Sie sind auf Seite 1von 4

commentary

Too cool to work


Xavier Moya, Emmanuel Defay, Volker Heine and Neil D. Mathur
Magnetocaloric and electrocaloric effects are driven by doing work, but this work has barely been explored,
even though these caloric effects are being exploited in a growing number of prototype cooling devices.

an solid materials be used as the basis


for economically viable heat-pump
devices that cool food, beverages,
medicine, electronics or populated spaces?
This is a question that occupies universities,
national laboratories and companies. It has
resulted in four large projects within Europe,
and it represents the basis for the biennial
THERMAG conference series, dedicated
primarily to magnetocaloric (MC) cooling
near room temperature using materials
that display thermal changes in response
to magnetic field changes. It was revealed
at THERMAG VI that the many existing
MC prototypes have now been joined by
a third prototype based on electrocaloric
(EC) materials1 (Fig.1a,b), and a second
a

prototype based on mechanocaloric (mC)


materials2 (Fig.1c,d), with thermal changes
driven by changes of electric and stress
fields, respectively. (Caloric materials and
previous prototypes are discussed in ref.3
and references therein.) Unfortunately,
the efficiency of these prototype heat
pumps is rarely discussed, even though
the main motivation is to outperform
existingrefrigerators.
We are led to ask an even more
fundamental question: how much work
is done to drive the nominally reversible
thermal changes in the materials themselves?
These thermal changes are large near ferroic
(often structural) phase transitions in the
vicinity of room temperature, for example,
b

Electrical connector

Ceramic plates

300

SMA
Heat sink

Heat sink

298
296

Heat source
Heat source

T (K)

SMA

294
292

Figure 1 | Recent electrocaloric1 and mechanocaloric2 prototype coolers. a,b, Electrocaloric cooler (a)and
the back part of the cooler (b). A voltage is cycled across 30 ceramic electrocaloric plates while a fluid
flows in alternate directions. This establishes a 3.3K temperature difference that exceeds the adiabatic
temperature changes in the 200-m-thick plate material (Pb[(Mg1/3Nb2/3)O3]0.9[PbTiO3]0.1) by a factor
of 3.7. c,d, Prototype heat pump based on mechanocaloric effects in a ribbon of the superelastic shape
memory alloy (SMA) NiTi (c). The thermal changes are driven by changes in uniaxial stress, and therefore
fall into the subset of mechanocaloric effects known as elastocaloric effects. An infrared thermal
image (d) reveals the temperature differential established by repeatedly using the SMA to absorb heat at
the source and then dump it after moving to the sink. The colour scale shows the temperature in kelvin.
The images in a and b are courtesy of U. Plaznik. The images in c and d are courtesy of S. Seelecke.
202

in ferromagnets and ferroelectrics near the


Curie temperature. Given that work is such a
well-established thermodynamic concept, it
is perhaps surprising that it has been highly
neglected in the caloricliterature.
The thermodynamics of polarizable
bodies in electric or magnetic fields is
inherently tricky because the general
formula for work derived from Maxwells
equations involves an integral over all
space4. It is therefore virtually impossible to
do work on the polarizable body alone, as
additional work must be done to generate
the fields arising from induced poles.
Another complication is that the work can
be done electrically or mechanically. If the
work is done electrically then additional
work must be done to generate the applied
magnetic (electric) field through the flow
(redistribution) of charge. In contrast,
no additional work is performed if the
work is done mechanically by relative
motion with respect to a pre-existing field,
even though the response of the body
isindistinguishable.
Building on the formalism for calculating
work that was established by VolkerHeine
over half a century ago4, we find that it is
indeed more efficient to drive MC effects
mechanically, whereas due to screening it is
more efficient to drive EC effects electrically.
The mechanically driven MC effects are
generically an order of magnitude more
efficient than the electrically driven EC
effects, but these intrinsic efficiencies do
not ultimately limit the extrinsic efficiencies
that may be achieved by recovering as
much work as possible for subsequent use.
In what follows, work will be calculated
without considering mundane contributions
associated with Joule heating, friction
andgravity.
Our figure of merit will be the
dimensionless materials efficiency 5,
=|Q/W|, where electrical or mechanical
work, W, is done to drive highly reversible
caloric effects in an isothermal body,
whose entropy is thus modified such that
heat, Q, flows to (Q<0) or from (Q>0)
the thermal bath. For a given material,
nonlinear field dependences force to

