Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

July 7, 2009 11:59 00059

Functional Materials Letters


Vol. 2, No. 2 (2009) 73–78
© World Scientific Publishing Company

THE USE OF SHAPE-MEMORY ALLOYS FOR MECHANICAL REFRIGERATION

LLUÍS MAÑOSA∗ , ANTONI PLANES and EDUARD VIVES


Departament d’Estructura i Constituents de la Materia, Facultat de Física,
Diagonal 647, 08028 Barcelona, Catalonia, Spain
ERELL BONNOT†
Departament de Física i Enginyeria Nuclear, ETSEIB, Universitat Politècnica de Catalunya, Diagonal 647,
08028 Barcelona, Catalonia, Spain
RICARDO ROMERO‡
IFIMAT, Universidad del Centro de la Provincia de Buenos Aires and CICPBA, Pinto 399, 7000 Tandil, Argentina
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com

Received 2 February 2009


by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

This letter reports on stress–strain experiments on a Cu–Zn–Al single crystal performed using a purpose-built tensile device which
enables the load applied to the specimen to be controlled while elongation is continuously monitored. From the measured isothermal
tensile curves, the stress-induced entropy changes are obtained at different temperatures. These data quantify the elastocaloric effect
associated with the martensitic transition in shape-memory alloys. The large temperature changes estimated for this effect, suggest
the possibility of using shape-memory alloys as mechanical refrigerators.

Keywords: Elastocaloric effect; martensitic transition; shape-memory alloy; magnetocaloric effect.

Shape-memory alloys are an important class of functional exhibiting other effects such as the electrocaloric effect5 and
materials which are being used for sensor and actuator appli- the barocaloric effect6 also appear as good candidates for clean
cations. Their functionality is linked to the large reversible refrigeration.
deformations (up to 10%) they can experience under appro- Caloric effects are inherent to any thermodynamic sys-
priate changes of temperature and stress which arise from the tem. They are expected to be particularly large in the vicinity of
martensitic transition undergone by these alloys. In the present first-order phase transitions where tiny variations of external
letter we report on a different functional application, namely control parameters lead to significant changes in the extensive
the possibility of these alloys being used as mechanical refrig- thermodynamic quantities. For instance, in magnetic mate-
erators. This property is based on the caloric response under rials, the giant magnetocaloric effect is a consequence of a
the application of an external stress (the elastocaloric effect), first-order magnetostructural transition, which encompasses a
which is expected to be large in the vicinity of the martensitic large entropy change, related to the latent heat of the tran-
transition.1 sition.7 The martensitic transition in shape-memory alloys
The need for environmental-friendly cooling methods meets these requirements and therefore large elastocaloric
has prompted an intense search for materials exhibiting effects take place in these materials. The similarity between
large caloric effects at and around room temperature, which the elastocaloric effect in shape-memory alloys and the widely
could be an alternative to the classic compressor-based reported giant magnetocaloric effect in materials undergoing
technologies involving gases that are harmful to the envi- magnetostructural transitions will be presented.
ronment. Among several caloric effects, the magnetocaloric In a thermodynamic system described by generalized
effect has received increasing interest,2–4 although materials forces {Yi } and displacements {xi }, a differential change in
entropy is expressed as
∗lluis@ecm.ub.es
n  
C ∂xi
†erell.bonnot@upc.edu dS = dT + · dYi , (1)
‡rromero@exa.unicen.edu.ar T ∂ T {Y j =1,...,n }
i=1

73
July 7, 2009 11:59 00059

74 L. Mañosa et al.

where generalized Maxwell relations temperature change, induced by a change of the field from
    0 to Y :
∂S ∂xi  Y
= (2) δq
∂Yi T ,Y j =i ∂T Y j =1,...,n S(0 → Y ) ≥ (6)
0 T

have been used and C is the heat capacity at a constant value and
of the fields. Each pair of conjugated variables have the same
T irr (0 → Y ) ≥ T (0 → Y ). (7)
tensorial order.
A caloric effect associated with a differential change of For a first-order transition with a given associated hys-
a field Yi is quantified by teresis, the dissipated energy can be estimated as E diss =
   (Y − Y0 )x, where Y0 is the equilibrium transition field at
Yi ∂xi temperature T , and Y is the actual transition field at this tem-
S(0 → Yi ) = · dYi , (3)
0 ∂T {Y j =1,...,n } perature. x is the discontinuity in the generalized displace-
ment. Consideration of dissipative effects leads to a modified
when the field is applied isothermally, and Clausius–Clapeyron equation which (under the assumption of
   constant x) reads
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com

