Thermomagnetic liquid metal switches with fast bidirectional response

Article https://doi.org/10.1038/s41467-025-58015-x Thermomagnetic liquid metal switches with fast bidirectional response Haodong Chen 1 ,YuyuXie 1 ,KaimingQiao 1 , Longlong Xie 1 ,ZiyuanYu 1 , Chenyu Xu 1 , Jingyi Liu 1 , Mingze Liu 1 , Yongyu Lu 2 , Jing Liu 2 ,FengxiaHu 3 ,BaogenShen 3 , Ke Chu 4 ,JiaYanLaw 5 ,VictorinoFranco 5 & Hu Zhang 1 Traditional thermal switches usually suffer from long-time and unidirectional response. Herein, we combined liquid metals with magnetic Ni 2 Mn 1.4 In 0.6 particles and developed a thermomagnetic liquid metal. These materials not only show excellent fluidity and electrical conductivity, but also exhibit fast response at a tuneable temperature that traditional magnetic liquid metals do not have. The ultimate application is designed as micro-channel thermal switch. Particularly, our thermal switch features bidirectional response through the droplet displacement, thus simultaneously cutting off the working circuit while turning on the fire extinguishing. Its response time of 1.2 s is 3.3 −5.6 times faster than typical commercial thermal switches under the same hot source temperature of 75 °C, and it can be further reduced to 660 ms under the optimal environment. Moreover, this fast-response thermal switch offers the fastest recovery time, low cost, and long-cycle stability, showing a huge potential as a generation of thermal switches for diverse applications. Localized overheating of electric circuits can be seen almost everywhere in daily life and industrial production (Fig. 1a). However, localized temperature buildups can often go unnoticed until they escalate and transform into high-temperature environments or large-scale flames1, leading to devastating consequences. In such cases, the rapid spread of flames hinders timely control and extinguishment, resulting in significant casualties and irreparable property losses2. Analysis of firestatistics over the past decade found that electrical fires accounted for the highest proportion of all fire types (see Fig. 1b), and have remainedat about 30% for long time, posing a serious threat to human life and property safety as well as ecological environment3. Rapid detection and response to electrical fire hazards, especially in the critical few seconds after ignition, are essential for minimizing damage and saving lives. Even a few seconds’delay in detection can substantially increase fire size, smoke inhalation, and property damage4. However, most commercial thermal switches and smoke detectors respond within several seconds to over 100 seconds, this highlights the urgent need for faster thermal switches to enable early detection and rapid suppression, thus preventing loss of life and property. Thermal switches, serving as a pivotal thermal response control element, offer a crucial and widespread solution in many scenarios, such as equipment protection and fire suppression4. By promptly triggering (switch ON/OFF) in response to elevated temperatures, thermal switches help mitigate the risk of overheating hazards and issue timely warnings of overheating. Generally, thermal switches can be classified into three categories based on their working principles: gas/smoke sensors5,6, resistance-shift switches7,andshape-change switches8,9. The gas/smoke sensors are the most common thermal switches used in many practical applications of fire warning5,which can not only trigger the sprinkler system but also alert the fire department. However, they are prone to false alarms caused by nonfire-related smoke (e.g., cooking or dust), leading to unnecessary disruptions and desensitization to alarms. Moreover, their response time usually takes longer than 100 s, limited by the relatively low Received: 18 September 2024 Accepted: 10 March 2025 Check for updates 1 School of Materials Science and Engineering, University of Science and Technology Beijing, Beijing 100083, P R China. 2 Key Laboratory of Cryogenic Science and Technology, Technical Institute of Physics and Chemistry, Chinese Academy of Sciences, Beijing 100190, P R China. 3 Beijing National Laboratory for Condensed Matter Physics, Institute of Physics, Chinese Academy of Sciences, Beijing 100190, P R China. 4 School of Materials Science and Engineering, Lanzhou Jiaotong University, Lanzhou 730070, P R China. 5 Multidisciplinary Unit for Energy Science, Dpto. Física de la Materia Condensada, ICMS-CSIC, Universidad de Sevilla, P.O. Box 1065, 41080 Sevilla, Spain. e-mail: [email protected] Nature Communications | (2025) 16:2634 1 1234567890():,; 1234567890():,; sensitivity6, resulting in a failure to provide a fast response to incipient fires. In contrast, the resistance-shift switches undergo a transition from insulator to conductor, significantly increasing electrical conductivity, thus boasting a relatively fast response time of ~8 s with the “sandwich-like”structure7, or even 0.25 s for an amino-functionalized carbon nanotube (CNT) aerogel10. Unfortunately, they usually suffer from unidirectional response, non-linear resistance changes, and poor long-term stability, which limits their practical application in complex circuits. On the contrary, shape-change switches can effectively disconnect one circuit while simultaneously connecting another, thanks to their shape-change abilities in response to temperature variations8. However, the low thermal conductivity and recovery stress of their working materials cause slow response times, which can extend to as long as 5 min11. In addition, due to the mechanical wear and fatigue issues, they also suffer from unstable contact resistance and limited cycle life12. Therefore, it is crucial to develop a thermal switch that can effectively address the limitations of the current traditional thermal switches. Recently, liquid metals (LMs) have been widely reported to have many properties, including high electrical and thermal conductivity13,14,highdeformability 15,16, andlowtoxicity17,18.Inaddition, researchers have added Fe19–21,ironoxides 22,andNdFeB 15,23 particles into LMs to obtain magnetic liquid metals (MLMs), which can be positioned by the attraction of a magnet. The MLMs can be driven by moving the magnet to function as magnetic switches21,24. However, these magneticswitches cannot be used asthermal switches due to the lack of thermosensitive property. Ni 2 MnZ(Z=Al,GaandIn)Heusler alloys have attracted much attention due to their interesting thermalinduced magnetic and