Surface dynamics of glassy polymers with different molecular weight distributions

Abstract

The surface dynamics of polymers are of great interest in the field of polymer physics. It is generally believed that a ‘liquid-like’ layer exists at the glassy polymer surface, exhibiting enhanced segmental dynamics compared to the bulk. While much effort has been dedicated to elucidating the underlying mechanism of this simple yet fascinating phenomenon, most studies have focused on polymers with a single molecular weight or a narrow molecular weight distribution. In practical applications, polymeric products consist of a broad range of molecular weights. Consequently, composition of polymers may not be expected to be uniform in space, especially in the presence of an interface. In this study, we investigate the surface dynamics of glassy polymers with a bimodal molecular weight distribution using a droplet wetting-dewetting experiment. The results reveal that the surface dynamics of polymer blends are predominantly influenced by a layer of segregated short-chain polymers. However, upon the removal of this segregated layer, the liquid-like layer begins to dictate the surface behaviors.

Investigations of the surface dynamics of polymers focused on polymers with a narrow molecular weight distribution and do not account for polydisperse polymers found in most polymer products. Here, the authors investigate the surface dynamics of glassy polymers with a bimodal molecular weight distribution using a droplet wetting-dewetting experiment.

Introduction

When glassy polymers are confined to a nanoscopic regime, they exhibit many abnormal behaviors, such as a reduced thermal glass-transition temperature (T_g), easily accessing low energy states by aging¹, and increased plasticity². This phenomenon, drawing significant attention in the field of polymer physics, is commonly attributed to a liquid-like or, precisely, a rubbery-like layer³ on the glassy polymer surface^4–8. Such a layer, exhibiting enhanced segmental and molecular mobility, has also been observed in various other glass systems, such as metallic glasses and ordinary molecular glass formers^9–13.

Therefore, understanding the molecular origin of this liquid-like layer, its associated abnormal dynamic and thermodynamic behavior, and how it affects material properties remains an important task in the glass community^14–17. While substantial progress has been made in recent decades, most studies have primarily focused on polymers with a monodisperse molecular weight distribution. In practice, however, many polymeric products have a wide range of molecular weights, resulting in enrichment of shorter chains at the surface because surface-induced entropic loss is smaller for shorter chains than longer chains, as shown in many recent studies^18–23. When such segregation occurs, the surface dynamics would be dominated by shorter chains. Given the distinct dynamic behaviors between short-chain and long-chain polymers, it is crucial to determine what governs the surface dynamics of glassy polymer blends: the liquid-like layer or the surface-segregated layer^24,25.

Despite both the liquid-like layer and the surface-segregated layer originating from the presence of a surface, they arise from fundamentally different physical mechanisms. Here, the liquid-like layer refers to a reduction in intermolecular constraints at the surface due to the absence of neighboring segments, resulting in enhanced segmental mobility. In contrast, the surface-segregated layer arises from compositional inhomogeneity, in which shorter chains preferentially accumulate near the surface.

Importantly, these two mechanisms are not mutually exclusive and may coexist at polymer surfaces. For example, a surface region enriched with shorter chains may simultaneously exhibit liquid-like dynamics. However, because these mechanisms have distinct physical origins, it remains unclear which effect primarily governs the observed surface dynamics in glassy polymer systems, and under what conditions one mechanism dominates over the other. Addressing this question is essential for a quantitative understanding of surface mobility in realistic polymer systems with polydisperse molecular weight distributions.

To clarify the effect of surface segregation, we prepared polymer blends with a bimodal molecular weight distribution and investigated the surface dynamics of thin polymer films by observing the distinctive interfacial wetting behavior of an immiscible ionic liquid droplet on their surfaces, as we previously reported³. This study aims to unravel the distinctions between the effects of surface segregation and the liquid-like layer on the dynamics at the polymer surface near T_g.

