Progress in Energy and Combustion Science 96 (2023) 101043

Contents lists available at ScienceDirect

Progress in Energy and Combustion Science

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

Modelling treatment of deposits in particulate filters for internal combustion

emissions
Chung Ting Lao a , Jethro Akroyd a,b , Markus Kraft a,b,c ,∗
a Department of Chemical Engineering and Biotechnology, University of Cambridge, West Cambridge Site, Philippa Fawcett Drive, Cambridge CB3 0AS, United
Kingdom
b
CARES, Cambridge Centre for Advanced Research and Education in Singapore, 1 Create Way, CREATE Tower, #05-05, Singapore, 138602, Singapore
c
School of Chemical and Biomedical Engineering, Nanyang Technological University, 62 Nanyang Drive, Singapore, 637459, Singapore

ARTICLE INFO ABSTRACT

Keywords: Internal combustion in transport vehicles is still one of the biggest contributors to ultrafine particle emissions
Emission which have been proven to have many adverse effects on human health and the environment in general. To
Ash mitigate this problem a variety of particle filters have been developed and along with these filters a whole
Computational model
range of models aiming to optimise filter performance. This paper reviews a wide variety of particulate filter
Diesel particulate filter (DPF)
models for vehicular emission control and presents the volume of work in a unified and consistent notation.
Gasoline particulate filter (GPF)
Particle filtration models are examined with respect to their filtration efficiency, the way they handle particle
deposits within the filter wall, the formation of filter cake and the role of catalytic conversion and the effect
of gaseous emission. Further, the impact of the chemical and physical properties of particulate deposits on
the filter regeneration process is analysed and reaction pathways and rates are presented. In addition the
accumulation of ash deposits and its impact on the filter behaviour is critically reviewed. Finally, various
measures are identified that can potentially improve the current particle filter models.

1. Introduction mode particles if they deposit on the surface of engine cylinders, which
are re-entrained into the exhaust at a later stage.
1.1. Particulate emissions from engines The major component of engine-out particles is soot [13]. This
can be loosely defined as hydrocarbon solid substances with an 8:1
Vehicular emission is often the biggest contributor towards ultrafine carbon-to-hydrogen ratio [14]. It is the product of the inception of
particle emissions in urban areas [1]. Particulate matter emitted by polyaromatic hydrocarbons (PAHs) in fuel-rich regions inside engine
vehicles can have adverse health and environmental impacts [2,3]. cylinders, where fuel pyrolysis occurs [5,15]. The major steps of soot
There is a great variety in the size, shape and composition of such formation are shown in Fig. 2. Primary particles formed at the early
particulates [4–6]. The properties of the particulates are influenced stages are approximately spherical [16]. Primary particles can grow
by the fuel composition, engine speed, engine load and combustion in size via condensation of gaseous species; larger aggregates can be
mode of the engine [7–10]. A typical particle size distribution of formed due to the coagulation of smaller primary particles [14,17].
particles emitted from an internal combustion engine (ICE) is shown This leads to a fractal-like aggregate structure as shown in Fig. 3. The
in Fig. 1. Three size classes may be identified: the nucleation mode
morphology of the aggregate may be further modified by oxidation
(below 50 nm), the accumulation mode (50 nm to 1 micron) and the
which can also lead to fragmentation [13].
coarse mode (above 1 micron) [11]. The relative proportion of particles
In addition, inorganic ash may be found in engine-out particu-
in each mode can vary significantly depending on the engine and its
lates [18]. This originates from additives doped in the fuel and the
operating condition [12]. Each particle mode can be approximated
lubricant oil. Ash may be found on the surface of aggregates or is
well by a log-normal distribution [13]. The nucleation mode is formed
embedded within the aggregate [19]. In addition, wear and corrosion
from nucleation of hydrocarbon and sulphates when the exhaust is
cooled down. The accumulation mode is composed of soot aggregates can lead to the formation of individual ash particles [20]. ICEs may
formed during the combustion process. They may grow into coarse also emit liquid condensate and semi-volatile particles as opposed to

∗ Corresponding author at: Department of Chemical Engineering and Biotechnology, University of Cambridge, West Cambridge Site, Philippa Fawcett Drive,
Cambridge CB3 0AS, United Kingdom.
E-mail address: [email protected] (M. Kraft).

https://doi.org/10.1016/j.pecs.2022.101043
Received 10 March 2022; Received in revised form 5 October 2022; Accepted 6 October 2022
Available online 8 February 2023
0360-1285/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Fig. 1. A typical particle size distribution from an Internal Combustion Engine (ICE).
Source: Adapted from Ref. [13].

Fig. 4. The structure of a typical Wall Flow Monolith (WFM) [37].

the respiratory system more easily than larger particles and are there-
fore more harmful [26]. Effective control of ultrafine particle emission
will become more important as the cut-off particle size for the PN limit
Fig. 2. Schematic diagram of soot formation process.
Source: Adapted from Ref. [14].
is expected to reduce from 23 nm to 10 nm [27].
In addition to more stringent emission standards, the emission
testing procedure has been improved to minimise differences between
laboratory emission measurements and real emission. Several studies
have indicated that over-simplistic drive cycles, for example, the New
European Drive Cycle (NEDC), can lead to underestimation of real
emissions [28–30]. Drive cycles developed based on data collected from
actual vehicles such as the Worldwide harmonized Light vehicles Test
Cycles (WLTC) were introduced to address this issue [31]. Furthermore,
the development of portable emission measurement systems (PEMS)
has enabled real driving emission (RDE) tests that measure the on-road
emissions from vehicles [32].
The updates of emission standards and testing protocols have been
motivating for wider applications of particulate filters and newer par-
ticulate filter technologies. Gasoline Particulate Filters (GPFs) are be-
coming more popular for GDI engines to comply with the Euro 6 PN
limit [33,34]. Moreover, port fuel injection (PFI) engines may need to
be installed with GPFs in the future as RDE studies have shown that PFIs
Fig. 3. A typical aggregate from Internal Combustion Engines (ICEs) [18]. Ash may can have comparable PN emission to GDI in cold environments [35,36].
be embedded within soot particles or attached to their surface.
The most common structure of particulate filters is the Wall Flow
Monolith (WFM) structure as shown in Fig. 4 [38]. The monolith is
composed of a set of alternately plugged porous walled channels. The
solid particles [21]. The semi-volatile particles are often composed of inlet channels are plugged at the rear and the outlet channels are
soluble organic fractions (SOFs) derived from unburnt fuel or lubricant. plugged at the front of the monolith. The engine-out exhaust enters the
Inorganic species such as sulphates and phosphates can also be of filter via the inlet channels and exits through the outlet channels. The
significant quantity depending on the fuel formulation [22]. porous wall between inlet and outlet channels traps the particles carried
by the exhaust gas. Particle deposition inside the porous space of the
1.2. Particulate filters wall leads to an increase in filtration efficiency but also an increase in
the pressure drop across the filter. As this process continues, further
in-wall particle deposition is prevented by pore closure due to already
As the negative impacts of particulates were realised, governing
deposited particles. As a result, a particle cake layer is formed on the
bodies around the world imposed limits on particulate emissions from
interface between the inlet channel and the porous wall instead.
ICEs [23]. The application of particulate filer was found to be effective
Regeneration refers to the process of burning off the trapped par-
in controlling engine-out particulate emissions. The Diesel Particulate
ticles. Regeneration is necessary to avoid high filter pressure drop
Filter (DPF) is now an essential component of diesel engine systems to (and associated fuel penalty) and overfilling of the filter [39]. ‘‘Active
comply with emission standards [24]. However, particular filters are regeneration’’ describes regeneration strategies that require a tempo-
not just relevant to diesel engines, but also to gasoline direct injection rary change in the operating condition of the filter. Typical active
(GDI) engines. regeneration increases the filter temperature to promote the oxidation
Emission standards have been continuously revised to be more strin- of trapped particles, usually with O2 as the oxidant. The additional heat
gent as the understanding of the negative impacts of particulates and may be provided by post-fuel injection [40], electric heaters [41] or
relevant measuring technologies improve [23]. The European Union microwaves [42,43]. It should be noted that ash is incombustible and
(EU) introduced limits on Particle Number (PN) emission in addition therefore remains in the filter after regeneration.
to the Particle Mass (PM) limit in an attempt to reduce emissions of On the contrary, ‘‘passive regeneration’’ occurs spontaneously when-
ultrafine particles [25]. This is because smaller particles may penetrate ever the operating condition is favourable without active control.

2
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Passive regeneration usually relies on NO2 as the oxidant, since it is approach is often adopted for both the gas phase and particle phase
a stronger oxidising agent than O2 which allows particle combustion in order to reduce the computational cost [69].
to occur at a lower temperature [44]. An oxidation catalyst is needed Channel-scale models focus on the relevant phenomena occurring
to produce NO2 for passive regeneration since engine-out NO2 con- at the scale of the channel diameter, O (μm) − O (mm). Single-channel
centration is insufficient [45]. It should be noted that GPFs operate models consider an inlet channel and an outlet channel of the wall-
at different conditions than DPFs and therefore their regeneration flow monolith. In the case of filters with asymmetric channel designs,
strategies differ significantly. Gasoline exhaust gas is generally at channel-scale models should simulate the ‘‘minimum repeating unit’’
higher temperatures and has lower oxygen contents than diesel exhaust of the filter; for example, three hexagonal channels, three rectangular
gas. Passive regeneration of GPFs with O2 is possible since the exhaust channels and a triangular channel are needed to be considered for the
temperature from gasoline engines is usually higher than that of diesel HRT filter design [59]. The gas flow and associated heat and mass
engines [46]. Spontaneous regeneration of GPFs can occur during transfer may be solved by commercial CFD software packages; this ap-
deceleration, at which point additional air flows through the engine proach is called the ‘‘3D single-channel model’’. Since the cross-section
and provides sufficient O2 for regeneration reactions [47]. of filter channels usually exhibit rotational symmetries, the channel
DPFs with catalytic coatings i.e. catalysed DPFs (CDPFs) were devel-
model may be simplified into 2D or 1D single-channel models [63].
oped in order to improve the regeneration performance [48]. Oxidation
Empirical correlations are often employed to describe convective heat
catalysts such as Precious Group Metals (PGM) may be coated inside
and mass transfer processes in 1D single-channel models [70]. The
or on the surface of the porous wall to promote passive regeneration
detailed structure of the porous wall is not fully resolved in channel-
via in-situ NO2 generation [49,50]. In addition, oxidation catalysts can
scale models. It is usually approximated as an assemble of simple
increase the reactivity of O2 regeneration and thus lower the ignition
geometries i.e. the ‘‘cell model’’ [71]. These are discussed in more detail
temperature of soot [51]. NOx removal (deNOx ) catalysts including lean
in the next section.
NOx trap (LNT) and selective catalytic reduction (SCR) catalysts can
be coated on DPF walls to reduce the volume of the overall exhaust Monolith-scale models consider an ensemble of multiple inlet and
after-treatment system [52,53]. Similarly, three way catalysts (TWC) outlet channels. They are developed to study the impact of inlet maldis-
are often coated on GPFs for the same reason [54]. tribution and interaction between filter channels [63]. The behaviours
In contrast to the conventional filter design shown in Fig. 4, alter- of inlet channels may not be identical if they are subjected to differ-
native structures have been developed in order to further improve the ent inlet conditions. Single-channel models cannot describe the filter
performance of particulate filters [45]. Increasing the cross-sectional behaviour adequately in these cases. The spatial domain of the filter
area of inlet channels can increase the particle storage capacity and model may need to be extended beyond the filter channels in order
achieve lower pressure drop at the same particle loading with respect to study the flow distribution effect [72]. Monolith-scale models are
to filters with symmetrical channel designs [55,56]. Filters with trian- generally expensive to solve given the large spatial domain involved.
gular and hexagonal cross-sections have been studied and they exhibit The choice of model depends on the purpose, expected level of
different heat and mass transfer behaviour to channels with square detail, accuracy required and computational cost limitation of the
cross-sections [57]. More sophisticated filter designs where channels particular application. Single-channel models are the most popular in
may have more than one type of cross-section have also been ex- the scientific literature since they provide a good compromise between
plored, for example, the Octo-Square (OS) design [58] and Hexagonal– predictive power and computational cost [61]. Pore-scale models are
Rectangular–Triangular (HRT) design [59]. They have shown better becoming more popular as designers are interested in optimising the
performance than the traditional design and they are not limited to the catalytic coating and microstructure of the porous wall for better
1:1 ratio of inlet and outlet channels [60]. filtration and pollutant conversion [68,73]. Multi-channel models are
valuable in assisting designs of inlet cones as well as investigation of
1.3. Particulate filter models non-uniform filter behaviour [74].

Many particulate filter models have been developed in order to

1.4. Scope
reduce the cost of the design process and to interpret experimental
results and gain more insight into filter behaviour [48,61–63]. Ade-
quate consideration of phenomena occurring at multiple length scales Many balance equations, including those of momentum, thermal
is essential for accurate prediction of filter behaviour. Particulate filter energy, species and particle population need to be solved by filter mod-
models can be put into three main categories based on the length-scale els. The process described by these equations can be influenced by the
of interest, namely pore-scale, channel-scale and monolith-scale [62]. properties of incoming and deposited particles. Adequate description of
Depending on the length-scale, the balance equations describing mo- the particle phase is essential for an accurate filter model. This is not
mentum, thermal energy, species, and particle population are solved in an easy task given the wide range of size, morphology and composition
different levels of detail. exhibited by engine-out particulates.
Pore-scale models focus on processes happening within individ- This review seeks to answer the following questions:
ual pores of the filter wall. A prerequisite of meaningful pore-scale
simulations is to consider ‘‘a sufficiently large domain that contains 1. How have the state and the impact of particles been considered
statistical representation of the morphological features present in the stud- in existing filter models?
ied system’’ [64]. Experimental techniques such as X-ray tomography 2. What level of detail is needed in order to adequately describe
(XRT) can be used to provide a detailed three-dimensional topology the impact of particles on filter behaviour?
of the porous wall as an input to pore-scale models [65,66]. The gas
flow through the porous network can either be solved by Computa- The rest of this paper is structured as follows: Modelling of the
tional Fluid Dynamics (CFD) approaches or Lattice Boltzmann Methods filtration process and the influence of trapped particles on the filter
(LBM) [64,67]. The trajectories of individual particles can be solved behaviour are described in Section 2. Filter regeneration models and
by Lagrangian particle tracking [68]. Since the characteristic length the treatment of particle deposit during regeneration are described in
of pore-scale models is O (nm) − O (μm), high spatial resolution is Section 3. The accumulation of incombustible ash deposits and their
needed for convergence, and hence pore-scale models are generally impact on the filter behaviour is described in Section 4. Finally, the
very demanding of computational resources. The Eulerian–Eulerian review is summarised in Section 5 with suggestions for future work.

3
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

𝐶3 ≈ 1.1 [84]. The Knudsen number Kn is the ratio of the mean free
path of gas molecules to the particle radius [83]
2𝜆
Kn = , (4)
𝑑p
where the mean free path 𝜆 is calculated by
Fig. 5. The build-up of the soot cake is initially non-uniform.
√
𝜇 𝜋𝑀 r
𝜆= , (5)
𝜌 2𝑅𝑇
2. Modelling particle filtration
where 𝜌 is the gas density, 𝑀 r is the average molecular weight of the
The soot deposition process in particulate filters may be divided gas, 𝑅 is the universal gas constant and 𝑇 is the gas temperature.
The Brownian diffusion force 𝐹B may be described as [67]
into four stages as shown in Fig. 5. The first stage is called the ‘‘deep
( )
bed filtration’’ or ‘‘depth deposition’’ stage where particles are trapped 3𝜋𝜇 𝑣 − 𝑣p 𝑑p √
within the filter wall. As particles deposit within the filter wall, they can d𝑊
𝐹B = 2𝐷p , (6)
block the pores on the surface of the wall. This forces incoming particles 𝐶c d𝑡
to build upon the wall surface instead. At first, lumps of particles where 𝑊 is a 3D Wiener process and 𝐷p is the diffusion coefficient of
are observed over the blocked pores [75]. These are referred to as particles [85]
‘‘particle trees’’ and this stage is called the ‘‘particle tree growth’’ stage. 𝑘B 𝑇 𝐶 c
The thickness of the cake layer (which is non-uniform at this point) 𝐷p = , (7)
3𝜋𝜇𝑑p
is increasing most rapidly at this stage [76]. As particle deposition
where 𝑘B is the Boltzmann constant.
continues, the particle trees being to grow laterally, where particles are
Eq. (1) may be called the Langevin equation when the Brown-
depositing on the ceramic surface [75]. The thickness of the cake layer
ian diffusion force is included [67]. The thermophoretic force may
grows more slowly at this ‘‘particle tree connection’’ stage. The soot
be included as a deterministic term in Eq. (1) as shown by Matte-
cake becomes uniform by the end of the connection stage, at which
Deschênes et al. [86], but its contribution is only significant when
point the rate of soot cake thickness growth becomes steady [76]. The
extreme temperature gradient occurs, which is rarely seen in practical
onset soot cake thickness at the beginning of this ‘‘soot cake layer’’ stage
devices. Pore-scale models often assume particles are trapped once they
has been found to be between 20–50 μm [75,76].
make contact with the filter wall with 100% efficiency [68,86]. This
is generally true except for sub−10 nm particles which can bounce off
2.1. Filtration efficiency of clean filters from the wall surface due to the high kinetic energy-to-mass ratio [87].
Solving for the particle mass concentration is an alternative to
The filtration process is controlled by the properties of the filter solving Eq. (1) in order to reduce computational demand [73,88,89].
medium, the particles being filtered and the carrying fluid [77]. The The local mass fraction of trapped particles is given by the following
ideal gas law is commonly used to describe the gaseous exhaust car- equation, which is used in conjunction with the Lattice Boltzmann
rying engine-out particulates [78–80]. The gas flow across the filter Method (LBM) model [90]
wall is determined by its tortuous internal pore network which is highly ( ) ( ) ∑ ( )
𝑌C,s x, 𝑡 + 𝛥𝑡 = 𝑌C,s x, 𝑡 + 𝑓C,g x, 𝑡 ⋅ 𝑝D , (8)
related to the manufacturing process [81]. g
The movement of particles may be simulated by solving the La- ( )
grangian equation of motion for the particles [82] where 𝑌C,s x, 𝑡 is the mass fraction of soot particles in the solid phase
at position x and time 𝑡. 𝑓C,g is the distribution function of soot mass
d𝑣p ∑ fraction in the LBM model. The sum of all 𝑓C,g gives the soot mass
𝑚p = 𝐹, (1)
d𝑡 fraction in the gas phase. The deposition probability 𝑝D determines the
where 𝑚p is the mass of a particle, 𝑣p is the vector of particle velocity number of particles trapped when they flow through a filtering node.
∑ Particles that are not trapped will bounce back into the gas flow [89].
and 𝐹 is the resultant force acting on the particle, which includes a
drag force that opposes the motion of the particle, a Brownian motion The deposition probability is calibrated based on experimental data and
force that causes random movement of the particle, and a gravitational treated as a constant [73].
force that pulls the particle towards the Earth. These forces can affect In channel-scale models, particles are usually assumed to distribute
the filtration efficiency of a particle by affecting its movement and homogeneously in the carrying gas and follow the streamlines of the
hence the probability of a collision between the particle and the fil- gas [80,91]. The ‘‘unit-cell’’ approach is commonly used to describe
tering medium. The contributions to the force from drag and Brownian the filtration process in channel-scale particulate filter models [61,69].
motion are usually considered [67], whereas gravitational force can be The complex internal structures of the filter wall are represented by
ignored for engine-out particulates [82]. an ensemble of particle ‘‘collectors’’ with simple geometry e.g. spheres
(for extruded ceramic filters) or cylinders (for fibrous filters) [92,93].
The drag force of a particle 𝐹 D may be represented by [83]
( ) The diameter of a spherical collector 𝑑c can be estimated from the pore
3𝜋𝜇 𝑣 − 𝑣p 𝑑p diameter 𝑑pore and the porosity of the filter wall 𝜖w . The pore diameter is
𝐹D = , (2) defined as a hydraulic diameter i.e. the diameter of a sphere conserving
𝐶c
the wet surface area and the void volume of the filter wall [77,84]:
where 𝜇 is the dynamic viscosity of the carrying fluid, 𝑣 is the ve- ( )
3 1 − 𝜖w
locity of the carrying fluid, 𝑑p is the particle diameter and 𝐶c is the 𝑑c = 𝑑pore . (9)
2𝜖w
Cunningham correction factor
( [ ]) The rest of this paper only considers models for filter walls with
𝐶
𝐶c = 1 + Kn 𝐶1 + 𝐶2 exp − 3 , (3) homogeneous microstructure i.e. uniform pore size and porosity ev-
Kn
erywhere within the filter wall. Real filter wall substrates can be
where 𝐶1 , 𝐶2 and 𝐶3 are empirical parameters. Their values depend highly non-uniform with great local variations of both porosity and
on the carrying gas. It is common to assume that the carrying gas has pore size [94,95]. The readers are recommended to read publications
the same properties as air, in which case 𝐶1 ≈ 1.257, 𝐶2 ≈ 0.4 and by Gong and Rutland [77], Gong et al. [96] and Wang et al. [71] for

4
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Fig. 7. The unit collector model is essentially one-dimensional because the particle
flow across the filter wall is perpendicular to the gas flow in channel.

Fig. 6. Summary of filtration mechanisms in filter wall [99].

2.1.1. Unit collector model
Various expressions of the single collector efficiencies for each
filtration mechanism have been presented by different authors in the lit-
details of filtration models that consider variations in porosity and pore erature. A general expression for diffusional single collector efficiency
diameter. 𝜂B is
The instantaneous filtration efficiency of a filter wall represented as ( )1
𝜖 3 𝐶2
a bed of spherical collectors can be derived from classical filter model 𝜂B = 𝐶1 Pe , (12)
𝐾
theory [97,98]
[ ] where 𝐶1 and 𝐶2 are numerical coefficients. The Kuwabara function
3𝐶 𝜂 (1 − 𝜖) 𝑤 𝐾, which derives from the Kuwabara model of the flow around the
𝐸 = 1 − exp − S , (10)
2 𝜖𝑑c collector, is a function of the porosity 𝜖 [77]
where 𝐸 is the instantaneous filtration efficiency, 𝜂 is the collection 1
𝐾 = 1 − 1.8 (1 − 𝜖) 3 + (1 − 𝜖) − 0.2 (1 − 𝜖)2 , (13)
efficiency of a single collector, 𝜖 and 𝑤 are the porosity and the thick-
ness of the filter medium. The sticking coefficient 𝐶S is an empirical and the Peclet number Pe is
parameter introduced to calibrate the single collector efficiency 𝜂 to fit 𝑣𝑑
Pe = c , (14)
against experimental data [99], although it is frequently assumed to be 𝐷p
1. The instantaneous filtration efficiency of a filter wall with a fibrous
where 𝑣 is the gas velocity and 𝐷p is the particle diffusion coefficient
structure can be calculated as [100]
calculated as per Eq. (7). It should be noted that the particle flow
[ ]
4𝐶 𝜂 (1 − 𝜖) 𝑤 across the filter wall is perpendicular to the gas flow in channel as
𝐸 = 1 − exp − S , (11)
𝜋 𝜖𝑑f shown in Fig. 7. Consequently, unit collector models are essentially
one-dimensional, and hence all velocities mentioned here and below
where 𝑑f is the diameter of a fibre. In some modelling works, the
are also one-dimensional, in the direction of the particle flow.
effective thickness of the filter medium 𝑤 in Eqs. (10) and (11) may
Literature models for diffusional filtration differ in the following
not be the actual thickness of the entire medium. It has been observed
aspects:
in experiments that soot particles only penetrate filter walls superfi-
cially [101]. A filter wall may be divided into two layers in general, 1. The values of 𝐶1 and 𝐶2
where the top layer facing the inlet channel is responsible for soot 2. Definition of gas velocity 𝑣 within Pe in Eq. (14)
filtration and the bottom layer is effectively clean (traps no particles). 3. Length scale used for slip-correction in the calculation of 𝐷p
For example, Serrano et al. [99] assumed that filtration only occurs ((3) and Eq. (7))
in the top 2% to 10% of the filter wall. In such models, the filtration
( )2
process is omitted in the remainder of the filter medium. This modelling According to the derivation by Lee and Gieseke [97], 𝐶1 = 2 3𝜋
3
4
approach will be called the ‘‘two-layer’’ approach in the rest of this ≈ 3.5. Most existing models adopted the approximate value of 3.5
paper. whilst Gong and Rutland [77] has used the exact value. An additional
The gas flow around a single collector or fibre can be described empirical multiplier was introduced by Karamitros and Koltsakis [80]
analytically by the Kuwabara and Happel models [77]. These two in order to adjust the value of the collector efficiency to capture the
models assume different boundary conditions at the surface of the behaviours of different substrate materials. The index of the Peclet
collector. the Happel model assumes that the tangential stress vanishes number 𝐶2 is usually − 23 [99,103,104]. Recent studies have shown that
at fluid–solid boundary whereas the Kuwabara model assumes that the 𝐶2 = − 12 is better suited for non-spherical collectors [105].
vorticity vanishes [93]. The filtration efficiency of a single collector via The Peclet number may be defined by either the superficial velocity
a specific filtration mechanism can then be derived. The three major 𝑈 which is the through-wall gas velocity at the face of the filter wall,
filtration mechanisms, namely Brownian diffusion, interception and or the interstitial velocity 𝑢 which is the actual gas velocity within the
inertial impaction, are shown in Fig. 6. The influences of gravitational pores. They are related by
forces and thermophoresis are neglected in the unit collector model. 𝑈
𝑢= . (15)
Brownian diffusion considers particle deposition onto the collector 𝜖
due to random collision with gas molecules. Interception considers Both definitions are used in the literature. Gong and Rutland [77],
particles that are trapped on the collectors because they follow along Liu et al. [106] and Tan et al. [107] all used superficial velocity
gas streamlines that lie closely to the surface of collectors [102]. Inertial whereas Konstandopoulos et al. [38], Serrano et al. [108] and Payri
impaction considers the filtration of particles that are too large to et al. [103] used interstitial velocity. Since the original formulation
follow the streamlines that cannot change direction fast enough to developed by Lee and Gieseke [97] defined the Peclet number using the
avoid colliding with the collector. velocity at the boundary of the collectors, it may be more appropriate

