WorksPhysics and Astronomy
Shape of semiflexible polymer chains: Lévy walks vs. Persistent random walks
Abstract
The conformational size and shape characteristics of semiflexible polymer chains, modeled via two distinct frameworks: persistent random walks (PRW) and Lévy walks (LW), are analyzed through numerical simulations. We introduce a mapping that treats the mean jump length of a Lévy distribution () as an effective persistence length of the semiflexible polymer chain, controlled by the exponent . While the conformational characteristics of the PRW model statistically exhibit a sharp crossover from flexible coil to the rigid-rod-like behavior in the immediate vicinity of the maximum persistence probability , the LW model shows a much richer conformational behavior. We obtain the quantitative estimates for universal shape parameters, such as asphericity and prolateness , which gradually increase with decreasing parameter (and thus with an increase of the persistence length), thus supporting previously observed findings for semiflexible polymers. Our results demonstrate the effictiveness of applying Lévy-walk statistics to capture the universal conformational properties of semiflexible polymer structures.
Introduction
Random walks (RW) constitute one of the central paradigms of statistical physics by describing motion as a sequence of random steps [1]. They provide, for example, a way to model biological movement and transport on networks [2, 3]. For unbiased, independent steps with finite variance, the mean-square displacement grows linearly with the number of steps. When each step takes equal time, this gives for the mean-square displacement the Brownian diffusion law [4]:
This behavior follows from the central limit theorem and forms the basis of the conventional theory of diffusion. Many complex systems, however, exhibit transport properties that deviate substantially from Brownian behavior. Such systems display anomalous diffusion, characterized by a nonlinear growth of the mean-square displacement [5, 6, 7, 8, 9, 10, 11]:
Of particular interest is the superdiffusive regime, , which emerges when rare but exceptionally long displacements dominate the dynamics. A natural description of such processes is provided by Lévy statistics, in which the distribution of step lengths possesses a heavy power-law tail,
leading to in Eq. (2) (see e.g. [12]). While idealized Lévy flights allow instantaneous, arbitrarily long jumps, physically realizable systems are more naturally described by Lévy walks, in which the propagation velocity remains finite and long displacements require correspondingly long travel times. In a Lévy walk, the discrete time required to make a jump of a given size is taken to be proportional to the jump size [13, 14, 15]. This modification preserves the essential heavy-tailed character of the motion while maintaining physical consistency. As a result, Lévy walks have become a standard model of superdiffusive transport in complex systems.
In contrast to Gaussian random walks, Lévy processes generate trajectories containing occasional long jumps and lead to fundamentally different scaling properties. They have been observed or invoked in diverse contexts, including turbulent transport, animal foraging, bacterial motion, transport in heterogeneous media, and various non-equilibrium systems [8, 16, 17, 18, 19, 20].
In this paper we intend to discuss the above reasonings in the context of polymer physics. The RW model has been successfully applied to describe flexible polymer molecules in a so-called solvent (Gaussian polymers) [21, 22]. Within this framework, each consecutive step of the RW trajectory represents a single monomer of the polymer chain, so that the propagation time in Eq. (1) is equal to the number of monomers in the corresponding macromolecule. In the present study, we focus only on Gaussian polymer structures, neglecting excluded-volume effects arising from mutual-avoidance interactions between monomers. Despite its simplicity, RW model perfectly captures the universal conformational properties of flexible polymers. In particular, for a size measure such as the mean-square displacement of an -step RW, the relation in Eq. (1) holds.
Many synthetic polymers, such as polyethylene, possess simple carbon backbones with monomers that can rotate around single bonds, making the chains highly flexible. On the other hand, complex organic macromolecules such as DNA, collagen, proteins [23, 24, 25, 26, 27] contain bulky repeating units with complicated molecular interactions that restrict bond rotation and are typical representatives of a class of semiflexible polymers. Theoretically, such polymers are described within the framework of the worm-like-chain (or Kratky–Porod polymer-chain) model [28, 29]. The key structural characteristic here is the persistence length , which is proportional to the bending rigidity of the chain. In the range , macromolecules statistically behave as flexible chains with an effective monomer size of the order of and obey the scaling relation in Gaussian regime
On the other hand, for the case of extremely large values of (), the polymer approaches the limit of a completely extended rigid rod with
To take the bending rigidity of semiflexible polymers into account, the model of biased, or persistent, random walks (PRW) is used [30, 31, 32]. In this model, higher probability is introduced for “trans” steps (taken in the same direction as the previous step) as compared to “gauche” steps (leading to bending of the trajectory). The constructed trajectories thus constitute the sequences of rigid segments of average length with random relative orientations, where each segment, in turn, comprises a number of monomers proportional to .
Our idea is to capture the physics of semiflexible polymers using a Lévy walk (LW) model. In this representation, the trajectory is viewed as a sequence of straight segments with lengths drawn from the Lévy distribution in Eq. (3). The propagation time along each straight segment is equal to the number of monomers it comprises. The average jump length , evaluated according to the distribution in Eq. (3), is therefore directly related to the concept of the persistence length of semiflexible polymers.
