Negative temperature coefficient metamaterial absorber for uniform microwave heating Authors: Jing Zhou, Di Li, Lingfeng Xia, Xicheng Heng, Shuting Liu, Guido Link, James Gao, Xiaozhong Hao, Lei Zhou, Yingguang Li Published: 2026 ABSTRACT AbstractMicrowave heating is widely used in daily applications but is fundamentally limited by non-uniform temperature distribution. Despite extensive efforts to manipulate energy distribution around materials, achieving uniform heating remains elusive due to the intrinsic inhomogeneity of electromagnetic fields. Here, we report a self-regulating solution that adaptively modulates the absorbance distribution using a Negative Temperature Coefficient (NTC) Metamaterial Absorber (MA). Distinct from electromagnetic field-shaping strategies, our approach intrinsically suppresses overheating in high-temperature regions and redistributes energy to cooler areas. We demonstrate, for the first time, that uniform heating—quantified by over 90% reduction in the coefficient of variation—can be achieved across diverse configurations, including planar, polyhedral, curved, and multiple objects, as well as under power variations spanning two orders of magnitude. This work not only provides a theoretical resolution to the longstanding challenge of non-uniform microwave heating but also opens new avenues for development and application of temperature-adaptive metamaterials.Subject terms: Structural materials, Electrical and electronic engineering Introduction Since the 1960s, microwave heating technology has emerged as a promising thermal processing method due to its advantages of rapid, volumetric, and selective heating1–4. Today, its applications have expanded into various fields closely tied to everyday life, such as food processing5,6, chemical catalysis7,8, materials synthesis or processing9–11, waste treatment12,13, environmental engineering14,15, energy industry16,17, and biomedical applications18,19. However, for decades, longstanding challenges arising from inherent non-uniform heating patterns have not only affected user experience in existing applications but also limited the technology’s potential in advanced applications20–22. To address this issue, numerous efforts have been devoted to manipulating the electromagnetic field around materials23–30. Representative strategies—including rotating turntables, mode stirring, and conveyor-assisted motion averaging—have achieved practical success in applications with moderate uniformity requirements and cost constraints. Nevertheless, realizing a genuinely uniform microwave field remains intrinsically challenging, as microwave antennas exhibit specific radiation patterns, while resonant cavities support discrete modes of microwaves, both of which naturally lead to spatial variations in field intensity rather than a uniform distribution. Theoretically, the temperature distribution is governed by the interaction between the material and the electromagnetic field, thereby retaining the possibility of controlling the temperature field via the material’s absorption characteristics. Motivated by this perspective, here we investigate the feasibility of improving temperature uniformity by dynamically tailoring the spatial absorbance distribution. In recent years, metamaterial absorbers (MAs) have attracted widespread interest due to their flexibly tunable absorption properties31–37. A typical MA structure consists of a layer of patterned metallic units and a conductive substrate, separated by a dielectric spacer. Its microwave absorption can be easily tuned by engineering the characteristics of constituent materials and by adjusting geometric parameters38–47. Building on these unique advantages, we propose a negative temperature coefficient (NTC) MA capable of adaptively redistributing microwave absorption during heating, thereby enabling self-regulated homogenization of the temperature field. A detailed description of the design framework and operating mechanism of the NTC MA is provided in Supplementary Note 1. As schematically illustrated in Fig. 1a, the MA used in this work also features a sandwich structure and exhibits narrowband absorption, with a resonant frequency of fres. The NTC absorption is achieved by designing the temperature-dependent permittivity (εr) of the dielectric spacer. According to the equivalent circuit model48, the relationship between fres and εr satisfies fres∝1/εr. Thus, as long as εr changes monotonically with temperature T, fres will also vary monotonically with T. In this work, as shown in Fig. 1b, εr is designed to decrease with T, and thus fres is positively correlated with T. Initially, when microwave power is applied at a frequency of fwave1, the absorbance reaches nearly 100% at a lower temperature (T1), but gradually decreases to a very low level at a higher temperature (T2), thereby achieving NTC microwave absorption. If we want to further heat the material, the frequency of the microwave power can be adjusted to fwave2, which matches the MA’s resonant frequency at T3, thereby restoring strong microwave absorption