Layer-Resolved Frictional Mechanisms of Engineering Polymers: Contrasting Spatial Modes of Dissipation in Polyoxymethylene and Polytetrafluoroethylene Authors: Tomohiro Kinjo, Aguru Yamamoto, Takeshi Aoyagi, Shunsuke Mieda, Hitoshi Washizu Published: 2026 ABSTRACT A fundamental molecular understanding of polymer friction remains elusive despite its critical engineering importance. Although all-atom molecular dynamics (AAMD) simulations can provide molecular-level insights, many studies have focused on individual polymers and/or sliding rates near or beyond the upper end of experimental conditions, often without the depth-resolved perspective needed to distinguish material-specific responses. Here, we use large-scale AAMD to identify two spatially distinct modes of frictional energy dissipation in polyoxymethylene (POM) and polytetrafluoroethylene (PTFE). After a brief interfacial transient, each system reaches a persistent quasi-steady response. POM exhibits a distributed mode that extends into the slab interior: deviations from the ideal Couette response and elevated orientational order persist several nanometers into the slab, and dissipation is dominated by bonded conformational relaxation through bond-angle and dihedral motions. By contrast, PTFE exhibits a mode localized near the interface: shear accommodation and the associated nonbonded energy dissipationdriven by van der Waals and electrostatic interactionsare concentrated within a narrow near-interface zone. In PTFE, this localized mode is accompanied by a pronounced subsurface temperature maximum. Notably, this thermal maximum is spatially offset from a deeper densification band, indicating distinct length scales for heat generation and structural packing. Together, these results establish spatial dissipation pathways and subsurface transport bottlenecks as key descriptors of polymer friction. This framework provides a molecular basis for tuning how and where shear work is absorbed in polymer tribomaterials. Introduction Polymeric materials are widely used in modern industries because they are lightweight, easy to process, and resistant to corrosion and degradation. , Their tribological advantagesincluding low friction, self-lubrication, and excellent wear resistancemake them well suited for sliding components such as bearings, seals, and gears. Consequently, polymer-based tribomaterials are increasingly adopted across the aerospace, automotive, biomedical, precision machinery, and electronics sectors. − The tribological performance of polymers directly influences the energy efficiency and durability of mechanical systems. Improving frictional interfaces is therefore important for global efforts to reduce CO2 emissions. Recent analyses estimate that frictional losses in manufacturing and transportation account for approximately 23% of global energy consumption. − Given the widespread use of polymer components in modern machinery, their contribution to these losses is unlikely to be negligible, making polymer-focused friction-reduction technologies an important lever for sustainable engineering and environmental stewardship. Polymer friction remains among the least understood phenomena in tribology, despite its central role in modern technologies. − This knowledge gap reflects the fact that friction arises from coupled interfacial adhesion, multiscale segmental relaxation, and nonequilibrium thermal transport under strongly driven conditions. For example, molecular dynamics (MD) studies of nonpolymeric liquids under elastohydrodynamic lubrication conditions show that friction can depend strongly on molecular architecture and interfacial structural organization. − In polymers, these sensitivities are further amplified by numerous internal degrees of freedom and nonlinear viscoelastic responses under prolonged shear. ,,− Consequently, bulk and subsurface relaxation processes can influence frictional dynamics beyond the immediate interface, yet they remain comparatively underexplored. Several engineering polymers used in tribological components are semicrystalline. However, sliding and heat generation are frequently governed by an amorphous interphase near the counterface, where chain mobility, local packing, and thermal transport jointly control dissipation. ,− This motivates an explicit depth-dependent description of frictional response, rather than treating the polymer slab as depth-uniform. To address these challenges, MD simulations have become essential tools. Coarse-grained (CG) and reactive MD approaches have elucidated conformational relaxation and tribochemical reactions, − and more recently, all-atom MD (AAMD) has enabled detailed analyses of chain orientation and energy dissipation, increasingly supported by experimental validation. ,, Nonetheless, current computational studies face two critical limitations. (i) Several studies have employed sliding velocities far above experimental conditions or have focused on specific, idealized polymers. This leaves a gap in our understanding of polymers under realistic, prolonged shear. , (ii) Most analyses emphasize interfacial phenomena and effectively adopt depth-uniform or depth-averaged treatments of the slab. This depth-averaging assumption is also shared by several continuum models. , Such depth-averaging obscures how dissipation is partitioned across the interface–subsurface region: whether shear localizes at the sliding interface or penetrates subsurface layers, and how this spatial partitioning couples to heat generation and transport barriers. AAMD is particularly well suited to address this gap because it enables depth-resolved characterization of stress transmission and energy relaxation that is not readily accessible with continuum or depth-averaged models. In this study, we tested the hypothesis that intrinsic molecular structure governs depth-resolved dissipation pathways by performing large-scale AAMD simulations against a rigid iron counterface. We focused on polyoxymethylene (POM) and polytetrafluoroethylene (PTFE) as archetypal low-friction polymers with contrasting molecular architectures that differ in chain polarity, backbone flexibility, and intermolecular interactions. This contrast allowed us to isolate chemistry-dependent contributions to stress transmission, structural relaxation, and heat transport across the interface–subsurface region. We performed depth-resolved, multimodal analyses that combine velocity-based shear-localization metrics relative to ideal Couette flow with chain orientation and temperature fields, and we use these results to define spatial dissipation modes as generalizable descriptors of polymer tribology. Materials and Methods Model Configuration Sliding Model The friction model comprises two contacting components: an Fe slab and a polymer slab (Figure ). The Fe slab serves as a controlled, computationally tractable representation of a steel counterface, modeled here as a pure Fe(001) surface. The Fe slab (10.3 × 7.18 × 2.30 nm3; 14,400 atoms) is treated as a rigid counterface in the present simulations. Chemical structures of POM and PTFE, polymer–Fe friction model, and simplified coordinate definition. (a, b) Repeating units (n = 40) and representative atomistic single-chain conformations for (a) POM and (b) PTFE. (c) Simulation model for friction evaluation (representative POM system). The polymer slab thickness is approximately 15.0 nm, and the Fe slab thickness is 2.30 nm; the lateral box dimensions are 10.3 nm × 7.18 nm. The bottom-most cyan region represents fixed atoms, followed by a white thermostat-controlled (NVT) region (0.5 nm thick) and an NVE region; the same settings are applied to all polymer models. The top Fe slab is subjected to a normal load in the z-direction and a sliding velocity in the x-direction. Atom color code: Fe (orange); carbon (gray); oxygen (red); hydrogen (white); and fluorine (light blue). Periodic boundary conditions are applied in the x and y directions. (d) Simplified coordinate-definition schematic used in the main text, showing the box-fixed surface-normal coordinate Z lab, the fixed reference interface Z int(0), and the material-anchored depth d mat = Z int(0) – Z lab, where positive d mat increases into the polymer. Full operational definitions are provided in Section S2.1 and Figure S2 of the Supporting Information. In practical environments, Fe-based surfaces are typically covered by hydroxide/oxide films before the underlying metal is exposed. , However, under high-load boundary sliding, these