High thermal conductivity in noncrystalline enriched polyethylene fibers

Abstract

Polyethylene represents an exceptional thermal conductor in theory: The perfectly extended chain is predicted to conduct heat extremely well. However, its practical scalable forms fall far below this limit because noncrystalline structures disrupt heat transport across multiple length scales. Here, we identify that a partially ordered, noncrystalline transitional phase in highly aligned polyethylene is not a negative by-product of processing but a key contributor to heat conduction. Guided by this insight, we develop a gel-state intermittent slow stretching method that directs noncrystalline evolution during polyethylene fiber formation. This approach enables chain relaxation and structural reorganization in regions typically regarded as amorphous and interfacial, promoting their conversion into the transitional phase and their seamless integration with crystalline domains. The resulting structure extends the continuity of ordered shish segments within the period structure, increasing the distance over which heat can travel quasi-ballistically. As a result, polyethylene fibers with 26.79% noncrystalline content achieve thermal conductivities up to 70.61 watts per meter per kelvin, representing 1.45 to 2.60 times of leading commercial polyethylene fibers. These findings establish control of noncrystalline structure as an essential route to unlocking high thermal conductivity in polymers and open a pathway toward lightweight, fully organic thermal conductors.

Noncrystalline enriched polyethylene fibers have thermal conductivity exceeding 70 W m⁻¹ K⁻¹.

INTRODUCTION

Thermally conductive materials are essential for managing heat across modern technologies, including advanced electronics, high-speed communication systems, energy devices, and emerging artificial intelligence hardware (1–6). Metals are widely used for heat dissipation, but their electron-based transport makes them electrically conducting, limiting use in systems that require insulation (7, 8). Ceramics transfer heat through phonons and offer electrical insulation, yet they are often brittle and hard to process (9–11). An ideal material would combine metal-like thermal conductivity with ceramic-grade insulation while retaining mechanical robustness and ease of processing. Polymers are attractive in this regard because they are lightweight, flexible, and intrinsically insulating. However, most polymers are strong thermal insulators because molecular and structural disorders severely restrict phonon transport (12–15). Achieving metal-level thermal conductivity in electrically insulating polymers remains a fundamental challenge.

Within this landscape, polyethylene provides a valuable model system for probing the limits of phonon transport in polymers. Its simple linear backbone and minimal chemical complexity allow unusually efficient vibrational transport. Simulations predict that an infinitely extended polyethylene chain could reach thermal conductivities near 350 W m⁻¹ K⁻¹ (16), and ultradrawn nanofibers approaching single-crystal order have already exceeded 100 W m⁻¹ K⁻¹ (17, 18). However, this intrinsic potential is not realized in macroscopic materials. When formed into materials with large-scale dimensions, polyethylene exhibits a sharp conductivity decline (Fig. 1A)—typically 20 to 22 W m⁻¹ K⁻¹ in continuous microfibers (19, 20), 10 to 60 W m⁻¹ K⁻¹ in thin films (21, 22), and only 0.3 to 7 W m⁻¹ K⁻¹ in bulk solids (23–25). This disparity reveals a persistent gap between theoretical limits and realizable polyethylene materials. Previous efforts to close this gap have focused on increasing crystallinity and chain alignment, assuming that heat travels primarily through crystalline domains (26–30). Yet even highly aligned, highly crystalline polyethylene still underperforms relative to theoretical expectations, indicating that a crystalline-only regulation should be insufficient. Between crystalline and amorphous regions lies a partially ordered yet noncrystalline transitional phase (Fig. 1B), also known as interphase or rigid amorphous fraction (31–33). Our experimental studies show that this phase can account for more than 20% of the structure in highly drawn fibers (Fig. 1, C and D), suggesting a potentially important and previously overlooked role in determining thermal transport. Whether this phase contributes positively to heat conduction and how it can be engineered to narrow the gap between intrinsic and realized thermal conductivity remains unclear.

Fig. 1. Hierarchical thermal conductivities and three-phase structures in polyethylene.

(A) Schematic illustrating the evolution of thermal conductivity in enlarging polyethylene structures, from an ideal single extended chain and a perfect orthorhombic crystal to experimental forms including nanofibers, continuous microfibers, thin films, and bulk samples. Thermal conductivity (table S1) declines sharply with increasing material dimensions, particularly as noncrystalline regions including the transitional zone (ii) and amorphous zone (iii) are introduced between ordered crystalline domains (i). (B) Selected area electron diffraction from a uniaxially stretched polyethylene sample confirms the coexistence of crystalline (i), transitional (ii), and amorphous (iii) phases. The arrow in region (i) indicates the stretching direction. Detailed indexing information was provided in text S1, fig. S1, and table S2. (C) Small-angle x-ray scattering and wide-angle x-ray diffraction (XRD) pattern from highly stretched fibers, revealing semicrystalline characteristics and pronounced anisotropy. (D) Proportion range of the three-phase structures in continuous polyethylene fibers at different stretching ratios, suggesting the weight and importance of transitional phase. Detailed calculating information was provided in texts S2 and S3, figs. S2 and S3, and tables S3 and S4.

