A Study of the Degradation of the Two-Way Effect in NiTi

Scott A. Skirlo, scott.skirlo@gmail.com

Abstract

The purpose of this study was to examine the degradation of the Two-way effect
in NiTi over several thousand thermal cycles with respect to the lowest and highest
temperatures reached during thermal cycling. The Two-way effect exists only in trained
shape memory alloy samples and consists effectively of the material remembering a hot
shape and a separate cold shape. The memory of each shape will degrade with repeated
thermal cycling. Knowing the impact of cycling temperatures is important for
applications and for understanding the mechanisms of degradation. This study utilized
stress-strain, resistance-temperature, strain-temperature, and strain-thermal cycle tests to
measure the degradation and understand the development of the microstructure. The
equipment for this study was constructed from scratch.
The results show that the degradation of the Two-way effect does not depend greatly
on the lowest temperature reached during thermal cycling. In contrast, changes in the
highest temperature reached significantly altered the degradation path. By increasing the
maximum cycling temperature, the rearrangement of the internal stress field causing the
Two-way effect increased, resulting in faster degradation. Lowering the maximum
cycling temperature subjected the trained dislocation array to less stress, better
maintaining the Two-way effect. The number of dislocations generated during cycling
did not appear to vary significantly throughout the tests, suggesting that internal stress
rearrangement is the most significant factor in the decrease of the Two-way strain.
Additional observations suggest that rearrangement involves softening.
 Introduction

Through a training procedure, shape memory alloys can exhibit a significant

dimensional change automatically with the temperature-induced Martensite
Transformation (MT) in the absence of an external stress. This is called the Two-way
shape memory effect, see Fig. 1. The training procedure introduces an anisotropic
internal stress field into the material. This stress field favors the growth of certain
properly oriented variants on cooling, resulting in the shape change [1]. For this study,
the test samples were Two-way trained by introducing dislocations from repeated cycles
of Martensite deformation followed by heating. This training method was selected
because it is easily implemented and because it is the most efficient [2].

Two-way Strain
2.5
As
2
Mf
1.5
Martensite
Strain(%)

Heating
1 Cooling

0.5
Austenite
Af
0
0 0.05 0.1 0.15 Ms 0.2 0.25 0.3
-0.5
Current (A)

Fig. 1 Two-Way Strain

The trained stress field will change with repeated thermal cycles, resulting in the
degradation of the Two-way effect. A few studies have been done, but the mechanisms
and relevant variables are still poorly understood [3-5]. This study will examine the effect
of the cycling temperatures, the maximum and minimum temperatures reached during
thermal cycling, on the degradation of the Two-way effect, since this is important for
applications and for understanding the degradation mechanisms. Interesting results from
other degradation tests will also be referenced in the discussion.

Experimental Design

All the wire used in this study was obtained free of charge from Dynalloy©.
Dynalloy© prepared their 90 °C AS, 0.1016 mm diameter wire, at request, by annealing it
for 2 hours at 500 °C. The purpose of the anneal was to remove some of the cold work
and make the wire more suitable for Two-way training.
This study was conducted on a custom built setup. The setup had to be able to do
several basic functions so that it could conduct resistance-temperature, strain-
temperature, strain-thermal cycles and stress-strain tests. The force sensor had a precision
of approx. 1 mN and was attached to one end of the wire. The moving end of the wire
was mounted to a slider on a linear potentiometer. The displacement resolution was about
10 um. A displacement controller was created to move the wire with a similar precision.
When measuring the length of the sample at a given temperature, a small load of
50 mN was applied to stretch the wire. Applying this load had the least possible impact
on the sample while maintaining accuracy in length measurement. This approx. 6 MPa
load is similar to the 5 MPa load employed for the same purpose in other studies [2].
Newtons are used throughout this paper instead of Megapascals because the small
diameter of the wire made it difficult to determine the cross sectional area precisely at
any given point in testing.
An electric current passing through the wire controlled its temperature. For test
samples with the same diameter and same history, surrounded by an atmosphere with a
stable, constant temperature, the amount of current passing through it, when at
equilibrium, correlates directly, and consistently to a temperature. The exact temperature
the wire was at was never known precisely since the test samples were very thin. When a
sample was undergoing thermal cycling 1 second was allowed to reach thermal
equilibrium when heating and 4 seconds for cooling. These times were considered
adequate since visibly the material had finished transforming and the current had
stabilized. When tests using 2 seconds for heating were conducted, no significant
variations in the results were observed. The precision of current measurement was about
0.1 mA, while the resolution of control was about 5 mA.
The electrical resistance of the material at a given temperature was determined by
passing a small, constant measurement current through the material briefly after allowing
it to reach equilibrium. The times the measurement current was passed through the
sample were small enough (approx. 1 mS) that there was negligible heat loss.

