Realization of large magnetocaloric effect in the Kagome

antiferromagnet Gd3BWO9 for Sub-Kelvin cryogenic refrigeration

Fangyuan Song1*, Xinyang Liu2*, Chao Dong1, Jin Zhou1, Xinlong Shi1, Yuyan Han3, Langsheng
Ling3, Huifen Ren2, Songliu Yuan1, Shun Wang1,4, Junsen Xiang2二, Peijie Sun2,三, Zhaoming Tian1 

1
Wuhan National High Magnetic Field Center and School of Physics, Huazhong University of
Science and Technology, Wuhan, 430074, China
2
Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese
Academy of Sciences, Beijing, 100190, China
3
Anhui Key Laboratory of Low-energy Quantum Materials and Devices, High Magnetic Field
Laboratory, Chinese Academy of Sciences, Hefei, Anhui 230031, China
4
School of Fundamental Physical quantum Physics, Huazhong University of Science and
Technology, Wuhan, 430074, China

Rare-earth (RE) based frustrated magnets have attracted great attention as excellent
candidates for magnetic refrigeration at sub-Kelvin temperatures, while the experimental
identification on systems exhibiting both large volumetric cooling capacity and reduced working
temperatures far below 1 K remain to be a challenge. Here, through the ultra-low temperature
magnetism and thermodynamic characterizations, we unveil the large magnetocaloric effect
(MCE) realized at sub-Kelvin temperatures in the frustrated Kagome antiferromagnet Gd3BWO9
with TN~1.0 K. The isothermal magnetization curves indicate the existence of field (B) induced
anisotropic magnetic phase diagrams, where four distinct magnetic phases for B // c-axis and five
magnetic phases for B // ab-plane are identified at T< TN. The analysis of magnetic entropy S(B, T)

data and direct adiabatic demagnetization tests reveal a remarkable cooling performance at sub-
Kelvin temperatures featured by a large volumetric entropy density 502.2 mJ/K/cm3 and a low
attainable minimal temperature Tmin~168 mK from the initial cooling condition of 2 K and 6 T,
surpassing most of Gd-based refrigerants previously documented in temperature ranges of 0.25-
4 K. The realized Tmin~168 mK far below TN ~ 1.0 K in Gd3BWO9 is related to the combined
effects of magnetic frustration and criticality-enhanced MCE, which together leave a substantial
magnetic entropy at reduced temperatures by enhancing spin fluctuations.
Keywords: magnetocaloric effect, frustrated magnets, quantum criticality, phase diagram

*
These authors contributed equally to this work.
二
Corresponding author: xiangjs@[Link]
三
Corresponding author: pjsun@[Link]

Corresponding author: tianzhaoming@[Link]
1
1. Introduction

Experimental realization of sub-Kelvin cooling is imperative for both fundamental research

and modern technology applications. Adiabatic demagnetization refrigeration (ADR) based on the
magnetocaloric effect (MCE) of magnetic refrigerants [1, 2], provides a valuable technique for
achieving ultra-low temperatures. In comparison with the prevailing cryogenic refrigeration based
on the liquid helium (3He-4He) dilution refrigeration, ADR cooling technique has the advantages
such as cryogen-free operation, environmental sustainability, compact design and ease of
operation, etc [3-5]. Thus, this technique has attracted growing attention for its potential of
widespread applications in solid-state based quantum devices/sensors and astronomical
cryogenic systems under microgravity conditions [6, 7]. Currently, commercial magnetic
refrigeration down to the sub-Kelvin temperatures uses the dilute paramagnetic salts with very
low ordering temperature (To) such as Ce2Mg3(NO3)12·24H2O (CMN) [8] and
Mn(NH4)2(SO4)2·6H2O (MAS) [9]. Here, the attainable minimum temperature (Tmin) is limited by
the To of magnetic refrigerants. However, the low volumetric magnetic ion density (Nmag) diluted
by the nonmagnetic ions restrict their cooling capacity and specific power. This limitation has
driven the identification of novel type of efficient ADR materials working at sub-Kelvin
temperatures, which need to combine a large Nmag and a low To. On the other side, larger Nmag of
ADR materials usually lead to a higher To and correspondingly smaller residual magnetic entropy
at low temperatures since the entropy release happens at a narrow temperature window near To.
Recently, this dilemma has been partially solved by employing the geometrically frustrated
magnets (GFMs) as the ADR coolants [10-15], where To can be suppressed by geometrical
frustration effect and strong spin fluctuations, leaving a substantial ground state entropy at
reduced temperatures [11,15-19].
In comparison with the 3d transition-metal based GFMs, rare-earth (RE) based compounds
are more competitive ADR coolants in sub-Kelvin temperatures due to the intrinsic weaker
exchange interactions among the 4f local moments, enabling them to easily reach lower working
temperatures in the same Nmag. Among them, Gd-based GFMs are especially attractive for
improving the cooling capacity due to the large magnetic entropy SM=Rln(2S+1)=Rln8 in contrast
to the SM=Rln2 in the other RE-based compounds that exhibit effective spin-1/2 moments, since
only the entropy of crystal-electric-field (CEF) ground state of RE ions can be effectively used in
this temperature window [16]. Indeed, the recent cooling performance tests have revealed that
Gd-based coolants can possess a larger magnetic entropy change (∆SM) and a longer cooling
duration time at temperatures around Tmin than the isostructural Yb-based compounds[11,18-21],
despite the much lower To for the latter. As typical examples, the frustrated hyperkagome lattice
Gd3Ga5O12 (GGG) [4,22,23] and triangular lattice KBaGd(BO3)2 [18,19] compounds have been
revealed to be excellent ADR coolants, which can achieve Tmin ~ 322 mK and Tmin~ 122 mK upon
demagnetization from the initial conditions of 2 K and 7 T. Stimulated by the effectiveness of spin
frustration on enhancing sub-Kelvin refrigeration, other Gd-based frustrated magnets including
2
pyrochlore-lattice Gd2Ti2O7 [24], face-centered cube (FCC)-lattice Ba2GdSbO6 [15] and frustrated
spin chain Gd9.33[SiO4]6O2 [25] have also been identified. To further promote the cooling
performance of Gd-based magnets, it is essential to seek for new types of frustrated lattice
systems with high Nmag and the suitable working temperature window with excellent cooling
performance. Meanwhile, new experimental approaches by utilizating novel magnetic states and
enhanced MCE near quantum criticality are essential for advancing magnetic cooling.
Kagome-lattice antiferromagnets as typical GFMs with high degree of geometric frustration,
can be used as ADR coolants. As one candidate, the recently discovered RE-based kagome
compounds RE3BWO9 (RE=Pr-Ho) have been intensely studied for hosting exotic magnetic states
[26]. In these series, magnetic RE3+ ions occupy on the kagome lattices and are stacked in the
“ABAB”-type fashion along the c-axis [see in Fig.1(a) ]. In Gd3BWO9, the small intralayer and
interlayer’s Gd-Gd distances (3.872-4.799 Å) indicate a magnetically dense structure enabling a
large cooling capacity. Here, we report the low-temperature anisotropic magnetism, specific heat
and ADR cooling performance of Gd3BWO9 antiferromagnet ordered at TN~ 1 K. Based on the
systematical magnetization measurements, the field-temperature magnetic phase diagrams are
established. More importantly, both the analysis of magnetic entropy data and direct ADR test
reveal Gd3BWO9 to be an excellent MCE material, featured by the combination of a large
volumetric magnetic entropy, an attainable quite low Tmin~168 mK and a long holding time after
the ADR process from the initial condition of 2 K and 6 T.