Here, we show that the partially ordered transitional phase is not a passive structural by-product but a key contributor to heat conduction in polyethylene. Nonequilibrium molecular dynamics (NEMD) simulations indicate that, in highly aligned structures, this phase can reach a thermal conductivity of 14.33 W m⁻¹ K⁻¹ at 100 K, far above what is typically expected for noncrystalline regions, and it can support extended phonon pathways for quasi-ballistic transport. Building on this insight, we introduced a gel-state intermittent slow drawing method to tune noncrystalline evolution during fiber processing. This approach enables chain relaxation and structural reorganization within regions typically considered amorphous and interfacial, allowing them to transform into the transitional phase and integrate smoothly with adjacent crystalline domains. As a result, the continuity of ordered segments is largely extended, as reflected by an increase in the fibrillar shish-crystal length (L_shish), which in turn enlarges the distance over which heat can propagate with minimal scattering. Consequently, polyethylene fibers, despite with 26.79% noncrystalline content, can reach thermal conductivities up to 70.61 W m⁻¹ K⁻¹, representing a 1.82-fold enhancement over a conventional quasi-molten drawing method and a 1.45- to 2.60-fold enhancement over leading commercial polyethylene fibers. Because of their low density, these fibers exhibit specific thermal conductivities comparable to those of ceramic and metallic conductors such as boron nitride, copper, and silver. Combined with additional flexibility, electrical insulation, and low dielectric constant, they can be woven into arbitrary geometries to direct heat flow, opening opportunities for high-performance thermal management in systems or composite materials where both high conductivity and strong dielectric insulation are essential.

RESULTS

Theoretical thermal transport in highly aligned polyethylene

To elucidate the thermal transport mechanism in highly aligned polyethylene and to quantify the role of the transitional phase, we first performed NEMD simulations supported by first-principles calculations. A representative structural model was constructed to reflect the hierarchical structure of a highly drawn polyethylene fiber (M_v = 3.66 × 10⁶; draw ratio = 27.5), consisting of alternating crystalline, transitional, and amorphous domains aligned along the chain axis (Fig. 2A and fig. S4). The crystalline domains consist of closely packed chains with long-range order, forming a well-defined lattice. The transitional domains contain highly aligned, axially extended chains but lack a regular lattice due to random rotational disorder. In contrast, the amorphous domains are composed of randomly coiled and entangled chains with disordered packing. A localized heat source was then introduced at the center of the model to establish steady-state heat flow. The resulting temperature profile (Fig. 2B) exhibits nearly linear gradients within each domain and only slight discontinuities at phase boundaries, indicating efficient heat transfer across interfaces. At 100 K, the calculated thermal conductivities are 53.23 W m⁻¹ K⁻¹ for the crystalline phase, 14.33 m⁻¹ K⁻¹ for the transitional phase, and 1.83 m⁻¹ K⁻¹ for the amorphous phase (Fig. 2C). Notably, the transitional phase exhibits a conductivity substantially higher than typically associated with noncrystalline regions (34–37). The interfacial thermal resistances within the continuous-chain structure at 100 K—8.39 × 10⁻¹⁰ m² K W⁻¹ at the amorphous-transitional boundary and 4.59 × 10⁻¹⁰ m² K W⁻¹ at the transitional-crystalline boundary (Fig. 2D)—are much lower than those commonly reported for solid-solid interfaces between dissimilar materials (~10⁻⁹ to 10⁻⁷ m² K W⁻¹, near room temperature) (38).

Fig. 2. Theoretical thermal transport in highly stretched polyethylene fibers.

(A) Schematic of the NEMD simulation model conducted at 100 K, representing the structure of a highly stretched polyethylene fiber featuring alternating crystalline, transitional, and amorphous regions. The model dimensions are 22 Å by 20 Å by 880 Å, with phase composition and chain orientation designed to replicate the structure of an experimental continuous fiber (molecular weight M_v = 3.66 × 10⁶; draw ratio = 27.5). (B) Temperature profile along the fiber axis, illustrating the thermal gradient across the heterogeneous structure. (C) Calculated thermal conductivities of the crystalline, transitional, and amorphous phases at 100 K. (D) Interfacial thermal resistance at the boundaries between different phases. (E) PDOS of crystalline, transitional, and amorphous phases. (F) Frequency-dependent phonon group velocities of the crystalline and transitional phases.

To interpret these results, we calculated the phonon density of states (PDOS) together with the phonon dispersion and group velocities for the amorphous, transitional, and crystalline phases (Fig. 2, E and F, and fig. S5). Typically, thermal transport in aligned polyethylene is dominated by low-frequency modes below ~30 THz, associated with bond bending and dihedral motions. The amorphous phase shows broadened and attenuated features in this range, indicating a lack of well-defined long-wavelength modes and strong phonon scattering. In contrast, the crystalline phase exhibits sharp, intense low-frequency peaks and clear acoustic branches with linear dispersion near Γ and high group velocities, consistent with coherent phonon propagation. The transitional phase displays intermediate characteristics. Its PDOS overlaps with both amorphous and crystalline phases, and its acoustic branches remain continuous in the main heat-carrying range, with only gradual flattening at higher frequencies. This shared vibrational window enables effective mode matching across phases while maintaining sufficient group velocity. As a result, phonon reflection at interfaces is reduced, interphase energy transfer is enhanced, and both the intrinsic conductivity of the transitional domains and cross-phase heat transport are improved.

