al-4V bar. It is significant that

such anisotropy can occur in material processed for isotropic super-

plasticity and emphasises the need for sheet production techniques

capable of achieving equiaxed microstructures.

5.3.2 Effect of test piece geometry

The curves in Fig 55 show that the constraints due to the test piece
heads were more pronounced for larger gauge widths (Fig 55a) and smaller
gauge lengths (Fig 55b). To ensure that geometrical constraints are
minimised and natural width and thickness strains are measured Fig 55a
indicates that the length to width ratio of the gauge length should be

1.25: 1 or greater (corresponding to the test piece of gauge length 10 mm

and gauge width 8 mm in Fig 55a) and according to Fig 55b it should be

at least 1.6: 1 (corresponding to the test piece of gauge length 25 mm

and gauge width 16 mm in Fig 55b). Therefore, for the 3.3 mm sheet used
1a
in this investigation at 925'C and a strain rate of 3.1 x 10-4 s- test

piece aspect ratio of 1.6: 1 or greater is required for the determination

48
of R as a unique function of strain. The limiting superplastic strain
to which meaningful R values can be measured depends on the strain at
which the variation in strain rate along the gauge length becomes

significant. For specimen 9 in Fig 56d this corresponded to a true

strain of about 2.0 (639%), although this in turn depends on the strain
rate sensitivity of the material and the machining tolerance of the test
piece as any waisting or irregularity before testing will cause some
localisation of strain.

For material of lower strain rate sensitivity than the Ti-6Al-4V

used in this work the influence of the test piece head would be greater

and the tapering of the gauge length more pronounced as the tendency for
the strain to localise in the narrowest region of the gauge length would
be stronger. In such cases the use of a high aspect ratio test piece
(large length to width ratio) becomes increasingly important. This is
illustrated by the R value measurements made on the Ti-15V-3Cr-3Al-3Sn
(Fig 58), which indicated some anisotropy despite the fact that the
starting microstructure in this alloy appeared to be completely isotropic
(Fig 34a). The deviation of R from unity was therefore due to geometrical
constraint even though the gauge aspect ratio was sufficiently high
(1.6: 1) to avoid significant end effects in the Ti-6Al-4V sheet test

pieces. It can be seen from Fig 33 that the strain variation along the
gauge length of a typical Ti-15V-3Cr-3Al-3Sn test piece was considerable
and that the centre of the gauge length was not free from the influence

of the test piece heads.

The standard procedures (153-155) for room temperature R value

measurement specify test pieces of high aspect ratio, typically 10: 1,

and of length approximately 200 mm. This aspect ratio is sufficient to

minimise the geometrical effects noticed during superplastic R value

measurement (Fig 55), but in practice these standard test pieces would
be too long for uniaxial superplastic testing when strains of several
hundred percent may be required (the hot zone of a laboratory furnace

being typically 100 mm long). The test pieces used in other work on the

superplasticity of Ti-6Al-4V eg refs 55,124,148,156 have gauge aspect

ratios of 1.6: 1 or greater and therefore, based on the criterion

established in this investigation, would have been suitable for R value

determinations.

49
 The choice of a standard test piece for the measurement of R
values after superplastic strain is therefore a matter of compromise;
a high aspect ratio is required in order to minimise end effects, but
there are practical restrictions on the gauge length and width.

The divergence of the two sets of curves in Fig 57 at e1 <1.5

arose because the width and thickness strain measurements on the failed
specimen were made close to the end of the gauge length. Further along
the gauge length ie at higher e the curves converged as the constraint
imposed by the test piece head became less pronounced, but for 61 >2.0
the curves diverge again. It is likely that as the shallow neck
developed during testing the local
rate increased such that
strain
deformation by slip became significant (61,64,65). It was shown in Fig
15 that the texture intensity of the alpha phase in Ti-6Al-4V decreases
slowly with superplastic strain. Therefore, deformation of Ti-6Al-4V at
the superplastic temperature by slip even following a certain amount of
superplastic strain can be influenced by the initial alpha texture. The
3.3 mm sheet had a basal edge texture (Fig 13) and therefore through-
thickness slip was preferred in L orientation test pieces. This is

consistent with the behaviour shown in Fig 57 for e1 >2.0. It is

expected that if these high strains could be achieved in an untapered

test piece eg by remachining a parallel gauge length at eI=2.0 then
Cw and et would lie on the extrapolated (dashed) curve in Fig 57.

Another feature of the curves in Figs 55 is that anisotropic

straining only occurred for E, <1.0. At higher strains ew and et lie

parallel to the line depicting isotropic behaviour. The gradient de /del

w
for the 4 mm gauge width in Fig 55a increased from 0.41 at eI=0.5 to

close to 0.5 (isotropy) at eI=2.0. This behaviour, representing R

value increasing with strain, is related to the reduction in alpha phase

aspect ratio on the TS section (as shown in Figs 17-19 for the LS section).

The definition of R based on total s.trains, =sw /E: means that

t,
_R
R values measured after high, superplastic strains are influenced by the

anisotropic strain which is accumulated for eI less than about 1.0.

Consequently, even when the microstructure has become equiaxed and the

material behaves isotropically this is not reflected in the R value which

remains less than 1.0. An alternative definition (106) of R using

incremental strains, R' = de Me provides a more useful description

w t,
of the normal anisotropy in those cases where R changes with strain.

50
R values calculated from Fig 55a (using data from the 4 mm gauge width
test pieces) are plotted against strain in Fig 67. Approximate values
of R' calculated from R are also given. The curve for R' is initially

steeper than that for R and approaches R' = 1.0 which indicates the
nearly isotropic behaviour of the material at high strain. Fig 67 also
includes the average R' value R' = (R' L+ R' )/2.
T
The superplastic flow stress data (section 4.1) for each of the
alloys indicate very little anisotropy in the plane of the sheet and the
L and T orientation data for the IMI 550 sheet are almost identical
(Fig 20). The flow stress in the two orthogonal directions in the plane
of the sheet aL and aT can be related (106) to the two corresponding
R values R' and R' as follows:
L T

CY
L1+R"
crTR +R
T(l LT)

Substitution of R' and R' Tý0.5 from Fig 67, which gives data
Lý0.6
for the Ti-6Al-4V sheet, produces aL /(y Thus, even though the
Tý1.06.
microstructure is measureably anisotropic in the plane of the sheet
(Fig 54), the superplastic flow stress is expected to be only marginally
greater in one direction. The IMI 550, for which the flow stress at
900*C was isotropic in the plane of the sheet (Fig 20), also showed very
little variation in R between the L and the T directions (Fig 58).

5.4 Application of uniaxial data to hemisphere forming

Plastic anisotropy in sheet metal can, in certain circumstance, improve

cold formability (106,157,158) and, along with the strain hardening

exponent n, the plastic strain ratio R is often measured to determine

cold forming characteristics. For cold sheet forming the normal anisotropy
is particularly pronounced for strongly textured hcp alloys (105,107)

where thinning of the sheet can be inhibited by appropriate orientation

of the slip systems. Under these conditions "texture hardening" (159)

occurs and critical thickness strains for failure (107) can only be

obtained at relatively high total strains.

Similarly, superplastic formability can be characterised by R

values, strain hardening and strain rate sensitivity. However, under

conditions the strain is accommodated largely by grain

superplastic
boundary sliding and slip and crystallographic texture may be of much
51
less importance. Consequently, flow stresses are mainly dependent on
grain size and shape and anisotropic flow during forming is due to
directionality in the microstructure. The R value gives an indication

of the through-thickness strength of the sheet relative to the in-plane

strength. Biaxial stretching is equivalent to uniaxial thinning.

Therefore, the main effect of R on bulge forming operations is to
influence the gas pressure required for deformation.

This section describes the application of the strain hardening,

strain rate sensitivity and R value data determined uniaxially on 3.3 mm

Ti-6Al-4V sheet to the forming of a hemisphere using the concepts of
equivalent stress and equivalent strain. The influence of each of these

parameters is considered and an optimum pressure cycle is given for a

constant strain rate is at the pole of the dome. Although the hemisphere

has several practical applications and has already been the subject of

analysis (160-165) the effects of superplastic anisotropy have so far

not been considered.

A statement of the equivalent stress function ae derived (106)

from Hill's general analysis (166) for plastically anisotropic material
is:

R1( cy CY R1 (a a3- CT
1- 2) 2-a 3) 1
cy a fl
e 1ý (1 +R1)1+R1
-1+R1R2

where gl, a 2' and cy3 are the principal stresses with a3 acting through

the thickness, a fl is the flow stress in the X1 direction and R1 and

R2 measured along the X1 and X2 axes respectively are:

dE dE
2 1
R' R'
dE LR2 dE T
3 3

If thin shell properties are assumed (162) then cr3 is small and
can be ignored. At the pole of a hemisphere a, =a2 and equation (2)

becomes:

R' R'
L+ T
CY CrLý -1
(3)

aL is the flow stress in the L direction.

where

52
 The tangential sress at the pole of a hemisphere is given by:

Pr
a=- (4)
Zt

where P is the pressure, r is the radius of curvature and t is the

thickness. Combining (3) and (4):

2ta R' R'

L L+ T
(5)
r R' (1 + R'
T L

where each of the variables is a function of strain. The experimentally

determined values of R' R 'T (Fig 67) and flow stress, aL (Fig 4) of
L,
the 3.3 mm thick sheet were used to evaluate equation (5) for various
stages of forming of the hemisphere. The calculation was based on a
hemisphere diameter of 200 mm and the radius of curvature, r, was
calculated for each strain increment by assuming that the bulge surface
formed part of a sphere. The polar thickness at each stage of forming

was estimated by interpolating the results of Cornfield and Johnson (162)

who calculated bulge thickness profiles for various strain rate sensiti-
vities. An m value of 0.67 (from Fig 7) was used.

The strain hardening, based on equivalent strain at the pole Ee

was calculated for each thickness strain increment as follows:

The work done for a small strain increment is given by the sum
of the work done in each of the principal directions ie

d
cy F- CY F- F- F- 2a F-
= 1d 1+G2d
2+G3d 3= 1dE1=-a1d 3

Substituting in equation (3) and assuming proportional straining

(106) :

[R 'L(l
+ R'
T)
R'L+ R'
T

where e3 is the thickness strain.

53
 The strain increments were related to forming time by imposing
1
a condition of constant strain rate of 3.1 x 10-4 s- at the pole. This

strain rate corresponded to maximum strain rate sensitivity during

uniaxial testing.

Fig 68 shows four pressure cycles calculated for the forming of

a hemisphere. For each curve different material properties were assumed
as follows:

(1) Uniform thinning was assumed, a constant flow stress of 9 MPa

and an R value of 1.0 (isotropic deformation) were used. This

type of calculation is probably typical of many used commercially.

(2) A strain rate sensitivity of m=0.67 was imposed. A

constant flow stress of 9 MPa and an R value of 1.0 were assumed.

(3) A strain rate sensitivity of m=0.67 was used and strain

hardening was included. The R value used was 1.0.

(4) As in (3) except that the R values R' and R' were assumed
L T
to vary with strain. This calculation makes full use of the
uniaxial experimental results.

Compared with the uniform thinning case the effect of including

the strain rate sensitivity was to extend the forming time and reduce the

maximum pressure (Fig 68), curve 2). Introducing strain hardening

caused a rise in the maximum pressure and displaced the peak to a longer
time (Fig 68, curve 3). Inclusion of the experimentally determined R

values in the calculation caused a reduction in pressure (Fig 68, curve (4).

The choice of a fixed flow stress of 9 MPa for two of the curves resulted
in an overestimation of pressure for the first half of the cycle and an

underestimation for the second half when compared with the curves
incorporating strain hardening. It is significant that for times up to
20 minutes the greatest discrepancy in pressure (up to 40%) was between

curve I for uniform thinning and curve 4 incorporating all the measured

variables.

