Phase Segregation of Passive Advective Particles in An Active Medium
Phase Segregation of Passive Advective Particles in An Active Medium
Phase Segregation of Passive Advective Particles in An Active Medium
have diffuse interfaces and low interfacial tension. The phase segregated steady state shows strong
macroscopic fluctuations characterized by multiscaling and intermittency, signifying rapid reorga-
nization of macroscopic structures. We expect these unique nonequilibrium features to manifest in
the actin-dependent molecular clustering at the cell surface.
Collections of active particles exhibit large concentra- surface context [18, 19]. Second, the domain coarsening
tion fluctuations, leading in many cases to clumping and dynamics shows strong departures from Porod behaviour
segregation into high and low density phases [1, 2] - the [17], which implies that the interfacial widths are diffuse
dynamics towards segregation and the space-time fluc- with low interfacial tension. Third, the active segregated
tuations of the order parameter, in single [37] and two phase shows strong fluctuations of the integrated order
component [8, 9] active systems, are found to be very parameter and domain size, characterised by broad non-
different from their equilibrium counterparts. Gaussian distributions, in contrast to the sharply peaked
Recent studies on actomyosin-dependent molecular distribution in closed systems. These strong fluctuations
clustering at the cell surface [1014] motivate us to ex- manifest as multiscaling and intermittency in the steady
plore a novel kind of phase segregation driven by activ- state, features seen in many driven nonequilibrium sys-
ity. These studies suggest that the cortical layer of acto- tems such as turbulence [20]. As a result, the steady state
myosin generates active stochastic stresses and currents exhibits a continual breakup and reformation of macro-
at the cell membrane, driving local clustering of several scopically large structures. This fluctuation dominated
membrane components [10, 12, 14]. The spatiotemporal phase ordering (FDPO) is similar to those observed in a
statistics of such clustering, driven by an active noise, variety of semiautonomously coupled dynamical systems,
is well described using an active hydrodynamics frame- such as passive sliders on a fluctuating surface [21], Burg-
work [14], where membrane components that (un)bind to ers fluids [22] and active nematics [23].
actomyosin are driven by the active filaments - we will Dynamics of active fluids - asters as emergent struc-
call these components passive advective scalars or simply tures The basic hydrodynamic equations describing the
passive particles [14, 15]. The simultaneous presence of a polar contractile active fluid in terms of the active fila-
dense stationary actin meshwork at the cortex [16] pro- ment concentration c, the polar orientation n, the apolar
vides a momentum sink to the dynamics of the cell mem- nematic orientation Q and the hydrodynamic velocity
brane components. Membrane curvature effects appear field v are described in recent reviews [2]. The precise
unimportant, for such clustering occurs predominantly in form of the equations relevant to the present context are
the flat regions of the cell surface [10]. described in [14, 24]. Many studies, including [14, 24]
Here we study the potential scale-dependent segrega- have shown that the homogeneous, orientationally disor-
tion of passive molecules from inert molecules (which do dered configuration of active polar filaments is unstable
not bind to cortical actin) at the cell surface. Our model to clumping and local orientation order to form spatially
system consists of passive and inert particles embedded localized contractile regions these emergent structures
in an active medium in two dimensions (2d). While we [25, 26] have been referred to as asters [14, 24].
observe both micro and macro phase segregation upon Introducing noise parametrised by an active tempera-
varying the active noise strength and active filament con- ture Ta , results in a stochastic breakup and re-formation
centration in the medium, the segregation dynamics and of asters (remodeling), with an exponential distribution
the nature of the active phase segregated state are dra- of lifetimes [14, 15, 24] whose mean r , is set by Ta . This
matically different from conventional phase segregation leads to a picture of a finite density of contractile struc-
[17]. First, the active phase segregation is driven by ac- tures or asters, whose spatiotemporal fluctuations corre-
tive advection and not by gradients of chemical potential, spond to birth-death processes [15].
and so occurs at temperatures higher than the equilib- Passive particles that (un)bind to the filaments are
rium segregation, a feature that is relevant to the cell driven by these contractile flows; when bound they are
2
(a) (b)
filaments with rates kb (ku ), get advected with speed v0
and timescale v = R /v0 in a direction pointing radi-
ally inwards, when the filaments are part of an aster.