Together, these simulations reveal that the partially ordered transitional phase is not a structural by-product but an active contributor to heat conduction. Its spectral and kinetic continuity with the crystalline phase enables phonons to traverse otherwise discontinuous and noncrystalline regions. This insight further suggests that processing strategies for highly aligned polyethylene should intentionally regulate the transitional phase to enlarge ordered segments in the periodic three-phase structure, extending quasi-ballistic phonon pathways and increasing the effective mean free path (MFP).

Structural engineering in highly drawn polyethylene fibers

Conventional polyethylene drawing is typically performed in a quasi-molten state at temperatures just below the melting point (18, 20, 39, 40). Under these conditions, molecular relaxations, such as chain sliding along crystalline surfaces and rearrangement between neighboring crystalline domains, proceed too slowly to accommodate the imposed drawing rates (text S4 and fig. S6). Therefore, molecular chains or fibrillar crystals are prone to rupture at high draw ratios. To overcome this limitation, we introduce a gel-state intermittent slow drawing strategy (fig. S7) that deliberately regulates phase evolution during unidirectional stretching. By allowing sufficient time for chain relaxation and structural reorganization within disordered and interfacial regions, this approach preserves molecular continuity and prevents fracture of fibrillar crystals (fig. S6). Typically, polyethylene powder was first blended with white petrolatum and extruded into continuous gel fibers to achieve preliminary chain alignment (Fig. 3, A and B, and fig. S8). The gel fibers were then directly subjected to intermittent slow drawing at 125° ± 0.5°C, slightly below the melting point, to prepare highly aligned polyethylene fibers with controlled draw ratios (Fig. 3C).

Fig. 3. Structural evolution and its correlation with thermal conductivity in highly drawn polyethylene fibers.

Photographs showing the physical form of ultrahigh molecular weight polyethylene (UHMWPE) at different processing stages: (A) raw powder, (B) gel fiber, and (C) highly drawn fiber. (D) Composition of crystalline, transitional, and amorphous phases as a function of draw ratio. Crystallinity was determined as the average of values obtained from XRD and differential scanning calorimetry (DSC). The molecular weight of polyethylene is 3.66 × 10⁶. (E) Fibrillar shish crystal length (L_shish) and orientation factor of the (200) crystal plane as functions of draw ratio. (F) Molecular chain orientation factors in the crystalline and amorphous regions as functions of draw ratio. (G) Axial thermal conductivity (k) of polyethylene fibers as a function of draw ratio. (H) Plotting L_shish against k for both the polyethylene fibers prepared in this study and commercial references. Solid symbols represent fibers synthesized by the gel-state intermittent slow stretching method in this work (polyethylene _x, where x indicates M_v), while hollow green symbols denote commercial fibers (polyethylene _y, where y refers to product specification), and hollow black symbols represent fibers prepared by the conventional quasi-molten drawing strategy.

During fiber processing, we quantitatively tracked structural evolution as a function of draw ratio, including phase fractions (figs. S3 and S9), shish length (fig. S10), crystalline orientation (fig. S11), and molecular chain alignment (fig. S12). As plotted in Fig. 3 (D to F), these changes follow three distinct regimes: an initial stage of chain alignment and crystallization (light gray), an intermediate stage marked by growth of the transitional phase (dark gray), and a final stage characterized by rupture of molecular chains and fibrillar shish structure (blue). In the initial regime, both crystalline and disordered chains progressively align along the drawing direction, leading to increased crystallinity, enhanced crystal-plane and chain orientation, and the formation of fibrillar shish structure. In the intermediate regime, molecular relaxation becomes increasingly constrained, as only a limited fraction of amorphous and interfacial chains can respond to further deformation (Fig. 3D and fig. S6). As the draw ratio increases from 20 to 27.5, both crystalline and amorphous regions progressively transform into a partially ordered transitional phase, accompanied by further extension of the fibrillar shish crystal structure (Fig. 3E). In the final regime, intergranular chains between crystalline domains begin to rupture, as evidenced by concurrent declines in crystallinity and crystalline orientation. Although the transitional phase fraction continues to increase, the L_shish decreases sharply, indicating structural degradation at excessive draw ratios.

Similar trends are also observed in polyethylene fibers with another molecular weight (M_v = 4.67 × 10⁶; fig. S13), whereas conventional drawing in a quasi-molten state produces only limited increases in L_shish (fig. S14). These comparisons indicate that the gel-state intermittent slow drawing method more effectively promotes structural reorganization during stretching. Specifically, disordered regions are able to transform into the transitional phase and integrate seamlessly with neighboring crystalline domains, leading to a substantial extension of ordered segments within the periodic three-phase structure.

Thermal conductivity and its quasi-linear correlation with shish-crystal length

The thermal conductivity of the polyethylene fibers was measured using laser flash analysis (LFA) with an extrapolation approach (fig. S15). Previous studies have attributed thermal transport in polyethylene mainly to crystallinity and chain alignment, assuming that heat flows predominantly through crystalline domains. However, in this study, the increase in thermal conductivity with a draw ratio does not closely follow the trends in either crystallinity or molecular chain orientation (Fig. 3, D to G, and fig. S13). Instead, for fibers with M_v = 3.66 × 10⁶ and M_v = 4.67 × 10⁶, the evolution of thermal conductivity mirrors the three-stage structural evolution observed during drawing, exhibiting a stronger link to the development of fibrillar shish crystal structure.

