betaFeedback

WorksPhysics and Astronomy

Polymer-polymer interdiffusion: effects of entanglements and a polymeric source

Abstract

Many industrial applications and biological scenarios involve the interdiffusion of two polymeric species. Motivated by biological subcellular source-driven processes, we study polymer-polymer interdiffusion problems in the absence or the presence of a polymeric source, for both unentangled and entangled scenarios. Utilizing a two-fluid formalism, we arrive at scaling relations, self-similar reductions, and analytical solutions, which are confirmed with one- and two-dimensional numerical simulations. The introduction of a source term breaks the self-similar structure, modifying the boundary conditions and the domain of integration. Nevertheless, we show that the front characteristics of the diffusing droplet exhibit similar spatial structures as in the absence of a source. Our results allow deeper understanding of polymer-polymer interdiffusion and nonlinear transport, especially in the presence of a source.

Introduction

Polymers play a crucial role in many technological and biological applications [1, 2, 3]. A single polymer exhibits complex dynamics that span multiple time and length scales [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]. Multiple polymers put together can give rise to entanglements, leading to dynamics such as reptation [16, 17, 18]. Mixing two different polymeric species considerably complicates the dynamics, which subsequently depend on the underlying interactions between the polymer types, their degree of entanglements, as well as their mechanical properties. In such a problem, the polymers can phase-separate, or diffuse into one another — a scenario termed polymer-polymer interdiffusion [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]. Below, we study the effects of entanglement on the spatiotemporal evolution of polymeric droplets undergoing interdiffusion.

Besides their technological value, concepts from polymer physics can help in understanding many biological systems. One of the predominant current frameworks to understand the formation of biological condensates is via liquid-liquid phase separation, which utilizes a similar framework as for polymer phase separation [19, 20, 21, 22, 23]. For example, phase separation in polymer-fluid mixtures exhibits biologically relevant morphologies, such as gel networks and condensates [24, 25, 26, 27, 28, 29, 30, 31]. Consequently, the application of ideas from polymer science to biological scenarios has motivated measurements of mechanical properties of biological condensates, improving our understanding of their rheological responses [32, 33, 34, 35].

While passive processes such as diffusion and phase separation are ubiquitous, many biological processes involve active production of polymers, for example in transcription and translation. One prominent example of such an active source is the nucleolus — a subnuclear condensate synthesizing and exporting ribosomal RNA (rRNA) [36, 37, 38, 39, 40]. The nucleolus acts as a persistent source, driving an outward flux of biopolymers into the surrounding nucleoplasm. Clearly, once such a source is present, the polymer concentration (of rRNAs in the above example) is also affected by the presence of the source.

Here, inspired by such biological condensates, we use a two-fluid formalism [41, 42, 43, 44, 45, 46, 47, 48, 49, 50] to study two types of polymer-polymer interdiffusion problems — one of passive droplet diffusion, and another of source-driven diffusion, both depicted schematically in Figure 1(a)-(b) respectively. We first review in Sec. II the equations of motion of the two-fluid formalism [41, 42, 43, 44, 45, 46, 47, 48, 49, 50], which yield a reduced set of one-dimensional (1D) equations of motion. Then we analyze the passive droplet diffusion cases in Sec. III. We show that under certain approximations, the polymer droplet interdiffusion is equivalent to a nonlinear diffusion problem, allowing us to cast the equations of motion in a self-similar form. We use the 1D formalism to obtain full numerical solutions, and compare those with the self-similar approach. We also demonstrate that our results hold in a two-dimensional (2D) case.

Figure 1. Passive (droplet) diffusion, and source-driven diffusion. (a) A sketch of an A-mer droplet interdiffusing into a B-mer surrounding medium. The A-mers are depicted as red polymers, and B-mers as gray ones. (b) A source-driven problem, where the source is depicted as a small black circle at the center of the droplet. The source generates A-mers, that then diffuse into the B-mers. (c) Plots of f(ϕ)f(\phi) in the absence of interactions (χ=0\chi=0) and for monomer surroundings (NB=1N_B=1) for a wide range of polymer lengths NAN_A. Increasing the polymer length increases the energetic minima, and develops an asymmetry from ϕ=0.5\phi=0.5. (d) Plots of the friction function μ(ϕ)\mu(\phi) for various values of α\alpha (the B-mers are assumed to be unentangled, μB=1−ϕ\mu_B=1-\phi). As α\alpha increases, the friction function approaches 1−ϕ1-\phi (drawn as dashed black line).

Then, we introduce a source and analyze the emerging scaling, the resulting reduced equations of motion, and the corresponding dynamics in Sec. IV. Specifically, we demonstrate that the source introduces an additional flux term, and breaks the self-similar structure of the passive problem. Consequently, instead of using a self-similar ansatz, we perform a perturbation around the expanding droplet’s tip, and show that the local geometry at the front of a source-driven droplet resembles the one observed in the passive diffusion scenario. We compare our analytical predictions against the 1D and 2D full numerical solutions, demonstrating the spatial similarity to the passive case. Finally, we discuss our results and potential implications in Sec. V.

Equations of motion

We consider a two-component mixture, of polymeric species A and B, referred to as A-mers and B-mers, both of the same density [41, 42, 43, 44, 45, 46, 47, 48, 49, 50]. With the A-mer and B-mer velocity fields uA(r,t)\mathbf{u}_A(\mathbf{r},t) and uB(r,t)\mathbf{u}_B(\mathbf{r},t), respectively [below we suppress the explicit (r,t)(\mathbf{r},t) dependence for readability], the volume fractions ϕA\phi_A and ϕB\phi_B evolve via the continuity equations

∂tϕA+∇⋅(uAϕA)=S,∂tϕB+∇⋅(uBϕB)=0,(1)\begin{aligned} \partial_t\phi_A+\nabla\cdot(\mathbf{u}_A\phi_A)&=S,\\ \partial_t\phi_B+\nabla\cdot(\mathbf{u}_B\phi_B)&=0, \tag*{(1)} \end{aligned}

where here we introduced an A-mer source S≡S(r,t)S\equiv S(\mathbf{r},t). Because we consider a binary mixture, ϕA+ϕB=1\phi_A+\phi_B=1, and we denote ϕA=ϕ\phi_A=\phi.

The quasi-static momentum conservation balances the pressure pp, the osmotic pressure Π\mathbf{\Pi}, and the material stresses σA\boldsymbol{\sigma}_A and σB\boldsymbol{\sigma}_B as [49, 50]

0=∇⋅(−pI−Π+σA+σB),(2)0=\nabla\cdot\left(-p\mathbf{I}-\mathbf{\Pi}+\boldsymbol{\sigma}_A+\boldsymbol{\sigma}_B\right), \tag*{(2)}

where I\mathbf{I} is the identity tensor.

The motion of one component immediately implies the relative motion of the other component. This relative motion is obtained as [42, 45, 49]

uA−uB=−1n0ζ0μ[(1−ϕ)∇⋅(Π−σA)+ϕ∇⋅σB],(3)\mathbf{u}_A-\mathbf{u}_B=-\frac{1}{n_0\zeta_0\mu}\left[(1-\phi)\nabla\cdot(\mathbf{\Pi}-\boldsymbol{\sigma}_A)+\phi\nabla\cdot\boldsymbol{\sigma}_B\right], \tag*{(3)}

where n0n_0 is the mixture’s number density (monomers per volume), ζ0\zeta_0 is a monomer friction coefficient (assumed to be the same for both A-mers and B-mers), and μ≡μ(ϕ)\mu\equiv\mu(\phi) is a dimensionless friction function that depends on the polymers and their surroundings [9, 10, 42]. The functional form of μ\mu is expressed via the friction function of the A-mers and B-mers, μA\mu_A and μB\mu_B respectively, as

μ=μAμBμA+μB.(4)\mu=\frac{\mu_A\mu_B}{\mu_A+\mu_B}. \tag*{(4)}

The individual friction functions μA\mu_A and μB\mu_B depend crucially on the polymer’s lengths, NAN_A and NBN_B, compared to an entanglement threshold NeN_e. Polymers shorter than NeN_e are considered unentangled, implying μ(i)=ϕ(i)\mu_{(i)}=\phi_{(i)} (for species ii). Entangled polymers are subjected to increased friction, μ(i)=α(i)ϕ(i)\mu_{(i)}=\alpha_{(i)}\phi_{(i)}, where α(i)≡N(i)/Ne>1\alpha_{(i)}\equiv N_{(i)}/N_e>1 [9, 10].