for the subsequent NTC heating process. Schematic illustration of the NTC MA and its contribution to improving microwave heating uniformity. a The adopted sandwich MA unit with a thermosensitive dielectric spacer. b Realization of NTC absorption through temperature-dependent absorption peaks. c Uniform microwave heating resulted from the adaptively complementary absorbance distribution. d Preferential energy redistribution to cooler areas under the selective heating effect of microwaves. e Comparison of heating uniformity in both simulation and experiment between MAs with constant absorption and NTC absorption. f Programmable pattern-selective uniform heating enabled by NTC MAs. Through NTC microwave absorption, a self-regulating mechanism can be achieved, as illustrated in Fig. 1c. When exposed to an uneven electromagnetic field, a non-uniform temperature field is generated, which activates the MA’s NTC absorption response. Specifically, regions with higher temperatures exhibit reduced absorption, while cooler areas retain higher microwave absorbance, adaptively generating a complementary absorbance distribution. As a consequence, the temperature distribution across the MA will eventually become uniform. Furthermore, in microwave cavities, the selective heating nature of microwaves preferentially redirects energy toward higher absorbance regions (i.e., cooler areas), further accelerating the convergence toward a uniform temperature distribution (Fig. 1d). As shown in Fig. 1e, uniform heating has been achieved in both simulations and experiments for NTC MAs, even when some faces of the polyhedron are oriented away from the wave port. This stands in stark contrast to conventional MAs, which exhibit constant microwave absorption. Fig. 1f further demonstrates the potential of the NTC MA for programmable, pattern-selective uniform heating. It can be observed that the temperature distribution sequentially forms the Chinese characters “一”, “十”, and “田” (meaning “one,” “ten,” and “field,” respectively). At each stage, uniform temperature within the target pattern is maintained, owing to the NTC absorption effect. Practically, the proposed NTC MA can be integrated with both conductive and dielectric target materials, functioning either as the conductive substrate or as part of the dielectric spacer, thereby enabling adaptive regulation of local microwave absorption while partially preserving the intrinsic advantages of volumetric microwave heating. More generally, the NTC MA may also be implemented as an ultrathin functional film laminated onto object surfaces for indirect heating applications. In this configuration, electromagnetic energy is primarily absorbed and redistributed by the NTC MA layer rather than by the object itself, thus avoiding object-specific electromagnetic redesign and enabling compatibility with a broad range of materials, including those with intrinsically weak microwave absorbance. As a complementary strategy to established field-engineering approaches such as rotating turntables and mode stirring, the proposed absorption-control strategy may be particularly valuable in advanced manufacturing scenarios where stringent temperature uniformity is essential, including high-performance fabrication of composite structures, high-precision processing of dielectric components, and high-value treatment of biomedical products. In the following sections, we first investigate two primary mechanisms for improving temperature uniformity: (i) suppressing temperature rise in hot regions (high field strength) and (ii) enhancing temperature rise in cold regions (low field strength). Guided by these mechanisms, we design an NTC MA and validate its effectiveness in improving temperature uniformity through both simulations and experiments. Finally, we present a representative application and a limit analysis of the NTC MA. Results Theoretical analysis of temperature distribution influenced by NTC absorption Theoretical analysis was first conducted on a series of representative NTC absorbance cases. In classical electromagnetism, information about microwave absorbance is inherently embedded within physical parameters such as material thickness, complex permittivity, and complex permeability49,50, making it difficult to directly isolate and analyze the impact of the absorbance distribution. In this study, we establish an explicit mathematical relationship among the temperature field T(r→), absorbance distribution A(r→), and external electromagnetic field E0(r→) or H0(r→), under the assumption that microwaves are incident on the top surface (at arbitrary point r→) of the thin NTC MA, with no transmission through the bottom surface and negligible edge effects. The corresponding derivation is provided in Supplementary Note 2. Accordingly, the temperature field is determined by both the absorbance distribution and the external electromagnetic (EM) field from the MA. However, in practical multimode cavity applications, variations in material absorbance inherently cause EM field redistribution, which poses significant