Issue
Natl Sci Open
Volume 3, Number 4, 2024
Special Topic: Active Matter
Article Number 20230069
Number of page(s) 12
Section Physics
DOI https://doi.org/10.1360/nso/20230069
Published online 01 April 2024

© The Author(s) 2024. Published by Science Press and EDP Sciences.

Licence Creative CommonsThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

INTRODUCTION

Active matter systems, consisting of self-propelled units that are able to convert stored or surrounding energy into their persistent motion, provide a fresh opportunity for applications of nonequilibrium statistical mechanics [1-4]. In particular, active colloidal suspension can serve as an active bath that can significantly influence the motion and dynamics of passive objects submerged within them [5-12]. Understanding the behavior of tracer particles in the active bath is a fundamental pursuit in statistical physics and plays a crucial role in various biological[10, 13-16], chemical[17, 18], and physical phenomena[8, 9, 12, 19-32]. Notably, the tracer particle immersed in the active bath exhibits a distinct diffusion profile compared to its equilibrium counterpart[26, 33-39]. Furthermore, studying the diffusion behavior of such a tracer is essential for unraveling the intricate dynamics that govern systems away from thermal equilibrium, providing a valuable tool to investigate the collective behavior [8, 40-46] and transport properties in such media [47-51]. Because of the importance and wide range of applications, understanding the dynamics of the tracer in such an active bath is desirable.

As one already knows, the classical work by Einstein laid the foundation for understanding Brownian motion, providing a framework for diffusive behavior in passive media. Subsequent advancements, such as the Langevin equation, have enriched our understanding of stochastic processes, diffusive phenomena, and the fluctuation-dissipation theorem in thermal equilibrium. Nevertheless, the dynamics of particles in active baths introduce additional complexity[37, 45, 51-57]. Several studies have investigated this question, often employing analytical and numerical techniques to model and characterize the motion of tracer particles within nonequilibrium media, including the active bath. For instance, Maes et al. established a generalized fluctuation-response relation for thermal systems driven out of equilibrium [58-63], utilized this method to investigate the fluctuation-dissipation relation for nonequilibrium bath [35], and further studied the dynamics of a tracer immersed in such bath, gave the friction and noise properties [53], the Langevin description [64], and correlation functions of the tracer variables to study the fluctuation properties[54]. Speck and Seifert et al. formulated a fluctuation-dissipation theorem (FDT) within a nonequilibrium steady state of a sheared colloidal suspension system [65, 66], and subsequently investigated the mobility and diffusivity of a tagged particle within this system, determined the velocity autocorrelation functions and response functions with small shear force, found that a phenomenological effective temperature recovers the Einstein relation in nonequilibrium [67]. Esparza-López et al. [68] proposed a stochastic fluid dynamic model to describe analytically and computationally the dynamics of microscopic particles driven by the motion of surface attached bacteria, analytically calculated expressions for the effective diffusion coefficient through a run-and-tumble model, found that the short-time mean squared displacement is proportional to the square of the swimming speed while the long-time one only depends on the size of the particle. Burkholder and Brady [22] studied the diffusion of a tracer in a dilute dispersion of active Brownian particles (ABPs), by employing the Smoluchowski equation and averaging over bath particles and orientation variables, obtained tracers single-particle probability distribution function, found that the active contribution to the diffusivity scales as U0 (characteristic swim speed of ABP) for strong swimming and U02 for weak swimming. Furthermore, they [69] derived a general relationship between diffusivity and mobility in generic colloidal suspensions, provided a method to quantify deviations from the FDT and express them in terms of an effective SES relation. More recently, Granek et al. [51] studied the long-time dynamics of a tracer immersed in a one-dimensional active bath, derived a time-dependent friction and noise correlation with power law long tails that depend on the symmetry of tracers, and found that shape asymmetry of the tracer induces ratchet effects and leads to super-diffusion and friction that grows with time.