Results

Segregation of short and long chains in simulations

Before experiments, we first conducted molecular dynamics simulations of a mixture containing short and long chains to demonstrate surface segregation, specifically, the enrichment of short chains at the surface. In the initial states, half of the surface beads are from long chains, and the other half from short chains. We also performed an additional simulation with a different initial configuration containing more long-chain than short-chain beads at the surface, which yielded similar surface segregation (see Fig. S1 in SI). Figure 1b shows the final state of one simulation, which reveals the accumulation of short chains at the surface, i.e., the polymer-vacuum interface. To quantify this accumulation, we calculated the bead densities at the surface ($z_{\mathrm{bead}}\in \left[\mathrm{25,\; 28}\right]$) for short and long chains, denoted as ${\rho }_{\mathrm{short}}^{\mathrm{surface}}$ and ${\rho }_{\mathrm{long}}^{\mathrm{surface}}$ (Fig. 1c). Figure 1d shows the ratio of the two densities, which was averaged over six simulations. This ratio increases from 1 to 4 during simulations.

Fig. 1. Segregation of short and long chains in simulations.

a,b Initial and final states of the simulation box. The short and long chains are marked in blue and red, respectively. c Time evolution of bead density at surface ($z_{\mathrm{bead}}\in \left[\mathrm{25,\; 28}\right]$) for short and long chains, denoted as ${\rho }_{\mathrm{short}}^{\mathrm{surface}}$ and ${\rho }_{\mathrm{long}}^{\mathrm{surface}}$. d Time evolution of the ratio ${\rho }_{\mathrm{short}}^{\mathrm{surface}}/{\rho }_{\mathrm{long}}^{\mathrm{surface}}$. e Examples of long and short chain conformations at the surface and bulk. $⟨z⟩$ is the average z-coordinate of the beads of a chain conformation, and $z_{std}$ is the standard deviation of the z coordinates. f We grouped short and long chains in different bins according to their $⟨z⟩$ values and then calculated the average $z_{std}$ in each bin.

Next, we provide a simple physical picture to understand this segregation phenomenon, complementary to the mechanisms in previous studies, such as the chain-end effect^20,26–28. The chain conformations are restricted or compressed by the surface. A longer chain experiences a stronger surface compression and has more entropic loss, which causes less abundance at the surface. To elaborate on this idea, Fig. 1e presents examples of long and short chain conformations at the surface and bulk. For each chain conformation, we label 〈z〉, the average z-coordinate of the beads, and $z_{std}$, the standard deviation of the z coordinates. In Fig. 1f, we grouped short and long chains in different bins according to their $⟨z⟩$ values and then calculated the average $z_{std}$ in each bin. Moving a long chain from bulk $\left(z\approx 15\right)$ in Fig. 1f to surface ($z>25$) dramatically reduces $z_{std}$, while moving a short chain from bulk to surface has a weaker effect on $z_{std}$. Obviously, the surface-compression effect is stronger for the longer chain.

To quantitatively understand ${\rho }_{\mathrm{short}}^{\mathrm{surface}}/{\rho }_{\mathrm{long}}^{\mathrm{surface}}\approx 4$, we can perform a very rough free energy analysis. In the case of an ideal chain, compression by a slit increases the free energy by ${\left(R_{\mathrm{bulk}}/R_{\mathrm{slit}}\right)}^2{k}_{B}T$, where $R_{\mathrm{slit}}$ and $R_{\mathrm{bulk}}$ are the confined and free-space conformational size along the confinement dimension. Because surface beads with $z_{\text{bead}}>25$ come from polymer conformations spanning a relatively wide range of depths to the surface, it is not straightforward to calculate the surface-confinement free energy rigorously. Instead, we provide a rough estimate. Specifically, we approximate that surface beads with $z_{\text{bead}}>25$ effectively come from polymer conformations with a depth of 4.5, and we estimate the confinement free energy using ${\left(R_{\mathrm{bulk}}/R_{\mathrm{slit}}\right)}^2{k}_{B}T\approx {\left(6/4.5\right)}^2{k}_{B}T\approx 1.8k_{B}T$. The surface-confinement free energy for short chains is much less than that for long chains and is neglected here. The logarithm value of the enrichment ratio: $\ln \left({\rho }_{\mathrm{short}}^{\mathrm{surface}}/{\rho }_{\mathrm{long}}^{\mathrm{surface}}\right)\approx 1.4$, which is close to the estimated surface-confinement free energy of 1.8 $k_{B}T$. See the Supplementary Information for additional calculation details, including the rationale for using an average depth of 4.5 in the above estimate.