To clarify the origin, we have extended the quantitative study of thermal conductivity and structural evolutions to polyethylene fibers with other molecular weights (e.g., M_v = 8.10 × 10⁶) and to available commercial PE fibers (table S4). Correlation analysis across all samples reveals that thermal conductivity shows its strongest dependence on the L_shish, with a correlation coefficient as high as 0.934 (fig. S16). When thermal conductivity is plotted against L_shish, a clear quasi-linear relationship emerges, particularly at high draw ratios (Fig. 3H). This trend is consistently observed across fibers of different molecular weights and preparation processes, indicating a general structural-transport relationship rather than a sample-specific effect.

In an ideal crystalline solid, thermal conductivity scales with the distance over which lattice vibrations can travel without scattering. To examine this link with L_shish, we measured cryogenic thermal conductivity as a function of temperature and extrapolated the data toward absolute zero to predict the phonon MFP in highly aligned polyethylene fibers (text S5 and fig. S17). For fibers with different molecular weights, the extracted MFP values are all very close to, but slightly exceed, the measured L_shish. Therefore, the quasi-linear scaling between thermal conductivity and L_shish indicates that the shish structure consisting of crystalline and transitional phases sets the effective length scale for heat transport in drawn polyethylene. Steady-state cryogenic measurements further show that the phonon MFP closely follows this structural length scale, implying that heat is carried primarily by quasi-ballistic phonons traveling through a continuous crystalline-transitional pathway. In this regime, strong chain alignment narrows the distribution of transport lengths, concentrating heat flow into phonons with long propagation distances. Furthermore, the observation that the MFP slightly exceeds the L_shish suggests that phonons are not confined to individual shish crystals (fig. S18). Instead, they partially penetrate into highly oriented amorphous regions, where low-frequency acoustic modes retain substantial transport capability beyond fibrillar shish crystal. This behavior is consistent with earlier transient-grating studies showing that thermal phonons can travel ballistically within and across nanocrystalline domains, even when their MFPs exceed the sizes of individual crystalline regions (41, 42).

As a result, by controlling the content of the transitional phase and preserving molecular continuity, the gel-state intermittent slow drawing process effectively extends the L_shish and the distance over which heat carriers can travel with minimal scattering. This structural control increases the phonon MFP and strengthens quasi-ballistic thermal transport along the fiber axis. Consequently, polyethylene fibers with a highest M_v of 8.10 × 10⁶, despite retaining 26.79% noncrystalline content (transitional and amorphous phases), achieve a shish length of 974.67 nm and an axial thermal conductivity as high as 70.61 W m⁻¹ K⁻¹ (Fig. 3H and fig. S19), corresponding to a 1.82-fold enhancement over a conventional quasi-molten drawing method (38.61 W m⁻¹ K⁻¹; fig. S14) and a 1.45- to 2.60-fold improvement over leading commercial polyethylene fibers.

Thermal transport demonstration of continuous polyethylene fibers

The practical heat transport performance of continuous polyethylene fibers was evaluated using the setup shown in Fig. 4A. To ensure a fair comparison, polyethylene fibers and reference materials were selected with similar diameters, as confirmed by scanning electron microscope images (fig. S20). The tested samples included nickel wire, stainless steel wire, alumina fiber, and two representative polyethylene fibers (polyethylene_810w prepared in this study and a commercial counterpart polyethylene_gr810w). All samples were coated with a thin gold layer to promise uniform surface emissivity. Heat was supplied by a ceramic heater, and the temperature distribution along each sample was monitored in real time using an infrared thermal imaging camera.

Fig. 4. Thermal transport behavior of continuous polyethylene fibers.

(A) Photograph and schematic of the experimental setup for thermal transport testing. A ceramic heater was used to heat one end of the fiber, while an infrared thermal imaging camera recorded the temperature distribution along its length. To ensure uniform emissivity, all wires and fibers were coated with a thin gold layer. (B) Infrared thermal imaging pictures showing temperature distributions along five representative wires and fibers. (C) Corresponding axial temperature profiles highlighting differences in thermal transport among the five samples. (D) Photograph of a three-dimensional woven structure made from continuous polyethylene fibers, suggesting their potential for thermal management applications such as cooling fabric, composite substrate for electronic circuits, and lightweight heat dissipation structural parts. (E) Comparison of various macroscopic materials in terms of specific thermal conductivity (thermal conductivity normalized by density) and electrical insulation performance (electrical resistivity).

