International Journal of Heat and Mass Transfer 135 (2019) 685–695

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

journal homepage: www.elsevier.com/locate/ijhmt

Lattice Boltzmann method simulation of SVOC mass transfer with

particle suspensions
Yun-Feng Mao, Zhuo Li, Wen-Quan Tao ⇑
State Key Laboratory of Pollution Control and Resources Reuse, College of Environmental Science & Engineering, Tongji University, Shanghai 200092, China
Shanghai Institute of Pollution Control and Ecological Security, China

a r t i c l e i n f o a b s t r a c t

Article history: Semi-volatile organic compounds (SVOC) exist in indoor air at very low concentrations (usually less than
Received 2 September 2018 100 mg/m3), but do great harm to human health. SVOC pollutants transfer in indoor air with suspended
Received in revised form 19 January 2019 particles so complicatedly that traditional experimental methods are hardly to reveal the mass-
Accepted 2 February 2019
transfer mechanism. This study employs Lattice Boltzmann method (LBM) to investigate the mass trans-
Available online 13 February 2019
fer behaviours of two-phase SVOC pollutants (gas- and particle-phase), and compare two widely-used
gas/particle partition models (the instantaneous equilibrium model and the dynamic model). LBM is used
Keywords:
to simulate the fluid flows and SVOC mass transfer, and the cell automation (CA) probabilistic model is
Semi-volatile organic compounds
Lattice Boltzmann method
applied to simulate the particle transport. The interaction between the gas- and particle-phase of
Airborne particles SVOC is considered in the source term of the Lattice Boltzmann equations. One particle is firstly simulated
Indoor air quality in order to focus on the particle-phase concentration, then a large number of particles are simulated to
focus on the effect of particles on the concentration field of SVOC. The simulation results show: (i) par-
ticles in the instantaneous equilibrium model for gas/particle partition always help gas-phase SVOC
transport in indoor air by particle-mediation effect, while such an effect in the dynamic model is greatly
determined by Kp and residence time of particles; (ii) particle-mediation effect can also increase the emis-
sion rate, and a non-uniform distribution of particles predicted by our numerical model indicates that
such an additional emission rate increases exponentially with particle concentration, because a small
number of particles have little possibility of approaching emission surface. (iii) Our numerical model
can effectively estimate SVOC concentration in the presence of airborne particles by tracking each parti-
cle. These results highlight the roles of airborne particles in the transport of indoor SVOC pollutants, and
are expected to be useful to consider dust deposition and resuspension when particles are in a realistic
size distribution.
Ó 2019 Published by Elsevier Ltd.

1. Introduction containing phthalates such as vinyl flooring threatens human

Semi-volatile organic compounds (SVOCs) are the subset of
organic compounds with boiling point from 240–260 to 380–
400 °C [1]. This kind of organic compounds existing indoors
include nicotine, polycyclic aromatic hydrocarbons (PAHs), poly-
chlorinated biphenyls (PCBs), flame retardant (FR) chemicals,
phthalates, etc. [2]. Among them, phthalates, as a typical kind of
SVOCs in indoor air, have been widely used as plasticizers and
adhesive agents. The extensive usage of building materials
⇑ Corresponding author at: State Key Laboratory of Pollution Control and
Resources Reuse, College of Environmental Science & Engineering, Tongji Univer-
sity, Shanghai 200092, China.
E-mail address: wqtao@mail.xjtu.edu.cn (W.-Q. Tao).
health seriously.
Many artifacts release SVOC into indoor air slowly and sustain-
ably, which contributes to main human exposures to indoor SVOC.
Due to their high boiling point, SVOCs are found to interact
strongly with surfaces. Many indoor materials could readily adsorb
them and then slowly desorb them, and hence act as a long-term
source of secondary SVOC pollutants. Partitioning between the
gas phase and airborne particles is another significant feature of
SVOCs, and it has a complicated influence on human exposure
via ingestion, inhalation and dermal pathways [3]. The SVOC inter-
action with both adsorption surface and airborne particles makes
in-depth experimental studies on SVOC transport mechanism in
indoor environment much more complicated. In terms of surface
adsorption, many chamber experiments were conducted to study

