Eur. Phys. J.

E (2016) 39: 22
DOI 10.1140/epje/i2016-16022-9
THE EUROPEAN
PHYSICAL JOURNAL E
Regular Article

Solute based Lagrangian scheme in modeling the drying process

of soft matter solutions
Fanlong Meng1,2,3,a , Ling Luo3 , Masao Doi3,b , and Zhongcan Ouyang1,2,4
1
State Key Laboratory of Theoretical Physics, Institute of Theoretical Physics, Chinese Academy of Sciences, Beijing 100190,
China
2
Kavli Institute for Theoretical Physics China, Beijing 100190, China
3
Center of Soft Matter Physics and its Applications, Beihang University, Beijing, China
4
Center for Advanced Study, Tsinghua University, Beijing 100084, China

Received 17 July 2015 and Received in final form 15 August 2015

Published online: 26 February 2016 – 
c EDP Sciences / Società Italiana di Fisica / Springer-Verlag 2016

Abstract. We develop a new dynamical model to study the drying process of a droplet of soft matter
solutions. The model includes the processes of solute diffusion, gel-layer formation and cavity creation. A
new scheme is proposed to handle the diffusion dynamics taking place in such processes. In this scheme,
the dynamics is described by the motion of material points taken on solute. It is convenient to apply this
scheme to solve problems that involve moving boundaries and phase changes. As an example, we show
results of a numerical calculation for a drying spherical droplet, and discuss how initial concentration and
evaporation rate affect the structural evolution of the droplet.

1 introduction soft matter solutions. In this scheme, we describe the dif-

The drying of soft matter solutions (colloidal solutions or
polymer solutions) is a complex process since it involves
several physico-chemical processes, such as solvent evapo-
ration, solute diffusion, formation of gel layers [1–4], buck-
ling and wrinkling of the gel layer [5–8], and creation of
cavities or bubbles [9, 10], etc. Understanding these pro-
cesses is important in many industrial processes such as
coating, printing, particle making, etc.
Attempts have been made to simulate the drying pro-
cess by simultaneously solving the transport equations for
mass and energy. However, previous works have been lim-
ited to solving the transport equations in a single phase,
for example to solving the diffusion equation for solute
under boundaries which is moving due to solvent evapo-
ration [11–14]. While such simulations have given us in-
sights in understanding the correlation between drying
conditions and structural changes of the final products,
they are not complete since the complex phase change
processes, such as gelation, solidification, and cavitation
are not taken into account in their model.
In this paper, we propose a new scheme to handle the
complex diffusion processes taking place in the drying of

Contribution to the Topical Issue “Wetting and Drying:
Physics and Pattern Formation”, edited by Duyang Zang,
Ludovic Pauchard and Wei Shen.
a
e-mail: flm@itp.ac.cn
b
e-mail: masao.doi@buaa.edu.cn
fusion by the motion of material points taken on solute.
In the simplest case (i.e., diffusion in a single phase), the
scheme gives the same answer as that of the conventional
diffusion equation. The new scheme has advantages that
1) the phase change (such as the sol-gel transition) is easily
included in the model, and that 2) the moving boundaries
(such as the surface of the solution or the boundary be-
tween fluid phase and gel phase) are naturally and easily
handled.
This paper is organized as follows. In sect. 2, we will
give an overview of the phenomena seen in the drying
of a droplet of soft matter solutions. In sect. 3, we will
describe our new scheme, and discuss the diffusion process
in solution phase. In sects. 4 and 5, we will discuss how to
include the gels phase and cavities in our model. in sect. 6,
we will show an example of our simulation, and in sect. 7,
we will conclude.
2 Drying behavior of a droplet of soft matter
solutions
We consider a spherical droplet of a soft matter solution
made of non-volatile solute and volatile solvent. The solute
can be polymers or colloids, but the subsequent model is
developed with colloidal solutions in mind.
Figure 1 shows a schematic picture of a drying droplet
of such a solution. Initially the solute concentration in the
droplet is uniform (fig. 1(a)). As solvent evaporates at the
Page 2 of 10
rd Rd(t)
r R(r,t)
(a) (b)
Fig. 1. A drying process of a droplet of a soft matter solution. (a) Initial state, (b) diffusion in solution, (c) formation of a gel
layer, and (d) cavity formation.
surface, the solute concentration at the surface increases
(fig. 1(b)), and the solute starts to diffuse towards the cen-
ter of the droplet. When the solute concentration at the
surface becomes high, the surface loses fluidity and forms
a gel layer (fig. 1(c)). With further evaporation of solvent,
various instabilities can take place [5–10]. Here we consider
cavitation: a cavity appears inside the droplet and grows
(fig. 1(d)). This creates a hollow particle in the end. In the
following we will discuss such structural changes. For sim-
plicity, we assume that spherical symmetry is maintained
throughout the whole drying process.
We denote the droplet radius at time t by Rd (t). The
time derivative of Rd (t) is equal to the evaporation rate J,
which is defined by the solvent volume evaporating from
the droplet surface per unit time per unit area,
Ṙd (t) = −J,
J depends on various factors, ambient vapor pressure,
droplet radius, solute concentration, etc., but here, for the
sake of simplicity, we assume that J is a given constant
independent of time.
We denote the solute concentration by the volume frac-
tion of solute. Let φ(R, t) be the volume fraction at a point
R (which denote the distance from the center) at time t.
Initially, φ(R, t) is a constant φ0 , independent of position
φ(R, 0) = φ0 .
Since the solvent is taken out at the surface, the solute
concentration at the surface increases faster than in bulk,
and solute starts to diffuse towards the center. The time
evolution of φ(R, t) is described by the diffusion equation
 
