Introduction

Electronics in aerospace systems, specialized battery modules, and electric vehicles frequently experience material aging, performance degradation, and thermal fatigue due to exposure to severe thermal cycling under dynamic operating conditions1,2,3. In aerospace environments, where constraints, such as weightlessness and volumetric limitations, impose significant design challenges, conventional thermal management systems (TMS) often fail to provide adequate thermal regulation4,5. To address this, phase change material (PCM)-based thermal buffers have been proposed as a viable solution, leveraging their inherent ability to absorb and release latent heat, thereby achieving thermal load leveling through peak shaving and valley filling6,7. However, the inherently low thermal diffusivity of PCMs necessitates the incorporation of high thermal conductivity reinforcement media to enhance heat dissipation efficiency8,9,10. Current enhancement strategies predominantly focus on mitigating thermal runaway phenomena during steady-state operation, yet they exhibit limited adaptability to extreme thermal transients and fluctuating ambient conditions11,12,13. To ensure sustained operational stability of electronic devices under prolonged exposure to dynamic thermal loads, the development of thermal buffer systems with dynamically tunable heat transfer properties is imperative. Such systems must incorporate mechanisms capable of actively modulating thermal resistance in response to transient thermal demands, thereby ensuring robust thermal management under extreme operational variability.

Dynamic tuning of thermal management characteristics is expected through two main approaches: one involves implementing a thermal switch between electronics and thermal buffer, while the other directly regulates the heat conductance within thermal buffer14,15,16. Concerning electrical switches that regulate the on-off of circuits, previous efforts have focused on manipulating the interfacial thermal resistance between components to control the activation and deactivation of heat transfer paths17. Substantial switching ratios can be achieved by dynamically adjusting the interfacial thermal resistance18, which is controlled by driving the movement of millimeter-scale liquid metal droplets within the voids of the heat transfer path19,20. As the nonuniform strain notably impacts phonon transport21, recent research has also attempted to regulate heat transfer paths by embedding thermally expandable microspheres between layers22. Additionally, nanoscale reversible electrochemical23, hydration24 as well as electric field25,26-driven approaches can enable practical tuning of thermal resistance27,28. Although thermal switches achieve a switching ratio greater than 10, they operate with only two states, on and off, and lack the ability for dynamic performance tuning29. Attempts to regulate heat transfer in thermal buffers have focused on inducing the solid-liquid phase transition of PCMs at temperatures below their melting point using various substances30,31,32,33. However, due to the low thermal conductivity of PCMs, it is more efficient and cost-effective to directly modify the morphology and arrangement of the heat transfer enhancement medium within the PCM34,35. It is demonstrated that incorporating nanoparticles can create more versatile heating surfaces36. However, as shown in Fig. 1a, current composite PCMs suffer from the anisotropic heat transfer properties of nanoparticles, such as carbon nanosheets and nanotubes37,38. The random distribution of these nanoparticles hinders precise control over the heat transfer process. Past attempts have drastically improved the heat transfer capacity of composite PCMs in a single direction by strategically orienting carbon-based nanoparticles39,40,41. These approaches further rely on van der Waals forces42 or magnetic forces43 to achieve structurally stable thermal buffer materials. Nevertheless, dynamic regulation of heat transfer performance often requires the application of an external force field1. Leveraging the intrinsic electrical conductivity of carbon-based nanoparticles, the application of a D.C. field can facilitate the formation of chain-like structures, resulting in a 1.35× switching ratio44. When combined with magnetic materials, the aggregation and arrangement of carbon-based nanoparticles can also be regulated using magnetic fields45,46,47.

Fig. 1: Thermal management process of PCM-based thermal buffer under dynamic operating conditions.
Fig. 1: Thermal management process of PCM-based thermal buffer under dynamic operating conditions.
Full size image

a Schematic representation of the distribution of nanoparticles within composite PCM in the original state. The nanoparticles show a random and uniform distribution, which only enhances the equivalent thermal conductivity of the PCM. b Magnetic field regulates nanoparticle aggregation for rapid heat dissipation under high heat flux status. c Magnetic field regulates the direction of the nanoparticle aggregation structure in the low-temperature standby status. The insulation process can be achieved through the weakening of the thermal conductivity along the mainstream heat transfer path.

Conventional methods for magnetically aligning thermally conductive fillers are confined to a static, one-time optimization during material fabrication, yielding composites with fixed thermal properties. This work transcends this paradigm by creating a dynamically reconfigurable thermal management system using Fe3O4@carbon nanotube (CNT) nanoparticles dispersed in n-eicosane. Moving beyond the prevailing macroscopic property enhancement, this work tends to precisely characterize and leverage the mesoscale aggregation structure of nanoparticles under a uniform magnetic field as a tunable thermal network. By dynamically orienting the nanoparticle aggregation structure through varying the tilt angle between the magnetic induction line and the dominant heat transfer path, we demonstrate in situ and reversible tuning of the effective thermal resistance with a 1.8× switching ratio. This enables the same material to function on demand as an efficient heat spreader (Fig. 1b) or a potent thermal insulator (Fig. 1c), thereby mitigating the temperature excursions under dynamic and intermittent loading. Cycling tests further confirm the reliability of the dynamic regulation method, demonstrating a reduction in temperature fluctuations by 10.8 °C. Furthermore, it is noteworthy that the dynamic reorientation of nanoparticle aggregates is primarily effective within the melted-phase PCM during the apparatus operating, since the solidified PCM matrix significantly hinders particle movement. This inherent behavior is strategically leveraged in our design: the pre-formed aggregated structure is “locked-in” upon solidification during standby, contributing to the thermal insulation effect, and is dynamically reconfigured only when the PCM melts again during the next operational cycle.