https://doi.org/10.1016/j.ijheatmasstransfer.2019.02.007
0017-9310/Ó 2019 Published by Elsevier Ltd.
686 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695
the mechanism of SVOC mass transport in indoor environment. Xu
et al. [4] found that the reduced area of adsorption surface could
decrease the time to reach the steady state, and hence developed
a specially-designed chamber to reduce the period for reaching
steady-state time for di-2-ethylhexyl phthalate (DEHP) from
150 days in FLECs to 30 days. It is also found that by changing
the chamber structure to improve the velocity field the time period
can be reduced. In the further specially-designed chamber by Liang
and Xu [5], the time to reach steady state for DEHP was sharply
reduced to 5 days. Moreover, chamber experiments indicate that
surface adsorption also has a strong influence on SVOC steady-
state concentration. Liang and Xu [6] conducted experiments
respectively in a stainless-steel chamber and a wood chamber,
and found that the gas-phase DEHP concentration (nearly 0.5 lg/
m3) in the wood chamber is greatly lower than it in the
stainless-steel chamber (nearly 1.5 lg/m3).
In terms of the SVOC mass-transfer especially for indoor air
with airborne particles, the experimental papers published in
peer-reviewed journals are limited. Regarding SVOC mass transfer
with settled dust, Clausen et al. [7] investigated the emission of
DEHP from PVC flooring into air and uptake in dust on the flooring,
and found that the dust layer could increase the emission rate of
SVOC from the flooring. Schripp et al. [8] conducted chamber
experiments with a dust layer placed on top of plasticized PVC foil,
and another dust layer on a shelf above the foil. Their experiments
indicated that there was no significant uptake of DEHP in the dust
layer separated from the source, but higher mass transfer was
found in the dust layer in direct contact with the source. As far
as airborne particles are concerned, Benning et al. (2013) [9] exper-
imentally studied SVOC transport in the presence of ammonium
sulfate particles in a specially-designed chamber. Lazarov et al.
[10] studied the influence of seed particles in air on flame retardant
emissions from building materials in the laboratory emission test
cell (FLEC). Wu et al. [11] developed a tube chamber to measure
the particle/gas partition coefficient. Up to now, the major research
method in this regard is theoretical analysis [12–14] based on
lumped-parameter models. Inhalation or even dermal exposures
to various phases of SVOCs were estimated based on either instan-
taneous equilibrium [15–20] or dynamic model [21–23] for gas/-
particle partition. Those models are helpful to assess human
exposure and to propose strategies to reduce exposure. Besides fre-
quent cleaning dust [17], it was revealed that the SVOC concentra-
tion associated with particles on the order of microns tends to
decrease as ventilation increases [14]. Despite the two gas/particle
partitioning models are both widely employed, no studies aimed to
compare the resulting difference. Shi and Zhao [24] compared the
predicted concentrations of SVOCs only originated from outdoor
sources by the instantaneous equilibrium model and dynamic
model for gas-particle partition. Liu et al. [13] compared the gas/
particle-phase concentrations of indoor originated SVOC, assuming
SVOCs were well-mixed in the indoor environment (lumped-
parameter method). In terms of SVOC mass-transfer mechanism,
lumped-parameter model tends to mis-estimate SVOC concentra-
tions in the presence of particles, because in this model (i) disper-
sion of SVOC and particles in indoor space is assumed to complete
in an instant; (ii) particles are assumed an identical residence time
as indoor air; and (iii) particles are assumed to be uniformly dis-
tributed in indoor environment, which may weaken the competi-
tive adsorption between adjacent particles. It is likely that
particle-phase concentration and particle’s effect are over-
estimated, and some mass-transfer details are ignored. So a
numerical model, considering SVOC spatial distribution, is
required.
Lattice Boltzmann method (LBM) has been developed into an
alternative and promising numerical scheme for simulating fluid
flows and modeling physics in fluids [25]. However, its application
in the SVOC mass transfer is quite limited, and to the authors
knowledge only several papers are found [26,27]. In terms of SVOC
mass transfer associated with particles, Mu et al. [26] employed
the LB method to study the adsorption/desorption process
between air and a particle, and found that the amount of adsorbed
DEHP on element carbon (EC) was predominant compared with the
gas phase in pores and dissolved phase in particulate matter (PM).
But their paper only focused on the gas/particle partition process of
SVOCs. Mao et al. [27] developed an adsorption boundary condi-
tion based on linear adsorption isotherm for Lattice Boltzmann
(LB) model, and validated the simulation results with experimental
results in an axisymmetric device. Thus, it can be seen from the
above brief review that there is a plenty room on the applications
of LBM in SVOC mass transfer. This paper aims to present a numer-
ical model, considering spatial distribution of SVOC and airborne
particles, to compare the results predicted by instantaneous equi-
librium and dynamic model for gas/particle partition, as well as to
find the particular SVOC transport phenomenon under the influ-
ence of the typical particle (PM2.5). It is a basis for future compre-
hensive study focused on respective effects of particles on
emission, adsorption and human exposure, which will further con-
sider dispersion, deposition and resuspension of particles in vari-
ous diameters and in a realistic size distribution by improving
the present model with a more practical boundary condition
instead of a simple bounce-back condition for the particle/wall
interaction. In the paper the LB method for fluid flow/mass trans-
port and the cellular automata (CA) probabilistic method for simu-
lating movement of airborne particles are combined to predict the
transport process of indoor SVOCs.
2. Physical/mathematical models and numerical method
2.1. Physical model
The numerical model to estimate the emission rate of DEHP
from vinyl flooring in indoor environment with certain gas- and
particle-phase concentration is shown in Fig. 1. At the bottom,
DEHP gas emits constantly from the vinyl flooring with a constant
equivalent gas-phase concentration C0. The released DEHP trans-
ports in the air (Cg in Fig. 1). Part of them is adsorbed on the wall
surface (Cs in Fig. 1), and some other partitions on airborne parti-
cles (Cp in Fig. 1). Some physical parameters of the model are
shown in Table 1.
The constant emission rate of DEHP can be explained as follows.
Firstly, SVOC emission rate from vinyl floorings is considerably
slow. The overall loss of DEHP from the vinyl slab over the exper-
imental duration of hundreds of days is only 0.002% according to
Fig. 1. Schematic numerical model of the indoor environment.
 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695 687

Table 1 2.2.1. The lattice Boltzmann model for fluid flow

Physical parameters. The evolution equation of the distribution function is written as
Parameter Value
1 eq
Dimensions of the computational domain 0.48 m  0.36 m
f a ðx þ ei Dt; t þ Dt Þ  f a ðx; t Þ ¼  ½f a ðx; t Þ  f a ðx; tÞ ð1Þ
s
Temperature, T 25 °C
eq
Air density, q 1.1728 kg/m3 where s is the relaxation time, f a ðx; tÞ is the equilibrium density
Air velocity at the inlet 0.001 m/s distribution function and can be calculated as
Kinematics viscosity of the air, v 1.579  105 m2/s " #
DEHP diffusion coefficient in the air 2.64  106 m2/s ðea  uÞ ðea  uÞ2 juj2
ðeqÞ
Concentration from the source, y0 1.0 mg/m3 fa ¼ xa q 1 þ þ  2 ð2Þ
Equilibrium sorption coefficient, Ks 1800 m c2s 2c4s 2cs
Density of airborne particles 2000 kg/m3
Particle diameter 2.5 mm where xa is the fixed weighting coefficient, depending on the
length of the corresponding lattice vector. They are given as fol-
lows: x0 = 4/9, x14 = 1/9, and x58 = 1/36. cs is the sound lattice
pffiffiffi
speed and is equals to c= 3 with c ¼ dx=dt. dx and dt are space
the experiments [20]. Secondly, in contrast to VOC emission, the and time increments, respectively. The velocity of mesoscopic fluid
emission of SVOCs is generally subject to ‘‘external” control (SVOCs particles is divided into nine directions as shown in Eq. (3). a
firstly diffuse from the emission material into the boundary layer denotes the types of particles moving to different directions
adjacent to the emission surface, which is called internal diffusion; (Fig. 2), and particles of type 0 are the rest particles.
then they further diffuse into the bulk air, which is called external 8
>
<
0 a¼0
diffusion). The overall emission process can be regarded as the     
ea ¼ c cos ða  1Þ p2 ; sin ða  1Þ p2 a ¼ 1; 2; 3; 4 ð3Þ
emission from the boundary layer above the emission surface, >
: pffiffiffi     
and the loss of SVOC in the boundary layer is replenished quickly 2c cos ð2a  1Þ p4 ; sin ð2a  1Þ p4 a ¼ 5; 6; 7; 8
by the internal diffusion. So the SVOC concentration in the bound-
The fluid density q and the fluid velocity u ¼ ðu; v Þ are given by
ary layer can be assumed to be stable and constant over a reason-
ably long period. For details, please refer to Ref. [28] of the present X
8
1X
8

paper. q¼ f a; ui ¼ ea f a ð4Þ
a¼0
q a¼0

2.2. Numerical model and the relationship between the fluid kinetic viscosity v and the
relaxation time s is
Solutions of the fluid flow and SVOC mass transport are imple- v ¼ c2s Dtðs  0:5Þ ð5Þ
mented by using LBM. The basic theory and application examples
in fluid flow and heat transfer have been documented in many
Refs. [29,30] and will not be restated for simplicity. Only a very 2.2.2. The lattice Boltzmann model for mass transport
brief description is given below. The LBM uses a regular grid and The evolution equation of the distribution function of mass
divides the fluid domain into a set of lattice nodes. Each node transport is written as
has sets of probability distribution functions representing the den- 1 
eq
sity of fluid particles and the concentration of SVOC particles. In g k ðx þ ei Dt; t þ DtÞ  g k ðx; t Þ ¼  0 g k ðx; t Þ  g k ðx; t Þ  wk DtS
s
this study, the single relaxation Bhatnagar-Cross-Krook (BGK)
ð6Þ
scheme proposed by Qian et al. [30] is employed to obtain the
collision-propagation equation. The most common lattice model where S is the source term representing the interaction between the
D2Q9 (Fig. 2) for two-dimensional velocity field simulation is used gas- and particle-phase SVOCs. g eq
k ðx; tÞ is the equilibrium distribu-
in the study while a reduced D2Q5 (Fig. 2) lattice model is chosen tion function for mass transport, which can be calculated as:
for the diffusion process of SVOC concentration without losing the  
ek  u
accuracy. The cellular automata probabilistic model is used for the g eq
k ¼ wk C 1 þ ð7Þ
c2s
particle transport.

