JOURNAL DE PHYSIQUE IV
Colloque C8, supplkment au Journal de Physique III, Volume 6, dicembre 1996 C8-371
Damping Properties of Shape Memory Alloys During Phase Transformation
J. Van Humbeeck
Dep. MTM, W. de Croylaan 2, 3001 Hevei-lee,Belgium
Abstract. A high damping capacity is considered as one of the important functional properties
of shape memory alloys. Those properties are related to a thermoelastic martensitic
transformation. As a consequence of this transformation, the internal friction or damping can
be investigated for three different states: 1. during thermal transformation cycling, 2. during
martensite induced strain cycling at constant temperature, 3. in the martensitic state.
The difficulty of formulating a unified theory to those three approaches is related to the
complexity of the microstructure. Phase boundaries, dislocations, precipitates, impurities and
point-defects interact strongly with each other, thereby influencing the macroscopic
observation.
However, improved models have been recently formulated to describe the observed internal
friction during transformation, taking into account the change in defect mobility. This has also
lead to a new approach of analyzing the observed damping in an amplitude dependent and
amplitude independent part.
Special attention is also given to superelastic damping with the aim of improving earthquake
resistance of constructions, concentrating reversible hysteretic behaviour in detailed regions of
the structure. Moreover, new relations between other functional properties and the internal
friction of the martensite have been established.
This paper will describe the recent findings, the importance of the results, fundamentally and
practically, concurrently pointing to some interesting topics for further research.
INTRODUCTION
The functional properties of shape memory alloys are related to a (largely) reversible (non-)
thermoelastic martensitic transformation, generally in a temperature region between -lOO°C and +200°C.
A high damping capacity is considered as an important functional property of the present shape memory
alloys. As a consequence a lot of research has been undertaken to quantify and optimize those damping
properties as function of composition and especially the thermomechanical processing. Indeed, since the
high internal friction, especially during transformation and in the martensite state, is strongly related
with the mobility of martensitelmartensite and parent/martensite interfaces and structural defects inherent
to the martensitic phase, the distribution and concentration of several types of defects (vacancies,
dislocations, grain boundaries, precipitates), controlled by the thermomechanical processing, play an
important controlling role in the global internal friction spectrum. Since this internal friction spectrum is
thus strongly linked to the microstructure and the presence of defects, it is not always evident to come to
a unique conclusion for the explanation of the observed behaviour.
Article published online by EDP Sciences and available at [Link]
C8-372 JOURNAL DE PHYSIQUE IV
At the other side, the study of the internal friction has lead to a better understanding of the relation
between the microstructure including the defects and the other functional properties of shape memory
alloys such as the one- and two way memory effect and the pseudoelasticity.
A part that is hardly touched in research is using internal friction to reveal the internal structure and
structural defects especially of the martensitic phase. A lot of activity related to internal friction of shape
memory alloys was going on until the mid eighties, then became steadily decreasing but gained during
recent years again a significant interest as well from a fundamental approach as well for application
reasons.
This paper will review the latest contributions and models and discuss some interesting features of using
the capacity of high energy absorption during vibrations. Earlier more detailed reviews can be found in
ref.[1,2,3,4].
1. INTERNAL FRICTION DURING MARTENSITIC TRANSFORMATION
During the martensitic transformation, an internal friction peak is observed concurrently with a strong
modulus minimum.
As pointed out by Bidaux et al. [5], in materials in which the two phases connected by the phase
transformation can coexist over a certain temperature range, one should consider three separate
contributions to the total internal friction, illustrated in Fig.1.
XI* >Is
Figure 1: Schematic representation of Q" and its contributions during martensitic phase transformations.
QT: is the transient part of Q:, and it exists only during cooling or heating (T # 0). It depends on
external parameters like temperature rate (T), resonance frequency ( f ) and oscillation amplitude oo.Q;:
depends on the transformation kinetics and is therefore proportional to the volume fraction which is
transformed per unit of time. Q& is related to mechanisms of the phase transformation (PT), which are
independent of the transformation rate, such as the movement of parent/martensite and
mmensitefn~artensiteinterfaces. Q& exhibits a small peak when the interface mobility is maximum.
Qz:is composed of the IF contributions of each phase and is strongly dependent on microstructural
properties (interface density, vacancies), especially in the martensitic phase.