∂φ(R, t) 1 ∂ 2 ∂φ(R, t)
= 2 R D(φ) , (3)
∂t R ∂R ∂R
where D(φ) is the diffusion coefficient, which can be a
function of φ. Since the solute cannot evaporate, the con-
servation equation for solute gives the following boundary
condition:
∂φ
D = φJ at R = Rd (t). (4)
∂R
The above set of eqs. (1)–(4) determines the evolution
of the system when the entire droplet is a solution. The
evolution equation will change as time goes on since gel
Eur. Phys. J. E (2016) 39: 22
gel solution cavity
(c) (d)
layer and cavity appear. In the following, we shall discuss
such a dynamics using a new scheme which we call solute
based Lagrangian scheme.
3 Dynamics in solution phase
We first explain the solute based Lagrangian scheme by
using a simple example in which the whole system is a
solution.
Suppose that we color a set of solute particles located
in a small region at r at time t = 0. The solute particles
will spread in time, but their average position R can be
defined, and is written as a function of r and t. The diffu-
(1) sion dynamics of solute can be described by this function
R(r, t). This is the solute based Lagrangian scheme in the
diffusion dynamics.
Based on the least energy dissipation principle [15] to
determine the steady state in linear systems such as vis-
cous flow in Newtonian fluid by L. Rayleigh in 1873, L.
Onsager proposed a variation principle [16,17] to calculate
the time evolution equation of non-reversible systems, by
taking account of the entropy change of the system. This
variation principle was recently reviewed in ref. [18–21],
named as Onsager Principle. In a soft matter system, there
(2) are usually several modes of motion, where we can define
slow and fast variables, correspondingly. The relaxation
time of slow variables are distinctively longer than that of
fast variables. Free energy of the system can be calculated
as a function of slow variables, by assuming that the fast
variables are not far from their equilibrium state. Under
these conditions, Onsager reciprocal relation can be guar-
anteed, by using Onsager’s argument, and then Onsager
Principle can be applied. The slow variables in our case
are the positions of the solutes. For applications of this
principle to other systems, one can refer to ref. [21] for
more details.
Onsager principle [20] is briefly stated in the following.
Consider a mechanically isolated soft matter system out
of equilibrium. Let x = (x1 , x2 , . . . xN ) be a set of slow
variables which characterises the non-equilibrium state.
The Onsager principle states that the time derivative of x,
i.e., ẋ = (ẋ1 , ẋ2 , . . . ẋN ) is determined by the minimization
of the following quantity called Rayleighian [20] which is
Eur. Phys. J. E (2016) 39: 22 Page 3 of 10

a quadratic function of ẋ: where rd = Rd (0) is the initial radius of the droplet. Then
Ȧsol can be calculated to be
(ẋ; x) = Φ(ẋ; x) + Ȧ(ẋ; x), (5)
  
kB T φ0 rd 2 1 
where Φ(ẋ; x) is the energy dissipation function which rep- Ȧsol ([Ṙ]; [R]) = dr4πr − Ṙ −  Ṙ
2
resents half of the energy dissipated in the system when vm 0 R R
  