Fig. 2. Discrete velocity vectors in D2Q9 and D2Q5.

688 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695

where w14 ¼ 1=6 and w0 ¼ 1=3 are the corresponding weight coef- k = 0,4 Eq. (12) is summed up with
ficient for mass transport. C is SVOC concentration. P4
g k ðx þ ei Dt; t þ DtÞ ¼ g k ðx; t þ Dt Þ, and k¼0 g k ð x; t þ Dt Þ can be
The interaction between the gas- and particle-phase SVOCs is obtained as
included in the source term. Two different kinds of particle parti- P4  eq 
tion models are used here, namely instantaneous equilibrium X
4
k¼0 g k ðx; t Þ  g k ðx; t Þ
X
4
g k ðx; t þ DtÞ ¼ 0 Kp 0
þ g k ðx; t Þ ð13Þ
model (Eq. (8)) and dynamic models (Eq. (9)): k¼0 s þ qp TSPs k¼0

Kp By substituting Eq. (13) into Eq. (12), the final distribution func-
qp ¼ C g TSP for the instantaneous model ð8Þ
qp tions for the instantaneous equilibrium model with an implicit
source term can be determined from the data of the previous time
8
< V p dC p ¼ hp Ap C g  C p step without iterations as follows.
dt Kp
for the dynamic model ð9Þ 1 
: qp ¼ TSP g k ðx þ ei Dt; t þ DtÞ  g k ðx; tÞ ¼  g k ðx; tÞ  g eq
q Cp k ðx; t Þ
p s0 " #
wk K p TSP X
4 X4
where TSP is the particle concentration; hp is the mass-transfer  0 0 g eq ð x; t Þ  g k ð x; t Þ
qp s þ K p TSPs k¼0 k
coefficient on the particle surface, and it may be influenced by many k¼0

factors, such as particle size, gas/particle partition coefficient and ð14Þ

air velocity around it. A uniform coefficient is used in the study
2Dg
based on an empirical correlation [31] hp ¼ 1þKn
dp 1þ1:71K n þ1:333Kn2
. Here 2.2.3. The cell automation (CA) probabilistic model for particle
Dg is the molecular diffusive coefficient of SVOCs, dp is particle transport
diameter, K n is Knudsen number determined by K n ¼ 2k ; k ¼ 3 cg ,
D Airborne particles transport in the indoor air under the effects
dp
qffiffiffiffiffiffi of the drag force, the buoyancy force, the gravity and the Brownian
and c ¼ 8RT
pM (k is the mean free path of SVOCs, c is the mean molec- force. The original CA probabilistic model was first proposed by
ular speed of SVOCs, R is the molecular weight of SVOCs.); In Eq. (9) Masselot and Chopard [32] to describe particle transport. Wang
Ap is the surface area of a particle; Vp is the volume of a particle; qp et al. [33,34] further improved it by quantitatively calculating the
is the density of a particle; Cp is the particle-phase SVOC mass per particle velocity and displacement under the combined effect of
unit particle’s volume; qp is the particle-phase SVOC mass per unit various forces, and their improved model successfully simulated
air volume; K p is the gas/particle equilibrium partition coefficient; the particle capture process of fibrous filters. When random Brow-
P nian force [35] is introduced to consider the Brownian diffusion,
and Cg is the gas-phase SVOC concentration (C g ¼ 4k¼0 g k ).
the particle motion is thus described by the Newtonian equation
In the discretized form, the source terms in Eq. (6) will take
under the effects of drag force FD, the Brownian force FB, the buoy-
either the instantaneous equilibrium model (Eq. (10a)) or the
ancy force FBUO and the gravity FG [33] as follows:
dynamic model (Eq. (10b)). It should be noted that K p is the gas/- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
particle equilibrium partition coefficient defined by
Cp
and is dup u  up 216lkB T q
Cg ¼ FD þ FB þFBUO þ FG ¼ þ1 þ gg ð15Þ
dimensionless. dt sp pq2p d5p Dt qp
8 DC g 2
< TSP Kqp Dt
instantaneous equlibrium model ðaÞ where sp ¼ qp dp C c =ð18lÞ is relaxation time scale of particle; l is
S¼ ð10Þ
p

:  TSP DC p dynamic viscosity; dp and qp are particle diameter and density

q Dt
dynamic model ðbÞ
p respectively; Cc is an empirical correction factor called Cunningham
An explicit treatment of the source terms in Eq. (6) would bring slip correction factor, here Cc = 1 for simplicity [40]; 1 is zero-mean,
instability and inaccuracy for the instantaneous equilibrium model unit-variance independent Gaussian random number; kB is Boltz-
especially when Kp value is relatively large such as 1012. Hence an mann constant; T is environmental temperature.
implicit scheme is used to deal with the mass source term for the In the Lagrangian particle tracking approach, the particle veloc-
instantaneous equilibrium model (Eq. (10a)). For the instantaneous ity and displacement can be explicitly calculated by integrating Eq.
equilibrium model, the change of the gas-phase SVOC concentra- (15) over time t:
tion can be written as Dt
up ¼ up exp 
X
sp
4
! !
DC g ¼ C g ðx; t þ Dt Þ  C g ðx; t Þ ¼ ½g k ðx; t þ DtÞ  g k ðx; tÞ ð11Þ q Dt
k¼0 þ u þ FB þ g  g sp 1  exp  ð16Þ
qp sp
By substituting Eq. (11) into Eq. (10a) and then into Eq. (6), the
new kinetic equation for the mass distribution functions with an Dt
external source term can be written as xp ¼ xp þ ðup  uÞ 1  exp  þ uDt
sp
!
1  Dt q
g k ðx þ ei Dt; t þ DtÞ  g k ðx; t Þ ¼  0 g k ðx; t Þ  g eq
k ðx; t Þ þ Dt þ 1  exp  sp FB þ g  g sp ð17Þ
s sp qp
Kp
 wk TSP
qp The Lagrangian particle tracking approach could calculate the
P4 accurate particle displacement (Dxp ¼ xp  xp ) within the period
k¼0 ½g k ðx; t þ DtÞ  g k ðx; tÞ
 Dt Dt.
Dt
In the CA probabilities model, the transport probabilities of the
ð12Þ
particle towards the east (p1), north (p2), west (p3) and south (p4)
If g k ðx þ ei Dt; t þ DtÞ in Eq. (12) is replaced by g k ðx; t þ DtÞ, it is direction are calculated in the D2Q5 model:
the collision step. After g k ðx; t þ Dt Þ is obtained, g k ðx þ ei Dt; t þ Dt Þ   Dt Dxp
is equal to g k ðx; t þ Dt Þ in the streaming step. Considering the fact pi ¼ max 0; up  ei ¼ max 0; ei ; i ¼ 1; 2; 3; 4 ð18Þ
Dx Dx
that the source term only takes effect in the collision step, for
 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695 689