martensitic structural transitions25–27. Therefore, it can be expected to improve the thermosensitive property of MLMs by utilizing the thermomagnetic phase transition of Ni 2 MnZ Heusler alloys, thus preparing a type of thermomagnetic liquid metal switches. In this work, we chose Ni 2 Mn 1.4 In 0.6 (Ni-Mn-In) Heusler alloy with a ferromagnetic transition as thermosensitive magnetic material, and introduced its particles into LMs with oxidized mixing method to prepare a thermomagnetic liquid metal (TMLM). This TMLMs have good fluidity, appropriate viscosity, and high electrical conductivity. In addition, this TMLMs not only can be driven by a magnetic field like traditional MLMs, but also exhibits fast temperature sensitive characteristics that traditional MLMsdo not possess. Although a few works have been reported on mixing magnetocaloric materials with Galinstan28–30, they primarily focused on the magnetocaloric effect. Unlike those works, based on the special advantages of TMLMs, we designed a micro-channel thermal switch by using this TMLMs. As the TMLM thermal switch offers fast bidirectional response, it can be hypothesized as an ideal solution for promptly suppressing circuit fires. Unlike traditional circuit safety switches, which disconnect the circuit only after it overheats (but do not activate the fireextinguishing equipment), the TMLM thermal switch can mitigate the fire promptly by quickly activating the fire extinguishing equipment in a simple way. This approach would help prevent property damage and loss of life by addressing circuit fires more effectively. The concept of TMLM potentially addressing circuit fires is schematically illustrated in Fig. 1c, where the TMLM thermal switch is positioned on a socket. The upper end of the thermal switch is connected to the electrical circuit within the socket, while the lower end is connected to a fire extinguishing circuit. In the event of the electrical circuit catching fire, the TMLM thermal switch on the socket will deactivate the electrical circuit, concurrently, activating the fire extinguishing circuit through the release (falling) of the conductive TMLM droplet. This rapid response will promptly and effectively extinguish the fire. The temperature response time of this TMLM thermal switch is 3.3 −5.6 times faster than those of other typical commercial thermal switches. Moreover, it can recover quickly for long-term recycling. These advantages suggest that this TMLM has great potential for application in thermal switches, especially in the field of microcircuit thermal management. Results Synthesis, morphology, and structure of TMLMs The TMLMs with Ni-Mn-In particles are synthesized according to the schematic presentation in Fig. 2a. Firstly, Ni-Mn-In ingot was prepared Working mechanism of TMLM thermal switch for fire extinguishing Response 1: Cut off the circuit Response 2: Put out the fire TMLM thermal switch 2 cm TMLM thermal switch (initial state) Electrical circuit Fire extinguishing circuit Response 1 (on fire) Response 2 (on fire) a Thermomagnetic liquid metal (TMLM) 31.45% 19.20% 6.38% 3.86% 3.93% 3.45% 23.18% 6.81% 1.60% 27.6℃ 34.4℃ 28.9℃ 44.2℃ c Electrical fire Use fire carelessly Smoking Play with fire Production operation Spontaneous combustion Set on fire Lightning strike Unknown reason Others N SMagnet N S N S OFF ON OFF OFF N S ON OFF Socket Fig. 1 | The huge proportion of electrical fires, as well as the concept and advantages of the TMLM thermal switch for fire extinguishing. a Infrared photos of localized overheating of electric circuits. bThe average proportion of various fire types in the past decade. cSchematic illustration and working principle of TMLM thermal switch with fast bidirectional response for fire extinguishing. (Elements inFig. 1c, createdwith freepik.com,released under a license. Source data are provided as a Source Data file). Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 2 by induction-melting from elemental raw materials, followed by annealing. It was then crushed into particles and sieved, obtaining an average diameter of ~37.88 μm and with a uniform composition (Supplementary Figs. 1 and 2). An appropriate amount of Ni-Mn-In particles (5, 10, 15,and 20 wt.%) were introduced into Galinstan LM and vigorously stirred for about 15 −30min to achieve complete homogeneity. During the mixing process, the oxide layer on the LM surface is continuously broken to create new oxides, which will coat Ni-Mn-In particles. Because this new oxide coating is easilymiscible with LM, the Ni-Mn-In particles were successfully incorporated into the LM. Finally, sonication treatment was used to make the Ni-Mn-In particles evenly distributed in the LM. Figure 2b shows the photographs of the four TMLMs containing different load of Ni-Mn-In magnetic particles, where all maintain a metallic luster and bead shape at room temperature (RT) without exhibiting any curing behavior even for the higher packing fractions. This indicates that they all have good fluidity. Besides, their SEM images show that the Ni-Mn-In particles are uniformly distributed without obvious agglomeration in the twodimensional plane (Supplementary Fig. 3). An in-situ 3D micro-CT analysis of the TMLM-20% sample reveals the distribution of its different components in three-dimensional space (Fig. 2c): the Ni-Mn-In particles (marked in blue) are evenly dispersed in the LM (red region) without noticeable agglomeration or sedimentation, indicating that the TMLM exhibits both uniform and stable thermomagnetic particle distribution. Furthermore, their refined results show that the LM, Ni-Mn-In particles and pores have volume fractions of 83.01%, 12.83%, and 4.16%, respectively. These are in good agreement with the theoretical volume fractions. The robust Ni-Mn-In (12.63%) Pores (1.94%) LM (85.43%) Theoretical volume ratio: LM (84.05%); Ni-Mn-In particles (15.95%) f e TMLM droplets g Synthesis of TMLM droplets Micro-CT of TMLM-20% Pressure Grinding Stirring Mixing Sonication Crushing Sieving LM Ni-Mn-In particles Ni-Mn-In ingot LM  Oxides Ni-Mn-In particles TMLM Oxidized mixing TMLM-5% TMLM-10% TMLM-15% TMLM-20% 20 30 40 50 60 70 80 90 Intensity (arb. units) Pure LM Pure Ni-Mn-In TMLM-60% Angle ( 2θ ) d (111) (200) (220) (222) (400) (422) Ni2MnIn (PDF#40-1208) LM Ni-Mn-In 100 nm LM Ni-Mn-In 5 nm A d(200) = 3.304 Å Ni-Mn-In Zone = [100] 5 1/nm (022) (020) (002) (000) ba c Fig. 2 | Synthesis, morphology, and structure of the TMLM droplets. a Synthesis of the TMLM droplets. bPhotographs of TMLM droplets with different packing fractions of 5, 10, 15, and 20wt.