1.1 The Intrinsic IF Q:;
This factor takes into account the intrinsic damping generated in the two coexisting phases. It can be
written as:
Q : ~ ~and Q:~ are the intrinsic damping of the martensite and the P phase, respectively, and V is the
volume fraction of the martensite phase.
1.2. The Non-transient IF peak, Q&
For an explanation of the Q& peak one can find different approaches in the literature [6]. The model of
Dejonghe et al. [7] considers a maximum of the interface mobility in the transformation region. This
maximum is expressed by a maximum amount of transformed volume fraction per stress unit (an/ao(~)).
When the temperature rate is zero or if the oscillation frequency is very high (f > 50 HZ), the transient IF
goes to zero and only the amplitude dependent part remains.
Mercier and Melton [8] proposed a dislocation mechanism, which takes into account the effect of
anisotropy (elastic softening) on the elastic energy of dislocations. They showed that the dislocation
damping in a crystal is inversely proportional to the crystallographic anisotropy factor (A), but
proportional to frequency.
Koshimizu 191 calculated the IF by using a Landau approximation of a first-order phase transition. In a
thermodynamic treatment, the Gibbs free energy is developed as a function of stress (o),temperature (T)
and an internal variable (0,called the order parameter. Considering a relaxation type of damping, the
calculations show discontinuities of Q& and the relaxed modulus at the transition temperatures.
Existing concepts for QG can in fact be divided in two classes: theories in which Q; is considered to
be the result of the coexistence of macroscopic amounts of martensitic and austenitic phase [5,7] and the
"pretransformational" models, where QG originates from the existence of submicroscopical nuclei of
the martensitic phase [8,10]. Strongest contradictions between these models are related to the frequency
and the amplitude dependence of Q$ .
Recently, S. Kustov et al. [ l l ] studied Q$ in CuAlNi single crystals at ultrasonic frequencies, for the
amplitude dependent (ADIF) and the amplitude independent (AIIF) part of the phase transition internal
friction.
The experiments were carried out at frequencies of approximately 100 kHz by the PUCOT technique,
rendering negligible, QG .
The internal friction spectra measurements were performed for each temperature step in the amplitude
independent and amplitude dependent ranges, corresponding to respectively the strain amplitude 2*10-7
and 8*104 - 3*1u5.
Several observations were made:
1. The AIIF peak lies between As and A, on heating and between Msand M, on cooling.
2. In the range of strain amplitudes investigated no amplitude dependence is revealed in the austenite
far enough from the transformation temperatures. IF is amplitude dependent both in martensite and
in the transformation temperature range. ADIF exhibits peaks both on heating and on cooling.
3. The start of the increase of the resonance frequency during cooling corresponds to the Ms
temperature.
4. The ADIF peak on cooling is shifted towards higher temperatures compared to the AIIF peak. The
ADIF peak on cooling occurs close to Ms. The ADIF peak is accompanied by a maximum of the
amplitude dependent part of the resonant frequency change, also at Ms.
C8-374 JOURNAL DE PHYSIQUE IV
5. ADIF appears about 15OC above Ms. Below Ma, when the resonant frequency begins to change
rapidly, both ADIF and amplitude dependence of the resonant frequency decrease.
Based on a more detailed analysis of these observations, the hypothesis can be proposed that during
direct transformation two different AIIF, and ADIFptmechanisms coexist which, unfortunately, are quite
difficult to separate in the experimental data: the first one, purely premartensitic (IF1) and the second
one, due to the coexistence of two phases (IF2). Those results appear consistent with the assumptions,
that IF1 has comparatively low values for AIIFpt in CuAlNi single crystals, but rather pronounced
amplitude dependence, and that IF2, on the contrary, has high AIIF?, values and much less pronounced
.4DIFptat ultrasonic frequencies, as compared to IF1. A schematic representation of the behaviour of
these two IF components is given in Fig. 2. Further experiments are needed to check the validity of the
present model for polycrystalline material and for other types of alloys.
ADIFpt
Figure 2: Schematic representation of the temperature dependence of the amplitude-independent (a) and amplitude-
dependent @) IF components due to premartensitic phenomena (1) and coexistence of two phases (2).