Results

Magnetic composite PCM and their thermal features

In the current investigation, we prepared composite PCM by incorporating Fe3O4@CNT nanoparticles as the controllable heat transfer reinforcement medium and using n-eicosane, a commercially available material with high latent heat, as the matrix. The attachment of Fe3O4 nanoparticle to CNT is implemented by a two-step method48,49 to impart tunability under magnetic fields (details see Supplementary Note 1 and Supplementary Fig. 1). The obtained composite PCM exhibits good stability, and the scanning electron microscopy (SEM) images reveal that the Fe3O4@CNT nanoparticles are uniformly distributed within the PCM and are well-bound to the matrix (Supplementary Fig. 2). Quantitative image analysis of multiple micrographs using ImageJ software confirms this homogeneity (Supplementary Note 3 and Supplementary Fig. 6a, b). The Fe3O4 nanoparticles and CNTs showed narrow size distributions with diameters of 33.5 ± 8.3 nm and 30.6 ± 5.1 nm, respectively. The thermal characteristics of the composite PCM are assessed using differential scanning calorimetry (DSC), to investigate the effect of nanoparticle concentration (Fig. 2a) and mixing ratio R = \({m}_{{{{{\rm{Fe}}}}}_{3}{{{{\rm{O}}}}}_{4}}/{m}_{{{{\rm{CNT}}}}}\) (Supplementary Fig. 3a). The standard DSC curves of crude n-eicosane and composite PCM samples with different concentrations or mixing ratios are nearly identical, while the corresponding temperature turning points and latent heat values are labeled in Fig. 2b and Supplementary Fig. 3b. It can be observed that the addition of nanoparticles demonstrates a negligible effect on the PCM melting process, with the temperature points at the peak of melting remaining nearly unchanged around 36 °C. As the concentration of nanoparticles increases, the latent heat value of the composite PCM generally exhibits a decreasing trend. Notably, the composite PCM’s latent heat value is slightly higher than that of the pure n-eicosane when the nanoparticle concentration is 0.1 wt%. This phenomenon aligns with the results obtained by Wang et al.50 and Shadab et al.51 and is primarily attributed to the interaction between CNTs and carbon-based PCM molecules. The observed decline in latent heat becomes disproportionately greater than the nanoparticle mass fraction as concentration increases. This phenomenon, which has been consistently reported in other nanocomposite PCM systems50,52, is primarily attributed to interfacial interactions and nanoconfinement effects. The high surface area of the CNTs provides numerous sites for the adsorption of PCM molecules. These adsorbed molecules experience restricted mobility during the phase transition, which diminishes their effective contribution to the bulk latent heat. The thermal processes exhibited by the composite PCM under macroscopic conditions are less affected by the mixing ratio. This is because the density of Fe3O4 is significantly greater than that of CNT, and the volume fraction of Fe3O4 remains relatively constant when the mixing ratio is altered53. Figure 2c depicts the thermal conductivity of composite PCM analyzed through a laser thermal conductivity meter (LFA447, Germany) at 20 °C. The thermal conductivity of composite PCM demonstrates an upward trend as the nanoparticle concentration progressively increases at a given mixing ratio. Since Fe3O4 exhibits both lower thermal conductivity and a lower volume fraction compared to CNTs, increasing the mixing ratio does enhance the thermal conductivity, but the effect remains minimal. To quantitatively validate the magnetic field-mediated tuning of thermal transport, we directly measured the effective thermal conductivity of magnetic field regulated samples with the procedure implemented in the “Methods” section. As summarized in Supplementary Fig. 4, exhibits pronounced anisotropy, governed by the alignment of the nanoparticle aggregation structures. This results in a significant thermal conductivity tuning ratio of 1.71 between the 0° and 90° configurations. The X-ray diffraction (XRD) pattern of the composite PCM in Fig. 2d shows the diffraction peaks of both n-eicosane and Fe3O4@CNT particles, indicating that the nanoparticles are physically bonded to the matrix, with no chemical reaction occurring to form a new substance. With the mass fraction and mixing ratio of Fe3O4@CNT particles respectively fixed as 0.2 wt% and 1, Fig. 2e illustrates the mesoscopic characterization of the Fe3O4@CNT nanoparticles after aggregation, which occurs in the presence of a uniform magnetic field. The detailed regulation process is described in Supplementary Note 2. It can be observed that the nanoparticles have aggregated into uniformly distributed bundle-like structures along the direction of the magnetic induction line, forming an efficient heat transfer network. As is characterized in Supplementary Note 3, the bundle-like aggregates demonstrate an average length of 201.5 ± 39.7 μm and width of 20.9 ± 3.6 μm, confirming the formation of a uniform and highly anisotropic network for thermal transport. The aggregation structure was found to be highly repeatable over multiple melting-solidification cycles (≥100), confirming the structural stability and reliability of the tuning mechanism (see Supplementary Note 5 for post-cycling characterization). By altering the tilt angle between the direction of boundary heat flux and the magnetic induction line, it is possible to regulate the transient heat transfer performance within thermal buffers.