5
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

to define the Peclet number using the interstitial velocity for diffusional where 𝜌p is particle density and 𝜌g is gas density. Gong and Rutland
filtration. Using different definitions can lead to a factor of 2 difference [77] defined both Reynolds and Stokes numbers using the superficial
in the value of 𝜂B because the typical porosities of particulate filters are velocity 𝑈 .
around 50% [109]. Serrano et al. [99] has the following expression for 𝜂I which is
The correction for slip phenomena (Eqs. (3) and (7)) requires the inherited in later works [104,108]
Knudsen number to be calculated (Eq. (4)). Whilst most literature mod- St ′2
els have used the particle diameter to define the Knudsen number, Payri 𝜂I = , (23)
(0.25 + St ′ )2
et al. [84] used the pore diameter of the filter wall as the relevant
where a different definition was used for the Stokes number St ′
length scale for the slip phenomena. This practice is adopted by fol-
lowing modelling works [99,103,104,107,108]. Since the diffusion of 𝜌p 𝑑p2 𝑢
St ′ = 𝐶c , (24)
the particle is due to random collisions with gas molecules, the relevant 9𝜇𝑑c
interface when slip occurs would be the surface of the particle. Hence where the Cunningham correction factor 𝐶c is calculated based on
the particle diameter should be used in slip correction calculation. the pore diameter of the filter medium. In contrast to Eq. (21), the
The interceptional single collector efficiency has been derived by interstitial velocity 𝑢 is used in defining the Stokes number in Eq. (24).
Lee and Gieseke [97] to be This is thought to be more appropriate as the interstitial velocity would
𝑁R2 be the actual gas velocity ‘‘experienced’’ by the particle. The use of
3 𝜖
𝜂R = ( ) , (16) interstitial velocity also incorporates the influence of filter wall porosity
2 𝐾 1 + 𝑁R 𝑠
into the inertial filtration efficiency expression.
where the interception parameter 𝑁R is the ratio between the particle
diameter 𝑑p and the collector diameter 𝑑c 2.1.2. Fibrous filter filtration model
Significantly fewer modelling works have considered particulate
𝑑p
𝑁R = . (17) filters with fibrous walls. The following expression was used by Wang
𝑑c et al. [93] and Liu et al. [106] for diffusional single fibre efficiency 𝜂Bf
The interception exponent 𝑠 = 3−2𝜖3𝜖
is sometimes approximated to
( )1
be 2 [77], which implies 𝜖 ≈ 83 (lower end of typical wall poros- 𝜖 3 2
𝜂Bf = 1.6 Pe− 3 𝐶B1 𝐶B2 , (25)
ity [109]). Karamitros and Koltsakis [80] introduced an additional 𝐾f
empirical multiplier to 𝜂R in their model, similar to their treatment for
where 𝐾f is the Kuwabara function for fibrous filters
𝜂B to account for different substrate materials.
An alternative formulation for interceptional single collector effi- 𝐾f = −0.5 ln (1 − 𝜖) + (1 − 𝜖) − 0.25 (1 − 𝜖)2 − 0.75. (26)
ciency was proposed by Wurzenberger et al. [110]. The authors adapted
The coefficients of diffusional single fibre efficiency 𝐶B1 and 𝐶B2 are
expressions from Ohara et al. [102] (who only presented expressions
for single collector efficiencies for the limiting cases Re → 0 and
( )1
Re → ∞) and obtained the following expression based on fitting against 𝜖Pe 3
𝐶B1 = 1 + 0.388Kn , (27)
experimental data 𝐾f
( )1
𝜂R = 1.5𝑁R + 10−7 𝑁R2 . (18) 1 𝜖 3 2
= 1 + 1.6 Pe− 3 𝐶B1 , (28)
𝐶B2 𝐾f
This formulation does not consider the impact of the filter medium
porosity. As a result, Eq. (18) is expected to only work for highly porous where the Peclet number Pe and Knudsen number Kn are calculated us-
filter media with collectors that are far from each other and can be ing the particle diameter 𝑑p and the superficial velocity 𝑈 . The Knudsen
approximated to be isolated. number is included to account for the slip effect of gas molecules at the
Inertial impaction is often neglected in filtration models for par- surface of particles [106]. Ou et al. [112] used a variation of Eq. (25)
ticulate filters. Jiang et al. [111] showed that inertial impaction can without 𝐶B2 where the fibre diameter 𝑑f and the superficial velocity 𝑈
were used to define 𝐶B1 , Kn and Pe.
be neglected for particles below 1 μm. This means that the inertial
Wang et al. [93], Liu et al. [106] and Ou et al. [112] used the
impaction is only relevant for coarse mode particles which are a minor
following expression for the interceptional filtration in fibrous filters
fraction of engine-out particulates. However, Liu et al. [106] suggested
that inertial impaction is non-negligible due to the low porosities of ( 2
)( )
ceramic filters. Gong and Rutland [77] used the following expression 𝜖 𝑁Rf Kn
𝜂Rf = 0.6 1 + 1.996 , (29)
for single collector efficiency due to inertial impaction 𝐾f 1 + 𝑁Rf 𝑁Rf

St 3ef f where Kn is defined by the fibre diameter. The interception parameter

𝜂I = , (19) for fibrous filters 𝑁Rf is similar to that of the unit collector model
(0.0014 + St ef f )3
except for the collector diameter 𝑑c in Eq. (17) is replaced by the fibre
where the effective Stokes number St ef f incorporates the influences of
diameter 𝑑f .
porosity 𝜖 and Reynolds number Re into the original Stokes number. Babaie Rabiee et al. [82] used a different expression for 𝜂Rf
The effective Stokes number is
[ ] 1 + 𝑁Rf [ ( )
𝜖 𝜂Rf = 2 ln 1 + 𝑁Rf − 𝜖
St ef f = 1 + 1.75Re St, (20) 2𝐾f (30)
150(1 − 𝜖) ( )−2 ( 1−𝜖
)
1−𝜖 ( )2 ]
+ 1 + 𝑁Rf 1− − 1 + 𝑁Rf ,
where the original Stoke number St is 2 2
𝜌p 𝑑p2 𝑈 which is derived using the Kuwabara flow field by Liu and Wang
St = , (21) [113].
18𝜇𝑑c
The single fibre efficiency due to inertial impaction is modelled
and the Reynolds number Re is as [112]
𝜌g 𝑈 𝑑 c St
Re = , (22) 𝜂If = , (31)
𝜇 4𝐾f2

6
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

where the Stokes number is defined as may be treated as spheres in the model with equivalent particle di-
𝑚p 𝑈 ameter and density [67]. Viswanathan et al. [116] found that the
St = 𝐶, (32)
3𝜋𝜇𝑑p 𝑑f c mobility diameter is the determining parameter for filtration behaviour.
Additional consideration of the shape of the particle is unnecessary
where 𝑚p is the mass of a particle. Eq. (32) relaxed the assumption
since it is a function of the mobility diameter. The mobility diameter
that the particle is spherical. Eq. (21) can be derived from Eq. (32) by
𝜋𝑑 3 𝑑mob of engine-out aggregates may be measured using a differential
substituting 𝑚p = 𝜌p 6 p . mobility analyser (DMA). It may be estimated if the number 𝑛pri and the
Babaie Rabiee et al. [82] adopted expressions from Brown [100] for
diameter 𝑑pri of primary particles within the aggregate are known [117]
inertial impaction in fibrous filters:
St [( ) 2 ]
𝜂If = 29.6 − 28 (1 − 𝜖)0.62 𝑁Rf 2.8
− 27.5𝑁Rf , (33) 1
(2𝐾f )2 𝐷
𝑑mob ∝ 𝑑pri 𝑛primm , (37)
for low St and when 𝑁Rf < 0.4, and
where 𝐷mm is the mass-mobility exponent. The value of 𝐷mm depends
( )−1
𝜂If = St 3 St 3 + 0.77St 2 + 0.22 , (34) on the flow regimes as well as the number of primary particles within
the aggregate [117].
for moderate St. Brown [100] described scenarios of low and high
Stokes numbers without stating the exact ranges of low, moderate and There are concerns about whether the mobility diameter is the
high Stokes numbers. The inertia of the particle can cause a slight correct length scale for all filtration mechanisms. A study conducted
perturbation on the particle’s trajectory and affect the interception by Chen et al. [118] on particle filtration of a nuclepore filter suggested
efficiency at low Stokes numbers. When the Stokes number is large, that the maximum projected length of the particle should be used
the trajectory of the particle is almost a straight line. Babaie Rabiee to calculate the interceptional filtration efficiency to account for the
et al. [82] applied Eqs. (33) and (34) for 0.01 ≤ St ≤ 10 without further elongated shape of aggregates. This approach is adopted in the work
clarification. of Tan et al. [104] on DPFs. In their work, the maximum diameter
is calculated using an empirical correlation considering the engine
2.1.3. Combining different filtration mechanisms operating conditions.
In order to calculate the instantaneous filtration efficiency using In addition to the size of the particles, it is also important to describe
Eqs. (10) or (11), the overall single collector (fibre) efficiency needs the mass of the particles since it will affect the effect of forces acting
to be calculated by combining the single collector (fibre) efficiencies
upon the particles (in particle tracking schemes) and the mass-based
due to each filtration mechanism. Most filter models in the literature
filtration efficiency which is a common metric used in emission stan-
assume each filtration mechanism acts independently; this is adopted
dards. The material density of engine-out particles have been reported
owing to mathematical difficulties [77]. Ou et al. [112] considered the
or assumed to be in the range of 1500–2500 kg/m3 [44,68,91,111,119–
enhancement of interception due to Brownian motion, where particles
that would not have been trapped by pure interception are filtered be- 122]. The mass of the particle can then be calculated if the number and
cause they are ‘‘pushed’’ towards the collector surface due to Brownian the diameter of primary particles are also known.
motion. The additional single fibre efficiency 𝜂BRf is As mentioned above, most filtration models treat aggregates as
2 spherical particles with equivalent mobility diameters. Hence the ef-
1.24𝑁 3 fective density of a particle is required to correctly calculate the mass
𝜂BRf = √ Rf . (35)
𝐾f Pe of a particle. The effective density of aggregates is usually described as
a power-law relationship [104,121,123]
Many authors have combined single collector (fibre) efficiencies by
( )
treating them as probabilities [103,104,110,114], for example [108] 𝑑mob 𝑞−3
𝜌ef f = 𝐶𝜌pri . (38)
( )( )( ) 𝑑pri
𝜂 = 1 − 1 − 𝜂B 1 − 𝜂R 1 − 𝜂I . (36)
where 𝐶 is a constant and 𝑞 is fractal dimension according to Gong et al.
However, it can be seen from the equations above that the ex-
[123] and mass-mobility exponent according to Liu et al. [121] and Tan
pressions used to describe the single collector (fibre) efficiencies (Eqs.
(12), (16), (25) etc.) are not bounded between 0 and 1. As a result, et al. [104]. Since the effective density is a function of the mobility
applying the independence rule of probabilities may yield unphysical diameter, defining 𝑞 as the fractal dimension will be inappropriate since
values for the overall single collector efficiency. According to the early it is a pure geometric parameter [117]. Other empirical formulas to
work of Lee and Gieseke [97], the diffusional single collector efficiency relate the effective density and the size of aggregates have been used
was defined as ‘‘ the ratio of the rate at which particles diffuse to the in the literature [67,123].
sphere surface to that at which particles approach toward the sphere surface
within the cross-sectional area of the sphere’’. A simple additive rule
may be more suitable for combining single collector efficiencies since 2.2. Particle deposition within filter walls
they are not probabilities by definition, but rates of particle collection
normalised by the rate of particle approaching collectors. The void volume within the filter wall reduces as particles are
Some authors introduced additional multipliers to single collector trapped in the filter wall. This can influence the instantaneous filtration
efficiencies to consider the interaction between the particles and the
performance [95] and generally leads to an increase in both filtration
filter wall. Wang et al. [71] introduced calibrating constants to single
efficiency and pressure drop [38,124].
collector efficiencies for diffusion 𝜂B and interception 𝜂R respectively.
The authors suggested that the introduction of these constants allows In pore-scale models, the deposition of particles can be captured by
the model to account for the effect of the pore morphology of the filter redefining the ‘‘wall’’ region. For example, Yamamoto and Sakai [125]
wall. tracked the local particle mass concentration within the filter wall. As
The size of a particle is characterised by its diameter in filtra- it reaches unity, the local node is considered as a solid phase instead of
tion models. Treating particles as spheres in filter models is the most the gas phase. An additional frictional force is applied to the gas flow
straightforward approach [115]. However, engine-out particulates of- in the solid region (since soot deposits are porous) [126]. This then has
ten are fractal-like aggregates. To overcome this problem, real particles a subsequent impact on the gas flow pattern and the pressure drop.

7
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

2.2.1. Spherical unit collector model where 𝑃 is the system pressure, 𝑃0 is the reference pressure at which
In channel-scale models coupled with the spherical unit collector the clean wall permeability 𝜅w0 is calibrated and 𝐶 is a constant to be
model, the build-up of particles within the filter wall is reflected by an determined. Other empirical formulas exist to account for the reduction
increase in the collector size. The diameter of the collector 𝑑c is often in wall permeability due to soot loading, for example [137]
treated as a function of the mass of trapped particles [107,110] ( √ )
1 𝑃 𝑇
[ ( )] 1 𝜅w = 1 + 𝐶3 0 𝜇 , (46)
3 𝑚sw 𝑑c0 3 1
+ 𝐶 𝜌 + 𝐶 𝜌2 𝑃 𝑀
𝑑c = 2 + , (39) 𝜅w0 1 p 2 p r
4𝜋 𝜌sw 2
where 𝜌p is defined as the ratio of the mass of the particles to the pore
where 𝑑c0 is the collector diameter without any trapped particles, 𝑚sw
volume of the wall. 𝐶1 , 𝐶2 and 𝐶3 are constants that require calibration.
is the mass of particle deposits and 𝜌sw is the packing density of particle
deposits within the filter wall. The packing density 𝜌sw is calibrated to
match experimental pressure drop data [127]. Konstandopoulos et al. 2.2.2. The two-layer approach
[128] showed that the value of this parameter may scale with the local In channel-scale models, the filter wall is divided into slabs with
Peclet number. each slab having a local collector size in order to reproduce the particle
As the diameter of the collector increases, the filtration efficiency of deposition profile across the wall [38]. In the work of Serrano et al.
the wall changes according to equations presented in Section 2.1. The [99], the porous wall was divided into two slabs. Only the top slab
local wall porosity 𝜖w which is another important input for the filtration that is adjacent to the inlet channel is responsible for particle filtra-
model can be calculated from the local collector diameter tion whereas the remainder of the porous wall is always clean. This
( ) approach is based on the experimental observation that only superficial
( ) 𝑑c 3
𝜖w = 1 − 1 − 𝜖w0 , (40) soot penetration occurs in the filter wall [101]. Whilst this ‘‘two-layer’’
𝑑𝑐0
approach is able to reproduce the pressure drop trend during the fil-
where the subscript 0 denotes clean wall condition. tration process, its performance deteriorates significantly for scenarios
The corresponding pressure drop across the filter wall in a channel- that are not considered during calibration. Furthermore, the thickness
scale model can be generally described as [129,130] of the top slab i.e. the soot penetration thickness is an input to the
𝜇𝑤 filtration model. The penetration thickness may be estimated using a
𝛥𝑃 = 𝑣 + 𝛽𝜌g 𝑤𝑣2 , (41)
𝜅 correlation between the soot penetration thickness and the wall Peclet
where 𝛥𝑃 is the pressure drop, 𝑤 is the thickness of the porous number [99].
medium (wall), 𝜅 is the permeability of the porous medium (wall), 𝜌g All collectors within the top slab are assumed to behave identically
is the gas density and 𝛽 is the Forchheimer coefficient of the porous in the ‘‘two-layer’’ model. Serrano et al. [138] introduced a multiplier
medium (wall). The first-order term on the right-hand side represents called ‘‘shape factor’’ to account for irregular (non-spherical) increase
the viscous loss and the second-order term on the right-hand side rep- of the collector size. Eq. (39) then becomes [103]
resents the inertial loss [131]. The permeability 𝜅 and the Forchheimer [ ( )] 1
coefficient 𝛽 can be calibrated based on experimental pressure drop 3 𝑚sw 𝑑c0 3
𝑑c = 2 + , (47)
data measured at various flow rates [131]. 4𝜋 𝜒𝜌ef f 2
The Forchheimer coefficient of a porous medium can be derived where 𝜌ef f is the effective density of individual aggregates described
from Ergun’s equation [130] in Section 2.1. Serrano et al. [138] suggested that its value should be
𝛽=
1.75 1 − 𝜖
. (42) 345 kg/m3 to represent aggregates with medium fractal dimension and
𝑑c 𝜖 3 number of primary particles. The shape factor 𝜒 is bounded between
The inertial momentum loss across porous media is often omitted 0 and 1, where a value of 1 represents the perfect spherical growth of
in filter models due to low wall Reynolds number [69,79,132], albeit the collector. The shape factor 𝜒 was found to increase with the soot
some experiments has shown that it has a non-negligible impact [133]. load and is defined as [103]
As particles deposit within the filter wall, its permeability (𝜅 in −𝐶
𝜒 = 𝐶1 𝛷soot2 , (48)
Eq. (41)) will reduce and hence lead to an increase in pressure drop.
The unit collector model calculates the permeability of the clogged wall where 𝛷soot is a ‘‘soot density factor’’ and 𝐶1 , and 𝐶2 are fitting
𝜅w from the local porosity 𝜖w and collector diameter 𝑑c [127] coefficients. Their values were found to be 𝐶1 = 2.3136 and 𝐶2 = 0.864
( ) by Serrano et al. [138] after calibration against experimental results.
𝑑c 2 𝑓K (𝜖w )
𝜅w = 𝜅w0 , (43) The soot density factor is defined as ‘‘the ratio between the soot packing
𝑑c0 𝑓K (𝜖w0 )
density inside the porous wall and the soot mass to soot penetration volume
where the Kuwabara geometric function 𝑓K (𝜖) is [134] ratio’’ [138]
9 1 1 ( )
1 2 − 5 (1 − 𝜖) − 𝜖 − 5 (1 − 𝜖)
3 2 𝑉sp −1
𝑓K (𝜖) = . (44) 𝛷soot = 𝜌ef f , (49)
18 1−𝜖 𝑚sw
It should be noted that some authors have used another version of where 𝑚sw is the mass trapped inside the wall and 𝑉𝑠𝑝 is the soot
Eq. (44) which has a factor of 4 difference from the equation reported penetration volume, defined as ‘‘the volume of porous wall which is used
here [38, for example]. This has no impact on Eq. (43) since only the by soot in the whole monolith’’ [138]
ratio of Kuwabara functions matters. However, one needs to be careful
about its impact on modelling the permeability of the clogged wall 𝑉sp = 4𝑑ch 𝐿𝑤w 𝑁ic 𝑤̂ sat (50)
(see Eq. (52)). The rest of the paper assumes the Kuwabara geometric
where 𝑑ch is the side length of square inlet channels, 𝐿 is the length
function has the form presented above.
of channels, 𝑤w is the thickness of the filter wall, 𝑁ic is the number
The permeability of the porous wall has been observed to change
of inlet channels and 𝑤̂ sat is soot penetration thickness normalised by
with temperature [135]. An additional term may be added to Eq. (43)
the total wall thickness. The permeability of the entire filter wall is the
to account for the change in permeability due to change in the mean
harmonic mean of the permeabilities of the effective filtration slab and
free path of the gas [136]
( √ ) the clean slab, weighted by their respective thicknesses
( )
𝑑c 2 𝑓K (𝜖w ) 𝑃 𝑇 𝜅w,clog 𝜅w0
𝜅w = 𝜅w0 1+𝐶 0𝜇 , (45) 𝜅w = , (51)
𝑑𝑐0 𝑓K (𝜖w0 ) 𝑃 𝑀 𝑤̂ sat 𝜅w0 + (1 − 𝑤̂ sat )𝜅w,clog
r

8
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

where 𝜅w,clog is the permeability of the clogged part i.e. the top slab.
It is calculated based on the porosity and the collector diameter of the
clogged wall

𝜅w,clog = 𝑓K (𝜖w )𝑑c2 𝐶c , (52)

where variables in Eq. (52) can be calculated based on Eqs.

(3), (40), (44) and (47).

2.2.3. The kamp’s and sappok’s model

Kamp et al. [139] and Sappok [140] used a simple model to estimate
the permeability of the filter wall when it has soot deposits Fig. 8. The clogged portion and the clean portion of the fibrous filter wall is considered
1 1 1 to form parallel resistance to gas flow across the wall [93].
= + 𝑉̂soot , (53)
𝜅w 𝜅w0 𝜅soot
where 𝑉̂soot is the volume ratio of soot deposit and the total porous
space within the wall. This model requires knowledge of the permeabil- where subscript ‘‘sd’’ denotes soot dendrite. The single collector effi-
ity of soot deposits 𝜅soot since it treats soot deposits within the filter ciency 𝜂 may be calculated as described in Section 2.1.2 using the
wall as a layer itself. Since soot deposits are not physically a layer dendrite diameter and metal fibre diameter respectively. The dendrite
within the filter wall, the value of 𝜅soot may only be calibrated to match diameter was chosen to fit against experimental pressure drop by Wang
experimental pressure drop data. It was calibrated to be 9.6 × 10−16 m2 et al. [93]. The quality of the model remains to be checked against
in the work of Zhang et al. [141]. experimental observations of the dendrites.
As the metal fibres and the soot dendrites are assumed to act as fil-
2.2.4. Summary of in-wall deposition treatment in unit collector models ters in parallel, they are exposed to different portions of in-flow. Wang
Three groups of modelling approach for in-wall deposition which et al. [93] assumed that the fraction of flow across the dendrite is
occurs during the deep bed filtration stage have been described above. proportional to the volume it occupies in the filter wall. The pressure
The one used by Kamp et al. [139] and Sappok [140] is the only one drop across a slab of the clogged filter is
with a single degree of freedom (𝜅soot ). The other two models have two √( )(
√ )
√ 𝛼 𝛼 𝛼sd 𝛼f ( ( )3 )
degrees of freedom. The predictions of the two-layer model developed 𝛥𝑃 = 64𝜇𝑈 𝑤w √ sd + f + 1 + 56 𝛼sd + 𝛼f . (56)
by Serrano et al. [99] are controlled by the effective density of individ- 𝑑 2 𝑑 2 𝑑 𝑑
sd f sd f
ual aggregates 𝜌ef f and the ratio between soot penetration thickness
and the total wall thickness 𝑤̂ sat . The filter behaviour predicted by 2.3. Formation of the particle cake layer
the standard unit collector model with multiple wall discretisation is
controlled by the packing density of particle deposits within the filter A soot cake is formed on the interface between the inlet channel
wall 𝜌sw and the number of discretisations 𝑛wall . and the filter wall when particles deposited in the filter wall prevents
As 𝑛wall increases, the calibrated value of 𝜌sw is expected to converge further in-wall deposition. The soot cake can filter incoming particles
asymptotically. However, the numerical convergence with respect to effectively once it is established.
𝑛wall has not been studied extensively. Researchers often choose the The simplest model to describe the transition from deep bed filtra-
value of 𝑛wall a priori without justification [136]. One needs to be aware tion to soot cake formation would be the model used by Zhang et al.
of this practice and not assume that asymptotic convergence is achieved [141]. A parameter called ‘‘depth filtration threshold’’ is defined as the
without evidence. maximum amount (mass) of soot allowed to deposit within the filter
On the other hand, the parameters of Serrano et al. [99]’s model wall. All filtered particles are assumed to be trapped within the wall
can be experimentally measured. Their model however requires a good before the soot load reaches the depth filtration threshold. Once the
estimation of the penetration thickness of soot deposits as an input, filter wall is saturated, all subsequently filtered particles deposit on
which is supposed to be the output of the model. This limits the the soot cake. As a result, the deep bed filtration and the soot cake
predictability of the model. Furthermore, the soot deposition profile formation regime are distinct and non-overlapping.
within the filter wall cannot be reproduced by Serrano et al. [99]’s Most unit collector models describe the transition by defining a
model. dimensionless parameter called partition coefficient 𝛷 that lies between
0 (pure deep bed filtration) and 1 (pure soot cake formation). Two
2.2.5. Fibrous filter model
versions of the formula have been seen in the literature
Particles trap in fibrous filters will form dendrites in the wall which
2
𝑑c,interface 2
− 𝑑c0
have been shown to act like new fibres [142]. The fibrous filter wall
𝛷= ( )2 , (57)
was discretised into 35 slabs in the work of Wang et al. [93]. Each slab 𝜓𝑑c,max − 𝑑c0 2
of the fibrous filter wall was split into two regions, one representing
3
𝑑c,interface 3
− 𝑑c0
filter fibres without soot particles whereas the other representing soot
𝛷= ( )3 , (58)
dendrites within the fibrous wall. This is shown in Fig. 8. 𝜓𝑑c,max − 𝑑c0 3
Wang et al. [93] incorporated this effect into a model of a filter
composed on metal fibres. The filtration efficiency of the metal fibre where 𝑑c,interface is the collector diameter of the first filter wall slab i.e.
𝐸f is a variation of Eq. (11) the one that is closest to the inlet channel. The maximum collector
[ ] diameter 𝑑c,max is calculated by matching the porosity of the collector
−4𝛼f 𝜂f 𝑤w
𝐸f = 1 − exp , (54) bed to the porosity of the clean filter wall [127]
𝜋𝑑f 𝜖w
3
𝑑c0
where 𝛼f is the solid volume fraction of metal fibres, 𝜂f is the single 3
𝑑c,max = , (59)
fibre collection efficiency, 𝑤w is the filter wall thickness, 𝑑f is the 1 − 𝜖0
diameter of metal fibres and 𝜖w is the filter wall porosity. Similarly, and 𝜓 is a model parameter called the percolation constant. It controls
the filtration efficiency of the soot dendrites 𝐸sd is the collector diameter at the onset of pure soot cake formation and
[ ] its value is usually calibrated to match experimental pressure drop
−4𝛼sd 𝜂sd 𝑤w
𝐸sd = 1 − exp , (55) data [38].
𝜋𝑑sd 𝜖w