Procedure

Samples after mounting on the setup underwent two thermal cycles to remove any
deformation associated with handling. Following this all samples were Two-way trained.
This consisted of loading the sample with 3.5 N at room temperature in its Martensite
state, then unloading and then recovering to Austenite by heating with 0.3 A. This
procedure was repeated 10 times and created a Two-way effect of about 2% consistently.
After training, a strain-temperature test and a resistance-temperature test were completed.
All test samples then underwent three thousand thermal cycles between two cycling
temperatures TL and TH, the lowest and highest temperatures reached respectively. The
length of the material at each temperature was recorded periodically throughout cycling.
Following thermal cycling the test samples underwent additional strain-temperature,
resistance-temperature, and stress-strain tests to characterize the ending microstructure.

Results

Fig. 2 shows the stress-strain curves of the material during training. Every cycle
the internal stress distribution was altered because of dislocation introduction. The very
first cycle represents a typical stress-strain curve for an annealed sample [6]. First the
twinned Martensite elastically loads. This is followed by a detwinning plateau. After the
plateau, there is plastic deformation and some additional detwinning.
 The stress required to achieve a particular strain continuously decreased with
continued training, a change opposite of work hardening. In this instance dislocations
introduced from the detwinning operation serve to make detwinning easier, i.e.
thermodynamically stabilize variants “pointing” along the loading axis. The strain offset
each cycle represents plastic deformation and accumulated Two-way strain. At a certain
threshold the trained stress field effectively makes detwinning automatically happen on
cooling, which is the Two-way effect.
4

3.5

2.5
Stress (N)

1.5

0.5

0
0 2 4 6 8 10
Strain (%)

Fig. 2 Two-Way Training Cycles

Of the accumulated offset strain, only about 2% of it actually contributes to the
Two-way effect. The training process does not involve the material “memorizing” the
exact stress field applied during deformation. Different parts of the crystal will remember
different stresses depending on the local dislocation activity; hence the change from a
plateau, resulting from a nearly uniform internal stress field, to a sloped one, resulting
from varying local internal stresses.
The strain-thermal cycle tests recorded the degradation of the Two-way effect
with thermal cycling between TH and TL (these temperatures were always sufficient to
induce a MT). The Two-way effect is defined by certain parameters during thermal
cycling. The length of the material at TH is called the Austenite length (LA), while the
length at TL is called the Martensite length (LM). The Two-way strain (εTW) of the
material is defined with respect to the Austenite length for a given thermal cycle as:

εTW = (LM – LA)/ LA ×100

The changes in the Austenite and Martensite lengths for an Nth thermal cycle, called the
Austenite and Martensite strains (εA,N, εM,N), are additionally defined as:

εA,N = (LA,N – LA,1)/ LA,1 ×100 εM,N = (LM,N – LM,1)/ LM,1 ×100

Results early on indicated that varying TL did not have a noticeable impact on the
degradation of the Two-way effect. Multiple tests were done, but all possible
combinations of TL and TH were not investigated. The main focus of the work as a result
was on varying TH keeping TL at room temperature for simplicity. Varying TH
surprisingly had a very strong effect on the degradation. Plots of εTW, εA, and εM for
 samples are given in Fig. 3 and Fig. 4. TH was controlled by the steady current listed for
each line. Strain-temperature plots for samples before and after cycling, with THs that
caused a minimum and maximum degradation are given in Fig. 5 and 6.
2.5
1.9
0.298 A 0.267 A 0.298 A 0.267 A
2 0.247 A 0.235 A
0.247 A 0.235 A
1.7
0.218 A 0.20 A
0.218 A 0.20 A

1.5 1.5

Strain (%)
Strain (%)

1.3
1

1.1

0.5
0.9

0 0.7

0.5
-0.5
0 500 1000 1500 2000 2500 3000 3500
0 500 1000 1500 2000 2500 3000 3500
Number of Thermal Cycles Number of Thermal Cycles

Fig. 3 εA and εM Fig. 4 εTW

The ending Two-way strain was maximized with a TH that occurred at 0.218 A
and minimized with a TH that occurred at 0.3 A. Even though decreasing TH decreases the
rate of degradation, there needs to be a sufficient change in temperature to have the MT.
Cutting too far below AF causes any gains made by the cooler TH to be countered by less
strain change overall. Even though the 0.2 A sample had the slowest rate of degradation,
it started out with a low Two-way strain and consequently did not have the highest Two-
way strain after 3000 thermal cycles.
2.5 2.5