Fig. 2: Characteristics of magnetic composite PCM filled with Fe3O4@carbon nanotube nanoparticles.
Fig. 2: Characteristics of magnetic composite PCM filled with Fe3O4@carbon nanotube nanoparticles.
Full size image

a Differential scanning calorimetry (DSC) curves of pure n-eicosane and composite PCM with different loading of Fe3O4@carbon nanotube (CNT) nanoparticles. The mixing ratio (R = \({m}_{{{{{\rm{Fe}}}}}_{3}{{{{\rm{O}}}}}_{4}}/{m}_{{{{\rm{CNT}}}}}\)) is fixed as 1. b Comparison of fusion phase-change enthalpy, onset and peak melting temperatures of pure n-eicosane and composite PCM loaded with Fe3O4@CNT nanoparticles. Error bars represent the errors in DSC testing and the uncertainties present during the statistical processing of DSC test results. c Thermal conductivities of composite PCMs with different loading (0.1 wt%, 0.2 wt% and 0.3 wt%) and mixing ratio (R = 0.5, 1 and 1.5) of Fe3O4@CNT nanoparticles. Error bars represent the measurement uncertainty of experimental data. d Comparison of X-ray diffractometer (XRD) patterns among composite PCMs with different loading and mixing ratio of Fe3O4@CNT nanoparticles. e Mesoscopic characterization of the Fe3O4@CNT nanoparticle aggregation structure within the PCM matrix, formed under a uniform D.C. magnetic field of 200 mT applied parallel to the observed alignment direction. The sample was exposed to the field for at least 20 s prior to imaging to ensure a stable and uniform structure. Source data are provided as a Source data file.

Heat dissipation performance

The transient heat transfer process within the thermal buffer is simulated at the mesoscopic scale to investigate the impact of the magnetic field-induced nanoparticle aggregation structure on local heat transfer characteristics. A detailed description of the implemented physical model can be found in Supplementary Note 6. The heat flux is fixed as 10,000 W/m2, which is widely adopted for simulating the operation of electronic devices54,55,56. The images recorded by electron magnification are converted into binary format to facilitate the construction of a mathematical model. The transient phase transition behavior of the PCM is analyzed using a two-dimensional total enthalpy-based lattice Boltzmann method57 (Supplementary Note 7). Good consistency in comparison with experimental values and analytical solutions demonstrates its reliability in addressing solid-liquid phase change and conjugate heat transfer problems58 (Supplementary Note 8). The thermophysical data of applied materials is tabulated in Supplementary Table 1, while the thermal physical properties of the aggregated nanoparticle structure are determined according to the rule of mixture model59. It should be noted that, as the thermal conductivity of the dispersed CNTs is overestimated by the perfect interface assumption50,51,59, the thermal conductivity of CNT in the dense-packed mat state60,61 is applied in this study (Supplementary Note 9).

Figure 3a compares the transient melt-front evolution and temperature distribution characteristics within the PCM region to contrast the advantages of magnetic field regulation at different tilt angles. Dimensionless Fourier number (\({{\mathrm{Fo}}}=\lambda t/\rho {c}_{p}{l}^{2}\)) is implemented to indicate time progression58. When the direction of magnetic induction is parallel to the direction of boundary heat flux (tilt angle 0°), the magnetic field-induced aggregated structures form an efficient heat diffusion network along the primary direction of heat transfer. This formation facilitates the rapid migration of the partial melt front along the bundle-like aggregation structure, thus alleviating the overheating degree around the bottom heat source. However, as the volume fraction of liquid PCM increases, the superheated liquid hinders heat transfer between the aggregated structures, eventually forming an unavoidable superheated zone at the bottom at Fo = 0.25. As the tilt angle is switched to 90°, the heat transfer process along the vertical direction is significantly suppressed. In this scenario, the melt front migrates more slowly than in the first two cases but is more uniformly distributed in the horizontal direction. Simultaneously, a severe overheating phenomenon occurs at the bottom due to the low thermal conductivity of the PCM and the absence of a rapid heat transfer path. A significant temperature gradient is also present along the heat transfer direction within the PCM domain, indicating a noticeable thermal insulation process.

Fig. 3: Simulated thermal buffering performance after regulation of magnetic field.
Fig. 3: Simulated thermal buffering performance after regulation of magnetic field.
Full size image

a Melt-front evolution and temperature distribution characteristics under different tilt angles. The results are presented at various Fourier number (\({{{\rm{Fo}}}}=\lambda t/\rho {c}_{p}{l}^{2}\)). b Evolution of average thermal resistance within the PCM region versus the Fourier number. c Comparison of widely adopted PCMs’ figure of merit (FOM) as well as the analytical thermal resistance. Source data are provided as a Source data file.

Furthermore, the average equivalent thermal resistance (Rave) through the PCM domain is implemented to quantitatively characterize the thermal diffusion performance in the PCM domain, which can be formulated as62:

$${R}_{{{{\rm{ave}}}}}=\frac{{T}_{{{{\rm{e}}}}}-{T}_{{{{\rm{u}}}}}}{{q}_{{{{\rm{e}}}}}}$$
(1)