2. Experimental section

Single crystals of Gd3BWO9 were grown using a flux method as described in previous
reports [27, 28], the transparent single crystals with typical dimensions of 5mm×2mm×1mm can
be obtained [see Fig. 1(b)]. The crystal structure was determined by Bruker single-crystal X-ray
diffraction (SXRD) using graphite-monochromated Mo Kα radiation (λ = 0.71073 Å), the refined
crystallographic data are listed in Tables S1-S3 in the Supplemental Material [29]. Using the
crushed single crystals, powder X-ray diffraction (XRD) spectra were collected in a PAN Analytical
X’Pert Pro MPD diffractometer with Cu Kα radiation (λ = 1.5418 Å), the refined XRD spectra is
shown in Fig. S1. From that, we extract the information on lattice parameters and crystal structure.
Temperature dependence of magnetic susceptibility  (T) and isothermal field (B) dependent
magnetization M(B) measurements between 1.8 K and 300 K were conducted using the Quantum
Design superconducting quantum interference device magnetometer. At temperatures down to
0.4 K, magnetic characterizations were carried out by the MPMS3 equipped with a 3He option.
Specific heat data were collected using the physical properties measurement system (PPMS,
Quantum Design) at different magnetic fields. The ultra-low temperature specific heat with
temperature down to 80 mK was measured in PPMS using the heat capacity option equipped
with a dilution refrigeration system.
To characterize magnetic cooling performance in sub-Kelvin temperatures, the quasi-
isentropic temperature change with varying magnetic field and holding time at zero field of the
3
samples were performed on the PPMS based setup by using a homemade sample stage, as
described in Refs. [13,30]. Both Gd3BWO9 single crystals with a mass of 0.7 g and polycrystals
with a mass of 3 g samples were employed for the measurements, and the silver powders with a
mass of 1.5 g were mixed with the polycrystalline samples to improve the thermal conductivity.
During the quasi-adiabatic demagnetization (ADR) measurements, the initial cooling conditions
were set to Bi= 2-7 T and Ti = 2 K. Additionally, at temperatures above 1.8 K, the adiabatic
temperature changes in response to field have been measured using pulsed magnetic field under
quasi-adiabatic conditions [31,32], which were carried out at the Wuhan National High Magnetic
Field Center. Pulsed high fields up to 32 T were generated by a long-pulse magnet with duration
of 80 ms, and the exact temperature was detected by a thin film RuO2 thermometer on the
surface of the sample. Due to the small and monotonic magnetoresistance response in our
measured field ranges of RuO2 film, RuO2 thermometer can be calibrated to ensure the accuracy
of measurement [31]. Additionally, the Au16Ge84 thermometer is used for the measurement and
which gives the same magnetocaloric data in pulsed high fields. At various based temperature
(To), the temperature (Ts) of thermometer attached on the Gd3BWO9 samples versus magnetic
field were obtained for field applied along the c-axis.

3. Results and discussion

3.1 Magnetic ground state

Gd3BWO9 crystallizes in the hexagonal structure with space group P63. Based on the
structural refinements of XRD data, the resultant lattice parameters are a = b = 8.557(3) Å and c
= 5.4082(18) Å, respectively. The structure analysis confirms a single phase of the synthesized
single crystals. No distinct antisite occupations between magnetic Gd3+ and non-magnetic
B3+/W6+ cations are detected, which can be due to the large differences of ionic radii and distinct
local oxygen coordination environments between Gd3+ and B3+/W6+ cations. It is found that the
magnetic Gd3+ ions form GdO8 dodecahedrons with low local symmetry (point symmetry C1). The
magnetic Gd³⁺ ions form a distorted Kagomé lattice in the ab-plane that are stacked in the “AB”
type fashion along the c-axis [see Fig. 1(a)]. Within the Kagome plane, the two Gd3+ regular
triangles have different Gd-Gd bond lengths of ������
1 =4.228 Šand ������
2 =4.799 Å slightly larger
than the interlayer spacing separation of ������ =3.872 Å, in accordance with the results of
polycrystals [26]. Herein, the small nearest neighboring Gd-Gd distances enable a large magnetic
ion density Nmag=17.49×1021 cm-3, ~40% larger than the hyperkagome lattice Gd3Ga5O12 system
[22].
Figure. 1(c) shows the isothermal magnetization M(B) curves at 1.8 K for B along the c-axis
and ab-plane, respectively. For field oriented within the ab-plane, the b*-axis is selected as
schematically shown in Fig. 1(c). A weak magnetic anisotropy is observed between the two field
directions, in accordance with the magnetic susceptibility  (T) data, which nearly overlap at
temperatures above 2 K [ see Fig. 1(d)]. In temperature range of 5 K ≤ T ≤ 15 K, the Curie-Weiss
4
(CW) fitting on the inverse magnetic susceptibility 1/  (T) yields the effective magnetic moments
μeff,c= 8.31 μB/Gd (B // c-axis) and μeff,b* = 8.28 μB/Gd (B // ab-plane), which are slightly larger than
the free ion moments �free
��� = 7.94 �B of Gd ions (4f , S=7/2,L=0). The negative CW temperatures
3+ 7