The osmotic pressure Π\mathbf{\Pi} originates from the polymer concentration, i.e., chemical potential, gradients. We utilize the Flory-Huggins (FH) thermodynamic potential [51, 52]

FFH=kBTn0f(ϕ),F_{FH}=k_BTn_0f(\phi),

where kBk_B is the Boltzmann constant, TT is the temperature, and

f(ϕ)=ϕNAlog⁡ϕ+(1−ϕ)NBlog⁡(1−ϕ)+χϕ(1−ϕ).(5)f(\phi)=\frac{\phi}{N_A}\log\phi+\frac{(1-\phi)}{N_B}\log(1-\phi)+\chi\phi(1-\phi). \tag*{(5)}

Here, NAN_A and NBN_B are the number of monomers per A-mer and B-mer respectively, and χ\chi is an energy interaction parameter between the two species. As we are interested in the mixing of the two species, we omit surface tension. For simplicity, we consider non-interacting polymers, setting χ=0\chi=0. For the FH potential of (5), the resulting osmotic pressure is isotropic, obtained as

Π=kBTn0I(ϕ∂ϕf−f)\mathbf{\Pi}=k_BTn_0\mathbf{I}\left(\phi\partial_\phi f-f\right)

[1, 42]. Note that the osmotic pressure derivative ∇⋅Π\nabla\cdot\mathbf{\Pi} is equivalent to the spatial gradients of the chemical potential, as ∇[ϕ∂ϕf−f]=ϕ∇(∂ϕf)\nabla[\phi\partial_\phi f-f]=\phi\nabla(\partial_\phi f) (we use Π\mathbf{\Pi} to be consistent with the existing literature). We plot f(ϕ)f(\phi) for different values of NAN_A (in the non-interacting limit χ=0\chi=0, and for monomer surroundings, NB=1N_B=1) in Figure 1(c).

The stresses σA\boldsymbol{\sigma}_A and σB\boldsymbol{\sigma}_B evolve via an upper-convected Maxwell (UCM) model [49, 53],

σ▽(i)=G(i)(L(i)+L(i)T)−1λ(i)σ(i),(6)\overset{\triangledown}{\boldsymbol{\sigma}}_{(i)}=G_{(i)}\left(\mathbf{L}_{(i)}+\mathbf{L}_{(i)}^{T}\right)-\frac{1}{\lambda_{(i)}}\boldsymbol{\sigma}_{(i)}, \tag*{(6)}

where ii denotes either A or B (no summation implied), σ▽(i)≡∂tσ(i)+u(i)⋅∇σ(i)−[L(i)T⋅σ(i)+σ(i)⋅L(i)]\overset{\triangledown}{\boldsymbol{\sigma}}_{(i)}\equiv\partial_t\boldsymbol{\sigma}_{(i)}+\mathbf{u}_{(i)}\cdot\nabla\boldsymbol{\sigma}_{(i)}-\left[\mathbf{L}_{(i)}^{T}\cdot\boldsymbol{\sigma}_{(i)}+\boldsymbol{\sigma}_{(i)}\cdot\mathbf{L}_{(i)}\right] is the upper-convected time derivative, G(i)G_{(i)} is the shear modulus, λ(i)\lambda_{(i)} is the relaxation time, and L(i)≡∇u(i)L_{(i)} \equiv\nabla u_{(i)} is the velocity gradient tensor for the ith component. While below we focus on a regime where the material stresses σA\sigma_A and σB\sigma_B could be safely neglected, we retain those for completeness.

To ease the analysis of the above equations of motion, we rescale Eqs. (1)-(6) with λA\lambda_A as a time scale, kBTλAζ0\sqrt{\frac{k_B T \lambda_A}{\zeta_0}} as a length scale, and n0kBTn_0 k_B T as an energy density scale. Thus, we arrive at a unitless, 1-dimensional (1D) reduced form of the above equations, cast as (i=A,B)(i=A,B)

∂tϕ+∂x(uAϕ)=S(7)\partial_t \phi+ \partial_x(u_A\phi) = S \tag*{(7)}
−∂tϕ+∂x[uB(1−ϕ)]=0(8)-\partial_t\phi+ \partial_x\left[u_B(1-\phi)\right] = 0 \tag*{(8)}
∂x(−p−Π+σA+σB)=0(9)\partial_x\left(-p-\Pi+\sigma_A+\sigma_B\right)=0 \tag*{(9)}
uA−uB=−1μ[(1−ϕ)∂x(Π−σA)+ϕ∂xσB](10)u_A-u_B=-\frac{1}{\mu}\left[(1-\phi)\partial_x(\Pi-\sigma_A)+\phi\partial_x\sigma_B\right] \tag*{(10)}
∂tσ(i)+u(i)∂xσ(i)=2(σ(i)+G)∂xu(i)−1λ(i)σ(i)(11)\partial_t\sigma_{(i)}+u_{(i)}\partial_x\sigma_{(i)}=2\left(\sigma_{(i)}+G\right)\partial_xu_{(i)}-\frac{1}{\lambda_{(i)}}\sigma_{(i)} \tag*{(11)}

Below we study the passive diffusive dynamics, and source-driven diffusive dynamics, of asymmetric polymer mixtures, where we assume NA≫NBN_A \gg N_B. We also assume that NBN_B is less than the entanglement threshold NeN_e, implying the B-mers are not entangled throughout the dynamics. To allow the A-mers to transition from a possibly entangled state (if they are sufficiently long) to an unentangled state, we use μA=ϕ[αϕ+(1−ϕ)]\mu_A=\phi[\alpha\phi+(1-\phi)] as the A-mers friction function. This function reduces to μA=ϕ\mu_A=\phi in the unentangled limit (when α=1\alpha=1). If the polymers are sufficiently long to entangle, it allows for a transition μA→ϕ\mu_A\to\phi at low volume fractions, and μA→α\mu_A\to\alpha as ϕ→1\phi\to1. We plot three examples of μ(ϕ)\mu(\phi) for different α\alpha values in fig:1d. As α\alpha increases, μ\mu develops an asymmetry, increasing the friction at intermediate ϕ\phi regions and approaches 1−ϕ1-\phi — an effect that is due to entanglements.

As we would like to focus on the role of entanglement and the passive or source-driven cases, we will neglect the roles of the material stresses σA\sigma_A and σB\sigma_B. To estimate their relative importance, we note that polymeric stresses σ(i)\sigma_{(i)} are usually characterized by the elastic moduli G(i)G_{(i)}, and relaxation times λ(i)\lambda_{(i)}, as explained above [45, 49, 53] [see also (6)]. While precise measurements are challenging, biological condensates yield maximal storage moduli of ∼1\sim1–1010 Pa at the low-frequency limit, and maximal viscosity of ∼10\sim10–100100 Pa ⋅\cdot s [32, 33, 34, 35]. Comparing these estimates with the energy density scale n0kBT≃KPan_0k_BT\simeq\mathrm{KPa} at T=300T=300 K (using n0=1n_0=1 mM as a low estimate for the monomer density), clearly indicates that under typical conditions the osmotic pressure is the dominant source of stress.

The relaxation times of biopolymers may also be sufficiently fast for our interests. Biological polymers exhibit relaxation times of λ≃10\lambda\simeq10 ms [32, 35]. If we compare this to the expected (linear) diffusion time τD≃R2ζ0/kBT\tau_D\simeq R^2\zeta_0/k_BT where RR is a typical droplet radius R≃5R\simeq5 μ\mum, and ζ0=6πη\zeta_0=6\pi\eta with η=1\eta=1 mPa ⋅\cdot s and a≃1a\simeq1 nm being the monomer (protein) typical size, we get τD≃100\tau_D\simeq100 ms, again confirming that polymer relaxation occurs at shorter times than those of interest here. Interestingly, a reminiscent time-scale separation was observed in [35] for trapped droplets composed of different biopolymers and probed with optical tweezers. These separations of stress scales and time scales allow us to confine our analysis to cases in which these two assumptions hold, and the material stresses σA\sigma_A and σB\sigma_B can be safely neglected. While velocity gradients and polymer conformations near interfaces could induce mechanical stresses, considering the osmotic pressure captures the dominant driving force needed to describe the emerging nonlinear transport. As we show below, the simplifying assumptions above allow us to study polymer-polymer interdiffusion and its dependence on two central aspects — the degree of entanglement and the absence or presence of a source.

Droplet diffusion

We first consider a droplet of A-mers placed in a B-mers medium, depicted schematically in fig:1a. As the species are assumed to be non-interacting (χ=0\chi=0), entropic forces cause the two species to diffuse into one another. This process is characterized by an effective nonlinear diffusion equation.

To see this description, we utilize our 1D formalism, and add Eq. (7)-(8). We define v=uAϕ+uB(1−ϕ)v=u_A\phi+u_B(1-\phi), which can be interpreted as the mean velocity, and obtain ∂xv=0\partial_xv=0 in the absence of a source (S=0S=0), implying v=0v=0 due to isotropy. We define the relative speed w=uA−uBw=u_A-u_B, and using uA=v+(1−ϕ)wu_A=v+(1-\phi)w, we obtain that in the absence of a source uA=(1−ϕ)wu_A=(1-\phi)w. We then use Eq. (10) with Eq. (7) (neglecting the material stresses σA\sigma_A and σB\sigma_B), to write

∂tϕ=∂x[ϕ(1−ϕ)2μ∂xΠ](12)\partial_t\phi=\partial_x\left[\frac{\phi(1-\phi)^2}{\mu}\partial_x\Pi\right] \tag*{(12)}

which corresponds to a nonlinear diffusion equation, with

Π=−[log⁡(1−ϕ)+(1−NA−1)ϕ](13)\Pi=-\left[\log(1-\phi)+(1-N_A^{-1})\phi\right] \tag*{(13)}

where we set NB=1N_B=1 and χ=0\chi=0, and

μ=ϕ(1−ϕ)1−ϕ+αϕ1−ϕ+ϕ(1−ϕ+αϕ)(14)\mu=\phi(1-\phi)\frac{1-\phi+\alpha\phi}{1-\phi+\phi(1-\phi+\alpha\phi)} \tag*{(14)}

[see (4)].