films can be locally disrupted and removed, exposing nascent metallic regions. We therefore adopt pure Fe(001) as a severe-contact reference to provide a controlled baseline for isolating polymer-dependent dissipation pathways under conditions in which surface films are mechanically compromised. This choice avoids uncertainties associated with oxide/hydroxide stoichiometry, the degree of hydroxylation, and interfacial parametrization, thereby focusing the mechanistic analysis on the polymer response. Extending the present framework to oxidized/hydroxylated Fe and humid environments, where film stability and interfacial chemistry evolve, remains an important direction for future work. The polymer slab comprises linear chains with 40 repeat units (n = 40) of four polymers: POM, PTFE, polyvinyl chloride (PVC), and poly­(methyl methacrylate) (PMMA). The molecular weights of the nominal chains (based on repeat-unit molar masses) are approximately 1.20 kDa (1.20 × 103 g mol–1), 4.00 kDa (4.00 × 103 g mol–1), 2.50 kDa (2.50 × 103 g mol–1), and 4.00 kDa (4.00 × 103 g mol–1) for POM, PTFE, PVC, and PMMA, respectively. While POM and PTFE are the focus of our mechanistic analysis, PVC and PMMA are included to evaluate the applicability of the framework across diverse polymer classes. All systems are simulated under identical conditions to ensure consistency. Representative repeat units and atomistic single-chain snapshots for POM and PTFE are provided in Figure a,b. These polymers exhibit contrasting backbone chemistry and polarity: an acetal (C–O) backbone for POM versus a C–C backbone bearing fluorine substituents for PTFE. Accordingly, intermolecular interactions in POM include relatively stronger electrostatic contributions, whereas those in PTFE are dominated by dispersion. Although these polymers are typically semicrystalline in practical applications, all systems in this study are modeled as amorphous slabs for three reasons. First, amorphous configurations help isolate chemistry-dependent dissipation pathways without convolution from lamellar morphology or crystallite texture. Second, near-surface amorphous or disordered layers are frequently observed at polymer tribocontacts, making amorphous slabs a relevant representation of the mechanically active interfacial region. − Third, prior experimental and simulation studies consistently identify the amorphous phase as a primary site of shear accommodation, damage initiation, and frictional energy dissipation. ,,− Together, these considerations justify amorphous slabs as a controlled platform for comparing how molecular structure governs frictional response in the depth-resolved analyses. Model construction followed a multiscale workflow in which key polymer descriptors were established before assembling the interface. First, characteristic ratios (C ∞) were specified for each polymer as a dimensionless measure of chain stiffness (unperturbed chain dimensions). C ∞ is defined from equilibrium conformational statistics as follows: C∞=limnb→∞⁡⟨Ree2⟩nb⟨lb2⟩, where ⟨Ree2⟩ is the mean-squared end-to-end distance, nb is the number of backbone bonds (bonds along the main chain), and ⟨lb2⟩ is the mean-squared backbone bond length. In practice, C ∞ can be obtained from equilibrium sampling of sufficiently long chains (or from Rotational Isomeric State-based estimates and experimental chain-dimension data). Here, C ∞ was used as an input descriptor in the multiscale construction to reproduce realistic chain dimensions and stiffness in the back-mapped atomistic models. The C ∞ values used were 12.6 (PTFE), 3.85 (POM), 7.6 (PVC), and 8.22 (PMMA). , Values for PTFE, PVC, and PMMA were sourced from the literature, whereas that for POM was estimated using the Synthia module in Materials Studio. , To determine the equilibrium target bulk density for each polymer, an independent bulk-melt simulation was performed under fully periodic (3D) boundary conditions following the relaxation protocol of Hofmann et al., ensuring the absence of free surfaces or interfacial contacts. After initial equilibration, the final density was estimated from a 1 ns isothermal–isobaric (NPT) trajectory at 300 K and 1 atm using a sufficiently large cubic cell to minimize finite-size effects. Densities were obtained by time-averaging over the terminal segment of this trajectory after volume and pressure had converged, yielding 1.26 g cm–3 (POM), 2.24 g cm–3 (PTFE), 1.35 g cm–3 (PVC), and 1.13 g cm–3 (PMMA), all consistent with experimental ranges. − Using these C ∞ values and the target bulk densities as inputs, we constructed the models using the multiscale workflow. CG configurations were generated using the Kremer–Grest bead–spring model, equilibrated, and then back-mapped to an atomistic (all-atom) representation while preserving chain conformations and local packing. The number of chains and the initial polymer-slab dimensions were selected to reproduce the target bulk densities. The equilibrated polymer slab was then brought into contact with the Fe(001) slab, and periodic boundary conditions were applied only in the x and y directions, with a nonperiodic z direction to accommodate the interface and normal loading. Model construction was performed using J-OCTA (v10.0) , and OCTA (v8.4), and visualizations were generated with OVITO Pro (v3.12.4). Heat-Flux NEMD Model for Thermal-Transport Analysis To quantify the interfacial thermal resistance and the depth-resolved thermal transport on the polymer side, we performed dedicated heat-flux nonequilibrium MD (NEMD) simulations using load-equilibrated contact configurations extracted from the sliding simulations at 0 and 50 ns (shear-history states). For these thermal-transport runs, we increased the Fe slab thickness from 2.30 to 5.50 nm to stabilize the metal-side temperature gradient and reduce boundary artifacts in the highly conductive substrate. This NEMD-specific modification differs from the friction-evaluation model shown in Figure c. We also expanded the boundary regions to keep thermostats away from the interface during equilibration and to define well-posed heat-source and heat-sink zones during the heat-flux stage. Specifically, the topmost 0.5 nm of the Fe slab was held fixed, and the adjacent 1.0 nm region was designated as the Fe heat-source/heat-sink region. On the polymer side, the thermostat-controlled sink region was expanded from 0.5 to 1.0 nm, whereas the remaining polymer atoms adjacent to the interface were left unthermostated to evolve freely. Periodic boundary conditions were applied only in the x and y directions, with a nonperiodic z direction. Simulation Protocols Sliding Simulations MD simulations were conducted using the LAMMPS package with the PCFF force field. All simulations employed a nonreactive all-atom description. Because this framework does not allow bond formation/cleavage or explicit tribochemical reactions, we designed the simulations to isolate physical dissipation mechanisms in amorphous polymer slabs sliding against an idealized rigid Fe(001) counterface. Accordingly, the present sliding simulations were positioned as a mechanistic baseline that captured the physical organization of the sliding state prior to the dominance of long-term wear or chemical evolution. To check whether nonbonded adhesion alone could induce qualitative polymer pickup, we performed auxiliary separation tests by lifting the Fe slab from saved configurations (t = 0, 20, 50, and 80 ns). Polymer-derived residue on the Fe counterface was observed for all polymers studied (POM, PTFE, PVC, and PMMA). Because this behavior was not polymer-specific in our data set, it is not analyzed further here (see Supporting Information, Section S2.4). We focused instead on the contrasting depthwise dissipation modes. A time step of 1 fs was used throughout the simulations. Lennard–Jones interactions were truncated at 1.0 nm, and long-range electrostatic interactions were evaluated using the slab-adapted particle–particle particle–mesh method with a Coulomb cutoff of 1.13 nm. Throughout all stages, the entire Fe slab was treated as a rigid body. Within the polymer slab, a bottom layer of 1.0 nm was held fixed to provide a stable substrate, while an adjacent 0.5 nm layer was coupled to a thermostat (Figure ). The thickness of this thermostated (NVT) layer was set to 0.5 nm to provide thermal stabilization while minimizing thermostat intrusion into the frictional region, thereby maximizing the NVE (microcanonical) region (constant