While the universality of polymer chains in good solvents is well known through scaling exponents, such as the mean square displacement exponent [Eq. (2)], several universal aspects of polymer shape continue to draw significant research attention. Our primary goal will be to analyze such shape properties of semi-flexible polymer chains. The spatial shape attained by a typical polymer conformation in a solvent is of great importance, as it determines the mobility and folding kinetics [33], as well as the catalytic activity [34], of biopolymers. The hydrodynamics of polymer fluids is also strongly affected by the size and shape properties of individual macromolecules [35].
To characterize the effective shape of polymers in -dimensional space, it is standard to use quantities such as the average asphericity and prolateness ; here and below, denotes an average over an ensemble of all possible polymer configurations [36, 37]. The asphericity parameter is strictly constrained to the interval . A value of zero represents a perfectly symmetric spherical conformation, whereas the upper bound of unity characterizes a fully extended, rigid rod-like configuration. The quantity is defined only for . It takes positive values for prolate, ellipsoid-like configurations and negative values for oblate shapes, and is generally bounded by . Rigorous mathematical definitions of these quantities will be given below.
The size and shape parameters of a wide range of flexible polymer macromolecules with various topologies have been analyzed both analytically and numerically [38, 39, 40, 41, 42, 43]. This includes complex scale-free polymer networks in which the functionalities, or degrees, of the nodes obey a power-law decaying distribution [44]. The conformational shape properties of linear and circular semiflexible macromolecules have been extensively analyzed in Refs. [45, 46, 47, 48]. These works demonstrate that the shape parameters and change continuously as functions of the persistence length, capturing the structural transition from a disordered flexible coil to a fully extended rigid rod.
Lévy-based descriptions of polymers have previously been considered for stiff chains with self-avoidance [49] and for non-Gaussian chains with -stable statistics [50]. Here, we focus on the influence of the Lévy exponent on global conformational characteristics such as radius of gyration, asphericity, and shape anisotropy. These quantities provide important information about the spatial organization of macromolecular structures and serve as sensitive probes of the underlying stochastic dynamics. In the present work, we compare polymer conformations generated by persistent random walks and Lévy walks in three dimensions, neglecting excluded-volume interactions. We determine the radius of gyration and its effective scaling exponent , together with the ensemble-averaged asphericity and prolateness . We also analyze the probability distributions of , , and . By varying the persistence probability in the first model and the exponent and upper segment-length cutoff in the second, we examine how the segment-length distribution affects these conformational characteristics. The rest of the paper is organized as follows. Section II defines the shape parameters, and Section III compares the segment-length distributions of the two models. The simulation procedure and discussion of results obtained are presented in Section IV. We follow with the conclusions in § 0.5.
Polymer shape parameters
The spatial shape characteristics of any polymer macromolecule, represented by a sequence of monomers within a -dimensional space, can be effectively described [51] using the gyration tensor with matrix elements defined as follows:
In this expression, denotes the th Cartesian component of the position vector of the th monomer, and represents the th component of the center-of-mass vector . The trace of the gyration tensor yields the squared radius of gyration :
which serves as a direct measure of the spatial distribution of monomers relative to the center of mass, characterizing the entire chain. A conformation may have a small end-to-end distance while retaining a large radius of gyration.
The anisotropy and shape asymmetry of the polymer can be evaluated based on the eigenvalues of the gyration tensor (6). For perfectly isotropic, spherical shapes, all eigenvalues degenerate to identical values. A completely elongated, rod-like structure is characterized by a single non-zero eigenvalue, while all others vanish. By defining the arithmetic mean of the eigenvalues as , the asphericity is defined as [36]:
This structural parameter is dependent on the spatial dimension even within the framework of the ideal Gaussian chain approximation. To evaluate properties that are directly observable in physical experiments, the ensemble average over all available conformational states is to be performed. However, exact analytical calculations face difficulties while treating averaging (8) directly. An alternative mathematical quantity, , was introduced [36, 37, 43]:
where the numerator and the denominator are averaged independently. The “true” ensemble-averaged asphericity is typically extracted from computer simulations:
In what follows below, we will address the three-dimensional case, . The prolateness in three dimensions is defined by . It is positive for prolate conformations and negative for oblate ones. Both and are evaluated for each configuration before taking the ensemble average. Correspondingly, numerical calculations of the “true” averaged asphericity (Eq. (10)) and prolateness give [52], [53].
Analytical background
We consider two complementary models of semiflexible polymer chains: persistent random walks (PRW) and Lévy walks (LW). Both generate sequences of straight segments, but differ in their segment-length statistics: the distribution for PRW constitutes the sequence of independent trials whereas it follows a power law for LW. We first examine these distributions and their moments, which provide a basis for comparing the conformations generated by the two models.
Persistent random walk (PRW). In a process of growing the trajectory, the probability to perform the next step in the same direction as the previous one is , whereas the probability of changing direction is . Thus, the probability for a straight segment of length (consisting of steps) to appear reads:
The average length of a straight segment (in a case of infinitely long chain ) is given by:
In the semiflexible polymer interpretation, can be viewed as persistence length , the characteristic length scale of local rigidity of polymer chain. Moments of are given in a form:
and are finite at any . In fact, they can be calculated using the Stirling numbers of the second kind [54]:
where with being a binomial coefficient.