In the unentangled limit, we set α=1\alpha=1, reducing μA=ϕ\mu_A=\phi, and μB=1−ϕ\mu_B=1-\phi, and from (14) μ=ϕ(1−ϕ)\mu=\phi(1-\phi) [see also fig:1d for a plot of μ(ϕ)\mu(\phi) for different α\alpha]. Using the 1D osmotic pressure of (13), and NA≫1N_A\gg1, (12) becomes

∂tϕ≃∂x[(1+NAϕ)NA∂xϕ]⟶ϕ≫NA−1∂x(ϕ∂xϕ)(15)\partial_t\phi\simeq\partial_x\left[\frac{(1+N_A\phi)}{N_A}\partial_x\phi\right]\underset{\phi\gg N_A^{-1}}{\longrightarrow}\partial_x\left(\phi\partial_x\phi\right) \tag*{(15)}

For sufficiently low volume fractions ϕNA≪1\phi N_A\ll1, the above equation reduces to a linear diffusion equation. However, supposing that the initial volume fraction within the droplet is such that ϕ≫NA−1\phi\gg N_A^{-1}, Eq. (15) approaches a nonlinear diffusion equation, as indicated above.

Using ℓ\ell as a typical length scale, Eq. (15) suggests ℓ2∼ϕt\ell^2\sim\phi t. We combine this scaling relation with fact that the total volume fraction I≡∫−ℓℓϕ dx\mathcal{I}\equiv\int_{-\ell}^{\ell}\phi\,dx must be conserved throughout the dynamics, implying ℓϕ∼I\ell\phi\sim\mathcal{I}. Put together, these scaling relations suggest ℓ∼(It)1/3\ell\sim(\mathcal{I}t)^{1/3}, and ϕ∼(I2/t)1/3\phi\sim(\mathcal{I}^2/t)^{1/3}.

We now use the self-similarity ansatz, ϕ(x,t)≡(I2/t)1/3Φu(ξ)\phi(x,t)\equiv(\mathcal{I}^2/t)^{1/3}\Phi_u(\xi) (where Φu\Phi_u denotes the unentangled Φ\Phi), with the self-similar coordinate ξ≡x/(It)1/3\xi\equiv x/(\mathcal{I}t)^{1/3}. Substituting into the approximate nonlinear diffusion equation, Eq. (15), yields an ordinary differential equation (ODE),

(3ΦuΦu′+ξΦu)′=0.(16)(3\Phi_u\Phi_u' + \xi\Phi_u)'=0 . \tag*{(16)}

Integrating once, assuming a symmetric droplet profile Φu′(0)=0\Phi_u'(0)=0, we arrive at Φu=−ξ26+c\Phi_u=-\frac{\xi^2}{6}+c, where cc is an integration constant. The obtained solution is a parabola that intersects the ξ\xi axis Φu(ξ∗)=0\Phi_u(\xi_*)=0, at ξ∗=6c\xi_*=\sqrt{6c}. Integrating over the domain and demanding the volume fraction conservation ∫0ξ∗Φu=12\int_0^{\xi_*}\Phi_u=\frac{1}{2} gives c=(332)1/3c=\left(\frac{3}{32}\right)^{1/3}.

We simulate the 1D system of Eqs. (7)-(11), with a finite value of NA=104N_A=10^4 and setting α=1\alpha=1 (using NB=1N_B=1, χ=0\chi=0, and in the absence of σA\sigma_A and σB\sigma_B). We initialize the distribution of ϕ\phi as a Gaussian of typical length of

2, and normalize it such that at its peak, ϕ=ϕmax⁡\phi=\phi_{\max}, i.e., ϕ(x,t=0)=ϕmax⁡exp⁡(−x2/8)\phi(x,t=0)=\phi_{\max}\exp(-x^2/8), and follow its diffusion. We then solve Eqs. (7)-(11), first by calculating the velocity profiles and ww, followed by integrating ϕ\phi in time [54, 55].

We show in Figure 2(a) the ϕ\phi profiles obtained at different times during the diffusion dynamics. Then, we apply the self-similar transformation, and re-plot the solution in terms of the self-similar function Φu\Phi_u, and variable ξ\xi in Figure 2(b). We also plot the theoretical solution Φu\Phi_u on top. As time advances, the profiles approach the predicted solution, demonstrating the validity of the approximate nonlinear diffusion equation (15) for finite NAN_A values, and the emerging scaling relation above. In Figure 2(c) we plot trajectories of ϕ\phi profiles obtained for three different I\mathcal{I} values (modified by changing ϕmax⁡\phi_{\max}). Rescaling those profiles as shown in Figure 2(d), reveals the scalings and analytical solution applies to those different parameter values and at different times.

Figure 2. Numerical and analytical solutions for unentangled polymer interdiffusion with α=1\alpha=1. (a) The spreading A-mer droplet dynamics of ϕ\phi, as obtained from our 1D numerical simulations of Eqs. (7)-(11) (ϕmax⁡=0.3\phi_{\max}=0.3 corresponding to I≃1.5\mathcal{I}\simeq1.5). Early times are depicted as gray, and later times are drawn in red. (b) The same ϕ\phi profiles drawn using the self-similar Φu=ϕ/(I2/t)1/3\Phi_u=\phi/(\mathcal{I}^2/t)^{1/3}, and ξ=x/(It)1/3\xi=x/(\mathcal{I}t)^{1/3}. Dashed black line is the analytical solution Φu=−ξ26+(332)1/3\Phi_u=-\frac{\xi^2}{6}+\left(\frac{3}{32}\right)^{1/3}. (c) Snapshots of ϕ\phi from simulations with different initial ϕmax⁡\phi_{\max} values of 0.3, 0.5, and 0.8, corresponding to different I\mathcal{I} values (see legend). (d) The same snapshots cast as self-similar solutions, with black dashed line the same as in (b).

We repeat a similar analysis for the case of entangled A-mers. For α≫1\alpha\gg1, we obtain μ∼1−ϕ\mu\sim1-\phi [see Figure 1(d)], and Eq. (12) reduces to

∂tϕ≈∂x[(1+NAϕ)NAϕ∂xϕ]ϕ≫NA−1⟶∂x[ϕ2∂xϕ].(17)\partial_t\phi\approx\partial_x\left[\frac{(1+N_A\phi)}{N_A}\phi\partial_x\phi\right]_{\phi\gg N_A^{-1}}\longrightarrow\partial_x\left[\phi^2\partial_x\phi\right]. \tag*{(17)}

In this case, the scaling relations are ℓ2∼ϕ2t\ell^2 \sim\phi^2 t and ϕℓ=I\phi\ell= \mathcal{I}, yielding ℓ∼(I2t)1/4\ell\sim(\mathcal{I}^2 t)^{1/4}, and ϕ∼(I2/t)1/4\phi\sim(\mathcal{I}^2/t)^{1/4}. Using the ansatz ϕ(x,t)=(I2/t)1/4Φe(ξ)\phi(x,t)=(\mathcal{I}^2/t)^{1/4}\Phi_{\mathrm{e}}(\xi) (here with a subscript Φe\Phi_{\mathrm{e}} to denote the entangled case) with ξ≡x/(I2t)1/4\xi\equiv x/(\mathcal{I}^2t)^{1/4}, we reduce (17) to

(4Φe2Φe′+ξΦe)′=0.(18)\left(4\Phi_{\mathrm{e}}^2\Phi_{\mathrm{e}}' + \xi\Phi_{\mathrm{e}}\right)'=0 . \tag*{(18)}

Using a similar procedure as described above we obtain Φe=124π−ξ24\Phi_{\mathrm{e}}=\frac{1}{2}\sqrt{\frac{4}{\pi}-\frac{\xi^2}{4}}. We show in Figure 3 numerical solutions for a 1D system in the entangled case in similar fashion to Figure 2 (with NA=104N_{\mathrm{A}}=10^4 and α=104\alpha=10^4). Specifically, Figure 3 (a) shows an exemplary trajectory at different times, and the inset shows a cusp occurring once ϕ∼α−1/2\phi\sim\alpha^{-1/2} — a “tail” where the simplification to ϕ2∂xϕ\phi^2\partial_x\phi is no longer valid, and a different scaling law emerges. This tail signifies a transition between an entangled state to the unentangled state, characterized by a different diffusion scaling. Then, Figure 3 (b) shows the collapse of the same trajectory using the suggested scalings, with the theoretical solution on top.

Figure 3. Numerical and analytical solutions for entangled polymer interdiffusion with α=104\alpha=10^4. (a) The spreading A-mer droplet dynamics of ϕ\phi, as obtained from our 1D numerical simulations of Eqs. (7)-(11) (using ϕmax⁡=0.3\phi_{\max}=0.3 corresponding to I≃1.5\mathcal{I}\simeq1.5). Early times are depicted as gray, and later times are drawn in blue. Horizontal dashed line indicates ϕ=α−1/2\phi=\alpha^{-1/2}. Inset: a close-up on the transition region ϕ≃α−1/2\phi\simeq\alpha^{-1/2} showing the formed tails. (b) The same ϕ\phi profiles drawn using the self-similar Φe=ϕ/(I2t)1/4\Phi_e=\phi/(\mathcal{I}^2t)^{1/4}, and ξ=x/(I2t)1/4\xi=x/(\mathcal{I}^2t)^{1/4}. Dashed black line is the analytical solution Φe=124π−ξ2\Phi_e=\frac{1}{2}\sqrt{\frac{4}{\pi}-\xi^2}. (c) Snapshots of ϕ\phi from simulations with different initial ϕmax⁡\phi_{\max} values of 0.3, 0.5 and 0.8, corresponding to different I\mathcal{I} values (see legend). (d) The same snapshots cast as self-similar solutions, with black dashed line the same as in (b).

