processes

Article
Technoeconomic Evaluation of a Process Capturing
CO2 Directly from Air
Romesh Pramodya Wijesiri 1,2 , Gregory Paul Knowles 2 , Hasina Yeasmin 1 ,
Andrew Forbes Alexander Hoadley 1, * and Alan Loyd Chaffee 2, *
1 Department of Chemical Engineering, 16 Alliance Lane, Clayton Campus, Monash University, Clayton,
VIC 3800, Australia
2 School of Chemistry, 17 Rainforest Walk, Clayton Campus, Monash University, Clayton, VIC 3800, Australia
* Correspondence: Andrew.Hoadley@monash.edu (A.F.A.H.); Alan.Chaffee@monash.edu (A.L.C.);
Tel.: +61-3-9905-3421 (A.F.A.H.); Tel.: +61-3-9905-4626 (A.L.C.)




Received: 30 June 2019; Accepted: 28 July 2019; Published: 2 August 2019


Abstract: Capturing CO2 directly from air is one of the options for mitigating the effects global
climate change, and therefore determining its cost is of great interest. A process model was
proposed and validated using laboratory results for adsorption/desorption of CO2 , with a branched
polyethyleneimine (PEI) loaded mesocellular foam (MCF) silica sorbent. The model was subjected to
a Multi-Objective Optimization (MOO) to evaluate the technoeconomic feasibility of the process and
to identify the operating conditions which yielded the lowest cost. The objectives of the MOO were
to minimize the cost of CO2 capture based on a discounted cash flow analysis, while simultaneously
maximizing the quantity of CO2 captured. This optimization identified the minimum cost of capture
as 612 USD tonne−1 for dry air entering the process at 25 ◦ C, and 657 USD tonne−1 for air at 22 ◦ C
and 39% relative humidity. The latter represents more realistic conditions which can be expected
for subtropical climates. The cost of direct air capture could be reduced by ~42% if waste heat was
utilized for the process, and by ~27% if the kinetics of the sorbent could be improved by a factor of
two. A combination of both would allow cost reductions of ~54%.

Keywords: direct air capture; economic; cost; model; steam; temperature vacuum swing; adsorption;
polyethyleneimine; carbon capture

1. Introduction
The increase in anthropogenic CO2 emissions has been identified as a main cause of global
climate change. Close to half [1] of these emissions are accounted for by diffuse sources such as motor
vehicles, homes and offices. Capturing these emissions at the source would require the fitting of
capture systems to each of these, which would neither be economical, nor practical. Alternatively,
the effect of these emissions could be compensated for with “Direct Air Capture” (DAC) systems,
i.e., processes which capture CO2 directly from the atmosphere. The captured CO2 could either
be sequestered or utilized for application in various industries [2]. The technologies evaluated for
DAC include absorption with aqueous hydroxide solutions, adsorption with solid inorganic bases,
and adsorption with solid-supported amines (SSA). A review of the studies done on DAC has been
detailed in Sanz-Pérez et al. [3]. Of these technologies, adsorption with SSA has presented itself to be
promising for DAC.
SSAs are a group of sorbents made of various amines, physically or chemically supported on
porous solid materials. They are well suited for DAC applications, due to their high uptake capacity
and selectivity, resilience to moisture which is present in air, and the possibility of regeneration
under relatively mild conditions. While significant research has been done on SSA for DAC, its focus