% (Denoted as TMLM-x%, x% refers to the mass fraction of Ni-Mn-In particles in TMLM droplets). cMicro-computed tomography (micro-CT) of TMLM-20%, where the 3D images show the distribution of lowdensity liquid metal phase (red), high-density Ni-Mn-In particles (blue), and pores (gray), respectively. dBright-field image at room temperature for an area with coexisting LM and Ni-Mn-In particle in TMLM-20%. eHigh-resolution transmission electron microscopy (HRTEM) image for coexisting LM and Ni-Mn-In particle in TMLM-20%. fSelected area diffraction pattern (SADP) for the cubic L2 1 austenite along [100] zone axis, corresponding to the dashed frame A in panel (e). gXRD characterization of pure LM, pure Ni-Mn-In, and TMLM-60% (left for 21 days). Blue and yellow symbols represent the characteristic peaks of LM and Ni-Mn-In, respectively. (Source data are provided as a Source Data file). Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 3 interfacial bonding between the LM and Ni-Mn-In is observed in the high-resolution transmission electron microscopy (HRTEM) investigationswhereitsmicrographpresentsperfectlycoatedNi-Mn-Inparticles by the LM in the nanoscale (Fig. 2d and Supplementary Fig. 4). It is also revealed that the Ni-Mn-In phase exhibits evenly arranged atomic layers with an interplanar spacing of 3.304 Å, which corresponds to the austenitic phase (Fig. 2e). Its selected area diffraction pattern (SADP) image further confirms the austenite characteristic of the cubic L2 1 Heusler structure with a zone axis of [100] (Fig. 2f). The LM region, on the other hand, has an amorphous structure where an obvious phase boundary is observed between the LM and Ni-Mn-In regions. No impurity phase is observed near the phase boundary, suggesting the absence of alloying reactions between the LM and NiMn-In particles, a phenomenon typically associated with LM and doped particles. Such reactions have the potential to compromise the functional properties of the doped particles. The collective observations, in conjunction with the low porosity volume, not only indicate a robust interfacial bonding between the LM and Ni-Mn-In particles but also establish that the entire TMLM will exhibit a high degree of responsiveness when subjected to a magnet, thanks to the good dispersion of the thermomagnetic particles. In order to investigate the possibility of alloying reactions between liquid metals and Ni-Mn-In particles, a significantly higher load, 60wt.% Ni-Mn-In particles was used, as that would facilitate the detection of this effect. After leaving the TMLM-60% sample for 21 days, X-ray diffraction experiments were performed to compare their resulting phase with constituents, pure LM and Ni-Mn-In particles, as showninFig.2g. Pure LM shows a broad diffraction halo around 26.2° −47.6°, corresponding to the typical characteristic of an amorphous phase, which agrees with TEM observations. The Ni-Mn-In diffraction pattern shows a single austenitic phase with L2 1 structure (space group Fm3m) at RT, with lattice parameter a= 6.024 Å. As for the TMLM-60%, it exhibits both characteristic behavior of Ni-Mn-In (diffraction peaks) and amorphous LM (broad halo) with no other characteristic peaks present. This re-affirms that no alloying behavior occurs between LM and Ni-Mn-In particles even after a long time, proving the long-term stability of TMLMs. Characterization and magnetic actuation performance of TMLMs Figure 3a shows the elastic modulus G’and viscous modulus G’’ as a function of shear stress for TMLMs. At low shear stress, up to ~70 Pa, all the TMLMs maintain elastic-dominant behavior in the linear viscoelastic region where G’>G’’. Shear stress causes the decrease of G’while G’’ gradually increases, and they eventually intersect at the critical shear stress τ 0 , implying that the fluidity of TMLMs begins to surpass its elasticity. It is noted that τ 0 only increases slightly with the increase of packing fraction, indicating that the TMLMs maintain as good fluidity as the pure LM. Furthermore, the contact angles of TMLMs show a slight increase with higher packing fractions, as depicted in Fig. 3b, consistent with the rising trend of apparent viscosity with doping fractions (Supplementary Fig. 5). This arises as the increased apparent viscosity induces a decrease in surface tension, thus leading to an increase in contact angle. Moreover, since the TMLM still maintains the bead shape of a LM, it indicates that the TMLM shows similar fluidity to pure LM, despite having a higher apparent viscosity. These collective findings signify that the current TMLMs demonstrate excellent fluidity and viscosity comparable to pure LM. Figure 3c compares the electrical conductivities of pure LM and TMLMs with different thermomagnetic filler concentrations. Although Ni-Mn-In has a relatively low electrical conductivity (0.71 × 106Sm −1)31 compared to pure LM (2.12 × 106Sm −1at RT), incorporation of the NiMn-Inparticles intotheLMresultsinonly a minordecreaseinelectrical conductivity, with the overall electrical conductivity remaining relatively high at ~2.0 × 106Sm −1. Even when the filler content reaches 60wt.%, it still remains as high as 1.91 × 106Sm −1.Thehighfluidity of LM allows it to maintain a dense and continuous structure even after doping with a high particle fraction. As a result, the low-resistance conductive path will hardly be changed by the doped high-resistance particles, making the electrical conductivity of TMLMs remain high, which is advantageous for their effective performance in circuit control applications. Figure 3d shows the temperature (T) dependence of zero-fieldcooling (ZFC) and field-cooling (FC) magnetization (M)under0.4Tfor pure Ni 2 Mn 1.4 In 0.6 alloy. The Ni 2 Mn 1.4 In 0.6 alloy experiences a typical first-order martensitic transition (FOMT) from ferromagnetic (FM) austenite to weak-magnetic martensite around 0 ~ −20°C with distinct thermal hysteresis. In addition, another second-order magnetic transition (SOMT) of austenite from FM to paramagnetic (PM) states appears at Curie temperature (T C ) of 43°C above the martensitic transition. The inset shows that the thermal hysteresis of FOMT is as large as ~10°C. This large thermal hysteresis causes different magnetizationsaswellastransition temperaturesduringheating andcooling, which would lead to inconsistent response temperature and response time of the thermal switch. These drawbacks