The influence of strain rate sensitivity on the pressure cycle

for the forming of the hemisphere was further investigated by assuming

m values of 0.5 and 1.0 in the calculation including strain hardening

and anisotropy. The results are given in Fig 69. All the calculated

pressure cycles lie close to the curve drawn through the points for

The only significant difference between the pressure cycles is

m=0.67.

54
that forming time increases with decreasing m value. The optimum
forming cycles for the range of m values typical of Ti-6Al-4V are,
therefore, approximately the same despite the fact that the final

thickness profiles, as calculated by Cornfield and Johnson (162) vary

significantly. In practice the strain rate sensitivity of the sheet may
be more significant factor in minimising thinning at the pole than

adherence to the optimum pressure cycle, although the two factors are
related by the sensitivity of m to strain rate variations.

The other set of points included in Fig 69 are those calculated

assuming an average R value R' = (R' L+ R' )/2.
T
These points also lie

close to the curve and indicate that a single R value (varying with strain)
can be used to describe the material. Traditionally R is also measured
at 45' to the rolling direction and a mean R value defined (154) :
T= (R +R )/4. A simpler definition suffices for R values after
0+R 45 90
superplastic strain because the normal anisotropy does not vary substan-
tially between different directions in the plane of the sheet.

The pressure calculations have been based on two main assumptions:

(1) The in-plate stresses a1 and a2 are equal. In practice,

these two stresses would be slightly different as a result of the

planar anisotropy in the sheet, as discussed in section 5.3. This

means that the bulge profile would not be symmetrical about the

pole. However, Fig 69 shows that using the average R value R',

rather than R' and R' in the calculation does not significantly
L T
affect the pressure cycle. Thus, the effects of planar anisotropy

are minimal.

(2) Strain rate sensitivity does not depend on strain. Experiment

shows (Fig 10) that this is approximately true for strains up to

1.0 and Fig 69 indicates that varying m between 0.5 and 1.0

does not substantially affect the calculated pressures.

In view of the assumptions that have been made the calculated

pressure cycles can only be considered as approximations and strictly

only apply to the material from which the uniaxial data was derived at
925'C. Nevertheless they serve to illustrate the underlying principles

and demonstrate the influence of the various material parameters.

5.5 Analysis of necking of Ti-6Al-4V during superplastic deformation

The computer analysis of neck development has allowed the material

55
properties inside the neck to be estimated separately for
and outside
each stage of deformation. Such results couldnot be achieved directly

by experiment.

It was noticed that the neck strain rate decreased slightly

during the first few percent of deformation (Fig 61). This was the case
for each of the neck sizes examined and occurred because, in the early
stages of deformation, the strain in the neck'was similar to that in

the uniform region and a relatively high neck strain rate was required
to balance the load. Further deformation led to greater strain
hardening in the neck compared with the uniform region and a small

reduction in neck strain rate occurred.

The two terms on the right hand side of equation (1) represent
the contributions of strain and strain rate changes to the overall
increase in flow stress with strain. Thus, Fig 64 illustrates the

relative importance of strain rate sensitivity and strain hardening in

resisting neck development. For the first part of the deformation, when

neck strain rate changes were minimal (Fig 61), the strain hardening term
is largest, but beyond about 85% uniform strain the strain rate sensitivity
is dominant. However, strain hardening also occurs in the uniform

region of the gauge length and for much of the deformation this almost

exactly cancels out the strain hardening in the neck (Fig 64). The

uniform strain at failure or "limiting strain " is therefore strongly

influenced by the strain rate sensitivity and particularly by the residual

m value after high strains. The initial m value is likely to be of

little consequence by comparison. The practical significance of this

for superplastic forming is that an alloy, for which grain coarsening

is relatively slow, may well demonstrate more superplastic ductility

than another that exhibits higher m values at low strains, but is less

resistant to grain coarsening. Similarly, a reduction in forming

temperature may reduce the initial strain rate sensitivity, but does

reduce the limiting strain. Other authors (167-172)

not necessarily
have remarked on the strong influence of rate sensitivity in resisting

neck development particularly in the "post-uniform" stage. The results

presented here apply only to superplastic Ti-6Al-4V with a relatively

large pre-existing defect in the gauge length, but clarify the significance

of m and n.

56
 Ghosh (172) carried out a numerical analysis of neck development
incorporating strain and strain hardening based on the relation
n6m directly
a=K, The results of this work are not comparable with
.
those presented here because Ghosh only investigated n values as high
as 0.2. However, discrepancies would be expected because Ghosh assumed
constant values of m and n in his analyses which would clearly be a

gross approximation for superplastic Ti-6Al-4V, as shown by Fig 63.

Arieli and Mukherjee (173) suggest that total superplastic alongations

are more closely related to the terminal strain rate sensitivity than to
the initial value. This is confirmed by the present work as discussed
above.

The point of instability during tensile deformation has been

the subject of some discussion (167-172,174-176). According to Hart

(167) instability is reached when y+m=1, where y is a strain
2tlna
hardening parameter _j( - The criterion for instability in this case
E: 91ne)

is (SA/6A) >0. Where 6A is the variation in cross section area between

p
the necked and the uniform regions and the condition of constant load p

simply reflects that each section of the gauge length bears the same load

at any one time. Duncombe (168,169) chose the point at which the

variation in strain rate 6(A/A) increased as the point of instability

and Ghosh (172) chose the point of maximum load, which for constant strain

rate testing, is equivalent to the Considere criterion (y = 1). Each of

these definitions can be applied to superplastic deformation as follows:

Hart (167) discussed the point of instability by considering

small variations in strain and strain rate along the gauge length
associated with small inhomogeneities or incipient necks. However, it

was assumed that, at any instant, the values of y and m did not vary
along the gauge length. Therefore, as a neck develops the stability
criterion 6ý/6A 0 will be represented less and less accurately by the

expression y+m1, whether or not the deformation is considered

stable. For superplastic materials, where a long period of neck
development precedes failure or where a large defect exists prior to
deformation, this inaccuracy is likely to be significant. The value of
6ý/6A is shown in Fig 70 as a function of strain along with the value

of y+m (for the uniform region). The indicated points of instability

occur at significantly different amounts of strain (50% and 150%

respectively). Furthermore, the value of y+m decreases monotonically

57
towards instability whereas the curve for 6ý/6A indicates a region of
increasing stability up to 15% uniform strain. The sign of 6A will
not normally change during deformation and the 6ý/6A =0 criterion can
be more simply expressed as the point at which the rate of change of
area is the same in the uniform region and in the neck. Thus, Hart's

criterion can be applied provided that the general form 6A/6A =0 is

used rather than the derived form y+m=1.

The maximum load occurs when y=1. It also follows, from the
definition of n that 6=n at this point. Examination of Figs 61 and
63 shows that at maximum load e aLn aLO. 52. The value of n is often
assumed to be constant during room temperature testing, but this is not
necessary in order that the equality above is satisfied at maximum load.
However, it is important to note that the condition of constant strain
rate in the uniform gauge section implies that maximum load would occur
at the same uniform strain regardless of the size of the pre-existing
defect.

Applying the definition of instability 6(A/A)> 0 implies that

deformation is unstable from the outset in the present case. However,

neck strain rate decreased during the early stages of deformation.

Therefore, choosing the point at which strain rate variation (6ý/A)

increases (about 10% uniform strain) seems more sensible in this case.
Hutchinson and Obrecht (174) used the area variation 6A to monitor neck
development in rate sensitive materials. For the case presented here 6A

decreased with strain and reached a minimum at about 73% uniform strain.
Thus, increasing 6A could also be used to define instability.

Therefore, the definition of tensile instability is very much

open to interpretation and for the case discussed here a wide range of

values of instability strain are predicted, as summarised in Fig 70.

None of the definitions can be applied without detailed knowledge of

the neck development except for the maximum load criterion, which takes

no account of the defect size when constant strain rate in the uniform

region is assumed.

The imaginary specimen considered here is of infinite length with

the neck developing in a pre-existing defect. Real specimens may contain

such a defect, but usually necking takes place at the centre of the

gauge length where the material is least constrained. Tensile failure

is also likely to be influenced by cavitation, even for the superplastic

58
deformation of titanium alloys, and localised rather than diffuse necking

will probably occur at some stage depending on the specimen shape and
isotropy. No account has been taken of stress components on the neck

other than the axial component ie the neck radius is considered to be

infinite. It is also assumed that a constant strain rate is maintained
in the uniform region of the specimen. However, in practice, if the

average strain rate is maintained constant over the gauge length such
that ý= V/1, where V is the extension rate and I is the instantaneous

gauge length, then the strain rate in the uniform region will reduce as
the neck develops.

A further assumption is that the flow stress of the material can

be described by the curves in Fig 4 regardless of its strain history.

This is roughly equivalent to assuming that the effects of static

annealing on grain size and hence on flow stress are negligible compared

with the effects of strain. Other work (138) shows that this assumption
I
is reasonable, particularly as no strain rates lower than 3x 10-4 s-

are considered.

59
6 CONCLUSIONS

1 Superplastic deformation of the alpha/beta titanium alloys

Ti-6Al-4V, IMI 550 and Ti-8Al-1Mo-1V occurred largely by grain boundary

sliding. This produced random grain rotations leading to reductions in

alpha and beta phase texture intensities. Observations of grain

emergence from test piece surfaces were made.

2 The highest m values recorded for the Ti-6Al-4V and Ti-8Al-lMo-lV

alloys were at temperatures corresponding to equal phase proportions.
At higher temperatures the m value of the Ti-6AI-4V was sharply reduced
as a result of rapid grain coarsening.

3 The sheet test pieces machined from Ti-6Al-4V bar exhibited lower

strain rate sensitivity and higher flow stress in areas of microstructure

containing contiguous, aligned alpha grains than in non-aligned regions.
Therefore, the R values were influenced by directionality in the

microstructure and hence by test piece orientation. The sense of the

superplastic strain anisotropy for each orientation was consistent with

the measured order of flow stress at 875-975*C; a I- >aT >0
S*
4 The R values of the rolled Ti-6Al-4V sheets were influenced by

directionality in the microstructure. The lowest R values were measured

when the alpha phase aspect ratio on the section perpendicular to the

tensile axis was highest. The R values of Ti-6Al-4V sheet increased

with superplastic strain as the microstructure became more equiaxed.

5 The R values of the Ti-6Al-4V rolled sheet test pieces were also
influenced by the effect of the test piece heads on the width contraction

of the gauge length. This effect was reduced by increasing the length or

reducing the width of the gauge section. A gauge length to width ratio

of 1.6: 1 was apparently sufficient for the measurement of R values without

significant test piece eftd effects.

6 The only observed influence of crystallographic texture on R value

was for the Ti-6Al-4V sheet test pieces machined from bar and occurred

at the lowest test temperature (800'C) and at the highest strain rate
1
(1.5 x 10-3 s- ).

7 During superplastic deformation of the beta alloy Ti-15V-3Cr-3Al-3Sn

subgrains developed and sliding on subgrain boundaries occurred. This

refined the grain size and led to an increase in m value with

process

60
strain. The superplastic flow stress of the Ti-15V-3Cr-3Al-3Sn was
high and the m value low by comparison with the alpha/beta alloys.
This was a result of the large initial grain size. The texture intensitv

of Ti-15V-3Cr-3Al-3Sn was not reduced by superplastic strain.

8 Annealing at the forming temperature caused recrystallisation and

a reduction in room temperature strength in Ti-6Al-4V, IMI 550 and
Ti-8Al-1Mo-1V. Superplastic strain generally caused a further slight
loss of strength in these alloys. The room temperature 0.2% proof stress
and tensile strength of IMI 550 after superplastic deformation was
raised by increasing the cooling rate from the forming temperature and
by ageing. Room temperature strengths up to 18% greater than those of
as-formed Ti-6Al-4V were achieved in IMI 550 by heat treatment after
forming. Superplastic deformation reduced the room temperature strength
of Ti-15V-3Cr-3Al-3Sn.