The agreement between the full numerical solution and the self-similar solution is partial — the droplet obtained from the numerical solutions of Eqs. (7)-(11) spreads further, and the ODE solution predicts higher central concentrations and lower concentrations at the rims. We suspect this partial agreement is due to the tails referred to above. As these tails move material away from the central droplet, we anticipate that as their relative contribution decreases, the solutions should approach the self-similar solution of Eq. (18). In Figure 3 (c) we show solutions with different I\mathcal{I} (obtained by changing the ϕmax⁡\phi_{\max} values), and in Figure 3 (d) we show their rescaled counterparts. As the initial maximal concentration ϕmax⁡\phi_{\max} is increased, I\mathcal{I} increases, and the relative contribution of the tail to the dynamics decreases. The rescaled solutions indeed confirm that as the tail contributions become negligible (higher ϕmax⁡\phi_{\max} and I\mathcal{I} values) the solutions approach the expected self-similar form.

In two dimensions (2D) we can repeat the procedure above, to approximate the problem as a nonlinear diffusion equation, which could be cast in a self-similar framework (see App. A). Using rr as the position vector, and assuming isotropy, we use the radial distance r≡∣r∣r\equiv|\mathbf{r}| to get ϕ(r,t)≡(I/t)1/2Φu(ξ)\phi(r,t)\equiv(\mathcal{I}/t)^{1/2}\Phi_{\mathrm{u}}(\xi), and ξ≡r/(It)1/4\xi\equiv r/(\mathcal{I}t)^{1/4}, and ϕ(r,t)≡(I/t)1/3Φe(ξ)\phi(r,t)\equiv(\mathcal{I}/t)^{1/3}\Phi_{\mathrm{e}}(\xi), and ξ≡r/(I2t)1/6\xi\equiv r/(\mathcal{I}^2t)^{1/6} for the unentangled and entangled cases, respectively. Using the transformations we obtain

ξ2Φu+4ξΦuΦu′=0for α=1,(19)\xi^2\Phi_{\mathrm{u}}+4\xi\Phi_{\mathrm{u}}\Phi_{\mathrm{u}}'=0 \qquad\text{for }\alpha=1 , \tag*{(19)}
ξ2Φe+6ξΦe2Φe′=0for α≫1.(20)\xi^2\Phi_{\mathrm{e}}+6\xi\Phi_{\mathrm{e}}^2\Phi_{\mathrm{e}}'=0 \qquad\text{for }\alpha\gg1 . \tag*{(20)}

Solving those in a similar fashion to the 1D case yields Φu=12(1π−ξ24)\Phi_{\mathrm{u}}=\frac{1}{2}\left(\frac{1}{\sqrt{\pi}}-\frac{\xi^2}{4}\right), and Φe=12(2π)2/3−2ξ23\Phi_{\mathrm{e}}=\frac{1}{2}\sqrt{\left(\frac{2}{\pi}\right)^{2/3}-\frac{2\xi^2}{3}}. We show our numerical solutions to Eqs. (1)-(3) (neglecting σA\sigma_{\mathrm{A}} and σB\sigma_{\mathrm{B}}), and the comparison to the expected self-similar solutions in Figure 4.

Figure 4. Numerical and analytical solutions for unentangled polymer interdiffusion in 2D. (a) Snapshots of the radially-averaged ϕ\phi from 2D simulations of Eqs. (1)-(3), with α=1\alpha=1, and with different initial ϕmax⁡\phi_{\max} values of 0.3, 0.5 and 0.8, corresponding to different I\mathcal{I} values (see legend). (b) The same snapshots as in (a) cast as self-similar solutions, with the self-similar solution plotted as a thick dashed black line. (c) Same as panel (a) but for the entangled case with α=104\alpha=10^4. (d) The same snapshots as in (c) cast as self-similar solutions, with the self-similar solution in black dashed line.

Analysing the passive diffusion problems, we note the similarity in the attained spatial form between the 1D and 2D cases, i.e., between Figs. 2, 3, and 4. Specifically, the unentangled cases vary with ξ2\xi^2, and the entangled cases vary as ξe2−ξ2\sqrt{\xi_{\mathrm{e}}^2-\xi^2}. Those similar spatial profiles hint that dimensionality plays a rather minor role in terms of the observed spatial profiles.

Generally, the suggested simplifications reduce both the 1D and 2D cases to a nonlinear diffusion problem with ∂tϕ=∇⋅(ϕm−1∇ϕ)\partial_t\phi=\nabla\cdot(\phi^{m-1}\nabla\phi), where mm quantifies the degree of nonlinearity that directly depends on α\alpha — the unentangled case implies α=1\alpha=1 leading to m=2m=2, while the entangled case has α≫1\alpha\gg1 leading to m=3m=3. Interestingly, similar equations appear in the context of diffusion in porous media [56, 57]. The diffusive self-similar length-scale and amplitudes depend on the nonlinearity mm, and the dimensionality dd of the problem. A passive droplet diffusion problem in dd dimensions, with a nonlinearity of power mm, suggests the scaling ϕ∼Iγ1tδ1\phi\sim\mathcal{I}^{\gamma_1}t^{\delta_1} and ℓ∼Iγ2tδ2\ell\sim\mathcal{I}^{\gamma_2}t^{\delta_2}, with

γ1=2/p,δ1=−d/p,\gamma_1=2/p,\qquad\delta_1=-d/p,
γ2=(m−1)/p,δ2=1/p,(21)\gamma_2=(m-1)/p,\qquad\delta_2=1/p, \tag*{(21)}

with p=2+d(m−1)p=2+d(m-1). Transitioning from the unentangled to entangled scenario slows the dynamics down — mm increases causing pp to increase as well. This parallels with polymer reptation in entangled scenarios, absent from the unentangled case [9, 10, 16]. So, while reptation is a single polymer mode-of-motion, it affects the emerging diffusion on a coarse-grained scale. In both cases, approximating the equations of motion in the limit of sufficiently large NAN_A and ϕ\phi, and using the corresponding α\alpha values (while neglecting the material stresses σA\sigma_A and σB\sigma_B) allowed us to reformulate the interdiffusion problem as a self-similar droplet diffusion problem. The emerging equations were simple enough to solve analytically, and resulted in good agreement with both our 1D and 2D numerical solutions.

Source-driven diffusion

The passive droplet diffusion scenario investigated above may become modified in certain biological cases. Some biological condensates serve as effective sources, e.g., in the synthesis of RNAs and proteins, injecting polymers at a certain rate into the system [36, 37, 38, 39, 40]. We now modify the formalism above to account for these scenarios, treating the synthesis centers as external polymer sources, depicted schematically in fig:1b.

We employ a similar numerical solution method as for the passive case discussed in Sec. III to obtain numerical solutions for the source-driven diffusion. Here, we set ϕ(x,t=0)=0\phi(x,t=0)=0, and use a localized Gaussian as our source term, S(x,t)=f(t)Zexp⁡(−x2/2)S(x,t)=\frac{f(t)}{Z}\exp(-x^2/2), with the normalization factor Z≡∑exp⁡(−x2/2)ΔxZ\equiv\sum\exp(-x^2/2)\Delta x over our integration domain (here Δx\Delta x is the resolution of the discretized domain). We show two exemplary trajectories for the unentangled α=1\alpha=1 case in fig:5a, and α≫1\alpha\gg1 in fig:5c, for a constant-rate, localized source f(t)=Qf(t)=Q, as obtained from our numerical simulations of Eqs. (7)-(11). The A-mers accumulate near the origin, increasing ϕ\phi locally, hence increasing the osmotic pressure/chemical gradients, driving outward diffusion. The volume fraction profile evolves from its early time behavior, until it converges to the long time profile. While there are intricate spatial features at early times, and near the origin, e.g., the presence of a “tip”, in both cases the volume fraction distributions near the edges, or droplet’s “front shape”, attained at long times are reminiscent of their spatial features in the passive droplet expansion (see Figure 2-Figure 3).

To characterize the volume fraction redistribution dynamics, we measure two typical droplet radii, denoted by ℓin\ell_{\mathrm{in}} and ℓout\ell_{\mathrm{out}}, defined here, respectively, as the half-width of the ϕ>0.1\phi>0.1 domain for ℓin\ell_{\mathrm{in}}, and ϕ>0.01\phi>0.01 domain for ℓout\ell_{\mathrm{out}} [as annotated in fig:5a and fig:5c]. We plot their time-dependencies for various source rates QQ, as shown in fig:5b and fig:5d for the unentangled and entangled cases, respectively. At very early times ℓin\ell_{\mathrm{in}} approximately increases according to a linear diffusion equation response ∼t1/2\sim t^{1/2}, while ℓout\ell_{\mathrm{out}} diffuses according to the passive droplet cases analyzed in Sec. III, i.e., following (21). At longer times, the two radii converge into similar nonlinear scalings. These early-time, spatially-dependent scalings indicate the volume fraction redistributes heterogeneously throughout the droplet.