(c) (d) Outside the aster, the passive particle moves diffusively,
with a diffusion coefficient set by v0 and the orientational
decorrelation time of the filaments.
The kinetic Monte-Carlo (MC) dynamics [27] involve
(i) Kawasaki exchanges [28, 29] of components A and
B which obey detailed-balance, (ii) detailed-balance vio-
lating active advective updates whenever A moves into
FIG. 1. (a-b) Schematic of a binary system of passive A (blue)
and inert B (chrome) particles in an active medium. (a) The
an aster and (iii) detailed-balance violating stochastic
active medium consists of polar active filaments (red arrows) birth-death updates of asters with a lifetime distribu-
and asters (circled), within the asters the arrows point in- tion, exp[t/r ] [30]. One MC step is N -attempts at
wards, while outside the asters they are orientationally un- exchange, N being the total number of particles in the
correlated. (b) The active medium consists of actin filaments system. We have converted all time, length and energy
(red) which form a mesh of actomyosin contractile platforms units to s, m and kB T , respectively [27]. We explore the
(circled), which draw in A particles that are bound to it. (c) dynamics as a function of particle concentration xA and
PDF of the order parameter for = 0.128 and different
values of A at a fixed temperature T /Tc = 1.17, where the
the active parameters characterizing the medium, namely
equilibrium (Eqm) P () shows a peak at = 0. The bimodal = r1 /d1 , the ratio of the aster remodeling rate to
distribution for > 0 indicates phase segregation, even at the particle diffusion rate (d is the diffusion time of par-
T > Tc . (d) Phase diagram in (T, A , xA ) at = 0.128, ticles over the scale of an aster), aster area fraction A
shows micro (blue) and macro (red) phase segregation, when and the Peclet number Pe v1 /d1 . We fix Pe 187
T > Tc . and duty ratio K kb / (kb + ku ) = 1 [27].
We find that the steady state distribution is very sen-
sitive to both active parameters, and A . At = 0,
advected with a current J cn; when unbound, they the steady state is equivalent to particles diffusing in a
move diffusively. Thus the formation of contractile asters quenched random media of stationary asters, character-
drive the localised clustering of passive particles (Fig. ized by an equilibrium distribution. At the other ex-
1a,b). The stochastic remodeling of asters give rise to treme, 1, the asters remodel much faster than the
dynamical fluctuations of clusters of passive particles ex- particle movement, the steady state is essentially that of
plored in [14]. Here, we study the dynamics of phase equilibrium diffusing particles. It is only for 0 < 1,
segregation of passive-inert particles moving in an ac- when the particles move fast compared to aster remodel-
tive medium consisting of a finite density of remodeling ing, that we get a genuine nonequilibrium steady state,
asters - however, rather than using the continuum hy- where clusters form, fragment and move towards newly
drodynamic equations [14, 24], we work with an effective formed asters.
theory that incorporates all its relevant features. Phase diagram We first obtain the equilibrium phase
Effective dynamics using kinetic Monte Carlo Con- diagram in the (T xA ) plane and the critical tempera-
sider a collection of contractile platforms, represented as ture kB Tc /J = 1.43, by monitoring the probability dis-
discs of diameter R , distributed uniformly with area tribution function (PDF) P () of the coarse-grained seg-
B
fraction A , in a 2d system of size L. In one model of the regation order parameter, = A A +B
, where A , B are
active fluid medium, the filaments are distributed uni- the local densities of A and B, respectively. The order
formly and point radially inward within the aster, while parameter is coarse-grained over a scale 3 which is
they are oriented randomly outside the asters (Fig. 1a). smaller than the typical domain size [27, 28]. The phase
Alternatively, our model could describe a situation where boundary is determined by the appearance of a distinct
the active medium consists of patches of contractile ac- double peak in P () at = 1.