As shown in Fig. 4 (B and C), the polyethylene_810w fiber conducts heat at a level comparable to nickel wire and substantially outperforms stainless steel (≈15 W m⁻¹ K⁻¹) and alumina fibers (≈35 W m⁻¹ K⁻¹). Nickel, a representative metal with moderate-to-high thermal conductivity (≈90 to 91 W m⁻¹ K⁻¹), provides a useful benchmark. The infrared imaging results are consistent with LFA measurements, which yield an axial thermal conductivity of ~70 W m⁻¹ K⁻¹ for the polyethylene _810w. Notably, this value reaches about 70% of the theoretical upper limit reported for ultradrawn polyethylene nanofibers approaching ideal single-crystal order (~104 W m⁻¹ K⁻¹), underscoring the effectiveness of the present processing strategy in tuning the transitional phase and shish crystal structure.

Beyond metal-level heat conduction, the polyethylene_810w fiber fully retains excellent electrical insulation (<10⁻¹⁴ S m⁻¹) and dielectric properties that surpass those of most ceramics (table S5). Its mechanical flexibility and one-dimensional form allow the fibers to be woven into arbitrary geometries that guide heat flow in complex three-dimensional systems (Fig. 4D). Typically, conventional thermally conductive composites rely on high loadings of particulate fillers, which often degrade mechanical properties and introduce notable interfacial resistance, limiting heat transport (43, 44). In contrast, these intrinsically conductive fibers provide three-dimensionally continuous pathways for heat flow along aligned molecular chains, minimizing scattering (45, 46). This structural continuity enables composite designs with lower thermal resistance and improved performance while preserving mechanical integrity. In Fig. 4E, we compare a range of macroscopic materials in terms of specific thermal conductivity (thermal conductivity normalized by density) and electrical insulation. The polyethylene _810w fiber emerges as a distinct polymer system whose specific thermal conductivity rivals that of ceramic and metal conductors, such as boron nitride films, copper, and silver, while offering markedly superior electrical insulation. This unique combination of high thermal conductivity, low density, ideal dielectric performance, and excellent mechanical strength and flexibility (fig. S21) enables applications in lightweight heat dissipation textiles, flexible electronic composite materials, and electrically insulating structural components. Together, these attributes define a class of fully organic thermal conductors that bridge the traditional divide between metals, ceramics, and polymers.

DISCUSSION

In summary, this study reports a gel-state intermittent slow stretching strategy to tune the evolution and connectivity of crystalline, transitional, and amorphous regions, extending quasi-ballistic pathways for heat flow across multiple length scales. The partially ordered transitional phase emerges as a critical heat-carrying component, bridging ordered and disordered regions and reducing resistance at internal boundaries. This structural continuity increases the phonon MFP and allows continuous polyethylene fibers to reach thermal conductivities of 70.61 W m⁻¹ K⁻¹, approaching 70% of the theoretical limit of perfectly aligned crystalline polyethylene. Beyond achieving record performance for electrically insulating polymers, these results provide a general framework for designing heat-conducting polymer solids: targeted control of noncrystalline structural phases to extend continuous, low-resistance pathways, which is believed to inspire future advances in other polymer systems, including poly-p-phenylene benzobisoxazole, liquid-crystalline polyimides, and epoxies.

MATERIALS AND METHODS

Materials

Ultrahigh molecular weight polyethylene (UHMWPE) powders, with viscosity-average molecular weights (M_v) of 3.66 million, 4.67 million, and 8.10 million, were provided by the Shanghai Institute of Organic Chemistry. Commercial polyethylene fibers were provided by two suppliers: grades 8104515 and 8104510 from Shandong Laiwei New Materials Co. Ltd. and grades gr800w, br800w, 467-283d, and 366-285d from Jiangsu Jiujiujiu Technology Co. Ltd. White petrolatum and n-hexane were sourced from Sonneborn and Adamas, respectively. All materials were used without further purification.

Spinning of polyethylene gel fiber

Polyethylene gel fibers with M_v of 3.66 million, 4.67 million, and 8.10 million were prepared with the help of JiuZhouXingJi Technology Co. Ltd. Typically, UHMWPE powder (M_v = 3.66 million) was compounded with white petrolatum at a precisely controlled mass ratio of 19:1 (w/w) using a Banbury-type internal mixer (XSS-300 torque rheometer) at 190°C. This high-temperature mixing process facilitated the disentanglement of UHMWPE molecular chains via applied shear forces, achieving a homogeneous dispersion within the petrolatum matrix. The resultant viscoelastic gel was subsequently processed into continuous fibers using extrusion technology. A custom-designed laboratory extruder, equipped with a multihole spinneret, was used to shape the gel into fibrous form under controlled thermal conditions at 150°C.

Preparation of continuous polyethylene fiber

The continuous polyethylene fibers were prepared through a two-stage process by a typical gel-state intermittent slow drawing method. Stage 1 was an intermittent hot drawing process. Typically, gel-state fibers were mounted in a uniaxial stretching system (WBE-9000B, Guangdong WBE Instrument Technology Co. Ltd.) placed inside a temperature-controlled oven. After thermal equilibration at 125° ± 0.5°C, the fibers were stretched at a constant strain rate of 1%/s. This slow, controlled stretching allowed sufficient time for molecular chains to reorganize and align along the fiber axis. To enhance chain orientation, the fibers underwent a series of intermittent drawing cycles: In each cycle, they were stretched to a draw ratio of 200% (i.e., twice their original length), followed by a 5-min annealing period in the oven. After n such cycles, the fibers were subjected to a final stretching step with a draw ratio of m. The total draw ratio was calculated as 2ⁿ × m, with final values ranging from 5 to 35. For comparison, fibers prepared without hot drawing, by directly extracting the gel state, were designated as draw ratio = 1. Stage 2 involved solvent extraction and purification. Specifically, the drawn fibers were wound onto carbon fiber-reinforced composite mandrels and purified using Soxhlet extraction with n-hexane to remove residual white petrolatum. The extraction was conducted under continuous reflux at 110° ± 1°C for 12 hours.