Numerous theories based on various starting points have demonstrated the importance and attraction of studying tracer behavior in nonequilibrium baths. In this study, we propose an alternative theoretical method based on path-integral method to investigate the behavior of tracer diffusion in an active particle bath, and subsequent simulation results successfully validate our theory. The starting point of the theory is the generalized Langevin equation (GLE) for the tracer, which utilizes a generalized version of Dean’s equation to describe the active bath. The GLE contains a memory kernel function and complex effective noise terms, reflecting the complex interactions between the tracer and bath particles. We then employ the path integral method [70, 71] to calculate the diffusion coefficient. Numerical calculations show that the effective diffusion has a non-trivial dependence on bath parameters such as the number density and the persistent time of the active bath particle. Finally, we perform extensive computer simulations, which show very good agreement with our theoretical predictions.

This work is organized as follows: In Section “Model and theory”, we introduce the model system and derive the GLE of the tracer. Then, we utilize the path integral method to obtain the effective diffusion of the tracer formally. In Section “Simulation results”, we show the numerical solution of such diffusion and compare it with simulation results. The paper ends with conclusion in Section “Conclusion”.

MODEL AND THEORY

Active bath model

Considering a system consisting of a tracer particle and N active Ornstein-Uhlenbeck (OU) particles (Figure 1), the coordinates of the tracer x and the active bath particles ri are governed by the overdamped Langevin equations,x˙=μtx[i=1NU(|xri|)+Uext(x)]+2μtTξt,(1a)r˙i=μb{fi+i[jiV(|rirj|)+U(|rix|)]}+2μbTξi,(1b)τbf˙i=fi+2Db/μb2ηi,(1c)wherein μt and μb are the mobilities of the tracer and bath particles respectively, Uext is an external potential acting on the tracer particle, U is the interacting potential between tracer and bath particles, and V is potential between bath particles, ξt, ξi and ηi are white noises with zero means and unit variances, fi is the self-propulsion force acting on bath particle i with time correlation function fi(t)fj(t)=Dbμb2τbδije|tt|/τbI, where I is the unit matrix, τb is the self-propulsion correlation time and Db serves as the amplitude of such force, with the same dimension as diffusivity. We consider the situation that all potentials (U and V) are square integrable, meaning that they have well defined and limited Fourier transforms: Uk=U(r)eikrdr, Vk=V(r)eikrdr.

thumbnail Figure 1

Schematic diagram of a tracer particle with coordinate x in an active particle bath consisting of self-propulsion particles at ri, i=1,2,…, N. The tracer-bath interaction is U(|xri|), and the bath-bath interaction is V(|rirj|). The wave lines label the thermal noises.

Inspired by the idea of Dean’s studies [72, 73], we derive a self-consistent equation for the density profile of bath particles ρ(r,t)=j=1Nδ(rrj(t)) (See details in Appendix A in Supplementary information),tρ(r,t)=μb{ρ(r,t)[ρ(r',t)V(|rr'|)dr'                  +U(|rx|)]}+μbT2ρ(r,t)+{ρ(r,t)[ξT(r,t)+ξA(r,t)]},(2)where the noise terms ξ{T,A} has correlation functions ξT(r,t)ξT(r',t)=2μbTδ(rr')δ(tt)I, ξA(r,t)ξA(r',t)=Dbτbe|tt|/τbδ(rr')I. In the Fourier space (δρk(t)=eirk(ρ(r,t)ρ0)dk), Eq. (4) has a formal solution,δρk(t)=te(ts)Gk{μbk2ρ0Ukeikxs+ρ0ik[ξ˜kT(s)+ξ˜kA(s)]}ds,(3)wherein Gk=μbk2(T+ρ0Vk) can be considered as a characteristic frequency (a typical illustration of this term is shown in Figure S1A). Noise terms ξ˜k{T,A}(t) are the Fourier transform of ξ{T,A}, respectively, with correlation functionsξ˜kT(t)ξ˜kT(t)=2μbT(2)3δ(k+k')δ(tt)I,(4a)ξ˜kA(t)ξ˜kA(t)=Dbτb(2)3δ(k+k')e|tt|/τbI,(4b)wherein means the average over noises.