indicate that it is not suitable to utilize the FOMT to design TMLM thermal switch. On the contrary, the SOMT in Ni-Mn-In alloy exhibits a perfectly reversible FM −PM transition without thermal hysteresis. Moreover, it shows a large magnetization change and high magnetocaloric effect (Supplementary Fig. 6), which is conducive to the fast response of the TMLM thermal switch. In addition, the SOMT does not generate large dislocation density and internal stress that easily leads to fracture failure like the FOMT. So it is favorable to obtaining better mechanical stability and longer service life than FOMT32. The above results indicate that the SOMT in Ni-Mn-In alloy is suitable for the design of thermal switch. Furthermore, Fig. 3eshowstheM−Tcurves for TMLMs with different Ni-Mn-In contents. Similar to the pure Ni-Mn-In alloy, all the TMLMs also undergo a FOMT followed by a SOMT with increasing temperature. This suggests that the thermomagnetic properties of TMLMs are attributed to the incorporation of Ni-Mn-In particles into the liquid metal, which results in the rapid thermomagnetic response of the TMLM thermal switch. Meanwhile, the T C values of all TMLMs remain stable around 43°C, indicating the thermomagnetic transition of Ni-Mn-In particles is not affected upon mixing with LM for creating the TMLMs. Moreover, the magnetization of TMLMs increases with higher Ni-Mn-In particle packing fractions, revealing the enhancement of ferromagnetism. The above results prove that the magnetic properties of TMLMs are predominantly determined by the Ni-Mn-In component, and so the magnetization will keep rising with the Ni-MnIn content further increasing even higher than 20 wt.%. The magnetization of TMLM-20% changes sharply from the maximum value of 5.4 Am2kg−1to nearly zero during the SOMT, which is highly suitable for the design of a magnetic thermal switch. Figure 3fcomparestheDSCcurvesduringcoolingandheatingof pureLM, pure Ni-Mn-In,andTMLM-20%,respectively.ForpureLM, the small peak at −9.0°C (T 1 ) indicates the onset of crystal nucleation during cooling, while the sharp peak observed at −44.5°C (T 2 ) corresponds to the completion of LM crystallization, consistent with literature30,33. During heating, LM starts to melt at 18.9°C (T 3 ), indicating that LM is a completely molten liquid at RT and thus has high fluidity. The difference between the melting and freezing temperatures is due to the supercooling nature of pure LM. On the other hand, pure Ni-Mn-In shows a structural transition from martensite to austenite, and vice versa, at temperatures T M−A and T A−M , with a large hysteresis, as well as a reversible SOMT at T C =40.3°C. These are consistent with the thermomagnetic results. For TMLM-20%, its DSC results reveal the occurrence of structural transformation and T C (same as that of Ni-Mn-In), consistent with those observed from M(T) curves. While TMLM-20% has T 2 comparable to that of pure LM, T 3 Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 4 shifts distinctly from 18.9 (pure LM) to −4.5°C, which suggests that the introduction of thermomagnetic particles reduces the supercooling of LM30,34. Consequently, the melting temperature can be lowered, which will benefitahighfluidity of the TMLM around RT. Furthermore, through the inclusion of Ni-Mn-In particles, the least loaded TMLM droplet(TMLM-5%)iseasilychanneledupabent tubebyamagnet(and smoothly falls to the bottom upon removal of the magnet due to the gravity) as presented in Fig. 3g, further demonstrated for its magnetic actuation in crooked channels and petri dish (Supplementary Fig. 7 and Movie 1). In addition to the characteristic magnetic-driven behavior, our TMLM demonstrates good thermomagnetic responses: the TMLM-5% droplet placed at a cold side, i.e., temperatures below its T C , can be FM-held by a magnet and ultimately drops off as the temperature rises beyond the T C (the droplet transits to a PM state) as shown in Supplementary Fig. 8. This demonstrates that the TMLMs, tackling with their characteristic fast thermal-induced magnetic transition, can aid the response time of a thermal switch unlike typically reported magnetic liquid metals, which are not sensitive to temperature15,19–21,24. These attributes present distinct advantages not found in conventional LM and magnetic LM circuits, making them highly desirable for designing the next-generation magnetic liquid metal thermal switch. Since the T C can be tuned by changing the composition of thermomagnetic transition materials, the operating temperature of the TMLMs can also be tuned accordingly. Figure 3h shows the M−T curves for another Ni-Mn-In Heusler alloy with different composition of Ni 1.58 Mn 2 In 0.42 . The inset reveals that the T C of Ni 1.58 Mn 2 In 0.42 is 65°C, higher than that of Ni 2 Mn 1.4 In 0.6 (43°C). Then, we placed two types of TMLMs doped with Ni 2 Mn 1.4 In 0.6 and Ni 1.58 Mn 2 In 0.42 particles into a water bath, as shown in Fig. 3i. Initially, both TMLMs exhibited ferromagnetic behavior in the cold water at 18.5°C, allowing them to be attracted to the magnet. As the water temperature increased, the TMLM filled with Ni 2 Mn 1.4 In 0.6 detached from the magnet at 47.8°C, while the TMLM filled with Ni 1.58 Mn 2 In 0.42 detached at 71.6°C. This ability to adjust the T C allows for the customization of TMLMs to meet specific temperature requirements, enhancing their versatility and performance in various thermal management applications. 110100 10 0 10 1 10 2 10 3 G'' TMLM-5% G'' TMLM-10% G'' TMLM-15% G'' TMLM-20% G' TMLM-5% G' TMLM-10% G' TMLM-15% G' TMLM-20% G', G'' (Pa) Shear stress (Pa) abc d -60 -40 -20 0 20 40 60 80 0.0 1.0 2.0 3.0 4.0 5.0 6.0 Cooling Heating TMLM-5% TMLM-10% TMLM-15% TMLM-20% μ 0 H = 0.4 T Magnetization (Am 2 kg –1 ) Temperature (ȭ) 0.0 0.5 1.0 1.5 2.0 2.5 LM At room temperature of 20ȭ 602015105Ni-Mn-In Percentage (%) Electrical conductivity (×106 S m –1 ) -60 -40 -20 0 20 40 60 80 -1 0 1 2 3 T M − A T A − M T C T 3 T 1 T 3 exo Pure LM Pure Ni-Mn-In TMLM-20% Heat flow (mW mg– 1 ) Temperature (ȭ) Cooling Heating T 2 f e Critical shear stress τ0 TMLM-5% TMLM-10% TMLM-15% TMLM-20% 60q TMLM-5% droplet driven by the magnet Magnet TMLM-5% Lift Release Drop TMLM-5% TMLM-5% Pure LM 05101520 115 125 135 145 Contact angle Percentage (%) -60 -40 -20 0 20 40 60 80 0 20 40 60 80 Ni 2 Mn 1.4 In 0.6 T C = 43 ȭ Cooling Heating μ 0 H = 0.4 T Magnetization (Am 2 kg –1 ) Temperature (ȭ) -60 -40 -20 0 20 40 60 -2 0 2 4 6 (Am 2 kg –1 K –1 ) dM/dT T (ȭ) ghiTemperature