In particular, the second moment can be calculated as
With growing , is increasing as expected (see Figure 1), and the polymer chain extends towards the perfectly rigid rod state at , where for an infinite chain. At , the trajectory still forms a random coil on a large scale, however, the effective step size is lengthened. As we can see from Figure 2, trajectories of the same length become more extended and elongated in space as we increase the parameter . The nonlinear increase of becomes particularly pronounced as , providing a direct measure of the persistence introduced by the model. The divergence in Eq. (12) refers to an infinitely long trajectory, in a finite chain, the segment length is bounded by the chain length.

Figure 1. Averaged persistence length of PRW model as function of .
Figure 2. Snapshots of the first 50 steps generated by the two semiflexible polymer models analyzed in this paper: the PRW model (upper panel) and the LW model (lower panel). In the PRW model, parameter defines the persistence probability. In the LW model, segment directions are uniformly distributed, and straight-segment lengths follow a power-law distribution with decay exponent [Eq. (3)]. The parameters are related by [Eq. (18)], ensuring . Colors denote four different pairs of matching parameters .
Lévy walks (LW). The normalized probability density function of the straight segment length at large distances is given by Eq. (3). The average value of segment length is evaluated as:
The mean square is given by
and is divergent at .
We can roughly relate the average segment lengths as given by Eqs. (12) and (16) by:
claiming that by appropriate choice of parameters and in these two models, we can observe the same value of in both of them. Figure 2 presents snapshots of several LW trajectories related to PRW trajectories shown in Figure 2 (with the values of being the inverse of values there, correspondingly). Note that it does not imply the fundamental analogy between models (11) and (3). Even if we tune and in such a way that their first moments match exactly, the global geometry of the constructed polymer configurations will differ due to fluctuations and higher-order moments. In particular, in the Lévy model, the distribution possesses a “heavy tail”:
even when is finite (at ), the second moment diverges for .
Let us introduce the ratio of probabilities (3) and (11) taken under the condition :
and presented at a fixed value of in Figure 3. At small , the power-law decay of the numerator is balanced by the denominator, making the probabilities comparable. However, in the asymptotic regime of large (), the exponential growth of the numerator plays the dominant role. This effect is caused by the non-trivial probability of giant “jumps” to occur in Lévy flight trajectory that are exponentially suppressed within the PRW model.
Figure 3. The relative probability as given by Eq. (19) at fixed
In what follows below, we use numerical simulations to compare the global conformations generated within the two above described models, Eqs. (3) and (11), correspondingly. We generate ensembles of three-dimensional chains and evaluate the size and shape observables defined in Sec. II.
Results and discussions
The method
Both models were simulated in three-dimensional continuous space. For the Lévy-walk model, the length of each straight segment was sampled from the truncated probability distribution Eq. (3). In such distribution the upper cutoff specifies the largest allowed length of an individual straight segment. The density is zero outside the interval . Truncation preserves the power-law dependence within the allowed range but excludes segments longer than , making all segment-length moments finite. Varying allows us to assess how restricting the heavy tail affects the size and shape of the resulting polymer conformations. The optimal truncation is discussed in what follows.
The direction of each segment was chosen independently and isotropically by generating a random unit vector in spherical coordinates. Specifically, the angle was sampled uniformly from , while was sampled uniformly from . This procedure ensures a uniform distribution of directions over the unit sphere. In the PRW model, each unit step retained the direction of the preceding step with probability . With probability , a new isotropically distributed direction was generated using the same spherical-coordinate procedure as in the Lévy-walk model. Consequently, each trajectory consisted of straight segments whose lengths followed the distribution given by Eq. (11). For the main data set, we generated ensembles of independent configurations, each containing steps (monomers). For the LW model, the upper cutoff ranged from to . The observables were calculated separately for each configuration at fixed and subsequently averaged over the ensemble.
Additional calculations at were performed with substantially lower precision. These results are included in Figure 9 as a qualitative reference for the trend at large cutoffs.

Figure 9. Effective exponent as a function of the upper cutoff at variable . The points at were obtained with substantially lower precision and are included as a qualitative reference.
Using this procedure, we first examine PRW model as a reference, and then analyze how the size and shape of LW conformations depend on the distribution parameters.
Persistent random walks
Results for the scaling exponent and shape characteristics and are given in Figure 4 and Figure 5.
Figure 4. Finite-size scaling of the gyration radius (upper panel) and corresponding exponents (lower panel) for different values of persistence probability . The values of were chosen to match via relation (18) values of , , , , correspondingly. FSS shows parallel lines (ergo the same power-law exponents) up until values very close to 1. Even for very large , there is a tendency towards scaling with as the chain length grows.

Figure 5. Asphericity (upper panel) and prolateness (lower panel) of PRW for different values of the persistence probability . Error bars fit within the data points. Note that, on the given scale, the dashed RW line overlaps with the data fit over almost the entire range of .