Generalized Langevin equation

To achieve an effective equation of motion of the tracer which does not contain any bath particle variables, one can use an identity xi=1NU(|rix|)=1(2π)3ikeikxρkUkd3k, then has a GLE for x,x˙(t)=μtxUext(x)+μt1(2π)3ikeikx(t)δρkUkd3k+2μtTξt=μtxUext(x)tK(s,t)ds+ηT(t)+ηA(t)2μtTξt,(5)where the memory kernelK(s,t)=μtμbρ0(2π)3ik2kUk2e(ts)Gkeik(x(t)x(s))d3k=μtμbρ0(2π)3Im[eik(x(t)x(s))]kk2Uk2e(ts)Gkd3k,(6)and the colored noise terms,ηT(t)=μt1(2π)3keikxtρ0Ukte(ts)Gkkξ˜kT(s)dsd3k,(7)ηA(t)=μt1(2π)3keikxtρ0Ukte(ts)Gkkξ˜kA(s)dsd3k,(8)and their correlations areGT(t,t)=ηT(t)ηT(t)=μt2ρ0μbT(2π)3eik(xtxt)kkk2Uk2e|tt|GkGkd3k,(9)GA(t,t)=ηA(t)ηA(t)=μt2ρ0Db(2π)3eik(xtxt)kkk2Uk2τbe|tt|/τb(e|tt|Gk/Gk)[(τbGk)21]d3k.(10)

Path integral and effective diffusion

To calculate the transport coefficients, one needs to calculate several correlation functions first. Considering the coupling between the tracer position and the colored noises ηA,T, we propose a path integral method to calculate them.

We consider a path of the tracer in the time interval [ti, tf]. The partition function of such trajectory can be written asZ=tδ{x˙+μtUext+tK(s,t)dsηAηT2μtTξt}P[ξt]P[ηT]P[ηA]DxDξtDηADηT.(11)Using the identity of delta function, δ(x)=1(2π)eipxdp, we also haveZ=eip{x˙+μtUext+tK(s,t)dsηAηT2μtTξt}dtP[ξt]P[ηT]P[ηA]DxDpDξtDηADηT,(12)where p is an auxiliary real vector field. Next, utilizing eau=e12a2u2 for a Gaussian random variable u with zero mean, we have the partition function as a function of the action,Z=eS(x,p)DxDp,S(x,p)=S0(x,p)+Sint(x,p),(13)whereS0=ip(t)[x˙(t)+μtxUext]dt+μtT|p(t)|2dt,(14)for a free particle andSint=ip(t)[tK(s,t)ds]dt+12p(t)[GT(t,s)+GA(t,s)]p(s)dtds(15)counts for the tracer-bath interaction. Herein, we have assumed that ηA,T are both Gaussian, and their deviations from the Gaussian distribution are only weakly present in regions far from the mean. Since this approximation primarily reflects the properties of the second moments, it is reasonable in this context. In addition, since we have used the perturbative expansion, the interaction between tracer and bath particles should be weak. Therefore, we choose soft harmonic potentials as the inter-particle potentials. The perturbation also demands that the activity is not very large, which constitutes a condition for the application of our theoretical framework.

After introducing the partition function over the trajectory, the average over any operator A as function of x(t), t[ti,tj] can be defined asA=Ae(S0+Sint)DxDpZ=AeSint0eSint0,(16)where 0=Z01()eS0DxDp corresponds to a tracer particle only affected by external potential, not any particle bath. So far, the eSint term is still too complicated to handle. A common treatment is the linear truncation when Sint is weak. Herein, we treat the tracer-bath interaction U as a small perturbative quantity by assigning U(r)=ϵu(r), where ϵ is a dimensionless factor that scales the interacting strength, and can be used to the following perturbative expansion. The memory kernel k and correlators GA,T(t) are order 1 of ϵ. Therefore, Eq. (19) can be expand asA=A0ASint0+A0Sint0+O(ϵ2).(17)For studying the diffusion problem, we only need to consider the free particle situation, Uext=0. According to the symmetry, we have p(t)eik[x(t)x(s)]0=ip(t)[x(t)x(s)]0kek2μtT(ts)=0. Similarly, we also havep(t)GA,T(t,s)p(t)0{|p(t)|20(k(x(t)x(s))p(t)0)2}ek2μtT(ts)=0.(18)Therefore,Sint0=0.(19)

