Science - USA (2020-10-02)

DTspanof9K(Fig.2B,greenstar).Last,we
modeled the impact of using water instead
of dielectric fluid, which increased theDTspan
by 20% and reduced the period in half (Fig. 2B,
blue square and blue star). In comparison with
the dielectric fluid, water has six times more
thermal conductivity, for which heat exchange
is enormously enhanced, and three times more
specific heat, for which more heat can be stored.
In addition, water is five times less viscous,
which lowers fluid self-heating and pumping
pressure.
We used the modeling to guide changes to
our design. We removed all structural pieces
to reduce the inactive mass of the system,
provided insulation with polyurethane foam,
used 0.5-mm-thick PST-MLCs, and increased
the regenerator length (fig. S4 and table S4)
( 35 ). Our final structure, labeled as REG2,
consisted of 128 PST-MLCs, distributed in a
matrix of 16 columns by eight rows. The di-
mensions were 115 (L) by 10.4 (W) by 6.25 mm
(T). The total PST-MLC mass of 38.4 g was
only 60% active because of the overlap of the
inner MLC electrode sheets. We restricted
these to avoid short-circuiting. We used the
same dielectric fluid as in REG1 ( 35 ).
We measured theDTECof a 0.5-mm-thick
PST-MLC as a function of the starting tem-
peratureTsfor an applied electric field (E) of
15.8 Vmm−^1 (Fig. 3A). We observed a maximum
DTECof 2.3 K at 38°C. Below 25°C,DTECstarts
decreasing and reaches ~0.6 K at 10°C. Be-
tween 18° and 25°C, we found asymmetric
behavior forDTECbecause its amplitude is
different when turning the field on (heating)
or off (cooling). This range corresponds to a
region where the ferroelectric (FE) phase of
PST is stable and applying anEcannot fully
drive the desired phase transition from para-
electric (PE) to FE. The optimum working tem-
perature lies in the range of 25° to 50°C, in
which the PE phase is stable and the FE phase
can be fully driven by anE. Likewise, we mea-
suredDTECas a function ofEfor a starting
temperatureTs= 30°C (Fig. 3B). As we ex-
pected,DTECincreases with the applied field
and reaches a common linear regime once the
field has brought the material into the FE
phase. In this case, we observed no asymmetry
between heating and cooling. We measured a
DTECin the material of 2.37 K for a maximum
applied voltageVof 700 V (E=18.4Vmm−^1 ).
These values are in agreement with previous
results ( 31 ), in which better B-site–ordered
0.9-mm-thick PST-MLCs displayed aDTECof
2.6 K in the same temperature range forE=
15.8 Vmm−^1. For this comparison, we have
taken the value reported assuming total in-
ternal thermalization because active volume
represents only 60% of the MLC body. A max-
imum adiabaticDTEC> 5 K was reported at
E= 29.2 Vmm−^1 near 330 K in the MLC ac-
tive volume ( 31 ). Our samples break down at

this level of anEmagnitude, so we limited our

field to 15.8 Vmm−^1.

We measured a maximumDTspanof 13.0 K

after 1500 s of operation in REG2 (Fig. 3C).

Tswas set to 30°C to make use of the entire

EC effect temperature window (Fig. 3A) and

maximize theDTspan. Under these conditions,

the PST-MLCs displayed aDTECof 2.2 K, which

translates to a regenerator factor of 5.9. The

DTspanwe measured is more than one order

of magnitude higher than our initial design

and much higher than several other notable

EC (Fig. 3D). The large variation of temper-

ature we measured is very close to the most

competitive elastocaloric prototypes, which

have aDTspanof 15.3 K ( 11 ). We believe that

there is still room for improvement. Most of

our PST-MLCs did not have an even and flat

shape. These morphologies cause irregular-

ities in the fluid slits or even force complete

closure, which compromises heat exchange

and lowers heat regeneration. Moreover, the

thickness of our PST-MLCs (0.5 mm) is still

thicker than the EC bulk plates used in most

of the literature examples, reaching values as

low as 200mm( 11 , 19 , 22 , 24 , 26 , 28 ). Thus,

heat regeneration could likely be further

enhanced by the use of thinner and more

regularly shaped PST-MLCs. Highly ordered

PST-MLCs with correspondingly higher break-

down strength would allow increasing theDTEC

by more than 50% ( 31 ), increasing accordingly

theDTspanof the device. The use of water as a

working fluid instead of dielectric fluid could

also increase performance, as shown by our

model results. Currently, water is a challenge

for use in EC prototypes because it requires

electrically insulating the entire EC material.

Last, a third regenerator, REG3, with 32 0.9-

mm-thick PST-MLCs and an electrical heater

in the cold side, was built to report cooling

power values following the same design as

REG2. We measured 3 K of no-load tempera-

ture span and 0.26 W (12 W kg–^1 per mass of

PST-MLCs) of maximum cooling power. These

data were used to fit our numerical model to

predict the performances of other configura-

tions. For the REG2 device, simulations dis-

played a no-load temperature span of 12.8 K,

which is very close to the 13 K in the exper-

iment, and a maximum cooling power of 1.22 W

(32 W kg−^1 ). In addition, the hypothetical con-

figuration of 0.3- and 0.2-mm-thick PST-MLCs

in ( 31 ), withDTECof 5.5 K, and water as a cool-

ant fluid, was modeled. Simulations reported a

no-load temperature span of 33 and 47.5 K and a

maximum cooling power of 10.6 W (550 W kg−^1 )

and 16.3 W (850 W kg−^1 ), respectively (fig. S5)

( 35 ), which is similar to the best magneto-

caloric and elastocaloric coolers.