 CW,c = -0.87 K (B // c-axis) and  CW,b* = -1.13 K (B // ab-plane) reveal the dominant
antiferromagnetic (AFM) interactions between the local moments of Gd3+ ions.
To unveil the magnetic ground state, zero-field specific heat Cp(T) down to 80 mK is shown in
Fig. 2(a). A sharp peak at TN~1.0 K suggests the onset of long-range antiferromagnetic (AFM)
order of Gd3BWO9. After subtracting the lattice contribution CLatt using the specific heat of
nonmagnetic analog Eu3BWO9 as background, the resultant magnetic specific heat CM(T) is
presented in Fig. 2(b). The magnetic entropy SM(T) can be obtained by integrating Cm/T versus T,
which approaches saturation of Rln8 at 5 K as expected for Gd3+ ions with S = 7/2. Also, we can
notice that large magnetic entropy ~0.6Rln8 is released well below TN~ 1 K, rendering Gd3BWO9
suitable for sub-Kelvin magnetic refrigeration. Below TN, CM(T) curves display a crossover
behavior at temperature T*~0.37 K, above T* the specific heat follow a power law CM(T)∝T1.5
(0.37 K < T <1.0 K) and below that it changes to CM(T)∝T2 (T < 0.37 K) behavior [see the inset of
Fig. 2(a)]. This is different from the conventional three-dimensional antiferromagnets following the
CM(T)~T3 dependence. Here, CM(T) data drops out slower than the T3 dependence, thus lead to
larger amount of entropy released at temperatures below TN. Besides, the CM(T) ∝ T2

dependence at T< T* suggests the possible novel low-energy excitation with persistent spin
dynamics as the report in other frustrated Gd-based systems including the hyper-Kagome lattice
Gd3Te2Li3O12 and pyrochlore lattice Gd2Pb2O7 [33-35]. The observed magnetic crossover
indicates a possible change of spin dynamics above and below T* ~ 0.37 K.

3.2 Magnetic phase diagrams

The isothermal magnetization M(B) and its field derivative magnetization dM/dB curves at 0.4
K of Gd3BWO9 are presented in Figs. 3(a) and 3(b) for field along the c-axis and b*-axis,
respectively. Along both directions, magnetization approaches its saturation MS= 7.0 μB/Gd3+ at
high fields. Above ~ 2 T, the nearly overlapped M(B) curves indicate the isotropic field responses
[see Fig. S2]. This is expected for the Gd3+ ions with the 4f7 half-filled configuration (S=7/2, L = 0).
At low field regions (B < 1 T), M(B) curves exhibit anisotropic field responses. For B // ab-plane,
three magnetic anomalies are observed at critical fields Bab,1~0.045 T, Bab,2~0.27 T and Bab,3~0.55
T, where Bab,1, Bab,2 and Bab,3 are determined by the peak of the dM/dB curves in Fig. 3(b). At Bab,2
and Bab,3, magnetization reaches ~1/3Ms and ~2/3Ms, respectively. By comparison, for B // c-axis,
two magnetic anomalies appear at fields Bc1~0.35 T and Bc2~0.82 T, and magnetization at Bc2 is
larger than 2/3Ms. As field increases above the saturated field (Bc,s), magnetization approaches its
saturation.
To more clearly illustrate the magnetic behaviors, the field derivative dM/dB curves at
selected temperatures along both directions are shown in Figs. 3(c) and 3(d), respectively. At

5
temperatures near or above TN = 1 K, a single broad hump is observed in the dM/dB curves with
the maximum located at a critical field (Bcr) characterizing a field-induced magnetic crossover,
since short-range spin correlation has been developed in this temperature region from the
analysis of M(B) results ( see Fig. S2). Below TN, several sharp anomalies appear in the dM/dB
curves, signifying the field-induced successive magnetic transitions. Based on the dM/dB data,
the contour plots of field-temperature (B-T) are constructed as shown in Fig. 3(e) and 3(f),
respectively. The intensity of the contour plot corresponds to the values of dM/dB data, which well
depicts all the magnetic phase boundaries. At zero field, a long-range AFM ground state is
established below TN. Upon applying field along the c-axis, two metamagnetic transitions at
critical fields (Bc1, Bc2) are observed at temperatures below 0.88 K, between which an
intermediate AFM phase (IAFM1) can be identified. Alongside the magnetic crossover line
determined by Bcr, the above three critical lines converge at a multiple critical point located at
Bcp~0.41±0.02T and Tcp~0.89 ±0.01 K in the phase diagram. Considering that B is oriented out of
the kagome plane, the IAFM1 phase likely corresponds to a non-coplanar spin configuration,
similar to that observed in its isostructural Nd3BWO9 [36, 37] and triangular lattice
Na2BaCo(PO4)2[13]. As B > Bc2, it enters into another intermediate magnetic state (IAFM2), Gd3+
moments gradually align along the c-axis. Above Bc,s, it becomes fully polarized state (FP). While,
for B // ab-plane, five distinct magnetically ordered phases are identified at T < TN [see Fig. 3(f)].
Moreover, since the observed fractional magnetizations 1/3Ms and 2/3Ms are usually ascribed to
the formation of “up-up-down” and “up-up-up-up-up-down” spin configurations in the triangular
lattice, the inferred low-field AFM (LAFM) ordered state is expected to evolve into a coplanar
canted “Y-shaped” (Bab,2 < B < Bab,3) and canted 'V-shaped' spin structure (Bab,3< B < Bab,s) under
magnetic field in the ab-plane. For B > Bab,s, it enters into the FP state. For the observed
fractional magnetizations, since the Gd3+ ions have large S=7/2 effective moments with weak
quantum spin fluctuations, large thermal fluctuations possibly play an important role on their
appearance, similar to the report in triangular-lattice RbFe(MoO4)2 (S=5/2) [38] and Kagome
lattice Cs2KCr3F12 (S=3/2) magnets [39].