the system evolves with rate ẋ. Φ is a quadratic function kB T φ0 rd 2 1 ∂
of ẋ and is written as = dr4πr − Ṙ −  Ṙ .
2
vm 0 R R ∂r
1 (13)
Φ(ẋ; x) = ζij (x)ẋi ẋj . (6)
2 i,j
Using the integral by parts, we have
Ȧ in eq. (5) is the time derivative of free energy of the   
kB T φ0 rd 2 2 R
system A(x), and is written as Ȧsol ([Ṙ]; [R]) = dr4πr − +  − 2 Ṙ.
2
vm 0 R rR R
 ∂A (14)
Ȧ(ẋ, x) = ẋi . (7) where R = ∂ 2 R/∂r2 .
i
∂xi
Energy dissipation Φ of the system arises from the fric-
Minimization of eq. (5) with respect to ẋi gives the time tion caused by the relative motion between solvent and
evolution equation solute. Let ζ be the friction constant of a particle. Since
the energy dissipation per particle is written as ζ Ṙ2 , and
 ∂A the number density of particles is written as φ/vm , the
ζij ẋj = − . (8)
∂xi energy dissipation function is written as
j

In [20], many dynamical equations which appear in soft 1 Rd φ(R, t) 2
Φ([Ṙ]; [R]) = dR4πR2 ζ Ṙ
matter physics (diffusion equation, Smoluchowskii equa- 2 0 vm
tion, gel dynamics equation, Nernst Fokker Planck equa-  rd
1 φ0
tion etc.) have been derived based on this principle. = dr4πr2 ζ Ṙ2 , (15)
2 vm 0
In the present problem, the non-equilibrium state of
the system is described by the function R(r, t), the average Hence the Rayleighian is given by
position of particles at time t which were initially located
at r. (Notice that here we are considering the spherically  
φ0 rd 2 1
symmetric case.) Hence the function R(r, t) corresponds to ([Ṙ]; [R]) = dr4πr ζ Ṙ2
vm 0 2
the set of slow variables. The time evolution equation for   
R(r, t) can be derived by the Onsager principle. Given the 2 2 R
+ kB T − + − 2 Ṙ . (16)
function R(r, t), particle concentration at R is calculated R rR R
as
r2 By setting the functional derivative of the Rayleighian
φ(R, t) = φ0 2  , (9)
R R δ/δ Ṙ with respect to Ṙ equal to zero, we have the fol-
where R = ∂R/∂r. lowing time evolution equation:
The free energy of the system is then obtained as a   
R 2 2
functional of R(r, t): Ṙ = D − + , (17)
R2 rR R
 Rd
A[R] = dR4πR2 fsol (φ), (10) where D = kB T /ζ is the diffusion constant of the particle.
0
Equation (17) represents the diffusion equation in the
where fsol (φ) is the free energy density (the free energy Lagrangian scheme. In appendix A, it is shown that this
per unit volume) of the solution of concentration φ. If the equation is equivalent to the diffusion equation.
solution is dilute, fsol (φ) is given by In the above discussion, we considered a dilute solu-
tion. At finite concentration, particle interaction may be-
kB T come important, and this can be included in the form of
fsol (φ) = φ ln φ, (11)
vm free energy density and friction constant [20]. For exam-
where kB is Boltzmann constant, T is the temperature, ple, the free energy density of non-dilute solutions may be
and vm is the volume of one solute molecule. From represented by
eqs. (9)–(11), A[R] is rewritten as kB T
   fsol (φ) = (φ ln φ + εφ2 ), (18)
kB T Rd r2 vm
2 φ0
A[R] = dR4πr  ln φ0 2 
vm 0 R R R where ε is a parameter representing the particle-particle
 rd  
kB T 2 r2 interaction. In the following calculation, we will use this
= dr4πr φ0 ln φ0 2  , (12) form of free energy density.
vm 0 R R
Page 4 of 10 Eur. Phys. J. E (2016) 39: 22