9
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

The approaches mentioned above have considered the filtration

efficiency of the soot cake as a function of the state of the filter wall
only. They have neglected the effect of the cake layer itself on the
filtration process. These approaches will not be able to describe the
filtration process correctly when an appreciable soot cake is present but
with a clean filter wall, which can occur during filter regeneration or
after a partial regeneration event [92]. Karamitros and Koltsakis [80]
considered the contribution towards soot cake growth from the filter
wall and the cake itself separately

𝐸cake = 𝐸mech1 + 𝐸mech2 − 𝐸mech1 𝐸mech2 , (61)

Fig. 9. The effect of the percolation constant on filter calibration. The crosses in the where 𝐸mech1 is the cake filtration efficiency due to pore blockage
figure indicates the beginning of the soot cake formation regime.. and 𝐸mech2 is the filtration efficiency due to the presence of the cake.
𝐸mech1 is assumed to be a linear function of the filter wall porosity and
𝐸mech2 is assumed to be a linear function of the soot cake thickness
The effect of the percolation constant is demonstrated in Fig. 9. respectively. The introduction of 𝐸mech2 allows the filtration capability
Increasing the value of the percolation constant allow more soot to of a developing soot cake to be modelled.
deposit within the filter wall during the deep bed filtration regime. As Premchand et al. [145] considered the initial soot cake formation by
a result, the transition to the soot cake formation regime occurs later interpreting the partition coefficient as the filtration efficiency of the
and at a higher filter pressure drop. soot cake at the early stage of its formation. This continues until the
Filter models that divide the filter wall into multiple slabs often permeability of the filter wall reached a pre-defined value, at which
adopt Eq. (57) [96,110,114,136]. This formulation implies that soot point Eq. (10) was applied to describe the filtration efficiency of the
cake formation is caused by the blockage of surface pores [63]. On soot cake. An additional multiplier was introduced by Premchand et al.
the other hand, filter models that adopted the two-layer approach use [145] to Eq. (10) to limit the maximum filtration efficiency of the soot
Eq. (58) to describe the transition from deep bed filtration to soot cake cake which is chosen to be 0.937. The collector diameter in the soot
formation [99,103]. The physical interpretation is that soot cake forma- cake is calculated using Eq. (9) with a user-specified pore diameter
tion is caused by full saturation of the filter wall. Both Eq. (57) and (58) defined as 100 nm in that work.
have been shown to match well with experimental data. Nonetheless, For fibrous filters, Wang et al. [93] modelled the cake filtration
the value of the calibrated percolation constant will be dependent on efficiency using Eq. (11) with cake-specific parameters. The onset of
the form of the partition coefficient equation and researchers should soot cake formation is determined by the soot volume fraction of the
be careful when they compare the values of percolation constants from top wall slab. This is analogous to Eq. (58). Once the soot volume
different studies. fraction reaches a pre-defined limit, the soot cake is formed assuming
Both the partition coefficient approaches and the depth filtration dendrites begin to grow outside of the filter.
threshold method requires the calibration of a model parameter for Most filter models assume that the soot cake is locally uniform,
each set of experimental data. The depth filtration threshold is only which means the particle tree formation and connection regimes are
useful at estimating the total soot load within the filter wall but it can- ignored. Serrano et al. [99]’s model captured the impact of particle tree
not predict the filter performance under a different operating condition growth on the rate of increase of the cake thickness. The surface area on
other than the calibrated experimental dataset. Whilst the assumption the wall-channel interface that allows soot cake to grow on 𝐴cakegrowth
behind Eq. (57) resembles the experimental observations more closely, is
its predictive power is limited as the value of the percolation constant
can vary greatly between different sets of experiments [63]. Further- 𝑎cg = 4𝑑ch 𝐿𝜃, (62)
more, its value may also depend on the level of discretisation within where 𝜃 is the area correction factor
the filter wall as discussed in Section 2.2. In order to increase the [ ]
𝑤cake
predictive power and applicability of the partition coefficient approach, 𝜃 = 𝜖w0 + (1 − 𝜖w0 ) min ,1 , (63)
𝑤cake,lim
the dependence of the percolation constant on filter properties, lo-
cal flow conditions, the properties of incoming particulates and the assuming a linear change in available area for soot cake to grow.
spatial discretisation should be studied. For example, Uenishi et al. Initially, particle trees only grow on the pores of the porous wall, at
[143] found that the value of the percolation constant is related to this point 𝜃 = 𝜖w0 . Since there is less area for the soot cake to grow on,
the filtration efficiency of the first wall layer and has described their the rate of increase of the cake thickness is higher than when uniform
relationship with a third-order polynomial. Future studies should aim cake growth is assumed. The available area for cake growth increases
to develop a correlation for percolation constant that allows accurate as the soot cake is formed during the particle tree connection stage and
predictions across a wide range of particulate filters. eventually, uniform cake growth is observed (𝜃 = 1). The cake thickness
In addition to the build-up of the soot cake due to pore blockage, the when uniform cake growth begins 𝑤cake,lim is a model input that may
soot cake itself is an effective filter medium. The partition coefficient be measured experimentally [76].
is often interpreted as the filtration efficiency of the cake layer [96, The overall pressure drop of the filter is increased by the formation
144]. Serrano et al. [99] proposed that the soot cake filtration efficiency of the soot cake via two means. First, it reduces the available cross-
is proportional to that of the porous wall, and modelled the soot cake sectional area in the inlet channels for gas flow in the axial direction.
filtration efficiency as Second, an additional pressure drop is caused by gas having to flow
[ ( )] through the soot cake.
𝛷 − 𝐶sat
𝐸cake = 𝐸w max 0, , (60) The momentum balance in the axial direction of the inlet channel
1 − 𝐶sat
in channel-scale models is [146]
where 𝐶sat is a model parameter called the limit saturation coefficient.
The cake filtration efficiency is zero when 𝛷 ≤ 𝐶sat . After the partition 𝜕𝑃 𝜕 ( 2) 𝐹f 𝜇𝑣
+ 𝜌 𝑣 =− , (64)
coefficient has exceeded the limit saturation coefficient, the filtration 𝜕𝑧 𝜕𝑧 g (𝑑 − 2𝑤w )2
efficiency of the soot cake increases and approaches that of the wall where 𝑃 is the gas pressure, 𝑧 is the axial coordinate, 𝜌g is the gas
when 𝛷 = 1. density, 𝑣 is the axial gas flow velocity and 𝐹f is the friction factor. The

10
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Fig. 11. Macro- and micro-porosity in the soot cake [104].

gas density across the soot cake, the pressure drop across the soot cake
Fig. 10. The cross-section of a square inlet channel filled with soot cake [38]. The formed in square channels is [157]
axial coordinate 𝑧 is perpendicular to the page.. 𝑦=𝑤cake
𝜇
𝛥𝑃cake = 𝑣cake (𝑦)𝑑𝑦
𝜅cake ∫𝑦=0
( ) (70)
𝜇𝑣cake,top ( ) 𝑑
thickness of the soot cake can be calculated from the mass of the soot = 𝑑 − 2𝑤cake ln ,
2𝜅cake 𝑑 − 2𝑤cake
cake, assuming that the shape of the soot cake is geometrically similar
to the channel cross-section [147]. For filters with square channels as where 𝑦 is the distance from the top of the soot cake as shown in Fig. 10,
shown in Fig. 10, the thickness of the soot cake is 𝑣cake,top is the gas velocity at the top of the soot cake and 𝜅cake is the
( √ ) permeability of the soot cake. The permeability of the cake layer has
1 𝑚cake
𝑤cake = 𝑑 − 𝑑2 − , (65) been modelled by different approaches in the literature. The easiest
2 𝑁ic 𝐿𝜌cake
method would be to calibrate it directly to match experimental pressure
where 𝜌cake is the effective density of the soot cake, which is a function drop data [154,158,159]. However, the permeability of the soot cake
of the material density of soot 𝜌soot and the porosity of the soot cake can be affected by the temperature and the pressure of the system.
𝜖cake [148] The following equation has been used to capture the change in cake
( ) permeability due to slip phenomenon [80,137,160]
𝜌cake = 𝜌soot 1 − 𝜖cake . (66) ( )
√
𝑃 𝑇
Eq. (65) and (66) assume that the soot density and porosity of the 𝜅cake = 𝜅cake0 1 + 𝐶 0 𝜇 , (71)
soot cake are constant i.e. they do not vary within the soot cake. 𝑃 𝑀 r
In situ direct measurements showed that the porosity of the soot where 𝐶 is a numerical parameter, 𝑃0 is the pressure at reference
cake in DPFs lies between 0.93 to 0.97 [149]. The porosity of the soot condition (at which 𝜅cake0 is defined) and 𝑀 r is the average molecular
cake is expected to be no less than 0.83 [128]. The packing of the soot weight of the gas. Mahadevan et al. [136] considered the change in
cake depends on the local gas flow condition. If the gas flow velocity soot cake permeability due to the change in mean free path length of
is high, ballistic deposition of soot particles is expected and the soot the gas
cake will be dense; otherwise sparse diffusion-limited structures will
𝜆
form [150]. Konstandopoulos et al. [151] suggested that the soot cake [𝜌cake 𝜅cake ] = 𝐶 , (72)
𝜆ref
porosity is a function of the local Peclet number
( )( )
Pe −𝑞
where 𝜆 is the mean free path length and 𝜆ref is the mean free path
𝜖cake = 1 − 1 − 𝜖cake ,∞ 1+ 0 , (67) length at a specific condition. The calibration constant 𝐶 is fitted
Pe
against experimental data.
where 𝜖cake ,∞ is the minimum cake porosity which is achieved when Some works have attempted to relate the permeability of the soot
Pe → ∞ and Pe0 is the critical Peclet number that signifies the tran- cake with its microstructure. Konstandopoulos et al. [148] used Eq. (52)
sition between diffusion-dominant and convection-dominant particle to estimate the soot cake permeability assuming the soot cake is a bed
deposition. The value of the exponent 𝑞 is not reported. of spherical primary particles. The Cunningham correction factor is
Furthermore, the packing of the soot cake may be affected by the calculated using the primary particle diameter as the relevant length
local pressure [152] scale (𝑂(10 nm)). On the other hand, Serrano et al. [138] argued that
( )𝑞 the slip phenomenon is governed by the pore diameter of the soot
𝛥𝑃cake
𝜌cake = 𝜌cake0 1 + ∗ , (68) cake. In their work, the mean aggregate diameter (𝑂(100 nm)) was used
𝛥𝑃cake
in Eq. (52) instead of the primary particle diameter. As a result, the
where 𝜌cake0 is the uncompressed cake density, 𝛥𝑃cake is the pressure calibrated soot cake porosity ranged from 0.6 to 0.7. The discrepancy
∗
drop across the cake, 𝛥𝑃cake is the strength of the cake and 𝑞 is a between the low soot cake porosity calibrated by Serrano et al. [138]
tunable exponent. Konstandopoulos et al. [151] suggested that soot and the experimental data measured by Liu et al. [149] (0.93–0.97)
cakes may exhibit Bingham-like behaviour i.e. the packing of the soot may be explained by Fig. 11.
cake is unaffected until the pressure difference across it exceeds a Models used by Serrano et al. [138], Payri et al. [103] and Macián
critical value. They suggested using the following equation should be et al. [122] treated the soot cake as a spherical packed bed formed
used to describe the pressure dependence of the soot cake packing by soot aggregates. However, the soot aggregates are ‘‘porous’’ because
( [ ] )𝑞 they have fractal-like structures. The actual porosity of the soot cake de-
max 0, 𝛥𝑃cake − 𝛥𝑃cake,cr
𝜌cake = 𝜌cake0 1 + , (69) pends on both the macro-porosity between soot aggregates 𝜖cake,macro
∗
𝛥𝑃cake and the micro-porosity within the soot aggregates 𝜖cake,micro
where 𝛥𝑃cake,lim is the minimum pressure drop required to affect the ( )( )
𝜖cake = 1 − 1 − 𝜖cake,macro 1 − 𝜖cake,micro , (73)
packing of the soot cake.
The pressure drop across the soot cake is usually described by the true porosity of the cake becomes 0.93 if 𝜖cake,macro = 0.65 and
Darcy’s law (Eq. (41)), which requires the knowledge of the thickness 𝜖cake,micro = 0.81 (typical porosity of mean soot aggregate [138]) are
and the permeability of the soot cake [153–156]. Assuming constant assumed, which is consistent with experimental observations.

11
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

The equations above have assumed that all particles within the where 𝑦 is the through-wall coordinate, 𝐷m,cake is the effective diffu-
soot cake have the same size. In the case where there is a significant sivity of the soot cake and 𝜔̇ is the rate of species production due to
size variation, Endo et al. [161] proposed an equation to estimate the chemical reactions. It can be seen from Eqs. (76) and (78) that the
pressure drop across the soot cake if the soot particles within the cake presence of the soot cake will have no impact on the concentration
have a log-normal particle size distribution profile if 𝜔̇ = 0. This is usually the case in the soot cake when no
(1 − 𝜖cake )2 𝑘 regeneration occurs within the filters [165]. A notable exception is ob-
𝛥𝑃cake = 180𝜇𝑈 𝑤cake ( 2 ), (74) served in catalysed DPFs when there are notable chemical interactions
3 𝐶
𝜖cake 2
c 𝑑p exp 4 ln 𝜎p between the soot cake and SCR-related species under certain operating
where 𝑑p and 𝜎p are the average diameter and standard deviation of conditions. Adsorption of NH3 as well as fast SCR and NO2 -SCR activity
the particle size distribution. was found to occur on the surface of soot deposits [166,167]. If these
Given the high porosity in the soot cake, Thomas et al. [162] sug- reactions are appreciable, Eq. (78) may need to be solved in the soot
gested that it should be modelled as a fibrous structure with ‘‘particle cake. Furthermore, soot may react with ammonium nitrate and produce
chains’’ as the fibres. The pressure drop across the soot cake can then CO/CO2 , which suppressed the formation of N2 O [168]. Under these
be described as conditions, the interaction between ammonium nitrate and soot cannot
( )3 [ ( )3 ] be captured by direct integration of independently developed soot and
64𝜇𝑈 𝑤cake 1 − 𝜖cake 2 1 + 56 1 − 𝜖cake 𝐶F SCR kinetics [169]. It was suggested by Mihai et al. [168] that the
𝛥𝑃cake = , (75)
𝐶c 𝑑p2 presence of soot can inhibit the formation of ammonium nitrate which
is formed at low temperature with high NO2 at the inlet. This subse-
where 𝐶F ranges between 1 and 23 for cylindrical fibres and spherical quently leads to an increase in SCR performance since less ammonium
particle chains respectively. nitrate is inhibiting SCR activity.
Ribeyre et al. [163] used Eqs. (74) and (75) to study the effect of Significant diffusion can occur within the catalysed filter wall [164].
humidity on the pressure drop across the soot cake. It was found that Eq. (78) can be applied to the wall if 𝐷m,cake is replaced by the
whilst the thickness of the soot cake decreased slightly with increasing diffusivity of the filter wall 𝐷m,w . The diffusivity of the filter wall can
water content, the overall pressure drop increased with water content
be estimated using the parallel pore model which considers both bulk
due to reduced permeability as void space within the soot cake was
diffusion and Knudsen diffusion [137,164]
reduced. On the other hand, Konstandopoulos et al. [128] suggested ( )
that crack formation can occur if the wetted soot cake is left to dry. The 1 𝜏 1 1
= + , (79)
pressure drop across the soot cake will therefore be reduced as cracks 𝐷m 𝜖 𝐷m,bulk 𝐷m,Kn
provided a shortcut for gas flow and hence the overall permeability is where 𝜏 is the tortuosity of the porous medium, 𝐷m,bulk is the molecular
increased. diffusivity of the gas and 𝐷m,Kn is the Knudsen diffusivity
√
2.4. Impact on catalytic conversion of gaseous pollutants 𝑑pore 8𝑅𝑇
𝐷m,Kn = . (80)
3 𝜋𝑀 r
Particulate filters often have a catalytic coating to convert gaseous
pollutants and aid regeneration. Here we discuss the impact of par- In principle, the porosity and the pore diameter of the filter wall
ticle deposition on the performance of catalytic filters during non- under certain soot loading may be calculated by the unit collector
regenerating conditions. The efficiency of catalytic conversion depends model using Eqs. (9), (39) and (40). However, since tortuosity is usually
on the intrinsic activity of the catalyst and the transport of reactants calibrated to match experimental data [80,114], the impact of soot
towards the catalyst. Soot deposits within catalytic filters introduce deposit on in-wall diffusion via porosity reduction may be masked by
additional mass transfer resistances via various means. the uncertainty of the tortuosity calibration. Furthermore, no model
First, the soot cake can affect the convective mass transfer in the has been established to describe the variation in tortuosity as the wall
inlet channels. The species balance in the inlet channel is [164] soot load increases. This limits the accuracy of the effective diffusivity
𝜕𝑐g 𝜕𝑐g 4 ( ) 4 estimation.
= −𝑣 − 𝑘m 𝑐 − 𝑐g,w − 𝑣w 𝑐 , (76) In addition to the diffusion across the filter wall, Greiner et al. [170]
𝜕𝑡 𝜕𝑧 𝑑 − 2𝑤cake g 𝑑 − 2𝑤cake gas,w
found that pore diffusion within the catalyst particle is important and it
where 𝑐g is the molar concentration of the gas species, 𝑑 is the channel
can be more significant than the effect of advection–diffusion across the
diameter, 𝑤cake is the soot cake thickness, 𝑣w is the through-wall gas
wall as described in Eq. (78). The authors suggested that the application
velocity and 𝑘m is the convective mass transfer coefficient. In addition
of an effectiveness factor to catalytic reactions within the wall is a good
to modifying the effective channel diameter, the presence of the soot
cake can also affect the convective mass transfer coefficient. The mass choice to consider pore diffusion in channel-scale filter models. The
transfer coefficient may be calculated using the correlations such as the effectiveness factor 𝛾 is the ‘‘ratio of the effective to the maximum reaction
one developed by Bissett et al. [70] rate’’ [171]

𝑘m (𝑑 − 2𝑤cake ) 𝜔̇ = 𝛾 𝜔̇ max , (81)

Sh = = 2.98 + 0.6Pew − 0.143Rew , (77)
𝐷m,g
where 𝜔̇ and 𝜔̇ max are the actual and maximum rate of production of
where 𝐷m,g is the bulk diffusivity of gaseous species, Pew is the Peclet a chemical products. The maximum rate of production is calculated
number and Rew is the Reynolds number for the through-wall gas flow, assuming the system is kinetic-limited i.e. there is no mass transfer limi-
which means they are functions of 𝑣w . The presence of the soot cake tation. It should be noted that the effectiveness factor approach assumes
can affect the mass transfer coefficient implicitly due to its effect on that the chemical reaction is first-order with a single reactant. Chundru
the through-wall gas flow velocity profile via Eq. (64). et al. [114] modelled the impact of soot deposits on the mass transfer at
Furthermore, the soot cake acts as a diffusive barrier between the
the catalyst surface by adopting the unit collector concept. The model
in-wall catalyst and the bulk gas in the inlet channel. An advection–
is illustrated in Fig. 12.
diffusion–reaction equation is needed to resolve the concentration gra-
The thickness of the soot deposits on the surface of collectors can
dient across the soot cake [49]
be shown as
𝜕𝑐g 𝜕𝑐g 𝜕 2 𝑐g
= −𝑣w + 𝐷m,cake + 𝜔,
̇ (78) 1( )
𝛿p = 𝑑 − 𝑑c0 (82)
𝜕𝑡 𝜕𝑦 𝜕𝑦2 2 c

12
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Table 1
Summary of thermal properties of soot deposits from literature.
Reference Specific heat capacity Thermal conductivity
(J/kgK) (W/mK)
Zheng and Keith [172] 1510 0.84
Depcik and Assanis [184] 889 2
Meng et al. [181] 1500 0.1
Di Sarli and Di Benedetto 900 10
[185]

between the filter wall and the inlet/outlet channel, 𝛬w is the thermal
Fig. 12. Soot deposit can inhibit SCR reactions by increasing diffusional resistance conductivity of the filter wall and 𝑄̇ is the temperature source term due
between reactant and catalyst [114]. to regeneration reactions. Eq. (86) shows that soot deposits can affect
heat transfer in the following ways:

1. Additional heat capacity in solid phase (𝜌cake 𝑤cake 𝑐p,cake

according to Fig. 12. The effectiveness factor in this case is
+ 𝜌w 𝑤w 𝑐pw )
√
𝐷m,cat 𝑘ideal tanh 𝜙 2. Convective heat transfer with inlet channel (ℎ1 )
𝛾= √ , (83)
𝛿p 𝐷m,cat 𝑘ideal tanh 𝜙 + 𝐷m,cat (
3. Heat conduction ) in the soot cake and the filter wall
𝜕 𝜕𝑇 𝜕2 𝑇
(𝛬cake 𝜕𝑧 𝑤cake 𝜕𝑧w + 𝛬w 𝑤w 𝜕𝑧2w )
where 𝐷m,cat is the diffusivity within the catalytic washcoat, 𝑘ideal is the
rate constant of catalytic reaction under kinetic-limited condition and ̇
4. The heat of regeneration reaction (𝑄)
𝜙 is the Thiele modulus, which is the ratio of reaction and diffusion
The heat released from regeneration can be calculated from the
within the catalytic coating [39]
change in enthalpies of gaseous reactants and products [181,182]. The
√
2
𝑘ideal 𝛿w convective heat transfer coefficient in the inlet channel depends on the
𝜙= . (84) gas flow across the wall, characterised by the wall Reynolds number
𝐷m,cat
and the Prantdl number [57,70,183]. The presence of soot in and on
Whilst increased mass transfer resistance due to soot deposition is the wall can affect ℎ1 implicitly, similar to its impact on convective
generally considered to be negative, the overall performance of the mass transfer (Section 2.4).
catalyst may benefit from deteriorated species transport. For example, The density and the thickness of the soot cake have been inves-
the presence of soot can inhibit NH3 oxidation in SCR-coated filters tigated by many researchers and have been discussed in Section 2.3.
at high temperatures. This allows more NH3 to be available for NOx On the other hand, the specific heat capacity 𝑐p,cake and the thermal
reduction reactions [168]. conductivity 𝛬cake of soot deposits have not been subjected to the same
level of scrutiny. The impact of soot loading on the filter wall thermal
2.5. Impact on heat transfer conductivity 𝛬w is usually neglected. Many models have included 𝑐p,cake
and 𝛬p,cake in their model equations without explicitly stating the values
The heat transfer phenomena and temperature distributions within used for these parameters [91,114,120,154,158,159,179]. The values
particulate filters are important topics within the field of filter mod- of specific heat capacities of soot deposits and the thermal conductivity
elling studies. Many publications have studied the thermal behaviour of soot cakes are summarised in Table 1.
Literature values for the specific heat capacity of the soot cake
of particulate filters especially during active regeneration [172–178].
agree reasonably and are comparable to that of the filter wall substrate
However, the impact of deposits on the heat transfer process has not
material (1120 J/kgK for cordierite, 800 J/kgK for SiC [181]). On the
been studied in detail.
other hand, values of thermal conductivities span over two orders of
It is common to assume that the gas within the porous wall is at local
magnitude. If the thermal conductivities of the soot cake and the filter
thermal equilibrium with the filter substrate [119,179]. The tempera-
wall are vastly different, the assumption that the soot cake and the filter
ture variation across the soot cake and the wall is also usually neglected wall are at local thermal equilibrium breaks down as heat will dissipate
in filter models. This assumption has been investigated by Haralampous at different speeds in the soot cake and the filter wall.
and Koltsakis [180] and they found the following condition for this Since the soot cake is porous, its thermal conductivity would be a
assumption to hold function of its porosity [180]
𝑐pg 𝑚̇ in 𝑤cake
≤ 0.1, (85) 𝛬cake = 𝜖𝛬g + (1 − 𝜖) 𝛬soot . (87)
𝑎𝛬cake
Assuming for a moment that pure soot has the same thermal prop-
where 𝑐pg is the specific heat capacity of the gas, 𝑚̇ in is the inlet mass
erties as black carbon (1.6 W/mK), the thermal conductivity of a soot
flow rate to the filter, 𝑎 is the total area of the interface between the
cake with 97% porosity would be 0.108 W/mK [180]. This suggested
inlet channel and the porous wall and 𝛬cake is the thermal conductivity
that the values of thermal conductivities assumed by Depcik and As-
of soot deposits. When this condition is satisfied, only a single equation
sanis [184] and Di Sarli and Di Benedetto [185] are unlikely to be
is needed to be solved for the temperature of the solid phase of the filter realistic, as those values are closer to typical thermal conductivities of
in single-channel models [181] metal oxides.
( ) 𝜕𝑇w So far there has been little attention to how soot deposits can
𝜌cake 𝑤cake 𝑐p,cake + 𝜌w 𝑤w 𝑐pw =
𝜕𝑡 impact heat transfer as it is less influential than other factors such as
( ) ( ) ( )
ℎ1 𝑇1 − 𝑇w + ℎ2 𝑇2 − 𝑇w + 𝜌g 𝑣w 𝑐pg 𝑇1 − 𝑇w (86) the operating condition and substrate properties. This is reflected in
( ) literature modelling studies where very different values have been used
𝜕 𝜕𝑇 𝜕 2 𝑇w
+ 𝛬cake 𝑤cake w + 𝛬w 𝑤w ̇
+ 𝑄. for the thermal properties of soot deposits and yet all manage to get
𝜕𝑧 𝜕𝑧 𝜕𝑧2 good agreement with experimental data. This is reasonable since the
where 𝑐p,cake is the specific heat capacity of the soot cake, 𝜌w is the mass of soot deposits in particulate filters is small with respect to the
bulk density of the filter wall, 𝑐pw is the specific heat capacity of filter substrate, and it is not an excellent heat insulator nor a huge heat
the filter wall, ℎ1 and ℎ2 are the convective heat transfer coefficients sink.