tomyosin mesh (Fig. 1b). In either case, the contrac- We next determine the nonequilbrium phase diagram,
tile platforms (e.g., asters) remodel with lifetimes chosen as a function of (T, A , xA ) at different values of . Un-
from an exponential distribution exp[t/r ]. In this dy- like the equilibrium case, phase segregation here is driven
namical background, consider a two component system by the active advection of one of the components into
of passive particles A (at concentration, xA ) and inert asters, rather than by gradients in chemical potential,
particles B (concentration, xB = 1 xA ). The particles which implies that segregation can occur even at T > Tc .
interact via a Lennard-Jones (LJ) short-range repulsion; Its extent depends on A - for small A only micro-phase
in addition, A A and B B have a mutual attractive segregation takes place (the domain sizes are roughly the
potential, u(r) = J (/r)6 ,where J is in units of tem- aster size); as A increases to a finite fraction of the sys-
perature kB T and is the LJ-length scale. tem size, we start getting macro-phase segregation. Fig-
The passive particles that bind/unbind to the active ure 1c,d indicates that the nature of segregation depends
3
(a) 1 (b) 16
t = 500
1000
k 3 , known as Porods law [17], signifying that the in-
2000
12 terface separating the growing phases is sharp. Here, in
0.5 3000
g(r, t)
R(t)
4000 24
5000 0.25
6000 8 12
0.33
where depends on and lies in the range 0 < 1
0 7000
8000 6
101 100 101
(Fig. 2d). This departure from Porod signifies that the
4
0 1 2 3 4 0 2 4 6 8 10 interfaces are diffuse with very low interfacial tension,
r/R(t) t 104 the interfacial width being ` R1 (while still having a
(d) 1.2 well-defined domain, `/R 1 for large R). This also
1
(c) 10 t = 500
1000
2000 implies that g(r/R(t)) has a cusp nonanalyticity at small
S(k, t)/R(t)2
3000 0.8
4000 r/R(t), going as g(y) 1 y () + . . ..
102 5000
6000 0.4 Fluctuation dominated phase ordering The most dra-
(2 + ) 7000
8000 matic aspect of the active segregated steady state is in the
0
103
103 102 101 100
nature of fluctuations of a variety of statistical quantities.
100 101
kR(t) Supplemental Movies [27] show that the dynamics of -
profile is characterized by strong fluctuations both at the
FIG. 2. Active coarsening dynamics, shown here for = interface and bulk, unlike the equilibrium situation where
0.128, A = 0.3 and T /Tc = 1.17 : (a) Dynamical scaling of the fluctuations are appreciable only at the interface. To
the order parameter correlation function g(r, t) with domain get a quantitative handle, we study theRtime series of R,
size R(t). The finite intercept of g(r, t) at the scaled r/R(t)
0 is an indication of long-range order. (b) R(t) grows with
the region-wise order parameter, = (r) dr (where
time (red) and saturates to a value that depends on A and the integration is done over a large enough region of size
[27], unlike the equilibrium phase segregation at T < Tc (+ L), and the fourier amplitude of at low k (Fig. S5-
symbols), where the domain size keeps growing as t1/3 (black S7), (1) |(k = 2/L)|, (2) |(k = 4/L)|,
symbols in inset) until it reaches system size. In contrast, (3) |(k = 6/L)|, . . .. We find that the correspond-
for active phase segregation, R(t) t1/4 before saturating ing PDFs at steady state, show large variance and heavy
(purple line), here = 0.128, A = 0.6. (c) Structure factor non-Gaussian tails (Fig. 3, Fig.h S5). Figure 3d shows
S(k, t) also exhibits dynamical scaling but a violation of Porod
i
2
that the PDF P () 1 exp A() (/) , ex-
behaviour - S(kR(t)) k(2+) at large k, where 0 < 1.