Before 3000 2 Before 3000

2
After 3000 After 3000
1.5 1.5
Strain(%)
Strain(%)

1 1

0.5 0.5

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 0.05 0.1 0.15 0.2 0.25 0.3
-0.5 -0.5
Current(A) Current(A)

Fig. 5 Thermal Cycling from 0.0 to 0.3 A Fig. 6 Thermal Cycling from 0.0 to 0.218 A
Discussion

From 0.25 A to 0.3 A, the amount of degradation continuously increased, rather

than leveling off, even though the measurable strain change was finished at 0.25 A. If the
test samples were annealing, these results would be expected, since a higher temperature
 would accelerate the rate of dislocation annihilation. If the sample was annealing then the
degradation would also depend on the heating time. This is not the case since ageing at a
temperature TH for 2 hours caused by a current of about 0.3 A showed the impact on the
Two-way effect that only a single thermal cycle was expected to show. The degradation
of the Two-way effect is caused by the MT the samples underwent during thermal
cycling. This means that the transformation is not finished at 0.25 A, even though the
measurable strain change was finished there.
1.2 2

1 Two-Way Strain Two-Way Strain

Austenite Strain Austenite Strain
0.8 1.5
Martensite Strain Martensite Strain
0.6
0.4
1

Strain(%)
Strain(%)

0.2
0
0.5
-0.2 0 500 1000 1500 2000 2500 3000 3500

-0.4
0
-0.6
0 1000 2000 3000 4000
-0.8
-1 -0.5
Number of Thermal Cycles Number of Thermal Cycles

Fig. 7 Two-Way Effect Degradation Fig. 8 Two-Way Effect Degradation

According to the work of Scherngell et al. the degradation of the Two-way effect
is by two main mechanisms [3]. The first is dislocation rearrangement and the second is
dislocation introduction. The existence of these two mechanisms was drawn mainly from
results like those in Fig. 7 and Fig. 8. Two behaviors can be observed according to trends
in εA, εM and εTW.
Dislocation rearrangement marks the first period of rapid strain loss. The trained
dislocation array is constantly exerting a tensile stress on the material. This deforms the
Austenite phase and detwins the Martensite on cooling. If the internal stresses causing the
Two-way effect were to reorient, then the net tensile stress would decrease. This would
cause a significant decrease in εA, εM and εTW as is observed. The material is assumed to
undergo rapid rearrangement initially because of the removal of the external training
stress. This external stress served to create and to support the dislocation array. In its
absence, thermal cycling causes the dislocation array to change. The other basis for the
existence of dislocation rearrangement comes from the work of Miyazaki et. al [7].
Dislocation introduction causes the increase of εM and εA. When the material
undergoes a MT, because of the imperfect nature of the transformation a significant
number of dislocations is generated [7]. The actual reason dislocation introduction should
cause the continuous increase in εM and εA in Two-way trained samples is unclear. This
increase has not been observed in samples without Two-way training [4]. The detwinning
plateau stress increases significantly after thermal cycling, as can be seen in Fig. 9.
Properly oriented dislocations can lower the plateau stress in a given loading direction,
but improperly oriented dislocations introduced by thermal cycling raise it by impeding
the detwinning process. This is the main confirmation that the second stage of
 degradation is due to dislocations introduced from the MT [4]. The rate of dislocation
introduction correlates to the rate of length increase [4]. Hard NiTi samples resist
dislocation introduction from mechanical deformation and from thermal cycling, while
soft samples allow dislocations to be introduced more easily by both means. This means
that soft samples have higher rates of length increase in the second stage of degradation
than hard samples.
Even though the degradation of the Two-way effect appears to be separated into
two separate regions by two different mechanisms, the point where one mechanism takes
over and the other stops is not clear. Looking closely at Fig. 8 reveals that dislocation
rearrangement apparently ends at different points for the Austenite and Martensite strains,
even though in other results they appear to coincide [4]. The place where the Austenite
and Martensite strains begin increasing does not mark the moment where dislocation
rearrangement stops, it simply marks the point where the strain decrease caused by
dislocation rearrangement is less than the strain increase by dislocation introduction.
Therefore in determining which mechanism is dominant in the decrease in the Two-way
strain, it cannot be assumed that only one mechanism is operating at a time.