Across most of the investigated range of , the effective size exponent remains close to the RW (Gaussian) value (see Figure 4 for the finite-size scaling of ), and the mean asphericity and prolateness stay near their RW reference values, see Figure 5. This behavior holds even at high persistence probabilities. Significant structural changes appear only very close to , where the chain stretches into strongly elongated conformations that approach the rigid-rod limit (, , and at ). This trend aligns with a finite-chain crossover governed by the persistence length scale in Eq. (12), through the ratio
For , the trajectory behaves like an ordinary random walk on large length scales. When drops to order unity or below, direction changes become rare and the chain straightens out. Because of this, the crossover sharpens toward as increases. At any fixed , a sufficiently long chain eventually recovers the Gaussian regime, whereas yields a perfectly straight rod regardless of . The sharp transition near thus traces back to the interplay between chain length and persistence length. Consequently, one concludes that the PRW model does not properly describe the expected behavior of a semi-flexible polymer [45, 46, 47, 48], since for almost all values of it belongs to the universality class of a random walk. Therefore, we focus on the LW model in the following.
Lévy walks. Gyration radius and exponent $\nu`
We start with examining the scaling relation (2), remembering that analytical prediction for the case of infinitely long Lévy trajectory is [55, 56].
Figure 6 shows how the gyration radius scales with the chain length at a fixed upper cutoff value . On logarithmic axes, the curves become roughly linear for large , signaling power-law scaling. The variation in their slopes demonstrates that influences both the magnitude of and its effective growth exponent. To quantify this behavior, we fit the data to the two-parameter form . Across all the shown curves, the fitted exponent grows from at up to at . This steeper scaling at smaller marks a clear departure from the Gaussian baseline of . Note that these exponents describe the accessible chain-length window at fixed cutoff, and determining their true asymptotic limits requires a separate finite-size analysis.
Figure 6. Gyration radius as a function of LW trajectory length at various and fixed . Since the leading term of is power-law, in log-log scale we see straight lines as the walk length grows.
An appropriate value of cutoff can be estimated as follows. Based on Eq. (3), where an infinite upper cutoff is assumed, the probability that the step length will exceed some prescribed value denoted as is:
For the -step polymer chain, the average number of cases when exceeds is thus . Only the case when means the adequate choice of . At fixed , should be large enough to satisfy this condition. Taking into account that we are modeling an ensemble of independent trajectories (), each of the length , the total number of steps generated is . Thus, to observe the correct behavior, the cutoff parameter should satisfy the condition .
Figure 7 (and, consequently, Figure 8 and Figure 9) present two complementary views of the same fitted exponents: the dependence on at fixed , and the dependence on at fixed .

Figure 8. Effective exponent as a function of the mean segment length (controlled by parameter ) for different values of upper cutoffs . The red line shows the reference relation for , continued by the Gaussian value for .
Figure 7 shows the effective exponent as a function of for various values of upper cutoff in the range . For , the red line represents the reference relation ; for , it shows the Gaussian value . At the smallest cutoff, the fitted exponents remain close to the Gaussian reference . Increasing the cutoff makes the dependence on more pronounced, indicating that rare long segments affect the size growth within the simulated length range. For sufficiently large value of , the numerical estimates are merging towards the theoretical reference for . For large , the results remain close to the Gaussian value and depend only weakly on the cutoff. The cutoff dependence becomes more pronounced as decreases. Even at , the effective exponent deviates from the analytical prediction at . Thus, within the simulated range, the dependence on is nonmonotonic. The revealed discrepancy indicates that for small the rate of convergence to the asymptotic limit is extremely slow. The theoretical prediction is formulated for infinite trajectories without restrictions on the step length. In numerical simulations as well as in experiments, the introduction of any finite cutoff (even as large as ) creates a strong finite-size effect. The long Lévy “jumps”, which should determine the superdiffusion nature of the motion at , are artificially truncated, which leads to an underestimation of the actual exponent .

Figure 7. Effective exponent as a function of for different upper cutoffs . Lines connecting numerical points are eye guides. The solid red curve shows the reference relation for , continued by the Gaussian value for .
Figure 8 presents the effective exponent as a function of the numerically determined mean segment length . For the distribution with infinite upper cutoff, Eq. (16) gives for . In the range , rewriting the reference relation therefore yields
Together with the Gaussian value for , this
defines the red reference curve. The numerical curves are merging close to the reference curve near but deviate as the mean segment length increases. Thus, within the investigated range, the influence of truncation remains visible in this representation and cannot be neglected.
Figure 9 shows the dependence of the effective exponent on the upper cutoff for fixed values of . For , the curves approach nearly constant values at cutoffs of approximately –. These values are merging the reference relation . For , the fitted exponent remains close to the Gaussian value , albeit slightly above it, throughout the investigated range. For smaller , the cutoff dependence persists over a broader range. In particular, for and , increasing produces a substantial increase in the effective exponent, with no clear plateau within the main range . The additional results at extend this comparison to larger cutoffs, but only as a general trend due to the reduced precision. At small cutoffs the heavy tails of the distribution are completely cut off. The increase in activates the superdiffusion properties of the trajectories, which leads to a pronounced increase in exponent .