Characterizations

The selected area electron diffraction patterns of the uniaxially stretched polyethylene sample were obtained using a Titan Krios transmission electron microscope operated at 300 kV. To probe the structural evolution of polyethylene from powder to gel fiber and lastly to continuous fiber, small-angle and wide-angle x-ray scattering (SAXS-WAXD) measurements were performed on a Xeuss 3.0 system (Xenocs, France) using Cu Kα radiation and an Eiger2R 1M detector (75-μm pixel size, 600-s exposure time). The scattering vector (q) spanned 0.025 to 3.250 nm⁻¹, with a sample-to-detector distance of 60 mm. The x-ray source, operating at 30 W with a 30-μm focal spot, delivered a maximum flux of 4.5 × 10⁸ photons/s at the sample position. Differential scanning calorimetry (DSC) measurements were performed using a DSC 8000 instrument (PerkinElmer) under a nitrogen atmosphere, with a constant heating rate of 10°C min⁻¹. X-ray diffraction (XRD) analysis was conducted using an Ultima IV diffractometer (Rigaku, Japan) with Cu Kα radiation, operating at 40 kV and 40 mA. Data were collected in the 2θ range of 10° to 50°, in steps of 0.02°. Raman spectroscopy measurements were conducted using a laser microconfocal Raman spectrometer (HR Evolution, HORIBA) with a 532-nm excitation wavelength. Rheological tests were conducted using a rotational rheometer (Anton Paar MCR 302, Graz, Austria) equipped with a 25-mm-diameter parallel plate geometry and a 2-mm gap. Samples were compression-molded into disc shapes and placed onto the preheated lower plate at 125°C. Following a 10-min thermal equilibration, frequency sweeps were conducted from 0.001 to 100 Hz at a constant strain of 0.1%.

NEMD method

Thermal transport properties of polyethylene with different molecular configurations were investigated using the NEMD method. To this end, a giant polyethylene model structure with dimensions of around 22 Å by 20 Å by 880 Å was constructed. The model was based on a stretched fiber sample in this study, composed of crystalline, transitional, and amorphous regions with volume fractions of 74.4% (XRD result), 18.85%, and 6.72%, respectively. Here, the crystal structure is approximately orthorhombic. Typically, the crystalline exhibits a high degree of chain alignment along the z axis, with an orientation factor of 0.809. The transitional region shows slightly lower chain alignment with an orientation factor of 0.800. Meanwhile, the amorphous region, although disordered, maintains a relatively high orientation factor of 0.886 along the z axis. These structural features are based on experimental data of the stretched fiber sample (M_v = 3.66 × 10⁶, draw ratio of 27.5), ensuring the simulation’s consistency with the actual system. These chain orientation factors were calculated using the Hermans orientation function

$$f=\frac{3\langle {\text{cos}}^2\mathrm{\theta }\rangle -1}{2}$$ (1)

where θ is the angle between the local chain segment direction and the fiber axis (z direction). In each region, the local chain segments were identified from connected C─C bonds after structural relaxation and equilibration.

Periodic boundary conditions were imposed in the three orthogonal directions to mimic large structural sample. To describe the atomic interactions in as-investigated polyethylene polymer systems, the adaptive intermolecular reactive empirical bond-order potential was adopted (47), which is composed of three-body reactive empirical bond-order (48) and two-body 12-6 Lennard-Jones (LJ) and torsional functions. The cutoff distance of nonbonded interaction of the 12-6 LJ was assigned to be 9.0 Å. The standard Newton’s equations of motion in the models were integrated using the velocity-Verlet integration algorithm. The time step was set to 0.1 fs. Before the formal simulation, a preequilibration process was conducted, beginning with structural equilibration of 1,000,000 time steps under constant Number of particles, Volume, and Temperature (NVT) ensemble at 100 K, followed by molecular dynamics (MD) simulations of 1,000,000 time steps under NPT ensemble at 100 K, and concluding with MD simulations of another 1,000,000 time steps at 100 K under NVT ensemble. This equilibration protocol was designed to prepare a locally relaxed and thermally stable three-phase model for the subsequent NEMD calculations, rather than to reproduce complete long-timescale conformational evolution of a bulk semicrystalline polyethylene system. In the present model, the crystalline, transitional, and amorphous regions were predefined according to the experimentally derived three-phase structure of a highly drawn fiber, and the simulation temperature was set to 100 K, well below the melting temperature of polyethylene. Under these conditions, large-scale chain motion is strongly suppressed, and structural relaxation is mainly governed by local processes such as bond length and bond angle adjustment, dihedral relaxation, and short-segment cooperative rearrangement. In addition, the amorphous and transitional regions are spatially confined between ordered domains, which further limits long-range chain displacement. Therefore, the adopted equilibration procedure was considered sufficient to remove residual local stress and stabilize the heterogeneous structure before thermal transport analysis. Moreover, the thermal conductivity was extracted only from the steady-state portion of the subsequent NEMD run, namely, the last 10,000,000 time steps, thereby minimizing the influence of any initial transient relaxation on the final transport results. Last, the NEMD method was used to calculate thermal conductivity. The enhanced heat exchange algorithm was adopted to impose z-directional constant heat flow of 0.05 eV/ps. MD simulation results of the last 10,000,000 time steps were used to calculate the thermal conductivity. The thermal conductivity of polyethylene in the transport direction was determined as follows (49)