9 The calculated pressure cycles for the forming of a hemisphere with

a constant strain rate at the pole were significantly altered by the
inclusion of uniaxial experimental strain hardening and R value data.

Variation of strain rate sensitivity in the range 0.5-1.0 produced only

minimal changes in the optimum pressure cycle.

10 The analysis of necking in superplastic Ti-6Al-4V showed that, for

the case examined, strain hardening controlled the neck development

during the early stages of deformation and strain rate hardening became

the most significant factor at a later stage. The application of

different definitions of instability produced widely differing values

of uniform strain at the point of instability.

This thesis has paid particular attention to the causes of

anisotropy of superplastic deformation in both sheet and bar material.

The significance of microstructural directionality was established and it

was shown that., for the alpha/beta alloys, crystallographic texture did

not cause anisotropy under superplastic conditions.

12 The rolled sheet alloys Ti-6Al-4V, IMI 550 and Ti-8Al-1Mo-1V and

Ti-15V-3Cr-3Al-3Sn were assessed for superplastic formability and it was

that the alpha/beta alloys were superior to the beta alloys in this
shown

respect. The IMI 550 exhibited high strain rate sensitivity and relatively
high room temperature strength after forming, especially after post-
forming heat treatment. Therefore, this alloy appears to be particularly

attractive for superplastic forming operations.

61
7 SUGGESTIONS FOR FURTHER WORK

1 More detailed study of the role of each of the phases in the

superplastic deformation of two phase alloys is necessary. Light

microscopy is restricted to examination of the room temperature

structure after forming. Therefore, more in-situ observations of the

deformation, such as those by Ma and Hammond (44) using the

photoemission electron microscope (PEEM) are required.

2A detailed model of superplastic deformation in two phase materials,

leading to a new constitutive equation, is required. More accurate
grain and phase boundary diffusion and high temperature shear modulus
data would be needed in order to reliably predict the superplastic
strain rates of alpha/beta titanium alloys.

3 The IMI 550 alloy appears to be an attractive candidate for sheet

forming operations. However, more post-forming room temperature

mechanical property data, such as fatigue resistance and fracture

toughness, is required to confirm this.

It would be desirable to carry out a series of biaxial forming

trials of hemispheres and other shapes to further investigate the

application of uniaxial flow stress, m and R value data to gas pressure

forming. The purpose of this would be to develop a technique for

predicting the optimum forming pressure cycle for any shape.

5 The analysis of neck development has concentrated on the case of

constant strain rate in the uniform section and a neck of infinite

radius. In order to make the results more representative of real

specimens it would be desirable to allow the uniform strain rate to

decrease such that the average strain rate rather than the uniform

strain rate was maintained constant as discussed above. Another important

case to investigate is that of uniform extension rate, in which case an

imaginary gauge length would need to be specified. To incorporate the

effect of a finite neck radius the specimen length would have to be

divided into segments and the equivalent stress and strain calculated
for each segment for each strain increment as in the approach adopted by

Ghosh (171). However, superplastic failures do not normally show

substantial localised necking and this elaborate method may not be necessary.

62
ACKNOWLEDGEMENTS

I am extremely grateful to my University supervisor, Prof Alan

Crocker, for his help and encouragement and for many useful discussions.

My RAE supervisor was Dr Peter Partridge. Without his advice and

considerable enthusiasm this thesis would not have been possible.

I would like to thank the RAE for permission to collaborate with

Surrey University in this way and for printing and typing facilities for

the preparation of this thesis.

I am also grateful to Dr Alun Bowen and to Chris Gilmore for

preparation of the pole figures, to Bob Butt, Ian Porcher and Nigel
Kennet for assistance with the tensile testing and to Dr Malcolm

Ward-Close who wrote the curve-fitting computer subroutine.

63
 Table 1

EFFECT OF SUPERPLASTIC STRAIN AT 925*C AND 3x 10-4S-l ON THE TENSILE

PROPERTIES OF Ti-6Al-4V SHEET. INITIAL SHEET THICKNESS 3.3mm

SP Eu.% Et %
Orien- 0. lps 0.2PS 0.5PS TS E
Strai-n on on
tation MPa MPa MPa MPa GPa
% 20mm 24mm

700*C 2 hrs,, furnace cooled (as-received)

T 1053 1051 1043 1090 124 8.8 16.3

L 938 933 935 1003 106 - 17.9

Annealed at the forming temp 925'C Jhr, cooled 25*C min-'

T 945 935 928 1025 127 10 20.0

.8
L 825 816 818 914 107 9.8 20.0

Eu% Et%
Superplastically formed, cooled 25*C min- on on
lomm 14mm

T 32 917 917 921 1023 129 10.4 18.5

T 58 918 916 916 1003 131 11.0 -

T 128 919 919 922 987 121 8.5 11.5

L 18 847 845 847 938 105 9.2 -

L 153 807 809 814 910 95 10.3 20.0

PS Z proof stress, TS Z tensile strength, E -ý Young's modulus,

Eu = uniform elongation, Et total elongation
-=

64
 Table 2

EFFECT OF HEAT TREATMENT ON THE TENSILE PROPERTIES OF IMI 550 SHEET

Eu% Et%
Heat Orien- 0.1PS 0.2PS 0.5PS TS E
Treatment MPa MPa MPa MPa GPa 0 pl on
tation
20uun 25mm
1
700'C 2hr, air cooled T 1034 1052 1060 1060 123 5.8 14.3
(as received)
L 989 997 1011 1079 113 4.2 13.5

T 1025 1033 1034 1034 112 5.6 11.9

900% ihr, cooled
25'C min-1 L 953 948 948 1014 107 5.8 18.8

900% Jhr, cooled T 1096 1104 1111 1121 121 5.6 12.4
25% min-1
aged 500'C 24hrs L 1033 1028 1033 1118 111 4.9 14.5

T 1062 1088 1109 1125 118 4.3 13.2

900*C 1hr, cooled
150*C min-' L 997 1013 1027 1116 108 5.0 15.0

900*C Jhr, cooled T 1129 1142 1155 1174 122 4.8 12.5
150*C min-1
aged 500'C 24hrs L 1070 1072 1082 1175 114 4.9 13.6

T 1002 1016 1025 1044 121 6.0 17.8

900*C 31hrs, cooled
25*C min-1 L 922 922 928 1017 108 6.2 18.1

T 1023 1055 1082 1117 119 5.4 15.9

900'C. 31hrs, ' c-ooled
15GO'C min--" L 963 984 1000 1096 110 5.3 18.1

Each result is the average of two tests

65
 Table 3

EFFECT OF SUPERPLASTIC STRAIN AT 900'C AND 3x 10-4S-1 ON THE

TENSILE PROPERTIES OF IMI 550 SHEET

SP
IM
Et Z
Orien- 0. fps 0.2PS 0.5P S TS E
strain ýI on on
tation MPa MPa MPa I MPa GPA lomm 21mm
%

L 44 809 821 838 902 104 3.2 5.8

L 91 899 907 917 994 105 3.9 7.6
L 123 931 935 941 1027 104 5.2 8.7
L 190 832 841 848 926 98 5.3 7.5

T 37 961 959 959 1020 112 5.2 8.8

T 82 941 947 959 1001 ill 3.5 4.8
T 120 926 921 922 978 105 4.6 7.8
T 169 960 965 972 1003 121 2.7 5.9
T 172 963 974 963 1019 113 6.0 7.5

66
 Table 4

EFFECT OF POST-FORMING HEAT TREATMENT ON THE TENSILE PROPERTIES OF

IMI 550 SHEET (L ORIENTATION)

Sp Et%
Post-forming O. 1PS 0.2PS 0.5PS TS E
heat strain on on
treatment MPa MPa MPa MPa GPa
% 1Ornm 12mm

Cooled 25*C min- 190 832 841 848 926 98 5.3 7.5

Cooled 25*C min-

172 983 995 1010 1091 108 5.5 11.0
aged 500'C 24 hrs

171 942 945 948 1024 108 5.4 9.3

900*C 1hr, cooled 181 907 923 957 1035 98 3.0 10.8
150*C min-1

900*C ihr,, cooled

150*C min-1, aged 183 1004 1020 1040 1127 108 5.4 9.5
24hrs 500*C

191 997 1017 1037 1123 112 3.0 9.0

67
 Table 5

EFFECT OF SUPERPLASTIC STRAIN AT 940,970 AND 1010% AND 3x 4

10- S-
ON THE TENSILE PROPERTIES OF Ti-8Al-lMo-lV SHEET

orming/ SP E % E %
Orien- 0.1PS 0.2PS O. 5PS TS E U t
Annealing strain on on
tation MPa MPa MPa MPa Ga
temp"C % lomm 15mm

790'C 8hrs, furnace cooled, 790*C zhr, air cooled (as-received)

I-1 1 1 1
TI - 868 877 882 975 125 9.7 17.2

Annealed at the forming temp Jhr, cooled 25*C min-'

1
T 940 831 836 838 939 121 F8
10.2 .2
T 970 826 826 826 931 127 10.7 19.1

T 1010 823 823 823 921 125 11.7 22.1

Ft, % Et %
Superplastically formed, cooled 25*C min on on
20mm 25mm

T 940 187 821 842 863 877 126 5.2 15.4

T 940 267 784 808 829 908 119 6.2 12.0

L 940 221 787 806 850 913 116 8.2 15.8

T 970 210 773 778 789 844 120 8.0 11.8

T 1010 184 783 795 809 881 122 7.7 11.9

68
 Table 6

EFFECT OF SUPERPLASTIC STRAIN AT 810,860 AMD 9100C AND 3x 10-4S-1

ON THE TENSILE PROPERTIES OF Ti-15V-3Cr-3Al-3Sn SHEET

Forming SP Fil 7* Ft 7.
Orien- 0.1PS 0.2PS 0.5PS TS E
temp strain on on
tation MPa MPa MPa MPa GPa
OC % i20mm 25

T as- 787 787 785 802 89 23.3

received
L 754 762 762 779 81 25.9

T 340 690 694 696 724 76 2.0 10.3

810
L 348 628 628 632 648 74 2.5 6.6

T 316 657 660 663 698 77 3.4 9.7

860
L 275 639 639 639 667 72 4.7 10.5

T 310 636 638 641 674 74 2.2 6.4

910
L 272 631 643 648 681 71 4.0 7.7

* Average of two tests

69
 1 K. A. Padmanabhan Superplasticity.
G. J. Davies Springer-Verlag (1980)

2 J. W. Edington Superplacticity.
K. N. Melton Prog. Mat. Sci. 21,61-170 (1976)
C. P. Cutler
3 J. Gittus Superplasticity: a review of data.
Res. Mech. 7,127-201 (1983)

4 R. H. Johnson Superplasticity.
Met. Reviews 15,115-134 (1970)

5 C. Herring Diffusional viscosity of a polycrystalline solid.

J. App. Phys. 21,437-445 (1950)

6 R. L. Coble A model for boundary diffusion controlled creep

in polycrystalline materials.
ibid. 34,1679-1682 (1963)

7 R. Raj On grain boundary sliding and diffusional creep.

M. F. Ashby Met. Trans. 2,1113-1127 (1971)
8 P. Griffiths Superplasticity in large grained materials.
C. Hammond Acta Met. 20,935-945 (1972)

9 H. Naziri Extended plasticity inoommercial purity zinc.

R. Pearce,. J. Inst. Metals 97,326-331 (1969)

10 H. Naziri The effect of grain size on work hardening and

R. Pearce superplasticity in Zn/0.4% Al alloy.
Scripta Met. 3,811-814 (1969)

11 R. D. Schmidt-Whitley On the role of grain size in superplastic defor-

formation.
ibid. 8,49-54 (1974)