This analysis convincingly proves that effective scaling
Figure 10. Asphericity as a function of the chain length for different values of at . The bottom dotted line shows the Gaussian-chain value [52].
Figure 11. Prolateness as a function of the chain length for different values of at . Error bars represent standard errors of the mean, the bottom dotted line shows the Gaussian-chain value [53].
behavior is strongly affected by truncation, especially for small . For distributions with heavier tails, the fitted exponents remain sensitive to over a broader range of cutoffs. The cutoffs required to reduce these effects have therefore to be assessed for each and chain-length range. Determining the asymptotic regime requires a separate analysis of the chain-length dependence.
Lévy walks. Shape characteristics
Figures 10 and 11 show the dependence of the mean asphericity and prolateness on the chain length at fixed value of cutoff . After a short initial transient, both quantities gradually approach nearly constant values. Thus, although the characteristic size of a trajectory continues to grow with , its shape becomes effectively independent of the chain length. The limiting values of both shape parameters depend strongly on the Lévy exponent . For large , the results approach the corresponding values for a Gaussian chain, [53], and [53]. The specific dependence on also varies with the upper cutoff , as discussed below.
The upper panels of Figs. 12 and 13 show a nonmonotonic variation of both shape parameters with . For the smallest cutoff, the curves are nearly flat and stay close to the Gaussian-chain reference values. As grows, clear peaks emerge in both mean asphericity and mean prolateness, shifting toward smaller with increasing cutoff. Across the sampled parameters, these maxima are located near for and near for . Hence, a heavier step-length distribution tail does not automatically yield stronger average shape anisotropy within this range.
The lower panels of Figs. 12 and 13 highlight the cutoff dependence at fixed . At , both observables show little variation once reaches –. By contrast, the data for and remain sensitive to the cutoff up through . At , both shape parameters stay near their Gaussian limits across
Figure 12. Average asphericity as a function of (upper panel), (middle panel) at fixed , and as a function of at fixed (lower panel). For each parameter pair, the -independent ensemble means are further averaged over . Error bars show their standard deviation across this interval. Lines connect the points as eye guides.
all cutoff values. Although reflecting finite chain lengths, the leveling off observed in several curves hints at the asymptotic behavior.
The middle panels of Figs. 12 and 13 present the same shape parameters as functions of the measured mean segment length . This representation retains a strong dependence on : comparable mean segment lengths can correspond to substantially different asphericities and prolatenesses. Within the investigated chain-length and cutoff ranges, alone is insufficient to determine the mean shape characteristics. The full distribution must also be taken into account.
Figure 13. Average prolateness as a function of (upper panel), (middle panel) at fixed , and as a function of at fixed (lower panel). For each parameter pair, the -dependent ensemble means are further averaged over . Error bars show their standard deviation across this interval. Lines connect the points as eye guides.
Lévy walks. Probability distributions
The distributions of the observables provide information that is not contained in their ensemble averages alone. As shown in Figure 14, the distribution of is asymmetric and possesses a pronounced tail towards large values, reflecting the occurrence of configurations containing unusually long segments.
Figure 14. Probability density of the radius of gyration of LW trajectories for , , , and .
Much broader distributions are observed for the shape parameters, as presented in Figure 15. The asphericity distribution has most of its statistical weight away from , showing that nearly spherical configurations are uncommon. The prolateness distribution is predominantly concentrated at positive values of , indicating that prolate configurations occur more frequently than oblate ones. The broadness of both distributions demonstrates that the mean values and represent ensembles containing configurations with notably different geometries.
Figure 15. Probability densities of the asphericity (above) and prolateness (below) of LW trajectories for , , , and different from to . Symbols with error bars represent histogram estimates. Solid lines show empirical fits using an extended Lhuillier form for , and separate power-law () and stretched-exponential () forms for .
To describe the distributions in Figure 15 quantitatively, we use the empirical fitting forms employed in Ref. [57]. For the asphericity, we adopt the extended Lhuillier-like expression
where , , , , and are fitting parameters. This form extends an ansatz originally developed for polymer size distributions [58].
For the prolateness, the branches and are fitted separately by and , respectively [57]. These empirical fits provide compact descriptions of the distribution shapes and allow comparison between different cutoffs beyond their means.
The fitted distributions can also be used to estimate mean values from their normalized first moments:
where denotes the fitted function, , and . Such estimates may provide a complementary consistency check.