sensitivity of TMLMs with different TC -60 -40 -20 0 20 40 60 80 100 0 5 10 15 Ni1.58Mn2In0.42 T C = 65ȭ μ0H = 0.01 T Magnetization (Am2 kg–1) Temperature (ȭ) Heating Cooling -60 -20 20 60 100 -1.0 -0.7 -0.4 -0.1 0.2 (Am 2 kg –1 K –1 ) dM/dT T (ȭ) 15.9℃25.7℃17.1℃59.5℃17.9℃84.5℃ Magnet Ni2Mn1.4In0.6 Ni1.58Mn2In0.42 18.5 ℃ 47.8 ℃ 71.6 ℃ TC TC Fig. 3 | Physical properties and magnetic actuation performance of TMLMs. aOscillatory rheological test of TMLMs with different packing fractions. G′and G′′ are the elastic modulus and viscous modulus, respectively. bContact angle as a function of packing fraction for TMLM droplets. The second row shows the droplet shape of pure LM and TMLMs with different packing fractions. Scale bars, 1 mm. cElectrical conductivity of Ni-Mn-In alloy, pure LM, and TMLMs with different packing fractions. Error bars indicate the standard deviation of three measurements. dMagnetization (M) as a function of temperature (T)atamagneticfield of 0.4 T for Ni 2 Mn 1.4 In 0.6 alloy. The inset shows the corresponding dM −dT curves. Solid symbols: heating, empty symbols: cooling. eThe M−Tcurves at 0.4 T for TMLMs with different packing fractions. fDSC curves of pure LM, pure Ni-Mn-In, and TMLM-20% measured during heating and cooling (negative values for heating and positive values forcooling).gVertical control of the TMLM-5% droplet by using the magnet. hThe M−Tcurves for another Ni-Mn-In Heusler alloy with composition of Ni 1.58 Mn 2 In 0.42 . The inset shows the corresponding dM −dT curves. iExperimental demonstration and infrared thermal images of the temperature sensitivity for TMLMs with different T C . (Source data are provided as a Source Data file). Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 5 Micro-channel TMLM thermal switches Based on the thermomagnetic-sensitive property of TMLMs, an advanced TMLM thermal switch was designed for a practical configuration: a micro-channel thermal switch using the TMLM droplet. As shown in Fig. 4a, a small TMLM droplet is housed within a thin glass tube. Initially, when the temperature of TMLM (T LM )islowerthanits T C , the droplet is drawn towards the upper end of the tube by a small magnet, thus completing the LED circuit. Upon surpassing the T C ,the droplet turns from FM to PM (due to the characteristic thermalinduced magnetic transition of TMLM), causing it to detach from the magnet and descend to the bottom, subsequently breaking the LED circuit. Similarly, the LED circuit can also be configured as a thermaltriggered conduction switch by positioning it at the lower end. In this case, the LED would illuminate upon the droplet detaching from the magnet as a result of overheating. Leveraging this design, we developed a miniature TMLM thermal switch using a TMLM-20% droplet. In 1 cm sm046sm002 20℃85℃ Simulated temperature of TMLM droplet during heating a Magnet LED (On) Power b LED (Off) LED (Off) LED (On) c ef d 0 ms 1020 ms 1200 ms Recovery OnOff OffOn 520 ms 540 ms 560 ms 590 ms 620 ms Mechanism of micro-channel thermal switch Detach and recovery the attraction of magnet for TMLM droplet Unidirectional response of TMLM thermal switch 0 20 40 100 Thiswork Maetal. Chenetal. Xieetal. Luoetal. Liueta l. Cuietal. Davidetal. Response time t1 (s) 45 55 65 75 85 0 5 10 15 20 Φ 6 × 30 mm Φ 5 × 20 mm Response time t 1 (s) Temperature ( ȭ ) TMLM thermal switch Resistance-shift switch Shape-change switch Gas/smoke sensors This work (660 ms) 0 ms 500 ms Recovery Convection heating Laser heating Initial state 0 20 40 60 80 Temperature (ȭ 9.4 9.5 9.6 Magnetic flux density B ( ×10 – 3 T) 0 300 600 900 1200 0 2 4 Speed v (cm s – 1) Time (ms) (Hot source = 85℃) Falling stage Response temperature Fig. 4 | The unidirectional response of micro-channel TMLM thermal switches. aSchematic illustration of the working principle of micro-channel TMLM thermal switch. bExperimental demonstration of TMLM-20% droplet detach and recovery the attraction of magnet during heating and cooling. Two different heating methods are used: contact convection heating and non-contact laser heating. cExperimental demonstration of two unidirectional response modes of TMLM thermal switch. dComparison of response time t 1 of TMLM thermal switch with those of reported different types of thermal switches. The inset shows t 1 of TMLM20% droplet in two sizes of glass tubes as a function of hot source temperature. Error bars indicate maximum and minimum values. eSimulated temperature of a TMLM-20% droplet falling during heating. fThe simulation results of the average temperature, magnetic flux density, and drop speed of TMLM-20% droplet as functions oftime during heating, hot source = 85°C. (Source data are provided as a Source Data file). Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 6 the experimental setup, both contact fluid convection heating and non-contact laser heating methods were used to facilitate the release of the TMLM droplet, as shown in Fig. 4b. A small droplet (volume of ~30 µL) consisting of TMLM-20%, is encapsulated in a glass tube that is filled with a commercial thermal fluid to enhance heat transfer. Moreover, this thermal fluid can also prevent direct contact between liquid metal and oxygen, and then reduces the formation of the oxide layer, leading to a good fluidity and thermomagnetic response of TMLMs (Supplementary Fig. 9). The droplet is held by the magnet, whose maximum field at RT is 0.4 T. It should be pointed out that the droplet can be lifted by a magnetic field µ 0 Hlower than 200 mT, which is favorable for practical applications. Meanwhile, the distance between the magnet and TMLM droplet has been adjusted to 3.5 mm based on the relationship between the magnetic field and the distance between the magnet and the droplet (Supplementary Fig. 10). To assess the actuation of this tube thermal switch, we submerge it in a beaker containing hot water (at a temperature of 75°C) to replicate the heating process. Upon exposure to the external heat, the TMLM-20% droplet gradually detaches from the magnet’s attraction and then assumes an ellipsoidal shape at 450 ms as a result of reduced surface tension35. Subsequently, at 1020 ms, it initiates detachment from the top, ultimately settling at the bottom at 1200 ms. Next, the microchannel thermal switch is extracted from the beaker for air cooling. The TMLM droplet gets re-attracted to the top end, demonstrating its recovery to the FM state, which strongly indicates the potential for repeated utilization over multiple operational cycles. The entire process of actuating the micro-channel thermal switch has been recorded and is available for viewing in Supplementary Movie 2. The microchannel TMLM thermal switch is susceptible not only to contact convectionheattransfer,butalsotonon-contact heat transfer methods, such as radiation and laser heat transfers (Supplementary Movie 3). The laser-heating experiment demonstrates that the TMLM droplet takes only 500 ms to fall to the bottom when subjected to the laser, and subsequently returns to its FM state through air cooling, rapidly moving to the top end. These results prove the suitability of the TMLM thermal switch for diverse heating scenarios. Additionally, the tube thermal switch was further integrated into a LED circuit following the schematic diagram inFig. 