Thermochimica Acta, 203 (1992) 419-444 419

Elsevier Science Publishers B.V., Amsterdam

Martensitic transformations in shape-memory alloys.

Successes and failures of thermal analysis and calorimetry

V. Torra a and H. Tachoire b

aApplied Physics Department, ETSECCIP, UPC, Jordi Girona Salgado 31,
E-08034 Barcelona (Spain)
b Thermochemistry Laboratory, University of Provence (Air-Marseille I), Place Victor Hugo,
F-13331 Marseille Cedex 03 (France)
(Received 28 August 1991)

The development of adapted equipment for high-resolution thermal analysis (thermomi-

croscopy, acoustic emission, stress-strain and heat conduction calorimetry) provides macro-
scopically relevant information about the microscopic interactions and the effect of the
defects in martensitic transformations of Cu-based shape-memory alloys, e.g., levels of
resolution and evolution time scales, intrinsic thermoelasticity, separation between the
effects of nucleation and friction. Furthermore, making direct measurements and applying a
simple thermodynamic formulation directly relates the friction work (mechanical, analogous
to the entropy production) with the width of the calorimetrically calculated hysteresis cycle.
Thus, some thermodynamic difficulties found in the measurements and encountered in
recent interpretations in the literature are avoided.

INTRODUCTION

In metal alloys, in Ni-Ti (nitinol), copper- or iron-based, in organic

materials (queratine) or plastics, in ion crystals, etc., a martensitic transfor-
mation appears which is crystallographically reversible. Practical interest in
the transformation is associated with its complete recovery when retrans-
formation is made. For this, it is only necessary to reverse the mechanical
or thermal action (mechanical stress or cooling) which induces direct
transformation. The particular behaviour of the material is associated with
the existence of a first-order martensitic transformation without diffusion,
called thermoelastic, between two metastable phases. The austenite phase
(parent phase) is obtained by quenching from its region of stable existence.

Correspondence to: V. Torra, Applied Physics Department, ETSECCIP, UPC, Jordi Girona
Salgado 31, E-08034 Barcelona, Spain.
Dedicated to Professor Joseph H. Flynn in honour of his 70th birthday.

0040-6031/92/$05.00 0 1992 - Elsevier Science Publishers B.V. All rights reserved

420 I/: Torta and H. Tachoire/ Thermochim. Acta 203 (19921419 - 444

For example in the case of alloys of Cu-Zn-Al, a body-~entred cube (bee)

phase, frozen by quenching from the region of temperatures between 500
and 85O”C, is converted into a face-centred tetragonal phase (fct), which is
somewhat monoclinic, when it undergoes martensitic transformation (see
refs. l-7 and related references).
The phase change is accompanied by a relatively small volume change
(from /3 to martensite, AV/V= -0.08 in Cu-Zn-Al 18R [S]) and the
phenomenological interpretation of the transformation is made from the
dilations and contractions of the axes of the crystal, together with an
appropriate shear. The positions of the various atoms of the crystal involve
the pa~i~ipation of several atomic layers in the elemental cell. In this way,
martensites 2H (hexagonal), 3R, 6R, 9R and 18R appear. The crystallo-
graphic characteristics and phenomenologicai mechanisms of transforma-
tion are known for standard materials, and relevant data can be found in
the bibliography.
The industrial applications of memory alloys (shape-memory alloys) are
associated with the possibility of “educating” or training the material. This
makes it possible to have two forms: “hot” or austenite, and “cold” or
martensite. Furthermore, there is a mechanical force associated with the
heating process (up to 70 kp mm-‘), when passing from the cold form to
the hot form. The applications are associated with the construction of an
indicator which is activated by a slight change in temperature. In fact, a few
degrees are sufficient to pass from one phase to the other. From the
transformation characteristics, the construction of “micromotors” may be
considered, which would produce a progressive stress in relation to the
temperature. In particular, the macroscopic thermoelasticity or pseudo-
elastic effect involving the relation between stress, strain and temperature,
establishes that a continuous stress field is present depending on the values
of temperature and deformation.
The particular behavioural features of this type of material during
transformation have found a restricted area of application as actuators. Its
properties are equivalent to those of systems made up of elements cur-
rently available: the~ometric sensors, stepper motors, springs and in
complement, operational electronics and its batteries. Their replacement
by a single device, operated by an alloy, is very interesting, particularly in
aggressive environments. In fact, electronic equipment is very sensitive to
the action of ionising radiation. It is unable to work in electromagnetic or
particle fields.
Temperature control systems are available on the market that are based
on memory alloys of Cu-Zn-Al and which can be used a limited number
of times. Thus, connection systems are built in avionics and prototypes of
heat engines have been built which use small differences in temperature
for the development of replacement energy sources. But their use as
custom-built parts in standard equipment (car industry, home computer
K Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444 421

connection systems, etc.) involves having a high economic competitiveness

together with a remarkable reproducibility and high behavioural reliability.
For example, from the two memory effect, tens of thousands of complete
working cycles (austenite/martensite) in two to five years are expected. In
this sense, it is also necessary to have predictable knowledge of the surface
of state (stress, strain, temperature), such that the material will not degrade
with time (ageing, fatigue) any more than the materials currently used and
that it will be much more economical.
Experimental analyses of the behaviour of the alloys [9,10] (hysteresis
cycles, diagrams of stress, strain and temperature) have revealed the
importance of atomic order, of defects (dislocations, precipitation), of the
size of the grains in the polycrystals [ll], of the effects of the surface state,
etc., on the macroscopic characteristics of the material. In particular, the
actions of repetitive thermomechanical treatments (cycling) produce mate-
rial ageing and fatigue. This tells us that the microscopic state and its
evolution have a great influence on the macroscopic behaviour and, as a
result, on the “thermodynamic equation of state”. Given the industrial
interest in having a quantitative form of the equation of state, it will be
necessary to include in it the structural parameters. All this involves having
comparative, global and local analyses of the transformation, and system-
atized phenomenological information.
In recent years, a set of data, frequently contradictory, complementary
or irrelevant (see, for instance in ref. 12, how the increase in iUS after
quenching is explained by an isothermal transformation mechanism), have
progressively described in detail, by means of macroscopic measurements,
the behaviour of the material. Acquiring the measurements has required
the development of adapted instrumentation. New equipment has pro-
duced results which required new interpretations in structural terms.
Frequently the microscopic effects are very slight and very difficult to
detect with standard characterization techniques (TEM, X-rays). In fact,
causal and clearly visible changes in some degrees or tenths of a degree in
the characteristic transformation temperatures, which in some cases can be
associated with atomic evolutions, are practically undetectable by means of
crystallographic techniques. Also, the main lines of research, necessary for
the systematic and predictive descriptions of material behaviour, have
produced a series of papers leading to the formulation of an equation of
state [13-201. In some cases, formulations based on interpretations of the
hysteresis cycle from the calorimetric results violate the laws of thermody-
namics; for instance, in ref. 21, the efficiency of the thermal cycle (see the
cycling processes in Fig. 1 and the energy absorbed/released in Fig. 2 of
ref. 21) is ten times the efficiency of the equivalent Carnot cycle.
In this study we present an outline of the systems of high-resolution
thermal analysis adapted for studying memory alloys, some examples of
their possibilities for detecting small structural evolutions and a simple
422 K Terra and H. Tachoire f Thermochim. Acta 203 (1992) 419 - 444

thermodynamic analysis of the hysteresis cycles which correctly produces

the magnitudes of interest from the calorimetric thermograms.