Conclusions
The concept of scaling is central to polymer physics, with origins in the foundational works of Kuhn, Flory, Huggins and others. Building on the work of de Gennes and des Cloizeaux [21, 59], the scaling behavior of long, flexible macromolecules in a good solvent was linked to critical phenomena near continuous phase transitions, establishing the basis for applying field-theoretic methods to polymer physics. The random walk and self-avoiding walk models serve as the standard paradigm for modeling flexible polymer chains. The objective of our work is to introduce and test a similar paradigm for describing semiflexible polymers; to this end, we examine two models: the persistent random walks (PRW) and truncated Lévy walks (LW). Within the PRW model, the probability is introduced to enhance the trajectory local stiffness, see Eq. (11). Accordingly, the persistence length gradually increases with , Eq. (12). Within the LW model, the mean step length of a Lévy distribution, Eq. (3), is treated as an effective persistence length of a semiflexible polymer chain.
To test these models, in addition to calculating scaling parameters—specifically, the scaling of the mean-square distance—we also consider universal shape properties, such as asphericity and prolateness, which are essential for analyzing the rheological properties of polymer solutions. It is worth noting that, just as polymer scaling exponents are related to the universal critical exponents of continuous phase transitions, polymer shape parameters are related to universal combinations of critical amplitudes. We focused on the impact of the statistical properties of the segment-length distributions (11) and (3) on the universal size and shape characteristics of semiflexible polymer chains. For the PRW model, we confirmed a crossover governed by the persistence length scale: the trajectory exhibits the Gaussian regime with scaling exponent till the immediate vicinity of , where transition to the rigid-rod regime with occurs abruptly. The same trend was observed for the shape parameters, which are attaining the Gaussian values of a flexible polymer chain ( [52], and [53]) until the abrupt chain stretching at into strongly elongated rigid-rod conformation with and . On the other hand, in the case of LW model the observed behavior is much more rich. Decreasing the Lévy exponent leads to an increase in the numerical estimates of effective scaling exponent from a near-Gaussian value at up to superdiffusive values, merging towards the theoretical reference at sufficiently large values of upper cutoff . The crucial role of the parameter is observed: the extremely long Lévy jumps, determining the superdiffusion nature of the motion are artificially truncated as increases, which leads to an “underestimation” of the exponent for . The shape parameters exhibit a gradual increase with decreasing (increasing the mean length of straight segment of LW trajectory). From this perspective, the LW model confirms the continuous change in the shape parameters and of semiflexible macromolecules with an increasing persistence length, as observed in Refs. [45, 46, 47, 48].
The results obtained support the potential of application of the Lévy-walk models for capturing the geometric size and shape properties of semi-flexible polymers. In particular, we demonstrate that the extended structural segments can fundamentally impact the global shape anisotropy and folding kinetics of semiflexible polymer chains. On the other hand, the constraints connected with a finite cutoff and finite chain length should be carefully taken into account when interpreting simulations within the LW framework. While the obtained values of the universal scaling exponent are compatible with theoretical predictions and indicate a shift in the universality class, our results for the shape parameters and are, to the best of our knowledge, the first numerical observations of their kind and call for further analytical confirmation.
Acknowledgments
We thank Khrystyna Haydukivska, Jaroslav Ilnytskyi and Dmytro Yaremchuk for useful discussions. VB and YuIH acknowledge support from the National Research Foundation of Ukraine under Project No. 2025.07/0366, “Functional efficiency of complex multibranched polymers in adsorption-based applications: computer simulations and predictions.”
References
- [1]K. Pearson, The problem of the random walk, Nature 72, 294–294 (1905).
- [2]E. A. Codling, M. J. Plank, and S. Benhamou, Random walk models in biology, Journal of The Royal Society Interface 5, 813–834 (2008).DOI
- [3]N. Masuda, M. A. Porter, and R. N. Lambiotte, Random walks and diffusion on networks, Physics Reports 716–717, 1–58 (2017).DOI
- [4]A. Einstein, Über die von der molekularkinetischen theorie der wärme geforderte bewegung von in ruhenden flüssigkeiten suspendierten teilchen, Annalen der Physik 322, 549 (1905).DOI
- [5]P. Lévy, Theorie de l’Addition des Variables Aleatoires, Monographies des Probabilites / publiees sous la direction de M. Emile Borel (Gauthier-Villars, 1937).
- [6]J.-P. Bouchaud and A. Georges, Anomalous diffusion in disordered media: Statistical mechanisms, models and physical applications, Physics Reports 195, 127 (1990).
- [7]R. Metzler and J. Klafter, The random walk’s guide to anomalous diffusion: a fractional dynamics approach, Physics Reports 339, 1–77 (2000).DOI
- [8]G. M. Viswanathan, S. V. Buldyrev, S. Havlin, M. G. E. da Luz, E. P. Raposo, and H. E. Stanley, Optimizing the success of random searches, Nature 401, 911 (1999).pubmed.ncbi.nlm.nih.gov/10553906
- [9]G. Viswanathan, V. Afanasyev, S. V. Buldyrev, S. Havlin, M. da Luz, E. Raposo, and H. Stanley, Lévy flights in random searches, Physica A: Statistical Mechanics and its Applications 282, 1 (2000).DOI
- [10]I. Golding and E. C. Cox, Physical nature of bacterial cytoplasm, Physical Review Letters 96, 098102 (2006).pubmed.ncbi.nlm.nih.gov/16606319
- [11]W. Campeau, A. M. Simons, and B. Stevens, The evolutionary maintenance of lévy flight foraging, PLOS Computational Biology 18, e1009490 (2022).