3.3 Specific heat and magnetocaloric effect

The low-temperature CM(T) curves of Gd3BWO9 at selected fields for B // c are presented in
Fig. 4(a). Upon increasing B, the peak in the CM(T) curves shifts to low temperatures and is
gradually smeared out at B~1.5 T. At low magnetic fields (B ≤ 1.2 T), a crossover behavior is
observed in CM(T) curves at temperature T*~0.37 K. Below T*, the CM/T data show a linear
temperature dependence with a slight decrease in slope as field increases. Above 1.5 T, no sharp
peak but only a broad Schottky-like anomaly emerges, with its position gradually shifting to higher
temperatures [see Fig. 4(b)]. A similar trend is observed for B // ab-plane, the CM(T) curves at
high fields of 3 T and 7 T are nearly overlapped with the ones for B // c, indicating the nearly
isotropic field response in polarized magnetic states (B ≥2 T).
Figure. 4(c) displays the resultant magnetic entropy SM(T,B) at different fields, obtained by
6
integrating CM(T)/T versus temperature. The SM(T,B) gives the evolution of magnetic entropy of
Gd3BWO9 at different T and B, which can be used to estimate the MCE and minimum
temperature attainable in the adiabatic demagnetization refrigeration (ADR) process. At low fields
(B ≤ 1 T), the SM(T, B) curves approach saturation as T increases up to 20 K with SM=Rln8,
indicative of full release of magnetic entropy of Gd3+ ions. From the SM(T, B) data, the
magnetocaloric properties can be evaluated as illustrated by the grey arrows, which include an
isothermal entropy change (-∆SM) in a given field change ΔB = Bi-B0 and adiabatic temperature
change ΔTad = Ti-Tf in the isentropic process. In Fig.4(c), B0 = 0 T and Bi = 7 T are selected for
demonstration, with Ti and Tf marking the initial and final temperature in the ADR process from
the initial condition (T = Ti, B = Bi). Here, the -∆SM characterizes the cooling capacity of
refrigerants as the driving force in the ADR process, and Tf refers to the attainable temperature as
field is driven to zero. Setting Ti = 2 K and ΔB = 7 T [see Fig. 4(d) and Fig. 4(f)], Gd3BWO9 shows
a nearly isotropic magnetic cooling performance for B // c and B // ab-plane, with the calculated
∆SM and ΔTad being comparable in both directions. Under ΔB = 2 T, the extracted volumetric
entropy change –ΔSM=210 mJ/K/cm3 at ~ 1 K is considerably large compared to the -ΔSM=126
mJ/K/cm3 in triangular-lattice frustrated KBaGd(BO3)2 and -ΔSM=145 mJ/K/cm3 in benchmark
magnetic refrigerant material Gd3Ga5O12 (GGG) at ~1 K [18,22,40]. This large low-field MCE is
appealing for its practical applications working in the low field region accessible by permanent
magnet (B ≤ 2 T) rather than superconducting magnet. Under ΔB = 7T, the maximum value of -
ΔSM=445 mJ/K/cm3 (0.89Rln8) exceeds most of Gd3+ and Eu2+ based ADR materials with
effective S=7/2 moments as listed in Table 1 [18-25, 40-49].
Figure. 4(e) plots the expected Tf as function of Ti. Starting from the initial condition (Ti = 2 K
and Bi = 7T), Gd3BWO9 is expected to cool down to Tf ~200 mK as field is driving back to 0 T. The
attainable Tf∼ 200 mK is lower than minimum temperature Tmin ∼ 322 mK for the hyperkagome
magnet Gd3Ga5O12 (GGG) magnet [40] and Tmin ∼ 220 mK for NaGdP2O4 [21]. As listed in Table 1,
only the triangular-lattice frustrated KBaGd(BO3)2 realize a much lower Tmin∼ 122 mK than
Gd3BWO9. But the achievable Tmin~200 mK of Gd3BWO9 is well below the base temperature T ~
0.3 K for the standard 3He refrigerators using self-contained and compact 3He cooler [50, 51].
Except for the achievable Tmin in the ADR process, Gd3BWO9 is also characterized by significant
temperature change ∆Tad, defined by the difference between sample temperature (Ts) and based
temperature (T0) as denoted by the blue arrow in Fig.4(c). From the initial based condition (T0= 2
K, B0= 0 T), ∆Tad reaches 20 K as field is increased up to field Bs = 7 T in an isentropic process. At
different field intervals (ΔB=Bs-B0), the extracted ∆Tad at different ΔB are plotted in Fig.4(f). Under
ΔB = 7 T, the derived ∆Tad exceeds ~20 K in the temperatures (T0) of 2-5 K, which is comparable
with those of commercial GGG and LiREF4(RE=Gd-Dy) coolants [31,32,41].

3.4 Test of magnetic cooling performance

The ADR cooling performance in sub-Kelvin temperatures was directly tested using a
homemade sample stage inserted into the PPMS, the initial cooling conditions were set to Ti = 2
7
K and Bi = 4 T or 6 T. From this starting point, magnetic field is swept down to the final field Bf = 0
T in a rate of 0.3 T/min. During this procedure, the accessible sample temperature T(B) is
recorded under a quasi-adiabatic condition for B // c-axis and ab-plane [see Fig. 5(a)]. From the
temperature-field trajectories, the isentropic T(B) curves at high field regimes (B ≥ 2 T) are nearly
overlapped along both field directions indicative of the isotropic field response in the FP state of
Gd3BWO9, and it is confirmed by comparing the T(B) data of polycrystals [ see Fig. S4]. At low
fields (B < 2 T), the T(B) curves display anisotropic behavior with two distinct minimums for B // c-
axis and three broad valleys for B // ab-plane [see the enlargement of T(B) data in the inset of Fig.
5(a)]. After the demagnetization process from Bi= 6 T, the realized lowest Tmin ~168 mK (B // c)
and Tmin ~227 mK (B // ab-plane) are lower than that of most Gd-based refrigerants as well as the
recently reported Eu-based compounds [18-21, 43, 47, 48]. More interestingly, the detected
temperature minimums in the adiabatic T(B) curves appear at the finite fields instead of Bf = 0 T.
Checking the location of fields (Bc1 ~ 0.27 T and Bc2 ~ 0.96 T for B // c, Bab,1 ~ 0.092 T and Bab,2 ~
0.31 T for B // ab-plane), we can find that they coincide with the ones extracted from the dM/dB
data. That means the minimal Tmin happen near the phase boundaries in the B-T phase diagrams
[ see Figs. 3(e, f)], signifying the presence of criticality-enhanced MCE as proposed in the low-
dimensional quantum magnets [52, 53]. Compared to B // ab-plane, the better cooling
performance near Bc2 for B // c-axis can be due to the presence of stronger spin fluctuations,
which can lead to the much larger entropy [Link] further quantify the MCE performance of
Gd3BWO9 at sub-Kelvin temperatures, we have measured the holding time (th) of sample after
ramping the field down to the final field Bf = 0 T. When the external field starts to decrease from
the initial condition (Ti = 2 K, Bi = 6 T), the time is defined to t = 0. The evolution of sample
temperature versus time T(t) is recorded, the results for both single crystals and polycrystals are
shown in Fig. 5(b). As seen, the holding time below ~250 mK is over an hour longer than the
recently reports in Gd-based frustrated magnets KBaGd(BO3)3 and NaGdP2O7 [18,21]. The low
attainable Tmin and long hold time of Gd3BWO9 let it as an excellent ADR material working in the
sub-Kelvin temperatures.
Besides the good cooling performance in sub-Kelvin temperatures, it is instructive to
characterize the MCE of Gd3BWO9 in extended temperature ranges for its application in
continuous ADR system [23, 40, 54]. Thus, we carried out the adiabatic MCE measurements at
given initial based temperature (To) under pulsed magnetic field up to ~32 T. Taking advantage of
fast field-sweep rate, this method is a nearly adiabatic condition [31,55]. Fig. 5(c) shows the
measured field dependence of sample temperatures (Ts) at various based temperatures, in which
the calculated Ts (open symbols) from the SM(T, B) data is also presented for comparison. The
reasonable agreement at fields B ≤ 7 T validate our MCE measurements under pulsed field. As
denoted by the arrow in Fig. 5(c), ∆Tad = 24-28 K is reached under ΔB= 10 T in temperature
intervals of T0 = 2-10 K, which is close to ∆Tad = 24-29 K of GGG in the comparable T0 and ΔB
[31], but Gd3BWO9 has more than 30% increase of magnetic entropy than that of GGG in this