To try and understand these spatial and temporal behaviors, we revisit the continuity equations (7)-(8). Upon addition, and using v=ϕuA+(1−ϕ)uBv=\phi u_A+(1-\phi)u_B, we obtain ∂xv=S\partial_xv=S. We then use this together with v=uA−uBv=u_A-u_B, to express uAu_A, and substitute back into (7), leading to the evolution equation

∂tϕ=∂x[ϕ(1−ϕ)2μ∂xΠ−vϕ]+S(22)\partial_t\phi=\partial_x\left[\frac{\phi(1-\phi)^2}{\mu}\partial_x\Pi-v\phi\right]+S \tag*{(22)}

Compared to the passive droplet equivalent, (12), the additional source term not only contributes by adding more A-mers into the system, but also modifies the average velocity vv and contributes a flux term. We expect a similar effect happens in higher dimensions as well, as a consequence of a material source near the origin (see App. B).

We now suppose that the source is spatially-localized, and takes the general form S(x,t)=δ(x)f(t)S(x,t)=\delta(x)f(t). Combining the unentangled and entangled cases, in the limit of ϕ≫NA−1\phi\gg N_{\mathrm{A}}^{-1}, (22) reduces to

∂tϕ+v∂xϕ≃∂x(ϕm−1∂xϕ)+δ(x)f(t)(1−ϕ)(23)\partial_t\phi+v\partial_x\phi\simeq\partial_x\left(\phi^{m-1}\partial_x\phi\right)+\delta(x)f(t)(1-\phi) \tag*{(23)}

where we have used the fact that ∂xv≡δ(x)f(t)\partial_xv\equiv\delta(x)f(t). Irrespective of the nonlinearity exponent mm, the presence of the advection term v∂xϕv\partial_x\phi suggests transforming the above into a reduced ODE [e.g., Eqs. (16), (18), (19) and (20)] may involve translation in addition to rescaling. Additionally, combining the source together with the incompressibility ϕA+ϕB=1\phi_{\mathrm{A}}+\phi_{\mathrm{B}}=1 renders the source’s amplitude to depend on the local concentration ϕ\phi, as δ(x)f(t)(1−ϕ)\delta(x)f(t)(1-\phi).

With these features in mind, we consider again the dominant balance of the osmotic pressure with the concentration time-variation ℓ2∼ϕm−1t\ell^2\sim\phi^{m-1}t, together with the increase in mass ϕℓ∼∫0tf(t′)dt′\phi\ell\sim\int_0^t f(t')\mathrm{d}t'. To make progress, we assume ff has a temporal power-law variation f=Qtq−1f=Qt^{q-1}, allowing us to obtain the scaling relations for any power-law increase qq, and to generalize for more complicated time-variations (potentially approximating those as piecewise power-law functions). Taking Qtq−1Qt^{q-1} for our source, we obtain ϕℓ∼Qtq\phi\ell\sim Qt^q, from which we recover ϕ(x,t)≡Qγ~1tδ~1Φ(ξ)\phi(x,t)\equiv Q^{\widetilde{\gamma}_1}t^{\widetilde{\delta}_1}\Phi(\xi) and ξ≡x/Qγ~2tδ~2\xi\equiv x/Q^{\widetilde{\gamma}_2}t^{\widetilde{\delta}_2} with p≡2+d(m−1)p\equiv2+d(m-1) as defined above, and

γ~1=2p,δ~1=2q−dp,\widetilde{\gamma}_1=\frac{2}{p},\qquad\widetilde{\delta}_1=\frac{2q-d}{p},
γ~2=m−1p,δ~2=1+q(m−1)p.(24)\widetilde{\gamma}_2=\frac{m-1}{p},\qquad\widetilde{\delta}_2=\frac{1+q(m-1)}{p}. \tag*{(24)}

Comparing (24) with the scaling relations in the passive droplet diffusion case, (21), reveals that only the time dependence, via the δ~\widetilde{\delta} exponents, are modified by the presence of a source, in which case QQ plays a role reminiscent of II in the passive case. These scaling exponents are the ones shown in the late-time regimes in Figure 5(b) and (d), signifying the droplet’s expansion before saturation.

Figure 5. Numerical solutions for unentangled and entangled source-driven polymer interdiffusion with f(t)=Qf(t)=Q. (a) Unentangled α=1\alpha=1 source-driven A-mer dynamics ϕ\phi, as obtained from our full 1D numerical simulations of Eqs. (7)-(11) (with α=1\alpha=1, NA=104N_A=10^4, and Q=0.1Q=0.1). Early times are depicted as gray, and later times are drawn in red. We denote the two length scales ℓin\ell_{\mathrm{in}} and ℓout\ell_{\mathrm{out}}, as the half-width of the droplet’s domain that is ϕ>0.1\phi>0.1 for ℓin\ell_{\mathrm{in}} and ϕ>0.01\phi>0.01 for ℓout\ell_{\mathrm{out}}. (b) Time-dependence of ℓin\ell_{\mathrm{in}} and ℓout\ell_{\mathrm{out}} for different source amplitudes QQ. At short times, ℓin∼t1/2\ell_{\mathrm{in}}\sim t^{1/2} similar to a linear diffusion scaling, while ℓout∼t1/3\ell_{\mathrm{out}}\sim t^{1/3} as in the passive droplet scenario. At long times, both length scales converge to a nonlinear scaling ∼t2/3\sim t^{2/3}. (c) Entangled source-driven A-mer dynamics ϕ\phi, as obtained from our full 1D numerical simulations of Eqs. (7)-(11) (with α=104\alpha=10^4, NA=104N_A=10^4, and Q=0.1Q=0.1). Early times are depicted as gray, and later times are drawn in blue. We denote the two length scales ℓin\ell_{\mathrm{in}} and ℓout\ell_{\mathrm{out}} [following similar definitions as in panel (a)]. (d) Time-dependence of ℓin\ell_{\mathrm{in}} and ℓout\ell_{\mathrm{out}} for different source amplitudes QQ. At short times, ℓin∼t1/2\ell_{\mathrm{in}}\sim t^{1/2} similar to a linear diffusion scaling, while ℓout∼t1/4\ell_{\mathrm{out}}\sim t^{1/4} as in the passive droplet scenario. At long times, both length scales converge to a nonlinear scaling ∼t3/4\sim t^{3/4}.

Combining the scaling and taking into account the advection term, we propose to use the 1D transformation ϕ=Qγ~1tδ~1Φ(η)\phi=Q^{\widetilde{\gamma}_1}t^{\widetilde{\delta}_1}\Phi(\eta) and η=[x−12qQtq]/Qγ~2tδ~2≡ξ+ξs\eta=[x-\frac{1}{2q}Qt^q]/Q^{\widetilde{\gamma}_2}t^{\widetilde{\delta}_2}\equiv\xi+\xi_s, where we defined ξs≡−Q1−γ~2tq−δ~22q\xi_s\equiv-\frac{Q^{1-\widetilde{\gamma}_2}t^{q-\widetilde{\delta}_2}}{2q} as the rescaled source coordinates (see App. C). Using these in (23) (setting aside the source term, which we treat as a boundary condition to our equations), we obtain

[3ΦuΦu′+(1−2q)ηΦu′]+3qηΦu′=0for α=1(25)\left[3\Phi_{\mathrm{u}}\Phi_{\mathrm{u}}^{\prime}+(1-2q)\eta\Phi_{\mathrm{u}}^{\prime}\right]+3q\eta\Phi_{\mathrm{u}}^{\prime}=0\quad\text{for }\alpha=1 \tag*{(25)}
[4Φe2+(1−2q)ηΦe′]+4qηΦe′=0for α≫1(26)\left[4\Phi_{\mathrm{e}}^2+(1-2q)\eta\Phi_{\mathrm{e}}^{\prime}\right]+4q\eta\Phi_{\mathrm{e}}^{\prime}=0\quad\text{for }\alpha\gg1 \tag*{(26)}

Clearly, in both cases the emerging equations are cast as ordinary differential equations, however in the self-similar coordinates η\eta the source is moving back to ξs\xi_s, and the boundary conditions themselves are time-dependent. So, while the equations are ordinary differential equations, the domain of integration is time dependent. This explicit time dependence breaks the self-similar structure.

Instead of solving the ODEs of (25)-(26) with a time-dependent domain and boundary conditions, we perform an asymptotic expansion near the front η∗\eta_*, where Φ(η∗)=0\Phi(\eta_*)=0. We anticipate that while the solutions’ features near the source may vary, the droplets’ fronts will attain a similar diffusion-dominated, spatial dependence as their passive counterparts, as already demonstrated in Figure 5(a) and (c). Substituting the expansion Φ~≃a1y1/(m−1)+a2y2/(m−1)\tilde{\Phi} \simeq a_1 y^{1/(m-1)} + a_2 y^{2/(m-1)} (where y≡η∗−ηy \equiv\eta_* - \eta) in Eqs. (25)-(26), using m=2m=2 and m=3m=3 for Φu\Phi_u and Φe\Phi_e respectively, collecting the leading orders, and demanding that those vanish, lead to a1=23η∗a_1 = \frac{2}{3}\eta_*, and a1=32η∗a_1 = \sqrt{\frac{3}{2}}\eta_* for m=2m=2 and $m=3 respectively.