Processes 2019, 7, 503; doi:10.3390/pr7080503 www.mdpi.com/journal/processes

Processes 2019, 7, 503 2 of 23

has mainly been on sorbent development. There have only been a few reports [4–9] on the energy
requirements and the cost of capture of such systems, and there is significant discrepancy between
their results. These reports [4–9] evaluated DAC systems which used temperature concentration
swing adsorption (TCSA) [5,6], temperature vacuum swing adsorption (TVSA) [9] and steam-assisted
temperature vacuum swing adsorption (S-TVSA) [4,7,8] type processes. For TCSA, after the CO2 is
adsorbed on to the sorbent, the sorbent is heated and purged with a gas (steam in Krekel et al. [5]
and Zhang et al. [6]) to effect desorption. In TVSA, desorption is achieved by heating the sorbent and
applying a vacuum. S-TVSA is a hybrid of the two approaches, where in addition to applying heat and
a vacuum, a steam purge is also used.
In two of the early studies, Kulkarni and Sholl [7] and Zhang et al. [6] carried out an economic
analysis based only on the operating cost of the processes. They reported costs of CO2 capture
of 43–494 [7] and 91–227 [6] USD tonne−1 . Krekel et al. [5] expanded on the work carried out by
Zhang et al. [6], and reported that once the capital expenses are included, the cost of capture would be
increased significantly to 792–1200 USD tonne−1 . Sinha et al. [4] estimated a cost of capture of 60–190
USD tonne−1 .
Several companies are active in the field of DAC with SSA. The most well-known of these are
Climeworks [10] and Global Thermostat [11]. The cost of capture reported for the operations of
first-generation DAC system by Climeworks was estimated to be 600 USD tonne−1 [12]. This is an
important benchmark, as it is based on a commercially operating system, as opposed to results of
studies which are sensitive to the scope and assumptions used. Climeworks has further expressed
their confidence in reducing this cost down to 200 USD tonne−1 by 2021, and down to 100 USD tonne−1
by 2030 [12]. Similarly, Global Thermostat expects a cost of capture of 100 USD tonne−1 for their first
commercial DAC process [12]. Carbon Engineering [13] is another company which is developing a
DAC process, although they focus on an aqueous hydroxide-based process. The projected cost of
capture with this process has been reported to be 94–232 USD tonne−1 [14], but this process requires
very high temperatures for regeneration of the sorbent.
In comparison, removing CO2 from air via afforestation and forest management has a cost
of capture of 15–50 USD tonne−1 [15]. However, this approach carries a large land requirement,
which would compete with the land available for food production. Moreover, the land available also
enforces an upper limit on the scale at which CO2 can be removed. In comparison, CO2 removal by
DAC faces less stringent limitations on the degree of possible scaling. However, it is important to
stress that DAC is not envisioned to be an alternative for good forest management practices, but rather
as a technology to supplement the rate of CO2 removal by the natural carbon cycle.
The energy requirements of the DAC processes in prior works, is largely accounted for by the
thermal energy required for the desorption process. This mainly consists of the sensible heat required
to heat up the sorbent to the desorption temperature, and the large heat of desorption for CO2 .
When adsorption is considered under humid conditions, the heat of desorption of water, which gets
co-adsorbed on SSA, adds to the thermal energy demand [9]. Another big contributor to the energy
consumption is that needed for pushing air through the air-sorbent contactors. Due to the low
concentration of CO2 in air, large amounts of air need to be processed to capture the CO2 . This results
in large energy requirements, even for low pressure drop contactor configurations as are described in
some studies [4,7].
It was previously reported [16] that branched polyethyleneimine (PEI) loaded mesocellular foam
(MCF) silica sorbent in pelletized form is promising for CO2 adsorption under DAC conditions.
The study identified that while low temperatures were better for the thermodynamics of the reaction
between the CO2 and amine sites, the uptake of CO2 was limited by diffusional resistances in the
sorbent. In contrast, higher temperatures allowed for better diffusion of CO2 , but the thermodynamics
were less favored. The amount of CO2 adsorbed by the sorbent was determined by the combined effect
of these two factors. Due to the large diffusional resistances of the sorbent studied, the highest uptake
Processes 2019, 7, 503 3 of 23

was achieved under relatively warm conditions (46 ◦ C). The study further identified that low levels of
moisture in the gas (0.5 to 2% mol-H2 O), enhanced the CO2 uptake by up to 53%.
The same sorbent was evaluated using a S-TVSA adsorption/desorption cycle [17]. It was
identified that substantial CO2 desorption could be achieved with mild vacuum levels (12 to 56 kPa
abs), and temperatures (70 to 100 ◦ C). This indicated that the process could benefit from a reduced
electrical energy demand for vacuum generation and a lower capital cost owing to the use of smaller
vacuum pumps. It was further noted that as desorption is possible at relatively low temperatures,
the thermal energy requirement could be supplied with low grade heat. The desorption conditions
(pressure, temperature, steam flow rate) were seen to have a significant effect on the desorption
performance. Moreover, it was identified that while the presence of moisture enhanced CO2 adsorption,
it also resulted in a significant uptake of H2 O, which increased the thermal energy demand for the
desorption stage.
Building on this prior work, the present study aimed to (1) identify operating conditions which
yield the lowest cost of DAC, (2) determine the relative effects of varying operating conditions on its
technoeconomic performance and (3) identify promising directions for research on sorbent development
that could foster further cost reduction for DAC.
To address the first two aims, a DAC technoeconomic model was developed based on the results
of laboratory scale experimental data [17]. Next, the process model was subjected to a Multi-Objective
Optimization (MOO) to minimize the cost of capture, while simultaneously maximizing the amount
of CO2 captured. In contrast to prior studies [4–7] which fixed on particular process conditions,
the present study considers a range of conditions, seeking to minimize the cost. Moreover, adsorption
from dry versus and humid air was compared, to evaluate the effect of the water, which co-adsorbs,
on the process. Previous technoeconomic studies do not appear to have addressed this aspect [4–7].
The final aim was addressed through the use of case studies exploring the relative effect that sorbent
modifications could have on the cost of capture.