challenges for isolating the effects of NTC characteristics on the temperature field. To address this complexity, a two-stage analysis is adopted in this study. In the first stage, the effects of EM field redistribution are eliminated to investigate the influence of fundamental NTC characteristics. In the second stage, the impact of NTC absorbance on EM field variation (i.e., the EM field redistribution effect) is examined. Influence of pure NTC absorption characteristics. The absorbance-temperature (A–T) profile can be characterized by three NTC features: (i) the initial absorbance A0, (ii) the decline amplitude ΔA=A0−A1, and (iii) the decline pattern dA/dT (e.g, fast-to-slow or slow-to-fast), where A1 denotes the final absorbance at the target temperature. A theoretical analysis was performed to investigate the influence of these three NTC characteristics on the evolution of the temperature field, assuming a fixed power density ratio of 8:1 between the high-power and low-power regions. Meanwhile, the coefficient of variation (COV, defined as the standard deviation over the mean) was employed to quantify the temperature uniformity29. The detailed mathematical framework for several representative cases supporting this analysis is provided in Supplementary Note 3. The impact of the initial absorbance is investigated in Fig. 2a. The initial absorbance A0 at room temperature (around 25 °C) is set as 1, 0.66 and 0.33, while the absorbance decreases linearly with a fixed decline amplitude ΔA of 0.33. It can be observed that the temperature profiles remain divergent throughout the process as long as the final absorbance A1≠0. However, in the case of A0=0.33 and A1=0, full temperature convergence is achieved. The COV initially increases due to the uneven power distribution, but then drops rapidly as the temperature in the high-power region approaches the target temperature, triggering a self-limiting heating behavior and allowing the low-power region to catch up. Fig. 2b depicts the result for different absorbance decline amplitudes. Similarly, full convergence can only be achieved when A1=0. Compared with the A0=0.33 group in Fig. 2a, the temperature convergence is significantly accelerated, with the time required for full convergence reduced from 3150 s to 1185 s, indicating that a large absorbance decline amplitude is favorable for achieving temperature uniformity. Fig. 2c shows the result for different absorbance decline patterns, with the initial and final absorbance fixed at A0=1 and A1=0, respectively. It can be seen that, as the area enclosed by the A–T curve and the coordinate axes increases from Case 1 to Case 3, the temperature convergence rate progressively accelerates, with the time required for full convergence decreasing from above 10,000 s to 780 s. This implies that, in the absence of EM field variations, the optimal A–T curve would ideally exhibit a stepwise change from 1 to 0 at the target temperature. Influence of pure NTC absorption on heating uniformity. a Influence of the initial absorbance. b Influence of absorbance decline amplitudes. c Influence of absorbance decline patterns. Taking the EM field redistribution effect into account. For multimode microwave cavities, which are widely used in practical applications, EM field redistribution—induced by variations in microwave absorption—plays a crucial role in shaping the temperature distribution. As illustrated in Fig. 3a, when two heating targets with identical microwave absorbance are placed in two different regions of the cavity, the absorbed power in, for example, Zone 2 may be significantly higher than that in Zone 1 due to the extremely uneven power distribution. Consequently, the temperature in Zone 2 rises much faster than that in Zone 1. With continued heating, the NTC effect causes the absorbance of Zone 2 to decrease to a low level, giving rise to a reflective behavior, whereas the absorbance in Zone 1 remains high and thus sustains strong microwave absorption (see Fig. 3b). In this case, the energy reflected from Zone 2 will preferentially redistribute toward Zone 1, further accelerating the temperature convergence between the two regions. It should be noted, however, that part of the reflected energy may also propagate back toward the microwave source through the wave port. This issue can be effectively mitigated by employing an asymmetric waveguide configuration51,52. EM field redistribution induced by NTC absorption in multimode cavities. a Schematic of uneven power distribution in the microwave cavity. b Power redistribution induced by variations in material absorbance. c Simulation setup and electric field magnitude in the dielectric spacer for different absorbance distributions. d Absorbed power of individual MA units under varying absorbance distributions. e Average power absorbed by units in Zone 1 and Zone 2. f Reflection loss (S11) at the wave port and normalized load loss. To quantitatively investigate the EM field redistribution effect, we performed a series of simulations using MAs whose microwave absorbance can be flexibly tuned