where Te and Tu are the average temperature of the heat source and the upper surface; qe denotes the operating heat flux of the heat source. As summarized in Fig. 3b, it can be observed that although the increase in liquid fraction, which inhibits heat transfer during the melting process, causes the average thermal resistance to rise, the effect of magnetic field regulation remains highly noticeable. When rapid heat dissipation is warranted, aligning the direction of magnetic induction parallel to the boundary heat flux induces the generation of an efficient heat transfer network through nanoparticle aggregation. Compared to the original composite PCM, this adjustment alleviates thermal resistance within the PCM domains by 46.5% and 37.9% during the middle (Fo =  0.12) and terminal (Fo = 0.25) melting stages, respectively. As a result, the average thermal resistance through the PCM domain can be controlled below 1.1 × 10−3 m2 K/W. When the magnetic induction direction is deflected by 45°, the heat transfer network still demonstrates efficient heat diffusion at the bottom when Fo < 0.05. However, as the melting process advances, it exhibits slightly higher thermal resistance. This represents a transitional state in the thermal resistance regulation process, as it is not as effective in diffusing the heat in the vertical direction as in the case of the tilt angle = 0°. When fixing tilt angle = 90°, heat transfer along the vertical direction is significantly inhibited. The composite PCM with randomly dispersed nanoparticles exhibits a thermal resistance value close to that of the 90°-aligned configuration. This similarity stems from the limited vertical heat transfer efficiency in both systems. In the 90°-aligned case, horizontally aligned bundle-like aggregates are separated by a low-thermal-conductivity PCM matrix, which impedes efficient heat transfer in the vertical direction. Similarly, in the randomly dispersed composite, the tortuous and contact-resistant pathways within the anisotropic network restrict heat transfer in all directions. It is essential, however, to distinguish between vertical thermal resistance and in-plane heat spreading capability. Under the uniform bottom heat flux applied in our measurement system, the random network offers moderate, omnidirectional thermal enhancement. By contrast, the 90°-aligned configuration is specifically designed to enhance horizontal heat dissipation. This structure promotes temperature uniformity along the base and is particularly effective in mitigating localized hot spots under nonuniform heat loads, a common challenge in electronics thermal management. To ensure the robustness of our findings, a sensitivity analysis was conducted on the thermal conductivity assigned to the nanoparticle aggregates, a parameter that accounts for inter-particle contact resistance. As detailed in Supplementary Note 10, varying effective thermal conductivity over an order of magnitude (10–200 W m−1 K−1) exhibits a negligible impact on the calculated thermal resistance switching ratio of 1.8. This confirms that the observed dynamic tuning effect is governed primarily by the magnetically reconfigured architecture of the thermal pathways, rather than the absolute value of nanoparticle thermal conductivity, thereby solidifying the validity of our conclusions.

To comprehensively and quantitatively analyze the advantages of magnetic field regulation, considering various thermophysical parameters of the materials, the thermal buffering potential of regulated PCMs is compared with several commonly used PCMs63 using the figure of merit (FOM)64,65,66. It can be formulated as67:

$${{{\rm{FOM}}}}=\lambda \rho L$$
(2)

where λ, ρ and L represents the average thermal conductivity, density and latent heat of target PCM, respectively. As summarized in Fig. 3c, metal-based PCMs typically have FOMs in the order of 106 due to their relatively high thermal conductivity (>10 W m−1 K−1) and substantial volumetric latent heat, resulting in extremely low thermal resistance during heat transfer. Compared to them, the exhibited thermal resistance of organic PCMs is 100 times higher and with a FOM an order of magnitude lower. By incorporating CNTs with high thermal conductivity, the average thermal resistance of the composite PCM can be reduced to 1.2 m2 K/W, and the FOM is improved nearly 16-fold. With the introduction of magnetic field-induced regulation in this study, it is feasible to achieve dynamic tuning of the average thermal resistance by a factor of nearly 1.8. This approach effectively enhances the range of FOM variation, symbolizing tunable thermal buffering performance under dynamic operating conditions. This phenomenon can also be explained from a micro/nanoscale perspective, while the alignment of CNTs into bundle-like aggregates under the magnetic field creates low-thermal-resistance pathways for phonon propagation along the tube axis. Thus, the tilt angle directly governs the efficiency of phonon transmission across the composite by controlling the connectivity and orientation of the primary conduction paths.

Thermal insulation performance

For space equipment electronics, particularly antenna products, the high thermal power operating time often constitutes only 1/6 or less of the orbital operational cycle. Therefore, in the prolonged standby status within a low-temperature environment, the TMS must possess sufficient thermal insulation capabilities to mitigate the accumulation of thermal stress caused by extreme temperature differences. In this regard, the thermal insulation properties of magnetic field-regulated PCM are further investigated. The initial temperature within the PCM region is set as 60 °C, while the upper boundary is exposed to a constant temperature environment at −40 °C. Figure 4a illustrates the solidification process of liquid PCM and the temperature distribution within magnetic field-regulated PCM. It is evident that after fulfilling the heat dissipation requirements during high-power operation, while the tilt angle is 0°, the thermal buffer’s thermal insulation effect will be inadequate if the heat transfer performance is not dynamically adjusted. Driven by the significant temperature difference, the PCM at the top solidifies first, while the temperature of the bottom PCM near the electronics to be thermal shielded continues to decrease. As Fo exceeds 0.35, the thermal buffer only offers thermal insulation with sensible heat release of PCM, leaving the electronics vulnerable to a sharp temperature drop during extended standby cycles. On the other hand, setting the tilt angle to 90° during standby status will significantly extend the period for stable temperature control of the electronics. The horizontally distributed heat transfer network ensures a uniform heat release process of liquid PCM, maximizing the suppression of heat loss from the PCM domain to the low-temperature environment. As a result, the solidification process of liquid PCM proceeds very slowly, with half of the PCM remaining in the liquid state at Fo = 0.25. Even at Fo =  0.45, the temperature of the electronics remains stabilized around the melting point of the PCM.

Fig. 4: Simulated thermal insulation performance during low temperature status.
Fig. 4: Simulated thermal insulation performance during low temperature status.
Full size image

a Solidification process of liquid PCM and the temperature evolution within the PCM domain under different magnetic field regulation parameters. The results are presented at various Fourier number (\({{{\rm{Fo}}}}=\lambda t/\rho {c}_{p}{l}^{2}\)). b Transient temperature variation properties at the bottom of the PCM domain, which symbolizes the electronic device. c Average temperature drop rate of the electronic device under different regulation parameters. Based on the intensity of temperature variation, the process of thermal insulation is divided into three stages: hybrid heat release, latent heat release and sensible heat release. Source data are provided as a Source data file.