2. Methods

2.1. Adsorption/Desorption Model

To develop a process model and evaluate the economics, the laboratory scale data was fitted to a
simplified heat and mass transfer model, which accurately predicts the performance of the sorbent.
In this simplified model, it is assumed that the adsorption bed has no concentration or temperature
gradients in either the radial or the axial direction. With this assumption, it is possible to develop a
sorbent column transport model, without using isotherm data. This simple empirical model is used to
simulate the adsorption/desorption behavior of the sorbent for a specific sorbent bed configuration.
The model is limited to predicting the performance with a single bed thickness and is not able to
account for different bed thicknesses, as these would require isotherm and bed dispersion experimental
data. The following assumptions were made to develop the model:

• Ideal gas law is assumed for all gases;

• “Air” in this study comprises of 420 ppm CO2 in N2 ;
• Only CO2 and H2 O interact with the sorbent;
• Adsorption and desorption occur under isobaric conditions;
• There are no heat losses to the surroundings.

2.1.1. CO2 /H2 O Mass Transfer Kinetics

The mass transfer rate of CO2 and H2 O from/to the sorbent was approximated by the linear
dq
driving force (LDF) model [18,19] (Equation (1)) where dt is the mass transfer rate (mol kg-sorbent−1
s−1 ), qeq is the equilibrium adsorbed amount of CO2 or H2 O (mol kg−1 ) under the specified process
condition, and qt is the amount adsorbed on the sorbent (mol kg−1 ) at time t. k is the LDF mass transfer
Processes 2019, 7, 503 4 of 23

coefficient (s−1 ) which is a lumped parameter accounting for all the resistances to mass transfer. i refers
to the components H2 O or CO2 and j refers to ads (adsorption) or des (desorption).

dqi,j  
= ki,j qeq,i,j − qt,i, j (1)
dt
As the desorption kinetics are affected by the temperature and the partial pressure of CO2 , the LDF
mass transfer coefficient, kCO2 ,des , was expanded to account for temperature in Equation (2) and partial
pressure in Equation (3). Due to the lack of isotherm data, an empirical relationship (Equation (3)) with
the steam flow rate and the desorption pressure was used to predict the effect of the partial pressure
of CO2 on the desorption kinetics. In Equation (2), Ea is the activation energy (J mol−1 ), R is the gas
constant (J mol−1 K−1 ) and T is temperature (◦ C). k0,CO2 is a constant (s−1 ) which follows an empirical
.
relationship with the pressure (kPa abs), P, and steam flow rate (kg-steam h−1 kg-sorbent−1 ), msteam ,
according to Equation (3). k1–5 are empirical constants.
Ea,CO
− 2
kCO2 ,des = k0,CO2 e R(T +273) (2)
. !
msteam
− k3
k0,CO2 = k1 k2 − e (k4 P + k5 ) (3)

For H2 O mass transfer in the desorption stage, the LDF mass transfer coefficient, kH2 O,des ,
was expanded according to Equation (4), to account for the effect of temperature on the kinetics.
Ea,H O
− 2
kH2 O,des = k0,H2 O e R(T +273) (4)

For the adsorption stage, this study only considered a single adsorption temperature (27 ◦ C).
Therefore, the LDF constants, kH2 O,ads and kCO2 ,ads , were estimated as fixed values and not as relationships
to the process conditions.