(d) varies non-monotonically with , showing equilibrium pressed in the large deviation function form [31], where
behaviour (represented by the dashed line) at = 0 and = hi. These observations are consistent with
1. other driven non-equilibrium systems, such as passive
sliders on fluctuating interfaces [21] and Burgers fluids
[22].
on the values of and A . The time series of various statistical quantities X(t) al-
Coarsening dynamics We next study the dynamics ternates between periods of quiescence and large changes
of segregation following a quench from the high temper- over very short time scales (Fig. 4a), indicating a break-
ature mixed phase to a region between the active and down of self-similarity due to intermittency. We char-
equilibrium phase boundaries (red and blue shaded re- acterize this behaviour in terms of the n-point structure
n
gions in Fig. 1d) - a situation relevant to the cell mem- functions Sn = h(X(t) X(0)) i, where h. . .i is an aver-
brane, where equilibrium lipid phase segregation has only age over histories, as in studies of turbulent systems [20].
been observed at temperatures much lower than physi- Here we compute such structure functions for R(t).
ological temperatures [18, 19]. We study the dynam- These structure functions are expected to scale with
ics of the order parameter correlation function g(r, t) = time as Sn tnn (Fig. S8). In equilibrium phase segre-
h(0, t)(r, t)i as the system coarsens. As in conventional gation (away from critical points), the statistical quanti-
phase ordering [17], g(r, t) appears to obey dynamical ties are self-similar, implying n = const - called gap scal-
scaling (Fig. 2a, Fig. S2), g(r, t) = g(r/R(t)), where the ing (Fig. S8). In the active phase segregation, n is a non-
domain size R(t) is obtained from the first-zero of g(r, t). linear function of n - a phenomenon known as multiscal-
The mean domain size R(t) grows before saturating to a ing (Fig. S8). This breakdown of self-similarity is a conse-
value that depends on and A - a power law fit yields quence of intermittency (which is dominated by extreme
(Fig. 2b, Fig. S3), R(t) t1/4 , different from the conven- events), as seen by the behaviour of flatness or kurto-
tional phase ordering growth of t1/3 [17]. Note that the sis, (t) = S4 (t)/S22 (t) (Fig. 4b). Intermittence shows up
maximal domain size is much smaller than the equilib- as a divergence in as t/ 0 (Fig. 4b), where is a
rium case, even though it increases with increasing sys- time scale which characterizes the lifetime of the largest
tem size L (with A held fixed). The fourier transform structures in the system [34]. We argue that scales
of g(r, t), the structure factor S(k, t) also exhibits scal- as the remodeling time r for large r and is set by d
ing S(k, t) = [R(t)]2 S(kR(t)) (Fig. 2c, Fig. S4). In con- for small
r - this
suggests the following nonlinear form,
ventional phase segregation, S(kR(t)) at large k goes as = r 2 + 02 , where 0 is a numerical fit parameter
4
102
(a) +1
(b) [32] (Fig. 4c). The onset of divergence appears earlier for
A = 0.30 +0.5
100 1
Eqm
40 active > 1 (Fig. 4c). This is in agreement with our earlier
(t)
101
Eqm (T < Tc ) 0 assertion that these limiting cases correspond to steady
102
103 102 101 100 101 102 50 52 54 56 58 60 states of an equilibrium model.
t 104
Non-Porod behaviour, non-Gaussian distributions,
100
= 0.3L = 0.128 A = 0.6 = 0.3L multiscaling and intermittency are signature features of a
A = 0.3 0.320 = 0.128 0.4L
0.4
0.640 0.5L driven nonequilibrium steady state called the fluctuation-
P () 0.5
101 Eqm 1.280 0.6L
P ()
(t)
101
state (Fig. 4d).
A = 0.3, = 0.320 We have now extended this study to temperatures
8 A = 0.6, = 0.128 Eqm
100 T < Tc , where the system can undergo an equilibrium
2 4 6 8 10 100 101 102
phase segregation in the absence of active driving. Here
t 104 t
(c) 103
A = 0.6 = 0.192 (d) 100 again, we find that turning on activity results in a break-
10 2
0.256
0.320 FD down of conventional segregation dynamics - R(t) grows
and saturates to a value determined by A and and the
0.384 101
P
(t)t2
0.448
O
1
10