Yi T ∂xi
T (0 → Yi ) = − · dYi , (4) dY S 1 d E diss
∂T =− + .
by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

0 C {Y j =1,...,n } (8)
dT x x d T
when the field is applied adiabatically. In general, E diss is weakly dependent on temperature,
In a typical experiment, the control parameter is one of and thus the last term in the previous equation is small,
the components of the force tensor, say Y , while the mea- which shows that the Clausius–Clapeyron equation is still
sured quantity is the displacement component conjugated to a good approximation. Moreover, in shape-memory alloys,
this field, x. In the particular case of the elastocaloric effect, the the martensitic transformation takes place close to equilib-
field is a uniaxial tensile stress σ for which the corresponding rium conditions, and dissipative effects are small.9 Therefore,
generalized displacement is the strain ε (or relative elongation Eqs. (3) and (4) are expected to be valid in evaluating caloric
along the direction of the applied force). In the case of the effects associated with this transition.
magnetocaloric effect, the magnetization M (along the field) In the above description, we have considered the gen-
and magnetic field H are the generalized displacement and eralized displacements (and fields) to be independent of one
force, respectively. another. Coupling between different degrees of freedom is
It is worth noticing that in the vicinity of a phase tran- reflected by an explicit dependence of one coordinate on the
sition, ∂ x/∂ T is expected to be large, since x experiences remaining ones. In this case, the expressions given in this
large changes arising from discontinuities in the displace- section are still valid, but it must be taken into account that
ments at first-order phase transitions. This will lead to giant the entropy changes obtained with the above equations will
caloric effects (see Eqs. (3) and (4)). Actually, due to such contain contributions from all degrees of freedom, which are
a discontinuity, the use of Maxwell relations in computing modified when the field is changed from 0 to Y . This is partic-
the entropy change at first-order phase transitions gave rise ularly relevant in the case of giant magnetocaloric materials
to some controversy. It has been shown,4,8 however, that the for which there is strong coupling between structure and mag-
above expressions are in general valid, and they lead to the netism and the large magnetic field-induced entropy change
Clausius–Clapeyron equation in the particular case of an ideal has a significant contribution from the entropy change associ-
first-order phase transition for which Eq. (3) renders the dis- ated with the structural change.10–12
continuity in the entropy at the transition, associated with the The elastocaloric effect for a solid subjected to an applied
latent heat. uniaxial stress σ with an associated uniaxial strain ε is
So far, the thermodynamic analysis has assumed equilib- given by
 σ 
rium processes. When considering the effect of irreversibility, ∂ε
the entropy contains a reversible (d S) and irreversible (δSi ≥ S(0 → σ ) = dσ (9)
0 ∂T σ
0) contributions, so that the heat exchanged is expressed as
and
δq  σ  
= d S − δSi , T ∂ε
(5) T (0 → σ ) = − dσ. (10)
T 0 C ∂T σ

which leads to the following inequalities which are to be sat- The martensitic transition in shape-memory alloys is first
isfied by the isothermal entropy change and the adiabatic order, and is characterized by discontinuities in the strain ε
July 7, 2009 11:59 00059