N, V, and total energy). A short validation run (a few ns) using a thicker NVT layer (1.0 nm; fixed layer unchanged) showed no appreciable differences in early stage frictional response or depth-resolved temperature profiles compared with the 0.5 nm setup; therefore, 0.5 nm was adopted for 80 ns production simulations. The simulation protocol comprised two stages. First, a normal pressure of 100 MPa was applied to the rigid Fe slab during a 1 ns contact-formation stage to establish contact with the polymer surface. During this stage, all mobile polymer atoms were maintained at 300 K using a Nosé–Hoover thermostat. Second, during the 80 ns sliding stage, the thermostat was applied only to the 0.5 nm layer, whereas the remaining mobile polymer atoms evolved in the microcanonical (NVE) ensemble to capture natural frictional heating. Shear was imposed by translating the rigid Fe slab along the x-axis at a constant velocity of v = 1.47 m s–1. Over the t = 80 ns sliding stage, the total slip distance was vt = 1.47 m s –1 × 80 ns = 117.6 nm. Considering the polymer film thickness as h = 15.0 nm, the accumulated simple shear strain was γ ≈ (vt)/h = 117.6/15.0 ≈ 7.84. Heat-Flux NEMD Simulations Heat-flux NEMD simulations were initialized from load-equilibrated contact configurations extracted from the sliding trajectories at 0 and 50 ns. The same normal pressure (100 MPa) was maintained by applying a constant force to the rigid Fe block. Each NEMD run comprised two stages. Load Equilibration (No Heat Flux) The system was equilibrated under the applied normal load for 2.0 ns. During this stage, thermostatting was restricted to the designated sink regions only (polymer sink and Fe sink) using Nosé–Hoover dynamics at 300 K, whereas the remaining mobile atoms evolved under NVE, thereby minimizing thermostat intrusion into the interfacial region. Heat-Flux Stage (Steady-Gradient Formation and Averaging) All thermostats were disabled, and all nonfixed atoms (including the sink regions) were integrated using NVE. A constant heat flux was imposed by adding energy to the polymer sink and removing the same amount from the Fe sink using the fix heat command in LAMMPS. In practice, energy was exchanged at a constant rate with ΔE = 0.040 kcal mol–1 exchanged per time step (Δt = 1.0 fs), yielding an imposed heat flux of q ≈ 3.74 GW m–2 (computed from the exchange rate and the cross-sectional area A, as reported in the LAMMPS log output). The heat-flux simulations were run for 2.5 ns in total. The first 0.5 ns was discarded as a transient period to establish a steady temperature gradient, and the subsequent 2.0 ns was used as the analysis window; temperature profiles were averaged over the final 0.5 ns of this window, consistent with the protocol described in Section S4.1.1 for the depth-resolved transport analysis. Analytical Procedures Analysis of Molecular Orientation Molecular orientation was quantified using the orientational order parameter, P 2,x , defined from the second-order Legendre polynomial, P 2, as follows: P2=32⟨cos2⁡θ⟩−12, where θ is the angle between the reference axis and the chain long axis, which was approximated here by the principal axis of inertia associated with the smallest principal moment of inertia. The sliding direction (the x-axis) was used as the reference axis to compute P 2, x . The P 2 values of 1.0, 0.0, and −0.5 indicate perfect parallel alignment, random orientation, and perpendicular alignment, respectively. To evaluate depth-resolved orientation while preserving molecular integrity, we partitioned the slab into 15 chain-based layers (Li ) along the surface-normal direction using K-means clustering (k = 15) of the per-chain Z lab centroids at t = 0 (the onset of sliding). This Lagrangian-like assignment places each polymer chain, in its entirety, into a single layer and avoids the artificial segmentation that can arise with fixed-thickness slicing. Layer assignments were fixed at t = 0 and retained throughout the trajectory. The layer definition used here is specific to the orientational analysis; the layering used for velocity and temperature analyses is defined separately in Section 2.3.2. For the slab-averaged orientational analysis used to compare overall alignment kinetics (Figure ), P 2, x (t) was evaluated for every stored trajectory frame (2 ps intervals), and the slab-averaged P 2,x (t) was obtained by averaging over all chains at each time point. For the depth-resolved layer analysis, P 2,x (t) was computed at 10 ps intervals. This subsampling was adopted because the orientational evolution was smooth and monotonic; the reduced temporal resolution did not alter the observed trends while lowering the analysis cost. Coupled dynamics of molecular orientation and friction. (a) Time evolution of orientational order parameter P 2,x (with respect to the sliding direction) for POM (purple) and PTFE (green). Lines show the across-trial mean; shaded bands denote variability (±1 s.d.). The inset reports the final order parameter at 80 ns (mean ± s.d.). (b) Time evolution of the slab-averaged shear stress τx (t) (plotted in ×102 MPa, consistent with the axis label). (c) Time derivatives of shear stress and P 2,x : dτ x /dt (left axis) and dP 2,x /dt (right axis; scaled by 10–2). Open symbols denote dτx /dt; filled symbols denote dP 2,x /dt. Derivatives were estimated from the processed time series as local OLS slopes; the displayed curves were further lightly smoothed for visualization. Corresponding τ x (t) curves for all four polymers (POM, PTFE, PVC, and PMMA) are provided in Supporting Information, Figure S1b. From the layer-resolved P 2,x (t) data, we extracted two metrics per layer: (i) final P 2,x , defined as the value at 80 ns, and (ii) alignment time, defined as the time required to reach 90% of the final P 2,x . These metrics were averaged over at least three independent trials initialized from different structures. To probe alignment–stress coupling, we analyzed the temporal derivatives dP 2,x /dt and dτ x /dt using the slab-averaged P 2, x (t) and τ x (t) series shown in Figure . For each trial, these series were sampled at 2 ps intervals. Before slope estimation, each series was smoothed with a 5000-point moving average (∼10 ns) to suppress short-time fluctuations. Local rates were then estimated by ordinary least-squares (OLS) regression within nonoverlapping 150-point windows (∼300 ps). The resulting derivative profiles were interpolated onto a common time grid and averaged across trials. For visualization, the trial-averaged slope profiles were lightly smoothed using locally weighted scatterplot smoothing (LOWESS), , to highlight the underlying trend. Depth-Resolved Analyses: Coordinates, Density, and Local Dynamics For density analysis, we used all polymer atoms (including hydrogen and excluding iron). For analyses of velocity, temperature, and velocity-derived fields (e.g., Couette-deviation maps; see Section ), we used the polymer heavy-atom subset (excluding hydrogen and iron). Coordinate Conventions and Layering To streamline the notation, we distinguish one box-fixed coordinate and two application-specific depth/distance measures. The box-fixed coordinate, Z lab (nm), was measured upward from the lower boundary of the simulation cell (box bottom; Z lab = 0), independent of the constrained polymer base layer. We used Z lab for instantaneous structural analyses, such as mass-density profiles. For time-evolving, depth-resolved sliding fields, we used the material-anchored depth, d mat (nm), referenced to the polymer–Fe interface at the analysis time origin (t = 0): dmat≡Zint(0)−Zlab,Zint(0)≡max{Zpoly,heavy(0)} where Z int(0) was determined from polymer heavy atoms only (Fe excluded). Here, d mat = 0 corresponds to the interface at t = 0, and positive d mat increases into the polymer interior. Importantly, Z int(0) was held fixed throughout the analysis; thus, d mat did not move with the instantaneous interface position Z int(t). Throughout the manuscript, box-fixed positions are denoted by Z lab, whereas application-specific depth or distance coordinates are denoted by d with an informative subscript. Depth-dependent sliding fields were evaluated using initial-position-based labeling (fixed membership). Specifically, at t = 0, the polymer film was partitioned into slices for atom-based fields and into chain-based layers for molecular metrics, and each atom or chain retained its initial (t = 0) slice/layer assignment throughout the trajectory, even if it later