NATURE PHYSICS | VOL 11 | MARCH 2015 | www.nature.com/naturephysics

2015 Macmillan Publishers Limited. All rights reserved

commentary
undergo a trivial increase on decreasing
|Q|, reflecting the fact that in physics as
in life it is sometimes easier to be more
efficient when achieving a bit less. Therefore
it is only meaningful to compare values of
(Q,T) for similar values of |Q|, and it is
primarily interesting to do this when |Q| is
maximized by employing large fields and
operating at a temperature, T = T0, that is
closely associated with the zero-field phase
transitiontemperature.
Values of both |Q| and |W| will be
expressed per unit volume of material,
assuming specific but reasonably realistic
scenarios in which fields, flux densities
and order parameters are uniform. Direct
caloric measurements are challenging and
so we primarily use values of |Q| that were
previously obtained from the well-known
indirect method3. For each material,
these values of |Q| were calculated from
measurements of order parameter as a
function of applied field at selected values
of temperature. Here we use the subset of
measurements at T0 to calculate values of
|W| and hence.
The materials efficiency, , corresponds
to each of the two legs in a rather useless
isothermal cycle that incurs no net work,
and so it does not describe a useful
cooling device with different sink and load
temperatures. However, it represents a
guide to the selection of materials in terms
of energy efficiency, without forcing one
to specify arbitrary details to describe a
coolingcycle.

Magnetocaloric efficiency

When magnetizing a magnetic material,


each element of the electrical and/or
mechanical work done, dW, is given most
generally by the integral of HdB over all

space, where the magnetic flux density,


B=0(H+M), depends on the magnetic
field, H, and the sample magnetization,
M, and where 0 is the permeability of
free space. If electrical work is done to
generate an applied magnetic field, H0, by
incrementally varying the current through
an infinitely long solenoid that is closely
wound around an infinitely long MC sample,
then there is no external field and so dW
may be evaluated by integrating HdB over
the sample volume alone. Moreover, we may
write H=H0 as the sample has no poles with
which to generate any field.
This evaluation procedure may be used
for the long, thin MC samples typically
studied if we imagine a closely wound
solenoid of the same length, but work is
underestimated by neglecting the fringing
field generated by the solenoid, and by
neglecting the demagnetizing (stray) field
generated inside (outside) the sample
by the small number of induced poles at
each end. Here, for long, thin MC samples
magnetized isothermally at T0, we obtain
a lower bound on the volume-normalized
electrical work, |W|, by integrating HdB,
using literature data of the form M(H0) and
assuming H=H0 due to the small number
of well-separated poles (collinear vectors are
denoted as scalar quantities).
If, instead, mechanical work is done to
generate H0 in the sample by relative motion
with respect to an inalterable permanent
magnet, then dW can be mathematically
reduced in a non-trivial manner 4 to the
integral of 0MdH0 over the sample volume
alone. There are no assumptions about
sample shape, and the negative sign reflects
the attraction between permanent magnets
and magnetic materials. For the long, thin
MC samples magnetized isothermally at

T0, we obtain volume-normalized values of


|W| by integrating 0MdH0 from the same
M(H0) data asbefore.
Whether electrically or mechanically
generated, the magnetic field applied to
an isothermal MC material at T0 produces
a heat, Q, that is typically evaluated by
integrating 0T0(M/T)H with respect to
H. This method of evaluation which
follows from the Maxwell relation,
0(M/T)H=(S/H)T, assuming the
experimental condition of constant stress
was exploited in the literature615 without the
factor of T0 to calculate the corresponding
changes of entropy, S, from measurements
of M(H0) at selected values of T, assuming
H=H0 for long, thin samples. We will
obtain volume-normalized values of |Q|
from published values of mass-normalized
entropy change by introducing factors of T0
anddensity.
For selected MC materials, mechanical
work was evaluated by integrating between
H0=0and a value corresponding to the
2T flux density that may just be provided
by good permanent magnets. Values of
electrical work (MC effects) were evaluated
here (previously) by integrating between
the same limits. The resulting values of
reveal MC materials to be an order of
magnitude more energy efficient when
driven by permanent magnets rather
than solenoids (Table1 and upper blue
ellipsoids in Fig.2). This discrepancy
arises because the solenoid must generate
applied fields that are already present when
using the permanent magnet. Even if it
were practical for solenoids to generate
5T, the corresponding values of (Table1
and bottom blue ellipsoid in Fig.2) are
suppressed because |W| increases faster
than |Q|, as caloric effects tend tosaturate.