13
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

3. Modelling regeneration 3.2. Rate equations

3.1. Reaction pathways The rates of regeneration reactions depend on the filter temperature,
the concentration of gaseous reactants, the properties of the particles,
The major oxidants of soot deposits in particulate filters are O2 and and the catalysts involved. A general equation to describe the rate of
NO2 [186]. Soot oxidation by NO is usually neglected as NO is a weaker regeneration reaction is
oxidant than O2 under the expected operating conditions of particulate
d𝑚p
filters [187]. Regeneration of soot by O2 and NO2 can be expressed 𝑟=− = 𝑘𝐺𝛤 , (93)
as [37] d𝑡
( ) ( ) where 𝑘 is the reaction rate constant, 𝐺 is a function of the concen-
C + 𝛶1 O2 → 2𝛶1 − 1 CO2 + 2 − 2𝛶1 CO, (88)
trations of the gaseous reactants and 𝛤 is a function of particle state,
( ) ( ) usually its mass. It should be noted that the rate of regeneration is
C + 𝛶2 NO2 → 𝛶2 NO + 𝛶2 − 1 CO2 + 2 − 𝛶2 CO, (89)
often presented as the rate of oxidant consumption instead of the rate
where 𝛶1 and 𝛶2 are stoichiometric coefficients of the regeneration of soot mass consumption. These rates may be interconverted based on
reactions. Soot is usually represented as pure carbon (C) in regeneration the stoichiometry of regeneration reactions.
reactions [188]. The formation of N2 from NO2 -assisted soot oxidation Arrhenius-type equations are frequently used to describe reaction
has also been considered in the literature [189]. When both NO2 rate constants [157,204,205]
and O2 are available in the exhaust, a synergistic effect has been ( )
𝐸
observed [190]. The overall rate of soot oxidation is higher than the 𝑘 = 𝐴𝑇 𝑏 exp − , (94)
𝑅𝑇
sum of O2 -only oxidation and NO2 -only oxidation activities, which
where 𝐴 is the pre-exponential constant, 𝑏 is the temperature exponent
suggests the existence of NO2 -assisted O2 oxidations [186]
and 𝐸 is the activation energy of the reaction. The temperature expo-
1 nent 𝑏 is often assumed to be 0 [206–208] or 1 [153,158,209,210]
C + NO2 + O → CO2 + NO, (90)
2 2 before any calibration. Catalysed and thermal regeneration are often
1 NO2 modelled as independent reactions with different rate parameters [107,
C+ O → CO. (91)
2 2 153,185].
Pyrolysis may happen in the absence of oxygen. This commonly 𝐺 in Eq. (93) is usually a function of concentrations of gaseous
occurs in gasoline engines [47]. Mitsouridis et al. [160] considered the oxidants [115,207,209–211]. Since O2 and NO2 are the most common
formation of CO from soot and CO2 oxidants for filter regeneration, the following expressions can cover
most variations of 𝐺 used in the literature
C + CO2 → 2CO. (92)
𝑞O 𝑞NO
This reaction is limited by thermodynamics at low temperatures 𝐺 = 𝑐O 2 𝑐NO 2 , (95)
2 2
and is therefore not expected to occur in DPFs. Nonetheless, it may be where 𝑐 is molar concentration and 𝑞 is the reaction order of the oxi-
appreciable in GPFs since the exhaust temperature is generally higher dants. 𝐺 can also be defined in terms of partial pressures, mass fractions
in GPFs than in DPFs [191]. or mole fractions of oxidants [91,107,110]. Regeneration reactions are
The performance of regeneration can be improved by coating the
often assumed to be first-order in the involved oxidants [157,158,204,
particulate filter with oxidation catalysts. First, oxidation catalysts can
211,212]. However, experimental studies have shown that the reaction
produce NO2 from NO to support regeneration [49]. Second, catalysts
order of O2 regeneration ranges between 0.8 to 1 in the absence of
can greatly improve CO2 selectivity over CO [192,193] by oxidising CO
catalysts [213–215]. This implies that the rate of oxidation is deter-
during regeneration [194]. Third, the presence of catalysts can lower
mined by the adsorption of oxygen on active carbon sites [216]. The
the light-off temperature of regeneration reactions [195,196]. This can
presence of a catalytic coating in the filter can affect the reaction orders
reduce the peak temperature during active regeneration [185], which
in turn reduces the chance of filter failure due to thermal-induced with respect to oxidants since the rate-determining step is different. For
stress. example, the reaction order of O2 in a regeneration reaction catalysed
The soot–catalyst contact can influence the regeneration perfor- by a Pt/Pd catalyst is found to be 0.3 [107]. It should be noted that the
mance of the catalyst [51]. It is strongly dependent on the morphology influences of catalyst on different regeneration pathways are different
of the catalyst and support [197,198]. The morphology is particularly e.g. the reaction order of NO2 regeneration remains 0.6 regardless of
important for soot oxidation since trapped particles are immobile [199, the presence of the Pt/Pd catalyst [107].
200]. As most of the soot cake is not touching the filter wall, only ther- Chemical species other than the oxidants can affect the reaction
mal (uncatalysed) regeneration is appreciable in the soot cake [119]. rates as well. Schejbal et al. [189] used the following expression for
Nonetheless, the catalyst can improve regeneration of the soot cake by NO2 -only regeneration to include the catalytic effect of water vapour
producing oxidants such as NO2 that can diffuse towards the soot cake. 𝑞NO ( 𝑞H O )
The dependence of catalysed regeneration on soot–catalyst contact con- 𝐺 = 𝑌NO 2 1 + 𝐶𝑌H O2 , (96)
2 2
ditions can also vary with temperature. Piumetti et al. [201] found that where 𝑌 is the mass fraction. Parameter 𝐶 and reaction orders 𝑞 require
the impact of the soot–catalyst condition on catalysed soot oxidation calibration.
is more significant at high temperatures for self-prepared ceria-based The simplest and the most widely used expression of 𝑃 in Eq. (93)
catalysts. The catalysed soot oxidation can be insensitive to the surface is the 𝑛th-order model [153,207,209,211,215,217]
contact condition at low temperature if the surface of the catalyst is
𝑞
masked by adsorbed species. 𝛤 = 𝑚pp . (97)
Other species in the exhaust can also affect regeneration reactions.
The presence of H2 O and CO2 has been observed to slightly delay If the oxidation occurs solely on the surface of the particle, the rate
soot ignition [192]. On the other hand, H2 O has also been reported to of the regeneration should be proportional to the surface area of the
catalyse NO2 regeneration [195]. In addition, O2 -only soot oxidation particles and hence the reaction order 𝑞p = 32 . On the other hand,
may be catalysed by water vapour between 450 ◦ C and 600 ◦ C [202]. 𝑞p = 1 implies internal burning and hence the rate of regeneration is
It should be noted that soot does not only contain elemental carbon. proportional to the mass of the particles [206]. Kinetic studies some-
Whilst catalyst may be active for the oxidation of elemental carbon times include the initial mass of soot into the rate equation [213,214]
in soot particles, it may not be active for the oxidation of adsorbed
( )𝑞p
polyaromatics [203]. If this is the case, polyaromatics may be released 𝑚p 1−𝑞 𝑞
into the atmosphere during regeneration. 𝛤 = 𝑚p0 = 𝑚p0 p 𝑚pp . (98)
𝑚p0

14
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

It will be difficult to integrate such rate equations with simula- The reaction rate is set to be proportional to the total reactive
tions of continuously regenerating filters since the history of trapped specific surface area of soot particles 𝑆p,r
particles needs to be tracked to compute the rate of regeneration.
Some authors expressed the formation of CO and CO2 as separate re- 𝑆p,r = 𝑆p,ext + 𝛾p 𝑆p,int , (102)
actions with their own rate parameters. Alternatively, the temperature- where 𝑆p,ext and 𝑆p,int are the external and internal specific surface
dependence of the CO selectivity 𝛹CO may be modelled as such [157] area of soot deposits. The external specific surface area was measured
by Kandas et al. [223] experimentally. The internal specific surface
𝑐̇ CO = 𝛹CO 𝑘𝑐O2 , area was assumed to be the uncovered area after removal of SOF. The
( ) internal effectiveness factor 𝛾p was included to account for the effect of
𝑐̇ CO2 = 1 − 𝛹CO 𝑘𝑐O2 ,
( ) internal pore diffusion
𝑘 = 𝐴 exp −
𝐸
, (99) ( )
𝑅𝑇 1 1 1
( ′) 𝛾p = − , (103)
1 𝐸 ′ 𝜙 tanh 3𝜙 𝜙
= 1 + 𝐴′ exp 𝑐O𝑞 ,
𝛹CO 𝑅𝑇 2
The Thiele modulus 𝜙 for adsorption–reaction on spherical soot
where 𝑐̇ is the rate of production of chemical species in molar concen- primary particles is [224]
tration. The Arrhenius parameters of the CO selectivity 𝐴′ , 𝐸 ′ and 𝑞 ′ √
𝑑pri 𝑆p,int 𝑘𝛶 𝐾g 𝑝g
will need to be calibrated. 𝜙= ′
, (104)
6 2𝐷m,Kn (𝐾g 𝑝g − ln[1 + 𝐾g 𝑝g ]) 1 + 𝐾g 𝑝g
Once the expression of reaction rate is chosen, kinetic parameters
are typically calibrated based on experimental data. Engine bench ′
where 𝛶 is the stoichiometric factor in Eqs. (88) and 𝐷m,Kn is the
tests should be used to determine soot oxidation kinetics to ensure adjusted Knudsen diffusivity calculated using the Salatino’s proce-
that experimental results are obtained with real soot under realistic dure [225]
operating conditions [218], although thermogravimetric analysis (TGA) ( )
𝐸
is also used frequently. The importance of the methodology on the rate
′
𝐷m,Kn = 𝐷m,Kn exp − ads , (105)
𝑅𝑇
parameters was highlighted by Bogarra et al. [215] who found that
where 𝐸ads is the activation energy of the adsorption step. Using this
TGA with a heat ramp and isothermal TGA gave different estimates of
model, Macián et al. [122] found that the equivalent reaction order of
rate parameters. It should also be noted that Printex-U is often used
NO2 is around 0.3 and 0.4 in typical passive regeneration conditions,
as a surrogate of diesel soot with high elemental carbon content in
whereas the equivalent reaction order of O2 varied during the course
kinetics tests [47]. Whilst Printex-U oxidises similarly to ash-free GDI
of regeneration.
soot [219], its oxidation behaviour does not always represent real diesel
The evolution of specific surface area during regeneration may be
soot [220].
modelled as [213]
The reactive surface area of soot particles plays a role in controlling √
the rate of oxidation. Many models assume a fixed specific surface area 𝑆p = 𝑆p0 (1 − 𝜁) 1 − 𝛱 ln (1 − 𝜁), (106)
for the soot cake during regeneration [91,114,120,136,154,158,181,
where 𝑆p0 is the initial specific surface area and 𝜁 is the conversion of
204]. This is counter-intuitive as we expect particles to become smaller
soot. The structural factor 𝛱 describes the initial pore structure within
as it burns and hence an increase in the specific surface area [44].
soot aggregates [226]. According to this equation, the surface area will
This is expected to speed up combustion because of the increasing
increase initially due to pore growth. As regeneration continues, the
surface–volume ratio [217].
surface area will reach a maximum and then decrease as the remaining
The change in surface area of soot particles is dependent on the soot collapses into itself.
combustion mode of the particles. Different burning modes were ob- Some works have found that there is no correlation between the
served when soot particles were oxidised by different oxidants. When specific surface area and soot reactivity [219,227]. Zygogianni et al.
only O2 is present, significant internal burning can occur within soot [228] found that structural differences of soot (e.g. the size of crystal-
particles [221]. This is due to the preferential removal of amorphous lite) and the composition of particles are more correlated to the soot
carbon over less reactive graphitic carbon. On the other hand, NO2 has oxidation reactivity. For example, a higher degree of graphitisation
shown no preference. As a result, a shrinking sphere model is adequate (more ordered structure) leads to less reactive soot [229]. However,
to describe the change in specific surface area during soot oxidation most models neglect the difference in nanostructure of soot particles
by NO2 . When both NO2 and O2 are present, the degree of internal and treated them as pure carbon species [120,179,182]. Kastrinaki
burning is greatly suppressed due to cooperative oxidation (Eq. (90) et al. [209] proposed a multi-population kinetics to describe soot oxida-
and (91)). Whether internal burning occurs also depends on the soot tion behaviour. The soot samples were divided into three populations
composition. Jung et al. [222] found that acetylene soot is impermeable with their own reaction rate constants and activation energies. The
to O2 and hence combustion will only occur on the soot surface. authors attributed the difference in soot oxidation behaviour to distinct
Macián et al. [122] considered the mass transport phenomena families of surface oxygen complexes (SOC) formed on the carbon
within soot aggregates during regeneration. Diffusion of oxidants surface of each population; Kastrinaki et al. [209] hypothesised that
within the soot aggregate as well as the adsorption of oxidants onto the the differences in activation energies of the three soot populations
reactive sites are considered. The impact of the adsorption and reaction are caused by the different bond configurations on the carbon surface
of gaseous reactants on the rate of regeneration is described by the i.e. single, double or triple bonds.
Langmuir isotherm Furthermore, real engine soot also contains a soluble organic frac-
𝐾g 𝑝g tion (SOF) and ash. Soot oxidation activity may be enhanced by higher
𝐺= , (100) soluble organic fraction (SOF) content and weakly bonded carbon
1 + 𝐾g 𝑝g
(WBC) content [230]. Removal of SOF can lead to a rapid increase
where 𝑝g is the partial pressure of a gaseous reactant. The equilibrium in the specific surface area of soot particles at the early stage of
constant of adsorption 𝐾g is described by oxidation [231]. Depending on the ash composition, soot oxidation
( ) activity may be enhanced or reduced [208]. Ash particles may act as
𝛥𝐻ads
𝐾g = 𝐴 exp − , (101) oxygen carrier which improves soot oxidation [76]. On the other hand,
𝑅𝑇 ash originated from phosphate-doped fuel was found to inhibit soot
where 𝛥𝐻ads is the adsorption enthalpy. oxidation [232].

15
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Fig. 13. Three-stage oxidation behaviour of ash-containing GDI soot.

Choi and Seong [230] proposed a global soot oxidation kinetic Catalysts within the filter can assist NO2 regeneration by producing
scheme that considers the impact of ash and SOF/WBC content in GDI NO2 from NO, although no catalytic effect was observed with platinum
soot. The overall rate is catalyst on the rate of NO2 soot oxidation [189,234]. If NO2 is being
d𝜁 𝑚SW,0 𝑚C,0 produced within catalysed filter walls, it can aid oxidation of the soot
𝑟̂ = = 𝑟̂ + 𝑟̂ , (107)
d𝑡 𝑚soot,0 SW 𝑚soot,0 C cake. NO2 may reach the soot cake via back-diffusion [137]. In order
to capture the concentration gradient across the soot cake and the
where 𝜁 is the overall conversion and 𝑟̂ is a dimensionless reaction rate.
filter wall, Eq. (78) would need to be solved in channel-scale models.
The subscripts SW and C refer to SOF/WBC and carbon respectively.
The soot cake would need to be discretised in this case to resolve
The rate of SOF/WBC oxidation is
( ) the concentration gradient of NO2 across the soot cake [204]. On the
d𝜁 𝐸 ( )𝑞 other hand, competition for NO2 can occur between passive regen-
𝑟̂SW = SW = 𝐴SW exp − SW 1 − 𝜁SW SW , (108)
d𝑡 𝑅𝑇 eration and SCR reactions in SCR-coated filters [132]. SCR reactions
where 𝜁SW is the degree of conversion of SOF-WBC are more competitive for NO2 than soot oxidation [235], although at
𝑚 high temperatures, soot oxidation can outweigh SCR reactions [236].
𝜁SW = 1 − SW . (109) Furthermore, NH3 may block free-edge sites on carbon particles. This
𝑚SW,0
phenomenon has been modelled by Trandafilović et al. [226].
The rate of carbon oxidation 𝑟C is It is shown in this section that the rates of regeneration reactions
( )
d𝜁 𝐸 ( )𝑞 are influenced by many factors, ranging from the nanostructure of soot
𝑟̂C = C = 𝐴C exp − C 1 − 𝜁C C + 𝐶1 exp(𝐶2 𝑇 )𝑌ash,O , (110) deposits, the surface of soot deposits, their ‘‘non-soot’’ components and
d𝑡 𝑅𝑇
the presence of catalysts. Many researchers have developed models to
where 𝐶1 and 𝐶2 are tunable parameters of ash-assisted soot oxidation,
quantify the impact of these factors, but it remains a challenging task to
𝑌ash,O is the fraction of oxidation-derived ash and 𝜁C is the degree of
combine these developments into a coherent regeneration model that
conversion of carbon
can capture the effect of all factors mentioned above.
𝑚
𝜁C = 1 − C . (111)
𝑚C,0
3.3. Evolution of particle deposit
Ash may be deliberately introduced by doping the fuel with fuel-
borne catalyst (FBC) metal additives because this can reduce the tem- The changes of soot cake properties during regeneration are often
perature at which soot burns off in the filter [233]. FBC is expected ignored in modelling studies [91]. Many filter modelling studies have
to perform better than the catalytic coating on filter walls because of assumed that the packing density of the soot cake remains constant
close contact between ash and soot especially in the soot cake [232]. during regeneration [114,120,179,181,204,237]. The thickness of the
Ash accumulates in the filter while soot burns off. soot cake 𝑤cake under O2 regeneration may be obtained by solving the
Easter [219] observed a three-stage oxidation behaviour for ash-rich following equation [91]
GDI soot. This was attributed to the change in soot–ash contact during
d𝑤cake 𝑣𝜌g 𝑌O2,0 ( [
𝑆p 𝑘𝑤cake
])
oxidation. The proposed mechanism is shown in Fig. 13. =− 1 − exp − , (112)
Initially, soot and ash are in close contact and the rate of oxidation d𝑡 𝛶 𝜌cake 𝑣
is at its highest. As regeneration continues, the contact between soot where 𝑣 is the through-wall gas flow velocity, 𝜌g is the gas density,
and ash is lost, result in a lower oxidation rate. The rate of oxidation is 𝑌O2,0 is the mass fraction of oxygen at the top of the soot cake, 𝛶 is the
recovered slightly towards the end of the oxidation, as contact between stoichiometric coefficient of O2 , 𝑆p is the specific surface area of the
remaining soot and ash are re-established since the majority of soot is soot cake and 𝑘 is the reaction rate constant described by an Arrhenius-
already consumed. type equation (Eq. (94)). Bissett [238] has shown the steps to derive
Although the contact condition between soot, ash and catalyst can Eq. (112).
affect the regeneration performance significantly [51], quantification of Kostoglou and Konstandopoulos [239] examined how assumptions
the impact of soot-ash-catalyst contact on regeneration is rarely seen in about the microstructure of the soot cake can affect the calibration and
literature filter models. Konstandopoulos et al. [211] studied the effect prediction of the regeneration process. Four models were considered in
of the size of milled catalyst particles on their soot oxidation ability their work and are shown in Fig. 14.
using the multi-population model [209]. They found that reducing the The packing density and the specific surface area of the soot cake
size of ceria catalyst particles can increase the oxidation rate of highly are the two relevant parameters during the regeneration process [239].
and moderately reactive soot (Activation energies of 120 kJ/mol and The first model in Fig. 14 assumes constant cake density and specific
180 kJ/mol respectively), whereas the oxidation of the least reactive surface throughout regeneration. The second model treats the soot
soot (240 kJ/mol) was insensitive to the size of catalyst particles. Note cake as spherical particles held in stationary cells that all burn at the
that Konstandopoulos et al. [211] have kept the activation energies of same rate. The third model treats the soot cake as cylindrical columns
soot populations constant regardless of the presence of a catalyst. Other with the same diameters as the primary soot particles. The fourth
authors often use lower activation energy for catalysed regeneration model treats the soot cake as columns of spherical particles, where
pathways with respect to thermal regeneration pathways [119]. the movement of particles and hence a change in cake thickness is

16
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Fig. 14. Conceptual regeneration models to describe the change in packing density
and surface area of soot cakes during regeneration [239].
Fig. 16. The fixed spatial discretisation of the soot cake proposed by Wurzenberger
et al. [205]. A mobility term is introduced to describe the movement of soot due to
regeneration.

soot cake. The model proposed by Wurzenberger et al. [205] used a

fixed spatial grid to describe the soot cake as shown in Fig. 16.
A mobility term for soot mass is needed to describe the shrinking
of the cake layer during regeneration. The rate of migration of soot is
Fig. 15. Conceptual model of soot cake evolution during active regeneration based on expressed as
neutron tomography observations [240].
𝑚̇ mob,𝑖 = 𝐶mob 𝛼soot,𝑖 𝜖𝑖+1 , (115)
were 𝐶mob is a mobility factor, 𝛼soot,𝑖 is the volume fraction of soot of the
considered, in contrast to the second model. The fourth model was 𝑖th layer and 𝜖𝑖+1 is the void fraction of the (𝑖 + 1)th layer that is below
used to fit against experimental data and good agreement was found. the 𝑖th layer. On the other hand, Haralampous and Koltsakis [245]
The authors suggested that future soot cake models should track the discretised the soot cake into a fixed number of slabs. The soot cake
evolution of both the cake density and the cake thickness. is re-divided into elements of equal mass at every time step. Whilst a
However, the change in the microstructure of the soot cake during moving grid approach can reduce numerical diffusion [246], this model
regeneration has been experimentally observed to be more complicated does not consider the local variation of soot cake properties such as the
than cases considered by Kostoglou and Konstandopoulos [239]. Fig. 15 porosity, unlike the model used by Wurzenberger et al. [205].
shows a conceptual model proposed by Toops et al. [240] who studied Sometimes the soot cake may collapse during regeneration and
the evolution of the soot cake during active regeneration using neutron leads to the formation of a mid-channel soot plug. This will lead to
tomography. a rapid increase in pressure drop. Since the pressure drop of the filter
During the first stage of regeneration, the soot cake becomes thinner is frequently used to estimate the soot load of the filter by on-board
and more compact while maintaining its structural uniformity. At the controls [152], the formation of mid-channel soot plug will lead to
second stage, the soot cake becomes more porous without significant malfunction of the on-board control. Fukui et al. [247] found that this
change in the cake thickness. Many irregular scallops were observed phenomenon can be invoked with intermittent increases of flow rate
and engine torque. A later study by Kim et al. [248] showed that three
through SEM during this stage [76]. A similar phenomenon was also
factors would be needed to trigger the collapse of the soot cake. First,
observed for GDI soot [219]. The formation of these scallops may
passive regeneration by in-situ NO2 generation is needed to create a
be driven by the high attractive force between soot deposits, which
gap between the soot cake and the filter wall. Second, water needs
is stronger than the attractive forces between soot and filter wall
to adsorb onto the soot cake and create a concave shape in the soot.
substrate [241]. The packing density and the thickness of the soot cake
Finally, the collapse of the soot cake is triggered by a cold start with
continue to decrease gradually after the second stage until the soot cake
heavy acceleration. To the best of our knowledge, this phenomenon
is completely oxidised. has yet to be included in the after treatment models reported in the
The rate of pressure drop reduction was observed to be higher literature.
during the second stage of active regeneration than the first and the Various experimental studies have reported surges in the concen-
third stage [76,242]. This implies an increase in the permeability of the tration of particle being released from regenerating filters [249]. Large
soot cake. Mahadevan et al. [136] modelled the increase in soot cake (80 nm) solid particles have been observed to be released from par-
permeability using an analogy with cracked concrete [243]. The cracks ticulate filters during active regeneration [229]. They are likely to
in concrete allow gas to flow through with little resistance which results be unfiltered particles that pass straight through the filter due to a
in an increase in the permeability. In the case of the soot cake, the reduction in the filtration efficiency after regeneration [250]. Many
‘‘cracks’’ are the gaps between soot scallops formed during the second sub−23 nm particles have also been observed [251]. They are com-
stage of active regeneration. The permeability of the damaged soot cake posed of semi-volatile material since they are not detected under high
is modelled as dilution ratio conditions [22]. They may be sulphuric acid particles des-
[( )𝐶 ] orbed from the catalyst [250,252]. Passively regenerating filters were
𝜅cake = 𝜅cake,0 exp 𝐶1 𝜁d 2 , (113) found to have lower particle emission levels than actively regenerating
where 𝐶1 and 𝐶2 are calibrated parameters. The damage variable 𝜁d is filters [253] since the presence of soot cake can limit the reduction in
defined as filtration efficiency.
The increased emission of solid particles during regeneration have
𝑚
𝜁d = 1 − cake . (114) been modelled in the literature. Lao et al. [92] has proposed an empir-
𝑚cake,0
ical model to describe the reduction in filtration efficiency of the soot
In contrast to active regeneration, the combined permeability of cake due to non-uniform regeneration using a sigmoid function
the clogged wall and the soot cake is not altered significantly by [ ( )]
𝑚
exp 𝐶1 𝑚 cake − 𝐶2
passive regeneration [244]. The bottom of the soot cake has the highest cake,0
𝐸cake = [ ( )] , (116)
regeneration rate during passive regeneration of catalysed filters since 𝑚
exp 𝐶1 𝑚 cake − 𝐶2 + 1
NO2 is generated in the catalysed wall and then diffuses back to the cake,0

soot cake. The soot cake has been discretised into multiple layers in where 𝑚cake,0 is the initial mass of the soot cake at the start of re-
filter models in order to resolve the concentration gradient within the generation, 𝐶1 and 𝐶2 are numerical parameters that were calibrated