4 4
3.5 3.5 0.298 A
0.267 A
3 3 0.247 A
2.5 0.235 A
2.5
Stress (N)

0.20 A
Stress (N)

2
2
1.5
1.5
1 Before Training
1
After Training
0.5 After 3000 0.5
0
0
0 2 4 6 8
Strain (%) 0 2 Strain (%) 4 6

Fig. 9 Stress v. Strain at Different Stages Fig. 10 Stress v. Strain after 3000 Cycles

The introduction of dislocations from thermal cycling is not a significant factor in

the decrease of εTW. The stress-strain curves for all samples tested after thermal cycling
showed detwinning at similar forces, see Fig. 10. This suggests that a similar number of
dislocations were generated in each sample despite great variations in the amount of
degradation. Additional information about the role of dislocation introduction can be
found from closer observation of the stress-strain curves. During training, properly
oriented dislocations introduced during deformation lower the stress for detwinning by
adding an internal bias stress. The resulting distortion of the stress-strain curve is shown
in Fig. 11 for a sample that was not annealed prior to training.
When dislocations are introduced from thermal cycling, they inhibit detwinning
and effectively decrease the internal bias stress, raising the detwinning portion of the
curve. If this were the main cause of Two-way strain loss, then the detwinning regions for
all the ending stress-strain curves in Fig. 10 would be the same length, they would just

i
occur at different stresses. The stress-strain curve would be effectively just shifted up or
down with greater or fewer thermal dislocations. Dislocation rearrangement is not the
main cause of degradation since all the detwinning regions occur at approximately the
same stress and have varying lengths. Dislocation rearrangement misaligns and decreases
the trained stress field, giving strain back to the detwinning plateau at a set stress, rather
than starting from below 0 N and raising the detwinning region from dislocation
introduction over many thermal cycles.
4

3.5

2.5
Stress(N)

1.5

0.5

0
0 1 2 3 4 5 6 7
Strain(%)

Fig. 11 2nd Training Deformation for Non-Treated Sample

Some additional observations about dislocation introduction can explain more
about the mechanism of dislocation rearrangement. As shown in Fig. 3 and 7 a steep drop
of the Martensite and Austenite lengths is usually followed by a steep increase in those
lengths which tapers off with more thermal cycles. Since the rate of length increase in the
second stage correlates to the rate of dislocation introduction, samples that experienced
more dislocation rearrangement (steeper drops) experienced more dislocation
introduction, at least initially, in the second stage. This means that dislocation
rearrangement involves softening, which is supported by other observations [7]. The
length increase tapers off because the increased dislocation introduction rapidly hardens
the sample. How this softening actually occurs is unclear.
The changes in the resistance-temperature relationship are shown from Two-way
training in Fig. 12. The sample before training had a simple transformation from
Martensite to Austenite and back. The curve for the trained material had the same
resistance-temperature slopes for the Austenite and Martensite states, but had increased
complexity in the transformation region. The main reason for this is the R-phase
transformation, which here occurred on both heating and cooling.
 Resistance vs. Current for Sample Before and After Training
5

4.5 Bef ore Training

Trained

Resistance (Arbitrary Scale) 3.5

2.5

1.5

0.5
0 0.05 0.1 0.15 0.2 0.25 0.3
Current (A)

Fig. 12 Resistance-Steady Current Curves Before and After Training

The R-phase transformation usually appears only on cooling and is marked by a
large peak in resistance. It is commonly the result of dislocations introduced by thermal
cycling. The R-phase on heating is marked by a similar peak, except that it generally
appears only with an applied external stress. This peak’s strength has been shown to
increase with increasing applied stress [8]. The trained internal stress plays the same role
as an external stress in this case, since the R-phase on heating is present.

4 2
Cycle 1
Cycle 32
3.5 0.22 A Cycle 63
R esistance (A rbitrary S cale)

1.5
Resistance (Arbitrary Scale)

0.25 A
3 Cycle 94
0.3 A
1
Cycle 125
2.5 Cycle 156

0.5
Cycle 187
2
Cycle 248
1.5 0 Cycle 309
0 0.05 0.1 0.15 0.2 0.25 0.3 Cycle 370
1 Cycle 431
-0.5
Current (A)
Cycle 492
0.5
Cycle 993
0 Cycle 2994
0 0.05 0.1 0.15 0.2 0.25 0.3
Current (A)
Fig. 13 Thermal Cycling of Trained Sample with 0.32 A and 0.0 A
 2

0.22 A
Resistance (Arbitrary Scale) 1.5
0.25 A
0.3 A
1

0.5

0
0 0.05 0.1 0.15 0.2 0.25 0.3

-0.5
Current (A)

