scandium tantalate (PST) ( 14 , 15 ). This design
was inspired by the magnetocaloric regen-
erator that Brown invented in 1975 ( 16 ), im-
proved first by Steyert ( 17 )andlateronby
Barclay ( 18 ), and reported a temperature span
of 5 K. In 1995, Sinyavsky performed prelimi-
nary investigations on a larger prototype that
was 30 cm long and made of 400 g of PST.
Sinyavsky suggested that a temperature
span of 15 K should be achievable ( 19 ). After
the discovery in 2006 of intrinsic giant EC
temperature changes in ceramic and polymer
thin films ( 20 , 21 ), several prototypes were pro-
posed ( 22 – 28 ), most of them following this
same principle of regeneration. These devices
are typically classified as fluid- or solid-based,
depending on the substance that is used to
collect the heat generated by the EC material
( 29 ). Examples of fluid-based coolers are given
by Plazniket al.( 24 ) and Blumenthalet al.
( 26 ), in which lead magnesium niobate–lead
titanate (PMN-PT) plates and multilayer ca-
pacitors were submerged in silicone oil. Con-
cerning solid-state devices, Guet al.( 22 ) based
their cooler on the oscillatory movement of
stainless-steel sheets through PVDF modules,
and Wanget al.( 23 ) developed a heat switch
with BaTiO 3 commercial multilayer capacitors
(BTO MLCs). Despite the various EC materials
and working principles shown, the maximum
temperature span achieved by these prototypes
was 6.6 K ( 22 ).
We followed the principle of active regen-
eration in designing our EC cooling device
( 11 , 17 , 18 , 24 , 26 , 30 ). Active regeneration is
common in heat pumps that are based on
materials with a low intrinsic temperature
change because it permits the device to display
a temperature difference between the hot and
cold side of the regenerator (DTspan)largerthan
the adiabatic EC temperature change of the
material (DTEC). The corresponding ratio be-
tween these two temperature differences is de-
fined as the regeneration factor. The principle
of active regeneration in fluid-based coolers
requires the caloric material to be porous so
that a coolant fluid can be sent back and forth.
The fluid movement must be synchronized
with the heating and cooling steps of the corre-
spondingcaloriceffect.InECs,parallelplates
are the dominating structure because apply-
ing a uniform electric field to the assembly
is easy. Specifically, the structure consists of
thin plates made of the caloric material that
arestuckoneontopoftheother.Inbetween
these plates, spacers are positioned to cre-
ate void channels where the fluid can flow
through. We based the design of our regen-
erator on the circulation of a dielectric fluid
through parallel plates. We built the plates from
Pb(Sc,Ta)O 3 multilayer capacitors (PST-MLCs)
( 31 ) because of their sharp first-order phase
transition and availability. The bespoke sam-
ples were fabricated by Murata Manufactur-
ing. The multilayer structure consists of ceramic
thin layers sandwiched with previously studied
inner electrodes in between ( 32 , 33 ). The MLC
architecture combines the advantages of bulk
and thin-films samples. The inner electrodes
are tens of micrometers apart, which decreases
the voltage required to reach the same amount
of field as in the bulk. Furthermore, the sam-
ple breakdown field increases when reducing
thickness, allowing higher electric fields to beapplied that favor a higherDTEC. The overall
MLC object is millimeters in size, providing
more thermal mass than thin films. This pre-
vents the heat generated with the EC effect
from vanishing into the surroundings before
we can use it. Moreover, the inner electrodes
can also enhance heat exchange ( 34 ). We used
two types of MLCs with 19 and 9 PST layers,
each 38mm thick. The final sample dimen-
sions were 10.6 (L) by 7.2 (W) by 0.9 or 0.5 mm
(T) ( 35 ).
We designed our experiment with the EC
regenerator that we formed with a matrix of
PST-MLCs, a syringe pump that displaces a
fluid, a power supply to trigger the EC effect,
and type K thermocouples to monitor the tem-
perature (Fig. 1A). We enclosed the EC regen-
erator with a heating box to control the starting
temperature (optimally above 25°C) of the
experiment ( 35 ). The AER was connected to
the power supply, which we chose to run in
the current source mode. This supply charges
(and discharges) the EC capacitors according
to the compliance voltage (fig. S1) ( 35 ). The
fluid system consists of a nonclosed single loop
where one end of the regenerator is con-
nected to the syringe pump and the other end
is attached to an unsealed fluid reservoir. Our
operational system (Fig. 1B) is constituted by
cycles of four steps, with the first two and last
two steps occurring simultaneously to emu-
late an Ericsson-Brayton–like cycle ( 30 , 35 ).
We charged the EC capacitors in the first step,
which increased their temperature by means
of the EC effect. In the second step, we ac-
tivated fluid movement to transport the heat
generated to one side of the AER (hot side).126 2 OCTOBER 2020•VOL 370 ISSUE 6512 sciencemag.org SCIENCE
Fig. 2. Modeling results.(A) Temperature span (DTspan) comparison as a function of the applied electric fieldEand timet(inset) for REG1 (fig. S2) ( 35 ). (B) Modeled
DTspanas a function of the cycle periodt. We used a 15.8 Vmm−^1 electric field, corresponding to a 2.2 K adiabatic EC temperature change of the material (DTspan).
RESEARCH | REPORTS