4 Dynamics in gel phase Free energy of the whole system which includes sol
phase and gel phase can be written as
The solution may turn into a gel when the solute concen-  rd
tration becomes high. We assume, for simplicity, that this A[R] = dr4πg(R, R ), (21)
transition takes place when the concentration becomes 0
higher than a certain critical concentration φg , called gela- 
tion concentration [22, 23]. where g(R, R ) is defined by
Gel is an elastic material and is quite different from
g(R, R ) = gsol (R, R )Θ (φg − φ(R, t))
fluid. In gels, solvent diffusion is coupled with mechanical
stress applied to the gel. The set of equations that de- + ggel (R, R )Θ (φ(R, t) − φg ) . (22)
scribes such coupling becomes complicated [24], but it be-
comes simple in the present problem of a spherical droplet. Here Θ(x) is a unit step function and gsol and ggel are
The evolution equation for R(r, t) in the gel phase can defined by
also be derived from the Onsager principle. The form of kB T r2 φ0
the energy dissipation function will not be essentially al- gsol (R, R ) = (ln φ(R, t) + εφ(R, t)), (23)
tered from that in solution since the energy dissipation in vm
gels arises also from the relative motion between solute ggel (R, R ) = Rg2 Rg fgel (R, R ). (24)
and solvent. However, the form of the free energy density
will be altered. With the definition of the free energy, we can get the
Gel (or elastic material generally) has a natural state Rayleighian of the gelation system. From the Onsager
to which the gel comes back when all external loads are principle, the dynamical equation of the system, which is
removed. Any deformation of the gel from this state costs represented by the positions of the solutes [R], is expressed
elastic free energy. Suppose that the solute particles la- as
beled by its initial position r become a gel at time tg ,  
kB T ∂gsol (R, R ) d ∂gsol (R, R )
then φ(Rg , tg ) = φg , where Rg = R(r, tg ) is the position Ṙ = −
vm ζ ∂R dr ∂R
of the point when it becomes a gel. Rg and tg are functions
of r, and can be written as Rg (r) and tg (r). We assume × Θ (φg − φ(R, t))
that the gel is in a natural state when it is formed, i.e., a  
material point r in the gel is in the natural state at time ∂ggel (R, R ) d ∂ggel (R, R )
+ −
tg (r). ∂R dr ∂R
At time t, the point moves to a new position R(r, t). At 
this state, the gel is deformed from the natural state. The × Θ (φ(R, t) − φg ) . (25)
deformation of a gel is expressed by three stretch factors
λi = Δ i /Δ i in three principal direction (i = 1, 2, 3), If the whole system is in a gel phase, eq. (25) becomes
where Δ i and Δ i is the length of a line segment before equivalent to the Tanaka-Filmore equation [25] for the gel
and after the deformation. In the present problem, the dynamics of a spherical gel; in other words, diffusion pro-
principal directions are radial direction and two tangential cess and gel dynamics are unified in our framework, by
directions, and their stretch factor is given by tracing the average movement of solute.
R R
λr (r) = , λt (r) = , (19)
Rg Rg 5 Dynamics of cavity formation and its
where Rg = dRg (r)/dr. growth
If λr and λt are different from unity, the free energy of
the gel increases. For small deformation, this increase can When a gel phase is formed at the surface of the droplet,
be expressed by the shear modulus G and osmotic bulk cavities are often observed in the droplet [5]. This can
modulus of K of the gel, and is written as also be commonly observed in hollow particle production
by spray drying [26–28]. In the previous paper [9], we have
  shown that this is due to the deformation of the gel layer
φg 2
fgel (R, R ) = fg − Π
2
− 1 + G (λr − λt ) created at the surface of the droplet: once a gel layer is
φ 3
 2 formed, the subsequent evaporation of solvent compresses
1 φg the gel layer and increases the elastic energy. When this
+ K −1 , (20)
2 φ elastic energy becomes large, the gel layer creates a cavity
to reduce the elastic energy.
where φ = φg /λr λ2t is the solute volume fraction in the gel, If we know R(r, t), we can discuss when such a cavity
and fg is the free energy density of the gel in the natural appears. Suppose that a cavity of radius Rc appears in
state. The second term on the right hand of eq. (20) en- the droplet. Then the point r, which was located at R(r, t)
sures the continuity of the osmotic pressure, when a gel is before the cavitation will move to R̃(r, t) which is given
formed. Given the energy form of the solution as eq. (18), by
Π here is equal to φg + εφ2g . R̃3 = R3 + Rc3 . (26)
Eur. Phys. J. E (2016) 39: 22
The change of the free energy of a droplet when a cavity
of radius Rc is created is then calculated as
 rd
F (Rc ) = dr4π[g(R̃, R̃ ) − g(R, R )] + 4πRc2 γ, (27)
0
where g(R, R ) is given by eq. (22), and γ is the surface
tension of the solution. The last term in eq. (28) represents
the surface energy of the cavity.
Notice that cavitation does not change the volume
fraction at each solute material point. Therefore, the free
energy of solution part (which depends on volume frac-
tion only) is independent of Rc . The bulk modulus part in
eq. (20) is also unchanged. Only the shear modulus part
is changed by cavitation. Therefore, the cavitation free
energy F (Rc ) can be calculated by
 