Table 1 | Magnetocaloric materials.


Electrical work

Mechanical work

Material

T0 (K)

|0H0| (T)

|Q| (J cm3)

|W| (J cm3)

|W| (J cm3)

Ref.

Gd

294

2 (5)

12.7 (25.2)

2.2 (11.1)

5.8 (2.3)

1.0

12.6

2.0 (10.9)

14.3 (3.5)

1.1

27.2

2.0 (10.7)

14.5 (3.6)

1.1

27.2

2.8 (11.3)

13.5 (3.5)

0.7

50.0

2.4 (11.0)

15.7 (3.6)

1.0

37.6

Gd5Si2Ge2

H > 0

276

2 (5)

29.1 (38.2)

184

2 (5)

37.6 (40.0)

MnFeP0.45As0.55

308

2 (5)

31.9 (39.9)

2.1 (11.2)

14.9 (3.6)

0.3

96.7

Ni52.6Mn23.1Ga24.3

300

2 (5)

15.2 (43.7)

1.6 (10.1)

9.2 (4.3)

0.9

17.4

10

Ni50Mn37Sn13

299

2 (5)

18.7 (50.1)

1.7 (10.3)

11.2 (4.9)

0.3

66.8

11

Ni50Mn34In16

190

2 (5)

14.7 (18.9)

1.7 (11.9)

8.6 (1.6)

0.4

36.6

12

Ni45.2Mn36.7In13Co5.1

317

45.0

2.0

22.5

0.2

211

13

Mn1.30Fe0.65P0.5Si0.5

267

29.0

2.1

14.0

0.7

40.6

14

Fe49Rh51

317

43.5

3.0

14.4

1.2

35.9

15

LaFe11.7Si1.3

H < 0
H > 0
H < 0

7
8

T0, operating temperature; |0H0|, change in applied magnetic field; 0, permeability of free space; Q, heat; W, work; =|Q/W|, materials efficiency. Data for |0H0|=5T are presented parenthetically where |Q| is
available. The hysteretic behaviour exhibited by Gd5Si2Ge2 and LaFe11.7Si1.3 is identified by up-sweep (H > 0) and down-sweep (H < 0) measurements. Values of |Q| were obtained from the cited references and values
of |W| were obtained from data therein.

NATURE PHYSICS | VOL 11 | MARCH 2015 | www.nature.com/naturephysics

2015 Macmillan Publishers Limited. All rights reserved

203

commentary
Electrocaloric efficiency

100

t, 2

ma

e
gn

T)

ergy
(en
EC very)
reco

T)
n)
id, 2
rive
eno
l
o
s
lly d
(
a
C
c
i
M
tr
elec
mC
EC (
5 T)
,
d
i
eno
(sol
MC

10

t
en

an

erm

p
C(

When polarizing an electrically polarizable


material, each element of the electrical
and/or mechanical work done, dW, is
given most generally by the integral of
EdD over all space, where the electric field,
E=01(DP), depends on the electric flux
density, D, and the sample polarization,
P, and where 0 is the permittivity of free
space. If electrical work is done to generate
an applied electric flux density, D0, by
incrementally varying the charge on infinitearea capacitor plates that lie on either side
of an infinite-area EC sample, then there is
no external field and dW may be evaluated
by integrating EdD over the sample volume
alone. Moreover, whether or not the area
is infinite, we may assert that D=D0 by
assuming that charge does not flow between
the sample and theplates.
This procedure may be used for
evaluating the large EC effects in thin films
with high breakdown fields and low leakage,
but work is very slightly underestimated by
neglecting fringing fields around the thin
active slab between large-area electrodes.
For this type of EC sample, polarized
isothermally at T0, we obtain a lower bound
on the volume-normalized electrical work,
|W|, by integrating EdD, using data in the
form of D(E) from the literature. Data
presented as P(E) are converted into this
form from the resultant field, E=01(DP),
which is small due to screening. The
resultant field is the independent variable
because the experimental control parameter
is voltage, notcharge.
Alternatively, and perhaps somewhat
hypothetically (at least for solid rather
than fluid16 EC materials), mechanical
work can be done to generate D0 within
the sample through relative motion
with respect to a fixed arrangement of
unscreened charges that reside on capacitor
plates or in an electret (vacuum would be
required to prevent electrical breakdown
of air). One may then mathematically