Fig. 14 Resistance Curves for Samples Following 3000 Cycles

Fig. 13 shows the changes in the resistance-temperature curve of a sample over
3000 thermal cycles. Every thermal cycle the size of the peak for the R-phase
transformation on heating decreased. In Fig. 14 the ending resistance-temperature curves
for samples tested with different THs are given. The samples with lower cycling
temperatures showed stronger R-phase transformations on heating, while those with
higher cycling temperatures had the transformation practically eliminated after 3000
thermal cycles. Correspondingly the samples with low TH had less Two-way strain
degradation than samples with higher TH.
For all samples tested, the internal stress distribution became more isotropic with
thermal cycling, since the R-phase on heating always diminished. The samples with
higher TH showed a more rapid elimination of the R-phase on heating, showing that the
higher cycling temperatures increased the rate of dislocation rearrangement. The R-phase
is a high-resistance, angular distortion of the B2, Austenite phase. As the temperature
increases, the angle of the distortion decreases, decreasing the resistance of the phase [9].
This results in the gradual approach of the resistance-temperature slope to that of the
Austenite phase with heating which is observed in trained samples. The presence of this
intermediate phase shows that in the trained material it is less energetically favorable to
transform straight to the Austenite phase, than to bypass using the R-phase.
A good way to understand the results is to view the trained stress field as
consisting effectively of spring-like stresses between atoms. The transformation to
Martensite decreases the magnitude of the internal stress because the phase
accommodates by allowing the growth of different crystal variants depending on the local
stresses. The R-phase does this also to a lesser extent. Austenite does not accommodate,
so the stress inside the material is increased with the transformation to Austenite from
any other phase.
The higher internal stress associated with the transformation to Austenite
increases the probability of events where a dislocation changes, moves or is annihilated,
which is dislocation rearrangement, since there is more stored internal energy. This is
consistent with the observation that increasing the amount of Martensite transformed to
Austenite in a cycle, by using higher cycling temperatures, caused more degradation.
Changing the lowest temperature reached has little impact because the Martensite relaxes
the stress. However with the limited understanding of the actual process of dislocation
rearrangement a more thorough study of the influence of the lowest cycling temperature
should be conducted since with this rationale it cannot be ruled out that some
rearrangement events do not occur on cooling also.
Even with these conclusions it is difficult to ascertain why the damage done to the
internal stress field occurs over many thermal cycles rather than from just one cycle.
Additionally it is unclear whether or not the rate of degradation continues to increase with
higher THs or levels off because the transformation to Austenite completely finishes at
some temperature. Also it would be useful to understand how dislocation introduction
causes the test samples to increase in length. Future work in this area most importantly
should focus on identifying the actual process of dislocation rearrangement so that
avenues of degradation can be blocked to make the Two-way effect suitable for
applications.

Conclusions

1) The introduction of dislocations from thermal cycling does not significantly

degrade the Two-way strain, though causing dimensional instability.
2) The rearrangement of internal stresses, which cause the Two-way effect, is the
primary source of Two-way strain loss.
3) The amount of rearrangement during a thermal cycle is controlled by the highest
temperature reached during that thermal cycle (TH).
4) Dislocation rearrangement involves softening by some mechanism.
5) A high TH maximizes Two-way strain loss, while a low TH minimizes it.

References

1. Stalmans R, Humbeeck J Van, Delaey L, Acta Metall. Mater. 40: 2927 (1992)
2. Miller DA, Lagoudas DC, Smart Materials and Structures 9(5): 640-652 (2000)
3. Scherngell H, Kneissl AC. Scripta Mater. 39: 210 (1998)
4. Scherngell H, Kneissl AC. Acta Mater. 50: 335, 339, 340 (2002)
5. Stalmans R, Humbeeck J Van, Delaey L, Scripta Metall. Mater. 31: 205-212 (1994)
6. Liu Y, Xie Z, Humbeeck J Van, Delaey L, Scripta Mater. 41: 1274 (1999)
7. Miyazaki S, Igo Y, Otsuka K, Acta Metall. 34: 2049, 2051 (1986)
8. Uchil J, Mahesh KK, Ganesh Kumara K. Physica B 324: 419 (2002)
9. Novak V, Sittner P, Dayananda GN, Braz-Fernandes FM, Mahesh KK Materials
Science and Engineering A 481-482: 429 (2008)

Acknowledgements
I would like to thank my electronics teacher Mr. Bell and the good people at NRL for
taking the time to help me get started. I greatly appreciate Jeff Brown and Dynalloy© for
cooperating with me and getting me the material I needed free of charge. I thank Mr.
Latham for advising me on my project. Finally I thank my parents for letting me use their
basement for this work.