We rescale the numerical solutions obtained for the unentangled and entangled cases, at different times and for different amplitudes QQ, according to (24), as shown in Figure 6(a) and (b). For high QQ values, our numerical integration did not advance far enough to transition into the long-time scaling regime, but the other solutions converge onto similar curves. The obtained curves vary especially near the origin, but as the front is approached, the different solutions converge onto a similar profile. In the insets in Figure 6(a) and (b), we show the rescaled solutions in η\eta coordinates, shifted to the front η∗\eta_*, and plot the leading order expected behavior discussed above. Clearly, the well-developed fronts all agree with our predicted leading-order behavior.

Figure 6. Numerical solutions and analytical approximation for droplet fronts of the unentangled and entangled source-driven polymer interdiffusion under different time-dependent sources. (a) The rescaled solutions Φu=ϕ/(Q2t)1/3\Phi_{\mathrm{u}}=\phi/(Q^2t)^{1/3} in the rescaled coordinates ξ=x/(Qt2)1/3\xi=x/(Qt^2)^{1/3} for the unentangled case with time-independent source q=1q=1, and for different amplitudes QQ. Each different color shows profiles obtained for different QQ value, where for each QQ value, we show two different time instances in darker and lighter curves (the dark curves represent earlier times, and light curves later times). Inset: the same solutions shifted to their front coordinate η−η∗\eta-\eta_*, with the theoretical prediction of Φu∼23ηη\Phi_{\mathrm{u}}\sim\frac{2}{3}\eta\eta (with η∗≃1\eta_*\simeq1 obtained from our numerical simulations). (b) The rescaled solution Φe=ϕ/(Q2t)1/4\Phi_{\mathrm{e}}=\phi/(Q^2t)^{1/4} in the coordinates ξ=x/(Q2t3)1/4\xi=x/(Q^2t^3)^{1/4} for the entangled source-driven spreading, with q=1q=1 for different source amplitudes QQ. For each QQ value, we show two different time instances in darker and lighter curves (the dark curves represent earlier times, and light curves later times). Inset: the same solutions shifted to their front coordinate η−η∗\eta-\eta_*, with the theoretical prediction of Φe∼23ηη\Phi_{\mathrm{e}}\sim\sqrt{\frac{2}{3}\eta\eta} (with η∗≃0.5\eta_*\simeq0.5 obtained from our numerical simulations). (c) The rescaled solutions Φu=ϕ/(Q2/3t)\Phi_{\mathrm{u}}=\phi/(Q^2/3t) in ξ=x/(Q1/3t)\xi=x/(Q^{1/3}t) coordinates for the unentangled case, under a time-dependent q=2q=2 source-driven A-mer dynamics, as obtained from our 1D numerical simulations. For each QQ value, we show two different time instances in darker and lighter curves (the dark curves represent earlier times, and light curves later times). (d) The rescaled solution Φe=ϕ/(Q2t3)1/4\Phi_{\mathrm{e}}=\phi/(Q^2t^3)^{1/4} in the coordinates ξ=x/(Q2t5)1/4\xi=x/(Q^2t^5)^{1/4} for entangled α=104\alpha=10^4 A-mers for q=2q=2. For each QQ value, we show two different time instances in darker and lighter curves (the dark curves represent earlier times, and light curves later times). Inset: Φe2\Phi_{\mathrm{e}}^2 versus the front coordinates η−η∗\eta-\eta_* showing Φe∼η∗−η\Phi_{\mathrm{e}}\sim\sqrt{\eta_*-\eta}.

The constant source with q=1q=1 is a special case in which the source ξs\xi_s moves with the same time-dependence as the rescaled coordinates ξ\xi. We now show that similar spatial profiles arise for q=2q=2, i.e., with a time-dependent source. We plot the rescaled numerical solutions at various times and for different QQ values in Figure 6(c) and (d), showing an explicit linear front geometry for the unentangled case in panel (c). In panel (d) we include an inset plotting Φe2\Phi_e^2 versus η−η∗\eta-\eta_*, showing that Φe∼η∗−η\Phi_e \sim\sqrt{\eta_*-\eta}, similar to the q=1q=1 case above.

Finally, we perform 2D numerical simulations with q=1q=1, and compare our results in a similar fashion in Figure 7. Panels (a) and (b) show exemplary radially-averaged ϕ\phi dynamics. In 2D, taking q=1q=1 reveals an interesting case in which the source amplitude is time-independent [see δ~1\tilde{\delta}_1 in Eq. (24)], and δ~2=1/2\tilde{\delta}_2=1/2 independently of mm, implying the unentangled and entangled time-dependencies are similar. We show this scaling in the insets of panels (a) and (b) respectively (here probing only a single length-scale ℓin\ell_{\mathrm{in}}, as the half-width of the ϕ>0.01\phi>0.01 domain). Then, panels (c) and (d) (and the inset therein) show the rescaled solutions, revealing similar spatial features at the droplets’ front, as the 1D counterparts (see Figure 6).

Figure 7. Numerical solutions for unentangled and entangled source-driven polymer interdiffusion in 2D. (a) Radially-averaged ϕ\phi profiles for 2D unentangled source-driven A-mers obtained from the numerical solution of Eqs. (1)-(3) (α=1\alpha=1, NA=104N_A=10^4 and Q=0.01Q=0.01, early times are depicted as gray, and later times are drawn in red). Inset: the typical droplet radius as a function of time for different QQ values [see panel (c) for legends], with a power-law of t1/2t^{1/2} as a guide to the eye. (b) Radially-averaged ϕ\phi profiles for 2D entangled A-mers (α=104\alpha=10^4, NA=104N_A=10^4 and Q=0.01Q=0.01, early times are depicted as gray, and later times are drawn in blue). Inset: the typical droplet radius as a function of time for different QQ values [see panel (d) for legends], with a power-law of t1/2t^{1/2} as a guide to the eye. (c) The rescaled solution Φu=ϕ/Q1/2\Phi_u=\phi/Q^{1/2} in the rescaled coordinates ξ=r/(Qt2)1/4\xi=r/(Qt^2)^{1/4} (darker and lighter curves representing different times of the same trajectory). (d) The rescaled solution Φe=ϕ/Q1/3\Phi_e=\phi/Q^{1/3}, and ξ=r/(Q2t3)1/6\xi=r/(Q^2t^3)^{1/6}. Inset: Φe2\Phi_e^2 versus η−η∗\eta-\eta_* showing Φe∼η∗−η\Phi_e\sim\sqrt{\eta_*-\eta}.

Discussion

Inspired by biocondensate diffusive and mixing dynamics, we used a two-fluid formalism to analyze polymer-polymer interdiffusion problems. We presented the equations of motion and demonstrated that under certain approximations the interdiffusion of a polymer droplet could be cast as a nonlinear diffusion equation, similar to diffusion in porous media [56, 57]. We highlighted the fact that the entanglement parameter α\alpha can dramatically alter the nonlinearity exponent mm, changing the anticipated scaling and spatial structure of the solutions. The results obtained from our full numerical simulations in both 1D and 2D corresponded well with the anticipated self-similar analytical solutions for the two α\alpha limits considered for unentangled case m=2m=2 (corresponding to α=1\alpha=1), and the entangled case m=3m=3 (corresponding to α≫1\alpha\gg1).

Then, we introduced a source into the equations of motion — akin to a continuous synthesis of polymers, prevalent in biological contexts. We demonstrated that for this case there are multiple temporal scaling laws that characterize the spread at different time regimes, and that while the source alters the time-dependence of the solutions and breaks the self-similar structure, the local geometry near the front bears similarity to the “passive diffusion” cases. Using both 1D and 2D numerical solutions, we demonstrated this similarity persists for various qq values, and in different spatial dimensions dd.

We suspect that the local front geometry could affect polymer conformations [58]. The sharp interface of droplets hints that the velocity profile should vary dramatically there, and should converge to a constant sufficiently far away. This local variation at the rim induces strain-rate gradients, and induce compressive forces on the polymers, causing them to adopt folded conformations. We suspect such effect could partially explain the compression of ribosomal RNA conformations as they drift away from the nucleolus [38, 58].

In the above work we neglected elastic and viscous effects that could modify the shape of the formed droplets, and cause the solutions to deviate from the reduced self-similar solutions. Such coupling between osmotic pressure and mechanics lies at the heart of gel physics [24, 26, 29, 31, 47, 48, 49, 50]. While the source-driven polymer interdiffusion problem considered above is reminiscent of a driven Cahn–Hilliard system [59, 60], the presence of material stresses may introduce additional temporal dependencies and spatial patterns, especially near the center and edges of the expanding droplets.

It may be interesting to consider a source-driven problem for phase separating mixtures. In such cases, the osmotic pressure would oppose the outward flux, which would eventually lead to droplet formation. We suspect this scenario could also be relevant for biological phase separation [61, 62]. Overall, we believe the presented two-fluid framework, the analysis above, and future works would deepen our understanding of polymer interdiffusion physics and biophysical diffusion processes.

Acknowledgments. — A.M. acknowledges support from the James S. McDonnell Foundation Postdoctoral Fellowship Award in Complex Systems. We acknowledge support from the NSF Grant DMS/NIGMS 2245850 and from the Princeton Center for Complex Materials (PCCM), an NSF-supported Materials Research Science and Engineering Center under award DMR-2011750. The authors also thank N. Wingreen and C. P. Brangwynne for helpful comments and discussions.