by tailoring the properties of the constituent materials and the geometric parameters of the structure. The methodology involved deliberately reducing the absorbance of the high-power region while maintaining the high absorbance of the low-power region, followed by monitoring the absorbed power of each MA unit. Under such conditions, any variation in absorbed energy within the low-power region originates solely from changes in its local EM field. As shown in Fig. 3c, a planar MA composed of nine units was placed at the center of a cubic microwave cavity with a side length of 300 mm. Initially, the absorbance of each MA unit was set to A=0.95, and the resulting absorbed power distribution—represented by the electric field magnitude in the dielectric spacer—reveals distinct low-power (Zone 1) and high-power (Zone 2) regions. Subsequently, four additional simulations were carried out by decreasing the absorbance in Zone 2 (A2) to 0.75, 0.40, 0.20, and 0.05, while keeping the absorbance in Zone 1 (A1) fixed at 0.95. More simulation details are provided in Supplementary Note 4. It can be seen from Fig. 3c, d that as A2 decreases from 0.95 to 0.05, the power absorbed by each unit in Zone 2 decreases significantly, while the energy absorbed by the units in Zone 1 increases markedly, indicating a pronounced EM field redistribution effect. As shown in Fig. 3e, the case with A2≈0.6 represents a critical state. Before this point, Zone 2’s unit absorbs more power than Zone 1’s, leading to a faster temperature rise and thus a pronounced temperature difference between the two regions. Beyond this point, however, Zone 2’s unit absorbs less power than Zone 1’s, resulting in a slower temperature increase that helps reduce the temperature difference. Remarkably, when A2 is reduced to 0.20, the average power absorbed by the unit in Zone 2 drops to about 12% of its original value, while that in Zone 1 increases to approximately 7.2 times its initial value due to the EM field redistribution effect. This phenomenon is particularly advantageous, as it demonstrates that complete suppression of microwave absorption in the high-power region is unnecessary—a substantial reduction in absorbance (e.g., from 1 to 0.20) is sufficient to trigger strong field redistribution, which is more feasible for practical implementation. However, it can also be observed that the total power absorbed by the whole MA (i.e., normalized load loss) decreases as A2 decreases, accompanied by increasing reflection of power back to the microwave source (see Fig. 3f). Quantitatively, when A2 decreases to 0.20, approximately 35% of the incident energy is reflected back to the source. The feasibility of reducing wave-port reflection via an asymmetric waveguide is demonstrated in Supplementary Note 5. Validation of enhanced heating uniformity through NTC absorption A series of simulations, followed by experimental validation, were conducted to validate the effectiveness of the proposed method by comparatively analyzing two representative cases: one involving a constant MA as a benchmark (Scenario 1), and the other utilizing an NTC MA developed in this work (Scenario 2). Guided by the theoretical framework above, the desired A–T response was engineered through an NTC MA incorporating a thermosensitive dielectric spacer and careful selection of the operating microwave frequency (see Fig. 4a). The ideal step-like A–T profile was not adopted, as abrupt phase-change dielectric spacers—such as vanadium dioxide—feature intrinsically low phase-transition temperatures, which currently limit their applicability. In addition, the residual absorbance (9.3%) of the NTC MA at the target temperature originates from the intrinsic loss of the dielectric spacer, which, as discussed above, does not undermine the heating uniformity owing to the assistance of the EM field redistribution effect. Detailed simulation settings can be found in Supplementary Note 4. Simulation validation of enhanced heating uniformity enabled by the NTC MA. a A–T curves adopted in simulations for Scenario 1 and Scenario 2 in the planar configuration. b, c Temperature profiles of all MA units under Scenario 1 (b) and Scenario 2 (c). d Heating rates of 9 representative units located in one quarter of the MA under Scenario 2. e Temporal evolution of the electric field distribution (magnitude) inside the MA’s dielectric spacer under Scenario 1 and Scenario 2. f Temporal evolution of the MA’s surface temperature under Scenario 1 and Scenario 2. g A–T curves adopted in simulations for Scenario 1 and Scenario 2 in the cubic configuration. h, i Temperature profiles of all units on the cubic MA under Scenario 1 (h) and Scenario 2 (i). j Heating rates of all units on the cubic MA under Scenario 2. k Geometry and spatial orientation of the cubic MA. l, m Temperature distributions on different cube faces under Scenario 1 (l) and Scenario 2 (m). The simulation results for a planar MA with 5 × 5 units are presented in Fig. 4b–f. As observed, Scenario 1 exhibits continuously divergent temperature profiles due to the uneven