20

40
|Q| (J cm3)

60

Figure 2 | Caloric materials efficiency map. The


materials efficiency, , for materials in Tables13
possessing similar values of caloric heat, |Q|,
at operating temperatures T0. Magnetocaloric
(MC) effects may be driven electrically using
a solenoid to generate |0H0|= 2T or 5T
(light-blue ellipsoids), or mechanically using a
permanent magnet to generate |0H0|=2T
(dark-blue ellipsoid). Electrocaloric (EC) effects
may be driven mechanically using fixed charges
in a vacuum (<<1, not shown), or electrically
by varying the charge on capacitor plates without
(dark-green ellipsoid) or with (light-green ellipsoid,
partially occluded) electrical energy recovery. For
mechanocaloric (mC) effects driven by uniaxial
stress, only Table 3 entries with similar values
of |Q| are represented (red ellipsoid). EC effects
in each material are driven by fields of different
magnitude, as are mC effects.

reduce dW in a non-trivial manner 4 to the


integral of PdE0 over the sample volume
alone. As for MC materials, there are no
assumptions about the sample shape, and
the negative sign reflects the fact that
electric fields attract polarizable materials.
By considering the thin active slab of
each EC film polarized isothermally at T0,
volume-normalized values of |W| can be
obtained by integrating PdE0 from the
same D(E) or P(E) data as before. To do

this, E0was rewritten as the experimentally


accessible parameter 01D, using D=D0, as
explainedabove.
For the thin active slab of each film
at a temperature of T0, the isothermal
EC heat, Q, is typically evaluated by
integrating T0(P/T)E with respect to E.
This follows from the Maxwell relation,
(P/T)E=(S/E)T, assuming approximate
mechanical boundary conditions, namely
constant stress for a free-standing film with
thin electrodes of similar area, and constant
strain if there is clamping from a substrate
and/or unaddressed surrounding film.
As for MC materials, we report volumenormalized values of |Q| based on literature
values of mass-normalized entropy change.
The latter were typically calculated using
the above Maxwell relation with data in the
form of D(E) or P(E) at selected values of
T. Here, P and D are freely interchangeable
given that D=0E+P implies that
(D/T)E=(P/T)E.
For selected EC materials, electrical
and mechanical work were evaluated by
integrating between the limits previously
used to evaluate EC effects, with D0 ranging
from zero to the highest value that does
not induce breakdown (except when the
lower limit is finite17, or when we have
reduced the upper limit 19,21 to restrict |Q|
for comparison). The resulting values of
(Table2 and dark green ellipsoid in Fig.2)
reveal EC materials to be comparable to
MC materials when doing electrical work.
The mechanical work done when driving
EC effects is much greater (<< 1, Table2),
as the resultant electric field undergoes
large changes associated with screening.
In contrast, the pre-existing magnetic field
of a permanent magnet is not screened
by MC materials, leading to the most
intrinsically efficient caloric effects (Table1
and dark blue ellipsoid in Fig.2). A more
complete and detailed treatment of the
thermodynamics presented here is currently
in preparation.

Table 2 | Electrocaloric materials.


Electrical work

Mechanical work

Material

T0 (K)

|E| (kV cm )

|Q| (J cm )

|W| (J cm )

|W| (J cm3)

Ref.