The Use of Shape-Memory Alloys for Mechanical Refrigeration 75

and entropy St (associated with the latent heat of the tran- continuously monitored. The system uses special grips which
sition). We will restrict our analysis to non-magnetic shape- adapt to the heads of the specimen. The upper grip is attached
memory alloys, for which the entropy change has a vibrational to a load cell hanging from the ceiling and the lower grip
origin,13,14 and the strain is not coupled to any other degree holds a container that plays the role of a dead load. The load
of freedom. In the vicinity of the transition, the strain can be is increased or decreased at a well-controlled rate by supply-
expressed as ing or removing water by means of a pump. A cryofurnace
enables the experiments to be conducted at different temper-
ε(T, σ ) = ε0 + εF [(Tt (σ ) − T )/T ], (11)
atures. Details of the device can be found in Ref. 15.
where ε0 is the strain outside the transition region, F is a Typical stress–strain hysteresis loops at three selected
shape-function and T is a measure of the temperature range temperatures are illustrated in Fig. 1. On increasing the
over which the transition spreads. Tt (σ ) is the equilibrium stress, strain linearly increases (elastic behavior of the high-
transition temperature for a given applied stress σ . In strict temperature phase) until a certain stress where the martensitic
equilibrium, T → 0 so that F approaches the Heaviside transition starts, giving rise to a large increase of strain (about
step function. 8%) at an almost constant stress. The elastic behavior of the
Using expression (9) and assuming that ε0 and ε are martensite is obtained when the stress is further increased.
constant, in this equilibrium case the elastocaloric effect in Upon unloading, the strain decreases with a certain hystere-
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com

the vicinity of the transition is given by sis (about 10 MPa) at the martensitic transition. An increase
by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

 ε in the transition stress with increasing temperature is clearly


− for T ∈ [Tt (0), Tt (σ )] observed.
S(0 → σ ) = α , (12)
 In Fig. 2(a), we show the stress–strain curves correspond-
0 for T ∈/ [Tt (0), Tt (σ )] ing to the loading branches measured at different temperatures.
where α ≡ d Tt /dσ is assumed to be constant. Taking into Here the control parameter (σ ) is plotted on the horizontal
account the Clausius–Clapeyron equation, α = −ε/St , axis while the measured parameter (ε) is plotted on the ver-
where St is the transition entropy change, thus, as expected, tical axis. Note that with this representation, the curves are
S(0 → σ ) = St , and T = Tσ −Tt (0) = ασ . Notice that similar to the corresponding isothermal field-magnetization
these results confirm the validity of using Maxwell relations curves used to compute the magnetocaloric effect in magnetic
in computing the elastocaloric effect, even in the case of an materials.
ideal first-order phase transition (involving a discontinuity in From the curves in Fig. 2(a) it is possible to compute the
the entropy). elastocaloric effect by means of Eq. (9). In order to minimize
A single crystal of Cu–Zn–Al, obtained by melting metals numerical errors, we first numerically compute the area A(T )
of 99.999% purity, was grown by the Bridgman method. The below the curve up to a certain stress. Results are shown in
composition of the crystal was determined from EDX to be Fig. 2(b), where the molar value V = 7.52 cm3 mol−1 has
Cu68.13 Zn15.74Al16.13. The martensitic transition temperature been used to achieve the energy units. The entropy change
under cooling without external stress is Ms = 234 K. A rectan- is then obtained from the temperature derivative of this data
gular sample with cylindrical heads was machined from the set as
ingot. The body of the sample has flat faces 35 mm long, 3.95  σ
d dA
mm wide and 1.4 mm thick. The crystallographic direction of S = εdσ = . (13)
dT 0 dT
the tensile axis is close to the [001] direction. The sample was
annealed at 1073 K for 30 min, cooled down to room temper-
ature in air and finally aged for 2 h in boiling water. This heat
treatment ensures that the sample is in the ordered state, free
from internal stresses and the vacancy concentration is close
to the equilibrium state.
Typically, stress–strain experiments are carried out by
means of tensile machines that control the sample displace-
ment while the load is continuously measured. In order to per-
form mechanical experiments that can be directly compared
to the corresponding experiments in other caloric systems
(where the field is the control parameter and the generalized
displacement is the measured quantity), we have developed a Fig. 1. Stress–strain hysteresis loops at different temperatures. From top to
purpose-built experimental set-up which enables external con- bottom: T = 310.5 K (green), T = 305.4 K (red) and T = 296.1 K (blue). (color
trol of the load applied to the sample, while the elongation is online)
July 7, 2009 11:59 00059