Then, the mean value of the instantaneous temperature at the bottom of the PCM domain is extracted in Fig. 4b, thus examining the anticipated temperature change of the electronics. It can be observed that the temperature change process of electronics can be divided into three statuses, differentiated based on the curve slope. The first is the hybrid heat release process, during which the temperature of the electronic device primarily depends on the latent heat released during the solidification of the top PCM and the sensible heat released during the cooling of the entire thermal buffer to maintain stability jointly. This is followed by a latent heat-dominated heat release process, marked by a leveling-off interval in the temperature profile. At this point, the temperature of the PCM in the vicinity of the electronics has been completely lowered to near the melting point, and the substantial latent heat ensures stable temperature control for the electronics. Finally, the sensible heat release phase begins once the PCM within the thermal buffer is fully solidified. During this stage, the temperature profile of the electronics typically exhibits a steeper decline compared to the first stage. When the tilt angle is fixed as 0°, the efficient heat transfer network causes a drastic temperature drop during the first stage. After sustaining the latent heat release process for just 0.15 of Fo, the temperature of the electronics begins to drop steeply at Fo = 0.28 and already reaches 0 °C by Fo = 0.36, which is unfavorable during the long-term standby period. When the tilt angle is adjusted to 45°, the temperature decrease trend of the electronic device slows down. However, the aggregated structure still partially guides the heat diffusion, and the stabilized temperature control time on the electronic device is not significantly extended. Adjusting the tilt angle to 0° during the standby status provides thermal protection that closely resembles the performance of the original composite PCM. The duration of the first stage is significantly extended, and the temperature drop is slowed down, the temperature of the electronics gradually attains 36 °C until Fo = 0.4. With a longer stable temperature control period, the electronics temperature eventually approaches 0 °C at Fo = 0.85, while the thermal insulation time is 136% longer compared to the case with 0° tilt angle. This mainly originates from a multi-faceted suppression of heat transfer. Primarily, the horizontally aligned nanoparticle bundles force heat to conduct through the low-thermal-conductivity PCM matrix between them, creating a highly tortuous and resistive vertical path. Additionally, the dense-packed nanoparticle aggregates act as intrinsic radiation shields, attenuating infrared radiation more effectively than the translucent pure PCM. The long-term structural integrity of the nanoparticle networks is ensured by a “structural freezing” mechanism during PCM solidification. The forming solid matrix physically locks the aligned aggregates in their configured orientation, effectively preventing irreversible agglomeration and preserving the pre-formed network for subsequent cycles. This mechanism underpins the exceptional cycling stability, quantitatively confirmed over 100 cycles (Supplementary Note 5).

Furthermore, the transient temperature control capability is characterized by the average temperature drop rate (\({v}_{{{{\rm{a}}}}}\)) of the electronic device, which can be formulated as:

$${v}_{{{{\rm{a}}}}}=-\frac{{T}_{{{{\rm{stop}}}}}-{T}_{{{{\rm{start}}}}}}{t}$$
(3)

where Tstart and Tstop represent the temperature at the beginning and end of an investigated period t. The temperature drop rates of the electronic device during the three thermal buffering stages are calculated based on different regulation parameters, providing a visual comparison of the thermal insulation performance. During the hybrid heat release stage, with tilt angle 0°, the temperature drop rate is elevated by 159% with a large span, while the duration is also shortened by 55.6% compared to the original composite PCM. As adjusting the tilt angle, the temperature drop rate can be alleviated by 42% (tilt angle 45°) and 57% (tilt angle 90°), eventually approaching the thermal insulation performance of the original composite PCM. After entering the latent heat dominated heat release stage, the temperature drop rate of electronics falls off a cliff to around 1 °C/s. At a tilt angle of 0°, the thermal buffer transitions into this stage the earliest and remains in it for the shortest duration, with the substantial latent heat being fully dissipated within a brief period. The results at a tilt angle of 90° are similar to those of the original composite PCM, and the temperature of the electronics is maintained essentially unchanged over Fo spans of up to 0.2. For the cases with tilt angle 45°, on the other hand, the inclination of the heat transfer network relative to the prevailing heat release direction results in inhomogeneities in the solidification process. Once the PCM inside the thermal buffer is fully solidified, the electronics will undergo the most drastic temperature drop in the sensible heat release process, though the temperature drop rate still diminishes gradually with reduced temperature gradient.

Dynamic tuning test

To evaluate the performance of our magnetic field-induced tuning method, a verification of concept experiment is conducted under dynamic operating conditions. The schematic of the experimental setup is depicted in Supplementary Fig. 15, and details are described in the “Methods” section. During the operating status, the heat flux is fixed as 5000 W/m2, and the cooling plate supplied by 5 °C cycled liquid is implemented as the low-temperature environment. The operating and standby periods are set as 15 and 75 min, respectively, maintaining a ratio of 1:5.