THE EXPERIMENTAL SET-UP IN HIGH-RESOLUTION THERMAL ANALYSIS

(HRTA)

The calorimetric measurements of transformation-retransformation cy-

cles make it possible to establish that, for each type of alloy or composition,
the material passes from one phase to another within a region of some tens
of degrees and that, for Cu-Zn-Al, the energy released is around 6 J g-i.
When samples of around 100 mg (a volume close to 10 mm31 are used,
conventional devices can provide the thermodynamic info~ation which,
according to the aims, are sufficient. Commercial DSCs (differential scan-
ning calorimeters) provide a resolution close to 10 FV and have a calori-
metric sensitivity of around 200 mV W-‘. The resolution in temperature is
0.1 K and the working area extends from the temperature of liquid
nitrogen (around 100 K> to 850 K. In DTAs (differential thermal analysers),
the sensitivities are lower but the working area may reach or exceed 1500
K, surpassing the melting temperature of the alloys of interest.
The average characteristics of the systems available on the market make
it possible to reach a certain level in the study of the materials. The
suitabili~ of a given ins~ument will depend on the type of process required
and, where appropriate, on the type of material. When a study is made of a
martensitic transformation, phenomena are seen that depend on the type
of instrumentation used. In fact, acoustic emission has revealed the com-
plexity of the transformation process. The introduction of heat flux or
conduction calorimetry made it possible to establish correlations between
the acoustic emission and the thermokinetics of the process. The tests
carried out to calculate the correlation established that the complexity of
the acoustic emission demands that thermal analyses have an acceptably
“equivalent” separating power. This means a resolution in the temperature
of the order of (or higher than) 0.01 K. Similarly, calorimetric resolution
(heat flux or conduction ~alorimet~) in the measurement of energy is lower
than the possibilities of ultrasound detection. In impulse transformations it
is of the order of 25 p J. Acoustic impulses in the observations of j3 + j3’
(Cu-Zn-Al all oys ) are situated in a domain between 0.1 and 10e7 ,uJ
(resolution limit of the equipment used) [22,23]. Furthermore, experimental
analyses have made it possible to establish that both techniques produce
complementary information concerning the transformation process. Calori-
metrically, the energies associated with the transformed masses are de-
tected, and, acoustically, the release of the interphase pinnings (unpinning
processes) or the nucleation processes are detected.
The set of conditions imposed in carrying out the experiment suggests
that transformation should be studied so that the phenomena under study
K Terra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444 423

Fig. 1. High-resolution thermal analysis or HRTA, CTT-UPC patent: T*, temperature of

the external thermostat; A, thermobattery (Peltier-effect plate) powered by a programmable
current intensity I(t); B, working area made up of a copper plate at temperature T,(t); C,
Pt-100 resistance; a calorimetric system situated in the working space and operating at
temperature TO,;D and D’, calorimetric detectors; E and E’, sample and reference.

may be separated. In order to do this, we choose to operate in restricted

temperature areas situated close to room temperature. The basic aim is to
have rigorous control and automation that is easy to perform. With these
limitations, the Peltier effect is very suitable because the cooling/heating
processes are performed without difficulty from a programming of the
intensity of the Peltier current. The signal processing associated with a
systematic modelling of the experimental device makes it possible to
compensate for the instrumental effects and to smooth the effect of the
noise. The various inst~ments and devices for high-resolution thermal
analysis operate from PC-XT- or 286-AT-type computers by means of
programs written in MICROSOFT ASSEMBLER and QUICK-BASIC language.
Figure 1 shows a thermal analysis stage together with adaptations to
convert it into a calorimetric system. In this case a thermostat at tempera-
ture T * is used. Temperature control and programming are made from an
approximation to the causal ratio between the intensity of the current
produced by the Peltier effect and the temperature of the working space.
In order to correct the differences between the values predicted and those
obtained, a supplemental on-line correction is introduced.
The working space may be used for measuring various parameters. It is
possible to follow the thermal analysis with an optical microscope (thermo-
microscopy) with a videotape recording system for further study. A me-
chanical traction device may be built, making it possible to calculate the
surface of state, stress (a), strain (E) and temperature (T). The electrical
resistance of the sample under study may be measured, or the associated
acoustic emission may be detected, etc.
In order to have effective programming, it is necessary to know the
thermal parameters of the experimental device. This presupposes establish-
ing a minimally representative model of the physical characteristics of the
system. With this object, the model based on the transport of heat by
conduction used in the representation of the invariant systems (models at
424 K Torra and H. Tachoire/ Thermochim. Acta 203 (1992) 419 -444

localized constants) may be used. This type of representation provides a set

of N linear differential equations (i = 1, 2, . . . , N) of the form
~(t)=Ci d~/dt+ CP,k(~-T,)+Pi*(11:-T*)+Pio(~-_T,) (1)
k+i

In general, the coefficients are functions of temperature; T * is the temper-

ature of the external thermostat, To is the temperature of the working
stage, and in the q(t) model all the Peltier and/or Joule dissipations
should be included.
This formulation makes it possible to apply the algorithm AIRRT to
Joule or Peltier tests carried out at different constant temperatures [24,25].
From them, for each temperature, an approximation to the physical charac-
teristics of the experimental system can be obtained. In order to make
simulations that represent the effect of the temperature programming, it is
necessary to use a numerical approximation of the system of differential
equations (see, for instance, the Runge-Kutta algorithm). In fact, there are
different parameters that evolve with the value T,,.
In the region of low intensities, small temperature variations are ob-
tained and it may be considered that the intensity is proportional to energy
dissipation. In these conditions it may be assumed that the system is
governed by invariant system equations (models at localized constants).
The ratio existing between current intensity and temperature, determined
in the Pt-100 resistance from reference values (Iref, Rref) and derived from
eqn. (l), takes the form
dAI d2AI dN-‘AR
AZ+a-i-b-+...=+ +B dtN-’ -t “’ +AR (2)
dt dt2 P I
In the above expression, A I = I - Iref, AR = R - Rref, S, is the sensitivity
of the Peltier effect and the series of coefficients a, b, . . . , and A, B, . . .
are related to the series of poles (- l/~~) and zeros (-- l/~~*> of the
Transference Function TF(p) which, in the Laplace space, takes the form
[26-291
M

1 gCTj*pfl)
TF(p) = s I>’ (3)
’ zG(Tia + ‘>

The experimental study of the system response, for a set of values of T *

and T,, establishes that S, depends in a relevant manner on T* and on
the difference between T * and To. However, the evolution of the poles
and the zeros is relatively smooth with the temperatures. In this way, for
programmings of some tens of degrees, it is possible to consider them
constant and equal to their mean values. This permits a semi-quantitative
K Torra and H. Tachoire / Thermochim. Acta 203 (I 992) 419 - 444 425

SP

50
c * *
25
..
:’
t I

, 1
-550 I_,* +50
TO

Fig. 2. Sensitivity S, (K A-‘) of the Peltier effect versus the temperature T, (“0. The
thermal bath temperature T * = -3O.O”C.