2.1.2. Equilibrium Model for H2 O Uptake During the Desorption Stage

The equilibrium uptake of water by the sorbent under the desorption conditions were fitted to a
Freundlich isobar equation [20] (Equations (5) and (6)), where qeq is the equilibrium adsorbed amount
of water (mol kg−1 ), R is the gas constant (J mol−1 K−1 ), T is the temperature (◦ C) and P is the pressure
(kPa abs). K0 (mol kg−1 kPa−1/n0 ), α and A0 (J mol−1 ) are constants.
αR(T +273) 1
− A0
qeq,H2 O,des = K0 e P n0 (5)

Ao
n0 = (6)
R T + 273)
(

2.1.3. Heat Transfer Model

The heat transfer to/from the sorbent was modelled according to the energy balance given by
Equation (7), where msorbent is the mass of sorbent (kg), Cp,sorbent is the specific heat capacity of the
sorbent (J kg−1 ◦ C−1 ), Tbed is the bed temperature (◦ C), t is time (s), Theat/cool is the temperature (◦ C) of
the medium used to heat up or cool down the sorbent bed, U is the overall heat transfer coefficient
(W m−2 ◦ C−1 ) between the heating/cooling medium and the sorbent bed, A is the heat transfer area of
.
the bed (m2 ), m is mass flow rate (kg s−1 ) and h is the specific enthalpy (J kg−1 ).

dTbed
msorbent Cp,sorbent dt . .
= UA(Theat/cool − Tbed ) + min,H2 O hin,H2 O − mout,H2 O hout,H2 O (7)
. .
+min,CO2 hin,CO2 − mout,CO2 hout,CO2
Processes 2019, 7, 503 5 of 23

The specific enthalpies of the components were calculated according to Equation (8) where Hads
is the heat of adsorption (J kg−1 ), Cp is the specific heat capacity (J kg−1 ◦ C−1 ), T is the temperature
(◦ C) and Tre f is a reference temperature (◦ C), taken to be 50 ◦ C for this study. Hads,CO2 for this type of
sorbent was previously reported to be 2270 kJ kg−1 [21] and Cp,sorbent was measured to be 2 kJ kg−1 ◦ C−1
using a SETARAM micro DSC III. Hads,H2 O of 2611 kJ kg−1 was adopted from Wurzbacher et al. [9].
 
hi = Hads,i + Cp,i T − Tre f (8)

i refers to the component CO2 or H2 O.

2.1.4. Experimental Validation of Adsorption/Desorption Model and Parameter Estimation

The experimental data for the validation of the adsorption/desorption model was acquired using
the procedure described in Wijesiri et al. [17].
The parameters for Equation (1) for the mass transfer kinetics in the adsorption stage (ki,ads and
qeq,i,ads ) were determined by doing a least squares regression fit of the LDF model on the experimental
data for CO2 /H2 O uptake (qt,ads ) from 420 ppm CO2 at 27 ◦ C, under dry conditions and with 1%
mol-H2 O. The experimental data showed minimal changes in temperature (≤ 1 ◦ C) for both cases,
so the adsorption stage was assumed to be isothermal.
For the desorption stage, the CO2 mass transfer kinetics were experimentally determined for
the range of desorption conditions discussed in this study. Essentially all the CO2 was desorbed [17]
and therefore the equilibrium adsorbed amount (qeq,CO2 ,des term in Equation (1)) was assumed to be
0.00 mol kg−1 for all the desorption conditions. The constants Ea,CO2 and k1 to k5 , were estimated by
least squares regression of the experimental CO2 desorption data (qt,des ).
Similarly, the parameters for the Freundlich isobar (Equations (5) and (6)) were estimated by least
squares regression of the experimental equilibrium H2 O uptake data.
As it was not possible to obtain data on the H2 O mass transfer kinetics during the desorption
stage with the experimental set up used, it had to be approximated by using the bed temperature data
and the heat transfer model (Equation (7)). To do this, the heat transfer coefficient from the oven to the
sorbent bed, U was obtained by fitting the heat transfer model of the clean sorbent bed, Equation (7),
(with no CO2 or H2 O adsorbed) against the experimental bed temperature data (Tbed ).
The mass transfer kinetics of H2 O during the desorption stage were determined by coupling the
.
mass transfer model (Equations (1)–(4)) with the heat transfer model (Equation (7)). The mout,H2 O term
in Equation (7) was defined according Equation (9), where MWH2 O is the molar mass of H2 O (kg mol−1 )

. . dqH2 O
mout,H2 O = min,H2 O − × msorbent × MWH2 O (9)
dt
k0,H2 O and Ea,H2 O in Equation (4) were determined by least squares regression fitting Equation (7)
. . .
to the experimental bed and oven temperature data with the min,CO2 , mout,CO2 and min,H2 O data from
the experiments.