4a. The operation of thetube thermalswitch activating and deactivating the LED circuit is captured and shown in Fig. 4c and Supplementary Movie 4. The observation reveals that the TMLM droplet detaches when subjected to heating, thereby causing the LED circuit to switch ON or OFF. Theresponsetime(t 1 ), defined as the duration from the moment the switch contacts the heat source to the point at which the TMLM droplet detaches from the magnet, has been optimized by adjusting thesizeoftheglasstubeandthetemperatureoftheheatsource(T H ). The inset of Fig. 4dshowsthet 1 of TMLM-20% droplet within two glass tubes varies as a function of T H . For the same T H ,thet 1 in the smaller glass tube (Φ5×20mm)issignificantly shorter than that in the larger tube (Φ6×30mm).Thisisbecausethesmallerglasstubefacilitates better heat conduction, making it advantageous for the design of a micro-size TMLM thermal switch. With increasing T H ,thet 1 in both glass tubes significantly reduces. This is because the tube temperature rises more rapidly under a higher T H , and it takes a shorter time for the TMLM-20% droplet to reach T C (Supplementary Fig. 11), leading to the shortening of t 1 .Inaddition,thefastresponsetimet 1 for the Φ5× 20 mm tube under T H = 85°C can reach 660 ms under the optimal environment (Supplementary Fig. 12). It is noteworthy that the variance in response time among the tubes of different sizes diminishes significantly with increasing T H , implying that T H exerts a more pronounced influence on the response time. Furthermore, to ensure a fair comparison with different types of thermal switches,we selected three shape-change switches36–38 that exposed to similar T H ranging from 85°C to 100 °C. Notably, an alcohol lamp (with an outer flame temperature of ~500°C) was used to trigger another shape-change switch39 and two resistance-shift switches7,40. Detailed heating methods and temperatures are shown in Supplementary Table 1. Figure 4d presents a comparison of the t 1 of our TMLM thermal switch with that of other types of thermal switches. Notably, the t 1 of our TMLM thermal switch substantially outperforms the other reported thermal switches. Specifically, even the T H ( ~500°C) of the alcohol lamp is much higher than the 85 °C hot water bath used in our experiments, our TMLM thermal switch still exhibited a faster response time, which is between 1 and 2 orders of magnitude faster than those of resistance-shift and shapechange switches. Remarkably, this rapid t 1 exceeds that of gas/smoke sensors by 151 times. The exceptionally fast response time underscores the sensitivity and efficiency of our TMLM thermal switch. The temperature and shape variation of the TMLM-20% droplet as it falls at T H = 85°C were further analyzed by finite element simulation (detailed in Supplementary Figs. 13−16, Table 2, and Movie 5). Upon heat exposure for 200ms, the temperature of TMLM droplet within the tube experiences an initial increase (Fig. 4e and Supplementary Fig. 15). Notably, the temperature of TMLM droplet surpasses that of the thermal fluid, attributed to the high thermal conductivity of LM (25.4 W m−1K−1)14. Subsequently, the T LM quickly reaches ~60°C after 520ms, becoming ellipsoidal. It then starts detaching from the top at 540 ms, completing its descent to the bottom at 620 ms. Moreover, a further analysis of the magnetic flux density (B) within the droplet and drop speed (v) of TMLM-20% (Supplementary Fig. 16) as depicted in Fig. 4f presents the stability of the parameters T LM ,B,andvwithin the initial 110 ms. This stability is attributed to the transfer of external heat to the interior through the glass tube and thermal fluid. Then, the T LM gradually increases while Bdecreases, which is attributed to the thermal-induced magnetic transition. A minor T LM peak emerges around 62°C at 520 ms, signifying the critical response temperature of the TMLM droplet. At the same time, vshowsa sharp peakof4.5cm s−1, corresponding to the falling of TMLM droplet. The vthen diminishes to 0 at 620 ms, indicating the completion of the falling process. This modeled drop time of 620ms is in good agreement with the experimental value (660 ms, T H = 85°C), thereby affirming the reliability of the experimental response time. Consequently, the above results prove that our designed miniature micro-channel TMLM thermal switch can achieve two functions of thermal-triggered connection or disconnection with a fast response time. A fast bidirectional TMLM thermal switch technology for practical applications The aforementioned micro-channel thermal switch serves as the basis for our further design, which involves connecting two electrical circuits at the upper and lower ends. This modification allows for the bidirectional response of the TMLM thermal switch to be realized (see Fig. 5a). The experiment depicted in Supplementary Movie 4 demonstrates that the TMLM droplet initially moves to the upper end due to magnetic attraction, thereby completing the upper circuit and activatingLED1. Meanwhile, LED2atthelowerendremainsOFFduetothe lack of connectivity. Upon reaching the temperature above the T C of the TMLM, the droplet descends to the bottom, causing LED 1 to turn OFF and LED 2 to illuminate (ON). Figure 5b shows the corresponding experimental results of the system. At RT, the blue LED in the upper circuit is illuminated (ON) while the red LED in the lower circuit remains unlit (OFF). Upon the addition of hot water to the beaker, the thermal switch is heated, causing the TMLM droplet to detach. This results in the deactivation of the blue LED (OFF) and the activation of the red LED (ON). Notably, finite element simulation results indicate that