12 R. D. Schmidt-Whitley Reply to comments on the role of grain size in

superplastic deformation.
ibid. 8,927-930 (1974)

13 H. W. Hayden Superplasticity in the Ni-Cr-Fe system.

R. C. Gibson ASM Trans. Quarterly 60,3-14 (1967)
H. F. Merrick
J. H. Brophy
14 H. W. Hayden The inter-relation of grain size and superplas-
J. H. Brophy tic deformation in Ni-Cr-Fe alloys.
ibid. 61,542-549 (1968)

15 P. Chaudhari Deformation behaviour of a superplastic Zn-Al

Alloy.
Acta Met. 15,1777-1786 (1967)

16 P. Chaudhari A dislocation cascade mechanism in super-

plasticiýty.
Met. Trans. 5,1692-1693 (1974)

17 P. Shariat An evaluation of the roles of intercrystalline

R. B. Vastava and interphase boundary sliding in two-phase
T. G. Langdon superplastic alloys.
Acta Met. 30,285-296 (1982)

70
18 D. L. Holt The relation between superplasticity and grain
boundary shear in the Aluminium-Zinc eutectoid
alloy.
Trans. AIME, 242,25-31 (1968)

19 A. Ball Superplasticity in the aluminium-zinc eutectoid

M. M. Hutchi. son Met. Sci. J. 3,1-7 (1969)
20 R. B. Vastava An investigation intercrystalline
of and inter-
T. G. Langdon phase boundary layer sliding in the superplastic
Pb-62% Sn eutectic.
Acta Met. 27,251-257 (1979)
21 D. Lee The nature deformatio
of superplastic n in the
Mg-Al eutectic.
ibid. 17,1057-1069 (1969)
22 R. C. Gifkins Grain rearrangements during superplastic
deformation.
J. Mat. Sci. 13,1926-1936 (1978)
23 A. E. Geckinli Superplastic deforation of the Pb-Sn eutectic.
C. R. Barrett ibid. 11,510-521 (1976)
24 W. Hotz Observation of processes of superplasticity with
E. Ruedl the scanning electron microscope.
P. Schiller ibid. 1051 2003-2006 (1975)
25 D. M. R. Taplin Flow and failure of superplastic materials.
G. L. Dunlop Ann. Rev. Mat. Sci. 9,151-189 (1979)
T. G. Langdon
26 D. McLean Grain-boundary sliding: A finite but unbounded
limit.
Met. Sci. J. 4,144-145 (1970)
27 R. N. Stevens Grain boundary sliding and diffusion creep in
polycrystalline solids.
Phil. Mag. 23,265-283 (1971)
28 M. F. Ashby Diffusion accommodated flow and superplasticity.
R. A. Verrall Acta Met. 21,149-163 (1973)

29 R. C. Cook Superplasticity in a dilute zinc-aluminium. alloy.

N. R. Riseborough Scripta Met. 2,487-489 (1968)

30 C. H. Hamilton Microstructure and phase ratio effects on the

A. K. Ghosh superplasticity of Ti-6Al-4V.
M. W. Mahoney In Advanced processing methods for titanium,
129-144, AIME, qdited. by D. F. Hasson and
C. H. Hamilton (1982)

31 A. M. Garde Micrograin superplasticity in zircaloy at 850'C

H. M. Chung Adta Met. 26,153-166 (1978)
T. F. Kassner
32 B. P. Kashyap Review. Microstructural aspects of superplas-
A. Arieli ticity.
A. K. Mukherjee J. Mat. Sci. 20,2661-2686 (1985)

33 F. A. Mohamed Creep at low stress levels in the superplastic

T. G. Langdon Zn-22% Al eutectoid.
Acta Met. 23,117-124 (1975)

71
34 F. A. Mohamed Creep behaviour in the superplastic Pb-62% Sn
T. G. Langdon eutectic.
Phil. Mag. 32,697-709 (1975)
35 R. C. Gifkins Comments on theories of structural super-
T. G. Langdon plasticity.
Mat. Sci. Eng. 36,27-33 (1978)
36 A. E. Geckinli Grain boundary sliding model for superplastic
deformatiou
Metal Sci. 17,12-18 (1983)

37 T. H. Alden The origin of superplasticity in the Sn-5'/O Bi

Alloy.
Acta Met. 15,469-480 (1967)

38 R. C. Gifkins Grain boundary sliding and its accommodation

during creep and superplasticity.
Met. Trans. 7A, 1225-1232 (1976)

39 W. Beere Grain-boundary sliding controlled creep: its

relevance to grain rolling and superplasticity.
j. Mat. Sci. 12,2093-2098 (1977)

40 A. K. Mukherjee The rate controlling mechanism in superplasticity.

Mat. Sci. Eng. 8,83-89 (1971)
41 J. H. Gittus Theory of superplastic flow in two-phase mater-
ials: Roles of interphase-boundary dislocations,
ledges and diffusion.
J. Eng. Mat. Tech. 99,244-251 (1977)

42 F. Dyment Self and solute diffusion in titanium and titan-

ium alloys.
In: Titanium '80,1,519-528, AIME, edited by
H. Kimura and 0. Izumi (1980)

43 S. M. L. Sastry High temperature deformation of Ti-6Al-4V.

P. S. Pao ibid. 21,873-886
K. K. Sankaran

44 J. Ma Superplastic deformation mechanisms and alloying

C. Hammond element segregation in Ti-6%AI-4%V-1%Co and
Ti-6%Al-4%V-1.8%Ni alloys.
In: Titanium science and technology 2,703-709
Deutsche Gesellschaft fur Metallkunde, edited by
G. Lutjering, U. Zwicker and W. Bunk (1985)

45 J. R. Leader The effect of alloying additions on the super-

D. F. Neal plastic properties of Ti-6%Al-4%V.
C. Hammond Met. Trans. In the press.

46 B. Hildalgo-Prada Correlation between mechanical properties and

A. K. Mukherjee microstructure in a Ni-modified Ti-6Al-4V alloy.
ICSMA 7. Montreal 1985 In the press

47 M. Suery Deformation mechanism of two-phase superplastic

B. Baudelet alloys.
Res. Mech. 2,163-173 (1981)

48 M. Suery Hydrodynamical behaviour of a two phase super-

B. Baudelet plastic alloy: a/ý brass.
Phil. Mag. L1,41-64 (1980)

72
49 J. R. Springarn Diffusional creep and diffusionally accommo-
W. D. Nix dated grain rearrangement.
Acta Met. 26,1389-1398 (1978)

50 H. Naziri Microstructural-mechanism relationship in the

R. Pearce Zinc/Aluminium. eutectoid superplastic alloy.
M. Henderson Brown ibid. 23,489-496 (1975)
K. F. Kale
51 P. G. Partridge A deformation model for anisotropic superplas-
D. S. McDarmaid ticity in two phase alloys.
A. W. Bowen Ibid. 33,571-577 (1985)

52 G. L. Dunlop Caviation at grain and phase boundaries during

E. Shapiro superplastic flow of an aluminium bronze.
D. M. R. Taplin Met. Trans. 4,2039-2044 (1973)
J. Crane
53 T. Chandra Grain-boundary sliding and intergranular cavi-
J. J. Jonas tation during superplastic deformation of a/ý
D. M. R. Taplin brass
J. Mat. Sci. 13,2380-2384 (1978)

54 J. W. D. Patterson Effect of phase proportions on deformation and

N. Ridley cavitation of superplastic a/ý brass.
J. Mat. Sci. 16,457-464 (1981)

55 J. A. Wert Enhanced superplasticity and strength in modi-

N. E. Paton fied Ti-6Al-4V alloys.
Met. Trans. 14A, 2535-2544 (1983)

56 N. E. Paton Titanium base alloy for superplastic forming.

J. A. Hall US Patent 4,299,626 (1981)

57 1. Chen Superplastic flow of two phase alloys.

In: Superplasticity, CNRS, edited by B. Baudelet
and M. Suery (1985)

58 J. Ma Superplastic deformation in Ti-4%Al-4%Mo-2%Sn-

R. Kent 0.5%Si (IMI 550)
C. Hammond J. Mat. Sci. In the press

59 D. S. McDarmaid Superplastic deformation of strongly textured

A. W. Bowen Ti-6Al-4V. Part 2 Changes in texture and
P. G. Partridge microstructure.
ibid. 20,1976-1984 (1985)

60 D. S. McDarmaid Anisotropic superplastic deformation of strongly

A. W. Bowen textured Ti-6Al-4V alloy.
P. G. Partridge RAE Technical Report 83006 (1983)

61 P. G. Partridge The effect of texture and microstructure on the

A. W. Bowen superplastic deformation of metals.
C. D. Ingelbrecht In: Superplasticity, CNRS, edited by B. Baudelet
D. S. McDarmaid and M Suery (1985)

62 D. S. McDarmaid Superplastic deformation of strongly textured

A. W. Bowen Ti-6Al-4V. Part 1 Stress and strain anisotropy.
P. G. Partridge J. Mat. Sci. 19,2378-2386 (1984)

63 O. A. Kaibyshev Influence of texture on superplasticity of the

IN. Kazachov titanium alloy VT6.
P. M. Galeev ibid. 16,2501-2506 (1981)

73
64 D. S. McDarmaid The effect of strain rate, temperature and
P. G. Partridge texture on anisotropic deformation in
Ti-6AI-4V.
RAE Technical Report In the press
65 D. S. McDarmaid Superplastic anisotropy in Ti-6AI-4V.
A. W. Bowen In: Titanium science and technology 2,
P. G. Partridge 689-694
Deutsche Gesellschaft fur Metallkunde,
edited by G. Lutjering, U. Zwicker and
W. Bunk (1985)
66 C. M. Packer Evidence for the importance of crystallo-
R. H. Johnson graphic slip during superplastic deformation
O. D. Sherby of eutectic Zinc-Aluminium.
Trans. TMS-AIME 242,2485-2494 (1968)
67 R. H. Johnson Microstructure of superplastic alloys.
C. M. Packer Phil. Mag. 18,1309-1314 (1968)
L. Anderson
O. D. Sherby
68 H. Naziri Anisotropic effects in superplastic Zn-0.4%Al
R. Pearce sheet.
J. Inst. Metals 98,71-77 (1970)
69 H. Naziri Anisotropic superplasticity.
R. Pearce Scripta Met. 3,807-810 (1969)

70 U. Heubner Anisotropie bei superplastischer Umformung von

K. -H. Matucha Zink-Aluminium-Blechen.
H. Sandig Z. Metallkunde 63,607-614 (1972)
.
71 H. Naziri Superplasticity in a Zn-0.4%Al alloy.
R. Pearce Acta Met. 22,1321-1330 (1974)

72 K. Nuttall The room-temperature deformation characteristics

of the superplastic Zn-Al eutectoid alloy.
J. Inst. Metals 100,114-124 (1972)

73 K. N. Melton Anisotropy during superplastic deformation of the

C. P. Cutler Sn-Pb eutectic alloy.
J. W. Edington Scripta Met. 9,515-520 (1975)

74 B. P. Kashyap Influence of non equiaxed microstructure on the

G. S. Murty superplastic behaviour of the Sn-Pb eutectic.
Trans. Jap. Inst. Metals 22,258-266 (1981)

75 C. P. Cutler Textures in the superplastically deformed tin-

J. W. Edington lead eutectic alloy.
Met. Sci. J. 5,201-205 (1971)

76 R. H. Bricknell Mechanical anisotropy and deformation mechanisms

J. W. Edington in an Al-Cu-Zr superplastic alloy.
Acta Met. 27,1313-1318 (1979)

77 R. H. Bricknell Textures in a superplastic Al-6Cu-0.3Zr alloy.

J. W. Edington ibid. 27,1303-1311 (1979)

78 G. L. Dunlop Anisotropic ductility of a superplastic alumin-

J. D. Reid ium bronze.
D. M. R. Taplin Met. Trans. 2,2308-2310 (1971)

79 A. J. Shakesheff Superplastic deformation of Al-Li alloys

P. G. Partridge RAE Technical Report 84020 (1984)

74
80 C. P. Cutler Quantitative texture studies of the
J. W. Edington superplastically deformed Al-Cu eutectic alloy.
J. S. Kallend Acta Met. 22,665-671 (1974)
K. N. Melton
81 K. N. Melton Textures in superplastic Zn-40wt%Al.
J. W. Edington ibid. 22,165-170 (1974)
J. S. Kallend
C. P. Cutler
82 K. N. Melton Crystallographic slip during superplastic
J. W. Edington deformation of the Zn-Al eutectoid alloy.
Scripta Met. 8,1141-1144 (1974)