Figure 5a presents a comparison of infrared thermal images illustrating the temperature distribution within the thermal buffer throughout an operating cycle, contrasting the original composite PCM with the dynamic magnetic field regulation approach. During the work period, the bottom PCM completes heat absorption first, gradually reducing the heat load along the vertical direction within the thermal buffer. Heat dissipates away from the heat source, facilitated by the reinforcement of nanoparticles. However, after 10 min, when the bottom PCM is completely melted, a distinct superheated layer emerged at the bottom of the thermal buffer in the case of original composite PCMs. After completing the operation, the superheated liquid at the bottom does not dissipate heat as efficiently as it would under a large temperature gradient, potentially leading to overheating of the electronics. When applying magnetic field with tilt angle of 0°, the mesoscopic-scale heat transfer network functions effectively, preventing the formation of a superheated layer at the bottom of the thermal buffer at 10 min and ensuring a more uniform temperature distribution in the bottom melted region. Eventually, after the electronics cease heat generation at 15 min, a stepped temperature distribution forms within the thermal buffer, while the degree of overheating at the bottom is significantly alleviated. The larger volume of PCM at around 50 °C under the 0° tilt angle configuration indicates more effective heat distribution throughout the buffer, as the aligned network draws heat away from the source and utilizes the latent heat capacity of a greater PCM mass. In contrast, the nonregulated composite PCM exhibits heat localization and a severe overheating layer at the bottom, failing to utilize the full thermal storage capacity and leading to inferior performance. During the standby status, the direction of the magnetic induction line is adjusted to fulfill the thermal insulation effect. It can be observed that the temperature distribution within the thermal buffer aligns closely with that of the original composite PCM after altering the heat transfer network’s orientation. The melted PCM attains its melting point at 30 min, gradually releasing latent heat to ensure temperature stability. By 90 min, the temperature at the bottom of the thermal buffer closely approaches the initial temperature prior to the start of the cycle.

Fig. 5: Experimental evaluation of the magnetic field induced dynamic tuning method.
Fig. 5: Experimental evaluation of the magnetic field induced dynamic tuning method.
Full size image

a Infrared thermal images of the thermal buffer, which are recorded within one operating cycle. The results obtained when applying the original composite PCM are compared with those achieved through the dynamic magnetic field regulation approach. b Transient temperature variation properties of electronic devices through continuous operating cycles. Analyzed by contrasting the original composite PCM with the dynamic magnetic field regulation approach. c Comparison of temperature gradients along the mainstream heat transfer direction at different range of locations within the thermal buffer. d Analysis of local thermal resistance of PCM in the range of 5–10 mm from the heat source under different conditions. Source data are provided as a Source data file.

In addition, the transient temperature variations of electronic devices during continuous operating cycles are analyzed in Fig. 5b. After the device starts generating heat, its temperature continues to rise. Then, after exceeding the melting point of PCM, the slope of the temperature curve is lowered due to the latent heat of fusion. It is revealed that the composite PCM with randomly dispersed nanoparticles exhibits a faster temperature rise than the sample with magnetic field regulation. This is attributed to the difference in the effective thermal diffusivity of the system along the vertical direction. The composite PCM possesses lower vertical thermal diffusivity, causing heat to accumulate locally at the bottom and resulting in a more rapid local temperature rise, which indicates inferior heat dissipation performance. In contrast, 0°-aligned sample rapidly distributes heat away from the source due to its high-diffusivity vertical pathways, thereby moderating the source’s temperature increase. At the end of the first cycle, the maximum temperature is reduced by 6.7 °C compared to the original composite PCM. Upon entering the standby phase, driven by the temperature gradient inside the thermal buffer, the electronics temperature initially undergoes a rapid decline. After the temperature drops to near the PCM melting point, the release of latent heat significantly contributes to temperature stabilization, leading the electronic device to enter a phase of relatively moderate temperature reduction. With the timely adjustment of the magnetic field orientation, the temperature variation of the electronics during the standby status resembles that of the composite PCM, and even ends the cycle with a slightly higher temperature, exhibiting smaller temperature fluctuations. In subsequent cycles, the maximum temperatures reached during the operation are consistently reduced, influenced by the continuous chilling effect of the cooling plate. From this point of view, excessive cooling during the standby status also emerges as a significant concern. Owing to the dynamic magnetic field regulation approach, which enhances heat diffusion within the thermal buffer during the heat generation status and ensures heat retention during the low-temperature standby status, the temperature fluctuation of the electronic device within one cycle is minimized by 10.8 °C compared to the composite PCM. According to the first-order analysis conducted in Supplementary Note 14 robustly confirms that the ~11 °C improvement observed experimentally is physically sound and aligns well with the fundamental thermophysical principles underlying our tuning strategy.

Further analysis of the temperature gradient between different regions (Fig. 5c) reveals that at 6 min, after the bottom PCM fully melts, the magnetic field regulation reduces the temperature gradient within the range of 5–10 mm from the heat source, effectively guiding the heat to diffuse away from the heat source. On the contrary, the composite PCM fails to achieve rapid heat diffusion inside the PCM even under larger temperature difference, highlighting its limitations in the thermal buffering process. The PCM in the range of 10–15 mm remains solid state, and since the magnetic field regulation cannot function, heat transfer within this portion of the PCM is only weakly affected.

To quantitatively analyze the influence of the mesoscopic heat transfer network formed by the dynamic magnetic field regulation on the macroscopic heat transfer process, the local thermal resistance (R) in the range of 5–10 mm is calculated using the transient thermal analysis method68.