2 
F (Rc ) = dRg 4πRg2 G (λ̃r − λ̃t )2 − (λr − λt )2
gel 3
+ 4πRc2 γ,
where λ̃r = R̃ /Rg and λ̃t = R̃/Rg . The cavity size is de-
termined by minimizing F (Rc ), and cavitation takes place
when there appears a local minimum of F (Rc ) located at
a non-zero Rc .
The above argument is for the criterion of cavitation.
If there is no cavity at time t, we calculate the new posi-
tion R(r, t + Δt) of each point at time t + Δt following the
procedure described in the previous section. We then ob-
tain the minimum value of cavitation free energy F (Rc ). If
there is not a new local minimum of it, no cavity appears,
and if there is, a cavity appears, and R(r, t) is replaced by
R̃(r, t).
After the cavity is formed, we repeat similar procedure
to account for the growth of the cavity. First we calculate
R(r, t + Δt) assuming that cavity size remains unchanged.
We then obtain the minimum of the free energy with vary-
ing the cavity size, and obtain the new value of R(r, t+Δt).
It should be noticed that in the above procedure so-
lutes moves with solvents when the cavity size is changed.
In other words, volume fraction φ is conserved locally
when the cavity appears. This assumption comes from the
fact that chemical response of the droplet is usually slower
than the mechanical one.
6 Examples
The above model is implemented in a numerical calcula-
tion. The detail is described in appendix B. We now give a
few examples of the results of such a calculation. Here we
change two parameters, initial concentration φ0 and evap-
oration rate J. In the numerical calculation, we choose the
unit of length, time and elastic constant as rd (the initial
droplet radius), rd2 /D (the diffusion time), and kB T /vm .
In the following the dimensionless evaporation rate Jr is
Page 5 of 10
defined by
Jr = Jrd /D, (29)
which is the Peclet number in the present problem.
Figure 2 shows how the initial volume fraction φ0 in-
fluences the drying behavior. Here the evaporation rate Jr
is fixed at Jr = 10. At the beginning of the drying pro-
cess, the volume fraction of the outer layers increases with
the evaporation of solvents, while the volume fraction of
inner layers remain almost unchanged. When gel layers
are formed, they actually promote the diffusion process of
the solute, as one see that the slope in fig. 2(a), (c), (e)
becomes smaller after gelation. This effect results from
the mechanical contribution of the gel layer, as the gel
has a shear modulus G. If the initial volume fraction of
solutes is relatively low, say φ0 = 0.05 (fig. 2(b)), there

is a cavity formed when the outermost layer becomes a
gel.
If we increase φ0 , there will be more gel layers when
(28) cavity is formed. For φ0 = 0.10 (fig. 2(d)), there are about
five layers forming a gel when cavitation happens, and for
φ0 = 0.15 (fig. 2(f)), there will be no cavity appearing at
all (here we do not discuss cavitation in a gel [29]). The
reason for this difference in cavitation phenomenon is a bit
complex. In drying process, different layers have different
reference states, as they form at different time, (as an
example, the N -th layer is not in its reference state when
the (N −1)th layer becomes a gel). When there are several
gel layers, the outer gel layer might prefer expanding, to
reduce its shear elastic energy as we discussed in cavitation
section. However, at the same time, the inner gel layer may
prefer not to expand, as the newly formed gel layer is not
far from its reference state. As a result of this competition
between different layers, the cavitation might not happen
when φ0 is large.
Figure 3 shows the results when the evaporation rate
Jr is changed. Here the volume fraction is fixed at φ0 =
0.1. If evaporation rate is small, say Jr = 1, the volume
fractions of the divided layers are almost the same at dif-
ferent time (fig. 3(a)). The difference of the volume frac-
tions between different layers increases with the evapora-
tion rate, as can be seen from other two cases, Jr = 5
(fig. 3(c)) and Jr = 10 (fig. 3(e)).
One can see that cavitation happens in fig. 3(d)(f),
while not in Fig. 3(b). When evaporation rate is small, the
volume fraction of separate pieces is almost the same, so
their gelation time is not too different. If the whole system
becomes a gel, then it is actually much more difficult for
a cavity to form [29] than that in a solution. For a larger
evaporation rate, the mechanism of cavitation is similar
with the above case when fixing Jr and varying φ0 .
Here we have not scanned the whole parameters space,
as our main aim of this work is to show the new dynam-
ical model to deal with a drying process of soft matter
solutions. This model can also be easily modified to meet
specific demand under different situations, such as skin
formation when drying soft matter solutions in a beaker,
or shape changing of a suspended drying droplet.
Page 6 of 10 Eur. Phys. J. E (2016) 39: 22

Fig. 2. Change of the drying behavior when the initial volume fraction φ0 is changed. φ0 is 0.05 in (a)(b), 0.1 in (c)(d) and
0.15 in (e)(f). The left column ((a), (c) and (e)) shows the time evolution of the solute volume fraction in each layer, and the
right column ((b), (d) and (f)) shows the evolution of the position of each layer. The time unit tref is given by tref = rd2 /D. The
parameters used in the calculation are Jr = 10, G = K = kB T /vm , ε = 2, φg = 0.2. The droplet is divided into N = 20 layers.