Based on the actual displacement of the particle within time model [20], as shown in Fig. 3. It can be seen that the simulated
step Dt in Eq. (17), the transport probabilities pi in different direc- results from the present model agree well with the results from
tions can be obtained from Eq. (18). The transport probabilities pi is the Xu’s model.
further used to determine whether the solid particle stays at the The model and self-developed program are further validated
original or jump to a nearest-neighbor node in directions of 1,2,3,4: through comparing one airborne particle migration by the cellular
X4 automata probabilistic model with that by the Lagrangian point-
xp ¼ xp þ i¼1
li ei Dt ð19Þ particle tracking approach. A particle gets in the simulation
domain with a zero SVOC concentration from the inlet, and moves
where mi is a Boolean variable which is equal to 1 with probability
in the indoor air until it gets out under the influence of the drag
pi. Based on two independent random numbers (r1 and r2) from a
force, the buoyancy force, the gravity and the Brownian force.
uniform distribution in the interval [0,1], the new position of the
The moving trajectories of the particle from the two methods are
particle, in a typical example where p1 > 0, p2 > 0, p3 ¼ 0, and
shown in Fig. 4(a). The SVOC concentrations associated with the
p4 ¼ 0, is determined by Monte Carlo method as follows:
particle obtained by the present model and the Xu’s model are
8
> if r 1 > p1 and r 2 > p2 ; xp ¼ xp shown in Fig. 4(b). From the validation results in Fig. 3 and
>
>
< if r 1 < p1 and r 2 > p2 ; xp ¼ xp þ e1 Dt Fig. 4, it can be seen that the present model and self-developed
ð20Þ code are feasible to predict the SVOC emission, adsorption and
>
> if r 1 > p1 and r 2 < p2 ; xp ¼ xp þ e2 Dt
>
: advection-diffusion in indoor air, the particle migration, and the
if r 1 < p1 and r 2 < p2 ; xp ¼ xp þ e1 Dt þ e2 Dt interaction between the gas- and the particle-phase SVOCs.
Once each particle position and velocity are determined by the It is worth noting that it would be better to compare our numer-
CA probabilities model, the number of real particles (N(x,t)) can be ical results with some experimental data. Because SVOC emission
obtained by statistics of all particles at each node. The velocity of process is slow and its concentration is quite small, most of exper-
real particles and the particle-phase SVOC concentration can be imental studies were conducted in a small chamber with a very
calculated by: large air exchange rate to reduce experimental period (e.g. 25 h1
in the specially-designed chamber [4], 780 h1 in the FLEC cham-
1 XNðx;tÞ j ber [7]). In that case, the effects of airborne particles with such a
up ðx; t Þ ¼ up ðx; tÞ ð21Þ
Nðx; tÞ j¼1
small residence time would be much weakened. So we conducted
our simulation in a medium size chamber, for which experiments
1 XNðx;tÞ j are difficult to conduct and no test data are available for
C p ðx; t Þ ¼ C p ðx; tÞ ð22Þ
Nðx; tÞ j¼1 comparison.

3.2. SVOC concentration associated with a single particle by the

2.3. Initial conditions and boundary conditions
instantaneous equilibrium model and the dynamic model
The velocity-inlet and pressure-outlet boundaries (Fig. 1) are
Xu and Little [20] employed the the linear instantaneous equi-
applied for the inlet and outlet of the airflow. The fluid inlet veloc-
librium model to characterize the gaseous SVOC partitioning on
ity is constant: u(x) = u0. Non-equilibrium extrapolation scheme is
particles. The instantaneous equilibrium model is also widely used
used to deal with the inlet and outlet boundary conditions, which
in the relationship between the gas- and particle-phase SVOCs in
has second-order accuracy and good stability [36]. The velocity
other studies [14,27,37,38]. A dynamic model considering the
boundary condition of the wall is non-slip boundary. The SVOC
dynamic interaction between the gas-phase SVOCs and airborne
concentration in the computational domain is initially assumed
particles is also developed to describe the transport processes of
to be zero. Constant SVOC concentration C0 is fixed on the bottom.
both gas- and particle-phase SVOCs for higher accuracy [12,13].
The SVOC adsorption boundary condition of the wall is imple-
Due to the conciseness and implementation simplicity of the
mented by:
instantaneous equilibrium model, the dynamic model isn’t applied
@C g @C g extensively in the analysis of particle effects on SVOC transport in
Dg ¼ flux ¼ K s ð23Þ
@n surf @t surf
indoor environment. In this section, only one particle is considered
in order to track the SVOC concentration change on the particle
where Dg is the molecular diffusion coefficient of SVOC in the air, n with its motion in order to compare the instantaneous equilibrium
is the normal direction of the adsorption surface.
When the particle-phase SVOC concentration on a single parti-
cle is studied, the particle number is fixed to be zero for the initial
condition, but a single particle is released in the inlet as for the
inlet boundary condition. When the SVOC concentration field
under the effects of a large number of particles is studied, the par-
ticle number is constantly N0 as the inlet boundary condition.