2.2. Process Model and Economic Model

2.2.1. Air-Sorbent Contactor Configuration

For the scaled-up process, an air-sorbent contact similar to that described in Patent
WO2014170184A1 [22] was considered. This consists of a series of stacked thin cylindrical adsorption
beds contained in a larger contactor (see Figure 1a). In such a contactor, air flows in through the
inlet on one side, flows up across the thickness of the sorbent bed, and exits through another side
of the contactor. The heating and cooling of the bed is envisioned to be done by heat transfer coils
directly under the sorbent beds. When scaling up, it was assumed that the ratio of heat transfer to
mass of sorbent would be kept the same as the lab scale experimental set up used. The beds were
Processes 2019, 7, 503 6 of 23

 
2.26 m in diameter dbed,contactor which is close to the maximum recommended for adsorption beds [23].
The height (Hbed,contactor ) was 0.10 m; this being the same height as the laboratory scale experiments.
The void factor () of the beds was assumed to be 0.5. It was also assumed that the diameter of
the sorbent pellets, dpellet , used in the contactor would be the same as the lab scale set up (1.8 mm).
These dimensions correspond to 250 kg of sorbent per bed. For the current study, a contactor with 16
beds, with
Processes a total
2019, of 4000
6, x FOR PEER kg of sorbent per contactor, was considered.
REVIEW 6 of 26

Figure 1. (a) A schematic of the proposed air-sorbent contactor configuration and (b) the process flow
Figure 1. (a) A schematic of the proposed air-sorbent contactor configuration and (b) the process flow
diagram of the proposed Direct Air Contact (DAC) system.
diagram of the proposed Direct Air Contact (DAC) system.
2.2.2. Process Description
2.2.2. Process Description
The process model considers two scenarios in order to determine the effect of moisture inherent in
The process
air collected model
from the considers two
surroundings: scenarios
adsorption in order
from dry air to and
determine the effect
adsorption from of airmoisture inherent
with a moisture
in air collected from the surroundings: adsorption from dry air and ◦
adsorption
content of 1% mol-H2 O. In both scenarios, the adsorption is carried out at 27 C. These two scenarios from air with a
moisture
are referredcontent
to as theof “dry
1% mol-H O. In“humid
case” 2and both scenarios,
case”. the adsorption is carried out at 27 °C. These two
scenarios are referred to as the “dry case”
A schematic of the process considered for the and “humid
currentcase”.
study is depicted in Figure 1b. The process
consists A schematic
of multipleofcontactors
the process considered
adsorbing COfor the current
2 from study is and
air in parallel, depicted in Figure
a single 1b. The
contactor process
desorbing
COconsists
. In theof multiple
adsorption contactors
stage, the adsorbing
incoming CO
air 2 from
is heated air
up in parallel,
to the and
adsorption a single contactor
temperature (27 ◦ C) and
desorbing
2
CO2. Inthrough
pushed the adsorption stage, the
the contactors usingincoming
a fan. Theair is
airheated
is heated,up toasthe adsorption
it was previously temperature
discovered(27 °C)that
[16] and
pushed
higher through the led
temperatures contactors
to higherusing
CO2a uptake,
fan. Thedue air istoheated,
improved as it diffusion
was previously discovered
kinetics. Heating [16] that
the air
higher
also temperatures
provides the coolinglednecessary
to higherfor CO 2 uptake,
the due to of
condensation improved
steam in diffusion kinetics.
the desorption Heating
stage. the air
A previous
also provides
study [16] reportedthe cooling
that thisnecessary
sorbent had forthe
thehighest
condensation
CO2 uptake at 46in◦ C.
of steam theHowever,
desorption stage.
as this A previous
temperature
study [16] reported that this sorbent had the highest CO2 uptake at 46 °C. However, as this
temperature corresponds to unrealistically warm climates, and heating the air to 46 °C would require
a significant amount of energy, a milder adsorption temperature of 27 °С is considered for the current
study.
The temperature at which the air could enter the process, 𝑇 (°C), was calculated according to
Equation (10), where 𝑄 , is the cooling duty of the condenser (W) (refer Section S1 of the
Processes 2019, 7, 503 7 of 23

corresponds to unrealistically warm climates, and heating the air to 46 ◦ C would require a significant
amount of energy, a milder adsorption temperature of 27 ◦ C is considered for the current study.
The temperature at which the air could enter the process, Tair (◦ C), was calculated according
to Equation (10), where Qcondenser , is the cooling duty of the condenser (W) (refer Section S1 of the
.
Supplementary Information), mair is the mass flow rate of air per contactor (kg s−1 ), Cp is the specific
heat capacity of air (J kg−1 ◦ C−1 ) and Ncontactors is the number of contactors in the system. The relative
humidity (RH) values corresponding to 1% mol-H2 O for the humid case were calculated according to
Equation (11), where Pair is the pressure of the air which is assumed to be 101 kPa abs, and psat,H2 O is
the saturation vapour pressure of water, calculated according to the Antoine correlation [24].