Two-dimensional passive droplet diffusion

In 2D we assume that the flow is radial, implying the radial velocities uAru_A^r and uBru_B^r obey ∇⋅(ϕuAr+(1−ϕ)uBr)≡∇⋅v=0\nabla\cdot(\phi u_A^r + (1-\phi)u_B^r) \equiv\nabla\cdot\boldsymbol{v} = 0, implying vr∼v0rv^r \sim\frac{v_0}{r}. From here we use similar simplifications as those used for the 1D case, neglecting the material stresses σA\boldsymbol{\sigma}_A and σB\boldsymbol{\sigma}_B, so that wr≡uAr−uBr≃−(1−ϕ)μ∂rΠw^r \equiv u_A^r-u_B^r \simeq-\frac{(1-\phi)}{\mu}\partial_r\Pi (Π≡IΠ\boldsymbol{\Pi}\equiv I\Pi, assumed to be isotropic). Expressing uAr=vr+(1−ϕ)2μ∂rΠu_A^r=v^r+\frac{(1-\phi)^2}{\mu}\partial_r\Pi, and using this back in (1) (without a source) yields

∂tϕ≃1r∂r(rϕm−1∂rϕ)−vr∂rϕ,(27)\partial_t\phi\simeq\frac{1}{r}\partial_r\left(r\phi^{m-1}\partial_r\phi\right)-v^r\partial_r\phi, \tag*{(27)}

where we used ∇⋅v=0\nabla\cdot\boldsymbol{v}=0. While vr≃v0rv^r \simeq\frac{v_0}{r} could contribute significantly near the origin, as we are examining a droplet expansion, we assume that at sufficiently late times when the droplet has extended, this term can be neglected, yielding a similar nonlinear diffusion equation as in the 1D case.

From here we proceed with the usual scaling arguments. The nonlinear equation implies ℓ2∼ϕm−1t\ell^2\sim\phi^{m-1}t, and the integral conservation law implies ϕℓ2∼L\phi\ell^2\sim\mathcal{L}, yielding ϕ∼(L/t)1/m\phi\sim(\mathcal{L}/t)^{1/m}, and ℓ∼Lm−12mt12m\ell\sim\mathcal{L}^{\frac{m-1}{2m}}t^{\frac{1}{2m}}, as obtained from (21) (with d=2d=2).

Higher-dimensional problems with a source

In higher spatial dimensions, we add (1) to obtain

∇⋅v=S.(28)\nabla\cdot\boldsymbol{v}=S . \tag*{(28)}

Now, using w\boldsymbol{w} and v\boldsymbol{v} to express uA=v+(1−ϕ)w\boldsymbol{u}_A=\boldsymbol{v}+(1-\phi)\boldsymbol{w}, we can rewrite

∂tϕ=∇⋅[(1−ϕ)2ϕμ∇⋅Π−vϕ]+S,(29)\partial_t\phi=\nabla\cdot\left[\frac{(1-\phi)^2\phi}{\mu}\nabla\cdot\boldsymbol{\Pi}-\boldsymbol{v}\phi\right]+S , \tag*{(29)}

analogous to (22), where here we again neglect the material stresses σA\boldsymbol{\sigma}_A and σB\boldsymbol{\sigma}_B. We can then express this also as

∂tϕ+v⋅∇ϕ=∇⋅[(1−ϕ)2ϕμ∇⋅Π]+S(1−ϕ),(30)\partial_t\phi+\boldsymbol{v}\cdot\nabla\phi=\nabla\cdot\left[\frac{(1-\phi)^2\phi}{\mu}\nabla\cdot\boldsymbol{\Pi}\right]+S(1-\phi) , \tag*{(30)}

analogous to (23), where we used (28). This explicitly shows that the addition of a source implies an advection term, and a modification to the source amplitude, as was shown in 1D above.

Transformation into self-similar variables for the source-driven problem

Here we provide detailed calculations that show how using the transformations ϕ=Qγ~1t~δ~1Φ(η)\phi=Q^{\tilde{\gamma}_1}\tilde{t}^{\tilde{\delta}_1}\Phi(\eta) and η=[x−12qQtq]/Qγ~2t~δ~2\eta=\left[x-\frac{1}{2q}Qt^q\right]/Q^{\tilde{\gamma}_2}\tilde{t}^{\tilde{\delta}_2} proposed above in (22) result in ordinary differential equation.

To begin we evaluate

∂tη=−12Q1−γ~2tq−1−δ~2−δ~2t−1η,(31)\partial_t\eta=-\frac{1}{2}Q^{1-\tilde{\gamma}_2}t^{q-1-\tilde{\delta}_2}-\tilde{\delta}_2t^{-1}\eta, \tag*{(31)}
∂xη=Q−γ~2t−δ~2.\partial_x\eta=Q^{-\tilde{\gamma}_2}t^{-\tilde{\delta}_2}.

Taking similar derivatives for ϕ\phi, we have

∂tϕ=δ~1Qγ~1tδ~1−1Φ+Qγ~1tδ~1Φ′∂tη,\partial_t\phi=\tilde{\delta}_1Q^{\tilde{\gamma}_1}t^{\tilde{\delta}_1-1}\Phi+Q^{\tilde{\gamma}_1}t^{\tilde{\delta}_1}\Phi'\partial_t\eta,
∂xϕ=Qγ~1tδ~1Φ′∂xη.(32)\partial_x\phi=Q^{\tilde{\gamma}_1}t^{\tilde{\delta}_1}\Phi'\partial_x\eta. \tag*{(32)}

Using the above relations in the left-hand side of (22) (focusing on x>0x>0 for simplicity), results in

∂tϕ+12Qtq−1∂xϕ=Qγ~1tδ~1−1(δ~1Φ−δ~2ηΦ′).(33)\partial_t\phi+\frac{1}{2}Qt^{q-1}\partial_x\phi=Q^{\tilde{\gamma}_1}t^{\tilde{\delta}_1-1}\left(\tilde{\delta}_1\Phi-\tilde{\delta}_2\eta\Phi'\right) . \tag*{(33)}

Similarly, using the above in the nonlinear diffusion term yields

∂x(ϕm−1∂xϕ)=Qmγ~1−2γ~2tmδ~1−2δ~2[(m−1)Φ′2+ΦΦ′′]Φm−2(34)\partial_x\left(\phi^{m-1}\partial_x\phi\right)=Q^{m\tilde{\gamma}_1-2\tilde{\gamma}_2}t^{m\tilde{\delta}_1-2\tilde{\delta}_2}\left[(m-1)\Phi^{\prime2}+\Phi\Phi^{\prime\prime}\right]\Phi^{m-2} \tag*{(34)}

Comparing the powers of these two expression reveals

(m−1)γ~1=2γ~2,(m−1)δ~1=2δ~2−1,(35)\begin{aligned} (m-1)\tilde{\gamma}_1&=2\tilde{\gamma}_2,\\ (m-1)\tilde{\delta}_1&=2\tilde{\delta}_2-1, \tag*{(35)} \end{aligned}

both obeyed by Eq. (24). Using m=2m=2 and m=3m=3 yield Eqs. (25)-(26) respectively.