7 conclusion tion, there are two main advantages of our scheme: we

In this paper, we propose a new dynamical model for han-
dling drying phenomena of soft matter solutions, by tak-
ing a spherical droplet as an example. The main charac-
teristic of this model is that, we use Lagragian descrip-
tion to describe the motion of solutes, rather than solving
diffusion equations which is based on Eulerian descrip-
tion. Compared with the conventional Eulerian descrip-
can calculate the evolution of the solutes, without han-
dling complex moving boundary conditions as Eulerian
description does; gelation in drying process is considered
naturally, whereas it is difficult to incorporate the gela-
tion process in Eulerian description. By applying this new
model, we also provide a simple example where a droplet
of soft matter solutions is being dried under varying con-
ditions.
Eur. Phys. J. E (2016) 39: 22 Page 7 of 10

Fig. 3. Change of the drying behavior when the evaporation rate Jr is changed. Jr is 1 in (a)(b), 5 in (c)(d) and 10 in (e)(f).
The left column ((a), (c) and (e)) shows the time evolution of the solute volume fraction in each layer, and the right column
((b), (d) and (f)) shows the evolution of the position of each layer. Initial volume fraction is fixed at φ0 = 0.1. Other parameters
are the same as in fig. 2.

Here, we have dealt with a simplified model of a drying
droplet of soft matter solutions. The realistic case is actu-
ally more complex. For example, when gelation happens,
evaporation rate of the solvents might change. Moreover,
gel layers can induce other instabilities, such as wrinkling,
buckling, etc., which gives various complex structures for
the final dried particles. Though there are many detailed
problems in drying soft matter solutions, we believe that
the new dynamical method given in this work can provide
a basic tool in studying these phenomena.
Author contribution statement
Fanlong Meng and Masao Doi propose the new scheme
and write the manuscript, Ling Luo and Zhongcan Ouyang
provide quite informational discussions and help check re-
sults.
F.L. M. and Z.C. O.Y. thank the support of the Na-
tional Basic Research Program of China (973 program, No.
2013CB932800). M. D. acknowledges the financial support of
the Chinese Central Government in the program of Thousand
talents and the NSFC grant (No. 21434001).
Page 8 of 10 Eur. Phys. J. E (2016) 39: 22

Appendix A. Proof of the equivalence ∂φ(R(r, t), t) ∂r

between the solute-based Lagrangian ∂r ∂t R
scheme and solving conventional  
2r r2 r2 R Ṙ
diffusion equation = −φ0 − 2 −
R2 R R3 R2 R2 R
    
2r r2 r2 R R 2 2
= −φ0 − 2 − − + .
It can be proved analytically, that the solute-based La- R2 R R3 R2 R2 R3 rR2 RR
grangian scheme is equivalent to the conventional method
solving diffusion equation, for diffusion process of dilute (A.6)
solutions.
Given the dynamical equation (eq. (17)) in solute- From eqs. (A.5) and (A.6), the left term of the diffusion
based Lagrangian scheme, equation (eq. (A.2)) can be arranged to be
  
R 2 2 ∂φ(R, t) r2 R 3r2 R2 2r2 R
Ṙ = D 2
− 
+ , (A.1) = Dφ0 − + +
R rR R ∂t R2 R4 R2 R5 R3 R3
R

diffusion equation (eq. (3)) is derived 2r2 4r 2 6rR
+ 4  − 3 2 + 2 3 − 2 4 .
R R R R R R R R
 
∂φ(R, t) 1 ∂ 2 ∂φ(R, t) (A.7)
= 2 R D (A.2)
∂t R ∂R ∂R

and vice versa. The details are presented as follows: At the same time, the right term of the diffusion equa-
tion (eq. (A.2)) can be calculated as
The left term of eq. (A.2) can be rewritten as
 
∂φ(R, t) ∂φ(R(r, t), t) ∂φ(R(r, t), t) ∂r 1 ∂ ∂φ(R, t)
2
= + , R D
∂t ∂t ∂r ∂t R2 ∂R ∂R
R r
(A.3)
R
 2 
1 ∂ R ∂φ(R(r, t), t)
where R(r, t) here is expressed as a function of its initial =D 2
R ∂R R ∂r
position r at time t. As shown in the text, concentration
 
of solutes φ(R, t) located at R can be written as 1 ∂ 2r 2r2 r2 R
= Dφ0 2  − −
R R ∂r R2 RR R3
r2  2 
φ(R, t) = φ0
R2 R
. (A.4) r R 3r2 R2 2r2 R
= Dφ0 − 2 4 + 2 5 + 3 3
R R R R R R
Together with ∂r/∂t = −Ṙ/R and eq. (A.1), the first- 
2r2 4r 2 6rR
term on the right-hand side of eq. (A.3) can be calculated + 4  − 3 2 + 2 3 − 2 4 , (A.8)
R R R R R R R R
to be

∂φ(R(r, t), t) which is the same with eq. (A.7).