Qcondenser
Tair = 27 − . (10)
mair Cp.air × (Ncontactors − 1)

Pair × 1%
RH = (11)
Psat,H2 O
In the desorption stage, water is boiled and superheated to produce steam, which is then passed
through the contactor. Downstream of the contactor, the steam is desuperheated via the addition of
water, and passed through the condenser, which cools the CO2 /H2 O mixture down to 45 ◦ C. Following
this, the condensate is separated from the gas and returned to the boiler. The gas stream from the
condenser, consisting of CO2 saturated with H2 O, is compressed to atmospheric pressure using the
vacuum pump. The calculations for the unit operations in the process are described in S1 in the
Supplementary Information.
The adsorption/desorption cycle the contactors are subjected to is shown in Figure 2a. The cycle is
split into three stages. Firstly, the contactor goes through the adsorption stage, until a predetermined
amount of CO2 (qads ) is adsorbed into the bed. The desorption stage starts by first evacuating the
contactor to the desorption pressure and heating up the sorbent to the desorption temperature.
The heating is provided by a heat transfer fluid, at 10 ◦ C higher than the desorption temperature,
passing through the heating coils shown in Figure 1. After the sorbent is heated to just above the
steam dew point temperature, the steam purge is started. Following the introduction of steam, heat is
continuously supplied to the bed, to maintain the desorption temperature, for the remainder of the
desorption stage. The desorption stage ends after a predetermined amount of CO2 (qdes ) is desorbed.
Following this, the bed goes into the cooling stage, where the bed is cooled down to the adsorption
temperature by passing water, at 25 ◦ C, through the cooling coils. Finally, the contactor is re-pressurized
and goes back into the adsorption stage. For this study, it was assumed that the evacuation and
re-pressurization steps would take negligible time. The operational sequence of the contactors is shown
in Figure 2b, which shows how the contactors are subjected to the cycles, with one bed in desorption at
all times.

2.2.3. Energy Consumption

The energy requirements considered for the evaluation are the electrical energy required to operate
the fan (Efan ) in the adsorption stage and the vacuum pump (Evacuum ) in the desorption stage, and the
thermal energy required to heat the sorbent beds (Ebed ) and to produce steam (Esteam ). The methods
for calculating these are presented in Section S2 of the Supplementary Information.

2.2.4. Capital Cost

The purchased equipment costs of the major equipment were estimated using cost
correlations [23,25,26]. The cost correlations were adjusted for inflation using the chemical engineering
plant cost index (CEPCI). The total plant cost was calculated by accounting for insulation, piping,
instrumentation, electrical work, civil and structures, and lagging [27]. The detailed calculations are
included in Section S3 of the Supplementary Information.
 Processes
Processes 6, x 7,
2019,
2019, FOR503PEER REVIEW 8 of 826of 23