8
temperature region.
For RE-based ADR refrigerants working at sub-Kelvin temperatures, the Yb- and Gd-based
magnets are relatively well studied. The former has the advantage on achieving lower Tmin down
below 20 mK, the latter possesses larger cooling capacity but works at higher temperatures
above 100 mK. Although pursuing the lowest possible Tmin is important, these two series of
compounds are basically suitable for the different stages in a continuous ADR system [56]. The
Yb-based materials are competitive in the first and second stages with temperature windows
below 0.05 K and 0.045- 0.3 K, while the Gd-based magnets are more suitable in the third and
fourth stages covering the temperature intervals of 0.25-1.1 K and 1.0-4.5 K. When the ADR
cooling temperature regions is limited to 0.25-4.5 K, Gd3BWO9 exhibits excellent cooling
performance among the representative Gd- and Eu-based refrigeration materials with Seff=7/2
reported to date [18, 21, 40-50], as listed in Table 1. Considering the importance of cooling power
per unit volume for practical application, the volumetric entropy density (SGS/vol, calculated by
dividing full entropy of magnetic ground state to unit cell volume) and the experimentally extracted
-ΔSm under field intervals (∆B = Bi-Bf) are used for the comparison. As seen from Table 1,
Gd3BWO9 has larger SGS/vol than the other Gd- and Eu-based compounds except Gd9.33[SiO4]6O2.
Moreover, the low minimum temperature Tmin ~168 mK enables it to fully cover the temperature
window of the third stage in continuous ADR system, in contrast to the other magnetically dense
materials with Tmin remaining above 250 mK. Among the materials with achievable Tmin <250 mK,
the values of SGS/vol are much smaller than that of Gd3BWO9. The above comparative results
highlight Gd3BWO9 as a superior magnetic refrigerant combining a quite large SGS/vol and a
relatively low Tmin.
To further evaluate the ADR cooling performance, we compare the extracted ∆SM in extended
temperature ranges (0.5-4 K) of Gd3BWO9 with other representative materials as listed in Table 2.
The ∆SM of Gd3BWO9 under ΔB = 7 T are comparable with the values of Gd9.33[SiO4]6O2 and far
exceeds those of other materials in temperature ranges of 2-4 K. Below 1 K, however, the ∆SM
values are lower than those of LiGdF4 and Gd9.33[SiO4]6O2 partially due to the reduced ordering
temperatures To~0.48 K in LiGdF4 [41] and the spin-disordered ground state in Gd9.33[SiO4]6O2
with the short-range spin correlation established at Tsr~0.52 K [25]. In this compound, the Gd3+

ions are randomly occupied on the frustrated one-dimensional spin chain with 2/3 occupation
ratio, thus both geometrical spin frustration and structural disorder effect are responsible for the
suppression of long-range magnetic order and the correspondingly enhanced sub-Kelvin MCE.
However, the presence of structural disorder complicates the understanding of how geometric
frustration alone regulates the MCE behavior. In this respect, Gd3BWO9 stands out as a disorder-
free frustrated system. The large differences in ionic radius of magnetic Gd3+ and nonmagnetic
B3+/W6+ cations prevent the antisite chemical disorder [26,27] in this material, making it a clean
system for examining the role of geometric frustration on optimizing the sub-Kelvin MCE.

9
 Besides the large SGS/vol, Gd3BWO9 exhibits a significantly low Tmin~168 mK compared to its
To~ 1.0 K. To highlight the large difference between Tmin and To of Gd3BWO9, we summarize the
comparative results on the reported sub-Kelvin Gd-and Eu- based ADR materials[18,22-25,
41,43,45,47], and plot the To and Tmin versus Nmag of materials in Fig. 6. Here, To denotes the
long-range AFM ordering (TN) or the onset of short-range magnetic order (Tsr), and Tmin is the
lowest attainable temperature after ADR process from initial condition of Ti = 2 K and Bi = 3- 12 T
[see the details in Table 1]. A general trend is that larger Nmag induces higher To with a roughly
linear dependence, which is reasonable due to the increased magnetic exchange and dipolar
interactions. On the other side, Tmin increase relatively slower with the variation of Nmag. Thus,
identifying materials with large Nmag and low To is challenging but essential for improving the sub-
Kelvin ADR cooling performance. Considering that the benchmark refrigerant GGG has
�GGG GGG GGG
��� =12.65 nm and �o =0.8 K [56], the identification of the materials with ���� > ���� and
-3

�0 < �GGG
o offers a promising route to surpass commercial GGG as the next-generation sub-Kelvin
ADR coolants. To satisfy the above criteria, the use of geometrically frustrated magnets (GFMs)
is effective to simultaneously increase Nmag and lower To because the To can be reduced or even
suppressed by spin frustration compared to the unfrustrated materials. For Gd3BWO9, its To = 1 K
is already reduced by spin frustration compared to its exchange energy scale ~3 K [see the
estimation based on magnetization data in Fig. S2], though being still higher than �GGG
o . Thus, the