crossed instantaneous slice boundaries. This material-following description tracks the same portions of the film over time. It also avoids rebinning artifacts and spurious advection effects that can arise in an instantaneous (Eulerian) analysis. Unless otherwise noted, this fixed d mat frame was used for all depth-dependent sliding analyses, including velocity, temperature, and velocity-derived fields. Orientation was evaluated using chain-based layers defined at t = 0 from per-chain z centroids (Section ). Velocity and temperature were evaluated on fixed atom-based slices of thickness Δd = 0.3 nm defined at t = 0 to resolve interfacial gradients while maintaining stable, material-following slice labels. When plotted against depth, d mat should therefore be interpreted as a nominal t = 0-referenced depth, because the instantaneous film thickness can evolve during sliding. Accordingly, depth windows reported in the main text (e.g., 0–4 nm) denote nominal t = 0-referenced ranges used to quantify localization trends, rather than instantaneous geometric distances at time t. For heat-flux NEMD transport analysis, we used a separate thermal signed distance, d th (nm), defined in Section S4.1.4 (Eq. S4-2), solely to extract RK and k eff(d th) from steady-state temperature profiles. dth≡Zth−Zlab where Z th was determined separately for each state/trial from the near-interface temperature transition. This coordinate should not be confused with d mat, which is the fixed material depth used for sliding-field analyses. A simplified coordinate schematic is provided in Figure d, and the full operational definitions are given in Section S2.1 and Figure S2 of the Supporting Information. Mass-Density Profiles To characterize structural layering, the mass density ρ­(Z lab, t) was computed by binning all polymer atoms along the box-fixed coordinate Z lab into equally spaced slabs of width ΔZ = 0.10 nm to resolve fine near-interface layering. Profiles at t = 0, 20, 50, and 80 ns were averaged over three independent trials. For visualization only, we applied a centered moving average of 0.6 nm over Z lab; the smoothed profiles were not used in subsequent quantitative analyses. Depth-Resolved Local Dynamics (Velocity and Temperature) To capture steep interfacial gradients, we computed velocity and temperature in fixed, atom-based slices of thickness Δd = 0.3 nm defined along t = 0. For slice Sk at time t, the sliding-direction velocity was calculated as the mass-weighted center-of-mass (CM) velocity as follows: vx(Sk,t)=∑i∈Skmivx,i(t)∑i∈Skmi[m⁡s−1]. After removing the slice CM drift V CM(Sk , t), the fluctuation velocity is δvi (t) = vi (t) – V CM(Sk, t). The instantaneous thermal kinetic energy was defined as follows: KTKE(Sk,t)=∑i∈Sk12mi∥δvi(t)∥2. Assuming equipartition and subtracting translational CM motion, the temperature was defined as follows: T(Sk,t)=2KTKE(Sk,t)(3Nk−3)kB where Nk is the number of polymer heavy atoms in Sk , and k B is the Boltzmann constant. The (3Nk – 3) degrees of freedom reflect the removal of translational CM motion; component-wise thermal kinetic energies (x, y, and surface-normal (z) components) were also recorded for diagnostics. For spatiotemporal visualization, each slice was associated with its center depth d mat, and trial-averaged maps vx (d mat, t) were constructed by aligning trials in physical time and averaging across at least three independent runs at each time point. For visualization only, we applied light LOWESS smoothing over time (span ≈ 0.10). As a reference for ideal Couette shear, we defined vC(dmat)=vtop⁡max(1−dmatHmob,0), where v top = 1.47 m s–1 and Hmob=Hpoly(0)−1.0nm, where H poly(0) denotes the polymer thickness at t = 0 (the Z lab-span of polymer heavy atoms; Fe and vacuum excluded); the bottom 1.0 nm is immobilized and excluded from H mob. The normalized deviation from ideal Couette shear was defined as follows: D(dmat,t)=vx(dmat,t)−vC(dmat)vtop Thermal-Transport Descriptors from Heat-Flux NEMD Steady-state temperature profiles were obtained by binning atoms along the surface-normal, box-fixed coordinate Z lab (nm) and evaluating the kinetic temperature in each bin. The coordinate conventions follow Section : Z lab denotes the box-fixed position, whereas application-specific depth or distance coordinates are denoted by d with an informative subscript. For thermal-transport analysis, the signed distance d th was used for constructing aligned temperature profiles and evaluating local temperature gradients. Full operational details of the averaging protocol, spatial-discretization robustness (bin widths of 0.1–0.5 nm), and the construction of T(d th) are provided in Sections S4.1.1 and S4.1.4–S4.1.5 and Figure S2. Unless otherwise noted, profiles were averaged over the final 0.5 ns of the transport-analysis window. The effective interfacial thermal resistance, RK , was evaluated as RK=ΔTintq, where q is the imposed heat flux and ΔT int is the interfacial temperature discontinuity. , The physical interface position Z int was determined from the density drop in the mixed Fe + polymer bin-occupancy profile, and ΔT int was obtained by linear extrapolation of the near-interface temperature profiles on the polymer and Fe sides to Z int (Sections S4.1.2–S4.1.3). , Interface-detection and fit-window sensitivity, together with trial-level diagnostics, are provided in Sections S4.1.2–S4.1.3 and Section S4.3 (Tables S2 and S3). To characterize polymer-side depth-resolved thermal transport, we defined a depth-dependent effective conductivity, κeff(dth)=q|dT/ddth|, where d th is the signed distance referenced to the per-trial thermal-center position Z th, dth≡Zth−Zlab, such that d th > 0 on the polymer side and d th < 0 on the Fe side. The procedure for determining Z th, constructing the aligned profile T(d th), and evaluating dT/dd th using a Savitzky–Golay derivative is described in Sections S4.1.4–S4.1.5. Reported values represent the mean ± SEM over independent trials (Section S4.1.6). Figure shows the steady-state temperature profiles plotted against the thermal signed distance d th, whereas RK is defined from the interfacial temperature discontinuity ΔT int at the detected physical interface Z int. This framework separates the contact-level entry resistance captured by RK (via ΔT int at Z int) from the polymer-side heat-spreading efficiency captured by k eff(d th). Results and Discussion Frictional Behavior and Initial Structural Response Validation of Simulated Frictional Behavior We quantified the coefficients of friction (COFs) of four representative polymers (POM, PTFE, PVC, and PMMA) from 2 ps block-averaged lateral shear stresses over the steady-state interval (72–80 ns) across ≥3 independent trials (see Supporting Information, Section S1.1). Under our simulation conditions, the simulations reproduced the experimentally established tribological ranking: PTFE < POM < PVC < PMMA (Supporting Information, Figure S1a). Across all systems, the lateral shear stress τ x increased rapidly after shear onset and then converged to a steady state (Supporting Information, Figure S1b). The simulated COF magnitudes broadly aligned with previously reported experimental ranges (Supporting Information, Figure S1a); minor deviations and their likely origins (e.g., rate mismatch and interfacial relaxation time scales) are discussed in the Supporting Information (Section S1.2). This validation is consistent with previously reported AAMD of amorphous PTFE–PTFE contacts without explicit tribochemistry, which reproduces macroscale tendencies such as a decreasing COF with increasing load. Collectively, these results support the utility of nonreactive all-atom simulations for capturing qualitative tribological trends. In this study, we similarly used an all-atom framework without bond formation or cleavage to isolate physical dissipation mechanisms, while noting that tribochemically anchored transfer films in reactive AA/CG studies at Fe/oxide interfaces can further lower friction under conditions outside the scope of our model. , We therefore focused our mechanistic analyses on POM and PTFE (Section ); results for PVC and PMMA are provided in the Supporting Information to benchmark generality. Dynamic Coupling of Chain Orientation and Friction To determine whether POM and PTFE achieve low friction through similar or distinct molecular processes, we analyzed their shear-induced chain-orientation dynamics. As established in Section (Supporting Information, Figure S1a), both polymers