17
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Deep-bed ash deposits are formed by the regeneration of ash-

containing soot particles that are trapped within the walls of a fil-
ter [255]. Whilst most experimental studies have shown that ash
usually only deposits on the surface of filter walls [133,256], there
is experimental evidence of deep penetration of ash into filter walls
(≥ 50% of the wall thickness [257]). Kamp et al. [258] found that
ash deposits penetrated the filter walls of sintered metal fibre filters
further if the filter undergoes active regeneration instead of passive
regeneration. This is because the persistent presence of the soot cake
during passive regeneration acted as an effective barrier for ash-bearing
soot to deposit within the filter wall.
Fig. 17. Different possible forms of ash deposits in a particulate filter.
The ash layer and the ash plug are formed from the combustion of
the soot cake. The ash plug is formed from soot fragments produced
during regeneration which migrate towards the back of the inlet chan-
based on the experimental filtration efficiency measurements during nels [133]. An optical study conducted by Matsuno and Kitamura [259]
active regeneration. A reduction in the wall filtration efficiency at found that lumps of soot can migrate towards the end of inlet channels
elevated temperatures was also discovered during the model calibration during active regeneration. Whether ash plugs form at the early stages
process. This reduction in filtration efficiency cannot be explained by of ash accumulation depends on the initial amount of soot [259]. Once
the removal of trapped particles. The cause of the reduction in filtration the ash layer is established, it makes soot migration during active
efficiency was not identified and remains an open question. regeneration occur more readily [259]. Shearing flow can also cause
Tan et al. [107] simulated the emission of particles during the ash to migrate towards the back of inlet channels. Whether ash particles
passive regeneration of a catalysed filter. The leaked particles were are re-entrained into the gas flow depends on their size [18]. Particles
hypothesised to be fragments of soot particles produced during catalytic with diameters around 100 μm are the easiest to be re-entrained [260].
oxidation in the filter wall. A correction factor was applied to the com- Larger particles are difficult to be re-entrained due to their inertia and
bined single collector efficiency predicted by the model (𝜂 in Eq. (10)). smaller particles are difficult to be re-entrained due to their smaller
The correction factor 𝐶frag was made to be a function of the partition surfaces [261].
coefficient 𝛷 (Eq. (57)) The size of ash particles depends on the regeneration conditions.
Low regeneration temperature, low soot loading and high soot-to-ash
⎧1 𝛷 < 𝛷low ratio before regeneration can lead to the formation of small, densely
⎪ ( ) 𝛷−𝛷low
𝐶f rag = ⎨1 − 1 − 𝐶frag,min 𝛷high −𝛷low
𝛷low ≤ 𝛷 ≤ 𝛷high (117) packed ash particles [261]. These ash particles may grow by sintering
⎪ with temperature excursion and become more sticky [262,263]. Bagi
⎩𝛿 𝛷 > 𝛷high ,
et al. [264] optically observed the densification and sintering of ash
where 𝛷high and 𝛷low are the limiting partition coefficients chosen precursor remained from soot oxidation. As ash particles sinter and
by Tan et al. [107]. The minimum correction factor 𝐶frag,min describes increase in size, they may be more susceptible to shearing flow which
the level of fragmentation and is a function of particle size and op- can lead to formation of the ash plugs in the mid-channel [265].
erating condition. Whilst the model is largely empirical, it allows Modelling the formation of ash is a challenging task. The following
quantification of the factor 𝐶frag and can help further the development three aspects need to be considered in modelling ash deposition in
of a more physical model. particulate filters:
Predicting the release of particles formed from semi-volatile mate-
rial during regeneration is difficult because they are difficult to mea- 1. Movement of ash fractions as part of soot particles due to
sure [254]. The state of the semi-volatile particles is strongly dependent filtration and regeneration-induced migration.
on the gas phase composition. Significant losses of semi-volatile parti- 2. Movement of ash deposits due to flow-induced migration.
cles can occur within the measuring system due to diffusional losses of 3. Change in size and structure of ash deposits due to agglomera-
particles to the walls of the measurement device. Lee et al. [21] pre- tion and sintering at high temperature.
sented a model that considers sulphates and SOF particles in addition to
solid soot particles. The model was shown to be successful in describing The filtration process can be modelled as described in Section 2. Ad-
the evolution of the particle population in both the engine and in the ditional variables such as ash mass fraction within soot particles need
to be tracked by the filter model in order to simulate the deposition of
particle sampling system. The release of semi-volatile particles during
ash contained within soot particles. The migration of soot cake during
filter regeneration may be captured if this type of modelling approach
regeneration can be described by the following equation [157]
is integrated with existing filter models.
𝜕𝑤cake 𝜕(𝑣𝑤cake )
+ 𝐶mob = 0, (118)
4. Modelling ash accumulation and migration 𝜕𝑡 𝜕𝑧
where 𝑤cake is the (local) thickness of the soot cake, 𝑣 is the axial
4.1. Formation mechanisms gas flow velocity and 𝐶mob is the mobility constant. Deng et al. [157]
did not specify how the mobility constant may be calibrated. Recent
work by Koltsakis et al. [266] has considered the detachment and re-
Ash are formed from incombustible residues that remains in partic-
deposition of the soot cake due to the shear stress experienced by the
ulate filters after regeneration. The ash can be distributed within the
soot cake. The shear stress 𝜄 can be calculated [267]
filter wall, form a layer on the filter wall and can form plugs at the
end of inlet channels. In some cases, ash plugs can also be found in the 𝐹f 𝜇𝑣
𝜄= , (119)
middle of inlet channels. These are illustrated in Fig. 17. 4(𝑑c − 2𝑤cake − 2𝑤ash )
The pattern of ash deposition is closely related to the regeneration where 𝐹f is the laminar channel flow friction factor, 𝜇 is the dynamic
process, the pattern is not dictated by whether the regeneration process viscosity, 𝑣 is the local axial gas flow velocity and (𝑑c − 2𝑤cake − 2𝑤ash )
is active or passive [255]. A passively regenerated filter with high NOx is the channel diameter of the open cross-section available for gas flow
content and high temperature can have a similar ash deposition pattern after accounting for the thicknesses of the soot cake and the ash layer.
to actively regenerated filters. Soot fragments will detach when the shear stress exceeds a critical

18
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

value. The detached fragment will flow along the inlet channel until The impact of ash deposits on the permeability of the filter wall
it reattaches to the soot cake. This happens in the model if the local is often neglected in modelling studies [141,269]. Kamp et al. [139]
shear stress drops below critical reattachment shear stress. No explicit published one of the few studies that did consider this. The following
formula or values regarding the critical shear stresses were reported equation has been used by Kamp et al. [139] to estimate the wall
by Koltsakis et al. [266]. The critical shear stress is expected to be a permeability with soot load and ash load
function of the size of the soot fragment and its adhesiveness to the 1 1 1 1
soot cake/filter wall surface. Although the latter could be measured by = + 𝑉̂soot + 𝑉̂ash , (123)
𝜅w 𝜅w0 𝜅soot 𝜅ash
atomic force microscopy (AFM) [241], no numerical model has been
proposed to predict the size of the soot fragment produced during the where 𝑉̂soot and 𝑉̂ash are the volume ratio of soot and ash deposit with
combustion of a soot cake. Since the detachment of soot fragments respect to the total porous space within the wall. In principle, Eq. (39)
may be modified to capture the impact of deep bed ash deposits in the
depends on adhesion with the surface, the likelihood of soot migration
unit collector model
will change depending on whether the ash layer is present. This is
[ ( ) ( )] 1
in agreement with experimental observations made by Matsuno and 3 𝑚sw 𝑚aw 𝑑c0 3

Kitamura [259]. 𝑑c = 2 + + , (124)

4𝜋 𝜌sw 𝜌aw 2
In addition to the regeneration-induced transport, migration of ash
where 𝑚aw is the mass of ash deposits and 𝜌aw is the packing density of
due to shearing flow on the ash layer has been modelled by Konstan-
ash deposits within the filter wall. It should be noted that the value of
dopoulos et al. [268]. The governing equations for the ash mass fraction
𝜌aw may increase during the lifetime of the filter as ash deposits undergo
suspended in the gas phase 𝑌ash and the ash layer thickness 𝑤ash are
sintering after multiple regeneration events.
𝑑𝑌ash 4 4 Similar to the soot cake, Darcy’s law is often used to describe pres-
𝜌g 𝑣 = 𝐽 − 𝜌 𝑣 𝑌 , (120)
𝑑𝑧 𝑑ch ash 𝑑ch g w ash sure drop across the ash layer (Eq. (41)) [103,154,155,179,210,269].
𝜕𝑤ash Whilst the inertial resistance of the ash layer is often ignored in models,
𝜌g = 𝜌g 𝑣w 𝑌ash − 𝐽ash , (121) experimental studies have shown that the pressure drop of ash-loaded
𝜕𝑡
GPFs can have a quadratic relationship with space velocity which sug-
where 𝑣w is the through-wall gas flow velocity. The ash re-entrainment
gests that ash-induced inertial pressure loss is not negligible [133]. The
flux 𝐽ash takes a non-zero pre-set value 𝐽0 if the local shear stress
permeability of the ash layer may be modelled analogously to that of
exceeded the critical detachment shear stress and the local temperature
the soot cake as shown in Eq. (52). Wurzenberger et al. [205] proposed
is lower than a critical value. The temperature constraint attempts
such a model. The model considers a particle cake layer composed of
to capture the impact of the ash stickiness. In the model proposed
both soot and ash. The particle cake layer is discretised and Darcy’s
by Koltsakis et al. [266], the migration of the ash layer can be inhibited
law is used to describe the pressure drop, with the permeability being
if a soot cake is present on top of the ash layer, subject to a critical soot
described as a function of the through-cake distance. The permeability
cake mass constraint. The critical detachment shear stress of the ash
of the hybrid cake is described by Carmen–Kozeny equation
layer is a function of the local velocity and the local ash agglomerate
3 2
size. The following expression was used by Koltsakis et al. [266] to 𝜖cake,hybrid 𝑑 cake,hybrid
describe the rate of agglomeration of primary ash particles 𝑟agglo 𝜅cake,hybrid = , (125)
(1 − 𝜖cake,hybrid )2 180
( )( )
𝐶 𝑉ash 𝐶3 where the porosity 𝜖cake,hybrid is calculated as follows
𝑟agglo = 𝐶1 exp − 2 , (122)
𝑇 𝑉cake ( ) ( )
where 𝐶1 , 𝐶2 and 𝐶3 are calibrated rate parameters. Eq. (122) is anal- 𝜖cake,hybrid = 1 − 𝛼soot 1 − 𝜖soot − 𝛼ash 1 − 𝜖ash , (126)
ogous to the Arrhenius equation, which 𝐶1 being the pre-exponential where 𝛼soot and 𝛼ash are the bulk volume fractions of soot and ash
constant, 𝐶2 being the ‘‘activation energy’’ of ash agglomeration and within the cake layer. The average particle diameter of the cake com-
the volume ratio of ash with respect to soot cake represents the impact posite 𝑑 p,cake,hybrid is calculated as
of ash ‘‘concentration’’.
𝛼soot + 𝛼ash 𝛼soot 𝛼ash
Last but not least, a multi-channel model would be needed for an = + . (127)
2 2 2
accurate description of real ash loaded filters because ash deposition 𝑑 p,cake,hybrid 𝑑 p,soot 𝑑 p,ash
patterns in real filters vary significantly between channels. Rubino et al.
Ash plugs at the end of inlet channels have a smaller impact on
[256] found that ash plugs had a half-elliptical distribution across filter
the filter pressure drop than the ash layer [111]. On the other hand,
channels, where the longest plug was found in the centre channel. It is
mid-channel ash plugs can lead to a drastic change in velocity profiles
apparent that more work is required to develop a filter model that can
and a huge increase in pressure drop [270]. If a mid-channel ash plug
accurately predicts the accumulation, migration and evolution of ash is impermeable to the gas flow, the effective length of the inlet channel
deposits. The development of such a model will be supported by large is reduced significantly. Despite its detrimental impact on the pressure
amount of experimental data. Non-destructive optical measurements drop at channel-scale, the severity of mid-channel plug formation is
such as MRI and XRT will be helpful. limited on the overall filter pressure drop as long as there are enough
clean channels because gas will preferably flow through paths with the
4.2. Impact of ash deposits on filter performance lowest resistances [261].
The formation of an ash layer generally leads to an increase in filtra-
Ash deposits in the filter can increase filter pressure drop via tion efficiency [111,121,271,272]. Deep bed filtration is assumed to be
different means depending on their form and location. Ash deposits negligible once the ash layer is established. When the ash layer has not
within the filter wall can lead to a reduction in the permeability of been established, ash deposition within the filter wall can reduce the
the filter wall. The ash layer is a porous medium that gas has to flow duration of deep bed filtration [273]. The impact of ash deposits on the
through in addition to the soot cake. The ash plug can influence the filtration process plays a crucial role in the ash deposition pattern since
momentum balance in inlet channels via reduction of the effective inlet ash precursors are transported within soot particles. However, there are
channel length. These can interfere with regeneration control strategies at least three problems associated with physic-based filtration models
that rely on the filter pressure drop as an input signal. In general, ash for ash deposits:
deposition causes active regenerations to occur more frequently due to
the overestimation of the actual soot load [237]. Unnecessary active 1. Deep bed ash deposits would have a different packing density
regenerations can reduce the fuel economy of vehicles. with respect to deep bed soot deposits as shown in Eq. (124).

19
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

2. The partition coefficient approach breaks down when an ash in the pressure drop and filtration efficiency are closely related to the
layer is present since the soot cake would build on the surface packing of the trapped particles. The channel-scale models available in
of the ash layer instead of the filter wall. the literature often consider the number of trapped particles on a mass
3. Whilst the filtration efficiency of the ash layer may be modelled basis instead of a volume basis. This may compromise their ability to
using Eq. (10), SEM images have shown that ash layers are quite predict filter behaviour when the incoming particles are polydisperse.
different from spherical packed beds [139,274]. It is challenging The transition from deep bed filtration to soot cake formation is an-
to adequately describe the microstructure of the ash layer. other important phenomenon to model. The deep bed filtration occurs
when particles are trapped in the walls of the filter. This eventually
Ash can hinder mass transfer in catalysed filters [275]. Bagi et al. gives way to cake filtration once a layer of particles is established on the
[264] optically observed the wetting of substrate by ash deposits. The surface of the wall. The rate of transition is dependent on the clogging
gaseous reactant will have difficulty accessing the catalytic surface of the surface pores and the filtration efficiency of the developing soot
due to ash masking. Ash layers can act as a diffusive barrier anal- cake. The saturation of the surface pores is usually modelled using a
ogously to soot cake. It should be noted that the impact of ash on partition coefficient approach by channel-scale models. It should be
the catalytic performance of filters depends on its composition. For noted that the calibration process of this approach is sensitive to the
example, magnesium-based ash can promote soot regeneration better spatial discretisation of the wall. Great caution should be taken when
than calcium-based ash [276]. quoting the values of the related model parameters. The filtration abil-
The impact of ash on heat transfer would be similar to that of soot ity of the developing soot cake has been considered by recent modelling
as described in Section 2.5 albeit with different material properties. studies. However, some models estimate the filtration behaviour of
Opposing comments about the role of ash on the thermal behaviour of the developing soot cake with equations that assume the presence of
regenerating filters have been made in the literature. Chen et al. [179] a uniform porous medium, which is not an accurate description of a
suggested that ash deposits can hinder heat transfer and lead to higher developing soot cake.
peak temperature during active regeneration, whilst other researchers The impact of mass transfer of gaseous species on the performance
have suggested that the overall heat capacity of the filter may be in- of catalysed particulate filters is a complex subject. Since current
creased by ash deposits, which reduces the peak temperature [76,273]. models rely on empirical tuning to obtain good agreement between
The impact of ash on heat transfer also depends on the distribution of model results and experimental measurements, a model that is capable
ash within particulate filters. of predicting the impact of soot deposits on the mass transfer of gaseous
Ash deposits can undergo substantial microscopic change during species for a wide range of systems and operating conditions is not
their lifetime in particulate filters. This substantially increases the expected to be developed in the near future. There is potential in
difficulty of modelling the impact of ash deposits. Whilst sintering linking the gas phase mass transfer models for in-wall soot deposits
and other re-structuring phenomena of ash have been widely reported to the filtration model. This will allow better quantification of the
in experimental studies, limited modelling studies have attempted to impact of soot deposits on the gas phase mass transfer phenomenon.
consider the evolution of the properties of ash deposits. Linking the The impact of soot load on heat transfer within particulate filters has
microstructure of the ash deposits to their impact on pressure drop, fil- not been studied in detail in the literature because it is not an important
tration and regeneration is vital for accurate filter models. In addition, factor in contrast to other factors such as operating condition and
the irregular shape of ash deposits further complicates the issue. SEM substrate properties. Current approaches in the literature are deemed
images have shown that the ash layer may not be formed by spherical adequate in this regard.
ash primary particles; rather it may be molten and have an irregular Regeneration reactions are described by Arrhenius-like equations.
structure [274]. Hollow ash particles can be formed due to sintering Details of many chemical and physical phenomena are rarely explicitly
over multiple regeneration events. Although most existing models using considered in channel-scale models. Their influences are lumped into
an analogous treatment for the ash and soot deposits, describing aged the values of calibrated parameters such as the pre-exponential con-
ash deposits as particles or aggregates is inconsistent with experimental stant. This severely limits the applicability of the calibrated kinetics.
observations. In particular, the change in the soot deposits during regeneration is
often oversimplified since many models only consider the mass of
5. Summary remaining soot deposits. Consideration of the change in the morphology
of soot deposits during regeneration will allow a better description
This paper has critically assessed how filtration, regeneration and of the regeneration process as well as the performances of partially
ash formation have been considered by particulate filter models. In regenerated filters.
particular, the description of particle deposits is discussed in detail. Furthermore, the change in soot deposits during regeneration is
The filtration efficiency of filter walls can be estimated precisely closely related to the formation of ash deposits. There is plenty of
with pore-scale models using knowledge of the internal structure of room for the development of ash formation models as they are scarce
the filter wall that can be obtained by tomography. Soot aggregates compare to experimental studies on ash formation in the literature.
can be reasonably approximated as equivalent spherical particles in The development of the ash formation models will need support from
pore-scale models. Correlations are available for the calculation of the extensive experimental data in addition to filter pressure drop mea-
diameter and the density of the equivalent spherical particles. However, surements. Non-destructive optical measurements will be extremely
the application of pore-scale models is limited due to computational helpful in assisting the development of ash models. In addition, the
cost. They are rarely applied to study the impact of soot deposits on impact of ash deposits on filtration behaviour should be studied in
the filtration behaviour of the walls. Channel-scale filtration models detail as it can feed back to the ash formation process. It is important to
remain the first choice for such tasks. They approximate the complex understand that approximating ash deposits as spherical particles can
internal structure of the porous wall using simple geometries, which be inappropriate due to sintering of the ash during regeneration events.
allow analytical expressions of the filtration efficiency to be derived. Filter models have been developed and applied successfully in
Many expressions have been developed in the literature with subtle the past, but further improvement can be made in various aspects.
differences between them. Simulations of pore-scale models can be a Regarding impacts of particulates on filter behaviour, focusing on
useful benchmark for the evaluation of various channel-scale filtration phenomena occurring at the smaller length scales i.e. pore size and
models. particle size will be needed. Since the high computational cost limits
As particles accumulate in a particulate filter, the pressure drop and the applications of pore-scale models to clean filter walls with relatively
the filtration efficiency generally increases. The rates of the increase simple particle models, channel-scale models will remain the major tool

20
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

for predicting filter behaviour during the entire life cycle of the filter. 𝑝 Partial pressure Pa
Extension of channel-scale models to consider pore-scale phenomena 𝑝D Deposition probability –
will be needed to improve the predictive power of models. Further 𝑞 Reaction order –
development of channel-scale models can benefit from support by 𝑟 Rate of regeneration kg s−1
𝑟agglo Rate of agglomeration of
experimental measurements and pore-scale models.
primary ash particles s−1
𝑟̂ Normalised rate of regeneration s−1
Nomenclature 𝑠 Interception exponent –
𝑡 Time s
Upper-case Roman 𝑢 Interstitial velocity m s−1
𝐴 Pre-exponential constant (varies) 𝑣 Velocity m s−1
𝐶 Numerical coefficient – 𝑤 Thickness m
𝐶c Cunningham correction factor – 𝑤̂ sat Normalised soot penetration thickness –
𝐶F Shape factor for fibrous medium – 𝑦 Spatial coordinate m
𝐶S Sticking coefficient – 𝑧 Spatial coordinate m
𝐷 Diffusion coefficient m2 s−1
Upper-case Greek
𝐷m Diffusivity m2 s−1
𝛤 Regeneration function of particle (varies)
𝐷mm Mass-mobility exponent –
𝛬 Thermal conductivity W m−1 K−1
𝐸 Filtration efficiency –
𝛱 Structural factor –
𝐸 Activation energy kg mol−1
𝛶 Stoichiometric coefficient –
𝐹 Force N
𝛷 Partition coefficient –
𝐹f Friction factor –
𝛷soot Soot density factor –
𝐺 Regeneration function of gas (varies)
𝛹 Selectivity –
𝛥𝐻 Enthalpy J mol−1
𝐽 Ash re-entrainment flux kg m−2 s−1 Lower-case Greek
𝐾 Kuwabara function – 𝛼 Solidity –
𝐾g Equilibrium constant – 𝛽 Forchheimer constant m−1
Kn Knudsen number – 𝛾 Effectiveness factor –
𝐿 Channel length m 𝛿 Thickness m
𝑀r Molecular weight kg mol−1 𝜖 Porosity –
𝑁ic Number of inlet channels – 𝜁 Conversion –
𝑁R Interception parameter – 𝜁d Damage variable –
𝑃 Pressure Pa 𝜂 Single collector efficiency –
Pe Peclet number – 𝜃 Area correction factor –
𝑄̇ Rate of heat generation W 𝜅 Permeability m2
𝑅 Universal gas constant J mol−1 K−1 𝜆 Mean free path m
Re Reynolds number – 𝜇 Dynamic viscosity Pa s
𝑆 Specific surface area m2 kg−1 𝜌 Density kg m−3
Sh Sherwood number – 𝜎 Standard deviation of particle size m
St Stokes number – distribution
𝑇 Temperature K 𝜏 Tortuosity –
𝑈 Superficial velocity m s−1 𝜙 Thiele modulus –
𝑉 Volume m3 𝜒 Shape factor –
𝑊 Wiener process – 𝜓 Percolation constant –
𝑉̂ Normalised volume – 𝜔̇ Rate of species molar production mol m−3 s−1
𝑉sp Soot penetration volume m3 Subscripts
𝐖 Wiener process – ads Denotes adsorption
𝑌 Mass fraction – aw Denotes ash in wall
Lower-case Roman bulk Bulk
𝑎 Area m2 B Denotes Brownian diffusion
𝑏 Temperature exponent – C Denotes carbon
𝑐 Molar concentration kg mol−3 c Denotes collector
𝑐p Specific heat capacity J kg−1 K−1 cake Cake
𝑑 Diameter m cat Catalytic
𝑑ch Channel diameter m cg Cake growth
𝑓 Distribution function – clog Clog
𝑓K Kuwabara geometric function – D Denotes drag
ℎ Heat transfer coefficient W m−2 K−1 D Denotes diffusion
𝑖 Counter – ef f Effective
𝑘 Rate constant (varies) ext Denotes external
𝑘B Boltzmann constant m2 kg s−2 K−1 f Denotes fibre
𝑘m Mass transfer coefficient m s−1 f rag Denotes fragmentation
𝑚 Mass kg g Denotes gas
𝑚̇ Mass flow rate kg s−1 I Denotes inertial impaction
𝑛 Number of particles – int Denotes internal