Figure 2. (a) The adsorption/desorption cycle the contactors are subjected to. (b) The operational
Figure 2. (a) of
sequence The
theadsorption/desorption cycle
contactors, with one bed the contactors
in desorption at all are subjected
times and the to.
rest(b) The operational
in adsorption. S1 to S3
sequence
refer toofthe
thestages
contactors,
in the with one bed in desorption
adsorption/desorption cycleat(adsorption,
all times anddesorption
the rest inand
adsorption.
cooling).S1 to S3
refer to the stages in the adsorption/desorption cycle (adsorption, desorption and cooling).
2.2.5. Operating Cost
2.2.3. Energy Consumption
The energy requirements of the process were assumed to be met by a solar thermal hot water
system and a solar
The energy photovoltaic
requirements system to
considered forminimize the CO2are
the evaluation emissions from the
the electrical process.
energy The energy
required to
costs the
were −1 −1
operate fantaken
(Efan) to
in be
the50 USD MWh
adsorption stageforandthermal energypump
the vacuum [28] and 100) USD
(Evacuum in theMWh for electrical
desorption stage,
andenergy [29]. The
the thermal energycostrequired
of watertowas
heattaken to be 3 USD
the sorbent m−3
beds (E [30] and the annual maintenance was
bed) and to produce steam (Esteam). The
assumed to be 2% of the total plant cost [27]. The sorbent
methods for calculating these are presented in Section S2 of the Supplementary was cost 8.1 USD kg−1 and
assumed to Information.
have a lifetime of 4 years. The cost of the sorbent was estimated by calculating the cost of the raw
materials
2.2.4. Capitalrequired
Cost and multiplying it by a factor of 3 to account for the production costs. More details
are included in Section S4 of the Supplementary Information.
The purchased equipment costs of the major equipment were estimated using cost correlations
[23,25,26]. Theofcost
2.2.6. Cost correlations were adjusted for inflation using the chemical engineering plant cost
Capture
index (CEPCI). The total plant cost was calculated−1by accounting for insulation, piping,
The cost of capture of CO2 (CCO2 ) in USD tonne is calculated from a discounted cash flow
instrumentation, electrical work, civil and structures, and lagging [27]. The detailed calculations are
calculation to obtain a zero Net Present Value (NPV) at the end of the project life using Equation (12),
included in Section S3 of the Supplementary Information.
where CCO2 is the cost of capture (USD tonne−1 ), Cplant , Cannual opex and Csorbent are the capital cost of
theOperating
2.2.5. plant, annual
Costoperating cost, and the cost of sorbent, respectively.
CR is the annual capture rate (tonne yr−1 ) as calculated according to Equation (13), where qdes
The energy requirements
is the amount of CO2 desorbed of the processtowere
(referred assumed
as the extent ofto desorption
be met by afromsolarhereon)
thermalper
hotcycle
water
(mol
system
kg−1and a solarisphotovoltaic
), msorbent the mass of system
sorbenttoinminimize
a single the CO2 emissions
contactor (kg), MW from theis process.
the molar The energy
mass of CO2
CO2
costs were taken
−1 to be 50 USD MWh −1 for thermal energy [28] and 100 USD MWh−1 for electrical
(kg mol ), Ncontactors is the number of contactors in the system, and tcycle is the total cycle time (h).
energy [29]. The
The process wascost of water
assumed was360
to run taken
daystoa year
be 3 for
USD24 m
−3 [30] and the annual maintenance was
h a day. More details are given in Section S5 of
assumed to be 2% of the total plant cost [27]. The sorbent was assumed to cost 8.1 USD kg−1 and have
a lifetime of 4 years. The cost of the sorbent was estimated by calculating the cost of the raw materials
Processes 2019, 7, 503 9 of 23

the Supplementary Information. For the NPV analysis, the plant operating lifetime is 20 years and a
discount rate of 10% is used.

(Cplant + Cannual opex 20 1 1

P P
x=1 (1+0.1)x + Csorbent y=0,4,...,16,20 (1+0.1) y )
CCO2 = (12)
CR × 20 1
P
x=1 (1+0.1)x

(qdes × msorbent × MWCO2 × Ncontactors )

CR = × 8.64 h year−1 t kg−1 (13)
tcycle

2.3. Multi-Objective Optimization

To determine the preferred operating conditions which would result in the lowest cost of capture
and the highest capture rate, a multi-objective optimization (MOO) was carried out on the process model
using the MATLAB gamultiobj function. This uses a controlled, elitist genetic algorithm (a variant of
NSGA-II [31]). The objectives for the MOO were to minimize (CCO2 ) and maximize (CR). The objectives,
variables and the constraints of the MOO problem are listed in Table 1.

Table 1. Definition of multi-objective optimization problem.

Objectives
Min (CCO2 ) and Max (CR)
Variables
Variable Range
.
Adsorption air flow rate to a single contactor (m3 h−1 kg-sorbent−1 ) V air 2 to 10
Total number of contactors Ncontactors 2 to 60
Desorption temperature (◦ C) Td 80 to 100
Desorption pressure (kPa abs) Pd 12 to 26
.
Desorption steam flow rate (kg h−1 kg-sorbent−1 ) msteam 0.09 to 1.86
0.25 to 1.50 (dry case)
Extent of adsorption (mol kg−1 ) qads
0.25 to 2.75 (humid case)
0.20 to 1.45 (dry case) 0.20 to 2.70
Extent of desorption (mol kg−1 ) qdes
(humid case)

Additionally, the constraint in Equation (14) was enforced to ensure that all the contactors can be
desorbed within a single cycle, where tads and tdes are the durations of the adsorption and desorption
stages (h) and Ncontactors is the number of contactors in the system.

tads ≥ tdes × (Ncontactors − 1) (14)

2.4. Sensitivity Analysis

Two sensitivity analyses were carried out. Firstly, the sensitivity of the results of the cost of capture
to some of the values in the model which carried uncertainty were tested. This was done by varying
these values by ±10% from the original values and observing the change in the cost of capture in the
lowest cost scenario. The values of interest chosen were cost of energy, cost of the contactor, discount
rate, sorbent cost and lifetime and the plant lifetime.
Next, the sensitivity of the results to the MOO parameters used was tested. This was done by
varying the parameters used for the original MOO.