1.3. The Transient Part of IF, Q:;
In the frequency range (f = 1 Hz) the transient part (Q;:) is the most important contribution to the IF
peak. The starting point for any IF modelling is the basic equation for Q-':
where J is the elastic compliance, oo is the oscillation amplitude, a is the angular frequency (w = 2x0
and dem is the anelastic or irreversible part of the deformation during one oscillation cycle.
The first attempt to explain the transient character of the IF peak during martensitic transformation dates
from 1956, when Scheil and Miiller [12] related QF: to the transformation rate (anfat). The anelastic
deformation (dean) in equation (3) arises from the lattice deformation that occurs when material is
transformed during the IF measurement:
an
de, = k.-dt
at
The basic idea which is expressed by a proportionality between dean and an/&, has been maintained in
all the following models. The crucial point for the further development of an expression for Q T ~has
been defining the influence of temperature (T) and oscillation amplitude (0,) on the anelastic
deformation (dean) as well as on the transformation rate (anlat) [13]. A thermally activated type of
anelastic behaviour is based on the model of Belko et al. [14]. dean is due to the thermally activated
formation and growth of nuclei of the new phase. The associated activation energy necessary to pass
from the state of nucleation to the state of growth is modified by the oscillation stress. However, this
model, which is based on a thermally activated type of anelasticity, does not seem to be appropriate for
describing the IF peak found in alloys, where the transformation is strictly athermic.
The IF models that are presented in the following are based on an athermic or hysteretic type of anelastic
deformation and can be applied for martensitic transformation in shape memory alloys. They are
classified in Table I with respect to their suppositions for the oscillation amplitude dependence of dean
and an~at.
Model of Transformation "Plasticity" : de, (0,
an/&)
Delorme and Gobin [15] have proposed an expression for dean, which is 6ased on the analysis of static
"transformation plasticity" experiments. They found that the transformation deformation (d?) increases
linearly with the transformed volume fraction and the applied stress,
and assumed that dean obeys the same law during dynamic measurements (Table I).
They derived the following expression:
As an extension of Delorme's model of transformation plasticity, Dejonghe et al. [7] added a stress-
dependent term of the transformation rate. Then anfat becomes :
This equation accounts for the possibility that material transforms under the action of the applied
oscillating stress. Furthermore, Dejonghe et al. [7] considered dn to include the amount of martensite
which retransforms during one oscillation. Assuming that there is a critical stress (oc) necessary to
reorientate the already existing variants or to stress-induce new variants, QT! becomes :
Incorporating the possibility of stress-induced transformation, this explains the experimental observation
that a small IF peak exists for T = 0 and that the height of this non-transient IF peak is dependent on oo.
Model of "Pure" Transformation strain : de, (an/at)
However, there is one discrepancy between the experimental results and the equations for Q: as
presented in the model of Delorme. Experimental results showed that does not increase linearly
with T/O. This problem led Gremaud et al. [16] to develop a new model, which defines the role of the
oscillating measuring stress for the transformation according to the Clausius-Clapeyron equation.
C8-376 JOURNAL DE PHYSIQUE IV
In the new approach of Gremaud et al. the anelastic deformation (dean) is related to a "pure"
transformation strain, as has been proposed by Delorme in the second part of his model. The important
point of the new model is that it takes into account the influence of the applied stress amplitude (oo) on
the transformation rate (an/&). This occurs in the sense that the thermodynamic equilibrium of the
material changes and the variants which are favoured by oo are induced to transform. Following this
idea an/& becomes equal to:
where a is the Clausius-Clapeyron factor (a = v~'/As), v is the molar volume and AS is the entropy
change of the transformation.
Finally J. Stoiber [6] studied in very much detail monocrystalline and polycrystalline materials. The
internal friction spectrum observed during martensitic transformation in Cu-Zn-A1 alloys show
significant differences when changing the microstructure of the specimen. J. Stoiber studied the effects
of grain refinement and two-way shape memory training on the internal friction and found that the
observed Q-' peak is dependent on the microstructural properties:
- Grain refined specimens show a smaller Q-' peak than specimens with a large grain size.
- Specimens with a two-way memory effect have a broad symmetric Q-' peak with a relatively low
Q-' maximum.