exhibited comparably low steady-state COFs under the present simulation conditions. However, given their contrasting molecular architectures, their frictional dissipation mechanisms may differ substantially. Despite these structural differences, both polymers showed a rapid and nearly identical increase in P 2,x (t) after shear onset and reached comparable steady-state values (final P 2, x ≈ 0.75–0.78; Figure a). The corresponding time evolution of the slab-averaged shear stress τx (t) is shown in Figure b. For both polymers, τx (t) rises rapidly to an early maximum within the first approximately 20 ns and then relaxes toward a steady value, broadly overlapping with the period of most rapid chain alignment. Because the P 2,x trajectories in Figure a are nearly indistinguishable, chain orientationalthough likely necessary for low friction in these systemsis not sufficient to identify how PTFE and POM dissipate shear work. This interpretation is consistent with recent reactive AA/CG studies on PTFE at Fe/oxide interfaces, which suggest that friction reduction can also involve interfacial anchoring and bond dynamics, as well as mesoscopic chain disentanglement; , thus, orientation alone does not uniquely specify the operative dissipation pathway. In the present nonreactive, amorphous models, such interfacial anchoring and tribochemical effects were excluded. This motivated a depth-resolved analysis of how purely physical interactions govern dissipation. We therefore examined where and how the imposed shear work was dissipated within the polymer to assess potential mechanistic differences (Section ). To probe the kinetics of the slab-averaged response, we analyzed the time derivatives dτ x /dt and dP 2,x /dt (Figure c). Both derivatives exhibit an early positive peak at approximately 15–20 ns; thereafter, dP 2,x /dt decays monotonically toward zero, whereas dτ x /dt decays, crosses zero, and exhibits a small transient undershoot before relaxing. This behavior indicates that stress relaxation can slightly lag behind, or partially decouple from, the final stages of alignment. The undershoot occurs slightly earlier in PTFE (≈30 ns) than in POM (≈35–40 ns); however, given its small magnitude relative to trial-to-trial variability and the smoothing and windowing applied in the derivative analysis, this timing difference is considered statistically insignificant. Overall, these slab-averaged signals indicate that, for polymers as structurally distinct as POM and PTFE, identical friction coefficients and similar P 2,x kinetics can mask fundamentally different internal responses. This apparent agreement at the slab-averaged level motivates the depth-resolved and energy-based analyses in Section , which are required to reveal the underlying, material-specific dissipation pathways. Mechanisms of Frictional Energy Dissipation in POM and PTFE To resolve the apparent macroscopic convergence identified in Section , we examined how the imposed shear work is dissipated in POM and PTFE within a nonreactive all-atom framework, with emphasis on molecular channels and depth localization that are not captured by slab-averaged metrics alone. We first performed bulk-level analyses that decompose the polymer potential energy into bonded (bond stretch, angle bend, dihedral/torsion) and nonbonded (van der Waals, electrostatic/Coulombic) components and relate these to slab-averaged structural signatures (Sections –). We then performed a depth-resolved analysis to localize these pathways across the slab thickness (Section ). Energy Component Analysis and Dissipation Mechanisms Decomposing each polymer’s total potential energy into bonded and nonbonded components quantifies how the imposed shear work is redistributed among molecular interaction channels during sliding (Figure ). In POM, reductions in the bonded angular terms dominate over 0–80 ns, whereas the bond-stretch term remains essentially unchanged (Figure a). Nonbonded energies also decreaseconsistent with the modest slab-wide densification described in Section but these changes remain secondary to the bonded angular contributions. Together, these trends indicate that intramolecular conformational relaxation (via angle and dihedral degrees of freedom) is the primary accommodation pathway in POM, with nonbonded interactions providing a secondary contribution. Component-resolved changes in potential energy during sliding. (a) POM; (b) PTFE. Polymer potential energy (Fe excluded) is decomposed into bonded (bond, angle, dihedral/torsion) and nonbonded (van der Waals, “vdW”; electrostatic/Coulombic, “ele”) terms. Curves show the component-wise change ΔE­(t) = E­(t) – E­(t 0 ) over 0–80 ns, where t 0 is the onset of sliding; negative ΔE indicates a decrease in the corresponding energy component. Lines represent across-trial means; shaded bands denote mean ± s.d across ≥3 independent trials. Curves are lightly smoothed for visualization; all statistics are computed from unsmoothed data. In PTFE, bonded terms change only slightly; instead, the largest energy decreases occur in the nonbonded components (Figure b), reflecting reorganization of interchain contacts and longer-range intrachain nonbonded interactions under shear. This nonbonded-dominated decrease is consistent with shear-induced rearrangements of interchain contact networks reported in nonreactive PTFE–PTFE MD, , which primarily reduce nonbonded (van der Waals/electrostatic) energies while leaving bonded terms largely unchanged. Accordingly, we interpret the PTFE response as governed mainly by nonbonded contact-network reorganization, in contrast to reactive Fe/oxide interfaces, where tribochemical anchoring can bias dissipation toward near-surface layers. , While tribochemical reactions are excluded in the present nonreactive framework, auxiliary liftoff checks (see Supporting Information) confirm polymer pickup/material transfer to the Fe counterface even without covalent bond formation. This observation is consistent with adhesion mediated by nonbonded interactions in the present model (Section S2.4). Importantly, these contrasting redistribution pathways can be rationalized by the intrinsic chemical architectures of polymers. POM has an acetal (C–O) backbone with small hydrogen substituents, providing accessible angular and torsional degrees of freedom that can respond under shear by redistributing conformational populations toward lower-energy regions of the bonded potential-energy landscape. By contrast, PTFE bears bulky fluorine substituents and exhibits a strong conformational preference that renders intramolecular distortions (angle/torsion) comparatively costly; thus, shear accommodation proceeds predominantly via the rearrangement of nonbonded contacts rather than large bonded conformational changes. Because these energy components reflect underlying structural evolution, we next examined shear-induced variations in slab-averaged density and intramolecular angle/dihedral distributions. Slab-Averaged Structural Signatures (Density and Intramolecular Distributions) To elucidate the structural origins of the dissipation pathways identified in Section , we analyzed mass-density profiles and slab-averaged intramolecular distributions (bond-angle and dihedral distributions) over the 0–80 ns shear window. The Eulerian profiles shown in Figure are plotted against the box-fixed coordinate, Z lab (origin at the simulation-cell base, Z lab = 0), rather than the material-anchored depth, d mat, thereby enabling direct tracking of absolute spatial evolution within the simulation cell. Polymer mass-density profiles during sliding plotted against the box-fixed coordinate, Z lab. (a) POM; (b) PTFE. Mass density ρ­(Z lab, t) was computed from all polymer atoms (H included; Fe excluded) on a fixed Eulerian grid along the surface-normal, box-fixed coordinate Z lab (ΔZ = 0.10 nm; origin at the simulation-cell base, Z lab = 0). Bins are not reindexed to the instantaneous interface. Curves correspond to t = 0, 20, 50, and 80 ns and show across-trial means (n ≥ 3); shaded bands denote mean ± s.d. For POM, the slab exhibits broadly distributed densification (Δρ ≈ 0.05 g/cm3 by 80 ns; Figure a). In parallel, the dihedral distributions sharpen and the primary bond-angle distribution shifts toward larger reported angles (Figure ), consistent with the dominant decreases in bond-angle and dihedral energies (Section ). Because J-OCTA reports supplementary bending angles (0° = linear), larger reported angles correspond to smaller