76 L. Mañosa et al.

(a)
Fig. 3. Stress-induced entropy change at selected values of the stress ranging
from 105 to 143 MPa. The color code (and symbols) correspond to the stress
value indicated in Fig. 2(a) by a vertical line. The continuous lines are fits
based on the Eq. 14. (color online)
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com
by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

(b)

Fig. 2. (a) Stress–strain curves at selected temperatures. From right to left,


T= 310.5, 309.4, 307.5, 306.0, 305.4, 303.1, 302.0, 299.9, 298.8, 297.9, 296.1,
295.3 and 294.6 K. (b) Area below the stress–strain curves. The color code
(and symbols) correspond to the maximum stress indicated as a vertical line Fig. 4. Transition stress as a function of temperature. The line corresponds
in Fig. (a). (color online) to a linear fit.

The obtained results are plotted in Fig. 3, which shows


the stress-induced entropy change S as a function of tem- Clausius–Clapeyron equation. Figure 4 shows the transition
perature, for the different levels of applied stress. stress as a function of temperature, where σt has been taken as
It is interesting to point out that the maximum stress- the inflection point of each isothermal stress–strain curve. A
induced entropy change remains constant within the errors linear rise in σt with increasing temperature is observed, with a
(independent of temperature and stress). This value corre- slope dσt /d T = 2.01 MPa/K. By taking an average value for
sponds to the whole entropy change of the martensitic transi- the discontinuity of strain ε = 0.075 ± 0.005 (see Fig. 5(a)),
tion St . we obtain St = −1.15 ± 0.05 J/mol K. The good agreement
As previously mentioned, the temperature and stress of this value with that obtained from the elastocaloric effect,
dependence of the strain can be expressed by Eq. (11). In our suggests that irreversible effects are negligible for the marten-
case, a convenient choice for the shape-function is F (z) = sitic transition under study. To support this assertion, we have
tanh−1 (z), which leads to the following expression for the computed the area of the hysteresis loops obtained at differ-
stress-induced entropy change ent temperatures, which quantify the dissipated energy in a
f
  full cycle E diss (to within a good approximation it is twice the
St Tt (σ ) − T
S(0 → σ ) = 1 + tanh . (14) energy dissipated in each branch, E diss). Results are shown
2 T in Fig. 5(b). It is evident that there is no significant depen-
This expression has been fitted to the stress-induced dence of the dissipated energy on temperature, which leads to
entropy change data given in Fig. 2(a), and the results for a vanishing second term in Eq. (8), thus confirming that the
each value of the applied stress are shown as continuous lines. Clausius–Clapeyron equation is valid to obtain the entropy
By averaging over the whole set of curves, we obtain the value change at the martensitic transition for this class of materials.
St = −1.21 ± 0.15 J mol−1 K−1 . In order to explore the possibilities of the elastocaloric
It is interesting to compare the value obtained for the effect for mechanical refrigeration, it is important to estimate
whole entropy change (St ) to the value given by the the temperature change associated with the application of a
July 7, 2009 11:59 00059

The Use of Shape-Memory Alloys for Mechanical Refrigeration 77

(a) (b)

Fig. 5. (a) Strain discontinuity at different values of temperature and (b) dissipated energy computed from the area of hysteresis loops, as a function of
temperature.
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com
by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