83 O. A. Kaibyshev The influence of texture on superplasticity of

I. V. Kazachkov the Zn-22%Al alloy.
S. Ya. Salikhov Acta Met. 26,1887-1894 (1978)

84 K. Matsuki Superplasticity in an Al-6wt%Mg alloy.

Y. Uetani Met. Sci. 10.235-242 (1976)
M. Yamada
Y. Murakami
85 K. Matsuki Superplastic behaviour in nominally single-phase
K. Minami and two-phase Al-Cu alloys.
M. Tokizawa ibid. 13,619-626 (1979)
Y. Murakami

-86 D. Lee The role of slip deformation in the superplastic

Zn-Al eutectoid.
J. Inst. Metals 99,66-68 (1971)

87 O. A. Kaibyshev Peculiarities of dislocation slip during

B. V. Rodionov superplastic deformation of Zn-Al alloys.
R. Z. Valiev Acta Met. 26,1877-1886 (1978)

88 N. E. Paton Microstructural influences on superplasticity in

C. H. Hamilton Ti-6Al-4V.
Met. Trans 10A, 241-250 (1979)

89 R. Pearce Superplasticity.
Metal Construction 11,506-509 (1979)

90 R. Sawle Commercial applications of superplastic sheet

forming.
In: Superplastic forming of structural alloys,
307-317, AIME, edited by N. E. Paton and
C. H. Hamilton. (1982)

91 D. B. Laycock Superplastic forming of sheet metal.

ibid. 257-271

92 E. D. Weisert Concurrent superplastic forming/diffusion

G. W. Stacher bonding of titanium
ibid. 273-289

93 New forming technique adopted by British

Aerospace. Metallurgist and Materials Technologist
297-299, June (1981)

94 E. D. Weisert Forming SPF/DB structure.

J. R. Fisher In: Advanced processing methods for titanium,
101-113, AINE, edited by D. FHasson and
C. H. Hamilton (1982)

75
95 J. R. Williamson Aerospace applications of SPF and SPF/DB.
In: Superplastic forming of structural alloys,
291-306, AIME, edited by N. E. Paton and
C. H. Hamilton (1982)

96 J. Peden New technique for superplastic forming of

stainless steel sheet.
Sheet Metal Industries 59,49 (1982)

97 G. B. Brook Superplastic forming of metallic materials

Part 2.
ibid. 581,801-809 (1981)

98 C. H. Hamilton Superplasticity in high strength aluminium alloys.

C. C. Bampton In: Superplastic forming of structural alloys,
N. E. Paton 173-189, AIME, edited by N. E. Paton and
C. H. Hamilton (1982).

99 N. E. Paton Method of imparting a fine grain structure to

C. H. Hamilton aluminium. alloys having precipitating constituents.
U. S. patent 4,092,181 (1978)

100 J. A. Wert Grain refinement in 7075 aluminium. by

N. E. Paton thermomechanical processing.
C. H. Hamilton Met. Trans. 12A, 1267-1276 (1981)
M. W. Mahoney

101 J. A. Wert Grain refinement and grain size control in

superplastic forming.
Journal of Metals 34,35-40 (1982)

102 C. C. Bampton The effect of superplastic deformation on

J. W. Edington subsequent service properties of fine grained
7475 Al.
J. Eng. Mat. and Tech. 105,55-60 (1983)

103 P. G. Partridge Superplastic deformation of Al-6.2%Zn-2.5%Mg-1.7%Cu

A. J. Shakesheff (7010) alloy sheet.
RAE Technical Report 82117 (1982)
104 C. C. Bampton Control of superplastic cavitation by hydrostatic
M. W. Mahoney pressure.
C. H. Hamilton Met. Trans. 14A, 1583-1591 (1983)
A. K. Ghosh
R. Raj
105 K. S. Chan Deformation and fracture of strongly textured Ti
D. A. Koss alloy sheets in uniaxial tension.
ibid. 14A, 1333-1342 (1983)

106 W. A. Backofen Deformation processing.

Addison-Wesley Publishing Company (1972)

107 K. S. Chan Stretch forming and fracture of strongly textured

D. A. Koss Ti alloy sheets.
Met. Trans. 14A, 1343-1348 (1983)

108 K. S. Chan Localized necking of sheet at negative minor

D. A. Koss strains.
A. K. Ghosh ibid. 15A, 323-329 (1984)

76
109 K. S. Chan Effects of plastic anisotropy and yield surface
shape on sheet metal stretchability.
ibid. 16A, 629-639 (1985)

110 C. D. Ingelbrecht The effect of superplastic strain on the tensile

D. S. McDarmaid properties of the titanium alloys Ti-6Al-4V and
P. G. Partridge IMI 550.
In: Titanium science and technology 2,761-767
Deutsche Gesellschaft fur Metallkund7e, edited by
G. Lutjering, U. Zwicker and W. Bunk (1985)
111 C. D. Ingelbrecht R values of Ti-6Al-4V sheet after superplastic
strain.
RAE Technical Report 85003 (1985)

112 F. Larson Properties of textured titanium alloys.

A. Zarkades Metals and Ceramics Information Center Report
MCIC-74-20 (1974)

113 I. L. Dillamore Preferred orientation in wrought and annealed

W. T. Roberts materials.
Met. Reviews 10,271-380 (1965)

114 C. S. Barrett Structure of metals.

T. B. Massalski McGraw-Hill (1966).

115 M. Peters Control of microstructure and texture in

G. Lutjering Ti-6Al-4V.
In: Titanium '80 2,925-935, AIME, edited by
H. Kimura and 0. Izumi (1980)

116 C. D. Ingelbrecht Superplasticity of sheet test pieces machined

from edge textured Ti-6Al-4V bar.
RAE Technical report 85053 (1985)

117 C. H. Hamilton Characterization of superplastic deformation

A. K. Ghosh properties of Ti-6Al-4V.
In: Titanium '80 2,1001-1014, AIME, edited by
H. Kimura and 0. Izumi (1980)

118 M. E. Rosenblum Microstructural aspects of superplastic forming

P. R. Smith of titanium alloys.
F. H. Froes ibid. 2,1015-1024

119 S. P. Agrawal Effect of small amounts of yttria on the

R. R. Boyer superplastic behaviour of Ti-6Al-4V.
E. D. Weisert ibid. 2,1057-1066

120 A. K. Ghosh Grain size distribution effects in superplasticity.

R. Raj Acta Met. 29,607-616 (1981)

121 R. Raj Micromechanical modelling of creep using

A. K. Ghosh distributed parameters.
ibid. 291,283-, 292 (1981)

122 H. W. Hayden The deformation mechanisms of superplasticity.

S. F loreen Met. Trans. 3,833-842 (1972)
P. D. Goodell

77
123 A. K. Mukherjee Experimental correlation for high temperature
J. E. Bird creep.
J. E. Dorn Trans. ASM 62,155-179 (1969)

124 A. Arieli The effect of strain and concurrent grain growth

B. J. Maclean on the superplastic behaviour of Ti-6Al-4V alloy.
A. K. Mukherjee Res Mech. 6,131-159 (1983)

125 J. J. Jonas Implications of flow hardening and flow softening

during superplastic forming.
In: Superplastic forming of structural alloys,
56-68, AIME, edited by N. E. Paton and
C. H. Hamilton (1982)

126 M. Suery Flow stress and microstructure in superplastic

B. Baudelet 60/40 brass.
J. Mat. Sci. 8,363-369 (1973)

127 G. Herriot Superplastic behaviour of two-phase Cu-P alloys.

B. Baudelet Acta Met. 24,687-694 (1976)
J. J. Jonas
128 L. B. Duffy Superplastic deformation of high strength titanium
alloy IMI 550.
PhD thesis, Manchester University (1985)

129 R. H. Bricknell Deformation characteristics of an Al-6Cu-0.4Zr

J. W. Edington superplastic alloy.
Met. Trans. 10A, 1257-1263 (1979)

130 F. Weinberg Grain boundary shear in aluminium.

Trans. TMS-AIME 212,808-817 (1958)

131 D. A. Porter Phase transformations in metals and alloys.

K. E. Easterling Van Nostrand Rheinholt (1981)

132 D. S. McDarmaid Superplastic forming and post forming tensile

properties of high strength titanium alloy
Ti-4At-4Mo-2Sn-0.5Si.
Mat. Sci. and Eng. 70,123-129 (1985)

133 IMI Titanium Private communications (1984)

Research Dept.
134 N. E. Paton Critical review. Superplasticity in titanium
C. H. Hamilton alloys.
In: Titanium science and technology 2,649-672
Deutsche Gesellschaft fur Metallkunde, edited by
G. Lutjering, U. Zwicker and W. Bunk (1985)

135 M. Peters Control of microstructures of (a+a)-titanium

G. Lutjering alloys.
G. Ziegler Z. Metallkunde 74,274-282 (1983)

136 D. S. McDarmaid Tensile properties of highly textured Ti-6Al-4V

A. W. Bowen alloy after superplastic deformation.
P. G. Partridge RAE Technical Report 82108 (1982)

137 W. -B. Busche Fatigue life of superplastically formed

H. -D. Kunze Ti-6Al-4V.
In: Titanium science and technology 2,725-732
Deutsche Gesellschaft fur Metallkunde, edited by
G. Lutjering, U. Zwicker and W. Bunk (1985)

78
138 D. S. McDarmaid Tensile properties of strongly textured Ti-6Al-4V
A. W. Bowen after superplastic deformation.
P. G. Partridge Mat. Sci. and Eng. 64,105-111 (1984)
139 J. Freed Superplastic forming of titanium 6Al-4V. Basic
I. L. G. Baillie mechanical properties.
BAe (Filton) Report R&D/B44/6243 (1982)
140 1. Bottomley Basic mechanical properties evaluation of
D. Mayo superplastic forming and diffusion bonding of
B. Millington titanium/6Al/4V material.
BAe (Warton) Report MDR 0792 (1985)
141 T. L. Mackay Metallurgical characterization of superplastic
S. M. L. Sastry forming.
C. F. Yolton McDonnell Douglas Technical Report
AFWAL-TR-80-4038 (1980)

142 British Sheet of titanium-aluminium-vanadium alloy.

Standards Aerospace specification 2TA 10
Institution

143 H. W. Rosenberg Ti-15-3: A new cold-formable sheet titanium alloy.

Journal of Metals 35,30-34 (1983)

144 G. Simmons Single crystal elastic constants and calculated

H. Wang aggregate properties: A handbook.
MIT press (1971)

145 A. Arieli Measurements of the strain rate sensitivity

A. Rosen coefficient in superplastic Ti-6Al-4V alloy.
Scripta Met. 10,471-475 (1976)

146 D. W. Livesey Activation energies for superplastic tensile and

N. Ridley compressive flow in microduplex a/ý copper alloys.
A. K. Mukherjee J. Mat. Sci. 19,3602-3611 (1984)

147 B. P. Kashyap On the models for superplastic deformation.

A. K. Mukherjee In: Superplasticity, CNRS, edited by B. Baudelet
and M. Suery (1985)

148 A. Arieli Superplastic deformation of Ti-6Al-4V alloy.

A. Rosen Met. Trans. 8A, 1591-1596 (1977)

149 D. Lee Superplasticity in some zirconium and titanium

W. A. Backofen alloys.
Trans. AIME 239,1034-1040 (1967)

150 N. E. W. De Reca Self-diffusion in ý-titanium and ý-hafnium.

C. M. Libanati Acta Met. 16,1297-1305 (1968)
1
151 R. W. Gardiner Effects of hydrogen content, test temperature and
environment on the fracture of highly tectured
Ti-6Al-4V.
RAE Technical Report 80097 (1980)

152 F. A. McClintock Mechanical behaviour of materials.

A. S. Ar. gon Addison-Wesley publishing company (1966)

153 M. Dripke Procedures formeasuring normal anisotropy (r) and

H. P. Worner plastic stress-strain exponent (n)-Part 1.
Sheet Metal Industries. 131-137 (1980)

154 ASTM Standard E517. Standard test method for plastic

strain ratio r for sheet metal.