electromagnetic field, while Scenario 2 shows profiles that initially diverge but gradually converge toward the target temperature, achieving a reduction of 92% in COV. Notably, in contrast to the non-intersecting temperature profiles depicted in Fig. 4b, the temperature curves in Fig. 4c exhibit pronounced intersections. This behavior arises because, unlike the static and uneven heating pattern in Scenario 1, Scenario 2 features an adaptive heating pattern (see Fig. 4d), where local heating rates dynamically adjust in response to temperature evolution. More intuitively, as shown in Fig. 4e, f and Supplementary Movie 1, the uneven and static electromagnetic field pattern under Scenario 1 leads to an increasingly non-uniform temperature distribution. In contrast, Scenario 2 exhibits a dynamic redistribution of electromagnetic fields, contributing to almost uniform temperature distribution—even though the initial power distribution across the planar NTC MA units spans two orders of magnitude (1:267). Fig. 4g–m present the simulation results for the MA with a cubic configuration. As shown in Fig. 4g, the A–T profile employed here resembles that in Fig. 4a, but with a different target temperature, highlighting the design flexibility of the proposed approach. Fig. 4h, i provide a comparative analysis of the simulated temperature profiles for all units distributed across the six faces under Scenario 1 and Scenario 2, revealing trends consistent with those observed for the planar MA. Quantitatively, a 94% reduction in COV is achieved at t=47 s, confirming the effectiveness of the adaptive heating pattern (Fig. 4j). To visualize the thermal evolution on all six faces, simulated temperature distributions are shown from two distinct viewing perspectives (Fig. 4k). As illustrated in Fig. 4l, Scenario 1 yields time-invariant and highly non-uniform thermal patterns, with certain faces—particularly those oriented away from the microwave port (i.e., View 2)—remain substantially cooler throughout the process. In sharply contrast, Scenario 2 exhibits dynamic thermal evolution, resulting in significantly improved temperature uniformity, as depicted in Fig. 4m and Supplementary Movie 2. Furthermore, Supplementary Note 6 demonstrates similar findings for MAs with a dodecahedral configuration, where a two-stage NTC strategy (Scenario 2 Plus) is also validated via microwave-frequency tuning. In addition, demonstrations on MAs with a dielectric target object and in cavities with different sizes and geometries have been conducted. Collectively, these results substantiate the capability of the proposed method to effectively handle complex situations and accommodate multi-stage heating requirements encountered in practical applications. Fig. 5a presents a photograph of the fabricated NTC MA, in which a key advancement was the thermo-responsive permittivity engineering of the dielectric spacer. The dielectric spacer consists of two materials: polyimide films (thickness N1d1), whose relative permittivity ε1 remains stable from room temperature to 120 °C, and organosilicon polymer films (thickness N2d2), whose relative permittivity ε2(T) decreases monotonically with temperature over the same range (Fig. 5b). According to effective medium theory, several NTC MAs with different thickness ratios (α) of organosilicon polymer films were fabricated, and their temperature-dependent absorption properties were measured. Further details regarding the design of the thermosensitive dielectric spacer and the microwave-absorption measurement setup are provided in Supplementary Note 7. As shown in Fig. 5c, temperature-dependent absorption peaks with varying resonant frequency shift ranges were realized by tailoring the value of α. A constant MA with stable microwave absorption was fabricated when α=0%, while two NTC MAs with resonant frequency shift ranges of 0.09 GHz and 0.30 GHz were obtained at α=17% and α=50%, respectively. Fig. 5d displays the measured A–T curves of the constant MA at 2.45 GHz (Scenario 1, α=0%) and the NTC MA at 2.43 GHz (Scenario 2, α=17%), which align closely with the results observed in the simulation (Fig. 4a). In addition, Fig. 5e shows the A–T curve of the NTC MA with α=50%, where two stages of NTC absorption are demonstrated by adjusting the microwave frequency from 2.40 GHz to 2.50 GHz at the beginning of the second stage (Scenario 2 Plus). Sample fabrication and measurement. a Photograph of a fabricated NTC MA composed of thermostable (polyimide) and thermosensitive (organosilicon polymer) dielectric layers. b Temperature-dependent permittivity of the polyimide and organosilicon polymer films. c Temperature-dependent absorbance-frequency spectra of fabricated MAs with different α values. d Measured A–T curves for Scenario 1 and Scenario 2. e Measured A–T curves for Scenario 2 Plus, showing two-stage NTC absorption at different frequencies. Microwave heating experiments were conducted on the fabricated MAs in various configurations, with detailed experimental settings provided in Supplementary Note 8. The results for the planar