$$R=\frac{\Delta {T}_{5-10}}{q}=\Delta {T}_{5-10}\cdot \frac{A}{{c}_{P}m}\cdot \frac{\partial t}{\partial T}$$
(4)

where ΔT5–10 presents the temperature difference between the location of 5 and 10 mm, A and m are respectively the heat transfer area and the PCM mass through this range, and the local heat flux q is dominated by the instantaneous temperature change \(\frac{\partial T}{\partial t}\) at 5 mm. To avoid the influence of latent heat of fusion on calculating heat flux, the local thermal resistance is selected for analysis during the period following PCM melting process and prior to the solidification process during standby status (Fig. 5d). For the original composite PCM, the presence of liquid PCM with low thermal conductivity leads to a continuous and rapid increase in thermal resistance within the 5–10 mm range, which is detrimental to the rapid heat dissipation process under high heat flux operation. Although the rate of increase in thermal resistance is gradually mitigated by the elevated temperature gradient, the overall heat dissipation efficiency remains compromised. After applying the magnetic field, the induced formation of mesoscopic-scale heat conduction networks exhibits a certain lag in its effect on reducing local thermal resistance at the macroscopic scale. This delay highlights the time required for the network to fully develop and influence the overall heat transfer performance within the PCM domain. At 10 min, the local thermal resistance is reduced by approximately 48% compared to the composite PCM. Despite the following tendency of rising as the melting progresses, the enhancement of heat transfer through magnetic field regulation is evident, demonstrating its significant impact on local thermal management performance. When the electronic device enters the standby status, adjusting the magnetic field tilt angle to 90° causes the thermal resistance in the 5–10 mm range to gradually increase, eventually reaching a level comparable to that of the composite PCM. At this point, the system begins to function effectively as thermal insulator, assisting to stabilize the temperature of the electronic device by reducing heat dissipation.

The dynamic tuning test was designed to emulate a demanding real-world application: thermal management for electronics in low-earth orbit. The 15-min operational and 75-min standby cycle with a 5 °C cold satellite’s internal environment mirrors a standard orbital duty cycle. The demonstrated 10.8 °C reduction in device temperature fluctuation under these conditions, achieved by switching between 0° (operation) and 90° (standby) configurations, validates the practical utility of our strategy in stabilizing temperature in dynamic, intermittent systems. The long-term reliability of the dynamic tuning system was rigorously validated over 100 thermal cycles. Post-cycling characterization confirmed the exceptional stability of the composite PCM, with negligible changes in latent heat (−1.4%) and thermal conductivity (−6.5%), and no degradation in the saturation magnetization of the Fe3O4 nanoparticles (Supplementary Note 4). Coupled with the quantitative analysis of the repeatable nanoparticle aggregation structures (Supplementary Note 5), these results underscore the robustness of the material for long-term service in adaptive thermal management.

Methods

Preparation of magnetic composite PCM and property measurement

Fe3O4@CNT nanoparticles are synthesized using the two-step method to serve as a controllable medium for enhancing heat transfer, while the mixing ratio between Fe3O4 and CNT are also considered. These nanoparticles are then mixed with melted n-eicosane to create magnetic composite PCM with stable dispersion, ensuring effective and consistent thermal performance. The surface morphology of prepared PCM is characterized with a (SEM) (VE9800S, Japan) operating at 10 kV. XRD patterns of different materials are obtained through an X-ray diffractometer (XRD) (Bruker D8 ADVANCE, Germany). The thermal characteristics of the composite PCM are assessed using DSC (DSC250, USA). During the testing process, the samples are heated from 0 to 60 °C, with the temperature increase rate set to 5 °C/min. The thermal diffusivity α of composite PCM is analyzed through a laser thermal conductivity meter (LFA447, Germany) at 20 °C, and the corresponding thermal conductivity can be calculated as \(\lambda=\alpha \rho {c}_{p}\) with an accuracy of ±3%.

Magnetic field regulation and mesoscopic characterization

The magnetic field is controlled using an eight-pole electromagnet, enabling precise adjustments in both the magnetic field strength and the orientation of the magnetic induction lines. Under the influence of a uniform magnetic field, magnetic nanoparticles align along the direction of the magnetic induction lines, driven by the combined effects of their magnetic moments and molecular cohesion. This alignment leads to the formation of uniformly distributed, bundle-like nanoparticle structures. However, the efficacy of the magnetic field regulation was found to be robust across a range of field strengths. A systematic investigation (Supplementary Note 11) revealed that the thermal resistance switching ratio saturates at its maximum value of 1.8 for magnetic field strengths above ~150 mT, indicating performance saturation. This saturation is attributed to the complete alignment of the magnetic moments of the Fe3O4 nanoparticles and the consequent full structural assembly of the CNT aggregates. In this regard, all results presented herein were obtained at 200  mT, well within this saturation regime, ensuring consistent and optimal performance. The final aggregation morphology, resulting from the magnetic field regulation, is characterized at the mesoscopic scale using a digital microscope, providing detailed insights into the structural configuration.

Measurement of sample thermal conductivity after magnetic field regulation

The effective thermal conductivity of the composite PCM under varying magnetic field orientations was quantified using a transient plane source method with a Hot Disk TPS 2500S system. Liquid composite PCM were prepared and encapsulated within a disk-shaped container (inner diameter: 30 mm, inner depth: 5 mm) in a temperature-controlled chamber. A uniform magnetic field was applied using the above-mentioned electromagnetic system. The field strength was set to 200 mT, and its spatial homogeneity was verified with a Hall-effect gaussmeter. The sample was positioned between the pole pieces, and the magnetic field vector was precisely oriented relative to the sample’s cylindrical axis (defining the primary heat flux direction) at angles of 0°, 45°, and 90°. For each orientation, the sample was exposed to the field for at least 20 s prior to ensure the formation of a stable nanoparticle aggregation structure. Subsequently, a liquid-cooled plate was employed at the bottom to rapidly and uniformly solidify the sample after magnetic field regulation. The hot disk sensor (Kapton-insulated, radius 2 mm) was sandwiched between two identical sample pieces, ensuring that the sensor plane was perpendicular to the applied heat flux. The magnetic field was applied in-plane (for 0° and 45° configurations) or through-plane (for 90° configuration) relative to the sensor. Each measurement consisted of a small, regulated power pulse through the sensor, and the subsequent temperature rise was recorded. The resulting thermal conductivity was derived by analyzing the temperature-time response using the hot disk proprietary software, which is based on the solution to the transient heat conduction equation for an anisotropic medium. For each configuration, a minimum of five measurements were performed, and the reported data represent the mean value. This methodology directly probed the directional heat transfer capability imparted by the magnetic field-induced nanostructures, providing definitive evidence of the tunable thermal anisotropy.