The next step is to calculate the mean square displacement (MSD), i.e., to calculate [x(tf)x(ti)]2 for a long time interval tf-ti. Then, the effective diffusion coefficient can be given asDeff=lim(tfti)12d(tfti)[x(tf)x(ti)]2,(20)where d is the dimension of the system. According to Eq. (20), the key step of calculating the MSD is handling [x(tf)x(ti)]2Sint0, i.e., calculating the following two correlation functions, one is[x(tf)x(ti)]2p(t)eik[x(t)x(s)]0=2ik[x(t)x(s)][x(tf)x(ti)]0[x(tf)x(ti)]p(t)0ek2μtT(ts)=4kμtTL([ti,tf][s,t])χ[ti,tf)(t)ek2μtT(ts)=4kμtT[tmax(ti,s)]χ[ti,tf)(t)ek2μtT(ts),(21)and the other is[x(tf)x(ti)]2p(t)p(s)eik[x(t)x(s)]0={4μtT[tmax(s,ti)]kk2χ[ti,tf)(s)1}ek2μtT(ts)χ[ti,tf)(t),(22)for t>s (the omitted mathematical details can be found in Appendix B in Supplementary information). After the calculation in Eq. (S9) (Supplementary information), we get[x(tf)x(ti)]2Sint0=2μtT(tfti)μtμbρ0(2)3k4Uk2Gk(k2μtT+Gk)d3k+μt2ρ0Db(2π)3(tfti)2k4Uk2(τbGk)21{τbμtTk21(μtTk2+τb1)2(μtTk2/Gk)1(μtTk2+Gk)2}d3k+o(tfti),(23)where Gk=μbk2(T+ρ0Vk) reflects the bath properties including background temperature T, number density ρ0 and interactions between bath particles. At last, taking the long-time limit of tf-ti, the o(tf-ti) term can be neglected, the effective diffusion coefficient of a tracer in an active bath is obtainedDeff=μtT{10gT(k)dk+0gA(k)dk},(24a)to the linear order, whereingT(k)=μtμbρ062k6Uk2Gk(k2μtT+Gk),(24b)gA(k)=μtρ0Db/T62k6Uk2(τbGk)21[μtTk2/Gk1(μtTk2+Gk)2τbμtTk21(μtTk2+τb1)2].(24c)This is the main result of the present work. In this equation, “1” in the brace of Eq. (24a) denotes the bare diffusion of a free tracer particle. The second term in the brace denotes the “passive part” of the tracer-bath interaction which is always a negative contribution to effective diffusion and recovers the results in Ref.[70]. In the absence of activity, the FDT holds since the effective mobility of the tracer satisfies μeff=Deff/T [70], wherein they have obtained an effective mobility coefficient μeff=μt(10gT(k)dk) (expressed with notations of the present work for convenience). The third term is a pure “active” contribution on the diffusion, which is a positive contribution and explicitly gives the nontrivial dependence of Deff on the bath parameters, ρ0, Db, τb and interactions U(r) and V(r). For this linear truncation, Deff is a linear function of Db. However, the dependence of τb and ρ0 (which is also contained in Gk) is illegible, which require numerical calculations to determine (see Section “Simulation results”). Further mathematical analysis of this expression is shown in Appendix C in Supplementary information.

SIMULATION RESULTS

In this section, we show numerical calculations of Eq. (24a) with the persistent time of active force τb and the number density of bath particles ρ, at small activity Db region. Then, by comparing these results with computer simulations, the validity and applicability of the theory can be verified.

In the present work, we choose V(r) and U(r) as both harmonic potentials, V(r)=κb2(rσbb)2 for r<σbb, and U(r)=κt2(rσtb)2 for r<σtb. We set σbb as the unit of length, κbσbb2 as the unit of energy, and 1/(μbκb) as the unit of time. The common parameters are set as: σtb=2.0, κt=1.0, μt=0.333, T=1.0. The other parameters are set as variables which are explicitly given in the following figures. In computer simulations, we construct a three-dimensional system with periodic boundary containing (1+4095) particles. The diffusion coefficient is calculated through a long-time simulation (~108 steps with 10-3 as the time step) and averaged over 20 samples with random initial configurations.