References

References

  1. [1]P. G. de Gennes, Scaling Concepts in Polymer Physics (Cornell University Press, 1979).
  2. [2]M. Rubinstein and R. H. Colby, Polymer Physics (Oxford University Press, 2003).
  3. [3]M. Doi and S. F. Edwards, The Theory of Polymer Dynamics (Oxford University Press, 1988).
  4. [4]P. G. de Gennes, Dynamics of fluctuations and spinodal decomposition in polymer blends, J. Chem. Phys. 72, 4756 (1980).
  5. [5]P. Pincus, Dynamics of fluctuations and spinodal decomposition in polymer blends. ii, J. Chem. Phys. 75, 1996 (1981).DOI
  6. [6]M. Doi and S. F. Edwards, Dynamics of concentrated polymer systems. part 1.—brownian motion in the equilibrium state, J. Chem. Soc. Faraday Trans. 74, 1789 (1978).
  7. [7]M. Doi and S. F. Edwards, Dynamics of concentrated polymer systems. part 2.—molecular motion under flow, J. Chem. Soc. Faraday Trans. 74, 1802 (1978).
  8. [8]M. Doi and S. F. Edwards, Dynamics of concentrated polymer systems. part 3.—the constitutive equation, J. Chem. Soc. Faraday Trans. 74, 1818 (1978).
  9. [9]F. Brochard, J. Jouffroy, and P. Levinson, Polymer-polymer diffusion in melts, Macromolecules 16, 1638 (1983).
  10. [10]F. Brochard-Wyart, Kinetics of polymer-polymer interdiffusion, in Fundamentals of Adhesion (Springer, 1991) pp. 181–206.
  11. [11]P. F. Green and E. J. Kramer, Diffusion in polymer alloy melts, MRS Bull. 12, 42 (1987).
  12. [12]E. A. Jordan, R. C. Ball, A. M. Donald, L. J. Fetters, R. A. L. Jones, and J. Klein, Mutual diffusion in blends of long and short entangled polymer chains, Macromolecules 21, 235 (1988).
  13. [13]F. Brochard and P. G. de Gennes, Polymer-polymer interdiffusion, EPL 1, 221 (1986).
  14. [14]H. H. Kausch and M. Tirrell, Polymer interdiffusion, Ann. Rev. Mater. Sci. 19, 341 (1989).
  15. [15]J. Klein, The interdiffusion of polymers, Science 250, 640 (1990).
  16. [16]P. G. de Gennes, Reptation of a polymer chain in the presence of fixed obstacles, J. Chem. Phys. 55, 572 (1971).
  17. [17]J. Klein, Evidence for reptation in an entangled polymer melt, Nature 271, 143 (1978).
  18. [18]T. Russell, V. Delaine, W. Dozier, G. Felcher, G. Agrawal, R. Wool, and J. Mays, Direct observation of reptation at polymer interfaces, Nature 365, 235 (1993).
  19. [19]C. P. Brangwynne, P. Tompa, and R. V. Pappu, Polymer physics of intracellular phase transitions, Nature Physics 11, 899 (2015).
  20. [20]A. A. Hyman, C. A. Weber, and F. Jülicher, Liquid-liquid phase separation in biology, Annu. Rev. Cell Dev. Biol. 30, 39 (2014).DOI
  21. [21]C. P. Brangwynne, C. R. Eckmann, D. S. Courson, A. Rybarska, C. Hoege, J. Gharakhani, F. Jülicher, and A. A. Hyman, Germline p granules are liquid droplets that localize by controlled dissolution/condensation, Science 324, 1729 (2009).DOI
  22. [22]Y. Shin and C. P. Brangwynne, Liquid phase condensation in cell physiology and disease, Science 357, eaaf4382 (2017).DOI
  23. [23]J. Berry, C. P. Brangwynne, and M. Haataja, Physical principles of intracellular organization via active and passive phase transitions, Rep. Prog. Phys. 81, 046601 (2018).
  24. [24]H. Tanaka, Unusual phase separation in a polymer solution caused by asymmetric molecular dynamics, Phys. Rev. Lett. 71, 3158 (1993).
  25. [25]T. Taniguchi and A. Onuki, Network domain structure in viscoelastic phase separation, Phys. Rev. Lett. 77, 4910 (1996).
  26. [26]H. Tanaka and T. Araki, Phase inversion during viscoelastic phase separation: roles of bulk and shear relaxation moduli, Phys. Rev. Lett. 78, 4966 (1997).DOI
  27. [27]T. Araki and H. Tanaka, Three-dimensional numerical simulations of viscoelastic phase separation: Morphological characteristics, Macromolecules 34, 1953 (2001).
  28. [28]J. Zhang, Z. Zhang, H. Zhang, and Y. Yang, Kinetics and morphologies of viscoelastic phase separation, Phys. Rev. E 64, 051510 (2001).
  29. [29]H. Tanaka and T. Araki, Viscoelastic phase separation in soft matter: Numerical-simulation study on its physical mechanism, Chem. Eng. Sci. 61, 2108 (2006).
  30. [30]M. Doi, Gel dynamics, J. Phys. Soc. Jpn. 78, 052001 (2009).
  31. [31]H. Tanaka, Viscoelastic phase separation in biological cells, Commun. Phys. 5, 167 (2022).
  32. [32]I. Alshareedah, M. M. Moosa, M. Pham, D. A. Potoyan, and P. R. Banerjee, Programmable viscoelasticity in protein-rna condensates with disordered sticker-spacer polypeptides, Nat. Commun. 12, 6620 (2021).
  33. [33]L. Javerth, E. Fischer-Friedrich, S. Saha, J. Wang, T. Franzmann, X. Zhang, J. Sachweh, M. Ruer, M. Ijavi, S. Saha, et al., Protein condensates as aging maxwell fluids, Science 370, 1317 (2020).
  34. [34]H. H. Cheng, J. V. Roggeveen, H. Wang, H. A. Stone, Z. Shi, and C. P. Brangwynne, Micropipette aspiration reveals differential rna-dependent viscoelasticity of nucleolar subcompartments, Proc. Natl. Acad. Sci. U. S. A. 122, e2407423122 (2025).DOI
  35. [35]A. Ghosh, D. Kota, and H.-X. Zhou, Shear relaxation governs fusion dynamics of biomolecular condensates, Nat. Commun. 12, 5995 (2021).DOI
  36. [36]F.-M. Boisvert, S. Van Koningsbruggen, J. Navascués, and A. I. Lamond, The multifunctional nucleolus, Nat. Rev. Mol. Cell Biol. 8, 574 (2007).
  37. [37]M. Carmo-Fonseca, L. Mendes-Soares, and I. Campos, To be or not to be in the nucleolus, Nat. Cell Biol. 2, E107 (2000).
  38. [38]J. A. Riback, J. M. Eeftens, D. S. Lee, S. A. Quinódoz, A. Donlic, N. Orlovsky, L. Wiesner, L. Beckers, L. A. Becker, A. R. Strom, et al., Viscoelasticity and advective flow of rna underlies nucleolar form and function, Mol. Cell 83, 3095 (2023).
  39. [39]E. B. Mougey, M. O’Reilly, Y. Osheim, O. L. Miller, A. Beyer, and B. Sollner-Webb, The terminal balls characteristic of eukaryotic rrna transcription units in chromatin spreads are rrna processing complexes, Genes Dev. 7, 1609 (1993).
  40. [40]O. L. Miller Jr, The nucleolus, chromosomes, and visualization of genetic activity, J. Cell Biol. 91, 15s (1981).
  41. [41]S. Milner, Hydrodynamics of semidilute polymer solutions, Phys. Rev. Lett. 66, 1477 (1991).
  42. [42]M. Doi and A. Onuki, Dynamic coupling between stress and composition in polymer solutions and blends, Journal de Physique II 2, 1631 (1992).
  43. [43]V. G. Mavratzas and A. N. Beris, Modeling of the rheology and flow-induced concentration changes in polymer solutions, Phys. Rev. Lett. 69, 273 (1992).
  44. [44]R. L. Larson, Flow-induced mixing, demixing, and phase transitions in polymeric fluids, Rheol. Acta. 31, 497 (1992).
  45. [45]A. N. Beris and B. J. Edwards, Thermodynamics of Flowing Systems: With Internal Microstructure, 36 (Oxford University Press, 1994).
  46. [46]A. N. Beris and V. G. Mavrantzas, On the compatibility between various macroscopic formalisms for the concentration and flow of dilute polymer solutions, J. Rheol. 38, 1235 (1994).
  47. [47]H. Tanaka, Universality of viscoelastic phase separation in dynamically asymmetric fluid mixtures, Phys. Rev. Lett. 76, 787 (1996).
  48. [48]H. Tanaka, Viscoelastic model of phase separation, Phys. Rev. E 56, 4451 (1997).
  49. [49]H. Tanaka, Viscoelastic phase separation, J. Phys. Condens. Matter 12, B207 (2000).
  50. [50]H. Tanaka, Viscoelastic phase separation in soft matter and foods, Farad. Discuss. 158, 371 (2012).
  51. [51]P. J. Flory, Principles of Polymer Chemistry (Cornell University Press, 1953).
  52. [52]M. L. Huggins, Theory of solutions of high polymers1, J. Am. Chem. Soc. 64, 1712 (1942).
  53. [53]E. Boyko and H. A. Stone, Perspective on the description of viscoelastic flows via continuum elastic dumbbell models, J. Eng. Math. 147, 5 (2024).
  54. [54]G. Van Rossum and F. L. Drake Jr, Python tutorial (Centrum voor Wiskunde en Informatica Amsterdam, The Netherlands, 1995).
  55. [55]C. R. Harris, K. J. Millman, S. J. van der Walt, R. Gommers, P. Virtanen, D. Cournapeau, E. Wieser, J. Taylor, S. Berg, N. J. Smith, R. Kern, M. Picus, S. Hoyer, M. H. van Kerkwijk, M. Brett, A. Haldane, J. F. del Río, M. Wiebe, P. Peterson, P. Gérard-Marchant, K. Sheppard, T. Reddy, W. Weckesser, H. Abbasi, C. Gohlke, and T. E. Oliphant, Array programming with NumPy, Nature 585, 357 (2020).
  56. [56]G. I. Barenblatt, Scaling, Vol. 34 (Cambridge University Press, 2003).
  57. [57]Z. Zheng and H. A. Stone, The influence of boundaries on gravity currents and thin films: drainage, confinement, convergence, and deformation effects, Annu. Rev. Fluid Mech. 54, 27 (2022).
  58. [58]A. Moriel and H. A. Stone, Nonlinear diffusion and compressive rims in source-driven biopolymer condensates, a link will be provided by the publisher.
  59. [59]J. W. Cahn and J. E. Hilliard, Free energy of a nonuniform system. i. interfacial free energy, J. Chem. Phys 28, 258 (1958).
  60. [60]J. W. Cahn and J. E. Hilliard, Free energy of a nonuniform system. iii. nucleation in a two-component incompressible fluid, J. Chem. Phys 31, 688 (1959).
  61. [61]S. Alberti, Phase separation in biology, Curr. Biol. 27, R1097 (2017).
  62. [62]D. M. Mitrea and R. W. Kriwacki, Phase separation in biology; functional organization of a higher order, Cell Commun. Signal. 14, 1 (2016).

Paper details

vlnc:W_m7uy2ady
Version
v1 · latest
Published
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents