betaFeedback

WorksPhysics and Astronomy

A single-particle energy-conserving dissipative particle dynamics approach for simulating thermophoresis of nanoparticles in polymer networks

Abstract

Thermophoresis is an effective method to drive the motion of nanoparticles in fluids. The transport of nanoparticles in polymer networks has significant fundamental and applied importance in biology and medicine, and can be described as Brownian particles crossing entropic barriers. This study proposes a novel extension of dissipative particle dynamics (DPD), called the single-particle energy-conserving dissipative particle dynamics (seDPD), which combines the features of single-particle dissipative particle dynamics (sDPD) and energy-conserving dissipative particle dynamics (eDPD) to simulate the thermophoresis of nanoparticles under temperature gradients. The reliability of the seDPD method is verified by considering the viscosity, thermal diffusivity, and hydrodynamic drag force on the nanoparticles. Using this method, the transport of nanoparticles driven by the thermophoretic force across the polymer network is simulated. The results show that the nanoparticles exhibit the phenomenon of giant acceleration of diffusion (GAD) in the polymer network, indicating that Brownian particles can exhibit GAD when crossing entropic barriers.

introduction

Thermophoresis[1, 2, 3, 4, 5] is the phenomenon of nanoparticle migration under a temperature gradient, often also referred to as the Soret effect. It is an effective method to move nanoparticles in fluids. Considering the size of nanoparticles ranging from tens to hundreds of nanometers[6, 7, 8], mesoscopic molecular dynamics methods are powerful tools for studying nanoparticle thermophoresis. One such mesoscopic method is dissipative particle dynamics proposed by Hoogerbrugge and Koelman[9], which has been applied to various systems such as DNA[10, 11], rheology[12, 13, 14], drug release[15, 16], and anomalous diffusion[17, 18, 19]. The standard DPD method follows Newton’s second law, where the particles experience conservative, dissipative, and random forces. The conservative force is a soft repulsive force that ensures the compressibility of the fluid. The dissipative force acts as the fluid viscosity, and the random force represents the stochastic effects at the mesoscopic scale. The dissipative force and the random force obey the fluctuation-dissipation theorem.

The DPD method is a flexible and extensible technique for mesoscopic simulations. Several extensions of the DPD method have been proposed to model different physical phenomena. For example, the many-body dissipative particle dynamics (mDPD)[20] method adds an attractive term to the conservative force to simulate vapor-liquid coexistence systems. The energy-conserving dissipative particle dynamics (eDPD)[21] method ensures the conservation of kinetic and internal energy by introducing temperature and heat flux in DPD, which enables the study of heat transfer and thermal convection. The single-particle dissipative particle dynamics (sDPD)[22] method introduces a non-central dissipative force and solves the conservation of the angular momentum of the DPD particle, which allows the simulation of spherical nanoparticles with finite size by a single DPD particle. Furthermore, these extended DPD methods can be integrated with each other. For instance, Zhang et al.[23] proposed a hybrid method of mDPD and eDPD, named mDPDe, which can simulate the thermocapillary motion of droplets on solid surfaces.

Despite the long history and versatility of the dissipative particle dynamics (DPD) method for simulating various systems, few existing method can model the thermophoretic behavior of nanoparticles using a single DPD particle. To address this gap, we propose a novel method that combines the sDPD method for nanoparticle simulation with the eDPD model for temperature and heat flux. Our method, which we call single-particle energy-conserving dissipative particle dynamics (seDPD). There are two technical issues for implementing the thermophoresis of nanoparticles in the seDPD method: (1) the viscous heat flux in eDPD needs to be modified due to the work done by the non-central dissipative and random forces in sDPD; (2) how to apply the thermophoretic force to the nanoparticles in the fluid with a temperature gradient.

We modify the conservative force in seDPD to account for the thermophoretic force FTF^{T} that the temperature gradient ∇Θ\nabla\Theta generates, which enables the nanoparticles to move along or against the gradient. The Soret force FTF^{T} depends on both the temperature gradient ∇Θ\nabla\Theta and the particle size RR. Some previous studies have established the relation between FTF^{T} and the size of nanoparticles as well as the temperature gradient. Duhr et al.[4] report experimental results for particles in liquids with temperature gradients, showing that the Soret coefficient ST=FT/∇ΘS_{T}=F_{T}/\nabla\Theta is proportional to the square of nanoparticle size (ST∼R2S_{T}\sim R^{2}). Mayer et al.[24] measure the thermophoresis of polystyrene beads over a wide range of temperature gradients. They find that the particle motion is dominated by fluctuations when the Peclet number Pe=RST∇Θ<1P_{e}=RS_{T}\nabla\Theta<1, and that Soret force varies linearly with the temperature gradient. For large Peclet numbers (Pe=RST∇Θ>1Pe = RST\nabla\Theta> 1), the Soret force varies sublinearly with the temperature gradient, the thermophoresis is dominated by drift and approaches the hydrodynamic model[25, 26]. The results of these studies will help us to implement the correct thermophoresis in the seDPD method, and provide theoretical and experimental basis for the parameterization in seDPD.

In this work, we propose a modified DPD method, seDPD, to model the thermophoresis of nanoparticles in fluids. This method combines the sDPD with the eDPD. The validation of viscosity, heat conduction and hydrodynamic drag force on nanoparticles in seDPD is applied. Furthermore, we develop a parameterization to control thermophoretic force on nanoparticles. Finally, we simulate the nanoparticle transport in the polymer network immersed with fluid with temperature gradients. This proposed model is not limited to thermophoresis of nanoparticles and can readily be extended to investigate various system including temperature and nanoparticles.

The paper is organized as follows. In section II, we introduce the governing equations and parameters of seDPD, the strategy of choosing physical units, and the methods of calculating the thermophoretic force. In section III, we validate the seDPD method by some benchmark cases to ensure its reliability in simulating the thermophoresis of nanoparticles, then we present the relation between the thermophoretic force and the seDPD parameters, and finally we simulate the transport of nanoparticles driven by the thermophoretic force in a polymer network. Finally, we conclude with a brief summary in section IV.

model and method

Governing equations of seDPD

In seDPD method, the simulation system contains a collection of finite-size particles with mass mim_i and mass moment of inertia IiI_i and heat capacity CviC_{v_i} for the iith particle. The governing equation of particle motion is the conservation of momentum and energy:

dr⃗idt=v⃗i,mid2r⃗idt2=F⃗i=∑jF⃗ij,(1)\frac{d\vec{r}_i}{dt}=\vec{v}_i,\quad m_i\frac{d^2\vec{r}_i}{dt^2}=\vec{F}_i=\sum_j\vec{F}_{ij}, \tag*{(1)}
Iidω⃗idt=T⃗i=−∑jλijr⃗ij×F⃗ij,(2)I_i\frac{d\vec{\omega}_i}{dt}=\vec{T}_i=-\sum_j\lambda_{ij}\vec{r}_{ij}\times\vec{F}_{ij}, \tag*{(2)}
Cvi∂Θi∂t=qi=∑i≠jqij,(3)C_{v_i}\frac{\partial\Theta_i}{\partial t}=q_i=\sum_{i\ne j}q_{ij}, \tag*{(3)}

where r⃗i\vec{r}_i, v⃗i\vec{v}_i, ω⃗i\vec{\omega}_i and Θi\Theta_i are the vector of iith particle’s position, velocity, angular velocity and temperature, r⃗ij=r⃗i−r⃗j\vec{r}_{ij}=\vec{r}_i-\vec{r}_j. The magnitude of the vector rij=∣r⃗ij∣r_{ij}=|\vec{r}_{ij}| is the distance between particle jj and particle ii. The factor λij\lambda_{ij} in angular momentum equation is a weighted parameter to account for the different contributions from the particles with different size for the conservation of the angular momentum[27], which is defined as:

λij=RiRi+Rj=11+RjRi,(4)\lambda_{ij}=\frac{R_i}{R_i+R_j}=\frac{1}{1+\frac{R_j}{R_i}}, \tag*{(4)}

where RiR_i and RjR_j denote the radii of particle ii and particle jj. If the size of small (solvent and polymer) particle jj is far smaller than NP ii, then Rj/Ri∼0R_j/R_i\sim0, yields λij=1\lambda_{ij}=1 and λji=0\lambda_{ji}=0, thus the angular momentum equation of solvent and polymer in equation (2) vanishes. F⃗i\vec{F}_i is the total force acting on the particle ii. F⃗ij\vec{F}_{ij} is the force exerted on particle ii by particle jj, which consists of three parts:

F⃗ij=F⃗ijC+F⃗ijD+F⃗ijR.(5)\vec{F}_{ij}=\vec{F}_{ij}^{C}+\vec{F}_{ij}^{D}+\vec{F}_{ij}^{R}. \tag*{(5)}

The conservative force is given by

F⃗ijC=aijwC(rij)e⃗ij(6)\vec{F}_{ij}^{C}=a_{ij}w_C(r_{ij})\vec{e}_{ij} \tag*{(6)}

In traditional DPD, aija_{ij} is the maximum of conservative force, e⃗ij=r⃗ij/rij\vec{e}_{ij}=\vec{r}_{ij}/r_{ij} is a unit vector, and the weight function is given by

wC(rij)={1−rijRCC,rij<RCC,0rij>RCC(7)w_C(r_{ij})= \begin{cases} 1-\frac{r_{ij}}{R_{CC}}, & r_{ij}<R_{CC},\\ 0 & r_{ij}>R_{CC} \end{cases} \tag*{(7)}

where RCCR_{CC} is the cut-off radius of conservative force. The dissipative force consists of three parts F⃗ijD=F⃗ijDc+F⃗ijDs+F⃗ijDr\vec{F}_{ij}^{D}=\vec{F}_{ij}^{Dc}+\vec{F}_{ij}^{Ds}+\vec{F}_{ij}^{Dr}: (1) The central dissipative force

F⃗ijDc=−γijcwD(rij)(e⃗ij⋅v⃗ij)e⃗ij;(8)\vec{F}_{ij}^{Dc}=-\gamma_{ij}^{c}w_D(r_{ij})(\vec{e}_{ij}\cdot\vec{v}_{ij})\vec{e}_{ij}; \tag*{(8)}

(2) The shearing dissipative force

F⃗ijDs=−γijswD2(rij)[v⃗ij−(e⃗ij⋅v⃗ij)e⃗ij];(9)\vec{F}_{ij}^{Ds}=-\gamma_{ij}^{s}w_D^2(r_{ij})[\vec{v}_{ij}-(\vec{e}_{ij}\cdot\vec{v}_{ij})\vec{e}_{ij}]; \tag*{(9)}

(3) The rotational dissipative force

F⃗ijDr=−γijswD(rij)[r⃗ij×(λijω⃗i+λjiω⃗j)],(10)\vec{F}_{ij}^{Dr}=-\gamma_{ij}^{s}w_D(r_{ij})[\vec{r}_{ij}\times(\lambda_{ij}\vec{\omega}_i+\lambda_{ji}\vec{\omega}_j)], \tag*{(10)}

where v⃗ij=v⃗i−v⃗j\vec{v}_{ij}=\vec{v}_i-\vec{v}_j is relative velocity and wD(rij)w_D(r_{ij}) is the weight function for dissipative force. γijc\gamma_{ij}^{c} and γijs\gamma_{ij}^{s} are the central and shearing dissipative coefficients. The rotational dissipative force has the same dissipative parameter γijs\gamma_{ij}^{s} with the shearing translational dissipative. For γijs=0\gamma_{ij}^{s}=0, the system recovers to traditional DPD.

The random force is defined by

F⃗ijRdt=wR(rij)[13σijctr[dWij]E+2σijsdWijA]⋅e⃗ij(11)\vec{F}_{ij}^{R}dt=w_R(r_{ij})\left[\frac{1}{\sqrt{3}}\sigma_{ij}^{c}tr[d\mathbf{W}_{ij}]\mathbf{E}+\sqrt{2}\sigma_{ij}^{s}d\mathbf{W}_{ij}^{A}\right]\cdot\vec{e}_{ij} \tag*{(11)}

where dtdt is the time step and tr[dWij]tr[d\mathbf{W}_{ij}] is the trace of symmetric independent Wiener increment matrix dWijd\mathbf{W}_{ij}, E\mathbf{E} is unit matrix. Anti-symmetric part of dWijd\mathbf{W}_{ij} is dWijAd\mathbf{W}_{ij}^{A}. σijc\sigma_{ij}^{c} and σijs\sigma_{ij}^{s} are the parameters in random force and wR(rij)w_R(r_{ij}) is the weight function of random force, which follows the fluctuation-dissipative theorem

(σijc)2=4kBγijcΘiΘjΘi+Θj,(σijs)2=4kBγijsΘiΘjΘi+Θj.(12)(\sigma_{ij}^{c})^2=\frac{4k_B\gamma_{ij}^{c}\Theta_i\Theta_j}{\Theta_i+\Theta_j},\quad(\sigma_{ij}^{s})^2=\frac{4k_B\gamma_{ij}^{s}\Theta_i\Theta_j}{\Theta_i+\Theta_j}. \tag*{(12)}
wD(rij)=wR2(rij)=(1−rijRCD)s(13)w_D(r_{ij})=w_R^2(r_{ij})=\left(1-\frac{r_{ij}}{R_{CD}}\right)^s \tag*{(13)}

where kBk_B is the Boltzmann constant, RCDR_{CD} is the cut-off radius for dissipative and random force, and ss is the exponent of weight function[28].

q⃗i\vec{q}_i is the total heat flux that acts on the particle ii. q⃗ij\vec{q}_{ij} is the heat flux exerted on particle ii by particle jj, which consists of three parts:

qij=qijC+qijV+qijR(14)q_{ij}=q_{ij}^{C}+q_{ij}^{V}+q_{ij}^{R} \tag*{(14)}

in which qijCq_{ij}^{C} and qijRq_{ij}^{R} is the collisional and random heat flux, given by

qijC=kijwCT(rij)(1Θi−1Θj),q_{ij}^{C}=k_{ij}w_{CT}(r_{ij})\left(\frac{1}{\Theta_i}-\frac{1}{\Theta_j}\right),
qijRdt=βijwRT(rij)dWije.(15)q_{ij}^{R}dt=\beta_{ij}w_{RT}(r_{ij})dW_{ij}^{e}. \tag*{(15)}

where kij=CviCvjκΘij2/kBk_{ij}=C_{v_i}C_{v_j}\kappa\Theta_{ij}^{2}/k_B in which κ\kappa is the mesoscale heat friction coefficient. βij2=kijkB\beta_{ij}^{2}=k_{ij}k_B. The weight function wCT(rij)w_{CT}(r_{ij}) and wRT(rij)w_{RT}(r_{ij}) are given as wCT(rij)=wRT2(rij)=1−rij/RCTw_{CT}(r_{ij})=w_{RT}^{2}(r_{ij})=1-r_{ij}/R_{CT} with the cut-off radius RCTR_{CT}.

The viscous heat flux contain three parts qijV=qijVc+qijVs+qijVrq_{ij}^{V}=q_{ij}^{Vc}+q_{ij}^{Vs}+q_{ij}^{Vr}: (1) viscous heat flux for central force

(Cvi+Cvj)qijVcdt=wD(rij)[(e⃗ij⋅v⃗ij)2−σijCm]dt−σijCwR(rij)(e⃗ij⋅v⃗ij)tr⁡[dWij]3(16)(C_{v_i}+C_{v_j})q_{ij}^{Vc}dt=w_D(r_{ij})\left[(\vec{e}_{ij}\cdot\vec{v}_{ij})^2-\frac{\sigma_{ij}^{C}}{m}\right]dt-\sigma_{ij}^{C}w_R(r_{ij})(\vec{e}_{ij}\cdot\vec{v}_{ij})\frac{\operatorname{tr}[dW_{ij}]}{3} \tag*{(16)}

(2) viscous heat flux for shearing force

(Cvi+Cvj)qijVsdt=wD(rij)[−γijs(e⃗ij⋅v⃗ij)2+γijs(v⃗ij⋅v⃗ij)−2(σijs)2m]dt−σijswR(rij)dWijA⋅e⃗ij⋅v⃗ij(17)(C_{v_i}+C_{v_j})q_{ij}^{Vs}dt=w_D(r_{ij})\left[-\gamma_{ij}^{s}(\vec{e}_{ij}\cdot\vec{v}_{ij})^2+\gamma_{ij}^{s}(\vec{v}_{ij}\cdot\vec{v}_{ij})-\frac{2(\sigma_{ij}^{s})^2}{m}\right]dt-\sigma_{ij}^{s}w_R(r_{ij})dW_{ij}^{A}\cdot\vec{e}_{ij}\cdot\vec{v}_{ij} \tag*{(17)}

(3) viscous heat flux for rotational force

(Cvi+Cvj)qijVr=γijrwD(rij)[r⃗ij×(λijω⃗i+λjiω⃗j)]⋅v⃗ij(18)(C_{v_i}+C_{v_j})q_{ij}^{Vr}=\gamma_{ij}^{r}w_D(r_{ij})[\vec{r}_{ij}\times(\lambda_{ij}\vec{\omega}_i+\lambda_{ji}\vec{\omega}_j)]\cdot\vec{v}_{ij} \tag*{(18)}

In this paper, we study the thermophoresis problem of nanoparticles, where the rotation of the particles is not a dominant factor. Therefore, we do not calculate Eq. (2). As a result, the rotational dissipative force (Eq. (10)) and the viscous heat flux for rotational force (Eq. (18)) are also not accounted for in the simulation.

Mapping of units

In seDPD, the energy is conservative, the summation of potential energy, kinetic energy and internal energy is a constant. In equilibrium, potential energy is minimized, the energy conserving equation is

dEk∗=dEI∗dE_k^*=dE_I^*
mv∗v∗dv∗=m∗ρ∗Cv∗dΘ∗(19)m_v^*v^*dv^*=\frac{m^*}{\rho^*}C_v^*d\Theta^* \tag*{(19)}

Using the following dimensionless variables

v=v∗TL;m=m∗M;ρ=ρ∗L3M;Θ=kB∗Θ∗ML2/T2(20)v=\frac{v^*T}{L};m=\frac{m^*}{M};\rho=\frac{\rho^*L^3}{M};\Theta=\frac{k_B^*\Theta^*}{ML^2/T^2} \tag*{(20)}

the dimensionless energy conserving equation is written as

1Cvvdv=CvdΘ.(21)\frac{1}{C_v}vdv=C_vd\Theta. \tag*{(21)}

and CvC_v can be written as[21]:

Cv=Cv∗L3ρ∗kB∗;(22)C_v=\frac{C_v^*L^3}{\rho^*k_B^*}; \tag*{(22)}

in which kB∗k_B^* is the Boltzmann constant, Cv∗C_v^* is the volumetric heat capacity of fluid. The equation (22) indicate that the large length scale will lead to a large dimensionless heat capacity CvC_v.

For water at 300k, kB∗=1.381×10−23 J K−1k_B^*=1.381\times10^{-23}\,\mathrm{J\,K^{-1}}, Cv∗=4.167×106 J m−3 K−1C_v^*=4.167\times10^6\,\mathrm{J\,m^{-3}\,K^{-1}}. when we choose the length unit L=100nmL=100nm, temperature of reference 300K and number density ρ=3.0\rho=3.0, we get the dimensionless volumetric heat capacity Cv=1×108C_v=1\times10^8.

Method to measure Soret force

Considering a particle driven by the force proportional to the temperature gradients FT=−ST∇ΘF^T=-S_T\nabla\Theta and subject to the stochastic force. The trajectory of Brownian particle xx is governed by an over-damped non-linear Langevin equations

dx=−D0ST∇Θdt+2D0dW(23)dx=-D_0S_T\nabla\Theta dt+\sqrt{2D_0}d\mathcal{W} \tag*{(23)}

in which W\mathcal{W} is the wiener process, D0D_0 is the diffusivity for particle diffusion in simple fluid. The unsteady state method to measure Soret force is to calculate the mean displacement of particles ⟨x⟩/t=−D0ST∇Θ\langle x\rangle/t=-D_0S_T\nabla\Theta.

A method to calculated Soret force in steady state is proposed by Duhr et al. [4]. The corresponding Fokker-Planck function of equation (23) is given by

{∂ρ(x,t)∂t=−∂JT+JD∂x,JT=−D0ST∇Θρ(x,t),JD=−D0∂∂xρ(x,t),(24)\left\{ \begin{aligned} \frac{\partial\rho(x,t)}{\partial t}&=-\frac{\partial J_T+J_D}{\partial x},\\ J_T&=-D_0S_T\nabla\Theta\rho(x,t),\\ J_D&=-D_0\frac{\partial}{\partial x}\rho(x,t), \end{aligned} \right. \tag*{(24)}

where JTJ_T and JDJ_D is the mass flux associated with a thermal gradient and concentration gradient. In steady state, the total mass flux is zero (JT=−JDJ_T=-J_D), yielding

dρ/ρ=−STdΘ(25)d\rho/\rho=-S_Td\Theta \tag*{(25)}

which gives a convincing method to obtain the Soret coefficient ST=−dlog⁡(ρ)/dTS_T=-d\log(\rho)/dT from the experimental or numerical results of the concentration and temperature in steady state.

Results and discussion

Validation of seDPD

The algorithmic validity of seDPD for simulating thermophoresis is examined by considering the effects of fluid viscosity, thermal conductivity, and hydrodynamic drag force on nanoparticles.

First, we simulate the Poiseuille flow to validate the hydrodynamic properties of the seDPD method. We implement the non-slip boundary condition using reverse Poiseuille flow, which was proposed by Backer et al. [29] and is widely used in DPD simulation[21, 23]. The seDPD parameters for the Poiseuille flow are listed in the caption of Figure 1. The simulation domain is 20×10×820 \times10 \times8 with DPD units. We apply two equal and opposite driving forces of magnitude F=0.01F = 0.01 along the yy−direction. The velocity profile (x∈[0,10]x \in[0,10]) is shown in Figure 1, which is in good agreement with the analytical solution of the Navier-Stokes equations.

The velocity profile for the Poiseuille flow in seDPD

Figure 1. The velocity profile for the Poiseuille flow in seDPD. The parameters is set to ρ=3.0\rho= 3.0, aij=25a_{ij} = 25, γc=4.5\gamma_c = 4.5, γs=0\gamma_s = 0, s=0.4s = 0.4, RCC=RCD=1.0R_{CC} = R_{CD} = 1.0. The driven force is F=0.01F = 0.01.

vf=ρFx2ηs(Lx/2−∣x∣),(26)v_f = \frac{\rho F x}{2\eta_s}\left(L_x/2 - |x|\right), \tag*{(26)}

where half of the simulation domain in xx−direction is Lx/2=10L_x/2 = 10, and the dynamic viscosity ηs=1.83\eta_s = 1.83, which agrees with the results 1.8368 in our previous study[17] for sDPD with the same parameters.

A steady state heat conduction is simulated to validate the thermal conduction in seDPD. The stationary fluid is confined between the hot wall at TH=1.1T_H = 1.1 and the cold wall at Tc=0.9T_c = 0.9. The solid wall is constructed by freeze the fluid particles. The fluid particle penetrates the solid wall and bounces back to the flow field. As shown in Figure 2, the temperature Θ\Theta is linear with position, which is the steady solution for the heat transfer equation. The thermal conductivity of the fluid is obtained by simulating the heat conduction analog of reverse Poiseuille flow, which was proposed by Li et al. [21]. Using the same simulation box of the RPF, one-half of the fluid domain is heated, and the other-half is cooled. As shown in Figure 3, the temperature profile (x∈[0,10]x \in[0,10]) agrees well with the analytical solution of the heat conduction equation.

The temperature profile for seDPD and eDPD

Figure 2. The temperature profile for seDPD and eDPD. The parameters are set to κ=10−8\kappa= 10^{-8}, Cv=108C_v = 10^8, RCT=1.0R_{CT} = 1.0.

The temperature profile for seDPD and eDPD

Figure 3. The temperature profile for seDPD and eDPD. The parameters are set to κ=10−8\kappa= 10^{-8}, Cv=108C_v = 10^8, RCT=1.0R_{CT} = 1.0.

Θ−Θ0=ρQx2λs(Lx/2−∣x∣),(27)\Theta- \Theta_0 = \frac{\rho Q_x}{2\lambda_s}\left(L_x/2 - |x|\right), \tag*{(27)}

where Θ0\Theta_0 is the temperature of the seDPD system before heating/cooling, and λs\lambda_s is the thermal diffusivity. As shown in Figure 2 and Figure 3, the temperature profile in seDPD agrees with those of the eDPD model, indicating that we can use the seDPD model to obtain the same thermal properties as eDPD.

Nanoparticles moving in a fluid under a temperature gradient experience a Soret force that balances the hydrodynamic drag force, i.e., FT=γvF^T = \gamma v, where γ\gamma is the friction coefficient. It is crucial to verify that the nanoparticles receive the correct hydrodynamic drag for studying thermophoresis. The friction coefficient can be calculated from the diffusivity of the nanoparticles using Einstein’s formula D0=kBΘ/γD_0 = k_B\Theta/\gamma. The friction coefficient should follow the Stokes formula for the hydrodynamic drag of nanoparticles, i.e., γ=6πRηs\gamma= 6\pi R\eta_s. As shown in the Figure 4, the self-diffusion rate of nanoparticles in the fluid is inversely proportional to the size of nanoparticles and follows the Einstein-Stokes formula. This demonstrates that the seDPD method can accurately capture the hydrodynamic drag of nanoparticles in the fluid.

Free diffusivity of nanoparticles in simple fluid

Figure 4. The free diffusivity of nanoparticles in simple fluid.

Soret effect in seDPD

To model the Soret effect in a DPD system, the key problem is to apply FTF^{T} on a nanoparticle with a finite size RR. A modified conservative force is utilized for NP and solvent:

F⃗ijC=aijwC(rij)e⃗ij+F⃗ijCT,F⃗ijCT=−aijT(Θi−Θj)wCT(rij)e⃗ij.(28)\begin{aligned} \vec{F}_{ij}^{C} &= a_{ij}w_{C}(r_{ij})\vec{e}_{ij}+\vec{F}_{ij}^{CT},\\ \vec{F}_{ij}^{CT} &= -a_{ij}^{T}(\Theta_i-\Theta_j)w_{CT}(r_{ij})\vec{e}_{ij}. \tag*{(28)} \end{aligned}

To satisfy Newton’s third law (F⃗ijCT=−F⃗jiCT\vec{F}_{ij}^{CT}=-\vec{F}_{ji}^{CT}), the parameter must be set to aijT=−ajiTa_{ij}^{T}=-a_{ji}^{T}. For a positive Soret coefficient of the particle ii, aijTa_{ij}^{T} should be positive, and when Ti>TjT_i>T_j the particle with higher temperature will move to the region with lower temperature. The weight function wCw_C in Eqs. (28) for the interaction of NP and solvent has an exponential form[30, 31]:

wC(rij)={11−e−bij(e−bijrijR+δ−e−bij),rij<R+δ,0,rij>R+δ.(29)w_C(r_{ij})= \begin{cases} \frac{1}{1-e^{-b_{ij}}}\left(e^{-\frac{b_{ij}r_{ij}}{R+\delta}}-e^{-b_{ij}}\right), & r_{ij}<R+\delta,\\ 0, & r_{ij}>R+\delta. \end{cases} \tag*{(29)}

where bijb_{ij} is the factor of weight function, RR is the hydrodynamic radius of Brownian particles, and δ\delta is the effective radius of fluid particles[32]. By applying the exponential form weight function, the fluid particle cannot penetrate into the NP.

The modified conservative force (Eqs. (28)) provides the in-homogeneous interaction between nanoparticles and fluid particles in the temperature gradient, which results in the thermophoretic force on the nanoparticles parallel to the temperature gradient. As shown in Figure 5, we obtain the following scaling law for the Soret coefficient and Soret force:

Numerical results of Soret coefficient

Figure 5. The numerical results of Soret coefficient STS_T is linear with aTR3a^{T}R^{3}. kask_{as} is the prefactor.

ST∼aTR3,FT∼aT∇ΘR3.(30)\begin{aligned} S_T &\sim a^{T}R^{3},\\ F^{T} &\sim a^{T}\nabla\Theta R^{3}. \tag*{(30)} \end{aligned}

A previous experiment[4] showed that the Soret coefficient STS_T grows linearly with the square of the particle size. Therefore, we set the parameter aTa^{T} to aT=αTR−1a^{T}=\alpha^{T}R^{-1}, where αT\alpha^{T} is a physical parameter that does not depend on the particle size.

Nanoparticle driven by Soret force transport in a polymer network

By applying modified conservative force (Eqs. (28)), the Soret force can be successfully applied on the nanoparticle in simple fluid. As a physical problem of great value for fundamental and applied research, the motion of the nanoparticle driven by the Soret force in the polymer network is worth analyzing. As shown in Figure 6, the polymer network with periodic structure is immersed in the fluid with a temperature gradient. The nanoparticle embedded in the polymer network driven by the Soret force. When the Soret coefficient is positive ST>0S_T>0, the Soret force FTF^{T} is opposite to the direction of temperature gradient, thus the nanoparticle move from the hot boundary to the cold boundary.

Nanoparticle driven by the Soret force in a polymer network

Figure 6. Nanoparticle (purple) driven by the Soret force moving in the polymer network (red). The fluid particles (blue) are filled to keep the number density ρ=3\rho=3.

In the polymer network, there are N=8N=8 DPD particles between every two neighbouring junctions, and the average distance between two neighbouring junctions is mesh size ax=2.33a_x=2.33. The adjacent DPD particles are connected by harmonic spring

Eb(rij)=Kb(rijlb−1)2,(31)E_b(r_{ij})=K_b\left(\frac{r_{ij}}{l_b}-1\right)^2, \tag*{(31)}

in which Kb=125K_b=125 stands for the energy of the bond, lb=0.45l_b=0.45 is the equilibrium distance between two DPD particles, thus the counter length of polymer stand is Lc=(N−1)lb=3.15L_c=(N-1)l_b=3.15.

The simulation coefficients are shown in Table I, where the first column indicates the type of particle: ww for water, nn for network, and pp for nanoparticle. Due to the exponential conservative force (Eqs. (29)), no DPD particles exist within the radius RR of the nanoparticle, and only a spherical shell with a thickness of δ\delta contains particles that interact directly with the nanoparticle. According to our previous work[17], the cut-off radius of the dissipative force RCDR_{CD} needs to be larger than that of the conservative force RCCR_{CC} for the interactions between nanoparticles and polymers as well as nanoparticles and fluids. Similarly, we chose a larger cutoff radius of the heat flux RCTR_{CT} than that of the conservative force, to ensure effective heat transfer and Soret force between the nanoparticles and the fluid and network particles. We also selected a large thermal friction κ\kappa coefficient to maintain the temperature gradient undisturbed by the motion of the nanoparticles, and to ensure the thermal equilibrium of the nanoparticles with the local temperature. Volumetric heat capacity Cv=1×108C_v=1\times10^{8} was determined by Eqs. (22), using a characteristic size of L=100nmL=100\mathrm{nm}, a number density of ρ=3\rho=3, and the physical parameters of water at 300K.

pairaija_{ij}γc\gamma^{c}γs\gamma^{s}ssRCCR_{CC}RCDR_{CD}κ\kappaCvC_vRCTR_{CT}aTa^{T}
ww/nn/wn254.500.41.01.010−810^{-8}10810^{8}1.00
wp/npRef.[17]3.55.50.4RRRef.[17]10−810^{-8}10810^{8}R+RsR+R_sEqs.30

Table I. The simulation coefficients

We studied the transport characteristics of nanoparticles in polymer network with different sizes (R=1.3,1.4,1.5R=1.3,1.4,1.5) under different thermophoretic force. When the size of a nanoparticle is larger than the mesh size of the polymer network 2R>ax2R>a_x, the diffusion of the nanoparticle is hop-dominated[33]. Thus, the local confinement that the polymer network exerts on the nanoparticle can be described by a periodic potential energy[17, 18, 33, 34]. The size of the nanoparticles is also chosen to be smaller than the contour length of the strands in the polymer network 2R<Lc2R<L_c, thus the potential energy is dominated by entropy. When driven by a constant force, the nanoparticle diffusion occurs in a tilted periodic potential, also referred to the washboard potential (WBP). In the present work, the driving force is realized by the Soret effect.

From the trajectories of the nanoparticle xn(t)x_n(t), the travel distance is ⟨xn(t)⟩\langle x_n(t)\rangle and the displacement variance is δ⟨xn(t)⟩=⟨xn2(t)⟩−⟨xn(t)⟩2\delta\langle x_n(t)\rangle=\langle x_n^2(t)\rangle-\langle x_n(t)\rangle^2 and can be calculated using the ensemble average. The transport characteristics of foremost interest is the drift velocity and diffusivity in long time stage

vL=lim⁡t→∞v(t)=lim⁡t→∞⟨xn(t)⟩/t,DL=lim⁡t→∞D(t)=lim⁡t→∞δ⟨x2(t)⟩2t.(32)\begin{aligned} v_L &= \lim_{t\to\infty}v(t)=\lim_{t\to\infty}\langle x_n(t)\rangle/t,\\ D_L &= \lim_{t\to\infty}D(t)=\lim_{t\to\infty}\frac{\delta\langle x^2(t)\rangle}{2t}. \tag*{(32)} \end{aligned}

As shown in Figure 7, the normalized drift velocity of nanoparticles confined in polymer networks, vL/v0v_L/v_0, is plotted as a function of the Soret force, FT∼aT∇ΘR3F^T\sim a^T\nabla\Theta R^3. Here, v0v_0 is the drift velocity of nanoparticles without confinement, which is given by

Normalized drift velocity of nanoparticles in polymer networks as a function of the Soret force

Figure 7. Drift velocity of nanoparticles in polymer network is converged to that in simple fluid, as the Soret force increases.

v0=D0FT∼D0aT∇ΘR3v_0=D_0F^T\sim D_0a^T\nabla\Theta R^3

The drift velocity of confined nanoparticles vLv_L increases gradually and approaches the free drift velocity v0v_0 as the Soret force FTF^T increases. When the Soret force is small, the confinement effect lowers the velocity of nanoparticles compared to the unconfined case, that is vL<v0v_L<v_0. This results is qualitatively consistent with previous studies on the diffusion of Brownian particles in a tilted periodic potential UU[35, 36, 37].

The confinement effect is determined by the height of the potential barrier, which depends on the ratio of the nanoparticle and mesh size[17, 33], that is, the confinement parameters 2R/ax2R/a_x. Therefore, the velocity reduction is more significant for larger nanoparticles, due to the higher potential barrier.

As shown in Figure 8, the normalized diffusivity DL/D0D_L/D_0 of nanoparticle diffusion in the polymer network is plotted against the drift velocity vLv_L, which monotonically increases with the Soret force FTF^T. The diffusion of nanoparticles in polymer networks exhibits a non-monotonic behavior as the function vLv_L, except for the nanoparticles with size R=1.3R=1.3, that is, the diffusion coefficient reaches a peak value that is larger than the free diffusion coefficient at a critical thermophoretic force. This phenomenon is known as giant acceleration of diffusion (GAD)[35, 36, 37], which is a non-equilibrium transport phenomenon in a tilted periodic potential. Since the potential energy exerted by the polymer network on the nanoparticles is dominated by entropy, this result numerically confirms that the Brownian particles crossing the entropy barrier can exhibit GAD[37, 38]. For higher entropy barriers, the enhancement of diffusivity is more pronounced for larger nanoparticles, which is qualitatively consistent with the GAD results in tilted periodic potentials[35, 39].