Wetting behavior of monodisperse polymers

Figure 2 shows the two-dimensional (2D) wetting profiles of PS thin films with a monodisperse molecular weight distribution at different temperatures, as measured by AFM. The change in contact angle during droplet contact with the polymer surface was also determined by a tensiometer. We found that the polymer surface deformation caused by the droplet changes with the temperature. When the temperature is close to the polymer’s T_g, an upward-growing wetting ridge appears on the polymer surface. Panels e and h illustrate the wetting ridge profiles formed on the surface of PS2.5k and PS601k films at temperatures slightly above T_g, respectively. Meanwhile, the contact angle between the droplet and the film remains roughly constant with time, meaning that the droplet was pinned on the film’s surface, such as $T\le 353K$ in Fig. 2d and $T\le 393K$ in Fig. 2g. With an increase of temperature to well above T_g, the liquidized polymers can no longer prevent the droplet from spreading, and the contact angle of the droplet decreases with time (Fig.2d, g), resulting in a wide trench forming on the ionic liquid side, as shown in 2f, i. At sufficiently high temperatures, the droplets wet downward until they reach the substrate. As the trench widens, a significant height difference appears on the two sides of the wetting ridge, corresponding to the film and the substrate, respectively. It is noteworthy that the pinning to the wetting transition of the ionic liquid droplet strongly correlates with the glass transition of PS, specifically the glass transition of PS chains in contact with the ionic liquid droplet, i.e., the surface/interface.

Fig. 2. AFM profiles and contact angles for the wetting on monodisperse polymers.

a–c Illustration of the wetting behavior at different temperatures. d Relative contact angle of the droplet on the short-chain sample (PS2.5k). e,f AFM profiles at 600 s for short-chain sample. g–i Results of the same experiments for long-chain samples (PS601k). Glass-transition temperatures of short-chain (PS2.5k) and long-chain (PS601k) samples are 337 K and 370 K, respectively.

Wetting behavior of polymer blends

To directly probe the surface dynamics of polymer blends, we investigated the wetting behavior of an ionic liquid droplet placed on their surfaces. In this study, we used φ to quantify the weight percentage of high molecular weight PS chains in the blend system.

Figure 3a–c illustrates the distinct wetting behaviors of the polymer blends at different temperatures. As shown schematically in Fig. 3c, when the temperature is above the blend glass-transition temperature, $T_{\mathrm{g},\mathrm{blend}}$, the wetting profiles (Fig. 3f) are similar to those observed for monodisperse polymers.

Fig. 3. Wetting behaviors for polymer blends.

a–c Illustration of the wetting behavior at different temperatures. d–f AFM profiles at different temperatures. g Height difference for temperatures below the respective T_{g, blend}. To avoid data overlapping, we randomly and slightly shifted the x-values of the data points when making the plot. h Variation of T_{g, blend,} and T_transition as a function of the blending ratio. The thin red dashed line is the calculated T_g by the Fox equation for two components: PS2.5k (T_g = 337 K) and PS601k (T_g = 370 K). Error bars represent standard deviation (SD) from n = 3 independent experiments.

When the temperature decreases below $T_{\mathrm{g},\mathrm{blend}}$, however, we observe an unusual wetting behavior for the polymer blends, as illustrated in Fig. 3b. The corresponding AFM profiles exhibit a distinct step-like feature with a height difference of approximately $d\approx 1\mathrm{nm}$ (Fig. 3e).

Upon further cooling, the system undergoes another transition at a temperature denoted as $T_{\mathrm{transition}}$, below which the droplet exhibits a pinning behavior similar to that observed for monodisperse polymers. The corresponding AFM profile is shown in Fig. 3d. The key distinction between the profiles in Fig. 3e and Fig. 3d is the presence of the step-like feature in Fig. 3(e). Accordingly, we define $T_{\mathrm{transition}}$ using the criterion that the height difference approaches zero, i.e. $d\to 0$.