$$k\left(L\right)=\frac{\mathit{dE}/\mathit{dt}}{S∆T/L}$$ (2)

where $S$ (22 × 20 Ų) and $L$ are the cross-sectional area and the effective axial length of model, respectively. The $\mathit{dE}/\mathit{dt}$ (0.05 eV/ps) and $∆T/L$ are the average energy exchange rate and temperature gradient in the thermal transport direction, respectively. All MD calculations were implemented using the Large-scale Atomic-Molecular Massively Parallel Simulator (LAMMPS) software package (50). Specific LAMMPS commands and tools used include fixing heat to impose a constant heat flux and computing temp/profile to measure the temperature gradient along the z axis. Custom scripts were used to calculate the energy exchange rate $\mathit{dE}/\mathit{dt}$ and the thermal conductivity $k\left(L\right)$ based on the extracted simulation data. These steps ensure reproducibility and accuracy in determining the thermal transport properties of the polyethylene fiber model.

PDOS calculation

The PDOS for amorphous, transitional, and crystalline polyethylene structures was calculated using a Green’s function-based molecular dynamics approach (51). This method enables the calculation of vibrational spectra by capturing atomic interactions and lattice dynamics, providing insight into phonon transport behavior across different structural phases of polyethylene.

Phonon spectrum and group velocity calculations

The phonon spectrum and group velocity calculations were conducted using density functional theory (DFT) within the projector augmented wave framework, as implemented in the Vienna ab initio simulation package (52). The generalized gradient approximation proposed by Perdew, Burke, and Ernzerhof was selected for the exchange-correlation potential (53). To account for long-range van der Waals forces, the DFT-D3 correction scheme was applied (54). The cutoff energy for plane wave was set to 600 eV. The energy criterion was set to 10⁻⁸ eV in iterative solution of the Kohn-Sham equation. The Brillouin zone integration was performed using a 3 by 2 by 1 k-mesh. All the structures were relaxed until the residual forces on the atoms declined to less than 10⁻⁵ eV/Å. The phonon spectrum and phonon velocity were calculated by phonopy and ShengBTE based on density functional perturbation theory (55, 56). Similarly, the transitional structure of polyethylene was constructed by creating a supercell (1 by 5 by 1) in the direction of -CH₂ chain and followed by quenching at 1000 K with the AIMD (Ab Initio Molecular Dynamics) scheme for 10 ps.

Thermal conductivity measurement

The thermal conductivity of the continuous polyethylene fibers was determined using a combination of LFA and a parallel model extrapolation method (fig. S15). To prepare the test samples, continuous fibers were vertically embedded in a polydimethylsiloxane (PDMS) matrix, forming a composite with well-aligned heat conduction pathways. Thermal diffusivity of the PDMS/fiber composites was measured using LFA (LFA 467, NETZSCH, Germany). The thermal conductivity of the composite was calculated using the following equation

$$k=\mathrm{\alpha }\times \mathrm{\rho }\times C_{\mathrm{p}}$$ (3)

where $\mathrm{\alpha }$ is the thermal diffusivity, $\mathrm{\rho }$ is the density, and $C_{\mathrm{p}}$ is the specific heat capacity measured by DSC, respectively. Given that the fibers spanned continuously from the bottom to the top of the sample, interfacial thermal resistance between the fiber and the PDMS matrix along the heat flow direction was considered negligible. Under this assumption, the thermal conductivity of the individual fibers was extracted using the classical one-dimensional parallel model

$$k_{\text{composite}}=\mathrm{\phi }\times k_{\text{fiber}}+(1-\mathrm{\phi })k_{\text{PDMS}}$$ (4)

where $\mathrm{\phi }$ is the volume fraction of the fibers, $k_{\text{composite}}$, $k_{\text{PDMS}}$, and $k_{\text{fiber}}$ are the thermal conductivity of PDMS composite, PDMS, and polyethylene fiber, respectively.

Determination of fibrillar shish crystal via SAXS

SAXS measurements were performed to investigate the fibrillar shish crystal using a Xeuss 2.0 system (Xenocs, France) with Cu Kα radiation. This technique provides quantitative characterization of hierarchical structural features in materials through its sensitivity to electron density fluctuations spanning multiple length scales from nanometer to submicrometer dimensions, typically involving measurements performed as a function of the scattering vector q. According to the Peter and Ruland method (57), modeling the fibrillar shish crystal as a cylinder whose length largely exceeds its diameter, the average length (L_shish) in the fiber can be determined. This determination used one-dimensional XRD intensity profiles acquired along direction of q_y (fig. S10A) at various scattering vectors (q), leveraging the electron density contrast between the shish crystals and the amorphous.