From the observations a model has been developed in which the fragmentation of the martensite is taken
into account [17,18]. The quantitative description of variant interactions can be conducted in terms of
the model of fractal geometry [19].
The non-linear increase of the transient IF as a function of n can be. explained in terms of the
"fragmentation" parameter, x (n), if one assumes that the additional anelastic deformation, which is due
to the migration of intervariant boundaries (martensitelmartensite interfaces), increases with each
fragmentation.
The equation for the transient part of the IF peak (QF;) that is observed during martensitic phase
transformations in shape memory alloys, Q;: is now formulated as follows:
The new Q;: equation accounts for the dependence of the transient IF peak on both the transformation
volume fraction per cycle (An) and the intensity of martensite variant interactions (x (n)), which depends
on microstructural properties of the present phase (e.g. grain size) or the transformation conditions.
The model of Stoiber can now be fitted correctly to any experimentally measured internal friction peak
and can explain the differences in width orland height of the different peaks, depending on the
underlying microstructure.
1.4. The ?-effect and its consequences
The previous models describe an inherent relation between T and Q.: Generally the internal friction or
damping capacity of shape memory alloys are measured at a constant heating or cooling rate and thus a
significant damping peak appears during transformation. However, at T=O, thus at constant temperature,
Q?: becomes zero. This is also actually seen during the experiment and the remaining damping capacity
remains a summation of Q;f and Q&. As a consequence, the damping capacity becomes a function of
the volume fraction of the martensite, and the very high damping capacity during transformation is lost.
Table 1: Existing models of the hig damping peak dunng martensite transformations
Q"' = j o"deari) o-a0 sincot
JtJoi
3n
Approach I: de an = k. a - ^ dt
at
(transformation "plasticity")
3n _ 3n •
3t _ 3 T
Delorme
1= k8nt
n -i
VTr
J 3T co
Approach II: de an = k. cr — dt
at
("pure" transformation strain)
3n _ 3n •
9t _ 3 T
Delorme Gremaud-Bidaux
os
4k 3n f
1 - - 2a
t
JCJ 3T coao i _ 2K 3n OJCTQ T T 2a
QF' = J 3T « CO «> 0"o 3it
1 +•
2a (O0O
k dn 2a 2a T 2a
Qn' = 4J3T
ro<5o » 3x <oo"o t
_k_3n_T_ t 2a
Qm =
2J dT <B<j0' roao «
(3-378 JOURNAL DE PHYSIQUE IV
For applications where a continuous vibration is applied at constant temperature a 100 % martensite
condition offers a more stable damping.
In the case of impact loadings with very low frequency, the two-phase region can be interesting since the
martensite will be now stress-induced concurrent with an exothermic heat effect.
2. RELAXATION PEAKS IN MARTENSITE
Several types of relaxation peaks have been reported. For Ni-Ti, Hasiguti and Iwasaki [20] and
Postnikow et al. [21] report a peak at 203 K with an activation energy of 0.38 eV and a frequency factor
of 6x10' s-'. This peak might be due to a movement of dislocations or a movement of point-defect-
pinned dislocations.
Tirbonod and Koshirnizu 1227 report a Bordoni type peak in an equiatomic Ni-Ti alloy. The presence of
this peak should be due to the motion of imperfect dislocations with a Burgers vector smaller than the
lattice parameter. These can be twinning dislocations but the dislocations situated at the interface
between the different variants can not be ruled out. Experiments on monovariant specimens could make
this matter clearer.
Cu-Al-Ni alloys with I'-martensite exhibit a twin relaxation peak as observed by Suzuki et al. [23]. The
intensity of this peak is even higher than the one of the transformation peak. It is also strain-amplitude-
dependent in a range from 5x10~'to 3x10". From the peak shift method an activation energy of 7.8x104
Jlmol was calculated.
A similar twin relaxation peak with an activation energy at 5 . 1 5 ~ 1 0Jlmol
~ has also been derived in Cu-
Mn [24].
Recently a broad peak in the Au-Cd system has been presented by Y. Murakarni et al. [25]. This peak is
also related with the twin boundaries motion and could be a relaxational effect. It appears that the twin
boundaries became less mobile with ageing, leading to changes in the internal friction spectrum.
Finally, the existence of an amplitude-dependent relaxational internal friction peak due to a competition
between depinning and dragging interactions between dislocations and pinning points has been revealed
by S. Kustov in 6 Cu-Al-Ni martensite 1261.
3. APPLICATION ORIENTED RESEARCH BASED ON THE MARTENSITIC
TRANSFORMATION
3.1 Damping by pseudoelastic loading
When shape memory alloys are loaded above the Aitemperature, martensite will be induced at a critical
stress (Fig. 3). This stress remains almost constant during further straining, until the sample is
completely martensite. Further straining will lead again to a stress increase, due to elastic loading of the
martensite.
Unloading leads to the reverse transformation endingat the initial shape of the sample. The unloading
stress-plateau however is below the loading stress-plateau, creating a hysteresis loop.
I 1
strain
Figure 3: A pseudoelastic loop showing the hysteresis after loading and unloading.
The area enclosed by the hysteresis loop is proportional to the energy absorbed and thus to the damping
capacity during this type of transformation.
Since large superelastic strains can be obtained up to 8% in polycrystalline Ni-Ti, this property is now
very much under investigation for applications in large structures such as buildings in seismic active
regions [27] or to damp low frequency vibrations in space structures [28].
For those applications a high loading and low unloading plateau is optimal. This is reached by alloys
with a large transformation hysteresis such as Ni-Ti-Nb alloys. However the system shows two
disadvantages:
1. The stresses at which the transformation occurs, increase linearly with temperature which can be
explained by the Clausius-Clapeyronequation:
Age-"'
-=- AS
AT Emu
where oP" is the critical stress during loading, T the absolute temperature, AS the entropy change
and E- the total transformation strain.
2. The damping is frequency dependent. The reason is that during the forward transformation
(beta+martensite, loading) the transformation occurs exothermic. During unloading, the
transformation occurs endothermic. The combination of the hysteresis and the heat effects influence
the total damping capacity that goes through a maximum as function of temperature [29].
3.2 Smart materials
Shape memory alloys can be used as actuators in so-called smart materials. One of the envisaged
application is the damping of resonance vibrations of panels. For this purpose, strained shape memory
wires are embedded in the composite structure. In case of resonance vibration, the wires can be
externally heated to transform. Since the wire is constrained in the panel it cannot exhibit the memory
effect, thereby creating high internal stresses during transformation. Those stresses change the internal
energy of the plate leading to a shift of the resonance frequency so that the actual vibration will stop.
For the same type of purpose also the significant modulus-defect or modulus change during the
martensitic transformation has interesting possibilities. Depending on material type, texture orientation
and microstructure, large modulus changes can be obtained 1301 leading to significant changes in
resonance frequencies in a small temperature range.
Conclusion
During recent years significant progress has been made in identifying the influencing parameters and the
modellisation of the internal friction behaviour during the martensitic transformation of shape memory
alloys.
The complexity of the microstructure, the concentration, distribution and interaction of different types of
defects however must be taken into account when results are compared for identical and different
systems.
The explosive growth of shape memory alloy applications has also raised the question of transformation
and phase stability, the influence of frequency and temperature and the influence of the
thermomechanical processing leading to different microstructures.
Observations of relaxational effects during the transformation and in the martensitic state can help in the
further understanding of these fascinating structures.
This paper has paid mainly attention to the internal friction during transformation. But it should be
stated that internal friction is still an underestimated tool for the study of the martensitic phase itself. As
an example, Y. Liu et al. [31] have recently shown that a linear relation exists between the reorientation
stress level in Ni-Ti alloys at strains between 2 and 6%, stresses between 150 and 200 MPa, and the
C8-380 JOURNAL DE PHYSIQUE IV
damping capacity of the martensitic phase at room temperatures where strains of only 10.' are applied
[281.
So there is an appeal to internal friction to become a handsome tool for the investigation of other
functional properties of shape memory alloys and the understanding of the role of the microstructure on
those properties.
References
[I] J. VAN HUMBEECK in S. Garczyca, L.B. Magalas (Eds.), Proc. Summer School on Internal
Friction in Solids, Wydawnitco AGH, Krakow (1984), 131.
[2] J. VAN HUMBEECK in B.B. Rath, M.S. Misra (Eds.), Proc. Int. Symp. on Role of Interfaces on
Material Damping, ASM, Metal Park, OH (1985), 5.
[3] J. VAN HUMBEECK, J. STOIBER, L. DELAEY, R. GOTTHARDT, Zeitschrift fur Metallkunde
86 (1995), 3, 176.
[4] W. YENING, C. XIAOHUA, S. HUIMIN, Proc. ICIFUAS-9 (1990, Jan), Intern. Acad. Publishers,
Pergamon Press, 305.
[5] J.E. BIDAUX, R. SCHALLER, W. BENOIT, Acta Met. 37 (1989), 803.
[6] J. STOIBER, Ph.D. Thesis, Lausanne-EPFL, These nr. 1115 (1993).
[7] W. DEJONGHE, R. DE BATIST, L. DELAEY, Scripta Metall. 10 (1976), 1125.
[8] 0. MERCER, K. MELTON, Scipta Metall. 10 (1976), 107.
[9] S. KOSHIMIZU, Ph. D. Thesis, Lausanne-EPFL, Thbse nr. 416 (1981).
[lo] P.C. CLAPP, Mat. Sci. Eng. 38 (1979), 193.
[ll] S. KUSTOV, J. VAN HUMBEECK, R. DE BATIST, Scripta Met. et Mat. 33 (1995), 1401.
[12] E. SCHEIL, J. M ~ ~ L L E R , Eisenhuttenwesen 27 (1956), 80.
Arch.
[13] W. YENING, Y. ZHENJI, Z. HE, M. JINCHAO, S. Nanying Univ. 2 (1963), 1.
[14] V.N. BELKO, B.M. DARINSKY, V.S. POSTNIKOV, I. SHARSHAKOV, Phys. Met. Metallogr.
27 (1969), 140.
[15] J.F. DELORME, P.F. GOBIN: MCtaux 573 (1973) 185,574 (1973), 209.
[16] G. GREMAUD, J.E. BIDAUX, W. BENOIT, Helv. Physica Acta 60 (1987), 947.
[17] J. STOIBER, J. VAN HUMBEECK, R. GOTTHARDT, Proc. M ~ D ,Mechanics and Mechanisms
of Materials Damping, ASTM STPl169, ASTM, Philadelphia (1992), 45.
[18] J. STOIBER, J.E. BIDAUX, R. GOlTHARDT, Acta Met. Mat. 42 (1994), 4059.
[19] E. HORNBOGEN, Zeitschrift fur Metallkunde 10 (1987), 352.
[20] R.R. HASIGUTI, K. IWASAKI, J. Appl. Phys. 59 (1968), 2182.
[21] V.S. POSTNIKOV, V.S. LEBEDINSKIY, V.A. YEVSYOKOV, I.M. SHARSKAKOV, M.S.
PESIN, Fiz. Met. Metallov. 29 (1970), 34.
[22] B. TIRBONOD, S. KOSHIMIZU, J. de Phys. 42 (1981), C5-1043.
[23] K. SUZUKI, N. NAKANISHI, H. MITAMI, JIM 44 (1980), 43.
[24] K. SUGIMOTO, T. MORI, S. SHIODE, Metal Science Journal 7 (1973), 103-108.
[25] Y. MURAKAMI, K. OTSUKA, H. MIZUBAYASHI, T. SUZUKI, Proc. Wayman Conf., May,
1996, Ed. TMS, to be published.
[26] S. KUSTOV, S. GOLYADIN, J. VAN HUMBEECK, R. DE BATIST, this conference.
[27] P.W. CLARK, I.D. AIKEN, J.M. KELLY, SPIE, VOL. 2445,241.
[28] Y.C. YIU, M.E. REGELBRUGGE, Amer. Inst. of Aern. and Astronautics, AIAA-95-1120-CP,
3390.
[29] J. VAN HUMBEECK, L. DELAEY, Journal de Physique, Tome 42 (1981), C5-1007.
[30] I. HURTADO, J. VAN HUMBEECK, MOO CHI-MEI, Spie Vol. 2441,13 1.
[31] Y. LIU, J. VAN HUMBEECK, R. STALMANS, L. DELAEY, to be published, Proc. Int.
Wayman-Symposium, May 1996, Ed. TMS.