Mesoscale simulation by lattice Boltzmann method

The schematic diagram of the mathematical model, which illustrates the impact of magnetic nanoparticle aggregation on the solid-liquid phase change process, is presented in Supplementary Fig. 9. The transient phase transition behavior of the PCM is analyzed using a two-dimensional total enthalpy-based model, with the corresponding energy governing equation expressed as follows57:

$${\partial }_{t}({\rho }_{0}{c}_{p}T)+\nabla \cdot ({\rho }_{0}{c}_{p}T{{{\bf{u}}}})=\nabla \cdot (\lambda \nabla T)-{\partial }_{t}({\rho }_{0}L{f}_{l})-\nabla \cdot ({\rho }_{0}L{f}_{l}{{{\bf{u}}}})$$
(5)

where ρ, cp, L and T represent density, specific heat, latent heat and transient temperature of PCM, respectively. These functions are discretized using a two-dimensional, nine-velocity (D2Q9) discrete velocity set. Consequently, the total enthalpy distribution function (\({g}_{i}({{{\bf{r}}}},t)\)) for the ith moment at lattice time t and position r can be formulated as69:

$${g}_{i}({{{\bf{r}}}}+{{{{\bf{e}}}}}_{i}\delta t,t+\delta t)={g}_{i}({{{\bf{r}}}},t)-\frac{1}{{\tau }_{g}}\left[{g}_{i}({{{\bf{r}}}},t)-{g}_{i}^{{{\mathrm{eq}}}}({{{\bf{r}}}},t)\right]$$
(6)

where \({\tau }_{g}\) denotes the collision time, with the collision process modeled using the BGK collision operator. the density distribution function (\({f}_{i}({{{\boldsymbol{r}}}},t)\)) with collision term and streaming term as well as the buoyancy force item (\({F}_{i}\)) can be formulated as69:

$${f}_{i}({{{\bf{r}}}}+{{{{\bf{e}}}}}_{i}\delta t,t+\delta t)={f}_{i}({{{\bf{r}}}},t)-\frac{1}{{\tau }_{{{{\rm{f}}}}}}\left[{f}_{i}({{{\bf{r}}}},t)-{f}_{i}^{{{\mathrm{eq}}}}({{{\bf{r}}}},t)\right]+\delta t{F}_{i}$$
(7)
$${F}_{i}={\omega }_{i}(1-\frac{1}{2\tau })(\frac{{{{{\bf{e}}}}}_{i}-{{{\bf{u}}}}}{{c}_{s}^{2}}+\frac{{{{{\bf{e}}}}}_{i}\cdot {{{\bf{u}}}}}{{c}_{s}^{4}}{{{{\bf{e}}}}}_{i})\cdot \rho {{{\bf{g}}}}\beta (T-{T}_{{{{\rm{m}}}}})$$
(8)

where cs, β and Tm denote the lattice speed of sound, thermal expansion coefficient and melting point of PCM.

Dynamic tuning test

The performance of the magnetic field-induced regulation method is tested under dynamic operating conditions within a room temperature environment (20 °C). The testing setup is shown in Supplementary Fig. 15, allowing for the evaluation of thermal buffering capabilities of the magnetic composite PCM under realistic thermal cycling scenarios, simulating conditions similar to those encountered in practical applications. The thermal buffer has inner dimensions of 60 × 60 × 20 mm, with acrylic plates placed on the sides for visualization purposes. Two flat copper plates, manufactured by laser cutting, are used as the bottom and top components to ensure effective heat transfer throughout the system. The film heater is affixed to the bottom copper plate to simulate the heat generation of an electronic device. The heating power is controlled by a direct-current regulated power supply, with a maximum thermal power of 100 W. The cooling plate is connected to a water-cooling bath, which provides constant water of 5 °C with a maximum refrigeration power of 1200 W. The bakelite box, with a thermal conductivity of less than 0.2 W m−1 K−1, is used as the outer shell of the experimental region to minimize heat loss. A uniform D.C. magnetic field with a strength of 200 ± 10 mT was generated by an eight-pole electromagnet. The spatial uniformity of the field across the sample volume was verified with a Gaussmeter. The entire electromagnet assembly is controlled by a four-channel DC reversing programmable power supply. The orientation of the magnetic field vector relative to the primary heat flux direction (defined as the tilt angle, e.g., 0° or 90°) was dynamically controlled by regulating the current through the electromagnet group, while the heat source and cooling plate remained fixed. All results presented in Fig. 5 were obtained using a single sample of the composite PCM with a Fe3O4@CNT concentration of 0.2 wt% and a mixing ratio of 1. The transient temperature at various locations within the experimental system is measured using T-type thermocouples (Omega), which offer an accuracy of 0.1 °C. The data from these thermocouples is collected and recorded using a 40-channel data acquisition system (Yokogawa, GM90PS, Japan) with the acquisition period of 0.1 s. An infrared camera (FLUKE, TV40, America) with an accuracy of ±1% for temperatures ranging from 20 to 150 °C, is fixed to record the temperature distribution within the thermal buffer.