Our work should raise interest not only in

the scientific community, but also in industry;

now that the 10 K barrier has been crossed,

large cooling powers are predicted, and en-

hanced efficiency can be obtained through

energy recovery ( 7 ). We hope that this will

promote the development of EC prototypes

in the years to come as an alternative to the

currently dominating vapor-compression

systems.

REFERENCES AND NOTES

1. IEA,“The Future of Cooling”(IEA, 2018); http://www.iea.org/reports/
   the-future-of-cooling.


2. S. Fähleret al.,Adv. Eng. Mater. 14 , 10–19 (2012).


3. L. Mañosa, A. Planes, M. Acet,J. Mater. Chem. A Mater. Energy
   Sustain. 1 , 4925–4936 (2013).


4. X. Moya, S. Kar-Narayan, N. D. Mathur,Nat. Mater. 13 , 439– 450
   (2014).


5. I. Takeuchi, K. Sandeman,Phys. Today 68 , 48–54 (2015).


6. X. Moya, E. Defay, V. Heine, N. D. Mathur,Nat. Phys. 11 ,
   202 – 205 (2015).


7. E. Defayet al.,Nat. Commun. 9 , 1827 (2018).


8. S. Qianet al.,Sci. Technol. Built Environ. 22 , 500– 506
   (2016).


9. M. Schmidt, A. Schütze, S. Seelecke,Int. J. Refrig. 54 , 88– 97
   (2015).


10. H. Ossmer, C. Chluba, S. Kauffmann-Weiss, E. Quandt, M. Kohl,
    APL Mater. 4 , 064102 (2016).


11. J. Tušek, K.et al.,Nat. Energy 1 , 16134 (2016).


12. H. Ossmeret al.,Smart Mater. Struct. 25 , 085037 (2016).


13. B. Yu, M. Liu, P. W. Egolf, A. Kitanovski,Int. J. Refrig. 33 ,
    1029 – 1060 (2010).


14. Y. V. Sinyavsky, N. D. Pashkov, Y. M. Gorovoy, G. E. Lugansky,
    L. Shebanov,Ferroelectrics 90 , 213–217 (1989).


15. Y. V. Sinyavsky, V. M. Brodyansky,Ferroelectrics 131 , 321– 325
    (1992).


16. G. V. Brown,J. Appl. Phys. 47 , 3673–3680 (1976).


17. W. A. Steyert,J. Appl. Phys. 49 , 1216–1226 (1978).


18. J. A. Barclay, W. A. Steyert, Active magnetic regenerator. U.S.
    patent no. 4.332.135 (1982).


19. Y. V. Sinyavskii,Chem. Petrol. Eng. 31 , 295–306 (1995).


20. A. S. Mischenko, Q. Zhang, J. F. Scott, R. W. Whatmore,
    N. D. Mathur,Science 311 , 1270–1271 (2006).


21. B. Neeseet al.,Science 321 , 821–823 (2008).


22. H. Guet al.,Appl. Phys. Lett. 102 , 122904 (2013).


23. Y. D. Wanget al.,Appl. Phys. Lett. 107 , 134103 (2015).


24. U. Plazniket al.,Appl. Phys. Lett. 106 , 043903 (2015).


25. D. Setteet al.,APL Mater. 4 , 091101 (2016).


26. P. Blumenthal, C. Molin, S. Gebhardt, A. Raatz,Ferroelectrics
    497 ,1–8 (2016).


27. T. Zhang, X.-S. Qian, H. Gu, Y. Hou, Q. M. Zhang,Appl. Phys.
    Lett. 110 , 243503 (2017).


28. R. Maet al.,Science 357 , 1130–1134 (2017).


29. P. Blumenthal, A. Raatz,EPL 115 , 17004 (2016).


30. A. Kitanovskiet al.,“Active Magnetic Refrigeration”in
    Magnetocaloric Energy Conversion(Springer, 2015), pp 97–166.


31. B. Nairet al.,Nature 575 , 468–472 (2019).


32. S. Kar-Narayan, N. D. Mathur,J. Phys. D Appl. Phys. 43 , 032002
    (2010).


33. X. Moya, E. Defay, N. D. Mathur, S. Hirose,MRS Bull. 43 ,
    291 – 294 (2018).


34. Y. Liu, B. Dkhil, E. Defay,ACS Energy Lett. 1 , 521– 528
    (2016).


35. Materials and methods are available as supplementary
    materials.

ACKNOWLEDGMENTS

We thank X. Moya, N. Mathur, and B. Nair for discussions. We

thank R. Faye, H. Strozyk, and S. Nicolau for their help in the

development of REG1. We also thank N. Furusawa, Y. Inoue, and

K. Honda for their kind assistance in fabricating the MLCs.

Funding:A.T., P.L., Y.N., M.G., O.B., and E.D. acknowledge the

Fonds National de la Recherche (FNR) of Luxembourg for supporting

this work through the projects MASSENA PRIDE/15/10935404/

Defay-Siebentritt, CAMELHEAT C17/MS/11703691/Defay, and

COFERMAT FNR/P12/4853155/Kreisel.Author contributions:

E.D. suggested the experimental study. S.H and T.U. prepared

the PST-MLC samples. M.G. designed REG1 with E.D. and

A.T.; A.T. developed and investigated the numerical modeling that

led to Fig. 2, A and B. E.D. and A.T. designed REG2 with M.G.;

A.T. and Y.N. ran the set of experiments that led to Fig. 3, A and

B. A.T. and P.L. ran the experiments with the prototype that led

to Fig. 3C. A.T. ran the experiments and simulations that led to

fig. S5. E.D., A.T., and P.L. interpreted the key findings. P.L. prepared

128 2 OCTOBER 2020•VOL 370 ISSUE 6512 sciencemag.org SCIENCE

RESEARCH | REPORTS