It is worth noting that $T_{\mathrm{transition}}$ does not correspond to the glass-transition temperature of the segregation layer itself. Even at temperatures below $T_{\mathrm{transition}}$, nanoscale deformation of the film surface can still be observed, indicating that the segregation layer retains finite mobility and remains slightly above the $T_{\mathrm{g}}$ of the segregation layer. The corresponding surface profiles measured below $T_{\mathrm{transition}}$ are shown in Fig. S16.

To systematically characterize the pinning-to-wetting transition shown in Fig. 3a, b, we performed measurements over a range of blending ratios. In each experiment, droplets were placed on films with a specified blend ratio for 600 s at different temperatures and then measured at 288 K with AFM (Fig. S15). The resulting values of $d$ are summarized in Fig. 3g. Remarkably, the transition corresponding to $d\to 0$, which marks the pinning-to-wetting transition, is largely insensitive to the blending ratio. Based on these measurements, we extracted $T_{\mathrm{transition}}$ for different blending ratios, as shown by the blue dots in Fig. 3h. Surprisingly, for all samples studied, $T_{\mathrm{transition}}$ remains nearly independent of φ and stays close to $~313$ K.

For comparison, Fig. 3h also shows the glass-transition temperatures of the long- and short-chain components, $T_{\mathrm{g},\mathrm{long}-\mathrm{chain}}=370$ K and $T_{\mathrm{g},\mathrm{short}-\mathrm{chain}}=337$K, respectively, together with the blend glass-transition temperatures. The values of $T_{\mathrm{g},\mathrm{blend}}$ exhibits the expected linear dependence on φ, consistent with the Fox equation.

A striking feature of Fig. 3h is that the pinning-to-wetting transition temperature lies far below $T_{\mathrm{g},\mathrm{blend}}$. This behavior can be naturally explained by surface segregation: the enrichment of short chains near the surface substantially enhances the surface mobility of the blend films.

Considering the lack of correlation between the T_transition and φ of the polymer blends, it is reasonable to conclude that this unusual pinning-wetting behavior is caused by a surface layer of polymer blends with a different chain composition compared to the bulk, rather than the presence of a free surface layer. Namely, the surface dynamics of polymer blends are dominated by a layer of short PS chains enriched at the sample surface, referred to as a surface segregation layer. It is worth noting that in our experiments, φ is limited to a range from 50% to 80%. For φ less than 50%, the difference between the segregation layer and the bulk is too weak to detect a step-change profile and measure the d value. It shows that the low-φ data will demonstrate no d → 0 transition. That will confirm that the PS601k component is essential for the d → 0 transition. However, for φ greater than 80%, the pre-annealing temperature would cause the evaporation of short-chain polymers.

In the preceding sections, the dwell time of the ionic liquid droplet was fixed at 600 s. We now investigate how the dwell time affects the wetting–pinning behavior of the ionic liquid droplet on the polymer thin films. Two temperatures were examined: one slightly above $T_{\mathrm{transition}}$($1.03T_{\mathrm{transition}}=323$ K) and the other slightly below $T_{\mathrm{g},\mathrm{blend}}$($0.97T_{\mathrm{g},\mathrm{blend}}=342$ K). The dwell time was varied from 60 s to 6000 s.

Figure 4 presents the evolution of the film surface profiles with increasing dwell time. The left panel of Fig. 4 shows that at 323 K, the complete process of trench formation, trench widening, and eventual development of a stepped profile can be clearly observed. This behavior resembles the widening trench observed in Fig. 2f, i, except that here the phenomenon occurs on the nanoscale due to the presence of the segregation layer.

Fig. 4. Time evolutions of the surface profiles.

The left and right panels are 323 K and 342 K, respectively. The blending ratio is φ = 0.5.

The right panel of Fig. 4 shows the behavior at 342 K. At this higher temperature, the material beneath the segregation layer, which contains a larger fraction of long chains, is no longer completely frozen. During the initial stage (within the first 30 min), the film surface also exhibits a stepped profile. However, at longer times, a wetting ridge gradually develops at the step boundary. We speculate that this ridge formation is associated with the creation of new free surfaces and the capillary-driven deformation of the material beneath the segregation layer under interfacial tension.

It is worth emphasizing again that the experimentally observed $T_{\mathrm{transition}}$ does not correspond to the glass-transition temperature of the segregation layer itself but lies slightly above it.