Variation of normalized diffusivity with drift velocity

Figure 8. Variation of normalized diffusivity DL/D0D_L/D_0 with drift velocity vLv_L.

Conclusion

This paper presents a novel extension of dissipative particle dynamics (DPD), called the single-particle energy-conserving dissipative particle dynamics (seDPD) method, which combines the features of single-particle DPD and energy-conserving DPD to simulate the thermophoresis of nanoparticles with a single DPD particle under temperature gradients.

The method accounts for the heat flux arising from the non-central dissipative and random forces, ensuring the conservation of kinetic and internal energy. The validity and reliability of the method and algorithm are verified. The thermophoretic force exerted on the nanoparticles is obtained by modifying the conservative force. Parameterization of the thermophoretic force is established on the basis of the DPD parameters and the nanoparticle size.

The transport of nanoparticles driven by the thermophoretic force in a polymer network is investigated. The phenomenon of giant acceleration of diffusion (GAD) of nanoparticles crossing the entropic barrier is observed. This paper demonstrates a reliable novel DPD method that can simulate systems involving nanoparticles and temperature fields simultaneously and shows numerically that Brownian particles can exhibit GAD when crossing the entropic barrier.

This model is not limited to thermophoresis of nanoparticles and can readily be extended to investigate various system including temperature and nanoparticles.

Acknowledgments

This research is supported by the Natural Science Foundation of China (Nos. 12332016, 11832017 and 12172209).