given stress, which under the assumption of constant C, can the strong dependence of transition temperature on stress (see
be obtained as Fig. 4), and also by the lack of a critical point (as occurring
T in the magnetic case) above which the first-order transition no
T ≈ − S. (15)
C longer occurs. Actually, the upper bound for the applied stress
In the temperature range of interest here, for Cu–Zn–Al, is imposed by the elastic limit of the cubic phase.
C  25 J K−1 mol−1 ,16 so for an adiabatic drop in stress σ In summary by means of stress–strain experiments, it has
(involving the whole transition) of about 30 MPa, the maxi- been shown that shape-memory alloys exhibit a large elas-
mum expected temperature change is 15 K. Notice that this tocaloric effect. The estimated temperature changes associ-
value is several orders of magnitude larger than the typical ated with this effect are larger than those reported for giant
values given by the conventional elastic behavior in solids magnetocaloric materials for fields achievable by permanent
far from any phase transition. It is worth mentioning that the magnets. In addition, the energy losses associated with the
expected temperature change for mechanical refrigeration is first-order character of the transition have been found to be
even larger than the values reported for the magnetocaloric weak. Present results suggest that shape-memory alloys have a
materials for magnetic fields up to 2 T (achievable by perma- potential use for close-to-room temperature mechanical refrig-
nent magnets). Hence, the 15 K value estimated is competi- eration. This new functional property adds to the well-known
tive with the values reported for pure Gd (6 K) and Gd–Si–Ge shape-memory properties in this class of alloys, and opens-up
(7 K).17 the possibility of designing multi-functional devices based on
Another figure of merit to be considered is the refrigerant shape-memory alloys.
capacity, which is defined as2

R= S(T )d T  ST, (16) Acknowledgments
T
We acknowledge financial support from CICyT (Spain) under
which in the case of elastocaloric effect is equal to R =
project MAT2007-62100 and from DURSI (Generalitat de
−εσ , where σ is the change of stress necessary to
Catalunya) through project 2005SGR00969.
change the transition temperature by T (S and ε are
assumed to be constants). For the investigated alloy, we obtain
R = 2.5 J cm−3 for σ = 30 MPa. The interest in mechanical References
refrigeration is that σ can be chosen in a broad range of val- 1. E. Bonnot et al., Phys. Rev. Lett. 100, 125901 (2008).
ues (see Fig. 3). In contrast, this is not the case for giant mag- 2. K. A. Gschneidner, V. K. Pecharsky and A. O. Tsokol, Rep.
netocaloric materials for which large entropy changes are only Prog. Phys. 68, 1479 (2005).
obtained in a relatively narrow temperature interval. The main 3. E. Brück, J. Phys. D: Appl. Phys. 83, R381 (2005).
difference is due to the fact that while tensile experiments can 4. A. Planes, L. Mañosa and M. Acet, J. Phys. Condens. Matter
21, 233201 (2009).
be performed at temperatures well above the transition tem-
5. A. S. Mischenko et al., Science 311, 1270 (2006).
perature at zero stress, in magnetic experiments, the field is 6. T. Strässle et al., Phys. Rev. B 67, 054407 (2003).
always applied close to the transition temperature at zero field. 7. O. Tegus et al., Physica B 319, 174 (2002).
Such a different experimental procedure is made possible by 8. F. Casanova et al., Phys. Rev. B 66, 100401 (2002).
July 7, 2009 11:59 00059

78 L. Mañosa et al.

9. J. Ortín, A. Planes and L. Delaey, The Science of Hysteresis, 13. L. Mañosa et al., Phys. Rev. B 48, 3611 (1993).
Vol. 3, ed. I Mayergoyz and G. Bertotti, pp. 467–553 (Elsevier, 14. A. Planes and L. Mañosa, Solid State Phys., Vol. 55 (Academic
2005). Press, New York, 2001), p. 159.
10. V. K. Pecharsky and K. A. Gschneidner Jr. Adv. Mater. 13, 683 15. E. Bonnot et al., Phys. Rev. B 76, 064105 (2007).
(2001). 16. J. C. Lashley et al., Phys. Rev. B 75, 064304 (2007).
11. F. Casanova et al., Appl. Phys. Lett. 86, 262504 (2005). 17. E. Brück (2008), Handbook of Magnetic Materials Vol. 17,
12. Y. Miyoshi et al., Rev. Sci. Inst. 79, 074901 (2008). Chapter 4, ed. K. H. J. Buschow (Elsevier 2008).
Funct. Mater. Lett. 2009.02:73-78. Downloaded from www.worldscientific.com
by UNIVERSITY OF SYDNEY on 04/30/15. For personal use only.

You might also like