∂t r
 
2r2 r2 
= φ0 − 3  Ṙ − 2 2 Ṙ
R R R R
   Appendix B. Numerical implementation
2r2 R 2 2
= Dφ0 − 3  − +
R R R2 rR R
 
r2 ∂ R 2 2 We divide the droplet into N pieces radially (shown in
− 2 2 − + fig. 4), Let R0j , R1j , R2j , . . . , RN
j
R R ∂r R2 rR R be the positions of the
 boundaries of the layers, where j = t/dt stands for the
2r2 R 4r 4r2 r2 R time step. At t = 0, values of Ri0 are equal to rd (i/N ).
= Dφ0 − 3 3 + 3 2 − 4  − 2 4
R R R R R R R R When volume fraction of the n-th piece φn is equal to φg ,
2 2  2 
 then this piece becomes a gel, and the inner radius of the
2r R 2 2rR 2r R
+ 2 5 − 2 3 − 2 4 + 4 2 , (A.5) n-th piece at this stage is defined as the gelation radius
R R R R R R R R Rgn−1 of the n piece (the outer radius is almost the same
with Rgn ). Correspondingly, the Rayleighian of the system
and the second term can be calculated to be can be discretized as
Eur. Phys. J. E (2016) 39: 22 Page 9 of 10

Ndr R1
RN RN
dr 2dr R2 R2 RN
0 R1 R1
R0
R0

R0 R2

(a) (b) (c) (d)

Fig. 4. A drying droplet which is divided into N pieces, corresponding to fig. 1.

 

N
2Rgi−1  
kB T 4π 3 φ0
({Ri }, t) = φ0 ri − ri−1
3
+ εφ0 × Θ φji − φg
v m 3 φji Rgi − Rgi−1
2 2
i=1
  j

˙
×φji Θ φg − φji 4 Ri+1 −Rij j
Ri+1 + Rij
+ 3
Rgi+1 −Rgi
3
G −
3 Rgi+1 −Rgi Rgi+1 + Rgi

N
4π   
+ 3
Rgi − Rgi−1
3
Θ φji − φg −2Rgi+1  
i=1
3 × Θ φ j
− φ g
  2
Rgi+1 − Rgi
2 i+1

4 Rij − Ri−1j
Rij + Ri−1
j
 
× G − φg  2  
3 Rgi − Rgi−1 Rgi + Rgi−1 + 3 −Π + K − 1 R j
Θ φj
− φ g
i i
j j j j
 φji
Ṙi − Ṙi−1 Ṙi + Ṙi−1   
× − φg  2  
Rgi − Rgi−1 Rgi + Rgi−1 + 3 Π −K −1 j j
Ri Θ φi+1 − φg . (B.3)
    φji+1
φg φg
− −Π + K −1 φ̇ji
φji (φji )2 From these equations, we can know how solutes distribute
 2 in the droplet, as well as mechanical properties of the gel
πdr3 3 parts.
+ ri+1 − ri−1
3
ζφ0 Ṙij , (B.1)
3 We can also get the discretized version of eq. (28), to
judge whether a cavity is formed or not,
where the changing rate of volume fraction is

N
4π  
3(Rij )2 Ṙij − 3(Ri−1 )2 Ṙi−1
j
F ({Rij }, vc ; t) = 3
Rgi − Rgi−1
3
Θ φji − φg
φ̇i = −φi . (B.2) 3
(Rij )3 − (Ri−1
j
)3 i=1
⎡ ⎡
j j
2
If we take the first derivative of ({Ri }) with respect 2 R̃ni − R̃ni−1 R̃j + R̃ni−1
j
×⎣ G ⎣ − ni
to Ṙi , then we can get the evolution equation of Ri (i = 3 Rgi − Rgi−1 Rgi + Rgi−1
1, 2, 3, . . . , N ),
j j j j
2 ⎤⎤
Rni − Rni−1 Rni + Rni−1
2 − − ⎦⎦
Rij+1 = Rij + 3 ) dt
Rgi − Rgi−1 Rgi + Rgi−1
ζφ0 (ri+1 − ri−1
3
  1 2
kB T 3 φ0 +γ(4π) 3 (3vc ) 3 . (B.4)
× ri − ri−1
3
+ εφ0
vm φji From the above numerical scheme, we can calculate diffu-

3(Rij )2   sion, gelation and cavitation process.
j
× −φi j j
Θ φg − φji
(Ri )3 − (Ri−1 )3

kB T 3 φ0 References
+ ri+1 − ri
3
+ εφ0
vm φji+1 1. P.G. de Gennes, Eur. Phys. J. E 7, 31 (2002).