MA are presented in Fig. 6a–e. It can be observed that the experimental results exhibit excellent agreement with the simulations. Under Scenario 1, a distinctly non-uniform and fixed heating pattern is observed in both the infrared thermal images and the temporal temperature profiles, along with the corresponding heating rates derived from these profiles. In contrast, Scenario 2 demonstrates adaptive thermal regulation, characterized by converging temperature trajectories, fluctuating heating rates, and dynamic temperature distributions. As a result, a 94% reduction in COV is achieved at t=6 min compared to Scenario 1. A visual comparison of the complete thermal dynamics between these two scenarios is provided in Supplementary Movie 3. It should be noted, however, that the proposed strategy inevitably reduces heating efficiency compared with conventional microwave heating. To ensure a fair comparison, we define an energy-efficiency coefficient as the average temperature rise of the entire target per unit microwave power. At t=6 min, the coefficient is 0.67 °C/W for Scenario 1 and 0.44 °C/W for Scenario 2, corresponding to a 34.3% decrease in energy efficiency for the NTC MA. This reduction stems intrinsically from the intentional suppression of microwave absorption at elevated temperatures. Nevertheless, such efficiency loss can be mitigated in practical implementations through complementary strategies, such as asymmetric waveguide configurations, as demonstrated in Supplementary Note 5. Experimental validation of enhanced heating uniformity enabled by the NTC MA. a, b Temperature profiles recorded at nine monitoring points, along with the associated COV curve, under Scenario 1 (a) and Scenario 2 (b) in the planar configuration. c, d Corresponding heating rates derived from the temperature profiles under Scenario 1 (c) and Scenario 2 (d). e Visual comparison of the thermal evolution between the two scenarios using infrared thermal imaging. f Photograph of the cubic MA and its arrangement within the microwave cavity. g, h Infrared thermal images of the cubic MA under Scenario 1 (g) and Scenario 2 (h). i, j Recorded temperature profiles and the corresponding COV curve under Scenario 1 (i) and Scenario 2 (j) in the cubic configuration. k Photograph and infrared thermal images of the cubic NTC MA in another arrangement. l Recorded temperature profiles and the corresponding COV curve of the cubic NTC MA in another arrangement. The heating performance of the fabricated MAs in the cubic configuration is presented in Fig. 6f–l. Fig. 6f shows a photograph of the cubic MA and its arrangement within the microwave cavity, where dual infrared cameras simultaneously monitor the temperature distributions across all faces. Infrared thermal images depicting the evolution of the temperature field under Scenario 1 and Scenario 2 are presented in Fig. 6g, h, with the complete thermal dynamics documented in Supplementary Movie 4. As observed, Scenario 1 exhibits a stable and uneven temperature distribution, with some regions remaining nearly unheated. In contrast, Scenario 2 demonstrates an almost uniform temperature field. The lower-temperature regions circled in the images arise from heating dissipation through the microwave- reflective support base. The recorded temperature profiles and corresponding COV curves for both scenarios are presented in Fig. 6i, j. Compared to Scenario 1, Scenario 2 achieves a reduction of 93% in COV at t=10 min. Moreover, we explored the influence of the arrangement of the NTC MA within the cavity. As shown in Fig. 6k, l, as well as in Supplementary Movie 5, Scenario 2 maintains consistent performance regardless of the position and orientation of the NTC MA, confirming the robustness of our method. Note that the apparent hot spots around the cubic MA in Supplementary Movies 4 and 5 are not actual temperature rises, but instead arise from infrared-signal reflections, as the metallic cavity walls behave as opposing mirrors that repeatedly reflect the emitted radiation. Beyond planar and polyhedral geometries, we further validated the effectiveness of the NTC MA on curved surfaces. As shown in Fig. 7a, b, the NTC MA is implemented on a cylindrical shell and a Möbius strip, representing directional and non-directional curvature, respectively. In both configurations, the NTC MA consistently achieves uniform microwave heating. The full thermal evolution is provided in Supplementary Movie 6, demonstrating the robustness of the approach against geometric deformation. It should be noted that apparent hot spots around the samples in the movie are artefacts caused by infrared reflections from the metallic cavity walls, as mentioned earlier. Furthermore, we validated the heating performance of the NTC MA after being cut into two smaller pieces (Fig. 7c), which maintains uniform microwave heating and facilitates conformal integration with objects of more complex geometries. Uniform heating under geometric deformation, mechanical reconfiguration, and programmable pattern-selective heating enabled by NTC MAs. a