Professional Documents
Culture Documents
A R T I C L E I N F O A B S T R A C T
Keywords: Aerosol-driven solvent losses have been identified as one of the major challenges of amine-based post-com
Process simulation bustion CO2 capture. In this work, a multi-component aerosol dynamics model based on the discrete-sectional
Post combustion CO2 capture approach, accounting for condensation and coagulation was coupled with a process simulation model devel
Aerosol dynamics
oped using ASPEN PLUS v10® to account for the multi-component mass and heat transfer, hydrodynamics,
Amine losses
thermodynamics, and electrolytic-solution chemistry. Particle losses inside the absorber were incorporated into
the model based on a cut-off size determined from experiments reported in the literature. Based on the results, it
was shown that neglecting particle losses inside the absorber resulted in a significant over-prediction of amine-
based solvent losses, while coagulation of particles resulted in reduced (~10%) amine emissions. Furthermore,
amine emissions increased when the number concentration and the geometric standard deviation of inlet par
ticles in the flue gas were increased. Moreover, it was shown that amine emissions were lower at lower solvent
concentrations and temperatures. CO2 capture efficiency dropped for decreasing solvent concentration, but
remained unchanged for lower temperatures, suggesting that decreasing the solvent temperature is an efficient
strategy to reduce amine emissions.
* Corresponding author.
E-mail address: pbiswas@wustl.edu (P. Biswas).
https://doi.org/10.1016/j.ijggc.2020.103179
Received 24 November 2019; Received in revised form 9 August 2020; Accepted 27 September 2020
Available online 23 October 2020
1750-5836/© 2020 Elsevier Ltd. All rights reserved.
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
2015; Mertens et al., 2014a; Mertens et al., 2017; Mertens et al., 2012; with an aerosol dynamics model. However, only Majeed et al., 2018
Moser et al., 2009; Moser et al., 2017) have performed both bench- and (Majeed et al., 2018) consider the particle size distribution in their
pilot-scale experiments that illuminate the phenomena involved. A good model, while the rest model the growth of single particle/droplet inside
summary on these experiments was provided by Harsha et al., 2019 the absorber solely by condensation. However, Majeed et al., 2018
(Harsha et al., 2019). The experiments of Khakharia et al., 2013 (Kha (Majeed et al., 2018) use a Moment-based model to track the average
kharia et al., 2013) clearly distinguish vapor-based and aerosol-based particle size and standard deviation, while setting the total number
emissions. Two important conclusions from their work were that concentration as a constant, because the model considers only conden
introducing particles (H2SO4 droplets) to the flue gas increases amine sation. No models account for coagulation or particle loss, and most of
emissions from 45 mg/Nm3 (vapor-based) to 1100 mg/Nm3 (aero these models are not validated against experimental data.
sol-based + vapor-based) and that amine emissions increase with an Therefore, in this work a coupled model is developed that integrates
increase in the number concentration of the particles. It can be inferred a multicomponent aerosol dynamics approach that accounts for the time
from these results that particles act as condensation nuclei for amine evolution of the particle size distribution with a process simulation
condensation, and as amine in the vapor phase is depleted, more amine model. The model predictions are validated with published experi
from the liquid phase vaporizes, leading to increased emissions. How mental data (Khakharia et al., 2013). The multi-component mass and
ever, it is critical to note that the vapor phase is not saturated with heat transfer between bulk gas and bulk liquid was modelled using
amine. As water vapor in the column is near saturation, water will be ASPEN PLUS v10®, and the multi-component aerosol dynamics was
also condense (Fulk and Rochelle, 2013). Hence, the concentration of modelled using the discrete-sectional approach. An empirical approach
amine on the particle’s surface is very small relative to its gas phase to account for the particle losses inside the absorber, developed based on
concentration, driving amine condensation. Furthermore, some experi the experimental data of Khakharia et al., 2013 (Khakharia et al., 2013),
ments conducted in the past (Harsha et al., 2019; Brachert et al., 2013) is presented. Furthermore, the influence of considering the vapor con
have shown reasonable evidence of aerosol coagulation inside the centrations in each stage of the absorber as a constant vs. depleting, was
absorber. Harsha et al., 2019 (Harsha et al., 2019) showed that in the investigated. The integrated model was used to investigate the effects of
case of H2SO4 aerosols, the number concentration reduced from 6.24 × particle number concentration and initial particle size distribution on
107 #/cm3 at the inlet of the absorber to 2.30 × 107 #/cm3 at the outlet. amine emissions. Particle number concentrations between 1.0 × 1014 -
They attribute this 63 % decrease in number concentration to coagula 1.8 × 1014 #/m3 (corresponding to the experiments of Khakharia et al.,
tion and depositional losses. 2013 (Khakharia et al., 2013)) were investigated and the predicted
Several researchers (Majeed et al., 2017a; Fulk and Rochelle, 2013; amine emissions were compared with the experimental data. Three
Majeed et al., 2018; Majeed et al., 2017b; Majeed and Svendsen, 2018; distributions: monodisperse, log-normal, and bi-lognormal, were
Majeed et al., 2017c; Zhang et al., 2017) have developed models to considered. Further, the integrated model was used to study the influ
investigate this system by coupling a process simulation package (to ence of solvent (30 wt.% Mono Ethanol Amine (MEA)) concentration,
model the multi-component reactive mass transfer inside the absorber) and solvent temperature on vapor- and aerosol-based amine emissions,
2
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
and CO2 capture efficiency. Solvent concentrations between 0.06 – model, 16 stages were used in APP to describe the column of Khakharia
0.125 (MEA mole fraction) and temperatures between 34 – 44 ◦ C, were et al., 2013 (Khakharia et al., 2013). Also, 20 discretization points were
considered in the study. used to model the liquid film. In the DISC model, the vapor phases
(amine and water) were considered as discrete sizes, while the particles
2. Model description - The framework of DISC_APP were treated as sections with a maximum of 400 sections. The DISC
simulations were performed in the Engineering Cloud Cluster of Wash
Amine-based CO2 capture is a thermal-swing reactive absorption ington University in St. Louis.
process. A schematic of the various physicochemical processes occurring
inside the absorber is shown in Fig. 1. The present model accounts for 2.1. ASPEN PLUS v10® model (APP)
these processes. The flue gas containing particles contacts the liquid
aqueous amine counter-currently, while the aerosols flowing with the The multicomponent mass and heat transfer between gas and liquid
gas grow by condensation and coagulation, and exit the column. The phases, the hydrodynamics, the thermodynamics, and the solution
reactive mass transfer process is kinetically limited (Schneider et al., chemistry were all modelled using APP. RadFrac was used to model the
1999) and hence, a rate-based model is required. Also, since the disso reactive mass transfer between the two phases, and the ELECNRTL
lution of CO2 in aqueous amine generates electrolytes, the non-ideality property method was used. In the rate-based model, separate mass and
due to ionic interactions should be accounted for. Since RadFrac, a heat balance equations (Eqn. 1 – 4) are written for each phase, while the
rate-based-model in ASPEN PLUS v10® (APP), has been extensively mass and heat transfer resistance are estimated by the two-film theory
used (Fulk and Rochelle, 2013; Zhang et al., 2017; Zhang and Chen, (Taylor and Krishna, 1993) through implicit calculations of interfacial
2013) and shown to predict this system reasonably well, it was used to fluxes (Eqns. 5 – 8). The bulk phase equations consider the axial change
simulate the multi-component mass and heat transfer between gas and of the total molar stream, composition, and enthalpy, and are written as
liquid phases.
The time-evolution of aerosols can be predicted by solving the gen dLxlb
i
= nlb i lb i l
(1)
i a AC + Ri ϕ AC , i = 1, ..., n
eral dynamics equation (GDE) for aerosol growth (Friedlander, 1977). dz
Although there are multiple modeling approaches to solve the GDE, the
discrete-sectional (DISC) approach was used to solve the partial dGygb
i
= ngb i
i a AC , i = 1, ..., n
g
(2)
integro-differential equations, given that it is the most versatile (in terms dz
of computational cost and accuracy) aerosol dynamics model (Wu and
dLhlb
Biswas, 1998; Wu and Flagan, 1988). Because the concentration and = qlfi ai AC (3)
dz
temperature of the different species in the vapor phase govern the
growth of aerosols, the predictions from APP were used as inputs to the dGhgb
DISC. The modelling framework of the integrated model (DISC_APP) is = qgf i
i a AC (4)
dz
shown in Fig. 2. The experimental system of Khakharia et al., 2013
(Khakharia et al., 2013) was used as the base case for simulation. In the In the film regions, differential mass and heat balances in the film
coordinate (ηi ) were performed, wherein the balance equations and
interfacial fluxes are written as
1 dnlfi
= Rlfi , i = 1, ..., nl (5)
δl d ηl
1 dngfi
= 0, i = 1, ..., nl (6)
δg d ηg
( )
clft Dleff ,i dxlfi F dφ
nlfi =− lf
+ xi zi + xlfi nlft , i = 1, ..., nl (7)
δl dηl RG T d η l
( )
cgf g
t Deff ,i dygf
ngf
i =− g
i
+ ygf gf
i nt , i = 1, ..., n
g
(8)
δ d ηg
lf
( ) nl
∑
λlf dTil
lf
q = − lf + n hlfi (9)
δ d ηl i=1
i
gf
( ) ng
∑
λgf dT g
gf
q = − gf + n hgf
i (10)
δ d ηg i=1
i
The flux due to ionic chemical species is accounted using the Nernst-
Plank expression (in Eqn. 7). Further, the thermodynamic equilibrium at
the gas-liquid interface, described by the relation below, relates the gas
and liquid phase compositions at the interface (Schneider et al., 1999
(Schneider et al., 1999)):
3
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
Fig. 2. The modelling framework developed in this work, showing the multi-component mass and heat transfer between bulk gas and liquid phases (modelled using
ASPEN PLUS v10®) and multi-component aerosol dynamics (modelled using the discrete-sectional approach).
The calculations for the non-idealities are based on the three-parametric 2.1.1. The discrete sectional model (DISC)
Electrolyte-NRTL method. The thicknesses of the films were estimated The general dynamics equation for aerosol growth due to conden
using the empirical correlations suggested by Taylor and Krishna, 1993 sation and coagulation can be written as (Friedlander, 1977)
(Taylor and Krishna, 1993). The holdup and pressure drop were esti ∫
∂n(v, t) ∂(Gn) 1 v
mated using the Billet and Schulte’s (Billet and Schultes, 1999) corre + = β(v − ṽ, v)n(v
∂t ∂v 2 0
lation. The interfacial area and mass transfer coefficients were estimated ∫∞
using the correlation suggested by Hanley et al., 2011 (Brian Hanley, − ṽ, t)n(ṽ, t)dṽn(v, t) β(ṽ, v)n(ṽ, t)dṽ (19)
2011), and the heat transfer coefficient was estimated using the Chilton 0
RNH2 OH + CO2 + H2 O = RNHOHCOO− + H3 O+ (17) where N (vk ) denotes the number concentration of section k. In this
work, the vapor species (amine and water) were modelled as discrete
CO2 + OH − = HCO−3 (18) sizes, the particles were considered as sections and number concentra
The activity-based kinetic expressions suggested by Zhang et al., tion was chosen as the conserved property (ζ = 0). Time step sizes were
2013 (Zhang and Chen, 2013) were used to describe the rate of the ki chosen as a function of time, wherein the size was finer when the driving
netic reactions. force for condensation was higher and vice-versa. The time-step sizes
ranged between 10-8 - 10-6. Some numerical diffusion was observed
4
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
when the particle sizes reached around 10-6 m, and hence the choice of
QG
the finest time-step size was made such that the results were not influ QG,eff = (30)
(1 − lh)∗vfrac
enced by a further decrease in time-step size.
The balance equations are written as: It should be noted that the mass transfer of CO2 in the particles
dqvapor,m (amine droplets) has been neglected. As the droplets are rather small
=0 (22) (micrometer sized), this assumption is fairly accurate and will not
dt
impact the results. Moreover, for the system considered, the mole frac
dqvapor,m ∑
max ∑
max
4 D ( ) tion of amine in the aerosol phase is in the order of 1 × 10-4, due to the
=− β1,k,m q1,m − Qk,sat,m Qk (23) significantly higher amounts of water in the aerosol phase. The gradients
dt
in concentration are rather small for the small droplets and can be
t=1 k=1
∑
MAX ∑
mmax assumed to be well mixed for the dilute concentrations. For conditions in
dQ1 13 16 2 D ( )
which the amine concentration is reasonably high (5 M, not this work),
4
= − β1 Q1 2 + β1 Q1 2 − Q1 β1,i Qi − β1,i,m q1,m − Qi,sat,m Qi
dt 2 2 i=2 m=1 the estimation and impact of radial concentration profiles of carbamate
∑ and amine are discussed in detail by Majeed et al., (Majeed et al., 2017b;
mmax
5 D ( )
+ β1,i,m q1,m − Qi,sat,m Qi
m=1 Majeed et al., 2017c).
(24)
3. Results and Discussion
dQk 1∑k− 1 ∑
k− 1
1 D ∑
k− 1
2 ∑
k− 1
5
dt
=+
2 i=1 j=1
βi,j,k Qi Qj + − Qk βi,k Qi + + Qk βi,k Qi 3.1. Simulation plan
i=1 i=1
5
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
3.2. Model development and validation the temperature profile in the upper-middle section of the absorber
satisfactorily well, wherein the simulation under-predicts the tempera
The experimental data of Khakharia et al., 2013 (Khakharia et al., ture near the second stage of the absorber. Note that Khakharia et al.,
2013) were chosen for validation, because it can be used to separately 2013 (Khakharia et al., 2013) describe these measurements as typical
validate the APP and APP_DISC results. The parameters and operating temperature profiles and do not explicitly mention how and for which
conditions relevant to this work are mentioned here, however a detailed case the profile was obtained. However, the temperature profiles
description can be found in (Khakharia et al., 2013). They used an ab measured and predicted by Zhang et al., 2013 (Zhang and Chen, 2013)
sorption column that was 3.5 m high and consisted of four packed beds are very similar to the APP model predictions (see Fig. 6 (a) of Zhang
(Sulzer Mellapak 2X, height = 0.5 m, diameter =4.5 cm). The column et al., 2013). The gas phase concentration of MEA and CO2 predicted by
did not include a water wash or demister, and the solvent was 30 wt.% the model agrees well with the profiles predicted by Majeed et al., 2017
aqueous MEA at 40 ◦ C at a typical lean loading of 0.25 mol CO2/mol (Majeed et al., 2017a). They have also used ASPEN to calculate the gas
amine. The comparison of model predictions with the corresponding phase compositions. While CO2 concentration showed a steady decrease
experiments are discussed in the following sections. with respect to increasing absorber height, MEA concentration increased
along the height of the column with the maximum increase near the top
3.2.1. Vapor-based emissions of the column. These trends can be observed in the work of Majeed et al.,
In the base line study, Khakharia et al. 2013 (Khakharia et al., 2013) 2017 (Majeed et al., 2017a). Further, note that a counter-current flow
measured vapor-based amine emissions using a FTIR analyzer (GASMET model was used in APP to describe the vapor and liquid phase flow in the
CX 4000) after the flue gas exiting the absorber was heated to 180 ◦ C. APP model because it predicted the extent of back-mixing in the column
The amine concentration was measured to be 45 mg/Nm3, with a re quite well. Also, the choice of the empirical correlations for interfacial
ported error of ±10%. In this case, the simulated flue gas did not include area, mass transfer coefficients (Hanley et al., 2011; (Chilton and Col
particles, and it consisted of CO2 (12.8 vol. %), O2 (14.6 vol.%) and CO burn, 1934)) and heat transfer coefficient (Chilton and Colburn, 1934)
(43.5 mg/Nm3). They reported a typical temperature profile inside the was made such the model represents the experimental system reason
bed which shows a temperature bulge at the upper-middle section of the ably well.
absorber, and reported the concentration of CO2 in the exiting flue gas to
be 2.5 vol.%. The APP system in our work was setup based on these 3.2.2. Accounting for vapor depletion
conditions (considering all the conditions mentioned in their work), and The simulated concentration of the vapors and the gas-phase tem
a comparison of simulated vapor-phase MEA, CO2, and temperature perature profiles by the APP model for the base-line case of Khakharia
profiles with the experimental data is shown in Fig. 3. While the APP et al., 2013 (Khakharia et al., 2013) is used as input to the DISC model.
simulations predict the measured temperature, concentration of MEA The amine vapor phase concentration predicted by the APP model is
and CO2 at the outlet of the absorber very well, the simulations capture shown in Fig. 4 (a). The mass transfer of vapors from the gas phase to the
Fig. 3. Comparison of ASPEN PLUS v10® simulation results using the experimental data of Khakharia et al., 2013 (Khakharia et al., 2013) showing (a) vapor phase
temperature profile (b) MEA vapor concentration, and (c) gas phase CO2 volume fraction inside the absorber.
6
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
Fig. 4. (a) The MEA vapor profile using APP model without including vapor losses, (b) the depleted MEA vapors calculated using a DISC model (with depleting
vapors), and (c) the MEA vapor profile using APP model while considering vapor loss compared with case (a).
particles surface due to condensation can lead to depletion of vapors DISC simulations were performed corresponding to the base-line case
inside the absorber when the rate of depletion of vapors due to discussed previously. The size distribution of the particles was chosen
condensation is significantly greater than the rate of vaporization from such that they are relevant to the experiments performed by Khakharia
the bulk liquid. However, if the rate of condensation is comparable or et al., 2013 (Khakharia et al., 2013) wherein they introduced H2SO4 acid
smaller than that of vaporization, then the vapor profiles inside the droplets into the flue gas. The number concentration was 1.016 × 108
absorber will remain constant during the process. In order to verify this, #/cm3, monodisperse at 100 nm. The DISC simulations were performed
7
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
for depleting vapors (Eqn. 23) and mass of vapors depleted in each stage in their flue gas and similarly measuring the heated exiting flue gas. In
per unit volume of gas was computed, and shown in Fig. 4 (b). It can seen this case, the amine emissions increased to 600 mg/Nm3 when the
that most of the amine vapor depletion is observed at the top of the number concentration of the particles was 1.016 × 1014 #/m3, and
column. These results agree with the model predictions of Majeed et al., further increased to 853 mg/Nm3 and 1163 mg/Nm3 when the number
2017 (Majeed et al., 2017b) wherein they have simulated the changes in concentration was increased to 1.3 × 1014 #/m3 and 1.73 × 1014 #/m3,
amine partial pressure due to the condensation of amine onto droplets respectively. They measured the particle number concentration using a
that are initially 0.15 μm in size (Case 5) and at different number con condensation particle counter (CPC; PALAS UFCPC with Sensor 200) at
centrations. They have shown that at NTOT = 107 #/cm3, amine deple the inlet to the absorber. Note that these amine emission values corre
tion is observable near the top of the column. Further, APP model was spond to total amine emissions (vapor + aerosol). However, the
modified to account for the depleting vapors at each stage (see Fig. 4 (c)) aerosol-based emissions can be calculated by subtracting the known
and the steady-state vapor and temperature profiles were computed. It vapor-based emissions (45 mg/Nm3). The temperature, vapor phase
can be seen in Fig. 4 (c) that the vapor profiles before and after ac amine, and water concentration profiles calculated using the APP model
counting for vapor depletion are similar. This result is because the sur were input to the DISC model. An initial particle size of 100 nm was used
face area available for condensation (0.0406 m2) is two orders of because it was reported by Khakharia et al., 2013 (Khakharia et al.,
magnitude lower than that available for gas-liquid contact (1.87 m2). 2013) (based on their simulations). The inlet particle size distribution
Therefore, for further cases, the DISC simulations were performed for a was assumed to be monodisperse, as there was no mention of the dis
constant vapor system (Eqn. 22). tribution. However, in our work the effect of initial particle size distri
bution was investigated separately (Section 3.2). We simulated the time
3.2.3. Particle losses and aerosol-based emissions evolution of particle size distribution using DISC for the conditions of
Khakharia et al., 2013 (Khakharia et al., 2013) measured the Khakharia et al., 2013 (Khakharia et al., 2013) wherein they studied
aerosol-based emissions by introducing particles (H2SO4 acid droplets) the effect of H2SO4 droplets’ number concentration on amine emissions.
Fig. 5. (a) Simulated amine mass in sections and the estimated cut-off size to account for particle losses based on the experimental data of Khakharia et al., 2013
(Khakharia et al., 2013), (b) estimated cut-off sizes at different initial number concentrations, and (c) the comparison of predicted amine emissions with experimental
measurements.
8
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
sponding to each number concentration was used to determine a cut-off 1 1.02 x 1014 601.58 556.58 29.7434
size such that the simulated amine emissions predict the experimentally 2 1.11 x 1014 603.17 558.17 29.7442
3 1.17 x 1014 831.13 786.13 29.7442
measured values. The cut-off size is defined in this work as the size (see
4 1.21 x 1014 846.97 801.97 29.7442
Fig. 5 (a)) at which the cumulative mass concentration of amine 5 1.31 x 1014 853.30 808.30 29.7434
condensed on these particles equals the measured aerosol-based amine 6 1.43 x 1014 1105.01 1060.01 29.7442
mass concentration (emission). The measured amine mass concentration 7 1.77 x 1014 1163.59 1118.59 29.7442
for different H2SO4 droplet number concentration was fitted such that
few representative data points (number concentration and amine
simulations, the aforementioned value was used as the cut-off size. It
emission) can be chosen for determining the cut-off size. The specific
should be noted for the case in which the solvent concentration was
reason for fitting the data was because at the same number concentra
decreased, the cut-off size approach did not work as the mean size of the
tion, Khakharia et al., 2013 (Khakharia et al., 2013) have reported
particles at the final time was smaller than the cut-off size. Hence for this
different amine emissions (see Fig. 5 (b)). The chosen number concen
case, the percent loss in the particles estimated from the base case at the
tration data points, measured amine emissions, fitted amine emissions
same number concentration was used to determine the cut-off size.
and estimated cut-off sizes are shown in Table 2. It can be seen in Fig. 5
Further simulations were performed to predict the effect of the number
(b) that the cut-off size for the different number concentrations was very
concentration of particles on amine emissions, and Fig. 5 (c) shows that
similar and it was found to be 29.74 μm (average cut-off size), based on
the simulations predict the amine emissions reasonably well. The
the simulations performed for the seven cases (see Fig. 5 (b)). The
number concentrations corresponding to the experimental measure
physical significance of this value is that it can be considered that the
ments of Khakharia et al. 2013 (Khakharia et al., 2013) were used, while
particles greater than this cut-off size will be lost inside the absorber and
the distribution was considered to be monodisperse at 100 nm. The
will not contribute to amine-based emissions. Therefore, for further
simulations predict the amine emissions reasonably well. It can be
Fig. 6. (a) Lognormal, bi-lognormal (b) experimentally measured (Mertens et al., 2014) size distributions considered to investigate the effect of the distribution on
amine emission. Effect of coagulation of particles on (c) amine emission and (c) water emission.
9
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
inferred from these results that aerosol-based amine emissions scale correspondingly higher.
proportionally with particle number concentration, due to the increase In order to understand the influence of coagulation on amine emis
in surface area available for condensation. This increase in amine sions, simulations were performed with and without coagulation terms
emission due to the increase in total number concentration agrees well in the balance equations. In both cases, condensation was considered.
with the model predictions of Majeed et al., 2017 (Majeed et al., 2017b). These simulations were performed for two different initial number
They have shown that as NTOT increases, the depletion of amine in the concentrations (lowest and highest in the range of interest). For these
gas phase increases due to condensation to the droplets. It can therefore simulations, the size distribution was considered to be monodisperse,
be inferred that at higher number concentrations, the surface area with particle size of 100 nm. Fig. 6 (b), shows that the model considering
available for condensation is higher, leading to increased amine deple coagulation and condensation predicted a lower amine emission
tion which results in higher amine emissions. (~10%) than a pure condensation model. This difference was observed
at all number concentrations considered, and it increased with an in
crease in number concentration. A very similar trend was also observed
3.3. Effect of particle size distribution and aerosol phenomena on amine
in the predicted water emissions (see Fig. 6 (c)). These results can be
emissions
explained by the data provided in Table 4, wherein the final number
concentration is smaller for the model that considers coagulation than
We investigated the influence of initial particle size distribution on
for the pure condensation model. As mentioned earlier, amine emissions
amine emissions, considering monodisperse, log-normal, bi-lognormal
are directly proportional to the surface area available for condensation.
distributions and experimentally measured size distributions reported
Coagulation decreases surface area hence there is a decrease in amine
by Mertens et al., 2104 (Mertens et al., 2014b) (for the system of Kha
emissions. Note that particle losses are also considered in these cases and
kharia et al., 2013 (Khakharia et al., 2013)). We have scaled down the
hence there is a decrease in number concentration in the pure conden
measured size distribution such that the total number concentrations
sation case also. Therefore, coagulation of particles in flue gas decreases
match the measurements of Khakharia et al, 2103 (Khakharia et al.,
amine emissions and hence it is important to include coagulation in the
2013). Also, Mertens et al., 2014 (Mertens et al., 2014b) have reported
model to accurately predict solvent emissions.
that their measurements possibly include artifacts due to dilution, in
clusion of a fan to assist ELPI sampling, and inclusion of a filter stage.
3.4. Effect of solvent concentration on vapor- and aerosol-based amine
Dilution and fan inclusion would shrink the particles and the inclusion of
emissions
the filter stage would over-estimate number concentration of particles in
the size < 10 nm. Table 3 shows the average parameters describing the
The DISC_APP model was further used to investigate the effect of
size distribution at the initial time, along with the amine emissions. For
solvent concentration on amine emissions. The mole fraction of MEA in
the bi-lognormal case, two log-normal distributions were added. For
the solvent was variously 0.06, 0.1 and 0.125. Fig. 7 (a) shows that the
each of the distributions, the number concentration was set equal to half
temperature profiles exhibit a bulge in each of the cases, while the
the NTOT of the other cases, while σg was set to 1.2. The dpg values were
temperature values showed an increase with increased MEA concen
set to 50 nm and 200 nm so that the dpg of the bi-lognormal distribution
tration with significant changes towards the bottom of the column. In
equaled 100 nm. The log-normal, bi-lognormal and the scaled-down
creases in the mole fraction of MEA resulted in higher absorption of CO2
version of experimentally measured distributions considered in the
(see Fig. 7 (d)), and hence more CO2 reacted with MEA in the liquid
simulation are shown in Fig. 6 (a). The results for these sets of simula
phase. Because the reactions between MEA and CO2 are exothermic,
tions are shown in Table 3. Between the monodisperse, log-normal and
more heat was released, increasing the temperature inside the column,
bi-lognormal cases amine emissions increased with σg in all the cases
which in turn increased the vapor phase concentrations of MEA and
and the maximum increase occurred in the bi-lognormal case wherein
water (Fig. 7 (b) and (c). The increased MEA concentration in the vapor
the initial σg was the largest. The amine emissions are proportional to
phase (vapor-based emissions), caused the increased amine emissions
the initial surface area concentration (STOT), and hence the bi-lognormal
(aerosol-based) shown in Fig. 7 (e). This critical result shows that
case shows the highest emissions, while the monodisperse case shows
aerosol-based emissions increase when vapor-based emissions increase,
the lowest emissions. The calculated STOT is shown in Table 3. From
and is a consequence of increased MEA concentration.
these results, it can be inferred that amine emissions increase with
Vapor-based water emissions show a bulge very similar to the tem
increasing standard deviation of the initial size distribution of the par
perature profile. However, as the MEA concentration was increased, the
ticles in the flue gas for a given initial total number concentration and
concentration of water in the solvent decreased (so that the mole frac
geometric mean diameter. However, the amine emissions for experi
tions add up to unity). However, vapor-based emissions of water are
mental case, showed significantly higher emissions which is because the
more sensitive to temperature profile than to concentration, and hence
size distribution includes extremely high concentration of particles in
an increase in vapor-based emission was seen for a decrease in water
the size range 10 – 40 nm. Since a cut-off size of 29.74 μm was used, the
concentration in the solvent. This is because the relative change in water
smaller particles would take-up a significantly higher volume of amine
concentration is lesser than the relative increase in temperature values.
to reach that size and hence the estimated amine emissions are
Consequently, the aerosol-based water emissions increase even as the as
water concentration in the solvent is decreased.
Table 3
Effect of initial particle size distribution on amine emission.
Initial Final
10
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
Fig. 7. Effect of solvent concentration on (a) vapor phase temperature profile, (b) MEA vapor concentration, (c) water vapor concentration, (d) gas-phase CO2 mole
fraction and (e) MEA and water emissions.
3.5. Effect of solvent temperature on vapor- and aerosol-based amine in temperature magnitudes when the solvent temperatures was
emissions increased, which is a direct consequence of the adiabatic absorber. The
increase in the temperature profile inside the absorber results in
The influence of the solvent’s inlet temperature was varied to increased vapor- and aerosol-based amine and water emissions and are
investigate its influence on vapor- and aerosol- based amine and water shown in Fig. 8 (b), (c) and (e), while the uptake efficiency of CO2 did
emissions. The temperature was varied as 34, 39 and 44 ◦ C. Fig. 8 (a) not change significantly (Fig. 8 (d)). It should be noted that the increase
shows that the temperature profiles inside the absorber show an increase in aerosol-based amine emission due to higher inlet solvent temperature
11
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
Fig. 8. Effect of solvent temperature on (a) vapor phase temperature profile, (b) MEA vapor concentration, (c) water vapor concentration, (d) gas-phase CO2 mole
fraction and (e) MEA and water emissions.
is in agreement with the model predictions of Zhang et al., 2017 (Zhang aerosol-based amine emissions.
et al., 2017). Given that the reactive mass transfer process is kinetically
limited, it can be inferred that the reaction rates of MEA and CO2 are 4. Conclusions
stronger functions of MEA concentration (based on the conclusions of
previous sections) than the gas-phase temperature (in the temperature An ASPEN PLUS v10® model accounting for multi-component mass
range considered). Therefore, decreasing the MEA’s temperature does and heat transfer, column hydrodynamics, solution thermodynamics,
not affect CO2 capture efficiency, while it can potentially reduce and electrolytic-solution chemistry was coupled with a multi-
12
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
component discrete-sectional-based aerosol dynamics model. The Sreedhar, I., Nahar, T., Venugopal, A., Srinivas, B., 2017. Carbon capture by absorption –
Path covered and ahead. Renewable and Sustainable Energy Reviews 76,
aerosol dynamics model was developed in FORTRAN 90, accounting for
1080–1107.
condensation, and coagulation, to predict amine emissions during post- J. N. Pitts, D. Grosjean, K. Van Cauwenberghe, J. P. Schmid, and D. R. Fitz,
combustion CO2 capture. Particle losses were considered by fitting the "Photooxidation of aliphatic amines under simulated atmospheric conditions:
experimental data Khakharia et al., 2013 (Khakharia et al., 2013). The formation of nitrosamines, nitramines, amides, and photochemical oxidant,"
Environmental Science & Technology, vol. 12, pp. 946-953, 1978/08/01 1978.
model’s outputs were validated using their experimentally measured Khakharia, P., Brachert, L., Mertens, J., Huizinga, A., Schallert, B., Schaber, K., et al.,
vapor- and aerosol-based amine emission data. The simulations showed 2013. Investigation of aerosol based emission of MEA due to sulphuric acid aerosol
that neglecting particle losses resulted in an order of magnitude and soot in a Post Combustion CO2 Capture process. International Journal of
Greenhouse Gas Control 19, 138–144.
over-prediction of amine-based solvent losses. The effect of including Harsha, S., Khakharia, P., Huizinga, A., Monteiro, J., Goetheer, E., Vlugt, T.J.H., 2019.
coagulation in the model was also investigated and it was found that In-situ experimental investigation on the growth of aerosols along the absorption
including coagulation in the model resulted in reduced (~10%) emis column in post combustion carbon capture. International Journal of Greenhouse Gas
Control 85, 86–99.
sions. Furthermore, it was shown that amine emissions increased with Mimura, T., Nojo, T., Ishida, K., Nakashoji, H., Tanaka, H., Hirata, T., 2004. Amine
an increasing number concentration of inlet particles. Different particle recovery method and apparatus and decarbonation apparatus having same. ed:
size distributions of inlet particles such as monodisperse, lognormal and Google Patents.
Khakharia, P., Brachert, L., Mertens, J., Anderlohr, C., Huizinga, A., Fernandez, E.S.,
bi-lognormal were considered, and it was shown that distributions with et al., 2015. Understanding aerosol based emissions in a Post Combustion CO2
higher geometric standard deviations at the same number concentration Capture process: Parameter testing and mechanisms. International Journal of
resulted in higher emissions. Furthermore, amine emissions were Greenhouse Gas Control 34, 63–74.
Mertens, J., Brachert, L., Desagher, D., Schallert, B., Khakharia, P., Goetheer, E., 2014a.
reduced when the solvent’s concentrations and temperatures were
Predicting Amine Mist Formation Based on Aerosol Number Concentration and Size
decreased. However, CO2 capture efficiency dropped for decreasing Measurements in Flue Gas. Energy Procedia 63, 893–901.
solvent concentration but remained unchanged at lower temperatures. It Mertens, J., Khakharia, P., Rogiers, P., Blondeau, J., Lepaumier, H., Goetheer, E., et al.,
can be concluded that aerosol-based emissions scale proportionally to 2017. Prevention of Mist Formation in Amine Based Carbon Capture: Field Testing
Using a Wet ElectroStatic Precipitator (WESP) and a Gas-Gas Heater (GGH). Energy
vapor-based emissions, which increases with increasing temperatures in Procedia 114, 987–999.
the absorber. Given that decreasing solvent temperatures reduce amine Mertens, J., Knudsen, J., Thielens, M.-L., Andersen, J., 2012. On-line monitoring and
emissions without affecting CO2 capture efficiency, this efficient controlling emissions in amine post combustion carbon capture: A field test.
International Journal of Greenhouse Gas Control 6, 2–11.
approach can reduce the absorber’s temperature and thereby reduce Moser, P., Schmidt, S., Sieder, G., Garcia, H., Ciattaglia, I., Klein, H., 2009. Enabling post
amine emissions during post-combustion CO2 capture. combustion capture optimization–The pilot plant project at Niederaussem. Energy
Procedia 1, 807–814.
Moser, P., Wiechers, G., Stahl, K., Stoffregen, T., Vorberg, G., Lozano, G.A., 2017. Solid
CRediT authorship contribution statement Particles as Nuclei for Aerosol Formation and Cause of Emissions – Results from the
Post-combustion Capture Pilot Plant at Niederaussem. Energy Procedia 114,
David Dhanraj: Conceptualization, Methodology, Software, Data 1000–1016.
Fulk, S.M., Rochelle, G.T., 2013. Modeling Aerosols in Amine-based CO2 Capture. Energy
curation, Writing - original draft, Visualization, Investigation, Valida Procedia 37, 1706–1719.
tion, Writing - review & editing. Pratim Biswas: Project administration, Brachert, L., Kochenburger, T., Schaber, K., 2013. Facing the Sulfuric Acid Aerosol
Resources, Funding acquisition, Conceptualization, Methodology, Su Problem in Flue Gas Cleaning: Pilot Plant Experiments and Simulation. Aerosol
Science and Technology 47, 1083–1091.
pervision, Writing - review & editing. Majeed, H., Hillestad, M., Knuutila, H., Svendsen, H.F., 2018. Predicting aerosol size
distribution development in absorption columns. Chemical Engineering Science 192,
25–33.
Declaration of Competing Interest Majeed, H., Knuutila, H., Hillestad, M., Svendsen, H.F., 2017b. Gas phase amine
depletion created by aerosol formation and growth. International Journal of
Greenhouse Gas Control 64, 212–222.
The authors declare that they have no known competing financial
Majeed, H., Svendsen, H.F., 2018. Effect of water wash on mist and aerosol formation in
interests or personal relationships that could have appeared to influence absorption column. Chemical Engineering Journal 333, 636–648.
the work reported in this paper. Majeed, H., Knuutila, H., Hillestad, M., Svendsen, H.F., 2017c. Characterization and
modelling of aerosol droplet in absorption columns. International Journal of
Greenhouse Gas Control 58, 114–126.
Acknowledgements Zhang, Y., Kang, J.-L., Fulk, S., Rochelle, G., 2017. Modeling Amine Aerosol Growth at
Realistic Pilot Plant Conditions. Energy Procedia 114, 1045–1060.
Schneider, R., Kenig, E.Y., Górak, A., 1999. Dynamic Modelling of Reactive Absorption
The authors would like to thank the sponsors: U.S. D.O.E and Linde,
with the Maxwell-Stefan Approach. Chemical Engineering Research and Design 77,
for funding this work through the grant entitled: Flue Gas Aerosol Pre 633–638.
treatment Technologies to Minimize PCC Solvent Losses (001365- Zhang, Y., Chen, C.-C., 2013. Modeling CO2 Absorption and Desorption by Aqueous
Monoethanolamine Solution with Aspen Rate-based Model. Energy Procedia 37,
39847). Further, the discrete-sectional simulations were performed in
1584–1596.
Engineering Cloud Cluster of Washington University in St. Louis. Friedlander, S.K., 1977. Smoke, dust and haze: Fundamentals of aerosol behavior. Wiley-
Interscience, New York, 333 p., 1977.
Wu, C.-Y., Biswas, P., 1998. Study of Numerical Diffusion in a Discrete-Sectional Model
Appendix A. Supplementary data and Its Application to Aerosol Dynamics Simulation. Aerosol Science and
Technology 29, 359–378.
Supplementary material related to this article can be found, in the Wu, J.J., Flagan, R.C., 1988. A discrete-sectional solution to the aerosol dynamic
equation. Journal of Colloid and interface Science 123, 339–352.
online version, at doi:https://doi.org/10.1016/j.ijggc.2020.103179.
Taylor, R., Krishna, R., 1993. Multicomponent mass transfer, vol. 2. John Wiley & Sons.
Billet, R., Schultes, M., 1999. Prediction of Mass Transfer Columns with Dumped and
References Arranged Packings. Chemical Engineering Research and Design 77, 498–504.
Brian Hanley, C.-C.C., 2011. New Mass-Transfer Correlations for Packed Towers. AIChE
58, 132–152.
Wang, M., Lawal, A., Stephenson, P., Sidders, J., Ramshaw, C., 2011. Post-combustion
Chilton, T.H., Colburn, A.P., 1934. Mass transfer (absorption) coefficients prediction
CO2 capture with chemical absorption: A state-of-the-art review. Chemical
from data on heat transfer and fluid friction. Industrial & engineering chemistry 26,
Engineering Research and Design 89, 1609–1624.
1183–1187.
Freund, P., 2003. Making deep reductions in CO2 emissions from coal-fired power plant
Biswas, P., Wu, C.Y., Zachariah, M.R., McMillin, B., 1997. Characterization of iron oxide-
using capture and storage of CO2. Proc. Inst. Mech. Eng. A: J. Power Eng 217, 1–8.
silica nanocomposites in flames: Part II. Comparison of discrete-sectional model
He, J.-K., 2015. Objectives and strategies for energy revolution in the context of tackling
predictions to experimental data. Journal of Materials Research 12, 714–723.
climate change. Advances in Climate Change Research 6, 101–107.
Majeed, H., Knuutila, H.K., Hillestad, M., Svendsen, H.F., 2017a. Characterization and
modelling of aerosol droplet in absorption columns. International Journal of
Greenhouse Gas Control 58, 114–126.
Davison, J., 2007. Performance and costs of power plants with capture and storage of
CO2. Energy 32, 1163–1176.
13
D. Dhanraj and P. Biswas International Journal of Greenhouse Gas Control 103 (2020) 103179
Pratsinis, S.E., 1988. Simultaneous nucleation, condensation, and coagulation in aerosol Mertens, J., Brachert, D., Desagher, M.L., Thielens, P., Khakharia, E., Goetheer,
reactors. Journal of Colloid and Interface Science 124, 416–427. Schaber, K., 2014. ELPI+ measurements of aerosol growth in an amine absorption
Fulk, S.M., 2016. Measuring and Modeling Aerosols in Carbon Dioxide Capture by column. International Journal of Greenhouse Gas Control 23 (2014), 44–50.
Aqueous Amines. PhD Thesis. The University of Texas at Austin.
14