79
155 Y. C. Liu on the R-value measurements.
Met. Trans. 14A, 1199-1205 (1983)

156 N. Furushiro Factors affecting the ductility of superplastic

H. Ishibashi Ti-6Al-4V alloy.
S. Shimoyama In: Titanium '80 2,993-1000, AIME, edited by
S. Hori H. Kimura and 0. Izumi (1980)

157 W. T. Lankford New criteria for predicting the press performance

S. C. Snyder of deep drawing sheets.
J. A. Bauscher Trans. ASM 42,1197-1232 (1950)

158 R. L. Whiteley The importance of directionality in deep drawing

quality sheet steel.
ibid. 52,254-169 (1960)

159 W. A. Backofen Texture hardening.

W. F. Hosford Jr Trans. ASM 551,264-267 (1967)
J. J. Burke

160 A. K. Ghosh Influences of material parameters and

C. H. Hamilton microstructure on superplastic forming.
Met. Trans. 13A, 733-743 (1982)

161 F. Jovanne An approximate analysis of the superplastic

forming of a thin circular diaphragm: Theory and
experiments.
Int. J. Mech. Sci. 10,403-427 (1968)

162 G. C. Cornfield The forming of superplastic sheet metal.

R. H. Johnson ibid. 12,479-490 (1970)

163 D. L. Holt An analysis of the bulging of a superplastic sheet

by lateral pressure.
ibid. 129 491-497 (1970)

164 D. M. Woo The analysis of axisymmetric forming of sheet

metal and hydrostatic bulging process.
ibid. 6,303-317 (1964)

165 A. R. Agab Determination of the equivalent stress versus

O. E. Habib equivalent strain rate behaviour of superplastic
alloys in biaxial stress systems.
Mat. Sci. and Eng. 64,5-14 (1984)

166 R. Hill A theory of the yielding and plastic flow of

anisotropic metals.
Proc. Roy. Soc. A193,281-297 (1948)

167 E. W. Hart Theory of the tensile test.

Acta met. 15,351-355 (1967)

168 E. Duncombe Plastic instability and growth of grooves and

patches in plates or tubes.
Int. J. Mech. Sci. 14,325-337 (1972)

169 E. Duncombe Analysis of diffuse plastic stability in tubes and

sheets.
Int. J. Solids Structures 10,1445 -1458 (1974)

170 F. A. Nichols Plastic instabilities and uniaxial tensile

ductilities.
Acta Met. 28,663-673 (1980)

80
171 A. K. Ghosh A numerical analysis of the tensile test for sheet
metals.
Met. Trans. 8A, 1221-1232 (1977)

172 A. K. Ghosh Tensile instability and necking in materials with

strain hardening and strain rate hardening.
Acta Met. 25,1413-1424 (1977)

173 A. Arieli Factors affecting the maximum attainable ductility

A. K. Mukherjee in a superplastic titanium alloy.
Met. Sci. Eng. 43,47-54 (1980)

174 J. W. Hutchinson Tensile instabilities in strain-rate dependant

H. Obrecht materials.
In: Proceedings of the fourth international
conference on fracture, 101-116, University of
Waterloo press, edited by D. M. R. Taplin (1977)

175 S. Sagat The stability of plastic flow in strain-rate

D. M. R. Taplin sensitive materials.
Metal Sci. 10,94-100 (1976)

176 M. A. Burke Plastic instabilities in tension creep.

W. D. Nix Acta Met. 23,793-798 (1975)

81
 Dimensions in mm

47

to

0
(D )0 0 M

0
rn

(D

_
!o 0 (D

Spot welding Reinforcing 12.5

plate

, rn
251 10 25__

wo -, -8-

rnrn 16 16 16

Fig 1 Test piece dimensions

0,

*-

Fig 2 Cutting diagram for the edge textured Ti-6Al-4V bar

 a) Quartz tube
removed
showing:
A test piece
B heat shields
C water and Ar
connections

b) Cooling pipes
connected and
furnace closed

Fig 3 Apparatus for suPerplastic tensile testing

 5.0
True
strain
C) Experimental data 18
4.5 :0
, 04,0.8
- 0.6
4.0 0.45
dl nor -- --
Known --- --*ýý4), "'\ 0.25
0.1
3.5 0
-di
CL
2ý
3.0
Ln
2.5

3: 2.0
o

1.5

1.0
- '-

0.5

0
-10 -7 -6 -5
In strain rate S-1)

Fig 4 Flow stress data for 3.3mm Ti-6Al-4V sheet (pulled at

4 1
9250C and 3x 10- s- in the L direction)

used for the computer modelling of neck development

 L Orientation Ti - 6AL-4V
100
92 5"C
0 3.3 mm
x 2.0mm Batch A ý_---x
,&1.6 mm
El 0.9 mm

10

IL ox,
U)
U,

U,

10 0 IT Orientationj
LL

10

11-11"
1_
4
1 10- 10 -3 10-2
Strain rate (5-1)

Fig 5 Effect of strain rate on flow stress of Ti-6Al-4V

sheet ot initial thickness 0.9-3.3mm pulled at 925"C

 100 L orientation Ti- 6AI -4V
925111C,2rnm
x Batch A x-x
* Batch B
* Batch C
10

a-

loo
LL

00.,

0*40
'OýOK
10
0

.00,

loe,'J.,
1
L-
10-5 10-4 10-3 10-2
Strain rate (s-1)

Fig 6 Effect of strain rate on flow stress of Ti-6Al-4V sheet

of initial thickness 2. Omm pulled at 9250C

 L orientation Ti- 6AI -4V
925*C
1.0
0 3.3 mm
x 2.0 Batch A
1.8
0-8 0.9
-x
x.., %%
4000

0.6 WOO,
do Boo,
.%

x-. oý - **%

0.4

>

:t0
Ln .2
in

0.'

IT I
L
orientation
c
Co
I-

X'-
1101
wol e-«

0.6
ýýe

0.4

0.21
10-5 10-4 10-1 10-1
Strafn rate (s-1)

Fig 7 Effect of strain rate on m value of Ti-6Al-4V sheet

of initial thickness 0.9-3.3mm pulled at 9250C

 L Orientation Ti-
1.0 6AL -4V
925 0C, 2mm
X Batch A
0 Batch B
0.8
A Batch C

0.6
x
E
0-4

in 0,2
c
IV
tn

1.0
IT Orientation I

0.8

0.6

0.4

0.2 1
10-1 10-1 10-3 10-2
Strain rate (s-1 )

Fig 8 Effect of strain rate on m value of Ti-6Al-4V sheet

of initial thickness 2. Omm pulled at 925 0C

 20
Ti- 6Al -4V
5

15 0
60

0 A
L orientation
01
013 9251C9 3x 10-4 S-1
10 - 0
0 3.3 mm
x 2.0 mm Batch A
e,
,
* 2.0 mm Batch 8
X 0 2.0 mm Batch C
5-- b 1.8 mm
x/ 0 0.9 mm
CL

U) 20
3:
0
LL

15

10

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Superplastic strain

4 1
Fig 9 Effect ot strain at 9250C and 3x 10- s-

on tlow stress ot Ti-6Al-4V sheet

 0.7
L orientation, Eý 3x 10-4s-I
0
0 Ti-6AL-4V 925*C 3.3mm
0 or X IMI 550 900 2mm
yo -N Ti-8AI-IMo-lV 1010
0.6 Ti-15V-3Cr-3At-3Sn 910
ooý

E )C-

0.5 - 0

in

4A
0\00
0.4 0

V) 0.3

1000/. 200 */o 300*/. LOO*/.500*/*

0.2 1. I. IhI
0 0.2 0.4 0.6 0-8 1.0 1.2 1.4 1.6 1.8 2.0

Superplastic strain

4 1
Fig 10 Effect of superplastic strain at 3x 10- s-

on m value of titanium sheet alloys

 1100

mpa Tensile strength

>Z
1005ýýx--
4
2
3
950
0v

875
7TTTT--I
ITT
F
0.2 0/a Proof stress
L Orientation

103.3 mm
2X2. Omm Batch A
1025

MPa

950

875

800 LL
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

t0
Mill annealed
Superplastic strain

4 1
Fig 11 Effect of superplastic strain at 925*C and 3x 10- s-

on room temperature strength of Ti-6Al-4V sheet

(L orientation)
 1100 LTensite-
strength

1025

mpa

ý3
950 -ýý13

x 2
IB75

Boo

0.2 Ve Proof stress

T orientation
1025
103.3 mm
2x2.0 Batch A
mpa
3&2.0 Batch 8
401.8
95 0 5v0.9

875
3
"I,
xv x25vx
Boo[ IX IIIIII -A
-II
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
t00.1

Mill annealed Superptastic strain

Fig 12 Effect of superplastic strain at 925 0C and 3x 10 s

on room temperature strength of Ti-6AI-4V sheet

(T orientation)
 Sheet surface Sheet centre
Contour intervals Ix random Contour intervals 1x random

0002)0(

01O)a

(110

Fig 13 Alpha and beta phase textures of 3.3mm Ti-6Al-4V sheet

in the mill annealed condition (as-received)

 Sheet surface Sheet centre
Contour intervals 1x random Contour intervals 3x random

0002

010)0(

(110)p

14 Alpha and beta phase textures of 3.3mm Ti-6Al-4V sheet

Fig

annealing at 9250C for lhr

after
 Sheet surface Sheet centre
Contour intervals 1xrandom Contour intervals 2x random

002)(

pd

tç

5.1

0, O)d

110)p

15 Alpha and beta phase textures of 3.3mm Ti-6Al-4V, Sheet

Fig
4s-1
200% strain at 9250C and 3x 10-
after superplastic
 tzl'
e,= -Q '-r, -
- JL,::

AW.

w %;ý, P ill
.-C,; 'x . 1% ...
-. 7- -%". " - ý$ -
- ".; _

_3L,

71
Or-

As-received

f-s--*,
20 pm
I

b) Annealed 925'C lhr

>
-m-, -,>-,
After 150% superplastic

strain at 9250C and

4 1
C x 10- s-

16 Microstructure of 2. Omm Ti-6Al-4V sheet (batch A)

Fig
 I-s-, *,

a)

20 prn
I

b)

Fig 17 Microstructure of mill annealed 3.3mm T1-6Al-4V

sheet quenched from 925 *C

 a) 50% strain b) 140% strain

20 pm
FS II

c) 435% strain d) 735% strain

Fig 18 Microstructure of 3.3mm Ti-6Al-4V sheet after

4 1.
6 superplastic strain at 9250C and 3x 10- s-

Material reheated to 9250C and quenched

 I 'a rý rv-, 4ki, - 1, Ti-r, AI-/ k/ - L. - -. L
2.2

2.0

(0-
-01.