3. Results and discussion

3.1. Model and code validation

In the part of model validation, a small inlet velocity 0.001 m/s

is used to make the particle reside longer in indoor environment,
and the gas/wall equilibrium partition coefficient (Ks) 1800 m,
which is often found in the steel-chamber experiments [4,5], is
employed. The parameters of the base case are shown in Table 1.
The mesh of the computational domain is 400  300. The averaged
SVOC concentration in the air is compared with the results of Xu’s Fig. 3. The validation of the DEHP mass transport in indoor environment.
690 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695
Fig. 4. Comparison results of single particle (a) Particle movement (b) Particle-phase DEHP concentration.
and dynamic models, excluding the influence of local gas-phase
SVOC concentration change due to other particles’ competitive
adsorption. In Section 3.3, the comparison of these two models is
also conducted by studying the influences on the SVOC concentra-
tion under a large number of particles in Eulerian coordinates. The
focus of the simulation is to study the gas-phase and particle-phase
SVOC concentration fields over the entire indoor space by the two
models. It should be noted here that the particle-phase SVOC con-
centration (Cp) in this section is calculated by dividing SVOC mass
on the particle by the particle volume while the particle-phase
concentration (qp) in Section 3.3 is the particle-phase SVOC mass
per air volume.
The changes of the particle-phase SVOC concentration with time
obtained by the instantaneous equilibrium and dynamic models are
shown in Fig. 5 for the case of K p = 1012. To compare the difference
from the two models, the results are displayed in the graph with
logarithmic ordinate (Fig. 5(a)) and arithmetic ordinate (Fig. 5(b),
(c) where the units are the same as Fig. 5(a)). From the figures,
we can find that the results of the two methods have a magnitude
difference as large as six orders. Usually the dynamic model is con-
sidered more accurate for the transient process. Thus, it indicates
that the instantaneous equilibrium model seriously overestimates
the SVOC transport between gas and particles. Such an overestima-
tion can lead to a large error in evaluation on human exposure to
particle-phase SVOC especially via dermal contact and dust inges-
tion, which, as well as inhalation, are the main potential pathways
of human exposure. Some particle-phase SVOC enters into human
body with air flow and thus inhalation exposure can not be evalu-
ated by Cp determined by the instantaneous equilibrium model.
Particles can also adhere to and accumulate on human skin and
other surfaces. Contaminants can penetrate the skin and directly
enter the blood. People can also ingest dust and associated contam-
inants accidently by food. Johnson-Restrepo and Kannan [39] found
that ingestion and dermal adsorption of house dust are the major
pathways of polybrominated diphenyl ethers (PBDEs) exposures,
accounting for 56–77% of the total PBDE intake. Beko et al. [40] also
pointed out that dust ingestion made the largest contribution of the
indoor exposure pathways for high molecular weight phthalates. So
for the case where dermal contact and ingestion of dust are non-
negligible, the choice between the instantaneous equilibrium
model and the dynamic model should be made very carefully, con-
sidering the fact that particle-phase SVOC mass is much larger in
the instantaneous equilibrium than in the dynamic model.
From Fig. 5(c), we could find that the SVOC concentration on the
particle keeps increasing in the dynamic model during the particle
moving in indoor environment. However, it fluctuates in the
instantaneous equilibrium model after the initial 75 s. According
to the mass conservation in the local air region, an increase of
the particle-phase concentration would cause a decrease of the
gas-phase concentration around the particle, and vice versa. The
particle in the dynamic model always traps the SVOC pollutants
from the air around it until it escapes the chamber. But the particle
simulated by the instantaneous equilibrium model not only cap-
tures the SVOC pollutants (the increasing part of the curve in
Fig. 5(b)), but also carries them to the region of low gas-phase con-
centration, and then releases them into the air (the decreasing part
of the curve). Hence the different models leads to different behav-
iors of a particle in the space. The particle in the instantaneous
equilibrium model has two roles: (1) works as a sink to remove
the gas-phase SVOC pollutants from the air; (2) helps the mass
transport of gas-phase SVOCs in the space. Because SVOC concen-
tration equilibrium between particles and air is more difficult to
achieve at a large Kp, mass transfer results at various lower Kp val-
ues for the dynamic model are further obtained to explore when
the particle in the dynamic model has the above two roles to affect
SVOC mass transfer. The results are shown in Fig. 6, where c
denotes the ratio of the cumulative desorbed mass over the
adsorbed mass at the corresponding time.
Fig. 6 shows that the variation of c with time does depend on
K p . When the K p value is relatively high, such as 1010 and 1012,
the particle doesn’t release SVOCs into the surrounding air. But
when K p decreases to 108, a fluctuating curve can be observed.
The lower the K p value is, the more the particle desorbs and the
more frequently the curve fluctuates, because the equilibrium state
can be reached more easily at lower values of Kp. At lower Kp, the
higher desorption/adsorption ratio (c) implies that the particle car-
ries less SVOC pollutants from the indoor air, and that the particle
helps the SVOC transport in indoor air more efficiently. So the par-
ticle in the dynamic model also plays the two roles simultaneously
at lower Kp. Above all, by focusing on a single particle motioning in
indoor air and the particle-phase SVOC concentration Cp, it is found
that the instantaneous equilibrium model exaggerates the
adsorbed amount on the particle. In that case, the dynamic model
can’t be simplified into the instantaneous equilibrium model when
evaluating the SVOC mass on a particle. Such an evaluation is often
conducted in the estimation of dermal adsorption and digestion of
particle-phase SVOCs in reference [40]. It is also observed that the
particle keeps exchanging SVOC pollutants with the nearby air,
which always happen in the instantaneous model, but happen in
the dynamic model only at a lower Kp. Whether the particle/air
SVOC exchange has a substantial effect on the overall gas-phase
concentration will be discussed in Section 3.3.
 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695 691

Fig. 5. The DEHP concentration associated with a particle (Cp) in the logarithmic ordinate and in the linear ordinate.

1.0 qb;ins  103 in Fig. 7 denotes the one thousandth of the particle-
Kp=106
phase concentration in the instantaneous equilibrium model, in
Kp=108
order to compare the gas-phase and particle-phase concentration
0.8 Kp=1010
in the same figure. Because human beings are directly exposed to
Kp=1012
air, qb instead of Cp should be used when we evaluate human expo-
0.6 sure to particle-phase SVOCs by inhalation. From Fig. 7, it can be
observed that the gas-phase SVOC concentrations (Cg) from the
instantaneous equilibrium and dynamic models have an apprecia-
γ

0.4 ble difference in qb when Kp value is 1012 and TSP is 100 mg/m3. Liu
et al. [41] indicated that particle-phase concentration (qb) domi-
0.2 nates airborne (gas- + particle-phase) concentration, but their
research was done based on the instantaneous equilibrium model.
Xu and Little [20] also found that particle-phase SVOC concentra-
0.0 tion (qb) was far larger than the gas-phase concentration (Cg).
0 50 100 150 200 250 Our research on the instantaneous equilibrium model comes to
time (s) the same conclusion, but results from the dynamic model indicate
that particle-phase concentration is far less than gas-phase con-
Fig. 6. The ratio of the desorbed DEHP mass from a particle over the adsorbed mass centration. At the time of 480 h, the qb value in the instantaneous
onto a particle at different Kp values in the dynamic model. equilibrium model is as much as 12.62 mg/m3, far larger than the