realization of lower Tmin ~168 mK than �GGG

min ~322 mK in the same ADR cooling condition is quite

intriguing.
A detailed analysis of magnetic entropy SM(T,B) landscape is crucial for understanding the
mechanism behind the exceptionally low Tmin in Gd3BWO9. In the ADR runs, the coolants
experience a quasi-isentropic process from the initial condition (Ti , Bi) to the final condition (Tf, Bf).
Once the Bi and Bf are fixed, ΔTad = Ti-Tf is fully determined by the entropy profiles of SM(T, Bi)
and SM(T, Bf) in the SM(T, B) curves [see Fig. 4(c)]. Since a larger ΔTad= Ti-Tf corresponds to a
lower Tmin, two approaches can be employed to reduce Tmin. One is to realize a smaller value of
entropy SM(Ti = 2K, B = Bi) at the starting point, the other is to retain a large entropy at the lower
temperatures, which will shift the Tf of endpoint to lower temperature in the SM(T,B) phase
diagram. For the former approach, the weak mutual interactions among Gd3+ ions in Gd3BWO9
lead to the fast shift of CM(T) curves in field to high temperatures and correspondingly small value
of SM(Ti, Bi). This is because that the movement of Schottky-like CM(T) feature roughly follows
gSeffµB(B-Bs) [17,18]. At the initial cooling condition of Ti = 2K and Bi = 7T, the integrated SM(Ti,
Bi)~0.035Rln8 of Gd3BWO9 is smaller than ~0.052Rln8 of Gd9.33[SiO4]6O2 [25] and 0.069Rln8 of
NaGdP2O7[21]. For the latter approach, by setting Bf = 0 T, since the attainable Tmin of Gd3BWO9
is well below TN, the value of Tmin is mainly determined by the zero-field CM(T) data. From this
viewpoint, the slow drop of CM∝T2 (T < 0.37 K) rather than CM∝T3 is favourable for the
accumulation of SM below Tf [ see Fig. 2(a)]. Thus, not only the lower To but also the slower decay
of CM/T below To can facilitate a large accumulation of SM(Tf, Bf), leading to a lower Tmin under Bf
10
= 0 T.
Tuning materials toward a quantum critical point can generate an additional enhancement of
spin fluctuations, leading to a marked temperature decrease under the adiabatic condition.
Checking the T-dependence of CM/T at different fields [see the inset of Fig. 4(a)], we can find that
the values of CM/T at 1 T are larger than those measured at fields below or above 1T in low
temperature regions (T < T*). This means the integrated entropy SM(T, B=Bf) reaches maxima if
setting Bf = 1 T. This can explain the lowest Tmin~168 mK (B // c-axis) is realized near the critical
field Bc2 ~0.96 T in the ADR cooling performance test [see Fig. 5(a)]. Thus, the lowest Tmin~168
mK realized at Bf = Bc2 rather than Tmin~232 mK at Bf = 0 T is facilitated by the criticality-enhanced
MCE. Compared to the zero-field spin fluctuation originating from geometric spin frustration of
Kagome lattice, the increased spin fluctuations near quantum critical point induce an additional
accumulation of entropy responsible for the ~26% decrease from Tmin~227 mK at Bf = 0 T. This
mechanism, known as criticality-enhanced MCE, has been proposed for low dimensional
frustrated magnets, especially the novel spin disordered system as a mean to improve the cooling
performance, as realized in the spin supersolid candidate Na2BaCo(PO4)2 with a quite low Tmin
~94 mK that is attained near its quantum critical point (QCP)[13]. The present cooling
performance results of Gd3BWO9 reveal that the criticality-enhanced MCE can also be used to
optimize the cooling performance of Gd-based frustrated antiferromagnets.

4. Conclusions

In summary, we have reported the anisotropic magnetic phase diagram and magnetic cooling
performance at sub-Kelvin temperatures of Kagome magnet Gd3BWO9. This material
demonstrates an excellent MCE performance in sub-Kelvin regime, featured by large volumetric
magnetic entropy in a wide temperature window from ~0.25-4 K, remarkable low Tmin~168 mK
and long holding time than an hour following the ADR process. Compared to most of dense Gd-
based antiferromagnets, the attainable quite low Tmin~168 mK in Gd3BWO9 is beneficial from the
enhancement of spin fluctuations arising from the geometric spin frustration and field-tuned
criticality effect, which together optimize the MCE at sub-Kelvin temperatures. Our work thus
provides a realistic example of utilizing field-tunable criticality to improve the ADR cooling
performance in frustrated magnets.

Acknowledgements
The work is supported by the National Key Research and Development Program (Grant No.
2024YFA1611200 & 2023YFA1406500). This work was supported by the National Natural
Science Foundation of China (Grant No. 12141002 and 52088101) and the Strategic Priority
Research Program of the Chinese Academy of Sciences (XDB1270000). A portion of this work
was carried out at the synergetic extreme condition user facility (SECUF). We would like to thank
the staff of the analysis center of Huazhong University of Science and Technology for their

11
assistance in structural characterizations.

Author Declarations
The authors have no conflicts to disclose.

Table 1. Comparison of key parameters for Gd- and Eu-based ADR materials at subKelvin
temperatures: To is the magnetic ordering temperature and short range order is labeled by (*),
Tmin is the attainable minimum temperature after the ADR process from the initial condition of (Ti,
Bi), Nmag is the density of magnetic ions, SGS is the entropy of the ground-state multiplet, and
SGS/vol is the volumetric magnetic entropy density, and -∆SM is the magnetic entropy change in 1
K and ∆B = 6 T or 7 T.
-∆SM
SGS/vol.
Tmin Ti,Bi Nmag [mJ/(K
Materials To (K) SGS [mJ/(K References
(mK) (K, T) (nm-3) cm3)]
cm3)]

Gd3BWO9 1 168 2,6 17.49 Rln8 502.2 303 This work

Gd3Ga5O12 0.8 (*) 322 2,6 12.65 Rln8 363.2 274 [22]

KBaGd(BO3)2 0.263 122 2,5 6.46 Rln8 185.8 149 [18]

Gd2Ti2O7 1.0 480 2,12 15.14 Rln8 435 115 [24]

K2GdNb5O15 0.31 175 2,5 3.29 Rln8 93.2 74 [44]

GdPO4 0.77 325 2,7 14.47 Rln8 415.6 290 [43]

Gd(HCOO)3 0.8 356 2,3 7.94 Rln8 228.1 192 [42]

Gd9.33[SiO4]6O2 0.5(*) 210 2,7 17.74 Rln8 509.4 344 [25]

NaGdP2O7 0.57 220 2,5 7.3 Rln8 210 93.2 [21]

[41]
LiGdF4 0.48 <500 2,5 13.36 Rln8 383.8 316

EuCl2 1.3 346 2,5 13.15 Rln8 377.1 151 [47]

EuB4O7 068 289 2,4 9.911 Rln8 284.5 186 [48]

Table 2. Comparison of To,Tmin, SGS/vol and -∆SM under different ∆B=Bf-B0 for the representative
Gd- and Eu-based ADR materials in selected temperature ranges from 0.5 K to 4 K.