Photographs and infrared images of a cylindrical shell. b Photographs and infrared images of a Möbius strip. c Photographs and infrared images of an MA film after mechanical cutting into two pieces. d Simulation configuration for the programmable pattern-selective heating demonstration. e Schematic and absorbance spectra of the stacked NTC MA comprising MA1, MA2, and MA3 units. f Sequential evolution of the dynamic heating patterns and the corresponding temperature profiles of each MA unit under frequency-selective excitation. Furthermore, simulations have been performed to investigate the potential of the NTC MA for programmable, pattern-selective uniform heating, which in this case requires dynamic tuning of the microwave frequency. As illustrated in Fig. 7d, a planar NTC MA with a stacked-metasurface configuration is positioned at the center of a cubic microwave cavity with a side length of 300 mm. The MA consists of three types of units—MA1, MA2, and MA3 (see Fig. 7e)—containing three, two, and one metal–dielectric–metal layers along the normal direction, respectively, and therefore exhibiting three, two, and one resonant peaks in the absorbance–frequency spectra. Additional details regarding the MA configuration are provided in Supplementary Note 4. As shown in Fig. 7f and Supplementary Movie 7, by tuning the microwave frequency to 2.04, 2.42, and 2.97 GHz, programmable pattern-selective heating is achieved, where the temperature distribution sequentially forms the Chinese characters “一”, “十”, and “田” (meaning “one,” “ten,” and “field,” respectively). At each stage, uniform temperature distribution within the target pattern is maintained, owing to the NTC absorption effect. This clearly demonstrates the strong programmability and versatility of the proposed approach. Finally, a representative application is showcased in Supplementary Note 9, where the developed NTC MA is applied to curing polymer composite structures. As expected, the enhanced temperature uniformity ensures uniform curing in all tested structures, including a planar laminate, a wing leading-edge panel, and a cylindrical shell. Limit analysis of NTC MAs in enhancing microwave heating uniformity To verify the effectiveness of the method under extreme conditions, simulation studies were carried out within a single-mode cavity (Fig. 8a). As is well known, single-mode cavities not only exhibit the most uneven electromagnetic field (with the strongest intensity at the center and zero at the edges) but also maintain a fixed field pattern, thereby eliminating contributions from field redistribution effects. Thus, they represent the extreme case in this study. As shown in Fig. 8b, the step-like A–T curve was adopted in Scenario 2 to ensure the optimal performance of the proposed method. However, a residual absorbance of approximately 4.8% beyond the target temperature of 120 °C is unavoidable due to the intrinsic absorption of the dielectric spacer. Detailed simulation settings are provided in Supplementary Note 4. Limit analysis of NTC MAs in enhancing microwave heating uniformity. a–f Simulation results within a single-mode cavity. a Illustration of the simulation configuration. b Adopted A–T curves for Scenario 1 and Scenario 2. c, d Temperature profiles on all MA units and the corresponding COV curve under Scenario 1 (c) and Scenario 2 (d). e Temporal evolution of the electric field distribution (magnitude) within the dielectric spacer under both scenarios. f Temporal evolution of the surface temperature field under both scenarios. g–k Simulation results within the spherical multimode cavity. g Spatial arrangement of five cubic MAs inside the cavity. h Spatial orientation of each MA. i, j Temperature profiles across six faces of the five cubic MAs under Scenario 1 (i) and Scenario 2 (j). k Final temperature distributions on the five MAs under both scenarios. As depicted in Fig. 8c, under Scenario 1, the temperature curves on each MA unit initially diverge and then become parallel under the influence of thermal conduction. In contrast, Scenario 2 displays temperature profiles that diverge initially but tend to converge toward the target temperature (Fig. 8d). Nevertheless, the COV does not ultimately reach zero, owing to the aforementioned residual absorbance. Even so, a 90% reduction in COV is achieved at t=52 s under this extreme condition. As illustrated in Fig. 8e, f, the electric field distribution within the dielectric spacer of the constant MA is consistently maintained in the form of a typical single-mode pattern, resulting in a highly non-uniform, sinusoidal temperature distribution. By comparison, the initially non-uniform electric field distribution within the NTC MA becomes progressively more uniform, accompanied by a gradual decline in overall field intensity. Correspondingly, the temperature field evolves from an initially uniform state (room temperature) to a non-uniform state during the heating process, and ultimately returns to uniformity at the target temperature. Moreover, microwave heating simulations involving multiple objects