0- 1.6

Q)
vi

0-

1.2

1.0

0 0.5 1.0 1.5 2.0

Superptastic str2in F-I

Fig 19 Effect of strain at 925"C and 3x lo-4s-1 on alpha

phase aspect ratio of 3.3mm Ti-6Al-4V sheet. Material

reheated to 925"C and quenched.

 100 Ti-4At 2Sn
900*C; 2 mm -4Mo - -0-5Si
jimi 550)
OL or ientation
-
IXT

10

V)

wooý

--p
LL eooý

'00ý
L--Or IIIIIaI
--I --tII
5 3
10- 10-1 10-
10-1
Strain rate (S-1)

Fig 20 Effect of strain rate on flow stress of IMI 550 sheet

pulled at 9000C

1.0

E
0.8

f) 0.6
C)

c: 0- 4
(Z

0.2 1---
10-5 10-4 10-3 10-2
Strain rate ( S-1 )

Fig 21 Effect of strain rate on m value ot IMI 550 sheet

pulled at 900 0C
 20
Ti -4 At -4 Mo - 2Sn - 0.5 Si
0L orientation (IMI 550)
xT
15
OL

x
x

-ý-to
cu
X0
Lq
10 .CrQ O'ýý
*0

0
U-

0 0.2 0.4 0.6 0.8 1.0

Superptastic strain

4 1
Fig 22 Effect of strain at 9000C and 3x 10- s-

on flow stress ot IMI 550 sheet

 11D0i iTensite
imi SSO strengt

1025 0
0

mpa
x
950
0

875

800 IL

110 0F
0.2 */9 proof stress
0L orientat .ion
xT

102 5[

MPa
x
x
950 0

875

IIII1111
1-- 11
6 DO __j
0.2 03 0.4 0.5 0.6 0.7 0.8 0.9 1.0
t00.1
Supe(plastic strain

Annealed

4s -1
Fig 23 Effect of superplastic strain at 9000C and 3x 10-

on room temperature strength of IMI 550 sheet

 a) As-received

20 pm
I

b) Annealed 9000C lhr

c) After 150% superplastic

strain at 9000C and

4 1
3x 10- s-

Fig 24 Microstructure of IMI 550 sheet

 L orientation
2mm
13 10109C Ti-8AI -I Mo -1V
A 970
100
0 940 O-X
x 910 1000ox
-0
e-0, ,, ,- -Z5,
ýa-
1--
00000'X'10
Ilooll D-43
.10
z Z
10
Y, ,Uo-oý
Zýý ooocý
m

in
ul

4n
10 0T
orientation
0
Li-

Oeoe
ooý
"(),
*,
_,
"Cr_-AD

x 01*1

X
zx 0,
A".
x
/0",oll
1r
0/ 6z
46roýooooisý

10-5 10-1 10-3 10-2

Strain rate ( s-1)

Fig 25 Effect of strain rate on flow stress of Ti-8Al-lMo-lV

sheet pulled at 910,940,970 and 10100C

 L orienlation
2 mm
El 10106C
A 970 Ti-SAL-lMo-lV
1.0 0 9410
x 910

a8

0.6

0.4

0.2

1.0

orientation]

f.
0.8
ch
a\
ý\.rz/I/,
0.6

0.4

0.2
10-5 10-4 103 10-2

Strain rate (S -1 )

Fig 26 Effect of strain rate on m value of Ti-8Al-lMo-lV

sheet pulled at 910,940,970 and 1010*C

 40 [ Tj-SAI- IM0-1v

35

30

25 -n

20

L orlentation
10 10 10"c
970
0 940
910
0

40-
x --------
1---ýx---
35

30

25 - 2--0
)
-(;
20

lo -
ýn
T orienta tion)

0.2 0.4 0.6 0.8 1.0 1.2

Superplastic strain

Fig 27 Effect of strain at 910,940,970 and 1010 *C and

4 1
3x 10- s- on flow stress of Ti -q ýL - Imc, - JV sheet
 ... 2. ,-. -- mrý-
-, .: ý -- -- -. ý. -- -f -. %.. -W' -- ýý'-"41 .-,

4.
,

IV

As-received

?ý, k
.. >0
7k --ý

ýoA

Fv
"',

20 pm
I

b) Annealed 1010"C lhr

c) After 150% superplastic

strain at 10100C and

4s
3x 10-

Fig 28 Microstructure of Ti-8Al-lMo-lV sheet

 L orientation I Ti-15V-3Cr-3AL -3Sn
2mm
9100C
100 ý 860
0- A
sio
0-4-4 A: r
m 1- ww-

10

Od
IL

W
IT ----I
orientation
100
3:
0
LL

10 cr
fr

%-
1 5
, o-
10-4 lo-
10-1
Strain rate (S -1)

Fig 29 Effect of strain rate on flow stress of

Ti-15V-3Cr-3AI-3Sn sheet pulled at 810,

860 910 .C
and
 L o6entation Ti- 15V -3Cr - 3AL - 3Sn
0.8
2mm
9100C
8 60
0.6 0 610

0.4
rL-

0.2

' il

0.8

0
.6

0.4

0.2

0 L-

10-1 lo-4 lo-3 10"

Strain rate (S-1)

Fig 30 Effect of strain rate on m value of Ti-15V-3Cr-3Al-3Sn

sheet pulled at 810,860, and 910"C

 50
Ti-15V -3Cr-3At-3Sn
45

40

35

30
- ----------
25

20
L orientation
15 9100C
10

m5
CL

50

45

iLL 440

35

30

25

20

15

10

0
0.2 0.4 0.6 0.8 1-0 1.2 1.4
Superplastic stra'in

4s -1
Fig 31 Effect of strain at 810,860 and 910 0C and 3x 10-

on flow stress of Ti-15V-3Cr-3AI-3Sn sheet

 (110)f

a) As-received b) After 238% strain at 810"C

4 1
and 3x 10- s-

Fig 32 Texture of Ti-15V-3Cr-3Al-3Sn sheet.

Contour intervals 1x random

a)

LO

Fig 33 Ti-15V-3Cr-3Al-3Sn sheet test pieces

a) Before testing
4s -1
b) After 200% strain at 9100C and 3x 10-
 (110)f

a) As-received b) After 238% strain at 810'C

1
and 3x 10-4 S_

Fig 32 Texture of Ti-15V-3Cr-3Al-3Sn sheet.

Contour intervals 1x random

b)

I OR

Fig 33 Ti-15V-3Cr-3Al-3Sn sheet test pieces

a) Before testing

4s -1
b) After 200% strain at 9100C and 3x 10-
 Y4

a) Annealed at 810"C b) 15% strain

(unstrained)
40 pm
II
--0,
FS
.0
Vt.

ire

./----
I ýý- -- 6--- --..

c) 46% strain d) 197% strain

-i-
4 1
Fig 34 Effect of strain at 8100C and 3x 10- s-

on the microstructure of Ti-15V-3Cr-3Al-3Sn sheet

 W.
I-I. '0. - --
I.

J4.,

WC,

e) 373% strain f) 634% strain

40 prn

SSS
_______________
/\

\L

1652% h) 20 Pm
g) strain Precipitation on

(necked region) subgrain boundaries

8100C 1
Fig 34 Effect of strain at and 3x 10-4 s-

on the microstructure of Ti-15V-3Cr-3Al-3Sn sheet.

Magnification 160x except h): (630x)

 0.12
Inn -in --r-l

0.10
cu
N
(A

E 0.08

4ý
CL

IL, 0.06
(U
C:

0.04

cru

0.02

0 0.5 1.0 1.5 2.0 2.5 3,0

Superplastic sirain

4s
Fig 35 Effect of strain at 810 and 860oC and 3x 10-

on grain size (measured in the L direction)

of Ti-15V-3Cr-3Al-3Sn sheet
 LT

200%

ST

200%

TL

200%

4s -1
a) Strain rate 3x 10-

LT

187%

ST
-, t, I co
140%
,
TL

180%

1.5 x 10 -3 s
b) Strain rate

Fig 36 Ti-6Al-4V sheet test pieces machined from bar. Test

piece shape after strain at 800*C. Magnification 0.6x

 LT

300%

ST

300%
ca

TL

273%

4 1
c) Strain rate 3x 10- s-

LT

340%

ST

240%

--Th_ 1
TL

200%

d) Strain rate 1.5 x 10 s-

Fig 36 Ti-6Al-4V sheet test pieces machined from bar. Test

piece shape after strain at 8750C. Magnification 0.6x

 LT

187%

ST

293%

I-'

/
TL

200%

4 1
e) Strain rate 3x 10- s-

1>13L% LT

MA c4 325%
-

ST

247%

TL

200%

3s -1
f) Strain rate 1.5 x 10-

Fig 36 Ti-6Al-4V sheet test pieces machined from bar. Test

piece shape after strain at 9250C. Magnification 0.6x

 I

LO LT
u
180%

ST

90%

um

TL

165%

4 1
g) Strain rate 3x 10- s-

LT

173%

ST

106%

TL
to

93%

3 1
h) Strain rate 1.5 x 10- s-

Fig 36 Ti-6Al-4V sheet test pieces machined from bar. Test

1.

piece shape after strain at 975"C. Magnification 0.6x

-
 /"--s
T

a) ST orientation test piece B7

mm

TL

b) LT orientation test piece C6

4 1
Fig 37 Effect of superplastic strain at 8000C and 3x 10- s-

Ti-6Al-4V sheet test pieces machined from bar

on
 Strain rate
-3 1
1.5 x 10 s-

ýi -- I..
-- -
,;z-

Strain rate
C,3
4 I
3x 10- s-

10 mm
I

---J . ---
a) ST orientation test pieces B4 and B6

Necking

B4

0.82
1 0.80 R values (necked regions)
0.78

86

II 1
0.99 1.01
R values (uniform strain regions)
0.25 0.98

b) Position of necks and R value measurements

Fig 38 Necking of ST orientation test pieces pulled at 9750C

 Lx

-,,,, k,; A-

q Jar

Nk
IZ
71

ý00

At
4-4 J

4k

SI,

Fig 39 Mill annealed (as-received) Ti-6Al-4V bar showing

aligned (A) and non-aligned (B) microstructure
 jot,

a) Quenched SL

20 prn
Ia
-

7/-
5 .? ---
-. w- -6g.,
-e' -
-eq2

: -/---L j
--
--
-- o--
-. --- ----1 f'>-
-- _-- -
4Y - ---- ----- .-
-- --
--
_T- _Ti': _
__'_/_

N:. _
-'

---: --
I-
___; -

--
J
-

b) Quenched from 8750C

Fig 40 Microstructures of quenched Ti-6Al-4V bar

 ,,
Af"

0 -*-4- -
A, Klýrl
I ý!
)IV'

-W IIt.,
W.., -r -i
t4; " lo's
,
'
lob I

Quenched 6 tý
c)
W ý.,
',
ýi, ý4 T

20 pm
I

14 - ", ,I--*, ".. --, ". ý, t 'A

ý;. --

14

-ZA
"60

Quenched from 975"C

Fig 40 Microstructures of quenched Ti-6Al-4V bar

 nitaining non-aligned
distributed a and 0 phase grains

ontaining aligned
sa and 0 phase grains

Fig 41 Schematic distribution of aligned (band A) and

non-aligned (band B) microstructure (reproduced

from ret. 51)

 7 -\

i
'-
DJJ

! l\.

A f -s-"ý

I '.

* '- t. t_J, 1 -"

-'-
_____________
S
mm
a)

"qgzt
1- FS
-.

20 Pm

Fig 42 TL orientation test piece after 200% strain at

9250C and 3x 10 -41 s. Tensile axis T

FL --0,

20 pm
II
Fig 43 ST orientation test piece after 200% strain at

875'C and 3x 10 -4-1s. Tensile axis S

 a) Annealed at 875"C

(unstrained)

FL--""
50 pm
k- A

b) 12% strain

c) 24% strain

1
Fig 44 Effect of strain at 8750C and 3x 10-4 s- on

microstructure of Ti-6Al-4V (reheated and quenched).