21
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

Kn Denotes Knudsen diffusion [10] Liu H, Wang Z, Long Y, Xiang S, Wang J, Fatouraie M. Comparative study
lim Denotes limit on alcohol–gasoline and gasoline–alcohol dual-fuel spark ignition (DFSI) com-
mob Denotes mobility diameter bustion for engine particle number (PN) reduction. Fuel 2015;159:250–8. http:
p Denotes particle //dx.doi.org/10.1016/j.fuel.2015.06.059.
[11] Anbari Attar M, Xu H. Correlations between particulate matter emissions and
pore Denotes pore gasoline direct injection spray characteristics. J Aerosol Sci 2016;102:128–41.
pri Primary particle http://dx.doi.org/10.1016/j.jaerosci.2016.09.006.
R Denotes interception [12] Reijnders J, Boot M, de Goey P. Particle nucleation-accumulation mode trade-
ref Reference off: A second diesel dilemma? J Aerosol Sci 2018;124:95–111. http://dx.doi.
sat Saturation org/10.1016/j.jaerosci.2018.06.013.
[13] Kittelson DB. Engines and nanoparticles: a review. J Aerosol Sci
sd Denotes soot dendrites 1998;29(5):575–88. http://dx.doi.org/10.1016/S0021-8502(97)10037-4.
soot Denotes soot [14] Tree DR, Svensson KI. Soot processes in compression ignition engines. Prog En-
SW Denotes SOF/WBC ergy Combust Sci 2007;33(3):272–309. http://dx.doi.org/10.1016/j.pecs.2006.
sw Denotes soot in wall 03.002.
w Denotes wall [15] Martin JW, Salamanca M, Kraft M. Soot inception: Carbonaceous nanoparticle
formation in flames. Prog Energy Combust Sci 2022;88:100956. http://dx.doi.
0 Denotes an initial condition org/10.1016/j.pecs.2021.100956.
[16] Burtscher H. Physical characterization of particulate emissions from diesel
Declaration of competing interest engines: a review. J Aerosol Sci 2005;36(7):896–932. http://dx.doi.org/10.
1016/j.jaerosci.2004.12.001.
[17] Überall A, Otte R, Eilts P, Krahl J. A literature research about particle emissions
The authors declare that they have no known competing finan-
from engines with direct gasoline injection and the potential to reduce these
cial interests or personal relationships that could have appeared to emissions. Fuel 2015;147:203–7. http://dx.doi.org/10.1016/j.fuel.2015.01.012.
influence the work reported in this paper. [18] Wang Y, Kamp CJ, Wang Y, Toops TJ, Su C, Wang R, et al. The origin, transport,
and evolution of ash in engine particulate filters. Appl Energy 2020;263:114631.
http://dx.doi.org/10.1016/j.apenergy.2020.114631.
Data availability
[19] Sappok AG, Wong VW. Detailed chemical and physical characterization of ash
species in diesel exhaust entering aftertreatment systems. In: SAE technical
No data was used for the research described in the article. paper. SAE International; 2007, http://dx.doi.org/10.4271/2007-01-0318.
[20] Mayer A, Czerwinski J, Kasper M, Ulrich A, Mooney JJ. Metal oxide parti-
cle emissions from diesel and petrol engines. In: SAE technical paper. SAE
Acknowledgements International; 2012, http://dx.doi.org/10.4271/2012-01-0841.
[21] Lee KF, Eaves N, Mosbach S, Ooi D, Lai J, Bhave A, et al. Model guided
This research was supported by the National Research Foundation, application for investigating particle number (PN) emissions in GDI spark
Singapore, Prime Minister’s Office, Singapore under its Campus for Re- ignition engines. In: SAE technical paper. 2019, 2019, http://dx.doi.org/10.
4271/2019-26-0062.
search Excellence and Technological Enterprise (CREATE) programme. [22] Yamada H, Inomata S, Tanimoto H. Mechanisms of increased particle and VOC
This project has received funding from the European Union’s Hori- emissions during DPF active regeneration and practical emissions considering
zon 2020 research and innovation programme under grant agreement regeneration. Environ Sci Technol 2017;51(5):2914–23. http://dx.doi.org/10.
814492. This work was supported by the UK Engineering and Phys- 1021/acs.est.6b05866.
[23] Giechaskiel B, Maricq M, Ntziachristos L, Dardiotis C, Wang X, Axmann H, et
ical Sciences Research Council (EPSRC) grant 1622599. The authors
al. Review of motor vehicle particulate emissions sampling and measurement:
would like to thank Royal Dutch Shell for their support. MK gratefully From smoke and filter mass to particle number. J Aerosol Sci 2014;67:48–86.
acknowledges the support of the Alexander von Humboldt Foundation, http://dx.doi.org/10.1016/j.jaerosci.2013.09.003.
Germany. [24] Millo F, Andreata M, Rafigh M, Mercuri D, Pozzi C. Impact on vehicle fuel
economy of the soot loading on diesel particulate filters made of different
substrate materials. Energy 2015;86:19–30. http://dx.doi.org/10.1016/j.energy.
References 2015.03.076.
[25] Regulation (EC) No 715/2007 on type approval of motor vehicles with respect
[1] Morawska L, Ristovski Z, Jayaratne E, Keogh D, Ling X. Ambient nano and ultra- to emissions from light passenger and commercial vehicles (Euro 5 and Euro
fine particles from motor vehicle emissions: Characteristics, ambient processing 6) and on access to vehicle repair and maintenance information. 2007.
and implications on human exposure. Atmos Environ 2008;42(35):8113–38. [26] Bogarra M, Herreros JM, Tsolakis A, York AP, Millington PJ. Study of
http://dx.doi.org/10.1016/j.atmosenv.2008.07.050. particulate matter and gaseous emissions in gasoline direct injection engine
[2] Ramanathan V, Carmichael G, Ramanathan V, Carmichael G, Ramanathan V, using on-board exhaust gas fuel reforming. Appl Energy 2016;180:245–55.
Carmichael G. Global and regional climate changes due to black carbon. Nat http://dx.doi.org/10.1016/j.apenergy.2016.07.100.
Geosci 2008;1(4):221–7. http://dx.doi.org/10.1038/ngeo156. [27] Bock N, Jeon J, Kittelson D, Northrop WF. Solid particle number and mass emis-
[3] Lelieveld J, Evans JS, Fnais M, Giannadaki D, Pozzer A. The contribution of sions from lean and stoichiometric gasoline direct injection engine operation.
outdoor air pollution sources to premature mortality on a global scale. Nature In: SAE technical paper. 2018, 2018, p. 1–10. http://dx.doi.org/10.4271/2018-
2015;525(7569):367–71. http://dx.doi.org/10.1038/nature15371. 01-0359.
[4] Qian Y, Li Z, Yu L, Wang X, Lu X. Review of the state-of-the-art of par- [28] Ho S-H, Wong Y-D, Chang VW-C. Developing Singapore driving cycle for
ticulate matter emissions from modern gasoline fueled engines. Appl Energy passenger cars to estimate fuel consumption and vehicular emissions. Atmos
2019;238:1269–98. http://dx.doi.org/10.1016/j.apenergy.2019.01.179. Environ 2014;97:353–62. http://dx.doi.org/10.1016/j.atmosenv.2014.08.042.
[5] Mohankumar S, Senthilkumar P. Particulate matter formation and its control [29] Sileghem L, Bosteels D, May J, Favre C, Verhelst S. Analysis of vehicle emission
methodologies for diesel engine: A comprehensive review. Renew Sustain measurements on the new WLTC, the NEDC and the CADC. Transp Res D:
Energy Rev 2017;80:1227–38. http://dx.doi.org/10.1016/j.rser.2017.05.133. Transp Environ 2014;32:70–85. http://dx.doi.org/10.1016/j.trd.2014.07.008.
[6] Matti Maricq M. Chemical characterization of particulate emissions from diesel [30] Marotta A, Pavlovic J, Ciuffo B, Serra S, Fontaras G. Gaseous emissions from
engines: A review. J Aerosol Sci 2007;38(11):1079–118. http://dx.doi.org/10. light-duty vehicles: Moving from NEDC to the new WLTP test procedure.
1016/j.jaerosci.2007.08.001. Environ Sci Technol 2015;49(14):8315–22. http://dx.doi.org/10.1021/acs.est.
[7] Ma Y, Zhu M, Zhang D. Effect of a homogeneous combustion catalyst on the 5b01364.
characteristics of diesel soot emitted from a compression ignition engine. Appl [31] Pavlovic J, Ciuffo B, Fontaras G, Valverde V, Marotta A. How much difference
Energy 2014;113:751–7. http://dx.doi.org/10.1016/j.apenergy.2013.08.028. in type-approval CO2 emissions from passenger cars in Europe can be expected
[8] Fiebig M, Wiartalla A, Holderbaum B, Kiesow S. Particulate emissions from from changing to the new test procedure (NEDC vs. WLTP)? Transp Res A:
diesel engines: correlation between engine technology and emissions. J Occup Policy Pract 2018;111:136–47. http://dx.doi.org/10.1016/j.tra.2018.02.002.
Med Toxicol 2014;9(1):6. http://dx.doi.org/10.1186/1745-6673-9-6. [32] Bishop JD, Molden N, Boies AM. Using portable emissions measurement systems
[9] Bermúdez V, Luján JM, Ruiz S, Campos D, Linares WG. New European driving (PEMS) to derive more accurate estimates of fuel use and nitrogen oxides emis-
cycle assessment by means of particle size distributions in a light-duty diesel sions from modern euro 6 passenger cars under real-world driving conditions.
engine fuelled with different fuel formulations. Fuel 2015;140:649–59. http: Appl Energy 2019;242:942–73. http://dx.doi.org/10.1016/j.apenergy.2019.03.
//dx.doi.org/10.1016/j.fuel.2014.10.016. 047.

22
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[33] Jang J, Lee J, Choi Y, Park S. Reduction of particle emissions from gasoline [56] Aravelli K, Heibel A. Improved lifetime pressure drop management for robust
vehicles with direct fuel injection systems using a gasoline particulate filter. Sci cordierite (RC) filters with asymmetric cell technology (ACT). In: SAE technical
Total Environ 2018;644:1418–28. http://dx.doi.org/10.1016/j.scitotenv.2018. paper. SAE International; 2007, http://dx.doi.org/10.4271/2007-01-0920.
06.362. [57] Wang W, Bissett EJ. Frictional and heat transfer characteristics of flow in
[34] Kontses A, Triantafyllopoulos G, Ntziachristos L, Samaras Z. Particle number triangle and hexagon channels of wall-flow monoliths. Emiss Control Sci
(PN) emissions from gasoline, diesel, LPG, CNG and hybrid-electric light-duty Technol 2018;4(3):198–218. http://dx.doi.org/10.1007/s40825-018-0093-7.
vehicles under real-world driving conditions. Atmos Environ 2020;222:117126. [58] Konstandopoulos AG, Kostoglou M. Analysis of asymmetric and variable
http://dx.doi.org/10.1016/j.atmosenv.2019.117126. cell geometry wall-flow particulate filters. SAE Int J Fuels Lubricants
[35] He L, Hu J, Zhang S, Wu Y, Zhu R, Zu L, et al. The impact from the direct 2014;7(2):489–95. http://dx.doi.org/10.4271/2014-01-1510.
injection and multi-port fuel injection technologies for gasoline vehicles on solid [59] Xiao G, Li B, Tian H, Leng X, Long W. Numerical study on flow and pressure
particle number and black carbon emissions. Appl Energy 2018;226:819–26. drop characteristics of a novel type asymmetric wall-flow diesel particulate
http://dx.doi.org/10.1016/j.apenergy.2018.06.050. filter. Fuel 2020;267. http://dx.doi.org/10.1016/j.fuel.2020.117148.
[36] Yusuf AA, Inambao FL. Effect of cold start emissions from gasoline-fueled [60] Nakamura K, Konstandopoulos A, Kostoglou M, Shibata T, Hashizume Y. New
engines of light-duty vehicles at low and high ambient temperatures: Recent asymmetric plugging layout of diesel particulate filters for the pressure drop
trends. Case Stud Therm Eng 2019;14:100417. http://dx.doi.org/10.1016/j. reduction. In: SAE technical papers. Vol. 1, 2014, http://dx.doi.org/10.4271/
csite.2019.100417. 2014-01-1512.
[37] Lao CT, Akroyd J, Eaves N, Smith A, Morgan N, Nurkowski D, et al. [61] Konstandopoulos A, Kostoglou M, Vlachos N. The multiscale nature of diesel
Investigation of the impact of the con fi guration of exhaust after-treatment particulate filter simulation. Int J Veh Des 2006;41. http://dx.doi.org/10.1504/
system for diesel engines. Appl Energy 2020;267:114844. http://dx.doi.org/10. IJVD.2006.009676.
1016/j.apenergy.2020.114844. [62] Fino D, Bensaid S, Piumetti M, Russo N. A review on the catalytic combustion
[38] Konstandopoulos AG, Kostoglou M, Skaperdas E, Papaioannou E, Zarvalis D, of soot in diesel particulate filters for automotive applications: From powder
Kladopoulou E. Fundamental studies of diesel particulate filters: Transient load- catalysts to structured reactors. Appl Catalysis A: General 2016;509:75–96.
ing, regeneration and aging. In: SAE 2000 world congress. SAE International; http://dx.doi.org/10.1016/j.apcata.2015.10.016.
2000, http://dx.doi.org/10.4271/2000-01-1016. [63] Yang S, Deng C, Gao Y, He Y. Diesel particulate filter design simula-
[39] Olowojebutu S, Steffen T. A review of the literature on modelling of integrated tion: A review. Adv Mech Eng 2016;8(3):1–14. http://dx.doi.org/10.1177/
SCR-in-DPF systems. In: SAE technical papers. 2017, http://dx.doi.org/10.4271/ 1687814016637328.
2017-01-0976. [64] Kočí P, Isoz M, Plachá M, Arvajová A, Václavík M, Svoboda M, et al. 3D
[40] Chen P, Wang J. Air-fraction modeling for simultaneous diesel engine NOx reconstruction and pore-scale modeling of coated catalytic filters for automotive
and PM emissions control during active DPF regenerations. Appl Energy exhaust gas aftertreatment. Catalysis Today 2018. http://dx.doi.org/10.1016/j.
2014;122:310–20. http://dx.doi.org/10.1016/j.apenergy.2014.02.031. cattod.2017.12.025.
[41] Stamatellou AM, Stamatelos A. Overview of Diesel particulate filter systems [65] Yamamoto K, Nakamura M, Yane H, Yamashita H. Simulation on catalytic
sizing approaches. Appl Therm Eng 2017;121:537–46. http://dx.doi.org/10. reaction in diesel particulate filter. Catalysis Today 2010;153(3):118–24. http:
1016/j.applthermaleng.2017.04.096. //dx.doi.org/10.1016/j.cattod.2010.02.064.
[42] Palma V, Ciambelli P, Meloni E, Sin A. Catalytic DPF microwave assisted active [66] Václavík M, Plachá M, Kočí P, Svoboda M, Hotchkiss T, Novák V, et al.
regeneration. Fuel 2015;140:50–61. http://dx.doi.org/10.1016/j.fuel.2014.09. Structure characterisation of catalytic particulate filters for automotive exhaust
051. gas aftertreatment. Mater Charact 2017;134:311–8. http://dx.doi.org/10.1016/
[43] Jiaqiang E, Zhao X, Liu G, Zhang B, Zuo Q, Wei K, et al. Effects analysis on j.matchar.2017.11.011.
optimal microwave energy consumption in the heating process of composite [67] Belot I, Vidal D, Votsmeier M, Hayes RE, Bertrand F. Numerical investigation
regeneration for the diesel particulate filter. Appl Energy 2019;254:113736. of the impact of washcoat distribution on the filtration performance of gasoline
http://dx.doi.org/10.1016/j.apenergy.2019.113736. particulate filters. Chem Eng Sci 2020;221:115656. http://dx.doi.org/10.1016/
[44] Yamamoto K, Kanamori Y. Measurements of size distribution and oxidation j.ces.2020.115656.
rate of PM with NOx . In: JSAE/SAE 2015 international powertrains, fuels & [68] Plachá M, Kočí P, Isoz M, Svoboda M, Price E, Thompsett D, et al. Pore-
lubricants meeting. 2015, http://dx.doi.org/10.4271/2015-01-1995. scale filtration model for coated catalytic filters in automotive exhaust gas
[45] Guan B, Zhan R, Lin H, Huang Z. Review of the state-of-the-art of exhaust aftertreatment. Chem Eng Sci 2020;226:1–10. http://dx.doi.org/10.1016/j.ces.
particulate filter technology in internal combustion engines. J Environ Manag 2020.115854.
2015;154:225–58. http://dx.doi.org/10.1016/j.jenvman.2015.02.027. [69] Bensaid S, Marchisio D, Fino D, Saracco G, Specchia V. Modelling of diesel
[46] Chan TW, Saffaripour M, Liu F, Hendren J, Thomson KA, Kubsh J, et al. particulate filtration in wall-flow traps. Chem Eng J 2009;154(1):211–8. http:
Characterization of real-time particle emissions from a gasoline direct injection //dx.doi.org/10.1016/j.cej.2009.03.043.
vehicle equipped with a catalyzed gasoline particulate filter during filter [70] Bissett EJ, Kostoglou M, Konstandopoulos AG. Frictional and heat transfer
regeneration. Emiss Control Sci Technol 2016;1–14. http://dx.doi.org/10.1007/ characteristics of flow in square porous tubes of wall-flow monoliths. Chem
s40825-016-0033-3. Eng Sci 2012;84:255–65. http://dx.doi.org/10.1016/j.ces.2012.08.012.
[47] Boger T, Rose D, Nicolin P, Gunasekaran N, Glasson T. Oxidation of soot [71] Wang Y, Gong J, Su C, Ou Q, Lyu Q, Pui D, et al. Theoretical framework of
(printex® u) in particulate filters operated on gasoline engines. Emiss Control a polydisperse cell filtration model. Environ Sci Technol 2020;54(18):11230–6.
Sci Technol 2015;1(1):49–63. http://dx.doi.org/10.1007/s40825-015-0011-1. http://dx.doi.org/10.1021/acs.est.0c02956.
[48] Koltsakis G, Haralampous O, Depcik C, Ragone JC. Catalyzed diesel particulate [72] Tang D, Zhao R, Wang S, Wang J, Ni L, Chen L. The simulation and
filter modeling. Rev Chem Eng 2013;29(1):1–61. http://dx.doi.org/10.1515/ experimental research of particulate matter sensor on diesel engine with diesel
revce-2012-0008. particulate filter. Sensors Actuators A 2017;259:160–70. http://dx.doi.org/10.
[49] York AP, Ahmadinejad M, Watling TC, Walker AP, Cox JP, Gast J, et al. 1016/j.sna.2017.03.036.
Modeling of the catalyzed continuously regenerating diesel particulate filter [73] Yamamoto K, Sakai T. Simulation of continuously regenerating trap with
(CCR-DPF) system: Model development and passive regeneration studies. In: catalyzed DPF. Catalysis Today 2015;242:357–62. http://dx.doi.org/10.1016/
SAE technical paper. 2007, http://dx.doi.org/10.4271/2007-01-0043. j.cattod.2014.07.022.
[50] Kaspera W, Indyka P, Sojka Z, Kotarba A. Bridging the gap between tight [74] Kostoglou M, Housiada P, Konstandopoulos AG. Multi-channel simulation of
and loose contacts for soot oxidation by vanadium doping of cryptomelane regeneration in honeycomb monolithic diesel particulate filters. Chem Eng Sci
nanorods catalyst using NO2 as an oxygen carrier. Catalysis Sci Technol 2003;58(14):3273–83. http://dx.doi.org/10.1016/S0009-2509(03)00178-7.
2018;8(12):3183–92. http://dx.doi.org/10.1039/c8cy00545a. [75] Meng Z, Fang J, Pu Y, Yan Y, Wu Y, Wang Y, et al. Experimental study on
[51] Bensaid S, Russo N, Fino D. CeO2 catalysts with fibrous morphology for soot the influence of DPF micropore structure and particle property on its filtration
oxidation: The importance of the soot–catalyst contact conditions. Catalysis process. J Combustion 2016. http://dx.doi.org/10.1155/2016/9612856.
Today 2013;216:57–63. http://dx.doi.org/10.1016/j.cattod.2013.05.006. [76] Du Y, Meng Z, Fang J, Qin Y, Jiang Y, Li S, Li J, et al. Characterization of
[52] Choi B, Lee K-S. LNT/CDPF catalysts for simultaneous removal of NOx and PM soot deposition and oxidation process on catalytic diesel particulate filter with
from diesel vehicle exhaust. Chem Eng J 2014;240:476–86. http://dx.doi.org/ ash loading through an optimized visualized method. Fuel 2019;243:251–61.
10.1016/j.cej.2013.10.100. http://dx.doi.org/10.1016/j.fuel.2019.01.103.
[53] Kang W, Choi B. Effect of copper precursor on simultaneous removal of PM and [77] Gong J, Rutland CJ. PDF-based heterogeneous multiscale filtration model.
NOx of a 2-way SCR/CDPF. Chem Eng Sci 2016;141:175–83. http://dx.doi.org/ Environ Sci Technol 2015;49(8):4963–70. http://dx.doi.org/10.1021/acs.est.
10.1016/j.ces.2015.11.007. 5b00329.
[54] Lambert C, Chanko T, Dobson D, Liu X, Pakko J. Gasoline particle filter [78] Lee H, Rutland CJ. Modeling uncontrolled regeneration of diesel particu-
development. Emiss Control Sci Technol 2017;105–11. http://dx.doi.org/10. late filters, taking into account hydrocarbon slip. Proc Inst Mech Eng D
1007/s40825-016-0055-x. 2013;227(2):281–96. http://dx.doi.org/10.1177/0954407012450837.
[55] Young DM, Hickman DL, Bhatia G, Gunasekaran N. Ash storage concept [79] Konstandopoulos AG. Flow resistance descriptors for diesel particulate fil-
for diesel particulate filters. In: SAE 2004 world congress & exhibition. SAE ters: Definitions, measurements and testing. In: SAE technical paper. SAE
International; 2004, http://dx.doi.org/10.4271/2004-01-0948. International; 2003, http://dx.doi.org/10.4271/2003-01-0846.