were also conducted within a spherical multimode cavity characterized by highly uneven electromagnetic field distributions. Fig. 8g, h illustrates the spatial arrangement and orientation of five cubic MAs inside the cavity. Specifically, MA-1, MA-4, MA-5, and the wave port were deliberately aligned along a diameter of the spherical cavity, resulting in pronounced mutual occlusion. MA-2, and MA-3 were placed on the equatorial plane to represent two distinct spatial offset configurations relative to the wave port. In addition, each MA was subjected to a controlled 45° rotation about different axes to thoroughly assess the robustness of our method. More simulation details can be found in Supplementary Note 4. For a comprehensive thermal analysis, temperature distributions were examined from both front and back perspectives. Fig. 8i, j present the temperature profiles across all six faces of the five cubic MAs under Scenario 1 and Scenario 2, respectively. Under Scenario 1, the temperature profiles still exhibit continuous divergence over time, while under Scenario 2, the initially diverging profiles progressively converge toward the target temperature, resulting in a 92% reduction in COV at t=180 s. A visual comparison of the thermal evolution in both scenarios is available in Supplementary Movie 8, with the final temperature distribution shown in Fig. 8k. These results clearly demonstrate the robustness and generalizability of our method. In the future, the proposed method could be further integrated with existing techniques—for example, electromagnetic field control and heat transfer optimization—to enable fully uniform heating across diverse application scenarios. Furthermore, the developed NTC MA concept may be extended to other electromagnetic frequency bands for adaptive wave absorption. Methods Simulation settings Electromagnetic–thermal coupling simulations were performed using COMSOL Multiphysics. Detailed configurations for the field-redistribution analysis, the heating of planar and polyhedral structures, the programmable pattern-selective heating in multimode cavities, and the limit-analysis simulations are provided in Supplementary Note 4. Experiment settings Details of the microwave-absorption measurements are given in Supplementary Note 7. The experimental setups for microwave heating of planar and cubic objects, as well as for the cylindrical shell and Möbius strip, are described in Supplementary Note 8. Supplementary information Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Supplementary Movie 5 Supplementary Movie 6 Supplementary Movie 7 Supplementary Movie 8 Transparent Peer Review file Source data Source Data Acknowledgements We thank Prof. Ke XU (NUAA), Prof. Huachen CUI (HKUST), and Prof. Huacheng ZHU (SCU) for their valuable discussions. Author contributions J.Z. and D.L. contributed equally to this work. J.Z., D.L. and Y.L. conceived the concept; J.Z., D.L., L.Z. and G.L. conducted the theoretical analysis; J.Z. and D.L. designed the prototype setup; J.Z., D.L. and S.L. conceived the simulation and experiment; D.L. and L.X. conducted the simulation; D.L., L.X. and X.C.H. conducted the experiment; J.Z., D.L., L.X., X.C.H., S.L. and X.Z.H. analyzed data; D.L. and J.Z. wrote the manuscript; J.G., G.L., L.Z. and Y.L. revised the manuscript; Y.L., L.Z. and X.Z.H. supervised the project; Y.L., X.Z.H, J.Z. and S.L. acquired the funding. All authors discussed the results and commented on the manuscript. Peer review Peer review information Nature Communications thanks Smriti Agarwal, Tie Jun Cui, Wonbin Hong, and the other anonymous reviewer for their contribution to the peer review of this work. A peer review file is available. Funding This project was funded by the Major Program of the National Natural Science Foundation of China (Grant no. 52090052), the Jiangsu Provincial Key Basic Research Program (Grant No. BK20253022), the National Science Fund for Distinguished Young Scholars (Grant no. 51925505), the Fundamental and Interdisciplinary Disciplines Breakthrough Plan of the Ministry of Education of China (Grant no. JYB2025XDXM205), the National Natural Science Foundation of China (Grant no. U2570244, 52405377 and 52105364), and the New Cornerstone Science Foundation through the XPLORER PRIZE. Data availability Data supporting the findings of this study are provided in the Source Data file. Source data are provided in this paper. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Jing Zhou, Di Li. Supplementary information The online version contains supplementary material available at 10.1038/s41467-026-74811-5. References Associated Data Supplementary Materials Supplementary Information Description of Additional Supplementary Files Supplementary Movie 1 Supplementary Movie 2 Supplementary Movie 3 Supplementary Movie 4 Supplementary Movie 5 Supplementary Movie 6 Supplementary Movie 7 Supplementary Movie 8 Transparent Peer Review file Source Data Data Availability Statement Data supporting the findings of this study are provided in the Source Data file. Source data are provided in this paper.