interior bond angles, indicating shear-assisted relaxation of configurational strain toward force-field equilibrium. Slab-averaged intramolecular conformational distributions in POM. Probability distributions of selected bond angles and dihedrals at t = 0 and 80 ns, averaged over independent trials (n ≥ 3). Bond angles are reported as supplementary (external) bending angles (0° = linear); thus, larger reported angles correspond to smaller interior bond angles. Dihedral angles are defined such that 0° corresponds to the trans conformation. For PTFE, densification is strongly nonuniform, reaching a pronounced subsurface peak (Δρ ≈ 0.15 g/cm3) at Z lab ≈ 7–11 nm beneath the contact layer (Figure b). This trend is qualitatively consistent with reactive MD reports that PTFE sliding on Fe/oxide forms an oriented near-surface shearing layer under reactive conditions, as well as with nonreactive PTFE–PTFE contact models that also show near-surface shear localization. , In contrast to POM, PTFE shows essentially invariant bond-angle distributions and only modest intramolecular changes at the slab-averaged level (Supporting Information, Figure S3). Although both polymers exhibit dihedral sharpening, the energetic implications differ. In POM, this sharpening reflects conformational redistribution associated with a substantial dihedral-energy decrease (Figure a). In PTFE, by contrast, it primarily indicates reduced fluctuations within existing wells and enhanced orientational coherence, consistent with minimal bonded-energy changes (Figure b). Taken together, these slab-averaged structural signatures reinforce the mechanistic contrast inferred from the energy decomposition. In POM, dissipation occurs predominantly through bonded conformational relaxation distributed across the slab, with a secondary nonbonded contribution. In PTFE, dissipation is governed by nonbonded contact-network reorganization that initiates near the interface and peaks in the subsurface. These trends further indicate polymer-specific biases in dissipation channels: the polar, conformationally flexible acetal backbone of POM more readily accommodates shear work via bonded relaxation involving angle and dihedral modes, whereas steric shielding by fluorine in PTFE suppresses this pathway and thereby favors intermolecular nonbonded reorganization near the interface. This motivates the depth-resolved analysis presented in Section . Depth-Resolved Dissipation Mechanisms Building on the structural contrasts between POM and PTFE, we next analyze depth-resolved mechanical, structural, and thermal responses during prolonged sliding along the material-anchored depth, d mat. Velocity Profiles and Shear Localization across Depth The depth–time maps of the normalized velocity deviation, D(d mat, t), show departures from the ideal Couette reference (Figure ) and thus indicate how shear accommodation is redistributed across the film. Persistent deviations confined in depth indicate localization within a narrow near-interface or subsurface region, whereas spatially dispersed and intermittent deviations indicate shear accommodation distributed across the bulk. Because γ̇(d mat, t) = ∂⟨vx (d mat, t)⟩/∂d mat, these deviations provide an operational indicator of where shear is redistributed away from the ideal Couette profile. Depth–time maps of normalized velocity deviation, D(d mat, t), for (a) POM and (b) PTFE. Color indicates deviations from an ideal Couette reference vC(d mat) (a linear profile based on the imposed wall speed v top): red denotes positive deviation (locally faster than Couette), whereas blue denotes negative deviation (locally slower than Couette). Maps are computed from atom-based layers of thickness Δd = 0.3 nm using nonhydrogen polymer atoms (Fe excluded) and are trial-averaged (n ≥ 3). The depth coordinate d mat is referenced to the polymer–slider interface at t = 0, with d mat = 0 at the interface and positive depth increasing into the polymer interior. It should therefore be interpreted as a nominal depth, t = 0-referenced to the initial partitioning. The two polymers exhibit fundamentally different shear-accommodation patterns. In POM, D(d mat, t) fluctuates dynamically in both sign and magnitude over time and depth, consistent with transient, spatially dispersed shear accommodation rather than the formation of a persistent shear band (Figure a). By contrast, PTFE shows a coherent pattern, with deviations confined to a narrow and persistent near-interface/subsurface band throughout the quasi-steady window (40–80 ns), together with plug-like motion in the interior (Figure b), indicating localized shear. To quantify this contrast, we evaluated depth-resolved shear-rate metrics defined along d mat (Supporting Information, Sections S2.3; Figure S4), including the shear-weight CDF (cumulative distribution of depthwise shear rate), median shear depth d mat,50, interfacial shear fraction in the nominal d mat =0–4 nm range, and localization factor Λ­(d mat). PTFE concentrates 32% of the cumulative shear within this near-interface region and reaches the median shear depth at d mat,50 = 5.8 nm. By contrast, POM shows a more diffuse distribution, with an interfacial shear fraction of 24%, d mat,50 = 6.9 nm, and a broader Λ­(d mat) profile. Consequently, shear in POM is distributed across the bulk, whereas in PTFE it is localized near the interface and adjacent subsurface. This near-interface/subsurface localization in PTFE is robust under the present conditions and is qualitatively consistent with near-surface shear bands reported in nonreactive PTFE–PTFE and reactive PTFE–Fe/oxide simulations. ,, Orientation Profiles and Dynamics These distinct shear-accommodation patterns leave clear depth-resolved signatures in the chain-orientation profiles and alignment kinetics (Figures and ). Although both polymers ultimately reach high local P 2,x values (layerwise maxima approaching 0.95) and exhibit similarly high P 2,x values over much of the analyzed depth range, this apparent agreement in absolute magnitude masks distinct modes of structural accommodation. The key difference is not the absolute magnitude of alignment, but the shape of the depthwise profile and the associated alignment dynamics, as highlighted in Figure . Layer indices are defined such that L1 corresponds to the sliding interface (Methods, Section ), and the shaded band in Figure marks the near-interface zone (≈0–4 nm). At L1, both POM and PTFE show suppressed P 2,x , consistent with interfacial pinning, with a more pronounced reduction in PTFE. Beyond this interfacially pinned layer, the profiles diverge: POM maintains a broader high-P 2,x region (>0.8) from L2 to L12, with a relatively shallow mid-depth maximum near L7–L8, whereas PTFE develops a more localized subsurface maximum around L3–L4, followed by a gradual decay from L5 into the interior. Layer-resolved final chain orientation. Layer-resolved orientational order parameter P 2,x at 80 ns for POM (purple) and PTFE (green). Error bars show the standard error of the mean (SEM) across n ≥ 3 independent trials. Layer indices are defined such that L1 corresponds to the sliding interface (Section ), with increasing layer number indicating increasing depth from the interface. The shaded band denotes the near-interface zone (approximately 0–4 nm). The inset shows representative P 2,x values: P 2,x = 1 indicates alignment parallel to the sliding direction, P 2,x ≈ 0 indicates random orientation, and P 2,x = –0.5 indicates orientation perpendicular to the sliding direction. Dashed curves are guides to the eye. The upper axis (“Approx. depth (nm)”) is based on a uniform order-parameter layer spacing of ΔL =1.0 nm, such that L1 corresponds to ∼1 nm and each subsequent layer increases by 1 nm. Layer-resolved alignment dynamics. Mean time for each layer to reach 90% of its final P 2,x (final defined as the 80 ns value) for POM (purple) and PTFE (green). Error bars show SEM across n ≥ 3 independent trials. These profile differences are reinforced by the alignment kinetics (Figure ). POM shows nearly uniform alignment times across depth (∼30 ns), consistent with a bulk-mediated response and, more generally, with the higher conformational flexibility of the acetal backbone. By contrast, PTFE shows pronounced layer dependence: the near-interface/subsurface response is delayed near L2, followed by faster, more uniform alignment from L5 into the interior, consistent with the near-interface localized response identified in Section . Thus, in PTFE, the localized subsurface enhancement in orientational order is accompanied by heterogeneous kinetics, whereas in POM, the broader high-P 2,x region is accompanied by nearly depth-uniform alignment dynamics. This layer-dependent heterogeneity in PTFE is qualitatively consistent with depth-dependent relaxation patterns reported under reactive shear, where near-surface constraints and mesoscale disentanglement can coevolve. In contrast, the near-uniform response of POM is consistent with bonded conformational accommodation distributed across the bulk. Thermal Profiles Differences in shear localization also govern the spatial distribution of frictional heat generation and the resulting thermal profiles (Figure ). Although the surface temperature of POM is slightly higher (Ts ≈ 304.4 K; Figure a), its temperature deviation from the surface, ΔT(d mat) = T(d mat) – Ts , remains nearly flat within the top ∼4 nm and then declines gently, indicating a diffuse thermal field consistent with distributed dissipation. In contrast, PTFE (Ts ≈ 301.5 K) exhibits a distinct subsurface thermal maximum: ΔT(d mat) remains positive up to d mat ≈ 0–3 nm before dropping more steeply than in POM (Figure b). Steady-state temperature profiles versus material-anchored depth, d mat. (a) Absolute temperature, T­(d mat), for POM (purple) and PTFE (green), averaged over 40–80 ns and across n ≥ 3 independent trials. Lines show the trial-averaged mean; open circles mark every second layer. Annotated values indicate the surface temperature, Ts , of the surface layer (L1, 0.0–0.3 nm). Here, d mat is referenced to the polymer–slider interface at t = 0. (b) Temperature deviation from the surface, ΔT(d mat) = T(d mat) – Ts , computed for each trial using its own Ts and then averaged across trials. Error envelopes (±SEM) are omitted because the variations are smaller than the line thickness. Here, ΔT is evaluated against the material-anchored depth, d mat (Section ), whereas the densification maximum is identified using the box-fixed coordinate, Z lab (Section ; Section and Figure ). This spatial offset indicates that the thermal maximum in PTFE (d mat ≈ 0–3 nm) is displaced from the region of maximum structural packing (Z lab ≈ 7–11 nm; see Section and Figure ). Although both signatures arise from the same shear-driven response, they peak at different depths: near-interface dissipation is strongest in the high-shear zone, whereas densification reflects cumulative packing and contact-network reorganization that extends deeper into the subsurface, yielding a more distant compaction maximum. A comparable enhancement of frictional heating in nonreactive PTFE–PTFE contacts has been reported previously, whereas reactive anchoring at Fe/oxide interfaces further reinforces near-surface confinement. Glass-Transition Proxy and Interfacial Thermal Transport Resistance The contrasting dissipation modes are also reflected in the depth dependence of local mobility and thermal transport. A region-resolved glass-transition proxy indicates pronounced near-interface mobility stratification in PTFE: the local T g decreases by ΔT g ≈ 88 K from the interfacial region (R1, d mat = 0.0–1.5 nm) to the subsurface region (R2, d mat = 1.5–3.0 nm), more than twice the corresponding decrease observed in POM (ΔT g ≈ 38 K; Figure S7). This sharp contrastconsistent with a dense, low-mobility interfacial skin transitioning to a more mobile interioraligns with the near-interface confinement of shear accommodation, P 2,x alignment, and frictional heating in PTFE (Figures –). To assess the transport implications of this stratification, we estimated the effective interfacial thermal resistance, RK , and the polymer-side, depth-dependent effective conductivity (k eff(d th) = q/|dT/dd th|) from heat-flux NEMD temperature profiles (Figure ). Additional k eff(d th) profiles are provided in Figure S8. Here, d th denotes the thermal signed distance referenced to the per-trial thermal-center position, Z th. Z th was determined for each trial from the extrapolated Fe- and polymer-side temperatures, and RK was defined at the detected physical interface position, Z int. RK shows pronounced shear-history dependence in PTFE, decreasing by approximately 40% from 6.61 ± 0.53 to 3.98 ± 0.66 m2 K/GW (0 to 50 ns). POM also shows a reduction in the mean RK (3.97 ± 0.37 → 1.79 ± 0.38 m2 K/GW); however, because RK is obtained by extrapolating the temperature fields in the near-contact regions, the absolute values should be interpreted as effective estimates (see Supporting Information Section S4.3 for the sensitivity analysis). Heat-flux NEMD temperature profiles and effective interfacial thermal resistance. Steady-state temperature profiles versus the thermal signed distance, d th ≡ Z th – Z lab, where Z th is the thermal-center reference determined for each trial/state (Supporting Information Section S4.1.4); d th < 0 corresponds to the Fe side and d th > 0 to the polymer side, for POM and PTFE at 0 ns (solid) and 50 ns (dashed) shear-history states under identical imposed heat flux q. Shaded bands denote mean ± SEM (n ≥ 3). Inset: effective interfacial thermal resistance, RK = ΔT int/q (m2 KGW–1); points show individual trials and error bars indicate mean ± SEM. ΔT int is defined as the interfacial temperature discontinuity at the detected physical interface Z int, obtained by linear extrapolation of the Fe- and polymer-side temperature fields to Z int (Supporting Information Sections S4.1.2–S4.1.3 for the interface-detection and extrapolation protocol). The same d th coordinate is used for polymer-side transport analysis in Section and for evaluating the effective local conductivity, k eff(d th). Crucially, the reduction in RK in PTFE coexists with restricted through-thickness heat spreading. After shear, k eff(d th) develops a pronounced subsurface minimum beyond the mixed/skin region (d th ≳ 1 nm), centered at d th ≈ 3.6 nm with k eff,min ≈ 0.134 W m–1K–1 (Figure S8). This behavior creates a series resistance consisting of an entry resistance and a subsurface transport bottleneck. The bottleneck occurs near the depth at which P 2,x exhibits a subsurface maximum (Figure ), indicating that the sharp near-interface structural transition is accompanied by locally reduced thermal-transport efficiency. Consistent with this series resistance, PTFE retains a substantially larger polymer-side temperature-drop magnitude, |ΔT 1→10| = |T(d th = 10 nm) – T(d th = 1 nm)| = 193.2 ± 10.2 K, compared with 110.3 ± 3.8 K for POM under identical heat flux (Figure ). Consequently, our T g proxy provides a chemistry-free baseline for identifying the propensity for stratified transport. Additional interfacial chemistry or longer-time interfacial evolution may further modify the absolute resistance values, but such effects are expected to reinforce the confined versus distributed dissipation modes identified here. , Integrated Interpretation: Contrasting Spatial Modes of Frictional Dissipation We synthesize the depth-resolved mechanical, structural, and thermal responses to identify two distinct spatial modes of frictional dissipation in POM and PTFE. These modes develop during the early transient (0–40 ns) and persist into the quasi-steady period (40–80 ns), as summarized in the integrated schematic (Figure ). Integrated schematic of contrasting spatial dissipation modes in POM (purple) and PTFE (green). (a–d) POM: (a) Shear distributed across the slab interior. (b) Shear-velocity profile relative to the Couette reference (gray dashed); the light-gray band marks the nominal near-interface zone (∼0–4 nm). Depth is normalized (0 = interface, 1 = interior). (c) Steady-state temperature deviation from the surface, ΔT, showing a gentle gradient; the vertical gray dotted line indicates ΔT = 0. (d) Dominant channel: bonded conformational relaxation (angle/dihedral; ○). (e–h) PTFE: (e) Shear localized near the interface. (f) Shear-velocity profile relative to the Couette reference (gray dashed), showing an interfacial shear band and a plug-like interior. (g) Steady-state temperature deviation from the surface, ΔT, with a subsurface maximum and a steep near-interface gradient. (h) Dominant channel: nonbonded contact-network reorganization (vdW/electrostatic; ◇). General: Qualitative schematic; not to scale. Dissipation Distributed across the Slab Interior in POM In POM, frictional energy is dissipated through a distributed pathway accommodated primarily by intramolecular reorganization. Energetically, relaxation