3.3. SVOC concentration under different gas/particle partition models

As it is indicated in Section 3.2, the instantaneous equilibrium

and dynamic gas/particle partition models make great differences
in the SVOC concentration associated with one particle. However,
because particles are often present in a large number, it is not clear
whether the particle-phase SVOC concentration per unit air vol-
ume (qb) can be affected greatly. Hence the gas-phase and
particle-phase SVOC concentration are studied respectively with
the instantaneous equilibrium and dynamic gas/particle partition
models when the total suspended particles (TSP) concentration
and the velocity are respectively 100 mg/m3 and 0.01 m/s in the
inflow air. The average gas- and particle-phase concentration with
the instantaneous model (Cg,ins and qb,ins) and the dynamic model
(Cg,dyn and qb,dyn) in the chamber are shown in Fig. 7.
As shown in Fig. 7, the particle-phase concentration per unit air
volume (qb) is greatly lower than that of the particle-phase concen- Fig. 7. Gas- and particle-phase concentration from the instantaneous equilibrium
tration per unit particle volume (Cp). It should be noted that and dynamic models (TSP = 100 mg/m3, Kp = 1012).
692 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695

gas-phase concentration (nearly 0.12 mg/m3), while qb in the 1

dynamic model is only 0.03 mg/m3. Hence the instantaneous equi-
librium model can cause a greatly large airborne SVOC concentra- 0.1
tion. High particle-phase concentration in instantaneous
equilibrium model is resulted from the increased emission rate. 0.01
Numerical simulation results show that at 480 h the emission rates
are 5.96 mg/(hm2) in the instantaneous equilibrium model and 1E-3
0.36 mg/(hm2) in the dynamic model, respectively.
instantaneous model

γ
The gas-phase concentrations and the emission rates at differ- 1E-4
dynamic model, Kp=108
ent K p values for the dynamic model are shown in Fig. 8 for the dynamic model, Kp=1010
1E-5
case of TSP = 100 mg/m3 in the inflow air with a velocity 0.001 m/ dynamic model, Kp=1012
s. It can be seen that both the gas-phase concentration and the
1E-6
emission rate don’t change greatly with K p , and they even get clo-
ser with increasing K p . In Fig. 8, the curves of the gas-phase con- 1E-7
centration (Cg) are smooth because the partition between 0 100 200 300 400 500
particles and air has less impact on the averaged gas-phase con- time (h)
centration. But the averaged emission rates fluctuate with time
Fig. 9. Time history of the ratio of the desorbed DEHP mass from particles over the
especially at a Kp value of 1012, because under such a condition
adsorbed mass on particles (TSP = 100 mg/m3).
the SVOC mass on a particle is larger and the discontinuous escap-
ing of particles from indoor environment affects strongly on the
total airborne SVOC concentration, which further affects the emis-
sion rate.
In Fig. 9 the variation of c with time is presented, where c is the
ratio of the desorbed SVOC mass from particles over the adsorbed
mass on particles. In the instantaneous equilibrium model, the des-
orbed SVOC mass and adsorbed mass grow synchronously, leading
to a constant c. However, in the dynamic model the desorbed SVOC
mass keeps decreasing relative to the adsorbed mass, which means
that particles are more inclined to remove SVOC pollutants from
indoor environment than enhancing gas-phase SVOC mass trans-
port in indoor air. When Kp increases, c greatly decreases. As we
can see in Fig. 9, the value c is slight below 0.1 at a Kp value of
108, while the c values are less than 103 and 106 at Kp values
of 1010 and 1012, respectively. It might be the reason why Cg at a
Kp value of 108 are a bit larger than Cg when Kp = 1010 and 1012
in Fig. 8. So we can conclude that air-borne particle has some
media effect in the transport of SVOC, that is, it adsorbs SVOC in
high-concentration region and desorbs SVOC in low- Fig. 10. The gas-phase SVOC concentration at different TSPs.
concentration region, which will obviously influence on human
exposure to gas-phase SVOC.
2.5 mm. The SVOC partition coefficient between air and particles
(Kp) is 1012. The variations of averaged gas-phase SVOC concentra-
3.4. The influence of TSP on SVOC concentration tions in the chamber over 20 days are shown in Fig. 10.
As shown in Fig. 10, for the four TSPs, the gas-phase SVOC con-
In this section another three cases with TSPs (20 mg/m3, 500 mg/ centration increases with time, but the increase trend gradually
m and 1000 mg/m3) in the inflow air are further simulated to study
3
becomes mild. The concentrations at different TSPs are different,
the influence of the TSP on the SVOC concentration with the same and their differences are not significant but become larger with
inlet air velocity (0.001 m/s). The diameter of the particles is also time. It indicates that residence time of particles greatly determi-
nes the influence of particle, so it is expected that airborne parti-
cles have a larger influence on gas-phase SVOC concentration in
a bigger realistic room. The more airborne particles, the lower
the gas-phase SVOC concentration is, implying that a larger TSP
may reduce the gas-phase SVOC exposure. It is also found that
gas-phase concentration at TSP = 100 lg/m3 is slightly lower than
at TSP = 20 lg/m3. The largest TSP in the study is only 1000 mg/
m3. However, in a seriously-polluted environment it may be much
larger. For example, the record shows that in the year of 2002 the
peak value of TSP was 11 mg/m3 in Beijing [42]. In central Europe, a
TSP of 100 mg/m3 was also observed [43]. In indoor environment,
particulate levels over 5 mg/m3 were observed during cooking
activities [44]. Particulate peek values of 5.5 mg/m3 were also
observed when cigarettes were smoked [45]. Hence in the presence
of more particulates, human health is exposed less to gas-phase
but more to particle-phase SVOCs by both inhalation and dermal
Fig. 8. Gas-phase concentration and emission rate at different Kp values contact, and the emission rate also increases, which will be dis-
(TSP = 100 mg/m3). cussed later.
 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695 693

Fig.11. The SVOC concentration contours at different TSPs in the 20th day.