- [12]A. V. Chechkin, V. Y. Gonchar, J. Klafter, and R. Metzler, Fundamentals of lévy flight processes (2006).
- [13]B. Dybiec, E. Gudowska-Nowak, E. Barkai, and A. A. Dubkov, Lévy flights versus lévy walks in bounded domains, Physical Review E 95, 10.1103/physreve.95.052102 (2017).DOI
- [14]V. V. Palyulin, G. Blackburn, M. A. Lomholt, N. W. Watkins, R. Metzler, R. Klages, and A. V. Chechkin, First passage and first hitting times of lévy flights and lévy walks, New Journal of Physics 21, 103028 (2019).DOI
- [15]O. Bénichou, C. Loverdo, M. Moreau, and R. Voituriez, Intermittent search strategies, Reviews of Modern Physics 83, 81–129 (2011).
- [16]T. H. Solomon, E. R. Weeks, and H. L. Swinney, Observation of anomalous diffusion and lévy flights in a two-dimensional rotating flow, Physical Review Letters 71, 3975 (1993).pubmed.ncbi.nlm.nih.gov/10055122
- [17]G. Ariel, A. Rabani, S. Benisty, J. D. Partridge, R. M. Harshey, and A. Be’er, Swarming bacteria migrate by lévy walk, Nature Communications 6, 8396 (2015).DOI
- [18]P. Barthelemy, J. Bertolotti, and D. S. Wiersma, A Lévy flight for light, Nature 453, 495 (2008).DOI
- [19]D. Brockmann, L. Hufnagel, and T. Geisel, The scaling laws of human travel, Nature 439, 462 (2006).DOI
- [20]R. N. Mantegna and H. E. Stanley, Scaling behaviour in the dynamics of an economic index, Nature 376, 46 (1995).DOI
- [21]J. des Cloizeaux and G. Jannink, Polymers in Solution: their modelling and structure (Clarendon Press: Oxford, 1991) p. 994.
- [22]K. Freed, Polymers and random walks - renormalization group description and comparison with experiment, Journal of Research of the National Bureau of Standards 90, 503 (1985).DOI
- [23]C. Bustamante, J. F. Marko, E. D. Siggia, and S. Smith, Entropic elasticity of λ-phage dna, Science 265, 1599–1600 (1994).DOI
- [24]G. Witz, K. Rechendorff, J. Adamcik, and G. Dietler, Conformation of circular dna in two dimensions, Physical Review Letters 101, 10.1103/physrevlett.101.148103 (2008).pubmed.ncbi.nlm.nih.gov/18851579
- [25]J. F. Marko and E. D. Siggia, Stretching dna, Macromolecules 28, 8759–8770 (1995).DOI
- [26]B. Biswas, D. Mitra, F. KP, S. Bhat, A. Chatterji, and G. Kumaraswamy, Ridigity dictates spontaneous helix formation of thermoresponsive colloidal chains in poor solvent, ACS Nano 15, 19702–19711 (2021).DOI
- [27]L. Gerrits, R. Hammink, and P. H. J. Kouwer, Semiflexible polymer scaffolds: an overview of conjugation strategies, Polymer Chemistry 12, 1362–1392 (2021).DOI
- [28]O. Kratky and G. Porod, Röntgenuntersuchung gelöster fadenmoleküle, Recueil des Travaux Chimiques des Pays-Bas 68, 1106–1122 (1949).DOI
- [29]M. Doi and S. F. Edwards, The Theory of Polymer Dynamics, International Series of Monographs on Physics (Clarendon Press, Oxford, 1988).
- [30]J. W. Halley, H. Nakanishi, and R. Sundararajan, Scaling in biased random walks, Physical Review B 31, 293–298 (1985).pubmed.ncbi.nlm.nih.gov/9935424
- [31]J. Masoliver, J. M. Porrà, and G. H. Weiss, Some two and three-dimensional persistent random walks, Physica A: Statistical Mechanics and its Applications 193, 469–482 (1993).
- [32]K. Havdukivska and V. Blavatska, Lattice models of directed and semiflexible polymers in anisotropic environment, Journal of Physics A: Mathematical and Theoretical 48, 425002 (2015).