According to the scattering theory for rod-like particles, the integral half-height width ($B_{\text{obs}}$) of the one-dimensional XRD intensity of a shish with finite length is inversely proportional to both its length $L_{\text{shish}}$ and the q value. If all the azimuthal distribution could be modeled by Lorentz functions, the relation of $B_{\text{obs}}$, $L_{\text{shish}}$, and q is as follows

$$B_{\text{obs}}=b_{\varnothing \; \varnothing }+2\mathrm{\pi }/(L_{\text{shish}}\times q)$$ (5)

where $b_{\varnothing \; \varnothing }$ is the misorientation factors of shish. $L_{\text{shish}}$ could be obtained from the slope of the equations. In this study, all azimuthal distributions were found to fit well with Lorentz functions. For example, as shown in fig. S10 (B and C), the slope of the fitting curves of the integral width versus different 1/q values is 0.00772 for the macro fiber with a draw ratio of 27.5 (M_v of 3.66 × 10⁶), the $L_{\text{shish}}$ = 2π/0.00772 = 813.47 nm.

Determination of crystalline orientation via WAXD

Crystalline orientation was evaluated using two-dimensional WAXD, performed on a Bruker D8 Advance system equipped with Cu Kα radiation operating at 40 kV and 40 mA. The degree of orientation was assessed from the azimuthal intensity distribution of the (002) diffraction peak, integrated over the 2θ range of 22.2° to 25.2° (fig. S11). The orientation factor was quantified using the Herman’s orientation parameter f, which describes the average alignment of crystallites relative to the fiber axis. It is calculated as

$$f=\frac{3\overline{\langle {\text{cos}}^2\mathrm{\varphi }\rangle }-1}{2}$$ (6)

$$\overline{\langle {\text{cos}}^2\mathrm{\varphi }\rangle }=\frac{{\displaystyle \underset{0}{\overset{\frac{\mathrm{\pi }}{2}}{\int }}}I(\varnothing \; \varnothing )\left({\text{cos}}^2\mathrm{\varphi }\right)d\varnothing \; \varnothing }{{\displaystyle \underset{0}{\overset{\frac{\mathrm{\pi }}{2}}{\int }}}I(\varnothing \; \varnothing )d\varnothing \; \varnothing }$$ (7)

where $I(\varnothing \; \varnothing )$ represents the azimuthal intensity around the (002) reflection, and $\mathrm{\varphi }$ is the azimuthal angle. An orientation factor f of 0 indicates a completely random crystalline distribution, while f = 1 denotes a perfect alignment of crystals along the fiber axis.

Molecular chain orientation analysis by polarized FTIR

The orientation of molecular chains in both crystalline and amorphous regions of polyethylene fibers was evaluated using polarized attenuated total reflection Fourier transform infrared spectroscopy (polarized FTIR, Nicolet IS-90, Thermo Fisher Scientific, USA). Two characteristic absorption bands were analyzed: the peak at 719 cm⁻¹, attributed to the in-plane rocking vibration of CH₂ groups in crystalline domains, and the peak at 723 cm⁻¹, associated with similar vibrations in amorphous regions. These two bands were selected as indicators of crystalline and amorphous orientation, respectively (fig. S12A).

Infrared absorption occurs due to changes in the dipole moment during molecular vibrations. The greater the dipole moment change, the stronger the absorption intensity. Because the vibrational dipole moment is directional, the absorption intensity also depends on the polarization of the incident light. When the dipole change aligns with the electric vector of the incident polarized light, absorption is maximized, known as a parallel band. If the dipole moment change is perpendicular to the light’s electric vector, then absorption is minimized, producing a perpendicular band. The variation in band intensity with polarization direction is known as infrared dichroism.

For a given sample, the infrared dichroic ratio (D) was calculated as follows (58, 59)

$$D=A_{\parallel }/A_{\perp }$$ (8)

where D is defined as the ratio of absorbance under parallel-polarized light (${\mathrm{\theta }}_{\parallel }$) to that under perpendicular-polarized light (${\mathrm{\theta }}_{\perp }$). $A_{\parallel }$ and $A_{\perp }$ are the measured integral absorbance value when electric vector of the polarized light is parallel and perpendicular to the fiber axis (i.e., M_v = 3.66 × 10⁶, draw ratio of 27.5; fig. S12B), respectively.

In addition, since the vibrational dipole moment of the characteristic functional group forms a specific angle with the molecular chain axis, angular correction must be applied to calculate the orientation degree of molecular chains within the fiber. Therefore, in this system, the orientation factor of the polyethylene molecular can be obtained by the following equation

$$f=\frac{D-1}{D+2}\times \frac{D_0+2}{D_0-1}$$ (9)

where f is the orientation factor, $D$ is the dichroic ratio, and $D_0$ is an angle-related parameter that is defined by the following equation

$$D_0=2{\text{cot}}^2\mathrm{\alpha }$$ (10)

where $\mathrm{\alpha }$ is the angle between the transition moment associated with the vibrational mode and the chain axis.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S21

Tables S1 to S5

References