is dominated by reductions in bonded angle and dihedral contributions (Figures a and d). This behavior is consistent with the polar acetal (C–O) backbone, which introduces local dipoles and provides multiple low-energy bond-angle and dihedral degrees of freedom to accommodate shear-induced strain. As a result, shear work can be absorbed efficiently through intramolecular conformational relaxation, allowing dissipation to extend into the slab interior via bulk molecular reorganization. This intramolecular pathway is also reflected in the kinematic and structural responses: deviations from the ideal Couette reference are dispersed across the slab thickness (Figures a and b), and the final chain orientation shows a broader high-P 2,x region across the mid-depth layers, with nearly depth-independent alignment kinetics (Figures and ). Consistent with this organization, the thermal response under the imposed heat flux remains diffuse: the NEMD profiles do not develop a pronounced bottleneck for subsurface heat transport, and the temperature deviation from the surface exhibits a comparatively gentle depth gradient (Figures and c), consistent with diffuse heat spreading under identical heat-flux conditions. Dissipation Localized near the Interface in PTFE In contrast, PTFE exhibits a confined dissipation mode localized within the near-interface zone (∼0–4 nm). Its energetic signature is dominated by nonbonded contributions (vdW/electrostatic; Figure b and h), consistent with interfacial packing and contact-network reorganization rather than primarily intramolecular angular relaxation. This confinement is consistent with the fluorinated PTFE backbone being sterically shielded by a dense fluorine sheath, which promotes a rigid, helical conformational preference and suppresses large-amplitude intramolecular angular relaxation. Consequently, frictional work is preferentially dissipated through intermolecular and interfacial packing/contact-network reorganization (nonbonded pathways) within the near-interface zone. This confinement is also evident in the kinematics: shear deviations localize within the near-interface zone, whereas the interior becomes more plug-like (Figures b and f). Structurally, PTFE develops a local subsurface maximum in P 2,x (Figure ), coincident with pronounced mobility stratification indicated by the layer-resolved glass-transition proxy (Figure S7; ΔT g ≈ 88 K from R1 to R2). Transport analysis further shows that sliding improves contact-level coupling and reduces the effective interfacial thermal resistance, RK (Figure ), yet this improvement coexists with restricted through-thickness heat spreading. Specifically, k eff(d th) develops a pronounced subsurface minimum beyond the mixed/skin region (Figure S8), establishing a subsurface transport bottleneck that acts in series with the entry resistance. This series resistance (entry resistance + subsurface transport bottleneck) provides a compact description of the PTFE thermal field: despite reduced entry resistance, the subsurface bottleneck limits heat penetration into the bulk and sustains a substantially larger temperature drop on the polymer side under identical imposed heat flux (Figures and g; |ΔT 1→10| ≈ 193 K vs ≈ 110 K in POM). Implications and Scope Collectively, the POM and PTFE results demonstrate that frictional dissipation is governed by polymer-specific spatial dissipation modes rather than by a universal depth scale or time scale. The characteristic depth scales identified here should be interpreted within the effective spatial resolution of the present analysis (∼0.3–1.0 nm). In PTFE, the emergence of a near-interface series resistance suggests a plausible tendency toward thermal confinement, whereby localized heating can be maintained near the interface despite a subsurface transport barrier, potentially stabilizing a sliding state that remains localized near the interface; testing this feedback across a broader range of operating conditions remains a topic for future study. These conclusions are based on idealized, nonreactive all-atom simulations with atomically smooth metal–polymer contacts and fully amorphous slabs. Although tribochemical reactions, long-term wear and interfacial evolution, surface oxides, and semicrystalline morphology are not explicitly resolved, the present simulations provide a controlled mechanistic baseline for the early organization of the sliding state and for the propensity toward confined versus distributed dissipation pathways. Conclusion Large-scale AAMD simulations reveal two distinct spatial modes of frictional energy dissipation in the low-friction polymers POM and PTFE. After a brief initial interfacial reorganization, each system establishes a dominant dissipation mode during quasi-steady sliding: a distributed mode extending into the slab interior in POM and a mode localized near the interface in PTFE. Although the present trajectories span a sliding distance of 117.6 nm, the persistence of layer- and depth-resolved mechanical, structural, and thermal signatures throughout the quasi-steady window suggests that, once the quasi-steady state is reached, these spatial modes reflect a stable organization of dissipation under the present conditions. Over longer sliding distances, slower structural or interfacial evolution may emerge; however, such changes are expected to refine rather than fundamentally alter the underlying dissipation architecture identified here. This divergence arises from distinct molecular mechanisms. In POM, dissipation is dominated by bonded conformational relaxation through angle and dihedral modes, with signatures of shear accommodation and frictional heating extending several nanometers into the slab. In PTFE, dissipation is driven primarily by nonbonded (van der Waals and electrostatic) reorganization within a dense near-interface region, leading to a confined shear band and a local subsurface maximum in P 2,x . Heat-flux NEMD analyses further support the PTFE mechanism by indicating two coupled sources of thermal resistance: an entry resistance at the interface and a subsurface transport bottleneck within the polymer. Shear history improves contact-level thermal coupling at the immediate Fe–polymer junction, reducing the effective interfacial thermal resistance RK (i.e., the entry resistance). In contrast, heat redistribution into the polymer interior remains constrained by a subsurface minimum in the depth-dependent effective conductivity, k eff(d th), which functions as a transport bottleneck. Together, these contributions form a series resistance. This resistance coincides with a steep near-interface mobility gradient and a localized subsurface thermal maximum that is notably offset from the deeper densification band, underscoring distinct length scales for heat generation and structural packing. Collectively, these layer-resolved analyses establish a molecular framework that links nanoscale dissipation pathways to macroscopic frictional performance. By mapping specific molecular featuresincluding chain polarity, backbone flexibility, and interchain interaction strengthto distinct dissipation modes, the present results provide practical design guidelines for controlling frictional energy dissipation. For applications requiring minimal bulk disruption and sustained lubricity, stabilizing slip localized near the interface while managing subsurface heat dissipation may mitigate excess heating without compromising low friction. Conversely, for energy-absorbing applications such as damping and impact mitigation, promoting dissipation pathways that extend into the bulk is advantageous. Finally, the “confined vs. distributed” modes identified here are derived from deliberately idealized, nonreactive simulations with atomically smooth contacts and fully amorphous slabs, and thus provide a mechanistic baseline. Although tribochemical reactions, long-term wear and interfacial evolution, surface oxides, and semicrystalline morphology fall outside the present scope, the layer-resolved framework introduced here provides a practical foundation for interpreting more complex tribosystems. Extending this framework through reactive and multiscale modeling, and benchmarking thermal-transport descriptors against thermoreflectance measurements (TDTR/FDTR) on well-defined metal–polymer contacts, will be important for quantifying how interfacial evolution modulates the intrinsic spatial dissipation modes identified in this study. Supplementary Material