the transition from the disconnection of the upper circuit to the conduction of the lower circuit occurs in a mere ~100ms. This brief timespan underscores the thermal switch’s ability to achieve instantaneous bidirectional response. Figure 5c, d presents a comparative analysis of the response process of a commercial thermal switch, referenced as KSD9700C, Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 7 versus our designed TMLM thermal switch in the context of a fire event. The process is visually detailed in Supplementary Movies 6 and 7. The KSD9700C switch primarily functions by utilizing the deformation ofits internal spring sheet during heating to interrupt the circuit. This commercial thermal switch design is widely used in our daily life and industrial applications. As shown in Fig. 5c, the LED electrical circuit remained operational (ON) initially. At 5.8s, a candle was introducd beneath the KSD9700C switch to emulate circuit overheating. Subsequently, the KSD9700C switch reacted at 15.3 s, disconnecting the LED electrical circuit. This observation indicates a 9.5 s response time for the commercial thermal switch to deactivate the electrical circuit following the onset of the fire. Moreover, its unidirectional response functionality restricts its capability to solely disable the LED electrical circuit without the capacity to independently initiate fire extinguishing actions. On the contrary, as shown in Fig. 5d and Supplementary Movie 7, upon introducing the candle under the TMLM thermal switch at 5.8 s, it promptly responded at 7.6 s to terminate the electrical circuit while simultaneously activating the fire shower, extinguishing the fire at 8.2 s. Compared to the KSD9700C switchwith aresponsetime of9.5s, the TMLM thermal switch achieves circuit deactivation 5.3 times faster, requiring a mere 1.8 s since the onset of the fire. Furthermore, it accomplishes fire suppression in only 2.4 s. Although the response time of a thermal switch may vary depending on the nature and environment of the fire, the above results obtained under the same conditions demonstrate that this TMLM thermal switch shows a significantly faster response time than conventional thermal switches. Thus, it allows for faster activation of alarms and protective measures (such as triggering fire suppression systems), significantly reducing the risk of damage or injury in critical situations. This outcome underscores the remarkably improved response capabilities of the TMLM thermal switch in swiftly terminating the circuit and triggering the fire suppression measures, indicating its potential to enhance fire safety systems. In addition to the fast bidirectional response, the TMLM droplet can convert back to FM state during cooling and return to its original state, thus realizing multiple recycling. An evaluation of our TMLM thermal switch, compared with five typical commercialthermal switches (see Supplementary Table 3) with Commercial thermal switches c aMechanism of bidirectional response LED1 (On) LED2 (Off) LED1 (Off) LED2 (On) )nO (d eR ,) ffO ( e u lB)ffO(deR, ) n O (e u lB e Demonstration of TMLM thermal switch for fire warning and extinguishing bBidirectional response of TMLM thermal switch fg d 0 s 5.8 s 8.2 s 7.6 s Nozzle LED Power Socket TMLM thermal switch Demonstration of the shape-change commercial thermal switch KSD9700C for fire warning Spout water h 0 50 100 180 200 KSD9700C KSD9700M KSD9700P KSD-301 KSD-01F Thiswork Recovery time t 2 (s) 0 2 4 6 8 KSD9700P KSD9700C KSD9700M KSD-01F KSD-301 Thiswork Response time t 1 (s) 0.0 0.5 1.0 1.5 2.0 KSD9700P KSD9700C KSD9700M KSD-01F KSD-301 Thiswork Unit cost C (USD) Hot source = 75℃Cooling method: air cooling TMLM thermal switch KSD 9700M KSD 9700P KSD 9700C KSD-301 KSD-01F 0 s 5.8 s 15.3 s Simulated fire Cut off the circuit Simulated fire Cut off the circuit KSD9700C thermal switch Power LED Power Final state 16.7 s Fig. 5 | Fast bidirectional response and practical applications of TMLM thermal switch, as well as the comprehensive comparison with typical commercial thermal switches. a Schematic illustration of the working principle of the bidirectional response of TMLM thermal switch. bExperimental demonstration of bidirectional response of TMLM thermal switch. cDemonstration of the shapechange commercial thermal switch KSD9700C for fire warning. dThe practical application demonstration of TMLM thermal switch for fire warning and extinguishing. eImages of TMLM thermal switch and typical commercial thermal switches. f–hComparison of (f) response time t 1 ,(g) recovery time t 2 ,and(h) unit cost Cof TMLM thermal switch with those of typical commercial thermal switches. (Source data are provided as a Source Data file). Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 8 operating temperatures (45 °C) close to the T C of our TMLMs (43 °C), indicates that the TMLM thermal switch is compact and suitable for use in confined space or micro-circuits as shown in Fig. 5e. Additionally, a comprehensive comparison of the response time (t 1 ), recovery time (t 2 ), and unit cost (C) between the TMLM thermal switch and typical commercial thermal switches (Fig. 5f) demonstrates that our TMLM thermal switch exhibits the shortest response time t 1 of 1.2 s under the same T H of 75 °C. Although the t 1 of 1.2 s may be slower than some millisecond-level devices, it is still 3.3 −5.6 times faster than typical commercial thermal switches. As the T H increases, the t 1 of our TMLM thermal switch can also reach milliseconds (Fig. 4d), thus showing a fast response characteristic. In applications such as fire safety or overheating protection systems, a faster response time means quicker disconnection of the electrical circuit or activation of fire suppression systems in the first few seconds after ignition. This significantly reduces the risk of property damage, injury, or even loss of life, effectively meeting the urgent need for fast responsed thermal switches. Moreover, in Fig. 5g, it is evident that the TMLM thermal switch demonstrates the fastest recovery time t 2 of 53.6 s among all thermal switches considered. Notably, this t 2 value is 3.5 times faster than that of KSD9700C switch. Moreover, we have demonstrated that this t 2 can be further reduced to 21 s or 4 s by using cold air or ice water cooling instead of air cooling (see Supplementary Movie 2). This fact highlights the substantial impact of the cooling environment on the recovery time. In addition, owing to the minimal amount