-∆SM
To Tmin SGS/vol.
Materials [mJ/(K cm3)] References
(K) (mK) [mJ/(K cm )] 3

∆B (T) 0.5 K 1K 2K 4K
This work
Gd3BWO9 1 170 502.2 7 110 303 438 408

12
 [25]
Gd9.33[SiO4]6O2 0.5 300 509.4 7 183 344 440 391

LiGdF4 0.48 <500 383.8 6 - 316 348 290 [41]

[22]
Gd3Ga5O12 0.8(*) 322 363.2 6 108 274 318 260
[45]
Gd(OH)CO3 0.7 508 394.7 7 106 280 348 317

Gd2Ti2O7 1.0 480 435 7 24.9 115 190 184 [24]

GdPO4 0.77 385 415.6 7 62 290 375 332 [43]

KBaGd(BO3)2 0.263 122 185.8 5 118 149 143 130 [18]

EuCl2 1.3 346 377.1 5 74 151 369 270 [47]

Fig. 1(a) The Kagome lattice formed by Gd3+ ions within the ab-plane and its stacking along the c-
axis in Gd3BWO9. (b) Picture of a typical Gd3BWO9 single crystal. (c) Isothermal magnetization
M(B) curves of Gd3BWO9 at 1.8 K with B // c axis and B // ab-plane, respectively. Inset shows the
definition of axes. (d) Temperature dependence of magnetic susceptibility  (T) and inverse
magnetic susceptibility 1/(T) under B = 0.1 T with B // c axis and B // ab-plane, respectively.

13
Fig. 2 (a) Zero field specific heat Cp(T) curves of Gd3BWO9 and its isostructural analog Eu3BWO9
for lattice contribution to Cp(T), inset shows the power law fit on the CM/T versus T. (b) Magnetic
specific heat CM(T) and entropy SM(T) data of Gd3BWO9.

Fig. 3 (a, b) The M(B) and dM/dB curves of Gd3BWO9 at 0.4 K with B // c axis and B // ab-plane,
respectively. (c, d) The dM/dB curves at different temperatures (0.4 K ≤ T ≤ 1.8 K) along c-axis
and b*-axis. (e, f) The B-T magnetic phase diagrams constructed by the dM/dB data for B along c
axis and b* axis, the stars denote the multi-critical points in the phase diagram.

14
Fig. 4 (a) Magnetic specific heat CM(T) curves of Gd3BWO9 at low magnetic fields (B≤ 1.5 T) for B
// c, inset shows the CM/T data plotted versus T. (b) Magnetic specific heat CM(T) curves of
Gd3BWO9 at selected magnetic fields. (c) The resultant magnetic entropy SM(B,T) under different
magnetic fields of Gd3BWO9, the arrows denote the isothermal suppression of entropy −SM and
adiabatic temperature change  Tad in the ADR processes. (d) Temperature dependence of
−SM(T) at different field changes (B). (e) The final temperature Tf versus initial temperature Ti
cooled from different initial field calculated from the SM(B,T) results. (f) The Tt versus T0 at
different B in an ADR processes.

Fig. 5 (a) The quasi-adiabatic demagnetization refrigeration test of Gd3BWO9 single crystals (0.7
g) started from two different initial cooling conditions (Ti = 2 K, Bi = 4 T) and (Ti = 2 K, Bi = 6 T) for
B // c axis and B // ab-plane, (b) Comparison of the holding time of Gd3BWO9 single crystals and
polycrystals under same initial condition Ti = 2 K, Bi = 6 T, (c) Field dependence of temperature
change Tt versus T0 of Gd3BWO9 single crystals measured with pulsed magnetic fields up to 32 T
applied along c-axis, the open symbols show the deduced Tt from the specific heat data, the
arrow corresponds to Tad (0→ 10 T).
15
Fig. 6 Comparison of magnetic ordering temperature (To) and minimum cooling temperature (Tmin)
versus magnetic ion density (Nmag) on the representative Gd and Eu-based magnetocaloric
materials.

References

[1] Debye P 1926 Ann. Phys. 386 1154

[2] Ambler E, and Hudson R P 1955 Rep. Prog. Phys. 18, 251
[3] Fruneau M, Lacaze A, Weil L 1967 Cryogenics. 7 135
[4] Wikus P, Canavan E, Heine S T, Matsumoto K, Numazawa T 2014 Cryogenics. 62 150
[5] P. J. Shirron 2014 Cryogenics. 62 130
[6] Shirron P J, KimballM O, Wegel D C, Canavan E R, DiPirroM J 2010 SPIE Astron. Telesc. Inst.:
Ultraviolet to Gamma Ray. 7732 273212
[7] Hagmann C, Richards P L. A 1995 Cryogenics. 35 303-309
[8] Fisher R A, Hornung E W, Brodale G E, Giauque W F 1973 J. Chem. Phys. 58 5584
[9] Fritz J J, Clark J T, Cesarano J 1975 J. Chem. Phys. 62 200
[10] Zhitomirsky M E 2003 Phys. Rev. B 67 104421
[11] Tokiwa Y, Bachus S, Kavita K, Jesche A, Tsirlin A A, Gegenwart P 2021 Commun. Mater. 2 1
[12] Liu X Y, Gao Y, Li H, Jin W, Xiang J, Jin H, Chen Z, Li W, Su G 2022 Commun. Phys. 5 1
[13] Xiang J, Zhang C, Gao Y, Schmidt W, Schmalzl K, Wang C W, Li B, Xi N, Liu X Y, Jin H, Li G,
Shen J, Chen Z, Qi Y, Wan Y, Jin W, Li W, Sun P, Su G 2024 Nature 625 270