Finally, we eliminated the pre-annealing step for the polymer blends to suppress the formation of a surface-segregated layer and repeated the experiment described above. As shown in Fig. 5, samples without pre-annealing exhibit profiles similar to those of monodisperse polymers. This further supports the conclusion that, under thermodynamic equilibrium conditions, intrinsic surface enrichment predominantly dictates the interfacial behavior of polymer blends.

Fig. 5. Comparison of annealed and unannealed samples.

The right panel shows the surface profiles of the annealed and non-annealed samples after wetting for 10 min at T = 343 K, which is lower than the T_{g, blend}. The left panel is the schematic diagram.

Surface segregation is not unique to polymers; it can arise in many multicomponent systems with size disparities²⁹. In addition to entropy-driven segregation, solvent evaporation during sample processing can also drive smaller components toward the surface, an effect particularly pronounced in colloidal systems³⁰. In our measurements, however, the segregation-induced wetting behavior is markedly stronger in thermally annealed samples than in samples prepared without pre-annealing. Since all blend films were annealed above T_g for an extended time (120 h), the nonequilibrium concentration gradients generated during spin coating should largely relax, allowing the system to approach thermodynamic equilibrium.

Comparison of simulations and experiments

Our experiments cannot directly quantify the surface enrichment of short chains observed in the simulations. To enable a semi-quantitative comparison, we evaluated the short-to-long chain ratio in the simulations as a function of distance from the free surface. The results suggest that the thickness of the surface-enriched layer is approximately four bead diameters (Fig. S6).

Experimentally, the thickness of the surface segregation layer is expected to be comparable to the trench height, such as the one shown in Fig. 3e. Using an estimated length of 0.25 nm per styrene repeat unit, the segregation-layer thicknesses inferred from the simulations and experiments are found to be of the same order, indicating reasonable consistency between the two approaches.

Discussion

We investigated the wetting-pinning behavior of an ionic liquid droplet on polymer thin films. We observed a wetting-pinning transition at a temperature higher than the bulk T_g for polymers with monodisperse molecular weight distribution. In contrast, for polymer blends with a binary molecular weight distribution, a much lower wetting-pinning transition temperature was observed. The transition temperature, T_transition, was found to be independent of the blend ratio of the high-molecular-weight component, φ. Guided by our simulation results on surface segregation of chains with different lengths, we conclude that in polymer blends composed of both short and long-chain polymers, the enrichment of short-chain polymers at the polymer blend’s surface dominates the overall surface behavior of polymer blends. By omitting the pre-annealing step during sample preparation, we eliminated surface segregation of short-chain polymers. All wetting-pinning behaviors of the ionic liquid on these samples were consistent with those of the monodisperse samples. In short, short-chain surface segregation significantly alters the surface composition of polymer blends, leading to a surface behavior dominated by short-chain polymers. In practical applications, considering the inherent polydispersity of all polymers, one should be very cautious in interpreting any unusual glassy surface behaviors observed.

Methods

The simulation box contains 30 long chains (length $L=160$) and 600 short chains ($L=8$), which correspond to 4800 beads in short chains and long chains, respectively (Fig. 1a). Each chain is modeled as a string of beads. A bending potential is applied to every three adjacent beads: $E_{\mathrm{bend}}/k_{B}T={\alpha }^2$, where $k_{B}T$ is the Boltzmann energy and $\alpha$ is the bending angle. The bead-bead interaction is described by Lennard-Jones potential interaction: $E_{LJ}\left(r\right)/k_{B}T=4\left[{\left(\sigma /r\right)}^{12}-{\left(\sigma /r\right)}^6\right]$ with a cutoff of 2.5σ, where σ is bead diameter and treated as the length unit. The simulation box has lengths of 20, 20, and 32 along x-, y-, and z-axes. Periodic boundary conditions are applied along the x- and y-axes. Along the z-axis, a wall potential was imposed at z = 0, while a free boundary condition was applied at z = 32 (see the SI for details). For each simulation, we run $5\times {10}^8$ steps in the LAMMPS program³¹ and save polymer conformations every $2\times {10}^3$ steps for analysis. The timestep is 0.01${\tau }_{LJ}$, and ${\tau }_{LJ}$ is the dimensionless time unit. We run six simulations with different random seeds.