approximation from the same differential equation relating the intensities

Al(t) to the resistances AR(t). The series of coefficients a, b, . . . , and A,
B , . . . are kept constant and a variable sensitivity S, = S,(T *, To - T *) is
chosen from a series of calibrations of the Peltier effect (Fig. 2). In order to
make calculation easier, these approximate to a rational fraction of the
form P(x)/&(x). If the degree of the numerator is M and that of the
denominator is N, Pade’s approximant is represented by {M, N]. For
temperatures T * near to room temperature, the overall form of function
S, is satisfactorily represented by means of a Pade’s approximant of the
type (3,1].
In order to ensure coincidence between the values predicted and those
obtained during programming, a local correction is carried out from the
same model used for performing the first open-loop calculation. The
correction uses the same differential equation - from the difference
existing between the value predicted and that obtained - and makes a
forecast of the actions to be carried out so that, some seconds later, the
difference may be cancelled. If the programmings (temperature rate) are
made with evolutions of some rnfl s-l, the differences between the
programming predicted and that obtained do not exceed IfiO.01 K. The
noise associated with the measurements and dynamic characteristics of the
system limit the temperature rates which are accessible for a certain
programming. These values can be analysed easily if the model (via the
transference function TF) is represented in module and frequency coordi-
nates using the Z transform and its subsequent conversion to frequency
space (see for instance refs. 30 and 31).
From two thermal analysis stages (29,32-341 a system can be constructed
which makes it possible to study the stress-strain-temperature diagram.
The system (computer controlled) has the same high resolution in tempera-
ture, close to fO.O1 K, in the region of room temperature (+ 1 rnCI is
equivalent to +2.5 mK>. In measuring strain, there is a resolution of +O.l
pm (maximum shift = + 1 mm) and the force is determined with a resolu-
tion of + 1 mN for total loads of less than 20 N [35].
426 I/: Torra and H. Tachoire / Tktmochim. Acta 203 (1992~419 - 444

TABLE 1
Working temperature T, (in fit, from the R-100) and calorimetric sensitivity 5, (mV W-l)
determined from Joule dissipations in steady state. The external thermal bath temperature
T” is -30°C

T0 S, Til &
79.08 558.0 88.35 628.6
79.18 558.0 88.61 628.7
79.40 560.1 124.25 748.1
79.93 563.9 124.79 757.4
86.63 617.3 133.48 747.3
88.33 628.4

When semiconductor plates are placed in the working space, there is a

heat flux or conduction calorimetric system. In this case, the same base
configuration is used [28,29,36-381 that has produced remarkable results
for transformation temperatures, entropy production and martensitic trans-
formation thermokinetics, and also (see below) yields results in terms of the
energies released. In these “classic” systems, the calorimetric detectors
were situated inside small copper blocks which, with a slight thermal
protection, were spontaneous cooled or heated directly by means of liquid
nitrogen vapour or by means of a heating resistance.
In order to use the calorimetry described in Fig. 1, a suitable calibration
of the Seebeck effect should be made by means of the Joule effect (see
Table 1). The system is automated 132-341. In order to calibrate the
calorimetric sensitivity S,, the computer generates voltage impulses which
produce the Joule effect when the response to Peltier’s effect (cooling or
heating) has reached a stationary state. In order to control the efficacy of
the calorimetric system during a given programming, Joule signals can be

c c
8
B A

A
- 10000s - - 10000s -
Fig. 3. Experimental calorimetric curves versus time with an automatic cycling between
-40°C and +41.7”C: A, calorimetric thermogram; B, Peltier intensity current; C, tempera-
ture; in arbitrary units. 1, Joufe effects; 2, martensiti~ transformation; the arrows indicate a
slight baseline deformation.
l? Terra and H. Tachuire / i’kermochim. Acta 203 (1992) 419 - 444 427

WORK1
NG ROUT1NE
Experimental Processing

step 1
si

step 2
/-

Joule SC - &CAR>
i time canstants

step 3

step 4
r
L
Preliminary

i----
studies: system control by Joule effects
I

L
I

step 5 Measurements

step 6 -1

Fig. 4. Flow diagram of the calorimetric routines; in high-resolution thermal analysis the
unnecesary steps, i.e. the Joule effects studies in steps I, 2 and 4, can be avoided.

generated in order to check the reliability of the equipment and of the

treatment routines (Fig. 3).
Baseline analysis and the analysis of its connection with the Peltier
intensity make it possible to eliminate its effect once a Pade’s approximant
is obtained which links both evolutions (baseline and Peltier intensity). The
thermogram obtained or, where appropriate, an auxiliary signal process in
order to compensate the inertia of the detectors, provides the thermody-
namic magnitudes of interest (see Fig. 4): in particular, characteristic
transformation temperatures, amounts of heat exchanged, an approach to
the entropy change by ]SQ/T, entropy production, hysteresis cycles (Fig.
51, etc.

succEssEs INHIGH-RESOLUTION THERMAL ANALYSIS (HRTA)

Using an HRTA set-up equipped with a system of observation under an

Olympus metallographic microscope (type BH2-UMA) and a Sony low-
priced recording system (video camera EVC-X10, video cassette recorder
SLV353 and TV Trinitron), together with possibilities of acoustic detection
(Bruel and Kjaer amplification system), of measuring electrical resistance
and of manually exerting a low mechanical stress, some experimental
observations have been made.
428 K Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444

0 +25 T/OC
Fig. 5. Hysteretic behaviour of transformation-retransformation processes obtained from
accumulated enthalpy (thermogram in Fig. 3B); X (% P-phase) versus temperature.

(1) The connection between acoustic emission and shifts or sudden

appearances (burst movements) of the interfaces [39].
When the high sensitivity of acoustic emission (AE) was coupled with
HRTA, the observations connected the accelerations in the interfaces
movement with the appearance of AE. Therefore, the diverging interpreta-
tions between the resistance measured and the AE observed [40] and the
existence of invisible nuclei [41] are questionable.
(2) A complete reproducibility of the behaviour for successive thermal
cycles which produces equal acoustic pulses for the same temperatures
[22,42,43].
The problems of stochasticity associated with transformation are proba-
bly associated with the evolution of the material (atomic order versus time
and the dislocations created by burst transformations) and with the uncer-
tainties concerning the external control parameters, e.g. the temperature
and/or stress control and temperature rate and/or stress rate.
(3) An evolution of the MS with time from the initial quenching by an
evolution of the different atomic order parameters (B2 and/or L2, in
Cu-Zn-Al 18R transformation). This depends on the type of initial heat
treatment, on the coexistence of both phases and on the processes of
martensite stabilization and recovery of the beta phase [44-471.
The analysis of the behaviour of an interphase p to martensite with
temperature and/or time makes it possible to examine in detail, where
appropriate, the separate different perturbing effects. This helps to distin-
guish the various microscopic mechanisms involved and, where appropri-
ate, to correlate them. In a simplistic way, they might be associated only
with “stochastic perturbations in the material”. Figure 6 (from ref. 48)
V. Torra and H. Tachoire / Thermochim. Acta 203 (I 992) 419 - 444 429

I , 1
19.80 20.00 20120 T(“(

Fig. 6. Partial hysteretic behaviour in a single, stress-free, martensite plate (air-quenched

sample). Plate width versus temperature; the position of the “internal” cycle (see circled
zone) is connected with the changes in transformation temperatures with elapsed time (from
ref. 48). The slope (dx/dT, see arrows) shows the action of the intrinsic thermoelasticity;
right arrow: increasing temperature is necessary to induce the retransformation in a
“gentle” stabilized martensite.