16
[14] Brasiliano D A P, Duval J M, Marin C, Bichaud E, Brison J P, Zhitomirsky M, Luchier N
2020 Cryogenics. 105 103002
[15] Koskelo E C, Mukherjee P, Liu C, Sackville Hamilton A C, Ong H S, Castelnovo C,
Zhitomirsky M E, DuttonS E 2023 PRX. Energy 2 033005
[16] Treu T, Klinger M, Oefele N,Telang P., Jesche A, and Gegenwart P 2025 J. Phys.: Condens.
Matter. 37 013001
[17] Liu A, Zhou J, Wang L, Cao Y, Song F, Han Y, Li J, Tong W, Xia Z, Ouyang Z, Zhao J, Guo H,
Tian Z 2024 Phys. Rev. B 110 144445
[18] Jesche A, Winterhalter-Stocker N, Hirschberger F, Bellon A, Bachus S, Tokiwa Y, Tsirlin A A,
Gegenwart P 2023 Phys. Rev. B 107 104402
[19] Xiang J, Su C, Xi N, Fu Z, Chen Z, Jin H, Chen Z, Mo Z, Qi Y, Shen J, Zhang L, Jin W, Li W,
Sun P, Su G, arXiv:2301.03571.
[20] Arjun U, Ranjith K M, Jesche A, Hirschberger F, Sarma D D, Gegenwart P 2023 Phys. Rev.
Appl. 20 014013
[21] Telang P, Treu T, Klinger M, Tsirlin A A, Gegenwart P, and Jesche A 2025 Phys. Rev. B 111
064431
[22] Daudin B, Lagnier R, Salce B 1982 J. Mag. Mag. Mater. 27 315-322
[23] Wikus P, Burghart G, Figueroa-Feliciano E 2011 Cryogenics. 51 555-558
[24] Sosin S S, Prozorova L A, Smirnov A I, Golov A I, Berkutov I B, Petrenko O A, Balakrishnan
G, Zhitomirsky M E 2005 Phys. Rev. B 71 094413
[25] Yang Z W, Zhang J, Liu B, Zhang X, Lu D, Zhao H, Pi M, Cui H, Zeng Y J,Pan Z, Shen Y, Li S,
and Long Y 2024 Adv. Sci. 2306842
[26] Ashtar M,Guo J, Wan Z, Wang Y,Gong G,Liu Y, Su Y, and Tian Z Inorg. Chem. 2020 59 5368
[27] Song F Y, Ge H, Liu A D, Qin Y Q, Han Y Y, Ling L S, Yuan S L, Ouyang Z W, Sheng J M, Wu
L S, Tian Z M, 2023 Phys. Rev. B 108 214410
[28] Majchrowski A, Michalski E, and Brenier A. 2003 J. Crystal. Growth. 247 467
[29] See Supplemental Material for the crystal structure refinements, low-T magnetization and the
MCE results of Gd3BWO9.
[30] Zhang C, Xiang J, Zhu Q, Wu L, Zhang S, Xu J, Yin W, Sun P, Li W, Su G, Jin W 2024 Phys.
Rev. Mater. 8 044409
[31] Kohama Y, Marcenat C, Klein T, Jaime M 2010 Rev. Sci. Instrum. 81 104902
[32] Liu P, Yuan D, Dong C,Lin G, Vílora E G, Qi J, Zhao X, Shimamura K, Ma J,Wang J, Zhang Z,
Li B 2023 NPG Asia Materials. 15 41
[33] Sosin S S, Prozorova L A, Bonville P, Zhitomirsky M E 2009 Phys. Rev. B 79 014419
[34] Hallas A M,Arevalo-Lopez A M,Sharma A Z, Munsie T, Attfield J P, Wiebe C R, Luke G M,
2015 Phys. Rev. B 91 104417
[35] Quilliam J A, Meng S,Craig H A, Corruccini L R, Balarkrishnan G, Petrenko O A, Gomez A,
Kycia S W, Gingras M J P, Kycia J B 2013 Phys. Rev. B 87 174421

17
[36] Flavian D, Nagl J,Hayashida S,Yan M,Zaharko O, Fennell T, Khalyavin D, Yan Z, Gvasaliya S,
and Zheludev A 2023 Phys. Rev. B 107 174406
[37] Nagl J, Flavián D, Duncan B, Hayashida S,Zaharko O,Ressouche E, Ollivier J, Yan Z,
Gvasaliya S, Zheludev A 2025 Phys. Rev. B 111, L180406
[38] Zelenskiy A, Quilliam J A, Shapiro A Ya, Quirion G 2021 Phys. Rev. B 103, 224422
[39] Goto M, Ueda H, Michioka C, Matsuo A, Kindo K, Sugawara K, Kobayashi S, Katayama N,
Sawa H, Yoshimura K 2018 Phys. Rev. B 97, 224421
[40] Barclay J A, Steyert W A 1982 Cryogenics. 22 73-80
[41] Numazawa T, Kamiya K, Shirron P, Dipirro M, Matsumoto K 2006 AIP Conf. Proc. 850 1579-
1580
[42] Lorusso G,Sharples J W, Palacios E, Roubeau O, Brechin E K, Sessoli R, Rossin A,Tuna F,
McInnes E J L, Collison D, and Evangelisti M 2013 Adv. Mater. 25 4653−4656
[43] Palacios E, Rodrıguez-Velamazán J A, Evangelisti M, McIntyre G J, Lorusso G, Visser D, De
Jongh L J, and Boatner L A 2014 Phys. Rev. B 90 214423
[44] Zeng Q, Ge H, Wu M, Ruan S, Li T, Wang Z, Li J, Ling L, Tong W, Huang S, Liu A,Zhou J,
Xia Z, Sheng J, Wu L, and Tian Z, 2024 Chem. Mater. 36 2867
[45] Chen Y C, Qin L, Meng Z S, Yang D F,Wu C, Fu Z, Zheng Y,Liu J L, Tarasenko R,
Orendáč M, Prokleška J, Sechovský V and Tong M L 2014 J. Mater. Chem. A 2 9851
[46] Yang Z, Zhang H, Bai M, Li W, Huang S, Ruan S, Zeng Y 2020 J. Mater. Chem. C. 8 11866-
11873
[47] Wang B, Liu X, Hu F, Wang J, Xiang J, Sun P, Wang J,Sun J, Zhao T, Mo Z,Shen J,Chen Y,
Huang Q, and Shen B 2024 J. Am. Chem. Soc. 146 35016−35022
[48] Wang Y, Xiang J,Zhang L, Gong J, Li W, Mo Z, and Shen J 2024 J. Am. Chem. Soc. 146
3315−3322
[49] Xu Q F, Chen M T, Wu R T,Long L S, Zheng L S 2024 J. Am. Chem. Soc. 146 20116−20121
[50] Shirron P J, Kimball M O, James B L, Wegel D C, Martinez R M, Faulkner R L, Neubauer L,
Sansebastian M 2012 Cryogenics. 52 165-171
[51] Cao HS 2021 J. Low. Temp. Phys. 204 175
[52] Wolf B, Tsui Y, Jaiswal-Nagar D, Tutsch U,Honecker A, Remović-Langer K,Hofmann G,
Prokofiev A, Assmus W, Donath G,Lang M 2021 PNAS. 204,175
[53] Li T, Liu X, Gao Y, Jin H, He J, Sheng X, Jin W,Chen Z, Li W 2021 Phys. Rev. Res. 3 033094
[54] Shirron P, Wegel D, Dipirro M 2006 AIP Conf. Proc. 850 1573-1574
[55] Kihara T, Kohama Y, Hashimoto Y, Katsumoto S, Tokunaga M 2013 Rev. Sci. Instrum. 84
074901
[56] Schiffer P, Ramirez A P, Huse D A, Valentino A J 1994 Phys. Rev. Lett. 73 2500-3503

18