We then sought to experimentally investigate the surface dynamics of glassy polymers with a bimodal molecular weight distribution using a droplet pinning-wetting experiment. Atactic polystyrene (PS) (PS601k, Mw = 601 kg/mol, PDI = 1.07, and PS2.5k, Mw = 2.5 kg/mol, PDI = 1.08) was procured from Polymer Source Inc. (Canada) and served as the high (long-chain) and low (short-chain) molecular weight components in the polymer blends. PS (PS3k, Mw = 3.0 kg/mol, PDI = 1.09; PS4.4k, Mw = 4.4 kg/mol, PDI = 1.05; PS7.1k, Mw = 7.1 kg/mol, PDI = 1.07, and PS10.9k, Mw = 10.9 kg/mol, PDI = 1.09) with different molecular weights and narrowly dispersed distribution from the same company were employed as the control groups for the experiments. The ionic liquid 1-ethyl-3-methylimidazolium tetrafluoroborate ([Eim]BF4), purchased from Sigma-Aldrich Chemical Co. (USA), was used as the testing liquid for the wetting experiments. [Eim]BF4 is immiscible with the PS; no plasticizing or swelling effects were observed at the surface of PS³.

Silicon (100) wafers were cut into 1 × 1 cm² pieces, serving as the substrates for the polymer films. The cleaning process of silicon wafers involved submersion in piranha solution (H₂SO₄/H₂O₂, 7:3) preheated to 373 K for 5 min, followed by rinsing with deionized water and drying with nitrogen gas. Thin films were fabricated by spin-coating. The weight percentage of the high molecular component in the polymer blends was set at 50%, 60%, 70%, and 80%, respectively. The monodisperse samples were annealed under vacuum at a temperature above their T_g for 24 h to eliminate residual organic solvents and release internal stress. Alternatively, the blend films with different composition ratios were annealed in a vacuum at 368 K for 120 h to induce short-chain surface segregation.

The thermal T_g of the film of monodispersed PS is approximately 370 and 337 K for PS601K and PS2.5 K, respectively, measured by an RC2 Ellipsometer (J. A. Wollman Inc.) during cooling at 1 K/min. These T_g values for the blend samples align with the Fox equation.

In the wetting experiment, 4 µL of the ionic liquid was placed on the surface of the polymer film, which had been preheated for 5 min on a heat stage to reach thermal equilibrium. The time-dependent contact angle of the ionic liquid was then measured using a surface tension apparatus (OCA 15EC from Dataphysics-Instruments Inc.). After 600 s of wetting experiments, the films with a droplet atop were quenched by quickly transferring the films from the heat stage to an ice bag. The ionic liquid droplet was then removed using filter paper, and the films were rinsed with deionized water.

The capillary force of the droplet induces nanoscale deformation on the mobile surface layer of the glassy films^10–12. The three-dimensional profile of the wetting-induced deformation on the film surfaces was imaged by an atomic force microscope (AFM) (Cypher-ES; Oxford Instruments; UK) under tapping mode at a temperature of 288 K.

Author contributions

Y.C. conceived the project. Y.C., L.D., and B.Z. designed and supervised the project. S.W., Y.Z., Y.F., and X.Z. performed the experiments and simulations. S.W. and Y.Z. analyzed the data and wrote the manuscript. All authors discussed the results and contributed to the final manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the Research Grants Council of Hong Kong (No. 21304421 to Y.C.; No. 11313322 and No. 11307224 to L.D.); the National Natural Science Foundation of China (No. 22003053 to Y.C.; No. 22273080 to L.D.; No. 22525303 to B.Z.); the Natural Science Foundation of Sichuan Province, China (No. 2026NSFSC0373 to Y.C.); and the Guangdong and Hong Kong Universities ‘1 + 1 + 1’ Joint Research Collaboration Scheme (to Y.C.).

Data availability

The experimental data that support the findings of this study are available from the corresponding author upon request.

Code availability

The LAMMPS input scripts and custom analysis codes used in this study are available from the corresponding author upon request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77601-1.