The concentration distributions of the gas- and particle-phase gas-phase SVOC. Because of the strong adsorption ability of parti-
SVOC at different TSPs are shown in Fig. 11. It can be seen from cles, the SVOCs are in the air rapidly adsorbed to particles, and
the figure that when the TSP is larger, humans are exposed to less newly-released SVOC pollutants are not able to fill the vacancy
694 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695
Fig. 12. Extra emission rates caused by particulate media.
quickly. Therefore, a considerably lower gas-phase concentration
can be observed at a larger TSP. In contrast, the particle-phase
SVOC concentration in the zone of low gas-phase concentration
is larger. In that case, an air cleaner can effectively reduce SVOC
exposure by capturing particles. Different from gas-phase concen-
tration, particle-phase concentration increases with TSP. But
human exposure to airborne SVOCs (gas- + particle-phase SVOCs)
is actually not reduced because two-phase SVOC pollutants can
both enter into human body through inhalation. Instead, airborne
SVOCs are increased by improved emission rates due to particles.
By comparing the concentration distributions of gas- and
particle-phase SVOC in Fig. 11, we can find that particle-phase
SVOCs accumulate in the lower left part of the chamber, and espe-
cially at the dead zone (Fig. 11(d), (f), and (h)), where recirculation
occurs (Fig. 11(a)). Hence more SVOC exposure can only be avoided
in the upper right of the space, and such a pattern of ventilation
isn’t helpful to reducing SVOC exposure. So extended investiga-
tions of ventilation patterns to reduce SVOC exposure are required,
and a realistic distribution of particle size should be taken into con-
sideration, in which particle flow, deposition and suspension may
occur simultaneously.
As we have indicated in Fig. 9, particles help SVOC transport in
indoor space at large gas/particle partition coefficient. If the same
effect occurs near the emission surface, particles may also enhance
SVOC emissions. Particle-phase SVOCs can rapidly reach equilib-
rium with the gas-phase because the large specific surface area
and the higher gas/particle partition coefficient. Afterwards, they
adsorb gas-phase SVOCs heavily in the thin layer above the emis-
sion surface, then carry them into the bulk air, and desorb them.
The emission rates accelerated by particle are calculated for the
cases of different TSPs, as shown in Fig. 12. Although the
particle-mediated emission rates are far less than the direct emis-
sion rate (nearly 0.3 mg/(m2h) for four cases), it increases with TSP
in the orders of magnitude. When TSP is 100 mg/m3, the additional
emission rate is only 2.2  106 mg/(m2h). When TSPs increase to
500 mg/m3 and 1000 mg/m3, the additional emission rates rise to
5.3  105 mg/(m2h) and 2.1  104 mg/(m2h), respectively. How-
ever, no additional emission rate is observed when TSP is 20 mg/
m3, because a small number of particles have little possibility of
approaching the emission surface. In fact, the additional emission
rate is an average value over the whole emission surface, and it
could be much larger in some zones above the emission surface
where particles flux toward and outward frequently. The above
results show that the additional emission rate can’t be ignored in
the room with a large number of particles like a kitchen and a
smoking area.
Finally, it is interesting to note that the present model could be
used to investigate the influence of air flow pattern such as inlet–
outlet configurations on then concentrations of SVOCs in the gas-
and particle phase. This will be our further work.
4. Conclusion
In this paper, the lattice Boltzmann method (LBM) is used to
study the SVOC mass transfer behaviors along with airborne parti-
cle. The cell automation (CA) probabilistic model is employed for
particle transport. The interaction between the gas- and particle-
phase SVOCs is included in the source term of the Lattice Boltz-
mann equations. The instantaneous equilibrium model and the
dynamic model for gas-particle partition are compared. Firstly,
one particle is simulated to focus on the particle-phase concentra-
tion (Cp), then a large number of particles are simulated to focus on
the effect of particle on the overall SVOC concentration field and
the particle-phase concentration per unit air volume (qp). An impli-
cit scheme for instantaneous equilibrium model is used to improve
computational stability. Following conclusions can be made, which
are often missed by lumped-parameter method:
(1) Because SVOC concentration is not uniformly distributed in
indoor environment and particles move faster than SVOC gas, it is
found that particles help gas-phase SVOC transport in indoor air by
adsorption in high-concentration zone and desorption in low-
concentration zone. This effect always occurs in the instantaneous
equilibrium model for gas/particle partition, but only occurs in the
dynamic model with a low Kp value. Such a particle-mediation
effect is determined by the residence time of airborne particles rel-
ative to the residence time of gas-phase concentration, as well as
Kp.
(2) Our simulation results based on a whole chamber indicate
that the additional emission rate due to airborne particles is far less
than the direct emission, and it grows by orders of magnitude with
the increase of particle concentration. A small number of particles
have no positive effect on the additional emission rate, because
these particles have little possibility of approaching the emission
surface, which differs from the uniform distribution of particles
in the lumped-parameter model.
(3) The model in this paper can effectively estimates the gas/
particle-phase SVOC concentration, in which information of each
particle (position, velocity, and SVOC concentration) is well
tracked. In addition, the model can be improved by considering
the deposition and resuspension of large particles to study a real-
istic situation of SVOC mass-transfer in indoor air.
Conflict of interest
We confirm there is no conflict of interest.
Acknowledgement
This work is supported by the National Key Research and Devel-
opment Program of China (2017YFC0702700), the Natural Science
Foundation of China (51676141) and the Fundamental Research
Funds for the Central Universities.
References
[1] P.A. Clausen, P. Wolkoff, Evaluation of automatic thermal desorption-capillary
GC for determination of semivolatile organic compounds (SVOCs) in indoor air,
J. High Resolut. Chromatogr. 20 (2) (1997) 99–108.
[2] M.D. Van Loy, W.J. Riley, J.M. Daisey, et al., Dynamic behavior of semivolatile
organic compounds in indoor air. 2. Nicotine and phenanthrene with carpet
and wallboard, Environ. Sci. Technol. 35 (3) (2001) 560–567.
[3] Y. Xu, J.S. Zhang, Understanding SVOCs, Ashrae J. 53 (12) (2011) 121–125.
 Y.-F. Mao et al. / International Journal of Heat and Mass Transfer 135 (2019) 685–695
[4] Y. Xu, Z. Liu, J. Park, et al., Measuring and predicting the emission rate of
phthalate plasticizer from vinyl flooring in a specially-designed Chamber,
Environ. Sci. Technol. 46 (22) (2012) 12534–12541.
[5] Y.R. Liang, Y. Xu, Improved method for measuring and characterizing phthalate