Firstly, we focus on the contribution of bath number density ρ on Deff, shown in Figure 2. In Figures 2 and 图3, dots and corresponding error bars are the direct simulations, and lines are numerical calculation of Eq. (24a). In general, Deff shows a diverse dependence on ρ. For small active force amplitude Db, Deff decreases with ρ as the tracer’s behavior in a passive particle bath. For large activity situation, Deff shows the opposite behavior. Consequently, one may expect that there is a moderate activity region that Deff has a non-monotonic dependence on ρ.

thumbnail Figure 2

The dependence of the effective diffusion on the number density of active bath particle ρ. Dots and corresponding error bars are simulation results of Deff. Solid lines are the numerical calculations of Deff. Both results show that, with the increase of activity Db, effective diffusion Deff gradually changes from monotonic decrease to monotonic increase with the number density of bath particle ρ, including a non-monotonic interval around Db=2.0. Herein, τb is set as 0.1.

thumbnail Figure 3

The dependence of the effective diffusion on the active force persistent time τb. Dots and corresponding error bars are simulation results of Deff. Solid lines are the numerical calculations of Deff. Both show that Deff monotonically decreases with the persistent time τb. The horizontal dash line denotes the bare diffusion coefficient of the tracer particle.

We then investigate the dependence of Deff on persistent time τb. According to Eq. (27), Deff decreases with τb since both (Gkτb)21 and τbμtTk21(μtTk2+τb1)2 increase with τb monotonically. This prediction has been confirmed in simulations, as shown in Figure 3. Physically, when the persistent time of the active force tends to zero, the active OU particle can be reduced to an ordinary Brownian particle under a higher temperature. If the activity amplitude Db is constant, the longer τb means a larger deviation of equilibrium. Based on the results here, we might conclude that an active OU particle bath that stays closer to equilibrium, is more conducive to the tracer diffusion, when the activity amplitude is given.

CONCLUSION

This study aims to shed light on the intricacies of tracer diffusion in an active particle bath. By employing the generalized Dean’s equation, incorporating the path integral method, and utilizing computer simulations, we characterize the impact of self-propulsion on the diffusion behavior of a passive tracer particle. In summary, the effective diffusion decreases with persistent time τb, and exhibits a variety of dependencies on bath density, depending on Db. The obtained insights expand our understanding of collective dynamics and transport phenomena in non-equilibrium systems, with potential applications in diverse scientific disciplines. For further studies, an extension of the active bath situation is straightforward since it has been confirmed to calculate the mobility of a tracer in particle bath [70, 71]. Additionally, after the effective mobility is achieved, the fluctuation-dissipation theorem can be further investigated to determine its validity or deviations with respect to the activity parameters.

Funding

This work was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (XDB0450402), and the National Natural Science Foundation of China (32090040 and 22373090).

Author contributions

H.Z. directed the project. F.M. derived the theory, wrote codes, and wrote the manuscript. All authors commented on the manuscript.

Conflict of interest

The authors declare no conflict of interest.

Supplementary information

The supporting information is available online at https://doi.org/10.1360/nso/20220069. The supporting materials are published as submitted, without typesetting or editing. The responsibility for scientific accuracy and content remains entirely with the authors.

References

All Figures

thumbnail Figure 1

Schematic diagram of a tracer particle with coordinate x in an active particle bath consisting of self-propulsion particles at ri, i=1,2,…, N. The tracer-bath interaction is U(|xri|), and the bath-bath interaction is V(|rirj|). The wave lines label the thermal noises.

In the text
thumbnail Figure 2

The dependence of the effective diffusion on the number density of active bath particle ρ. Dots and corresponding error bars are simulation results of Deff. Solid lines are the numerical calculations of Deff. Both results show that, with the increase of activity Db, effective diffusion Deff gradually changes from monotonic decrease to monotonic increase with the number density of bath particle ρ, including a non-monotonic interval around Db=2.0. Herein, τb is set as 0.1.

In the text
thumbnail Figure 3

The dependence of the effective diffusion on the active force persistent time τb. Dots and corresponding error bars are simulation results of Deff. Solid lines are the numerical calculations of Deff. Both show that Deff monotonically decreases with the persistent time τb. The horizontal dash line denotes the bare diffusion coefficient of the tracer particle.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.