PbZr0.95Ti0.05O3

499

480

31.2

10.3

3.0

9842

0.003

17

0.93PMN-0.07PT

308

720

24.0

4.3

5.6

2036

0.012

18

Pb0.8Ba0.2ZrO3

290

210

46.1

1.7

27.1

5193

0.009

19

P(VDF-TrFE)

353

2000

38.0

5.2

7.5

266

0.14

20

*P(VDF-TrFE)

323

1000

32.5

2.9

11.2

216

0.15

21

P(VDF-TrFE-CFE)

310

3100

38.0

7.9

4.8

333

0.11

20

*P(VDF-TrFE-CFE)

350

3500

61.0

8.7

7.0

187

0.33

22

T0, operating temperature; |E|, change in resultant electric field; Q, heat; W, work; =|Q/W|, materials efficiency. Values of |Q| were obtained from the cited references and values of |W| were obtained
from data therein. PMN=PbMg1/3Nb2/3O3; PT=PbTiO3; P(VDF-TrFE)=poly(vinylidene fluoride-trifluoroethylene) 55/45mol%; *P(VDF-TrFE)=poly(vinylidene fluoride-trifluoroethylene) 65/35mol%;
P(VDFTrFECFE)=poly(vinylidene fluoride-trifluoroethylene-chlorofluoroethylene) 59.2/33.6/7.2mol%; *P(VDFTrFECFE)=poly(vinylidene fluoride-trifluoroethylene-chlorofluoroethylene) 56.2/36.3/7.6mol%.

204

NATURE PHYSICS | VOL 11 | MARCH 2015 | www.nature.com/naturephysics

2015 Macmillan Publishers Limited. All rights reserved

commentary

Table 3 | Mechanocaloric materials.


Mechanical work
Material
Cu68.1Zn15.8Al16.1
NiTi

T0 (K)
> 0
< 0
> 0
< 0

Fe68.8Pd31.2

|| (GPa)

300

0.13

295

0.65

240

0.10

|Q| (J cm3)

|W| (J cm3)

8.7

5.5

7.9

6.1

91.2

34.0

2.7

60.4

11.5

5.3

9.0

1.1

8.4

47.8

Ref.
23
24
25

T0, operating temperature; ||, change in the applied/resultant uniaxial stress field; Q, heat; W, work. The hysteretic behaviour exhibited by Cu68.1Zn15.8Al16.1 and NiTi is identified by up-sweep ( > 0) and down-sweep
( < 0) measurements.

Mechanocaloric efficiency

For completeness, we include an analogous


analysis for mechanical work done on mC
materials. There are no pressurevolume
data for barocaloric materials, so our
analysis is limited to uniaxial stressstrain
() data for elastocaloric materials such
that dW=d (Table 3 and red ellipsoid in
Fig.2). Based on our very limited dataset, it
seems that the efficiency of mC materials is
relatively low.

Energy recovery

Knowledge of the second law of


thermodynamics might not be cool, as
lamented by C. P. Snow, but it is certainly
necessary to appreciate that net work
must be done to pump heat uphill and
achieve cooling. The work associated with
driving and undriving a caloric material
through any useful cycle is necessarily
larger, but much of the driving work may
be automatically or artificially recovered on
undriving. To illustrate automatic energy
recovery, consider permanent magnets that
address some angular fraction of a spinning
MC annulus, such that material moving
into the field region is compensated by
material moving out. This strategy is not
employed in all MC prototypes, suggesting
that it could be desirable to use mechanical
or other means to artificially recover as
much work as possible to drive subsequent
MCeffects.
Automatic energy recovery with solid EC
materials would be challenging, so if one
cannot use EC fluids16 that flow in and out
of a field region, it is important to artificially