emissions from building materials and its application to exposure assessment,
Environ. Sci. Technol. 48 (8) (2014) 4475–4484.
[6] Y.R. Liang, Y. Xu, The influence of surface sorption and air flow rate on
phthalate emissions from vinyl flooring: Measurement and modeling, Atmos.
Environ. 103 (2015) 147–155.
[7] P.A. Clausen, V. Hansen, L. Gunnarsen, et al., Emission of Di-2-ethylhexyl
phthalate from PVC flooring into air and uptake in dust: Emission and sorption
experiments in FLEC and CLIMPAQ, Environ. Sci. Technol. 38 (9) (2004) 2531–
2537.
[8] T. Schripp, C. Fauck, T. Salthammer, Chamber studies on mass-transfer of di(2-
ethylhexyl)phthalate (DEHP) and di-n-butylphthalate (DnBP) from emission
sources into house dust, Atmos. Environ. 44 (24) (2010) 2840–2845.
[9] J.L. Benning, Z. Liu, A. Tiwari, et al., Characterizing gas-particle interactions of
phthalate plasticizer emitted from vinyl flooring, Environ. Sci. Technol. 47 (6)
(2013) 2696–2703.
[10] B. Lazarov, R. Swinnen, D. Poelmans, et al., Influence of suspended particles on
the emission of organophosphate flame retardant from insulation boards,
Environ. Sci. Pollut. Res. 23 (2016) 17183–17190.
[11] Y. Wu, C.M. Eichler, J. Cao, et al., Particle/gas partitioning of phthalates to
organic and inorganic airborne particles in the indoor environment, Environ.
Sci. Technol. 52 (6) (2018) 3583–3590.
[12] K. Hu, Q. Chen, J.H. Hao, Influence of suspended particles on indoor semi-
volatile organic compounds emission, Atmos. Environ. 79 (2013) 695–704.
[13] C. Liu, S. Shi, C. Weschler, et al., Analysis of the Dynamic Interaction Between
SVOCs and Airborne Particles, Aerosol Sci. Technol. 47 (2) (2013) 125–136.
[14] C. Liu, Y.P. Zhang, C.J. Weschler, The impact of mass transfer limitations on size
distributions of particle associated SVOCs in outdoor and indoor
environments, Sci. Total Environ. 497 (2014) 401–411.
[15] D.H. Bennett, E.J. Furtaw, Fugacity-based indoor residential pesticide fate
model, Environ. Sci. Technol. 38 (7) (2004) 2142–2152.
[16] H.M. Shin, T.E. McKone, D.H. Bennett, Intake fraction for the indoor
environment: a tool for prioritizing indoor chemical sources, Environ. Sci.
Technol. 46 (18) (2012) 10063–10072.
[17] X. Zhang, M.L. Diamond, C. Ibarra, et al., Multimedia modeling of
polybrominated diphenyl ether emissions and fate indoors, Environ. Sci.
Technol. 43 (8) (2009) 2845–2850.
[18] C. Liu, B. Zhao, Y. Zhang, The influence of aerosol dynamics on indoor exposure
to airborne DEHP, Atmos. Environ. 44 (16) (2010) 1952–1959.
[19] H.M. Shin, T.E. McKone, N.S. Tulve, et al., Indoor residence times of semivolatile
organic compounds: model estimation and field evaluation, Environ. Sci.
Technol. 47 (2013) 859–867.
[20] Y. Xu, J.C. Little, Predicting emissions of SVOCs from polymeric materials and
their interaction with airborne particles, Environ. Sci. Technol. 40 (2006) 456–
461.
[21] S. Shi, B. Zhao, Modeled exposure assessment via inhalation and dermal
pathways to airborne semivolatile organic compounds (SVOCs) in residences,
Environ. Sci. Technol. 48 (2014) 5691–5699.
[22] Y. Liang, C. Bi, X. Wang, Y. Xu, A general mechanistic model for predicting the
fate and transport of phthalates in indoor environments, Indoor Air 29 (2019)
55–69.
[23] Y. Xu, E.A. Cohen Hubal, P.A. Clausen, et al., Predicting residential exposure to
phthalate plasticizer emitted from vinyl flooring: a mechanistic analysis,
Environ. Sci. Technol. 43 (2009) 2374–2380.
[24] S. Shi, B. Zhao, Comparison of the predicted concentration of outdoor
originated indoor polycyclic aromatic hydrocarbons between a kinetic
695
partition model and a linear instantaneous model for gas–particle partition,
Atmos. Environ. 59 (2012) 93–101.
[25] S. Chen, G.D. Doolen, Lattice Boltzmann method for fluid flows, Annu. Rev.
Fluid Mech. 30 (1998) 329–364.
[26] Y.T. Mu, L. Chen, Y.L. He, et al., Pore-scale modelling of dynamic interaction
between SVOCs and airborne particles with lattice Boltzmann method, Build.
Environ. 104 (2016) 152–161.
[27] Y.F. Mao, Z. Li, Y.T. Mu, et al., Numerical investigation of SVOC mass transport
in a tube by an axisymmetric lattice Boltzmann method, Build. Environ. 128
(2018) 180–189.
[28] Y.F. Mao, Z. Li, Y.P. Zhang, et al., A review of mass-transfer models and
mechanistic studies of semi-volatile organic compounds in indoor
environments, Indoor Built Environ. 27 (10) (2018) 1–15.
[29] Y.L. He, Y. Wang, Q. Li, Lattice Boltzmann Method: Theory and Applications,
Science Press, Beijing, 2009.
[30] Y.H. Qian, D. Dhumieres, P. Lallemand, Lattice Bgk Models for Navier-Stokes
Equation, Europhys. Lett. 17 (1992) 479–484.
[31] C. Liu, G.C. Morrison, Y.P. Zhang, Role of aerosols in enhancing SVOC flux
between air and indoor surfaces and its influence on exposure, Atmos. Environ.
55 (2012) 347–356.
[32] A. Masselot, B. Chopard, A lattice Boltzmann model for particle transport and
deposition, Europhys. Lett. 42 (1998) 259–264.
[33] H.M. Wang, H.B. Zhao, Z.L. Guo, et al., Numerical simulation of particle capture
process of fibrous filters using Lattice Boltzmann two-phase flow model,
Powder Technol. 227 (2012) 111–122.
[34] H.M. Wang, H.B. Zhao, Z.L. Guo, et al., Lattice Boltzmann method for
simulations of gas-particle flows over a backward-facing step, J. Comput.
Phys. 239 (2013) 57–71.
[35] M.M. Kim, A.L. Zydney, Effect of electrostatic, hydrodynamic, and Brownian
forces on particle trajectories and sieving in normal flow filtration, J. Colloid
Interface Sci. 269 (2004) 425–431.
[36] Z.L. Guo, C.G. Zheng, B.C. Shi, Non-equilibrium extrapolation method for
velocity and pressure boundary conditions in the lattice Boltzmann method,
Chin. Phys. 11 (2002) 366–374.
[37] J.L. Benning, Z. Liu, A. Tiwari, et al., Characterizing gas-particle interactions of
phthalate plasticizer emitted from vinyl flooring, Environ. Sci. Technol. 47
(2013) 2696–2703.
[38] C.J. Weschler, W.W. Nazaroff, SVOC partitioning between the gas phase and
settled dust indoors, Atmos. Environ. 44 (2010) 3609–3620.
[39] B. Johnson-Restrepo, K. Kannan, An assessment of sources and pathways of
human exposure to polybrominated diphenyl ethers in the United States,
Chemosphere 76 (2009) 542–548.
[40] G. Beko, C.J. Weschler, S. Langer, et al., Children’s phthalate intakes and
resultant cumulative exposures estimated from urine compared with
estimates from dust ingestion, inhalation and dermal absorption in their
homes and daycare centers, PLoS One 8 (2013) 18.
[41] C. Liu, B. Zhao, Y.P. Zhang, The influence of aerosol dynamics on indoor
exposure to airborne DEHP, Atmos. Environ. 44 (2010) 1952–1959.
[42] N. Sugimoto, I. Uno, M. Nishikawa, et al., Record heavy Asian dust in Beijing in
2002: Observations and model analysis of recent events. Geophys. Res. Lett. 30
(12) (2003).
[43] A. Peters, J. Skorkovsky, F. Kotesovec, et al., Associations between mortality
and air pollution in Central Europe, Environ. Health Perspect. 108 (2000) 283–
287.
[44] C.Y.H. Chao, T.C.W. Tung, J. Burnett, Influence of different indoor activities on
the indoor particulate levels in residential buildings, Indoor Built Environ. 7
(1998) 110–121.
[45] L. Wallace, Indoor particles: A review, J. Air Waste Manag. Assoc. 46 (1996)
98–126.