of liquid metal and low raw material cost, the unit cost (C) of the TMLM thermal switch, as shown in Fig. 5h, is the lowest at ~1.016 USD (detailed cost calculations are shown in the Supplementary Notes). This cost reflects raw material expenses in a laboratory setting and does not include processing or molding costs, and we expect thatthe cost of this TMLM thermal switch can be further reduced with large-scale manufacturing. So it would be more cost-effective than other thermal switches after commercialization. These benefits underscore the enormous potential of our TMLM thermal switch in the field of thermal management. Furthermore, we built a cycle life test device and carried out a cyclic test to verify the cyclic stability of the TMLM thermal switch (Supplementary Fig. 17). After 200 heating and cooling cycles, the TMLM thermal switch still retains good temperature sensitivity and fast response capability, demonstrating excellent durability and stable performance over extended use. It is worth noting that this ferromagnetic transition TMLM can also find application in many other various fields, such as direct 3D printing on various substrates and function as flexible thermal switches (Supplementary Figs. 18−21, Movies 8 and 9); fast response temperature sensors and full supervision of refrigerated transportation in industry (Supplementary Fig. 22); magnetic refrigeration technology based on the magnetocaloric effect, and efficient heat dissipation of micro-size devices (Supplementary Fig. 23). Therefore, the diverse capabilities of this ferromagnetic transition TMLM will have a broader range of potentialapplication beyond thermal switches in future. Discussion Thermal switches are important for detecting malfunctions in a broad variety of devices. They are instrumental in preventing significant economic damage and saving lives in the case of equipment overheating. The key parameters for optimal performance include response time and operation temperature. In addition, reactivation after the issue is resolved and the ability to independently activate countermeasures are significant enhancements for thermal switches. In this work, we designed a TMLM thermal switch with a fast bidirectional response by creating a thermomagnetic liquid metal. This material has good fluidity, appropriate viscosity, and high electrical conductivity. It can be tuned to actuate at different temperatures, and the proposed design enables simultaneous detection of thermal malfunctions and activation of countermeasures. It shows a fast response time t 1 of 1.2 s under the same T H of 75 °C, which is 3.3 −5.6 times faster than typical commercial thermal switches. Moreover, it can be further reduced to 660 ms under the optimal environment. Besides, the TMLM thermal switch also hasthe fastest recovery time t 2 of 53.6s compared with the commercial thermal switches, and this t 2 can be further reduced to 4 s using ice water cooling. Meanwhile, itstill retains good temperature sensitivity and fast response capability after 200 heating and cooling cycles. Despite its enhanced performance and added functionality, this thermal switch incorporating the TMLM is estimated to cost only ~1.02 USD for each. The intrinsic properties of the TMLM also make it suitable for applications beyond this thermal switch in the future. Methods Raw materials The raw materials of Ni, Mn, and In with purity higher than 99.5 wt.% were purchased from Beijing Jiaming Platinum Nonferrous Metals Co., Ltd. The commercial Galinstan (Ga 68.5 In 21.5 Sn 10 )liquidmetals(LMs) with a theoretical melting point of ~11°C were purchased from Dongguan Houjie Dingtai Metal Materials Co., Ltd. The thermal fluid was purchased from Sichuan Wuxiandian Technology Co., Ltd. This thermal fluid is a commercial diamond nanofluid with a thermal conductivity of 1.0917 W m−1K−1, and it is also non-conductive and noncorrosive and usually be used as a coolant. Customized quartz glass tubes of different specifications were purchased from Donghai County Shengyi Quartz Products Co., Ltd. Rubber stoppers of different specifications were purchased from Kunshan Wanjianguangmang Hardware and Electrical Co., Ltd. Customized N52 magnets were purchased from Ganzhou Ruitong Magnetic Materials Co., Ltd. Preparation and characterization of Ni-Mn-In particles Ni 2 Mn 1.4 In 0.6 (Ni-Mn-In) and Ni 1.58 Mn 2 In 0.42 alloys were prepared by induction-melting the pure components with a purity higher than 99.5 wt.% under an argon atmosphere. The as-cast ingots were both sealed in a high-vacuum quartz tube and annealed at 900°C for 7 days (Ni 2 Mn 1.4 In 0.6 ) and 800°C for 2 days (Ni 1.58 Mn 2 In 0.42 ), followed by quenching in ice water. The ingots were then crushed into large particles with an iron pestle, manually pounded into powders using a mortar, and sieved through a 300-mesh screen. The size distribution was measured by dynamic light scattering with Malvern Mastersizer 2000 laser particle size analyzer. The microstructure was studied by a Merlin VP Compact field emission scanning electron microscope (SEM), and the elemental analysis was investigated by energy-dispersive spectroscopy (EDS). Phase analysis was performed and lattice parameters were determined by roomtemperature X-ray powder diffraction (XRD) using Ultima IV diffractometer with Cu K α1 radiation, the scan rate was 10° min−1. Synthesis of thermo-magnetic liquid metals (TMLMs) The magnetic modification process for obtaining the TMLMs is illustrated in Fig. 2a, which can be specified as the following steps. First, weaddedabout10 g Galinstanliquid metalintoa clean beaker by using a pipette gun.Next, the suitable mass of Ni-Mn-In particles were added to Galinstan LM, and then the beaker was shaken until the surface of Galinstan was coated with a uniform layer of Ni-Mn-In particles. Then, TMLMs were made by vigorously stirring galinstan and Ni-Mn-In particles. It takes about 15 −30 min to fully internalize the particles, depending on the packing fraction of Ni-Mn-In particles. Finally, TMLMs were sonicated for 2 min to make the Ni-Mn-In particles evenly distributed in the LM and improve the stabilityof the TMLMs. It should be noted that we prepared the TMLMs by directly mixing LM and NiMn-In particles in air rather than in HCl or NaOH solutions as reported in most works28,41, in order to prevent LM and Ni-Mn-In particles from reacting in these solutions. Article https://doi.org/10.1038/s41467-025-58015-x Nature Communications | (2025) 16:2634 9