23
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[80] Karamitros D, Koltsakis G. Model-based optimization of catalyst zoning on SCR- [104] Tan P-q, Wang D-y, Yao C-j, Zhu L, Wang Y-h, Wang M-h, et al. Extended
coated particulate filters. Chem Eng Sci 2017;173:514–24. http://dx.doi.org/10. filtration model for diesel particulate filter based on diesel particulate matter
1016/j.ces.2017.08.016. morphology characteristics. Fuel 2020;277:118150. http://dx.doi.org/10.1016/
[81] Orihuela MP, Gómez-Martín A, Becerra JA, Chacartegui R, Ramírez-Rico J. j.fuel.2020.118150.
Performance of biomorphic silicon carbide as particulate filter in diesel boilers. [105] Long W, Huang H, Serlemitsos J, Liu E, Reed AH, Hilpert M. Pore-scale study of
J Environ Manag 2017;203:907–19. http://dx.doi.org/10.1016/j.jenvman.2017. the collector efficiency of nanoparticles in packings of nonspherical collectors.
05.003. Colloids Surf A 2010;358(1):163–71. http://dx.doi.org/10.1016/j.colsurfa.2010.
[82] Babaie Rabiee M, Talebi S, Abouali O, Izadpanah E. Investigation of the 01.043.
characteristics of particulate flows through fibrous filters using the lattice [106] Liu X, Kim J, Chanko T, Lambert C, Pakko J. A modeling analysis of fibrous
Boltzmann method. Particuology 2015;21:90–8. http://dx.doi.org/10.1016/j. media for gasoline particulate filters. In: SAE technical papers. 2017, http:
partic.2014.11.010. //dx.doi.org/10.4271/2017-01-0967.
[83] Friedlander SK. Smoke, dust and haze: fundamentals of aerosol behavior. Wiley; [107] Tan P-Q, Cao C-Y, Hu Z-Y, Lou D-M. Modeling of soot fragmentation that
1977. proceeds in a catalyzed diesel particulate filter of a diesel engine. Chem Eng J
[84] Payri F, Broatch A, Serrano J, Piqueras P. Experimental–theoretical method- 2019;375:122110. http://dx.doi.org/10.1016/j.cej.2019.122110.
ology for determination of inertial pressure drop distribution and pore [108] Serrano J, Bermúdez V, Piqueras P, Angiolini E. On the impact of DPF
structure properties in wall-flow diesel particulate filters (DPFs). Energy downsizing and cellular geometry on filtration efficiency in pre- and post-
2011;36(12):6731–44. http://dx.doi.org/10.1016/j.energy.2011.10.033. turbine placement. J Aerosol Sci 2017;113:20–35. http://dx.doi.org/10.1016/j.
jaerosci.2017.07.014.
[85] Eaves NA, Zhang Q, Liu F, Guo H, Dworkin SB, Thomson MJ. Coflame: A
[109] Sappok A, Santiago M, Vianna T, Wong VW. Characteristics and effects of ash
refined and validated numerical algorithm for modeling sooting laminar coflow
accumulation on diesel particulate filter performance: Rapidly aged and field
diffusion flames. Comput Phys Comm 2016;207:464–77. http://dx.doi.org/10.
aged results. In: SAE world congress & exhibition. SAE International; 2009,
1016/j.cpc.2016.06.016.
http://dx.doi.org/10.4271/2009-01-1086.
[86] Matte-Deschênes G, Vidal D, Bertrand F, Hayes RE. Numerical investigation of
[110] Wurzenberger JC, Triebl C, Kutschi S, Poetsch C. Particulate matter classifica-
the impact of thermophoresis on the capture efficiency of diesel particulate
tion in filtration and regeneration-plant modeling for SiL and HiL environment.
filters. Can J Chem Eng 2016;94(2):291–303. http://dx.doi.org/10.1002/cjce.
SAE Int J Engines 2017;10(4). http://dx.doi.org/10.4271/2017-01-0970.
22396.
[111] Jiang J, Gong J, Liu W, Chen T, Zhong C. Analysis on filtration characteristic
[87] Sirignano M, D’Anna A. Filtration and coagulation efficiency of sub-10 nm
of wall-flow filter for ash deposition in cake. J Aerosol Sci 2016;95:73–83.
combustion-generated particles. Fuel 2018;221:298–302. http://dx.doi.org/10.
http://dx.doi.org/10.1016/j.jaerosci.2016.01.009.
1016/j.fuel.2018.02.107.
[112] Ou Q, Maricq MM, Pakko J, Chanko TB, Pui DY. Design and evaluation of a
[88] Yamamoto K, Oohori S, Yamashita H, Daido S. Simulation on soot de- sintered metal fiber filter for gasoline direct injection engine exhaust aftertreat-
position and combustion in diesel particulate filter. Proc Combust Inst ment. J Aerosol Sci 2019;133:12–23. http://dx.doi.org/10.1016/j.jaerosci.2019.
2009;32(2):1965–72. http://dx.doi.org/10.1016/j.proci.2008.06.081. 04.003.
[89] Yamamoto K, Fujikake F, Matsui K. Non-catalytic after-treatment for diesel [113] Liu ZG, Wang PK. Pressure drop and interception efficiency of multifiber
particulates using carbon-fiber filter and experimental validation. Proc Combust filters. Aerosol Sci Technol 1997;26(4):313–25. http://dx.doi.org/10.1080/
Inst 2013;34(2):2865–75. http://dx.doi.org/10.1016/j.proci.2012.09.006. 02786829708965433.
[90] Yamamoto K, Yamauchi K. Numerical simulation of continuously regenerating [114] Chundru VR, Mahadevan BS, Johnson JH, Parker GG, Shahbakhti M. Develop-
diesel particulate filter. Proc Combust Inst 2013;34(2):3083–90. http://dx.doi. ment of a 2D model of a SCR catalyst on a DPF. Emiss Control Sci Technol
org/10.1016/j.proci.2012.06.117. 2019;5(2):133–71. http://dx.doi.org/10.1007/s40825-019-00115-4.
[91] Lupše J, Campolo M, Soldati A, Lupse J, Campolo M, Soldati A. Modelling soot [115] Ström H, Sjöblom J, Kannan AS, Ojagh H, Sundborg O, Koegler J. Near-wall
deposition and monolith regeneration for optimal design of automotive DPFs. dispersion, deposition and transformation of particles in automotive exhaust gas
Chem Eng Sci 2016;151:36–50. http://dx.doi.org/10.1016/j.ces.2016.05.008. aftertreatment systems. Int J Heat Fluid Flow 2018;70:171–80. http://dx.doi.
[92] Lao CT, Akroyd J, Eaves N, Smith A, Morgan N, Bhave A, et al. Modelling org/10.1016/j.ijheatfluidflow.2018.02.013.
particle mass and particle number emissions during the active regeneration of [116] Viswanathan S, Rothamer D, Zelenyuk A, Stewart M, Bell D. Experimental
diesel particulate filters. Proc Combust Inst 2019;37(4):4831–8. http://dx.doi. investigation of the effect of inlet particle properties on the capture efficiency
org/10.1016/j.proci.2018.07.079. in an exhaust particulate filter. J Aerosol Sci 2017;113:250–64. http://dx.doi.
[93] Wang Y, Kamp C, Saeid A, Jackson C, Ernstmeyer J, Kharraja B, et al. Modeling org/10.1016/j.jaerosci.2017.08.002.
study of metal fiber diesel particulate filter performance. In: SAE technical [117] Sorensen CM. The mobility of fractal aggregates: A review. Aerosol Sci Technol
papers. 2015, http://dx.doi.org/10.4271/2015-01-1047. 2011;45(7):755–69. http://dx.doi.org/10.1080/02786826.2011.560909.
[94] Yamamoto K, Toda Y. Numerical simulation on flow dynamics and pressure [118] Chen SC, Wang J, Fissan H, Pui DY. Optimizing filtration experiments
variation in porous ceramic filter. Computation 2018;6(4). http://dx.doi.org/ for length and fractal dimension characterization of non-spherical particles.
10.3390/computation6040052. Aerosol Sci Technol 2015;49(8):547–55. http://dx.doi.org/10.1080/02786826.
[95] Gong J, Rutland CJ. Filtration characteristics of fuel neutral particulates 2015.1052038.
using a heterogeneous multiscale filtration model. J Eng Gas Turbines Power [119] Di Sarli V, Di Benedetto A. Modeling and simulation of soot combustion
2015;137(11):111507. http://dx.doi.org/10.1115/1.4030282. dynamics in a catalytic diesel particulate filter. Chem Eng Sci 2015;137:69–78.
http://dx.doi.org/10.1016/j.ces.2015.06.011.
[96] Gong J, Viswanathan S, Rothamer DA, Foster DE, Rutland CJ. Dynamic
[120] Deng Y, Cui J, Jiaqiang E, Zhang B, Zhao X, Zhang Z, et al. Investigations on the
heterogeneous multiscale filtration model: Probing micro- and macroscopic
temperature distribution of the diesel particulate filter in the thermal regener-
filtration characteristics of gasoline particulate filters. Environ Sci Technol
ation process and its field synergy analysis. Appl Therm Eng 2017;123:92–102.
2017;51(19):11196–204. http://dx.doi.org/10.1021/acs.est.7b02535.
http://dx.doi.org/10.1016/j.applthermaleng.2017.05.072.
[97] Lee KW, Gieseke JA. Collection of aerosol particles by packed beds. Environ
[121] Liu X, Chanko T, Lambert C, Maricq M. Gasoline particulate filter efficiency
Sci Technol 1979;13(4):466–70. http://dx.doi.org/10.1021/es60152a013.
and backpressure at very low mileage. In: SAE technical papers. 2018, p. 1–9.
[98] Konstandopoulos AG, Johnson JH. Wall-flow diesel particulate filters—
http://dx.doi.org/10.4271/2018-01-1259.
Their pressure drop and collection efficiency. In: SAE technical paper. SAE
[122] Macián V, Serrano JR, Piqueras P, Sanchis EJ. Internal pore diffusion and
International; 1989, http://dx.doi.org/10.4271/890405.
adsorption impact on the soot oxidation in wall-flow particulate filters. Energy
[99] Serrano JR, Climent H, Piqueras P, Angiolini E. Filtration modelling in wall-flow 2019;179(2):407–21. http://dx.doi.org/10.1016/j.energy.2019.04.200.
particulate filters of low soot penetration thickness. Energy 2016;112:883–98. [123] Gong J, Stewart ML, Zelenyuk A, Strzelec A, Viswanathan S, Rothamer DA, et al.
http://dx.doi.org/10.1016/j.energy.2016.06.121. Importance of filter’s microstructure in dynamic filtration modeling of gasoline
[100] Brown RC. Air filtrationan integrated approach to the theory and applications particulate filters (GPFs): Inhomogeneous porosity and pore size distribution.
of fibrous filters. Pergamon Press; 1993. Chem Eng J 2018;338:15–26. http://dx.doi.org/10.1016/j.cej.2018.01.006.
[101] Fino D, Russo N, Millo F, Vezza DS, Ferrero F, Chianale A. New tool [124] Zhang Z, Yang SL, Johnson JH. Modeling and numerical simulation of diesel
for experimental analysis of diesel particulate filter loading. Top Catalysis particulate trap performance during loading and regeneration. In: SAE 2002
2009;52(13):2083. http://dx.doi.org/10.1007/s11244-009-9393-z. world congress & exhibition. SAE International; 2002, http://dx.doi.org/10.
[102] Ohara E, Mizuno Y, Miyairi Y, Mizutani T, Yuuki K, Noguchi Y, et al. 4271/2002-01-1019.
Filtration behavior of diesel particulate filters (1). In: SAE technical paper. SAE [125] Yamamoto K, Sakai T. Effect of pore structure on soot deposition in
International; 2007, http://dx.doi.org/10.4271/2007-01-0921. diesel particulate filter. Computation 2016;4(4). http://dx.doi.org/10.3390/
[103] Payri F, Arnau FJ, Piqueras P, Ruiz MJ. Lumped approach for flow-through and computation4040046.
wall-flow monolithic reactors modelling for real-time automotive applications. [126] Hayashi H, Kubo S. Computer simulation study on filtration of soot particles
In: SAE technical papers. 2018, p. 1–24. http://dx.doi.org/10.4271/2018-01- in diesel particulate filter. Comput Math Appl 2008;55(7):1450–60. http://dx.
0954. doi.org/10.1016/j.camwa.2007.08.012.

24
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[127] Kladopoulou EA, Yang SL, Johnson JH, Parker GG, Konstandopoulos AG. A [149] Liu J, Swanson JJ, Kittelson DB, Pui DY, Wang J. Microstructural and loading
study describing the performance of diesel particulate filters during loading characteristics of diesel aggregate cakes. Powder Technol 2013;241:244–51.
and regeneration - a lumped parameter model for control applications. In: SAE http://dx.doi.org/10.1016/j.powtec.2013.03.028.
2003 world congress & exhibition. SAE International; 2003, http://dx.doi.org/ [150] Houi D, Lenormand R. Particle deposition on a filter medium. In: Kinetics of
10.4271/2003-01-0842. aggregation and gelation. Elsevier; 1984, p. 173–6. http://dx.doi.org/10.1016/
[128] Konstandopoulos AG, Metallinou R, Dimitrakis DA. Growth and restructuring B978-0-444-86912-8.50041-9.
phenomena of deposits in particulate filters. In: SAE technical papers. 2018, p. [151] Konstandopoulos AG, Kostoglou M, Vlachos N, Kladopoulou E. Progress in diesel
1–8. http://dx.doi.org/10.4271/2018-01-1265. particulate filter simulation. In: SAE 2005 world congress & exhibition. SAE
[129] Masoudi M, Heibel A, Then PM. Predicting pressure drop of wall-flow diesel International; 2005, http://dx.doi.org/10.4271/2005-01-0946.
particulate filters - theory and experiment. In: SAE technical paper. SAE [152] Voutsi O, Tsinoglou D, Karamitros D, Koltsakis G. Pressure drop of particulate
International; 2000, http://dx.doi.org/10.4271/2000-01-0184. filters and correlation with the deposited soot for heavy-duty engines. In: SAE
[130] Konstandopoulos AG, Skaperdas E, Masoudi M. Inertial contributions to the technical papers. 2019, p. 692–701. http://dx.doi.org/10.4271/2019-24-0151.
pressure drop of diesel particulate filters. In: SAE 2001 world congress. SAE [153] Qu D, Li J, Liu Y. Research on particulate filter simulation and regeneration
International; 2001, http://dx.doi.org/10.4271/2001-01-0909. control strategy. Mech Syst Signal Process 2017;87:214–26. http://dx.doi.org/
[131] Watling TC, Ravenscroft MR, Cleeton JP, Rees ID, Wilkins DA. Development of a 10.1016/j.ymssp.2016.05.039.
particulate filter model for the prediction of backpressure: Improved momentum [154] Zhang B, Jiaqiang E, Gong J, Yuan W, Zhao X, Hu W. Influence of structural
balance and entrance and exit effect equations. SAE Int J Engines 2017;10(4). and operating factors on performance degradation of the diesel particulate
http://dx.doi.org/10.4271/2017-01-0974. filter based on composite regeneration. Appl Therm Eng 2017;121:838–52.
[132] Watling TC, Ravenscroft MR, Avery G. Development, validation and application http://dx.doi.org/10.1016/j.applthermaleng.2017.04.155.
of a model for an SCR catalyst coated diesel particulate filter. Catalysis Today [155] Wang Y, Pan Y, Su C, Srinivasan A, Gong J, Kamp CJ. Performance of
2012;188(1):32–41. http://dx.doi.org/10.1016/j.cattod.2012.02.007. asymmetric particulate filter with soot and ash deposits: Analytical solution
[133] Custer N, Kamp CJ, Sappok A, Pakko J, Lambert C, Boerensen C, et al. and its application. Ind Eng Chem Res 2018;57(46):15846–56. http://dx.doi.
Lubricant-derived ash impact on gasoline particulate filter performance. SAE org/10.1021/acs.iecr.8b02848.
Int J Engines 2016;9(3):1604–14. http://dx.doi.org/10.4271/2016-01-0942. [156] Zhong C, Gong J, Liu W, Liu G. Low temperature, medium temperature and
[134] Lee K, Reed L, Gieseke J. Pressure drop across packed beds in the low Knudsen high temperature performance of the continuous regenerative diesel particulate
number regime. J Aerosol Sci 1978;9(6):557–65. http://dx.doi.org/10.1016/ filter assisted by electric regeneration. Chem Eng Sci 2019;207:980–92. http:
0021-8502(78)90021-6. //dx.doi.org/10.1016/j.ces.2019.07.024.
[135] Versaevel P, Colas H, Rigaudeau C, Noirot R, Koltsakis GC, Stamatelos AM. [157] Deng Y, Zheng W, Jiaqiang E, Zhang B, Zhao X, Zuo Q, et al. Influence
Some empirical observations on diesel particulate filter modeling and com- of geometric characteristics of a diesel particulate filter on its behavior in
parison between simulations and experiments. In: SAE technical paper. 2000, equilibrium state. Appl Therm Eng 2017;123:61–73. http://dx.doi.org/10.1016/
http://dx.doi.org/10.4271/2000-01-0477. j.applthermaleng.2017.05.071.
[136] Mahadevan BS, Johnson JH, Shahbakhti M. Predicting pressure drop, tem- [158] Jiaqiang E, Zuo W, Gao J, Peng Q, Zhang Z, Hieu PM. Effect analysis on
perature, and particulate matter distribution of a catalyzed diesel particulate pressure drop of the continuous regeneration-diesel particulate filter based on
filter using a multi-zone model including cake permeability. Emiss Control Sci NO2 assisted regeneration. Appl Therm Eng 2016;100:356–66. http://dx.doi.
Technol 2017;3(2):171–201. org/10.1016/j.applthermaleng.2016.02.031.
[137] Tronconi E, Nova I, Marchitti F, Koltsakis G, Karamitros D, Maletic B, et al. [159] Jiaqiang E, Xie L, Zuo Q, Zhang G. Effect analysis on regeneration speed of
Interaction of NOx reduction and soot oxidation in a DPF with Cu-zeolite continuous regeneration-diesel particulate filter based on NO2 -assisted regener-
SCR coating. Emiss Control Sci Technol 2015;1(x):134–51. http://dx.doi.org/ ation. Atmospheric Pollut Res 2016;7(1):9–17. http://dx.doi.org/10.1016/j.apr.
10.1007/s40825-015-0014-y. 2015.06.012.
[138] Serrano JR, Arnau FJ, Piqueras P, Óscar García-Afonso. Packed bed of spherical [160] Mitsouridis MA, Karamitros D, Koltsakis G. Model-based analysis of TWC-coated
particles approach for pressure drop prediction in wall-flow DPFs (diesel filters performance. Emiss Control Sci Technol 2019;238–52. http://dx.doi.org/
particulate filters) under soot loading conditions. Energy 2013;58:644–54. http: 10.1007/s40825-019-00124-3.
//dx.doi.org/10.1016/j.energy.2013.05.051. [161] Endo Y, Chen D-R, Pui DY. Theoretical consideration of permeation resistance
[139] Kamp CJ, Zhang S, Bagi S, Wong V, Monahan G, Sappok A, et al. Ash of fluid through a particle packed layer. Powder Technol 2002;124(1):119–26.
permeability determination in the diesel particulate filter from ultra-high http://dx.doi.org/10.1016/S0032-5910(01)00479-X.
resolution 3D X-Ray imaging and image-based direct numerical simulations. SAE [162] Thomas D, Ouf F, Gensdarmes F, Bourrous S, Bouilloux L. Pressure drop
Int J Fuels Lubricants 2017;10(2). http://dx.doi.org/10.4271/2017-01-0927. model for nanostructured deposits. Sep Purif Technol 2014;138:144–52. http:
[140] Sappok A. The nature of lubricant-derived ash-related emissions and their im- //dx.doi.org/10.1016/j.seppur.2014.09.032.
pact on diesel aftertreatment system performance. (Ph.D. thesis), Massachusetts [163] Ribeyre Q, Charvet A, Vallières C, Thomas D. Impact of relative humidity on
Institue of Technology; 2009. a nanostructured filter cake – experimental and modelling approaches. Chem
[141] Zhang J, Wong VW, Shuai S, Chen Y, Sappok A. Quantitative estimation Eng Sci 2017;161:109–16. http://dx.doi.org/10.1016/j.ces.2016.12.013.
of the impact of ash accumulation on diesel particulate filter related fuel [164] Opitz B, Votsmeier M. A simulation study on the conversion efficiency of
penalty for a typical modern on-road heavy-duty diesel engine. Appl Energy catalytically active particulate filters. Chem Eng Sci 2016. http://dx.doi.org/
2018;229:1010–23. http://dx.doi.org/10.1016/j.apenergy.2018.08.071. 10.1016/j.ces.2016.04.010.
[142] Thomas D, Penicot P, Contal P, Leclerc D, Vendel J. Clogging of fibrous filters [165] Purfürst M, Naumov S, Langeheinecke K-J, Gläser R. Influence of soot on
by solid aerosol particles experimental and modelling study. Chem Eng Sci ammonia adsorption and catalytic DeNOx -properties of diesel particulate filters
2001;56(11):3549–61. http://dx.doi.org/10.1016/S0009-2509(01)00041-0. coated with SCR-catalysts. Chem Eng Sci 2017;168:423–36. http://dx.doi.org/
[143] Uenishi T, Tanaka E, Shigeno G, Fukuma T, Kusaka J, Daisho Y. A quasi two 10.1016/j.ces.2017.04.052.
dimensional model of transport phenomena in diesel particulate filters - the [166] Mihai O, Tamm S, Stenfeldt M, Olsson L. The effect of soot on ammonium
effects of particle and wall pore diameter on the pressure drop. In: JSAE/SAE nitrate species and NO2 selective catalytic reduction over Cu–zeolite catalyst-
2015 international powertrains, fuels & lubricants meeting. SAE International; coated particulate filter. Phil Trans R Soc A 2016;374(2061):20150086. http:
2015, http://dx.doi.org/10.4271/2015-01-2010. //dx.doi.org/10.1098/rsta.2015.0086.
[144] Tsujimoto D, Jin K, Fukuma T. A statistical approach to improve the accuracy of [167] Mihai O, Stenfeldt M, Olsson L. The effect of changing the gas compo-
the DPF simulation model under transient conditions. In: SAE technical papers. sition on soot oxidation over DPF and SCR-coated filters. Catalysis Today
2019, p. 1–10. http://dx.doi.org/10.4271/2019-01-0027. 2018;306:243–50. http://dx.doi.org/10.1016/j.cattod.2017.03.005.
[145] Premchand KC, Johnson JH, Yang S-L. Development of a 1-D catalyzed diesel [168] Mihai O, Tamm S, Stenfeldt M, Wang-Hansen C, Olsson L. Evaluation of an
particulate filter model for simulation of the oxidation of particulate matter integrated selective catalytic reduction-coated particulate filter. Ind Eng Chem
and gaseous species during passive oxidation and active regeneration. In: SAE Res 2015;54(47):11779–91. http://dx.doi.org/10.1021/acs.iecr.5b02392.
technical paper. SAE International; 2013, http://dx.doi.org/10.4271/2013-01- [169] Chigada PI, Ahmadinejad M, Newman AD, Ng AIP, Torbati R, Watling TC.
1574. Impact of SCR activity on soot regeneration and the converse effects of soot
[146] Haralampous OA, Koltsakis GC. Oxygen diffusion modeling in diesel particulate regeneration on SCR activity on a vanadia-SCRF® . In: SAE technical papers.
filter regeneration. AIChE J 2004;50(9):2008–19. http://dx.doi.org/10.1002/ (2):2018, p. 1–10. http://dx.doi.org/10.4271/2018-01-0962.
aic.10181. [170] Greiner R, Prill T, Iliev O, van Setten BA, Votsmeier M. Tomography based
[147] Watling TC. Geometric description of the soot cake in a one-dimensional model simulation of reactive flow at the micro-scale: Particulate filters with wall
of an octo-square asymmetric particulate filter. In: SAE technical papers. 2019, integrated catalyst. Chem Eng J 2019;378:121919. http://dx.doi.org/10.1016/
p. 1–23. http://dx.doi.org/10.4271/2019-01-0991. j.cej.2019.121919.
[148] Konstandopoulos AG, Skaperdas E, Masoudi M. Microstructural properties of [171] Mladenov N, Koop J, Tischer S, Deutschmann O. Modeling of transport and
soot deposits in diesel particulate traps. In: SAE 2002 world congress and chemistry in channel flows of automotive catalytic converters. Chem Eng Sci
exhibition. SAE International; 2002, http://dx.doi.org/10.4271/2002-01-1015. 2010;65(2):812–26. http://dx.doi.org/10.1016/j.ces.2009.09.034.

25
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[172] Zheng H, Keith JM. Ignition analysis of wall-flow monolith diesel particulate [197] Miceli P, Bensaid S, Russo N, Fino D. Effect of the morphological and surface
filters. Catalysis Today 2004;98(3):403–12. http://dx.doi.org/10.1016/j.cattod. properties of CeO2 -based catalysts on the soot oxidation activity. Chem Eng J
2004.08.008. 2015;278:190–8. http://dx.doi.org/10.1016/j.cej.2014.10.055.
[173] Chen K, Martirosyan KS, Luss D. Wrong-way behavior of soot combustion [198] Andana T, Piumetti M, Bensaid S, Veyre L, Thieuleux C, Russo N, et al. Ceria-
in a planar diesel particulate filter. Ind Eng Chem Res 2009;48(18):8451–6. supported small Pt and Pt3 Sn nanoparticles for NOx -assisted soot oxidation.
http://dx.doi.org/10.1021/ie900848d. Appl Catal B 2017;209(x):295–310. http://dx.doi.org/10.1016/j.apcatb.2017.
[174] Chen K, Luss D. Temperature excursions in diesel particulate filters: Response 03.010.
to shift to idle. Ind Eng Chem Res 2011;50(2):832–42. http://dx.doi.org/10. [199] Andana T, Piumetti M, Bensaid S, Russo N, Fino D, Pirone R. CO and soot
1021/ie101844r. oxidation over Ce-Zr-Pr oxide catalysts. Nanoscale Res Lett 2016;11(1). http:
[175] Galindo J, Serrano JR, Piqueras P, Óscar García-Afonso. Heat transfer modelling //dx.doi.org/10.1186/s11671-016-1494-6.
in honeycomb wall-flow diesel particulate filters. Energy 2012;43(1):201–13.
[200] Bensaid S, Piumetti M, Novara C, Giorgis F, Chiodoni A, Russo N, et al. Catalytic
http://dx.doi.org/10.1016/j.energy.2012.04.044.
oxidation of CO and soot over Ce-Zr-Pr mixed oxides synthesized in a multi-inlet
[176] Yu M, Luss D, Balakotaiah V. Analysis of flow distribution and heat transfer in
vortex reactor: Effect of structural defects on the catalytic activity. Nanoscale
a diesel particulate filter. Chem Eng J 2013;226:68–78. http://dx.doi.org/10.
Res Lett 2016;11(1):1–14. http://dx.doi.org/10.1186/s11671-016-1713-1.
1016/j.cej.2013.04.026.
[201] Piumetti M, Bensaid S, Russo N, Fino D. Nanostructured ceria-based catalysts
[177] Yu M, Luss D, Balakotaiah V. Analysis of ignition in a diesel particulate filter.
for soot combustion: Investigations on the surface sensitivity. Appl Catal B
Catalysis Today 2013;216:158–68. http://dx.doi.org/10.1016/j.cattod.2013.05.
2015;165:742–51. http://dx.doi.org/10.1016/j.apcatb.2014.10.062.
003.
[178] Yu M, Luss D, Balakotaiah V. Regeneration modes and peak temperatures in [202] Zouaoui N, Labaki M, Jeguirim M. Diesel soot oxidation by nitrogen dioxide,
a diesel particulate filter. Chem Eng J 2013;232:541–54. http://dx.doi.org/10. oxygen and water under engine exhaust conditions: Kinetics data related to
1016/j.cej.2013.08.006. the reaction mechanism. Comptes Rendus Chimie 2014;17(7):672–80. http:
[179] Chen T, Wu Z, Gong J, Jiaqiang E. Numerical simulation of diesel par- //dx.doi.org/10.1016/j.crci.2013.09.004.
ticulate filter regeneration considering ash deposit. Flow Turbul Combust [203] Ramdas R, Nowicka E, Jenkins R, Sellick D, Davies C, Golunski S. Using real
2016;97(3):849–64. http://dx.doi.org/10.1007/s10494-016-9717-6. particulate matter to evaluate combustion catalysts for direct regeneration of
[180] Haralampous O, Koltsakis GC. Intra-layer temperature gradients during re- diesel soot filters. Appl Catal B 2015;176:436–43. http://dx.doi.org/10.1016/j.
generation of diesel particulate filters. Chem Eng Sci 2002;57(13):2345–55. apcatb.2015.04.031.
http://dx.doi.org/10.1016/S0009-2509(02)00135-5. [204] Wurzenberger JC, Triebl C, Kutschi S. Diffusion supporting passive filter
[181] Meng Z, Zhang J, Chen C, Yan Y. A numerical investigation of the diesel regeneration- a modeling contribution on coated filters. In: SAE technical
particle filter regeneration process under temperature pulse conditions. Heat papers. 2018, p. 1–15. http://dx.doi.org/10.4271/2018-01-0957.
Mass Transf 2016. http://dx.doi.org/10.1007/s00231-016-1924-0. [205] Wurzenberger JC, Kutschi S, Nikodem A. Ash transport and deposition, cake
[182] Arunachalam H, Pozzato G, Hoffman MA, Onori S. Modeling the thermal formation and segregation-a modeling study on the impact of ash on particulate
and soot oxidation dynamics inside a ceria-coated gasoline particulate filter. filter performance. In: SAE technical papers. 2019, 2019, p. 1–16. http://dx.
Control Eng Pract 2020;94:104199. http://dx.doi.org/10.1016/j.conengprac. doi.org/10.4271/2019-01-0988.
2019.104199. [206] Vernikovskaya N, Pavlova T, Mokrinskii V, Murzin D, Chumakova N, Noskov A.
[183] Kostoglou M, Bissett EJ, Konstandopoulos AG. Improved transfer coefficients for Soot particulates abatement in diesel engine exhaust by catalytic oxidation
wall-flow monolithic catalytic reactors: Energy and momentum transport. Ind followed their trapping in filters. Chem Eng J 2015;269:416–24. http://dx.doi.
Eng Chem Res 2012;51(40):13062–72. http://dx.doi.org/10.1021/ie3011098. org/10.1016/j.cej.2015.01.129.
[184] Depcik C, Assanis D. Simulating area conservation and the gas-wall interface
[207] Bai S, Tang J, Wang G, Li G. Soot loading estimation model and passive regen-
for one-dimensional based diesel particulate filter models. J Eng Gas Turbines
eration characteristics of DPF system for heavy-duty engine. Appl Therm Eng
Power 2008;130(6). http://dx.doi.org/10.1115/1.2939002.
2016;100:1292–8. http://dx.doi.org/10.1016/j.applthermaleng.2016.02.055.
[185] Di Sarli V, Di Benedetto A. Operating map for regeneration of a catalytic diesel
[208] Fang J, Shi R, Meng Z, Jiang Y, Qin Z, Zhang Q, et al. The interaction effect of
particulate filter. Ind Eng Chem Res 2016;55(42):11052–61. http://dx.doi.org/
catalyst and ash on diesel soot oxidation by thermogravimetric analysis. Fuel
10.1021/acs.iecr.6b02521.
2019;258:116151. http://dx.doi.org/10.1016/j.fuel.2019.116151.
[186] Schejbal M, Marek M, Kubíček M, Kočí P. Modelling of diesel filters for
particulates removal. Chem Eng J 2009;154(1–3):219–30. http://dx.doi.org/10. [209] Kastrinaki G, Lorentzou S, Konstandopoulos AG. Soot oxidation kinetics of dif-
1016/j.cej.2009.04.056. ferent ceria nanoparticle catalysts. Emiss Control Sci Technol 2015;1(3):247–53.
[187] Lapuerta M, Rodríguez-Fernández J, Sánchez-Valdepeñas J. Soot reactivity http://dx.doi.org/10.1007/s40825-015-0021-z.
analysis and implications on diesel filter regeneration. Prog Energy Combust [210] Zhang J, Qi J, Shuai SJ, Wang L, Liu S, Wang G, et al. Numerical analysis on
Sci 2020;78. http://dx.doi.org/10.1016/j.pecs.2020.100833. the potential of reducing DPF size using low ash lubricant oil. In: SAE technical
[188] Kandylas IP, Koltsakis GC. NO2 -assisted regeneration of diesel particulate filters: papers. 2018, 2018, p. 1–11. http://dx.doi.org/10.4271/2018-01-1760.
A modeling study. Ind Eng Chem Res 2002;41(9):2115–23. http://dx.doi.org/ [211] Konstandopoulos AG, Pagkoura C, Lorentzou S, Kastrinaki G. Catalytic soot
10.1021/ie010842m. oxidation: Effect of ceria-zirconia catalyst particle size. SAE Int J Engines
[189] Schejbal M, Štěpánek J, Marek M, Kočí P, Kubíček M. Modelling of 2016;9(3). http://dx.doi.org/10.4271/2016-01-0968.
soot oxidation by NO2 in various types of diesel particulate filters. Fuel [212] Mahadevan BS, Johnson JH, Shahbakhti M. Development of a catalyzed diesel
2010;89(9):2365–75. http://dx.doi.org/10.1016/j.fuel.2010.04.018. particulate filter multi-zone model for simulation of axial and radial substrate
[190] Jeguirim M, Tschamber V, Brilhac J, Ehrburger P. Oxidation mechanism of temperature and particulate matter distribution. Emiss Control Sci Technol
carbon black by NO2 : Effect of water vapour. Fuel 2005;84(14):1949–56. 2015;1(2):183–202. http://dx.doi.org/10.1007/s40825-015-0015-x.
http://dx.doi.org/10.1016/j.fuel.2005.03.026. [213] Neeft J. Kinetics of the oxidation of diesel soot. Fuel 1997;76(12):1129–36.
[191] Bernardoff R, Hennebert B, Delvigne T, Courtois O, China P. A study of ash http://dx.doi.org/10.1016/S0016-2361(97)00119-1.
accumulation in the after-treatment system of a gasoline direct injection engine [214] Darcy P, Da Costa P, Mellottée H, Trichard J-M, Djéga-Mariadassou G. Kinetics
equipped with a gasoline particulate filter. In: SAE technical papers. 2017, of catalyzed and non-catalyzed oxidation of soot from a diesel engine. Catalysis
http://dx.doi.org/10.4271/2017-01-0879. Today 2007;119(1):252–6. http://dx.doi.org/10.1016/j.cattod.2006.08.056.
[192] Pérez VR, Bueno-López A. Catalytic regeneration of diesel particulate filters:
[215] Bogarra M, Herreros JM, Tsolakis A, Rodríguez-Fernández J, York AP, Milling-
Comparison of Pt and CePr active phases. Chem Eng J 2015;279:79–85. http:
ton PJ. Gasoline direct injection engine soot oxidation: Fundamentals and
//dx.doi.org/10.1016/j.cej.2015.05.004.
determination of kinetic parameters. Combust Flame 2018;190:177–87. http:
[193] Di Sarli V, Landi G, Lisi L, Saliva A, Di Benedetto A. Catalytic diesel particulate
//dx.doi.org/10.1016/j.combustflame.2017.11.027.
filters with highly dispersed ceria: Effect of the soot-catalyst contact on the
[216] Stanmore B, Brilhac J, Gilot P. The oxidation of soot: a review of experiments,
regeneration performance. Appl Catal B 2016;197:116–24. http://dx.doi.org/
mechanisms and models. Carbon 2001;39(15):2247–68. http://dx.doi.org/10.
10.1016/j.apcatb.2016.01.073.
1016/S0008-6223(01)00109-9.
[194] Piumetti M, Andana T, Bensaid S, Russo N, Fino D, Pirone R. Study on the CO
oxidation over ceria-based nanocatalysts. Nanoscale Res Lett 2016;11(1):165. [217] Luo Y, Zhu L, Fang J, Zhuang Z, Guan C, Xia C, et al. Size distribution, chemical
http://dx.doi.org/10.1186/s11671-016-1375-z. composition and oxidation reactivity of particulate matter from gasoline direct
[195] Jeguirim M, Tschamber V, Brilhac JF. Kinetics of catalyzed and non-catalyzed injection (GDI) engine fueled with ethanol-gasoline fuel. Appl Therm Eng
soot oxidation with nitrogen dioxide under regeneration particle trap con- 2015;89:647–55. http://dx.doi.org/10.1016/j.applthermaleng.2015.06.060.
ditions. J Chem Technol Biotechnol 2009;84(5):770–6. http://dx.doi.org/10. [218] Ahmadinejad M, Etheridge JE, Watling TC, Johansson A, John G. Computer
1002/jctb.2110. simulation of automotive emission control systems. Johnson Matthey Technol
[196] Piumetti M, van der Linden B, Makkee M, Miceli P, Fino D, Russo N, Rev 2015;59(2):152–65. http://dx.doi.org/10.1595/205651315X687876.
et al. Contact dynamics for a solid-solid reaction mediated by gas-phase [219] Easter J. Influence of fuel introduction parameters on the reactivity and
oxygen: Study on the soot oxidation over ceria-based catalysts. Appl Catal B oxidation process of soot from a gasoline direct injection engine. (Ph.D. thesis),
2016;199:96–107. http://dx.doi.org/10.1016/j.apcatb.2016.06.006. University of Michigan; 2017.

26
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[220] Yezerets A, Currier NW, Kim DH, Eadler HA, Epling WS, Peden CH. Differential [242] Choi S, Oh K-C, Lee C-B. The effects of filter porosity and flow conditions
kinetic analysis of diesel particulate matter (soot) oxidation by oxygen using a on soot deposition/oxidation and pressure drop in particulate filters. Energy
step–response technique. Appl Catal B 2005;61(1):120–9. http://dx.doi.org/10. 2014;77:327–37. http://dx.doi.org/10.1016/j.energy.2014.08.049.
1016/j.apcatb.2005.04.014. [243] Picandet V, Khelidj A, Bastian G. Effect of axial compressive damage on
[221] Singh M, Srilomsak M, Wang Y, Hanamura K, Vander Wal R. Nanos- gas permeability of ordinary and high-performance concrete. Cem Concr Res
tructure changes in diesel soot during NO2 –O2 oxidation under diesel 2001;31(11):1525–32. http://dx.doi.org/10.1016/S0008-8846(01)00546-4.
particulate filter-like conditions toward filter regeneration. Int J Engine Res [244] Foley R, Johnson J, Naber J, Rogoski L. Experimental measurements of
2019;20(8–9):953–66. http://dx.doi.org/10.1177/1468087418807608. particulate matter distribution in a catalyzed particulate filter. Emiss Control
[222] Jung H, Kittelson DB, Zachariah MR. Kinetics and visualization of soot oxidation Sci Technol 2015;1(1):32–48. http://dx.doi.org/10.1007/s40825-014-0005-4.
using transmission electron microscopy. Combust Flame 2004;136(4):445–56. [245] Haralampous OA, Koltsakis GC. Back-diffusion modeling of NO2 in catalyzed
http://dx.doi.org/10.1016/j.combustflame.2003.10.013. diesel particulate filters. Ind Eng Chem Res 2004;43(4):875–83. http://dx.doi.
[223] Kandas AW, Gokhan Senel I, Levendis Y, Sarofim AF. Soot surface area org/10.1021/ie034187p.
evolution during air oxidation as evaluated by small angle X-ray scattering [246] Wen JZ, Thomson MJ, Lightstone MF, Park SH, Rogak SN. An improved moving
and CO2 adsorption. Carbon 2005;43(2):241–51. http://dx.doi.org/10.1016/j. sectional aerosol model of soot formation in a plug flow reactor. Combust Sci
carbon.2004.08.028. Technol 2006;178(5):921–51. http://dx.doi.org/10.1080/00102200500270007.
[224] Hong J, Hecker WC, Fletcher TH. Improving the accuracy of predicting [247] Fukui R, Okamoto Y, Nakao M. Experimental analysis of sudden pressure
effectiveness factors for mth order and langmuir rate equations in spheri- increase phenomenon by real-time internal observation of diesel particulate
cal coordinates. Energy Fuels 2000;14(3):663–70. http://dx.doi.org/10.1021/ filter. J Eng Gas Turbines Power 2016;138(10):102803. http://dx.doi.org/10.
ef9902193. 1115/1.4033061.
[225] Salatino P, Zimbardi F. A fractal approach to the analysis of low temperature [248] Kim K, Mital R, Higuchi T. An investigative study of sudden pressure increase
combustion rate of a coal char. II: Model development. Carbon 1994;32(1):51–9. phenomenon across the SCR on filter catalyst. In: SAE technical papers. 2016,
http://dx.doi.org/10.1016/0008-6223(94)90008-6. 2016, http://dx.doi.org/10.4271/2016-01-2319.
[226] Trandafilović LV, Mihai O, Woo J, Leistner K, Stenfeldt M, Olsson L. A kinetic [249] Beatrice C, Costagliola MA, Guido C, Napolitano P, Prati MV. How much
model for SCR coated particulate filters—Effect of ammonia-soot interactions. regeneration events influence particle emissions of DPF-equipped vehicles? In:
Appl Catal B 2019;241:66–80. http://dx.doi.org/10.1016/j.apcatb.2018.08.076. 13th international conference on engines & vehicles. SAE International; 2017,
[227] Ye P, Sun C, Lapuerta M, Agudelo J, Wal RV, Boehman AL, et al. Im- http://dx.doi.org/10.4271/2017-24-0144.
pact of rail pressure and biodiesel fueling on the particulate morphology
[250] Yamada H, Inomata S, Tanimoto H. Particle and VOC emissions from stoi-
and soot nanostructures from a common-rail turbocharged direct injection
chiometric gasoline direct injection vehicles and correlation between particle
diesel engine. Int J Engine Res 2016;17(2):193–208. http://dx.doi.org/10.1177/
number and mass emissions. Emiss Control Sci Technol 2017. http://dx.doi.
1468087414564229.
org/10.1007/s40825-016-0060-0.
[228] Zygogianni A, Syrigou M, Konstandopoulos AG, Kostoglou M. Oxidative re-
[251] Cedric L, Goriaux M, Tassel P, Perret P, Andre M, Liu Y. Impact of af-
activity of particulate samples from different diesel combustion systems and
tertreatment device and driving conditions on black carbon, ultrafine particle
its relation to structural and spectral characteristics of soot. Emiss Control Sci
and NOx emissions for euro 5 diesel and gasoline vehicles. Transp Res Proc
Technol 2019;5(2):99–123. http://dx.doi.org/10.1007/s40825-019-00118-1.
2016;14(2):3079–88. http://dx.doi.org/10.1016/j.trpro.2016.05.454.
[229] Rothe D, Knauer M, Emmerling G, Deyerling D, Niessner R. Emissions during
[252] R’Mili B, Boréave A, Meme A, Vernoux P, Leblanc M, Noël L, et al. Physico-
active regeneration of a diesel particulate filter on a heavy duty diesel engine:
chemical characterization of fine and ultrafine particles emitted during diesel
Stationary tests. J Aerosol Sci 2015;90:14–25. http://dx.doi.org/10.1016/j.
particulate filter active regeneration of Euro5 diesel vehicles. Environ Sci
jaerosci.2015.07.007.
Technol 2018;52(5):3312–9. http://dx.doi.org/10.1021/acs.est.7b06644.
[230] Choi S, Seong H. Oxidation characteristics of gasoline direct-injection (GDI)
[253] Smith JD, Ruehl C, Burnitzki M, Sobieralski W, Ianni R, Quiros D, et al.
engine soot: Catalytic effects of ash and modified kinetic correlation. Combust
Real-time particulate emissions rates from active and passive heavy-duty diesel
Flame 2015;162(6):2371–89. http://dx.doi.org/10.1016/j.combustflame.2015.
particulate filter regeneration. Sci Total Environ 2019;680:132–9. http://dx.doi.
02.004.
org/10.1016/j.scitotenv.2019.04.447.
[231] Strzelec A, Vander Wal RL, Lewis SA, Toops TJ, Daw CS. Nanostructure
[254] Ruehl C, Herner JD, Yoon S, Collins JF, Misra C, Na K, et al. Similarities
and burning mode of light-duty diesel particulate with conventional diesel,
and differences between ‘‘traditional’’ and ‘‘clean’’ diesel PM. Emiss Control Sci
biodiesel, and intermediate blends. Int J Engine Res 2017;18(5–6):520–31.
Technol 2015;1(1):17–23. http://dx.doi.org/10.1007/s40825-014-0002-7.
http://dx.doi.org/10.1177/1468087416674414.
[255] Dittler A. Ash transport in diesel particle filters. In: SAE 2012 international
[232] Choi S, Seong H. Lube oil-dependent ash chemistry on soot oxidation reactivity
powertrains, fuels & lubricants meeting. SAE International; 2012, http://dx.
in a gasoline direct-injection engine. Combust Flame 2016;174:68–76. http:
doi.org/10.4271/2012-01-1732.
//dx.doi.org/10.1016/j.combustflame.2016.09.019.
[233] Stȩpień Z, Ziemiański L, Żak G, Wojtasik M, Jȩczmionek Ł, Burnus Z. The evalu- [256] Rubino L, Piotr Oles J, La Rocca A. Evaluating performance of uncoated GPF
ation of fuel borne catalyst (FBC’s) for DPF regeneration. Fuel 2015;161:278–86. in real world driving using experimental results and CFD modelling. In: SAE
http://dx.doi.org/10.1016/j.fuel.2015.08.071. technical papers. 2017, 2017, http://dx.doi.org/10.4271/2017-24-0128.
[234] Jeguirim M, Tschamber V, Ehrburger P. Catalytic effect of platinum on the [257] Seong H, Choi S, Matusik KE, Kastengren AL, Powell CF. 3D pore analysis of
kinetics of carbon oxidation by NO2 and O2 . Appl Catal B 2007;76(3):235–40. gasoline particulate filters using X-ray tomography: impact of coating and ash
http://dx.doi.org/10.1016/j.apcatb.2007.05.029. loading. J Mater Sci 2019;54(8):6053–65. http://dx.doi.org/10.1007/s10853-
[235] Marchitti F, Tronconi E. Experimental study of the interaction between soot 018-03310-w.
combustion and NH3 -SCR reactivity over a Cu–Zeolite SDPF catalyst. Catalysis [258] Kamp CJ, Folino P, Sappok A, Wong VW, Wang Y, Ernstmeyer J, et al. Ash
Today 2016;267:110–8. http://dx.doi.org/10.1016/j.cattod.2016.01.027. accumulation and impact on sintered metal fiber diesel particulate filters. SAE
[236] Schobing J, Tschamber V, Brilhac J-F, Auclaire A, Hohl Y. Simultaneous Int J Fuels Lubricants 2015;8(2):487–93. http://dx.doi.org/10.4271/2015-01-
soot combustion and NOx reduction over a vanadia-based selective catalytic 1012.
reduction catalyst. Comptes Rendus Chimie 2018;21(3):221–31. http://dx.doi. [259] Matsuno M, Kitamura T. Direct visualization of soot and ash transport in diesel
org/10.1016/j.crci.2017.03.002. particulate filters during active regeneration process. In: SAE technical papers.
[237] Liu Y, Su C, Clerc J, Harinath A, Rogoski L. Experimental and modeling study of 2019, http://dx.doi.org/10.4271/2019-01-2287.
ash impact on DPF backpressure and regeneration behaviors. SAE Int J Engines [260] Shao Y. Physics and modelling of wind erosion. Springer; 2008.
2015;8(3). http://dx.doi.org/10.4271/2015-01-1063. [261] Wang Y, Obuchi Y, Zhang J, Tracy I, Wong V. Experiments and analyses
[238] Bissett EJ. Mathematical model of the thermal regeneration of a wall-flow on stability/mid-channel collapse of ash-deposit wall layers and pre-mature
monolith diesel particulate filter. Chem Eng Sci 1984;39(7):1233–44. http: clogging of diesel particulate filters. In: SAE technical papers. 2019, 2019, p.
//dx.doi.org/10.1016/0009-2509(84)85084-8. 1–20. http://dx.doi.org/10.4271/2019-01-0972.
[239] Kostoglou M, Konstandopoulos AG. Effect of soot layer microstructure on diesel [262] Zarvalis D, Lorentzou S, Konstandopoulos AG. A methodology for the fast
particulate filter regeneration. AIChE J 2005;51(9):2534–46. http://dx.doi.org/ evaluation of the effect of ash aging on the diesel particulate filter performance.
10.1002/aic.10484. In: SAE technical paper. SAE International; 2009, http://dx.doi.org/10.4271/
[240] Toops TJ, Pihl JA, Finney CEA, Gregor J, Bilheux H. Progression of soot 2009-01-0630.
cake layer properties during the systematic regeneration of diesel particu- [263] Sappok A, Wong VW. Ash effects on diesel particulate filter pressure drop
late filters measured with neutron tomography. Emiss Control Sci Technol sensitivity to soot and implications for regeneration frequency and DPF control.
2015;1(1):24–31. http://dx.doi.org/10.1007/s40825-014-0008-1. SAE Int J Fuels Lubricants 2010;3(1):380–96. http://dx.doi.org/10.4271/2010-
[241] Kamp CJ, Sappok A, Wang Y, Bryk W, Rubin A, Wong V. Direct measurements 01-0811.
of soot/ash affinity in the diesel particulate filter by atomic force microscopy [264] Bagi S, Kamp CJ, Sharma V, Aswath PB. Multiscale characterization of exhaust
and implications for ash accumulation and DPF degradation. SAE Int J Fuels and crankcase soot extracted from heavy-duty diesel engine and implications for
Lubr 2014;7:307–16. http://dx.doi.org/10.4271/2014-01-1486. DPF ash. Fuel 2020;282:118878. http://dx.doi.org/10.1016/j.fuel.2020.118878.

27
C.T. Lao et al. Progress in Energy and Combustion Science 96 (2023) 101043

[265] Kamp CJ, Bagi S, Wang Y. Phenomenological investigations of mid-channel Chung Ting Lao joined the CoMo Group, Department
ash deposit formation and characteristics in diesel particulate filters. In: SAE of Chemical Engineering and Biotechnology, University of
technical papers. 2019, 2019, p. 1–11. http://dx.doi.org/10.4271/2019-01- Cambridge as a Ph.D. student in 2015 after he completed
0973. his M.Eng. in Chemical Engineering at the University of
[266] Koltsakis G, Mitsouridis M, Mylonidis I, Baumgard K, Duddukuri R, Zhou W, Cambridge. His research focuses on developing numerical
et al. Towards predictive ash accumulation and transport modeling. 2020, model for exhaust after-treatment system, an essential com-
Presented at CLEERS 2020 on 15/09/2020, URL https://cleers.org/cleers- ponent in internal combustion engine systems for emission
workshops/workshop-presentations/entry/2755/. reduction.
[267] Sappok A, Wang Y, Wang R-Q, Kamp C, Wong V. Theoretical and experimental
analysis of ash accumulation and mobility in ceramic exhaust particulate
filters and potential for improved ash management. SAE Int J Fuels Lubr
2014;7:511–24.
[268] Konstandopoulos AG, Kostoglou M, Housiada P, Vlachos N, Zarvalis D. Mul-
tichannel simulation of soot oxidation in diesel particulate filters. In: SAE
Dr Jethro Akroyd is a Fellow of Churchill College and a
technical paper. SAE International; 2003, http://dx.doi.org/10.4271/2003-01-
Senior Research Associate in the CoMo Group, Department
0839.
of Chemical Engineering and Biotechnology, University of
[269] Zhang B, Jiaqiang E, Gong J, Yuan W, Zuo W, Li Y, et al. Multidisciplinary
Cambridge. He has a strong interest in the area of com-
design optimization of the diesel particulate filter in the composite regeneration
putational modelling and optimisation targeted towards
process. Appl Energy 2016;181:14–28. http://dx.doi.org/10.1016/j.apenergy. developing carbon abatement and emissions reduction tech-
2016.08.051. nologies for the automotive, power and chemical industries.
[270] Wang Y, Kamp C. The effects of mid-channel ash plug on DPF pressure Together with the CoMo group, he has contributed signif-
drop. In: SAE 2016 world congress and exhibition. SAE International; 2016, icantly towards the detailed modelling of the growth and
http://dx.doi.org/10.4271/2016-01-0966. morphology of organic and inorganic nanoparticles.
[271] Lambert CK, Chanko T, Jagner M, Hangas J, Liu X, Pakko J, et al. Analysis
of ash in low mileage, rapid aged, and high mileage gasoline exhaust particle
filters. SAE Int J Engines 2017;10(4). http://dx.doi.org/10.4271/2017-01-0930.
[272] Porter Q, Strzelec A. Benchtop investigation of filtration efficiency and pressure
drop behavior of commercial high porosity gasoline particulate filters. In: SAE Prof Markus Kraft is a Fellow of Churchill College
technical paper series. 1, 2019, p. 1–11. http://dx.doi.org/10.4271/2019-01- Cambridge and Professor in the Department of Chemical
0054. Engineering and Biotechnology. He is the director of CARES
[273] Fang J, Meng Z, Li J, Pu Y, Du Y, Li J, et al. The influence of ash on soot ltd., the Singapore-Cambridge CREATE Research Centre. He
deposition and regeneration processes in diesel particular filter. Appl Therm Eng is also a principal investigator of ‘‘Cambridge Centre for
2017;124:633–40. http://dx.doi.org/10.1016/j.applthermaleng.2017.06.076. Carbon Reduction in Chemical Technology (C4T)’’. He has a
[274] Kamp CJ, Sappok A, Wong V. Soot and ash deposition characteristics at the strong interest in the area of computational modelling and
catalyst-substrate interface and intra-layer interactions in aged diesel particulate optimisation targeted towards developing carbon abatement
filters illustrated using focused ion beam (FIB) milling. SAE Int J Fuels Lubr and emissions reduction technologies for the automotive,
2012;5:696–710. http://dx.doi.org/10.4271/2012-01-0836. power and chemical industries. He has contributed sig-
[275] Seong H, Choi S, Lee S, Zaluzec NJ, Toops TJ, Lance MJ, et al. Deactivation nificantly towards the detailed modelling of combustion
of three-way catalysts coated within gasoline particulate filters by engine-oil- synthesis of organic and inorganic nanoparticles and worked
derived chemicals. Ind Eng Chem Res 2019;58:10724–36. http://dx.doi.org/10. on engine simulation, spray drying and the granulation of
1021/acs.iecr.9b00342. fine powders. More recently, he has been working on cyber
[276] Bardasz EA, Cowling S, Panesar A, Durham J, Tadrous TN. Effects of lubricant physical systems employing time varying knowledge graphs
derived chemistries on performance of the catalyzed diesel particulate filters. In: with the aim to build large cross domain applications that
SAE technical paper. SAE International; 2005, http://dx.doi.org/10.4271/2005- help to reduce energy consumption and harmful emissions.
01-2168.

28