ST orientation test piece B5. Tensile axis S

 d) 169% strain

FL,

50

e) 468% strain

f) 2000% strain
(necked region)

1
Fig 44 Effect of strain at 8750C and 3x 10-4 s- on

microstructure of Ti-6Al-4V (reheated and quenched).

ST orientation test piece B5. Tensile axis S

 :'
: if:

"1

-,0-
FS
100 pm
40-e II

-i-.
.
j-:
a)

/vz

vný

IL

vo

5C rT,

b)
IýF',,
neck I un 11 orat

region I strain

Fig 45 Microstructure inside neck and in adjacent region of

ST orientation test piece B4 pulled at 975'C and

1.5 x 10 -3-aS. Tensile axis, S
 ý..* ý.
-ý. " ý- ,I ýqpomqm
47
.,,,
,,

FL
--"

a) ST orientation test piece B4 pulled

3 1
at 9750C and 1.5 x 10- s-

prn

FS,

b) TL orientation test piece A3 pulled

4 1
at 975 C and 3x 10- s-

Fig 46 Necked regions of ST and TL test pieces

pulled at 9750C
 F7,

20

Fig 47 ST orientation test piece B2 pulled at 875'C and

1.5 x 10 -3-1
s. Cavitation inside necked

region. Tensile axis S

I *!ý-, lb ýmk %
ft-

FLO,

pm

Fig 48 ST orientation test piece B5 pulled at 875*C and

3x 1
10-4 s- showing test piece edge. Tensile

axis S. Scanninq electron microqraph

 100 -
Orientalion
Lo's
TL
5T
so - LT

4
CL
60 -t1.5
X 10-1 S-1

40 -

t3x 10-4s-1

20 -

01
875 925 975

Boo Temperature IC '1

Fig 49 Flow stress at maximum load for each orientation and

strain rate for Ti-6Al-4V sheet test pieces machined

from bar

Orientation
1.0 TL
SLT
T

0.8

0.6
1=3x 10-4 s7l \X

c 0.4

0.2

0
Boo 575 925 975
Tomperature OC

Fig 50 m values at maximum load for each orientation and

strain rate for Ti-6Al-4V sheet test pieces machined

from bar
 t -1 -1 1
=3x 10 5
1 500 875 925 975'C
Non-banded 0a0
Banded 0

1.4
R TL ST LT
1.3

1.2
0 AA
0

1.0 __j CL
0.5 1.0 1.5 a5 1.0 1.5 0.5 1.0 1.5
*0
0.9 -

0.8 -A
A0 0
OAA a
0.7 9
so
00
0.61 0
A
IPA A

0.5 a) Strain rate

0. 4 3x 10-4s-1

7---7
Cc 1.5-10 31
1100 1711 115
Non-banded 0AD
Banded 0Aa

1.44
-

R TL ST 0 LT
1.3 -

1.2 -
A
1.1 - -0A-
0
L
0.5 1.0 1.5 1.0 1.5 0.5 1.0 1.5
0-
0.9 -

0*
0.8 - 0

0.7 -
4

0.6 -
0
OLS -

0.4 - b) Strain rate

9
0.3 L 1.5 x 10-3s-l

Fig 51 Effect of strain on R value for Ti-6Al-4V

test pieces machined from bar

 -1.25 TL orientation, 875'C
t=3x 10-4s-1 E
x Banded (A)
B2
0 Non-banded (B)
-1.00

-0.75
A 3// ew

EW, ct A2 I

x B2
-0-50
Bl
xB4

-0-25 B3

0 0.5 1.0 1.5 2.0

'E

a)

Al A2 A3
I

Bl B2 B3 B4

b)

Fig 52 Strain distribution for TL test piece A4 pulled at

4 1.
8750C and 3x 10- s- The indices Al-A3 and BI-B4

in b) each refer to a set ot two points in a)

 L Orientation T Orientation

1.0 1.0 0 3.3 mm sheet

x 2. Ornm batch A
2.0 mrn batch B
0.9 - 0.9 1.11mrn
V 0.9mm

0.8
R R

0.7 0.7

0.6 F 0.6

10
0.25 0.5
ct
0.75 1.0
0.5
0
0.25 0.5
EL
0.25 1.0

Fig 53 Effect of strain at 925-C and 3x 10 -4 S-1

on R values of Ti-6Al-4V sheet

Sheet thickness (mrn)

0.80 Test piece orientation

0.75

0.70

R value
at E=0.4

0.65

0.60

0.55
1
3.3 (T )

0.50 111111
1.2 I. L 1.6 1.8 2.0 2.2

Initial a phase aspect ratio of specimen cross-seclion

Fig 54 Relationship between alpha phase aspect ratio and

R value for Ti-6Al-4V sheet after strain at 9250C

3x 10-4 s- I
and
 -1.2 -0 4mm gouge width 1

x8 ct

A 16
ID 19.5
-
Gauge length 10mm
, oq

cW

Cw, Ct
df,
-=0.48
dck
-0.6 -

4, Smrn gauge widli,

16
-0.4 r 19.5

ý- Zx
-0.2
a) Effect of
dc,
0.41
varying gauge
0 0.5 1.0 1.5 2.0
c width

0 25mm gauge length

x 10
Ct
L5
1.2 02

Gauge width l6mm 1ý16,11

-1.0
1
"I

cw

Ew Ct
,

-0.6

25mm gauge iength

-0.4
10
5
2

-0.2 b) Effect of

varying gauge
05101.5 20
ct length

Fig 55 Effect of test piece geometry on superplastic

strain anisotropy of Ti-6Al-4V sheet pulled at

4s -1
925"C and 3x 10-
 I

''

4
a) w 19.5mm b) w 16mm
00
tpl 88% tp3 252%
tp2 320% 20 mm tp4 320%
11

cog

c) w 8mm d) w 4mm
0 0
tp6 120% tp9 690%
tp7 400% tplO 460%
tp8 320% tpll 130%

Fig 56 Ti-6Al-4V sheet test pieces atter strain at 925*C and

3x 10 -4-1s. 1nitial gauge length 10mm, initial

gauge width w0
 -2 .L

-2.0

-1.6

Ew, Cl

-1.2

-0.6

-0.4

0 1.0 2. o 3.0 4.0 S.

Fig 57 Strain distribution along the gauge length of a

failed test piece (tp 9 in Fig 56d)

0.9

V
V

0.8

0.7
A
AA
A

LT
0.6
00 Ti -6AL - 4V 9251C

LAI mi 550 900

o2 Ti -BAI - Wo -1V 1010

0.6 vv Ti-15V -3Cr -3AL - 3Sn 910
a

0.4 11
0 0.5 1.0 1.5 2.0

Superplastic strain

4 1
strain at 3x 10- s-
Fig 58 Etfect oi superplastic

titanium alloy sheet test pieces

on P values ot
 10A

10

10
_x

lo,

100
0 1 10 2 10 3
10 10
Uniform strain (*/. )

Fig 59 Effect of uniform strain increment size on predicted

neck strain. Alternate points have been omitted for the

0.025 strain increment size for uniform strains < 200%.

lnitial normalised neck area 0.95

 10,

103

.ý;1ol
U
C)
z

10,

10,
10,10 1 10,10 3
Uniform strain COM

Fig 60 Effect of initial normalised neck area on predicted

neck strain. Uniform strain increment size 0.05

 1.8

1.6

1.4
'V
fe

1.2

1.0
C
E

in
0
0.8 z
Ile
u
(k)
z 0.6

0.4

0.2

100 10,10,10
Uniform strain ( */. )

Fig 61 Predicted load and neck strain rate for an initial

normalised neck area of 0.8

60 r

o Uniform f[ow stress

so I
01 Neck flow stress

40 0
EL
m

30
13
0
13
0
Ll- 20 13
13
13
13 000,10
13
0
13 C)ocooooo
0 000
10 [
00n00
0 0 0
0 0
0
ol
103

10, lo,
Uniform strain 10,
Fig 62 Predicted uniform and neck flow stresses for an

initial normalised neck area of 0.8

 0.7

0.6 o O'D
0, 00000
D0
OOODQD
* Uniform m & Ailip
0.5 * Neck M
4ý Vva
4b b

* Uniform n býý

0.4 * Neck n
v

0.3

E 0.2 V

17

0.1
V

loo 10, 10, 10

Uniform strain I'M

Fig 63 Predicted unitorm and neck m and n values for

an initial normalised neck area of 0.8

40

0 Uniform 1 8u dE)
mu
35 (au/a Eýi dE
. Uc
Neck
Ila
0 Mo. '
A KI., 1,
30 ý,-v I 4 WO
Neck da/dc
Ln
25
0
cl
E 3
C, 20 v
u
01 0 0 0
C 0
C: 15

m
00 80

10
0200
A 13 00
00 0 CbO

LO
Cý
0"
A 0
L 103
101 lo,
0
10 Uniform stfain (*/. )

Fig 64 Predicted strain and strain rate hardening

for an initial normalised neck area ot 0.8

 12
L ori entation 1
mm
Ti -6A1 -4V 9 25', C
100 IM 1 550 900
Ti-6A1-1Mo-1V 1010
Ti-15V-3Cr-3At-3Sn 910

ILiv
Y

10
4n

0
ir r

I L-
10-5 10-4 1()-3 1 0-ý2

Strain rate ( S4)

Fig 65 Effect of strain rate on flow stress for

each of the alloys

 Test ) mm
a) piece appearance
I

0.6ý

b) R value measurements

Fig 66 TL orientation test piece A9 atter strain

at 8750C and 1.5 x 10-

 1.0 0.9
EW
0R
ct
x

0.9 X R'
dEw
dEt "-, "X
0.8 -
/

0.8 0.7 -

R, R'

0.7 0.6 X

0.6 0.5 v2 R+R

1R

0.5L 0.4 'II

0 0.5 1.0 1.5 2.0 0 0.5 1.0
¬1 4E

at 925*C and 3x 10 -4 s
Fig 67 Effect of superplastic strain

R, RI RI for 3.3mm Ti-6Al-4V sheet

on strain ratios and
 0.5

d= 200)
34
to = 3.3mm

1% '%ýý
-'j 0.3 xc
a. x
14,
ýX

W
ift
go

It 0.2

1 00 UUniffOrmthinning, a, = 9MPa, R= I
2xm=0.67,
m a, = 9MPa, R=I
3am=0.67, a, =a (C). R
0"
4 13 m=0.67, cr, =a (c), RL, RT = RL, RT (4E)

0 10 20 30 40 so
Time (minutes)

Fig 68 Calculated optimum pressure forming cycles for a 200mm

diameter hemisphere from 3.3mm Ti-6Al-4V at 9250C

0.5 r-

0.4

¶
'.j

In
U)
a)
E 0.2
/0m=0.6 0m=0,67, at = Cr(C), Rý, Rý - WL, Rj (C)

xm=1.0 a, = or OE), R'L RT RL, Rý(C)

,
Am=0.5 af =a (C), Rý, RT RL, RT OE)
0.1 7, af = (7 W, P'= OC )

f Time for hemisphere

completion of

0 10 20 30 40 so so
Time (minutes)

Fig 69 Calculated optimum pressure cycles for various m values

 8

0
- 4.5

- 4.0

aA - 3.5
Increasing 0
X Max load 3.0
-
4 y
LO E
< 0 - 2.5 +

0 - 2.0
2
0
1.5
a (A/A 000 -y+rn
Inc rea sli ng a4laA=D 0*%, 4
1.0
0

0.5
00
00

10, 10, 102 10,

Uniform strain (*/. )

Fig 70 Predicted tensile instabilitY points for

an initial normalised neck area of 0.8