- [33]R. I. Dima and D. Thirumalai, Asymmetry in the shapes of folded and denatured states of proteins, The Journal of Physical Chemistry B 108, 6564–6570 (2004).DOI
- [34]K. W. Plaxco and D. Baker, Limited internal friction in the rate-limiting step of a two-state protein folding reaction, Proceedings of the National Academy of Sciences 95, 13591–13596 (1998).pubmed.ncbi.nlm.nih.gov/9811844
- [35]J. García de la Torre, O. Llorca, J. L. Carrascosa, and J. M. Valpuesta, Hydromic: prediction of hydrodynamic properties of rigid macromolecular structures obtained from electron microscopy images, European Biophysics Journal 30, 457–462 (2001).pubmed.ncbi.nlm.nih.gov/11718300
- [36]J. Aronovitz and D. Nelson, Universal features of polymer shapes, Journal de Physique 47, 1445 (1986).DOI
- [37]J. Rudnick and G. Gaspari, The asphericity of random walks, Journal of Physics A: Mathematical and General 19, L191 (1986).DOI
- [38]O. Kalyuzhnyi, K. Haidukivska, V. Blavatska, and J. Ilnytskyi, Universal size and shape ratios for arms in star-branched polymers: Theory and mesoscopic simulations, Macromolecular Theory and Simulations 28, 10.1002/mats.201900012 (2019).DOI
- [39]H. Ma and J. Xu, Mean-square radius of gyration of Poly(dimethylsiloxane) chain with side groups, Polymer Journal 26, 779 (1994).DOI
- [40]J. Patrzej, S. S. Sheiko, S. Panyukov, and M. Rubinstein, Molecular structure of bottlebrush polymers in melts, Science Advances 2, 10.1126/sciadv.1601478 (2016).
- [41]Y. Nakamura, Y. Wan, J. W. J. Vayss, H. Iatrou, and N. Hadjichristidis, Radius of gyration of polystyrene combs and centipedes in solution, Macromolecules 33, 8323 (2000).
- [42]P. E. Theodorakis and N. G. Fytas, Molecular dynamics simulations of bottle-brush polymers with a flexible backbone under Theta and good solvent conditions, American Journal of Condensed Matter Physics 2, 101 (2012).DOI
- [43]V. Blavatska and R. Metzler, Conformational properties of complex polymers: rosette versus star-like structures, Journal of Physics A: Mathematical and Theoretical 48, 135001 (2015).DOI
- [44]V. Blavatska and Y. Holovatch, On the shape of gaussian scale-free polymer networks, Journal of Physics: Condensed Matter 37, 145101 (2025).DOI
- [45]K. Alim and E. Frey, Shapes of semiflexible polymer rings, Physical Review Letters 99, 10.1103/physrevlett.99.198102 (2007).DOI
- [46]A. Chen, B. Zhang, and N. Zhao, A comparative study of semi-flexible linear and ring polymer conformational change in an anisotropic environment, Physical Chemistry Chemical Physics 22, 9137–9147 (2020).DOI
- [47]M. Fošnarič, A. Iglič, D. M. Kroll, and S. May, Monte carlo simulations of a polymer confined within a fluid vesicle, Soft Matter 9, 3976 (2013).DOI
- [48]L. Liu, W. Chen, and J. Chen, Shape and diffusion of circular polyelectrolytes in salt-free dilute solutions and comparison with linear polyelectrolytes, Macromolecules 50, 6659–6667 (2017).DOI
- [49]J. Moon and H. Nakanishi, Self-avoiding lévy walk: A model for very stiff polymers, Physical Review A 42, 3221 (1990).pubmed.ncbi.nlm.nih.gov/9904400
- [50]M. Majka and P. F. Góra, Non-gaussian polymers described by alpha-stable chain statistics: Model, effective interactions in binary mixtures, and application to on-surface separation, Physical Review E 91, 052602 (2015).pubmed.ncbi.nlm.nih.gov/26066189
- [51]K. Solc and W. H. Stockmayer, Shape of a random-flight chain, The Journal of Chemical Physics 54, 2756–2757 (1971).
- [52]G. Wei and X. Zhu, Shapes and sizes of arbitrary random walks at o(1/d3) ii. asphericity and prolateness parameters, Physica A: Statistical Mechanics and its Applications 237, 423 (1997).DOI
- [53]O. Jagodzinski, E. Eisenriegler, and K. Kremer, Universal shape properties of open and closed polymer chains: renormalization group analysis and monte carlo experiments, Journal de Physique I 2, 2243–2279 (1992).DOI
- [54]M. Abramowitz and I. A. Stegun, Stirling numbers of the second kind, in Handbook of Mathematical Functions with Formulas, Graphs, and Mathematical Tables, edited by M. Abramowitz and I. A. Stegun (Dover, New York, 1972) 9th ed., Chap. 24.1.4, pp. 824–825.
- [55]M. E. Fisher, S.-k. Ma, and B. G. Nickel, Critical exponents for long-range interactions, Phys. Rev. Lett. 29, 917 (1972).DOI
- [56]P. Grassberger, Critical exponents of self-avoiding levy flights, Journal of Physics A: Mathematical and General 18, L463 (1985).DOI
- [57]O. Kalyuzhnyi, M. Ilnytskyi, Y. Holovatch, and C. von Ferber, Universal shape characteristics for the mesoscopic polymer chain via dissipative particle dynamics, Journal of Physics: Condensed Matter 28, 505101 (2016).DOI
- [58]D. Lhuillier, A simple model for polymeric fractals in a good solvent and an improved version of the Flory approximation, Journal de Physique 49, 705 (1988).DOI
- [59]P.-G. de Gennes, Scaling Concepts in Polymer Physics (Cornell University Press, Ithaca, NY, 1979) p. 326.DOI