Data Availability

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

References

  1. [1]F Huang, P Chakraborty, CC Lundstrom, C Holdmen, JJG Glessner, SW Kieffer, and CE Lesher. Isotope fractionation in silicate melts by thermal diffusion. Nature, 464(7287):396–400, 2010.
  2. [2]Christoph J Wienken, Philipp Baaske, Ulrich Rothbauer, Dieter Braun, and Stefan Duhr. Protein-binding assays in biological liquids using microscale thermophoresis. Nature communications, 1(1):100, 2010.
  3. [3]Gerardo Dominguez, Gautam Wilkins, and Mark H Thiemens. The soret effect and isotopic fractionation in high-temperature silicate melts. Nature, 473(7345):70–73, 2011.DOI
  4. [4]Stefan Duhr and Dieter Braun. Thermophoretic depletion follows boltzmann distribution. Phys. Rev. Lett., 96:168301, Apr 2006.pubmed.ncbi.nlm.nih.gov/16712279
  5. [5]Maren Reichl, Mario Herzog, Alexandra Götz, and Dieter Braun. Why charged molecules move across a temperature gradient: The role of electric fields. Phys. Rev. Lett., 112:198101, May 2014.pubmed.ncbi.nlm.nih.gov/24877967
  6. [6]Hao Wang, Shutao Wang, Helen Su, Kuan-Ju Chen, Amanda Lee Armijo, Wei-Yu Lin, Yanju Wang, Jing Sun, Ken-ichiro Kamei, Johannes Czernin, et al. A supramolecular approach for preparation of size-controlled nanoparticles. Angewandte Chemie, 121(24):4408–4412, 2009.
  7. [7]Chundong Xue, Xu Zheng, Kaikai Chen, Yu Tian, and Guoqing Hu. Probing Non-Gaussianity in confined diffusion of nanoparticles. The Journal of Physical Chemistry Letters, 7(3):514–519, 2016.DOI
  8. [8]Chundong Xue, Xinghua Shi, Yu Tian, Xu Zheng, and Guoqing Hu. Diffusion of nanoparticles with activated hopping in crowded polymer solutions. Nano Letters, 20(5):3895–3904, 2020.DOI
  9. [9]PJ Hoogerbrugge and JMVA Koelman. Simulating microscopic hydrodynamic phenomena with dissipative particle dynamics. Europhysics Letters, 19(3):155, 1992.DOI
  10. [10]Yu Lu and Guo-Hui Hu. Linear polymer chain diffusion in semi-flexible polymer network: A dissipative particle dynamics study. Physics of Fluids, 35(1):012007, 01 2023.DOI
  11. [11]S. Kumar Ranjith, B. S. V. Patnaik, and Srikanth Vedantam. Transport of dna in hydrophobic microchannels: a dissipative particle dynamics simulation. Soft Matter, 10:4184–4191, 2014.DOI
  12. [12]E. S. Boek, P. V. Coveney, H. N. W. Lekkerkerker, and P. van der Schoot. Simulating the rheology of dense colloidal suspensions using dissipative particle dynamics. Phys. Rev. E, 55:3124–3133, Mar 1997.DOI
  13. [13]Dmitry A. Fedosov, George Em Karniadakis, and Bruce Caswell. Steady shear rheometry of dissipative particle dynamics models of polymer fluids in reverse Poiseuille flow. The Journal of Chemical Physics, 132(14):144103, 04 2010.DOI
  14. [14]Rossana Pasquino, Hermes Droghetti, Paola Carbone, Shadi Mirzaagha, Nino Grizzuti, and Daniele Marchisio. An experimental rheological phase diagram of a tri-block co-polymer in water validated against dissipative particle dynamics simulations. Soft Matter, 15:1396–1404, 2019.DOI
  15. [15]Yun Hao Feng, Xiao Peng Zhang, Ze Qiang Zhao, and Xin Dong Guo. Dissipative particle dynamics aided design of drug delivery systems: a review. Molecular Pharmaceutics, 17(6):1778–1799, 2020.DOI
  16. [16]Jihao Wang, Yufen Han, Zhiyang Xu, Xiaozhen Yang, Seeram Ramakrishna, and Yong Liu. Dissipative particle dynamics simulation: A review on investigating mesoscale properties of polymer systems. Macromolecular Materials and Engineering, 306(4):2000724, 2021.DOI
  17. [17]Yu Lu and Guo-Hui Hu. A potential barrier in the diffusion of nanoparticles in ordered polymer networks. Soft Matter, 17:6374–6382, 2021.DOI
  18. [18]Yu Lu, Xin-Yue Liu, and Guo-Hui Hu. Double-spring model for nanoparticle diffusion in a polymer network. Macromolecules, 55(11):4548–4556, 2022.DOI
  19. [19]Ziyang Xu, Xiaobin Dai, Xiangyu Bu, Ye Yang, Xuanyu Zhang, Xingkun Man, Xinghua Zhang, Masao Doi, and Li-Tang Yan. Enhanced heterogeneous diffusion of nanoparticles in semiflexible networks. ACS Nano, 15(3):4608–4616, 2021. PMID: 33625839.DOI
  20. [20]Zhen Li, Guo-Hui Hu, Zhi-Liang Wang, Yan-Bao Ma, and Zhe-Wei Zhou. Three dimensional flow structures in a moving droplet on substrate: A dissipative particle dynamics study. Physics of Fluids, 25(7):072103, 07 2013.DOI
  21. [21]Zhen Li, Yu-Hang Tang, Huan Lei, Bruce Caswell, and George Em Karniadakis. Energy-conserving dissipative particle dynamics with temperature-dependent properties. Journal of Computational Physics, 265:113–127, 2014.DOI
  22. [22]W. Pan, I. V. Pivkin, and G. E. Karniadakis. Single-particle hydrodynamics in dpd: a new formulation. Europhysics Letters, 84(1):548–551, 2008.DOI
  23. [23]Kaixuan Zhang, Jie Li, Wei Fang, Chensen Lin, Jiayi Zhao, Zhen Li, Yang Liu, Shuo Chen, Cunjing Lv, and Xi-Qiao Feng. An energy-conservative many-body dissipative particle dynamics model for thermocapillary drop motion. Physics of Fluids, 34(5):052011, 05 2022.DOI
  24. [24]Daniel B. Mayer, Thomas Franosch, Christof Mast, and Dieter Braun. Thermophoresis beyond local thermodynamic equilibrium. Phys. Rev. Lett., 130:168202, Apr 2023.DOI
  25. [25]Seyyed Nader Rasuli and Ramin Golestanian. Soret motion of a charged spherical colloid. Phys. Rev. Lett., 101:08301, Sep 2008.pubmed.ncbi.nlm.nih.gov/18851261
  26. [26]Daniel B. Mayer, Dieter Braun, and Thomas Franosch. Thermophoretic motion of a charged single colloidal particle. Phys. Rev. E, 107:044602, Apr 2023.DOI
  27. [27]Victor Priam ytsyn and Venkat Ganesan. A coarse-grained explicit solvent simulation of rheology of colloidal suspensions. The Journal of Chemical Physics, 122(10):104906, 2005.
  28. [28]Xijuan Fan, Nan Phan-Thien, Shuo Chen, Xuhong Wu, and Yong Ng Teng. Simulating flow of DNA suspension using dissipative particle dynamics. Physics of Fluids, 18(6):861–136, 2006.
  29. [29]JA Becker, C Plowe, HCJ Hoeflsoot, and PD Iedema. Poiseuille flow to measure the viscosity of particle model fluids. The Journal of Chemical Physics, 122(15):154503, 2005.pubmed.ncbi.nlm.nih.gov/15945641
  30. [30]Wenxiao Pan, Bruce Caswell, and George Em Karniadakis. Rheology, microstructure and migration in Brownian colloidal suspensions. Langmuir, 26(1):133–142, 2009.
  31. [31]H Hassanzadeh Afrouzi, Abouzar Moshfegh, Ashkan Javadzadegan, Maryam Mohammadi, and Mousa Farhadi. Systematic fine-tuning of dissipative particle dynamics for mesoscale simulate and dense colloidal suspensions. Physica A: Statistical Mechanics and its Applications, 510:492–506, 2018.
  32. [32]Yi Wang, Jie Ouyang, and Yanggui Li. Parametric study of fluid–solid interaction for single-particle dissipative particle dynamics model. Microfluidics and Nanofluidics, 22(8):78, 2018.
  33. [33]Li-Heng Cai, Sergey Panyukov, and Michael Rubinstein. Hopping diffusion of nanoparticles in polymer matrices. Macromolecules, 48(3):847–862, 2015.pubmed.ncbi.nlm.nih.gov/25691803
  34. [34]Zachary E. Dell and Kenneth S. Schweizer. Theory of localization and activated hopping of nanoparticles in cross-linked networks and entangled polymer melts. Macromolecules, 47(1):405–414, 2013.
  35. [35]P. Reimann, C. Van den Broeck, H. Linke, P. Hänggi, J. M. Rubi, and A. Pérez-Madrid. Giant acceleration of free diffusion by use of tilted periodic potentials. Physical Review Letters, 87:010602, Jun 2001.pubmed.ncbi.nlm.nih.gov/11461454
  36. [36]Peter Reimann and Ralf Eichhorn. Weak disorder strongly improves the selective enhancement of diffusion in a tilted periodic potential. Physical Review Letters, 101:180601, Oct 2008.
  37. [37]Daniel Kim, Clark Bowman, Jackson T Del Bonis-O’Donnell, Anastasios Matzavinos, and Derek Stein. Giant acceleration of DNA diffusion in an array of entropic barriers. Physical Review Letters, 118(4):048002, 2017.
  38. [38]P. S. Burada, G. Schmid, D. Reguera, M. H. Vainstein, J. M. Rubi, and P. Hänggi. Entropic stochastic resonance. Phys. Rev. Lett., 101:130602, Sep 2008.
  39. [39]Claudia Venditti, Alessandra Adrover, and Massimiliano Giona. Inertial effects and long-term transport properties of particle motion in washboard potential. Physica A: Statistical Mechanics and its Applications, 585:126407, 2022.

Paper details

vlnc:W_ncg2775e
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
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