shows a partial transformation-retransformation cycle which is partially

outside a larger cycle. The shift occurs as a result of opposing influences on
local temperatures of transformation with the permanence in martensite
(stabilization) or in the beta phase (recovery). The m-stabilizati.on [49-521
produces an increase in transformation temperatures and the P-recovery
produces a “recovery” to the standard values.
(4) The existence of an intrinsic thermoelasticity (see Figs. 6-8) associ-
ated with the concentration of dislocations of the material due to the
progressive creation of stacking faults with the movement of the inter-
phases in the martensite growth [53].
The intrinsic thermoelasticity makes a progressive cooling (or an in-
crease in mechanical stress) necessary in order to increase the amount of
martensite. Thus, it stimulates the spontaneous nucleation of other marten-
site plates. Industrial materials contain a high concentration of dislocations
during to the quenching necessary to freeze the parent phase. Therefore,
and in the absence of evolution (atomic order or dislocation creation), the
intrinsic thermoelasticity together with nucleation of the new martensite
plates control the forms of partial or overall transformation-retransforma-
tion hysteresis cycles.
(5) The spontaneous nucleation of the second beta phase [54] in particu-
larly remarkable crystallographic situations suggests a somewhat clearer
430 K Torra and H. Tachoire / Themnochim. Acta 203 (1992) 419 - 444

I J---x 283 TM

-I
200 LOO X&m I
_I

Fig. 7. Hysteretic behaviour of th: single martensite plate. Sample: Cu-Z&Al with y
precipitates (diameter approx, 100 A). (A) Temperature-induced transformation (cycles 3,4,
6 and 9); values af the nth slope dx/dT in cycle 9 are consistent with thermoelasticity in
air-quenched samples. (B) Stress-induced transformation without nucleation processes
(growing/shr~n~in~ only); oabdo, previously cycled zoned, maximum force (&,> at point b;
oacefo, increased zone (cyctes 1,2,3,5,9), maximum force CF,) at point c; the slopes (a and
f zones) in cycie 9 are consistent with the intrinsic therm~~ast~ci~~

interpretation of the existence of the memory effect. When a single

martensite domain reaches one end of a sample, it may nucleate the second
beta phase producing the breakdown of the shape-memory effect GME).

B F/N

200 400 x/pm 100 300 x/j.lm

Fig. 8. Force versus increased length in high-resolution stress-strain observations. (f-1
Air-quenched sample: A, single interface friction phenomena without nucleation; the
friction work in the hatched zone is 170 &J; the slope df/dx relates to the intrinsic
thermoelasticity; B, single interface nucleation processes. (R) Water~uen~hed sample:
repetitive cychng with N martensite pfates (10 < N < .%I).
V. Torra and II. Tachoire / Thermochim. Acta 203 (1992) 419 - 444 431

Second beta-phase nucleation only appears in perfect single crystals of

martensite.
This possibility indicates that reversion to the same “normal” form (the
memory effect) is the normal retransformation process. In a transformation
with several variants, these are only crystallographically coherent with the
initial phase. When a “standard single crystal” of martensite is produced,
the defects existing in the areas of coalescence (the intrinsic thermoelastic-
ity which is coupled with the concentration of dislocations assists in the
nucleation of martensite in different domains) bring about the reappear-
ance of the initial beta phase and, as a result, of the memory effect. Only a
low dislocation concentration in the sample can bring about the breakdown
of SME.
(6) The remarkable evolutions of the first transformation cycles (Cu-
Zn-Al) when an analysis is made of the behaviour of a martensite domain
under the action of a cyclic programming of th% temperature when it
contains small type-y precipitates (diameter = 100 A).
Thermomicroscopy establishes the existence of a very considerable re-
duction in the width of the hysteresis cycle with the number of cycles. After
the first cycle it is reduced by around 30% and after 10 cycles by around
70%. The change is associated with the increase in the M, temperature
and a slight decrease in A,. Similar results are obtained with stress-in-
duced transformation (see in Figs. 7(A) and 7(B) the temperature and
stress changes with cycling); they presuppose the appearance of a privi-
leged direction over the precipitates and, also, an increased thermodynamic
stability for the specific martensite variant. These modifications cannot be
seen by means of standard calorimetric techniques. In spontaneous trans-
formation, the nucleation of new martensite plates renders the growing/
shrinking effects irrelevant.

SOME FAILURES IN SHAPE-MEMORY ALLOYS: FRICTION PROCESSES, EN-

TROPY PRODUCTION AND THERMODYNAMICS

The hysteresis cycle may be obtained from the energies that provide the
calorimetric thermogram, e.g. the accumulated enthalpy versus tempera-
ture (Fig. 5). Then, a model of the behaviour of the material is required in
order to obtain some of its intrinsic properties from the thermal magni-
tudes, e.g. the friction energy or the elastic contributions associated with
transformation. The information generally provided by the literature from
calorimetric measurements appear not to distinguish sufficiently mechani-
cal actions (stress-induced transformations) from thermal actions (tempera-
ture-induced transformations). In this latter case, transformation occurs
spontaneously under the effect of temperature, and free of external stresses
( LX, = 0). In this sense the mechanical friction energy #zX, dl;k is zero. In
432 K Terra and H. Tachoire/ Ptmnochim. Acta 203 (I 992) 419 - 444

TABLE 2
Heat released (A), heat absorbed from calorimetric measurements (B) and friction work
CC) in J mof-’ M,, MI, A, and A, transformation temperatures (K); * indicates results
opposed to the second law of thermodynamics; qth (%o) calculated from experimental data
(if available)

A B C MS Mf A, A, 77th Ref. Observations

100 25 2 246 244 244 252 * 2.4 21
210 130 6 246 238 238 252 * 2.5 21 Data from Figs. 1 and 2
340 340 13 246 225 232 252 P 2.5 21
310 310 6 260 250 255 265 1.9 61 Quenching at 700 K
420 420 7 262 240 245 266 1.9 62
320 320 12 250 225 235 25.5 * 3.2 62 Samples with y
150 150 97 225 - - 285 * 21 62 precipitates

stress-induced transformation, the friction energy is the net heat released

in a complete cycle.
The hypotheses applied in the literature include.
(a) The energy dissipation obtained by calorimetry in transformation or
retransformation may only be identified with the change in chemical
enthalpy if there is no storage of elastic energy. That is, it is only strictly
related when the transformations are made by means of a free interphase
and without external mechanical stresses 1551.
(b) The total thermal energy measured in a complete calorimetric cycle
(assuming that there are no relevant changes in specific heat between the
two phases) represents the dissipation processes or the friction work
[21,56-601.
(c) Using an entropic balance in the hypothesis of zero entropy produc-
tion requires compensatory work [57]. This condition appears in order to
translate the width of the hysteresis cycle in terms of the amount of total
heat exchanged EQ,, i.e. as if the material is the fluid operating in a
reversible thermal machine (the more elementary machine is the Carnot
machine). The value of CQi, measured from the the~ograms of Cu-Zn-
Al, is frequently positive (heat absorbed). It has been assumed that this
energy leaves the material in the form of an acoustic emission [21,57] (see
Table 2).
In some cases, the arguments above were applied to sets of experimental
results. The situation becomes remarkably complicated when operating
with materials previously submitted to many cycles of education and/or
fatigue [63-671. Also, the application of this kind of model could lead to
notable difficulties from the thermodynamic point of view. In fact, when
partial cycles are made, increased energy differences appear (even higher
than 50%) for hysteresis cycle widths of around 5 R and working tempera-
tures around 245 K. These differences cannot be associated with work
V. Torru and H. Tachoire / Thermochim. Acfa 203 (1992) 419 - 444 433

TABLE 3
Calorimetric measurements (J) in Cu-&r-AI j3 J 2H (from ref. 68); e.c., electronic concen-
tration; composition in at.%; A, parent phase to martensite; B, martensite to parent phase;
C, mean value; D, standard deviation; N number of measurements

Sample Jl Sample 11 Sample K2

e.c. Zn Al e.c. Zn Al e.c. Zn Al
1.534 8.39 22.52 1.534 13.12 20.13 1.534 6.41 23.51
A I3 A B A B
C - 1.288 1.262 - 1.60 1.68 - 1.67 1.68
D + 0.008 4 0.004 f 0.02 i: 0.02 &-0.03 f 0.03
N 8 6 4 5 4 6

which the material provides on the outside. The performance of a thermal

machine operating in this region is 2%. Obviously, the thermodynamic
formulation used in ref. 21 and related references is wrong. The * in Table
2 indicates erroneous values.
Also, the calorimetric measurements made for electronic concentrations
above 1.40 (/3 -+ y’ transformations) suggest that the basic ideas deduced
from the measurements on p + j3’ are insufficient. The hypothesis of “null
entropy production” is, obviously, incorrect. Iiysteretic behaviour is strictly
connected to the entropy production: in stress-induced transfo~ation
from work lost or, in temperature-induced transformation, from degraded
heat. The acoustic emissions in this transfo~ation (/3 + y’) are very
intense (they are burst-like transformations). By contrast the energy aver-
ages are almost equal and within the margin of experimental uncertainty
(see in Table 3 some values obtained from ref. 68). This indicates that the
interpretation, based on experimental information, is also related to an
erroneous evaluation of the thermograms used to describe the thermoki-
netics of the transformation. This becomes an incorrect choice of base line
for the energy evaluation or undefined states of the samples [25].

When an analysis is made of the behaviour of the hysteresis cycle of a

material, a series of calorimetric measurements are made associated with a
set of complete, repetitive heat cycles (to avoid more difficult baseline
problems). The sample passes successively from one phase to the other
~+m+/3-+m+~+m...
and we assume that, for each each complete cycle, the same states of the
material (initial state I and final state Fl are recovered
I+F-+I-+F-+I+F...
434 K Torra and H. Tachoire/ Thermochim. Acta 203 (1992) 419 -444

In order to obtain a model of the behaviour, it has been assumed that the
material does not evolve when it has made around 20 cycles. Reproducibil-
ity is confirmed when a series of repetitive cycles of elemental transforma-
tions SISMP (single interface-single martensite plate) are observed by
means of thermomicroscopy and thermosonimetry [22,32,39] but, in sponta-
neous temperature-induced transformation, there is a marked evolution in
the calorimetric thermograms with cycling and a progressive increase in the
concentration of dislocations which has been determined at least up to
cycle 200 [731.
If we assume that the material does not evolve with cycling, the state
functions should recover their values. For internal energy or enthalpy, we
can write
#dU=O $dH=O (4)
The experimental situation of the calorimetric observations suggests that in
a complete cycle no external mechanical work is done, because the material
is not connected with any external mechanical system, In the absence of
sensor systems and external stresses, we should consider that the total work
exchanged is zero. That is
$SW=O (5)
Application of the first principle of thermodynamics tells us that
@Q=O (6)
If the heat capacities are equal (cp = cm), eqn. (6) reads CQi = 0 or
IQp-t,,,l = lQr,vp I. In any event, there are sets of results measured by
different laboratories using similar devices which provide slightly different
values between the processes of transformation and retransformation. If
there is a systematic divergence between experimental reality (CQi f 0)
and the energy formulation (CQi = 0) it should be completely clarified.
Thus, it is possible to consider two alternatives. In the first, an evaluation
has to be made of the uncertainties of behaviour of the experimental
devices when they are used at variable temperatures. Divergences could
arise from the evolution of calibration in dynamic situations (dT/dt # 0) or
from the uncertainties linked to the baseline. The second alternative is
associated with the lack of knowledge concerning the set of energy pro-
cesses undergone in the sample. Other alternatives could be considered:
reformulation of the thermodynamics applied to transformations, i.e. the
first and second laws of thermodynamics, or an assumption that there is an
internal and progressive accumulation of energy within the material during
hundreds or thousands of cycles.

Thermodynamic and phenomenological analysis

In order to clarify transformation phenomenology, it seems appropriate

to make direct measurements of the friction work and, in any event, to
I/. Torra and H. Tachoire / Thermochim. Acta 203 (I 992) 419 - 444 435

observe the most elementary phenomena possible and compare them with
the overall transformations. To do this, we made partial transformations
involving the appearance and disappearance and/or the growing/shrinking
of a single martensite domain. From a high resolution stress-strain system
the result is that the work per cycle that must be supplied is 0.005 J g-‘;
see zone A in Fig. 8 (left>, the stress-strain representation for air-quenched
samples. When nucleation has to be included (see zone B in Fig. S), the
friction work is around 0.03 J g- ‘. When water-quenched samples are used,
the results are similar but usually N martensite plates (10 < N < 50) are
studied with included nucleation (see Fig. 8, right). This tells us that, in a
monovariant transformation, the friction energy is less than 1% of the
thermal energy released during transformation (near 6 J g-l>. That is, even
if CQi represented friction energy (erroneous hypothesis), it would remain
below the calorimetric resolution available (near 1%) when we measure the
global energy obtained in the transformation.
Detailed analysis of the thermograms associated with the processes of
transformation-retransformation of a single plate of martensite indicates
that the classical calculation of the baseline is not sufficient. There appear
to be metastabilities that affect the “whole” sample and that deform the
baseline. A similar behaviour appears for the overall spontaneous trans-
formations. The thermograms associated with overall programmings show
slight deformation (see the arrow in Fig. 3) for the range of temperatures
below A,. It is clear that calculation of the baseline is decisive for
measuring energy or j 6Q/T, in particular, in zones with very weak and
long-lasting dissipations. Careful calculations of the energy contributions,
together with structural techniques, are necessary to understand the com-
plexity of the phenomenology associated with transformation. In any event,
it is important to note that conduction calorimetry is as well adapted to
measure the dissipations made at “constant temperature”, as those associ-
ated with first-order transitions, with mixing processes, and with chemical
reactions. By contrast, modifications of heat capacity only, introduce base-
line evolutions [25,69-721. Because their internal phenomenology is depen-
dent on the thermoelasticity effects, martensite transformations are made
under the action of the changes in temperature and, therefore, can be
interpreted as “abnormalities” in the heat capacity.
In general, the material evolves with time and cycling. After an air
quenching, the initial concentration of defects is very small and nucleation
is difficult. This produces burst-like transformations with a progressive
increase in the concentration of dislocations in relation to the number of
cycles. Transformation occurs more gently and extends into a wide range of
temperatures (temperature extension increased). The existence, on the
thermogram, of burst-like transformations and their progressive evolution
makes it possible to introduce an entropy production associated with these
pulses. In this way, the total entropy production (AS,) can be separated
436 K Torra and H. Tachoire/ Thermochim. Acta 203 (1992) 419 - 444

into two parts: firstly, AS;, associated with discontinuous contributions (b

or burst) with an evolutionary character [29,34,48]; and secondly, A&_,,
associated with the width of the hysteresis cycle (s or standard) and little
affected by the evolution of the sample.
A treatment by signal processing of the thermogram associated with the
jth cycle makes it possible to display the pulses more easily and to calculate
AS:(j) from
T=Mf

AS;(j) = CASE(j) = c m,c[log(T/(T+ AT)) + AT/T] (7)

i T=M,

The above expression relates to the increase in universal entropy associated

with a series of elemental (burst) transformations between MS and Mf. For
this, it is assumed that a material domain of mass m, undergoes a burst
transformation, The energy released in a pseudo-adiabatic form, changes
the temperature of the material locally (in m,) into AT (for instance, 4 K
which is equivalent to half a hysteresis width). The energy is transferred
progressively to the rest of the sample and to the temperature control
system (external thermal bath), which are at temperature T. The values of
AS$ j) (the jth cycle) obtained from an air quenching, decrease with the
number of cycles to a constant value AS:(m) (the mth cycle) in the “steady
state” (see Fig. 9). The change in AS: may be associated with the changes
inside the material via a “driving force”

TAS = T[ AS;(l) - AS;(w)] (8)

The introduction of defects progressively helps nucleation. In the case of
Cu-Zn-Al (electronic concentration e.c. = 1.481, the application of expres-
sion (8) to the measurements presented in Fig. 9 produces around 11 p J
g -’ K-l when T = 260 K and AT = 4 K. This is equivalent to around 24
mJ cmP3 and the energy necessary to create the corresponding dislocations

us, l 1
.
.
.
. (1

.
.
0 . l .‘.
1
30 60

Fig. 9. Entropy production ASL (in arbitrary units) versus cycle number. Sample after air
quenching; the arrow (a) indicates the change of the entropy production with spontaneous
cycling.
K Towa and Ii. Tachoire / Thermochim. Acta 203 (1992) 419 - 444 437

TABLE 4
Temperatures and heat quantities involved in calorimetric cycling processes: A, without
thermoelasticity; B, with thermoeiasticity
Temperature Heat absorbed Heat absorbed
by the sample by the bath
A
p to martensite M, (= Mf) Q rR-brn ... QM (=-Q,s-,m)
Martensite to p A,(=A,) Q m-+B QA(=-Qm-P)

B
p to martensite Iki, to Mf Qp-+m QM(=-Qp-tm)
Martensite to f3 A, to A* Q m+8 QA(=-Q m-a )

after 200 cycles is close to 80 mJ cmm3 1’731but the temperature rates used
in cycling are greater. The results show us that the irreversibilities (driving
forces AG) associated with burst transformations are close to the energies
for creation of dislocations. Furthermore, the description of the heat
“burst” (lasting several seconds) is much slower and “gentler” than the
mechanical process. Using a “full adiabatic” condition, AT = 14 K and the
driving forces approach 2 .I cm -3. This suggests that the entropy production
changes are only connected qualitatively to the permanent structural defor-
mation.

~~s~ere~ic~ehav~ur and the second law: the Camot cycle approach

In order to obtain an entropy balance and a calculation of the equivalent

“friction work” associated with spontaneous (external stress free) calori-
metrically induced transformations, it is necessary to model the process
undergone by the material (see Table 4 and Figs. 10 and 11). In an initial
approach, it may be considered that T,, the theoretical equilibrium temper-
ature of transformation, is Ta = (M, +A,)/2 and that the heat capacity of
martensite and the heat capacity of the P-phase are equal (cp = c,).
~e~hanicai stress-induced transfo~ations with one or more singie-variant
martensite domains are made at constant temperature (equivalent, in
temperature-induced transformation, to the process abed in Fig. 10 or the
reversible transformation bc in Fig. 11). Spontaneous temperature-induced
transformations, with a single variant, are related to the process abc’d’a in
Fig. 10 or, in a reversible situation, to the process bd in Fig. 11. In order to
pass from point c to point d in Fig. 11, a transformation over the
martensite-parent phase coexistence line can be made.
Calorimetric observations of spontaneous, stress-free cooling (o,,, = 0)
have revealed complex morphologies produced by the different variants.
This results in their representative point d not being able to be related to
point c (Fig, 11). With this type of process, points c and d are at external
438 I/: Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444

X
a beta
100 x
phase

C' i
OX
nnartensite phase
I I I 1 I r*
Mf A. H* To Af T
Fig. 10. Schematic hysteretic behaviour in X (% /3 phase) versus temperature; cycle abcda,
phase transformation with hysteresis; cycle abc’d’a, thermoelastic with hysteretic behaviour;
aefc’, reversible phase transformation (bc in Fig. 11); aef’c’, thermoelastic reversible
transformation (bd in Fig. 11).

stress zero (a,,, = 0), and at different temperatures. In order to construct a

thermodynamic ,cycle it is necessary to be able to pass with continuity from
one point to another. However, at point c we have a single variant, induced
by a very weak stress or by education of the material. At point d we have a
heterogeneous system which, by external stress perturbation (from u = 0,
increasing to c = Sa and decreasing to the initial value o = 01, may
become homogeneous (single variant). This shows us that the energy results
(or the entropy calculated from eqn. (18)) obtained calorimetrically in

/
b

sd % ‘b %
S

Fig. 11. Schematic temperature-entropy diagram in reversible transformations; ab, cooling

/3 phase; bc, phase transformation at constant temperature (/3 -+ m); bd, “thermoelastic”
transformation, cooling and stress-free (equivalent to the idealized spontaneous transforma-
tion in calorimetric processes); cd, cooling changing external stress (coexistence between /3
and martensite phases); de, cooling martensite phase.
K Torra and If: Tachoire / Thermockim.Acta 203 (f992) 419-444 439

spontaneous transformations are nut strictly linked to the change of phase

to a single variant of martensite.
The friction work in a mechanical stress-induced transfo~ation-re-
transformation cycle is obtained directly from $fzX,dl In order to calculate
the fictitious “friction work” associated with temperature-induced transfor-
mation, we shall consider only the cycle abcda in Fig. 10. In our case, eqn.
(6) reads ) QM I= 1QA I. The production of entropy ASb(tb) associated
with the changes in the thermal baths is
AS&(tb) = Q,/‘& + &A/‘% = Qd4 - ~~}/~~~~~) 69
The work, W,, (CC = Carnut cycle), which would produce a Camot
machine between A, and K3, absorbing an amount of heat 1Q, 1, would be
KY= tQAI(A,-~~)~~,=Q,(A,-~~)/ff, We
If this amount of work is released in the form of heat at temperature M,,
the corresponding entropy production AS$CC) is
AS&(CC) = W&./Ms = Q&l, - M,)/(A,M,) = AS;(tb) (11)
The above expression enables us to define IV*, a fictitious “friction work”,
in the temperature-induced transformation from
W&Z2 W,,= MsAS&(CC)= iQ,rtA,-~~>/A,=Q,cA,-M,)/~, (12)
The fictitious friction work is the work obtained in an equivalent Carnot
cycle working from the upper (A,) to the lower (MJ temieratures.
The free energy language (see Fig, 121, the Gibbs free energy (the
driving force) necessary to overcome the friction terms AC,_, and
AC,_+ are defined by
i3AG
A%, =: 7 AT= -(A&-- TJAS,,, (13)
aAG
AG T-A, = FAT= -(A,-T,)AS,_+

I
1
MS To A, T

Fig. 12. Gibbs function (parent with only one martensite plate) versus temperature at
constant stress (schematic). Coexistence between /3 and martensite at ~qu~l~b~urn tempera-
ture T,; aefc, reversible tr~sfn~ation path; aebcfda, hysteretic be&aviour (type abcda in
Fig. 101,
440 V; Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444

with
AS,,,= -AS,,,=%,-Sp=Qp_,m/T-,
The associated entropy production in a complete thermal cycle is
4-w
AS; = - [ AG,=,/M, + AG,+/Af] = QM M A = AS;@) (14)
s f
ASS is, obviously, equal to the ASb(tb) obtained from the heat transferred
to (and from) the heat sinks in the calorimetric cycle. The previous analysis
shows that the relevant quantities that explain the hysteretic behaviour are,
for instance, the entropy production: AS,, AS; and ASb.
When the hysteresis cycle has thermoelasticity (abc’d’a in Fig. lo),
partial cycles should be considered from zones of transformation-retrans-
formation, such as abx;x;a, x~x;x;x~x~, . . . , x~x~c’d’x~. If the width AT,,
of the hysteresis cycle is nearly constant, a straightforward (geometricj
approach gives us
M,AS;(CC) = Wf” = 2QMATh/(As +A,) (15)
If we define “thermal efficiency” qth%, from the associated Carnot cycle,
by
qt,., = 100Wf”/Q, = 2OOAT,/( A, + A,) (16)
immediate information is obtained on the relative importance of the
hysteresis cycle from the temperature involved. The values of 7th are
consistent with those obtained from direct mechanical measurements of the
friction cycle. For this, the “mechanical inefficiency” q,,,% may be defined
as

rim= lW..x,&'Q, (17)

For the equivalent cycles (in temperature and stress-induced transforma-
tions), the qth and q,.,, values are practically equal [68,74,75] and hysteretic
width dependent: 0.1% in single interface p to p’ growing/shrinking, 1%
in INSTRON-type measurements, single variant (in spontaneous calorimet-
ric measurements = 3%) and 10% in p to 2H measurements. Values of
qti, recalculated from experimental data in the literature (if available), are
shown in Table 2.
From analysis of the elemental cycle bed in Fig. 11, it is possible to
calculate thermoelastic energies, and also to “avoid” the irreversibility of
the cycle in order to obtain an approach to the entropy change of the
material in its spontaneous transformation, from the relationship
K Torra and H. Tachoire / Thermochim. Acta 203 (I 992) 419 - 444 441

which, in the case of Cu-Zn-Al (p + j3’ transformation) does not differ

significantly from the direct value of / 6&/T. For T = 300 K and 0.5AT, I
5 K, the AS,,, differs from / S&/T by approx. 2%. Speakly strictly,
/ S&/T is only connected with the entropy changes in the thermal baths.
Obviously, a transformation to a single variant does not have the difficul-
ties mentioned above: its temperature range is much smaller because it is
only affected by intrinsic thermoelasticity, but current reliability of the
results is poor. There is considerable uncertainty in the calculation of the
transformed mass and some metastabilities are observed. Similarly, direct
calorimetric tests on stress-strain systems are still inaccurate.

CONCLUSIONS

The development of the experimental systems of HRTA with automatic,

strictly reproducible programmings has made it possible to increase our
understanding of the processes that occur in memory alloys. For this, the
Peltier effect and a rigorous programming based on the analyses of the
dynamic behaviour of the device have been used. From basic HRTA
equipment, other auxiliary techniques have been adapted to analyse trans-
formation/retransformation processes with high resolution: thermomi-
croscopy, electrical resistance, heat conduction calorimetry, stress-strain
studies, etc.
The observations have shown that HRTA studies in shape-memory
alloys are very sensitive to minor structural changes. This is easily demon-
strated when reproducible behaviour and, where appropriate, evolution of
the material due to changes in the atomic order and/or creation of defects
caused by the transformation itself, are obtained. The relevance of the
processes of nucleation and growth in the hysteresis cycle is clearly estab-
lished. In this respect, the intrinsic thermoelasticity associated with the
concentration of dislocations is of highest importance in ensuring memory
effect.
Thermodynamic analysis establishes the strict coherence between the
friction obtained in entropic terms from the width of the hysteresis cycle,
determined calorimetrically, and the friction energy measured directly in
the stress-strain diagram. Its values are around 1% in the spontaneous
transformation at 18R (p + /?‘> and around 10% in the transformation at
2H (p + ~‘1. Obviously, recent thermodynamic interpretations using ther-
modynamic potentials and dissipative terms, identifying (in thermal cycles
and without external stresses) the friction work with the global heat
measured, are wrong. The relevant quantity in the thermal cycles is the
entropy production related to the hysteresis width. In addition, the calori-
metric thermograms, considered to be a result of the entropy of the “burst
processes”, are consistent with the creation of dislocations.
442 K Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444

ACKNOWLEDGEMENTS

Dr. Amengual &JIB-Spain), Dr. A. Isalgue (ETSAB/UPC-Spain) and

Dr. Lovey (CAB-Argentina) are gratefully acknowledged for their help in
the experimental measurements and for their enlightening discussions.
Contacts between the two countries are supported by DGICYT (H.T.,
sabatic program) and Univ. Provence (V.T. Associate Professor). V.T.
acknowledges the Lagrange Group (L.A. Garcia-Ramos) for aid in the
optimal choice of computer hardware. Experimental research in Spain is
carried out in the framework of the CICYT project MAT89-0407X03.

REFERENCES

1 V. Torra (Ed.), Science and Technology in Shape Memory Alloys, Proc. COMETS
Course 1989, Barcelona, Spain, Universitat de les Illes Balears, Spain, 1989.
2 B.C. Muddle (Ed.), Martensitic Transformations I and II, Proc. ICOMAT 89, Materials
Science Forum, Vol. 56-58, Trans. Tech. Publications, Zurich, 1990.
3 I. Tamma (Ed.), Proc. ICOMAT 86, The Japan Institute of Metals, Sendai, Japan, 1987.
4 E. Hornbogen and N. Jost (Eds.), The Martensitic Transformation in Science and
Technology, DGM, Oberursel, Germany, 1989.
5 Proc. COMETS II Course, Technologie et Mise en Oeuvre des Alliages a Memoire de
Forme, IMAGO, La Ciotat, France, 1991.
6 G.W. Lorimer (Ed.), Proc. Conf. Phase Transformations 87, The Institute of Metals,
Cambridge, UK, 1988.
7 M. Ahlers, Prog. Mater. Sci., 30 (1986) 135-186.
8 A. Caneiro and M. Chandrasekaran, Ser. Metall., 22 (1988) 1797-1800.
9 W. Zongguo and Y. Dazhi, Ser. Metall., 22 (1988) 1245-1249.
10 L. Li, A. Aernoudt and L. Delaey, Ser. Metall., 22 (1988) 1435-1440.
11 J.M. Guilemany and F.J. Gil; Mater. Res. Bull., 25 (1990) 1325-1332.
12 D.Z. Yang, M. Zhu, M. Qi and G.B. Li, Ser. Metall., 20 (1986) 1717-1720.
13 L. Delaey, J. Ortin and J.V. Humbeeck, ref. 6, pp. 60-66.
14 N. Ono and H. Shimanuki, Ser. Metall. Mater., 24 (1990) 2269-2272.
15 M. de Graqef and L. Delaey, ref. 2. pp. 423-428.
16 D.G. Maeder and M. Droz, ref. 2, pp. 119-124.
17 I. Takahashi and T. Suzuki, ref. 2, pp. 125-130.
18 E. Patoor, A. Eberhardt and M. Berveiller, ref. 4, pp. 133-140.
19 R. Stalmans, J.V. Humbeeck and L. Delaey, ref. 4, pp. 207-212.
20 I. Muller and H. Xu, Acta Metall. Mater., 39 (1991) 263-271.
21 A. Planes, J.L. Macqueron and J. Ortin, Philos. Mag. Lett., 57 (1988) 291-298.
22 F.C. Lovey, J. Ortin and V. Torra, Phys. Lett. A, 121 (1987) 353-356.
23 A. Amengual, F. Garcias, F. Marco, C. Segui and V. Torra, Acta Metall., 36(8) (1988)
2329-2334.
24 M. Rodriguez de Rivera, F. Socorro, J.P. Dub&, H. Tachoire and V. Torra, Ther-
mochim. Acta, 150 (1989) 11-25.
25 A. Amengual, F. Marco, M. Rodriguez de Rivera and V. Torra, Thermochim. Acta, 188
(1991) 63-76.
26 H. Tachoire, J.L. Macqueron and V. Torra, in M.A.V. Ribeiro da Silva (Ed.), Thermo-
chemistry and Its Applications to Chemical and Biochemical Systems, NATO AS1 Series
C 119, D. Reidel Publishing Co., Dordrecht, 1984, pp. 77-126.
V: Towa and H. Tachoire / Thermochim. Acta 203 (1992) 419-444 443