3(Rij )2   2. T. Okuzono, K. Ozawa, M. Doi, Phys. Rev. Lett. 97,
j
× φi+1 j j
Θ φg − φji+1 136103 (2006).
(Ri+1 )3 − (Ri )3 3. S. Arai, M. Doi, Eur. Phys. J. E 35, 57 (2012).
 4. C. Sadek, L. Pauchard, P. Schuck, Y. Fallourd, N. Pradeau,
4 Rij −Ri−1
j
Rj + Ri−1
j
+ 3
Rgi −Rgi−1
3
G − i C. Le Floch-Fouéré, R. Jeantet, Food Hydrocoll. 48, 8
3 Rgi −Rgi−1 Rgi + Rgi−1 (2015).
Page 10 of 10
5. L. Pauchard, Y. Couder, Europhys. Lett. 66, 667 (2004).
6. N. Tsapis, E.R. Dufresne, S.S. Sinha, C.S. Riera, J.W.
Hutchinson, L. Mahadevan, D.A. Weitz, Phys. Rev. Lett.
94, 018302 (2005).
7. R. Rizzieri, L. Mahadevan, A. Vaziri, A. Donald, Langmuir
22, 3622 (2006).
8. F. Boulogne, F. Giorgiutti-Dauphin, L. Pauchard, Soft
Matter 9, 750 (2013).
9. F. Meng, M. Doi, Z. Ouyang, Phys. Rev. Lett. 113, 098301
(2014).
10. C. Sadek, H. Tabuteau, P. Schuck, Y. Fallourd, N.
Pradeau, C. Le Floch-Fouéré, R. Jeantet, Langmuir 29,
15606 (2013).
11. R. Vehring, W.R. Foss, D. Lechuga-Ballesteros, J. Aerosol
Sci. 38, 728 (2007).
12. N. Reuge, B. Caussat, N. Joffin, J. Dexpert-ghys, M.
Verelst, H. Dexpert, AIChE J. 54, 394 (2008).
13. G.V. Jayanthi, S.C. Zhang, G.L. Messing, Aerosol Sci.
Technol. 19, 478 (1993).
14. R.W. Style, S.S.L. Peppin, Proc. R. Soc. A 467, 174
(2010).
15. L. Rayleigh, Proc. Math. Soc. London 4, 357 (1873).
16. L. Onsager, Phys. Rev. 37, 405 (1931).
Eur. Phys. J. E (2016) 39: 22
17. L. Onsager, Phys. Rev. 38, 2265 (1931).
18. M. Doi, Non-Equilibrium Soft Matter Physics (World Sci-
entific, Singapore, 2012) pp. 1–35.
19. M. Doi, J. Phys.: Condens. Matter 23, 284118 (2011).
20. M. Doi, Soft Matter Physics (Oxford University, New
York, 2013).
21. M. Doi, Chin. Phys. B 24, 020505 (2015).
22. D.S. Bolintineanu, G.S. Grest, J.B. Lechman, F. Pierce,
S.J. Plimpton, P.R. Schunk, Comput. Part. Mech. 1, 321
(2014).
23. M. Laurati, G. Petekidis, N. Koumakis, F. Cardinaux,
A.B. Schofield, J.M. Brader, M. Fuchs, S.U. Egelhaaf, J.
Chem. Phys. 130, 1 (2009).
24. M. Doi, J. Phys. Soc. Jpn. 78, 052001 (2009).
25. T. Tanaka, D.J. Fillmore, J. Chem. Phys. 70, 1214 (1979).
26. R. Vehring, Pharm. Res. 25, 999 (2008).
27. K. Masters, Spray Drying-An Introduction to Princi-
ples, Operational Practice, and Applications (Leonard Hill
Books, London, 1972).
28. D.E. Walton, Dev. Chem. Eng. Mineral Process 10, 323
(2002).
29. G.W. Scherer, D.M. Smith, J. Non. Cryst. Solids 189, 197
(1995).