recover electrical work to drive subsequent


EC effects. Given that 80% of this work can
indeed be recovered using simple circuitry,
values of 1/|W| and may be enhanced by
factors of five (E.Defayetal., manuscript
in preparation). Therefore extrinsic EC
efficiencies (light green ellipsoid in Fig.2)
can be comparable to the best intrinsic MC
efficiencies (dark blue ellipsoid inFig.2).
Our original question regarding caloric
materials in heat-pump devices raises many
issues beyond the isothermal materials
efficiency discussed here. For example, it
would, in practice, be difficult to reliably
achieve the maximum fields that we
have reported, because cost and design
considerations compromise magnetic field
strength, and because the need to avoid
breakdown compromises electric field
strength. Therefore, and more generally,
the key to commercial exploitation lies in
the development of caloric materials that
perform better, in smaller fields, over a wider
range of temperatures. It is our hope that
the parameter of materials efficiency, , will
play a small part in helping to parameterize
this cool future. And with apologies to
OscarWilde, it will serve to remind that
work is the curse of the coolingclasses.
Xavier Moya, Emmanuel Defay and Neil D.Mathur*
are in the Department of Materials Science, University
of Cambridge, CB3 0FS, UK. Emmanuel Defay is also
at CEA, LETI, Minatec Campus, 17Rue des Martyrs,
38054 Grenoble, France, and the Luxembourg
Institute of Science and Technology (LIST)
Materials Research and Technology Department,
41rue du Brill, L4422 Belvaux,Luxembourg.

NATURE PHYSICS | VOL 11 | MARCH 2015 | www.nature.com/naturephysics

2015 Macmillan Publishers Limited. All rights reserved

VolkerHeine is at the Cavendish Laboratory,


University of Cambridge, CB3 0HE, UK.
*e-mail: ndm12@cam.ac.uk
References

1. Plaznik, U. etal. Appl. Phys. Lett. 106, 043903 (2015).


2. Schmidt, M., Schtze, A. & Seelecke, S. Int. J. Refrigeration
(in the press).
3. Moya, X., Kar-Narayan, S. & Mathur, N.D. Nature Mater.
13, 439450 (2014).
4. Heine, V. Proc. Camb. Phil. Soc. 52, 546552 (1956).
5. Defay, E., Crossley, S., Kar-Narayan, S., Moya, X. & Mathur, N.D.
Adv. Mater. 25, 33373342 (2013).
6. Dankov, S. Yu., Tishin, A.M., Pecharsky, V.K. &
Gschneidner,K.A. Jr Phys. Rev. B 57, 34783490 (1998).
7. Pecharsky, V.K. & Gschneidner, K.A. Jr Phys. Rev. Lett.
78, 44944497 (1997).
8. Fujita, A., Fujieda, S., Hasegawa, Y. & Fukamichi, K. Phys. Rev. B
67, 104416 (2003).
9. Tegus, O., Brck, E., Buschow, K.H. J. & de Boer, F.R. Nature
415, 150152 (2002).
10. Hu, F-X., Shen, B-G., Sun, J-R. & Wu, G-H. Phys. Rev. B
64, 132412 (2001).
11. Krenke, T. etal. NatureMater. 4, 450454 (2005).
12. Krenke, T. etal. Phys. Rev. B. 75, 104414 (2007).
13. Liu, J., Gottschall, T., Skokov, K.P., Moore, J.D. & Gutfleisch, O.
Nature Mater. 11, 620626 (2012).
14. Dung, N.H. etal. Adv. Energy Mater. 1, 12151219 (2011).
15. Nishimura, K., Nakazawa, Y., Li, L. & Mori, K. Mater. Trans.
49, 17531756 (2008).
16. Qian, X-S. et al. Adv. Funct. Mater. 23, 28942898 (2013).
17. Mischenko, A.S., Zhang, Q., Scott, J.F., Whatmore, R.W. &
Mathur, N.D. Science 311, 12701271 (2006).
18. Correia, T.M. et al. Appl. Phys. Lett. 95, 182904 (2009).
19. Peng, B., Fan, H. & Zhang, Q. Adv. Funct. Mater.
23, 29872992 (2013).
20. Neese, B. et al. Science 321, 821823 (2008).
21. Li, X. etal. Appl.Phys.Lett. 101, 132903 (2012).
22. Liu, P.F. et al. NewJ.Phys. 12, 023035 (2010).
23. Bonnot, E., Romero, R., Maosa, Ll., Vives, E. & Planes, A.
Phys.Rev.Lett. 100, 125901 (2008).
24. Cui, J. etal. Appl. Phys. Lett. 101, 073904 (2012).
25. Xiao, F., Fukuda, T. & Kakeshita, T. Appl. Phys. Lett.
102, 161914 (2013).

Acknowledgements

X.M. is grateful for support from the Royal Society and


EPSRC EP/M003752/1.We thank B. Nair for discussions.

205

Das könnte Ihnen auch gefallen