27 W. Zielenkiewicz (Ed.), Thermokinetics. Signal processing in Calorimetric Systems,

Ossolineum, Warsaw, 1990.
28 H. Tachoire, J.L. Macqueron and V. Torra, Thermochim. Acta, 105 (1986) 333-367.
29 H. Tachoire and V. Torra, Can. J. Chem., 67 (1989) 983-990.
30 P. Thuillier and J.C. Belloc, Mathematiques - Analyse 3, Masson, Paris, 1989, pp.
181-194.
31 J.R. Rodriguez, C. Rey, V. Perez-Villar, V. Torra, J. Ortin and J. Viilals, Thermochim.
Acta, 63 (1983) 331-339.
32 A. Amengual and V. Torra, J. Phys. E: Sci. Instrum., 22 (1989) 433-437.
33 A. Amengual, V. Torra, A. Isalgue, and F. Marco, Thermochim. Acta, 1.55 (1989)
115-134.
34 V. Torra and H. Tachoire, J. Therm. Anal., 36 (1990) 1545-1577.
35 V. Torra, V.R. Torra, A. Isalgue and F. Marco, Thermochim. Acta, 200 (1992) 413-426.
36 C. Picornell, C. Segui, V. Torra, J. Hernaez and C. Lopez de1 Castillo, Thermochim.
Acta, 91 (1985) 311-320.
37 C. Picornell, C. Segui, V. Torra, C. Auguet, Ll. Mafiosa, E. Cesari and R. Rapacioli,
Thermochim. Acta, 106 (1986) 209-217.
38 G. Guenin, J.L. Macqueron, M. Mantel, C. Auguet, E. Cesari, Ll. Maiiosa, A. Planes, C.
Picornell, C. Segui, J. Ortin and V. Torra, ref. 3, pp. 794-799.
39 A. Amengual, Ll. Maiiosa, F. Marco, C. Picornell, C. Segui and V. Torra, Thermochim.
Acta, 116 (1987) 195-208.
40 N. Nakanishi, T. Shigematsu, M. Yasuoka, H. Oshima and S. Miura, ref. 3, pp. 210-215.
41 A. Steiner, R. Gotthard and D.G. Maeder, ref. 3, pp. 216-221.
42 F.C. Lovey, E. Cesari, V. Torra and J.M. Guilemany, Mater. Lett., 5 (1987) 159-162.
43 C. Picornell, C. Segui, V. Torra, F.C. Lovey and R. Rapacioli, Thermochim. Acta, 113
(1987) 171-183.
44 A. Planes, J. Viiials and V. Torra, Philos. Mag. A, 48 (1983) 501-508.
45 J. Viiials, V. Torra, A. Planes and J.L. Macqueron, Philos. Mag. A, 50 (1984) 653-666.
46 A. Planes, R. Romero and M. Ahlers, Acta Metall., 38 (1990) 757-763.
47 A. Amengual, F.C. Lovey and V. Torra, Ser. Metall., 24 (1990) 2241-2246.
48 A. Amengual and V. Torra, Thermochim. Acta, 198 (1992) 267-278.
49 M. Chandrasekaran, L. Cooreman, J.V. Humbccck and L. Delaey, Ser. Metall., 23
(1989) 237-239.
50 X. Duan and W.M. Stobbs, Ser. Metall., 23 (1989) 441-445.
51 T. Suzuki, R. Kojima, Y. Fujii and A. Nagasawa, Acta Metall., 37 (1989) 163-168.
52 C. Cooreman, J.V. Humbeeck and L. Delaey, Acta Metall., 38 (1990) 2663-2666.
53 F.C. Lovey, A. Amengual, V. Torra and M. Ahlers, Philos. Mag. A, 61 (1990) 159-165.
54 F.C. Lovey, A. Amengual and V. Torra, Philos. Mag. A, 64 (1991) 787-796.
55 R.J. Salzbrenner and M. Cohen, Acta Metall., 27 (1979) 739-748.
56 J. Ortin, Thermochim. Acta, 121 (1987) 397-412.
57 J. Ortin and A. Planes, Acta Metall., 36 (1988) 1873-1889.
58 J. Ortin and A. Planes, Acta Metall., 37 (1989) 1433-1441.
59 J. Ortin and A. Planes, ref. 4, pp. 75-82.
60 J. Ortin and A. Planes, ref. 2, pp. 139-144.
61 A. Planes, J.L. Macqueron, R. Rapacioli and G. Guenin, Philos. Mag. A, 61 (1990)
221-231.
62 C. Auguet, E. Cesari, R. Rapacioli and Ll. Maiiosa, Ser. Metall., 23 (1989) 579-583.
63 C. Auguet, E. Cesari and Ll. Maiiosa, J. Phys. D: Appl. Phys., 22 (1989) 1712-1720.
64 C. Picornell, M. Sade and E. Cesari, Ser. Metall., 23 (1989) 371-376.
65 C.M. Friend, J. Ortin, A. Planes, Ll. Manosa and M. Yoshikawa, Ser. Metall. Mater., 24
(1990) 1641-1645.
66 J. Pons, E. Cesari and M. Rota, Mater. Lett., 9 (1990) 542-546.
444 V: Torra and H. Tachoire / Thermochim. Acta 203 (1992) 419 - 444

67 C. Segui, E. Cesari and J.V. Humbeeck, Mater. Trans. Jpn. Inst. Metals, 31 (1990)
375-380.
68 J.L. Pelegrina, Ph.D. Thesis, 1990, Instituto Balseiro, CAB, Bariloche, Argentina.
69 SC. Mraw, Rev. Sci. Instrum., 53 (1982) 228-231.
70 C. Sandu and R.K. Singh, Thermochim. Acta, 159 (1990) 267-298.
71 E. Urbanovici and E. Segal, Thermochim. Acta, 159 (1990) 369-372.
72 J. Ortin, V. Torra and H. Tachoire, Thermochim. Acta, 121 (1987) 333-342.
73 J. Pons, F.C. Lovey and E. Cesari, Acta Metall. Mater., 38 (1990) 2733-2740.
74 M. Ahlers, F.C. Lovey, J.L. Pelegrina, A. Amengual and V. Torra, unpublished results,
1990.
75 A. Amengual, Ph.D. Thesis, 1990, Physics Dept., Univ. Illes Balears, Spain.