3. Results of Model Validation and Parameter Estimation

The experimental data and the model predictions for CO2 /H2 O mass transfer in the adsorption
stage are presented in Figure 3. The CO2 mass transfer, equilibrium H2 O uptake, H2 O mass transfer
kinetics and heat transfer in the desorption stage are presented in Figures 4–6. The figures show that
varying the parameters used for the original MOO.

3. Results of Model Validation and Parameter Estimation

The experimental data and the model predictions for CO2/H2O mass transfer in the adsorption
Processes
stage are2019, 7, 503
presented 10 of 23
in Figure 3. The CO2 mass transfer, equilibrium H2O uptake, H2O mass transfer
kinetics and heat transfer in the desorption stage are presented in Figures 4–6. The figures show that
the experimental data are in close agreement with the model predictions. The results of the parameter
the experimental data are in close agreement with the model predictions. The results of the parameter
estimations are listed in Table 2 along with the constants used.
estimations are listed in Table 2 along with the constants used.

Figure
Processes 2019,3. The
6, x experimental
FOR PEER REVIEWdata and the model predictions for CO2 and H2 O adsorption from 420 ppm
11 of 26
Figure 3. The experimental data and the model predictions for CO2 and H2O adsorption from 420
CO2 in N2 in the dry and humid cases.
ppm CO2 in N2 in the dry and humid cases.

Figure 4.
Figure Theexperimental
4. The experimentaldata
data and
and thethe model
model predictions
predictions for for
COCO 2 mass
2 mass transfer
transfer kinetics
kinetics in thein
the desorption stage. Legend for titles AA_BBB_CCC_DD (AA- desorption pressure
desorption stage. Legend for titles AA_BBB_CCC_DD (AA- desorption pressure (kPa abs), BBB- (kPa abs),
BBB-desorption temperature ◦ C), CCC- desorption steam flow rate (kg−1h−1 kg-sorbent −1 ), DD- amount
desorption temperature (°C),( CCC- desorption steam flow rate (kg h kg-sorbent−1), DD- amount of
of water adsorbed during adsorption stage (mol kg −1 )). More results are depicted in Figure S1 in
in the
the
water adsorbed during adsorption stage (mol kg )). More results are depicted in Figure S1
−1

Supplementary Information.
Supplementary Information.
 Figure 4. The experimental data and the model predictions for CO2 mass transfer kinetics in the
desorption stage. Legend for titles AA_BBB_CCC_DD (AA- desorption pressure (kPa abs), BBB-
desorption temperature (°C), CCC- desorption steam flow rate (kg h−1 kg-sorbent−1), DD- amount of
water adsorbed during adsorption stage (mol kg−1)). More results are depicted in Figure S1 in the
Processes 2019, 7, 503 11 of 23
Supplementary Information.

Figure
Processes 5.6,The
2019, experimental
x FOR data and the model predictions for equilibrium H2 O adsorption amount in
PEER REVIEW 12 of 26
Figure 5. The experimental data and the model predictions for equilibrium H2O adsorption amount
the desorption stage at desorption pressures of 12 kPa and 26 kPa abs.
in the desorption stage at desorption pressures of 12 kPa and 26 kPa abs.

Figure 6. The experimental data and the model predictions for heat transfer and H2 O mass transfer
Figure 6.
kinetics inThe
the experimental data and
desorption stage. the model
Legend predictions
for titles for heat transfer
AA_BBB_CCC_DD (AA-and H2O masspressure
desorption transfer
kinetics
(kPa abs),inBBB-desorption
the desorption temperature
stage. Legend(◦ C),
for CCC-
titles AA_BBB_CCC_DD
desorption steam flow (AA- (kg h−1 kg-sorbent
desorption
rate −1 ),
pressure (kPa
abs),amount
DD- BBB-desorption
of water temperature (°C), adsorption
adsorbed during CCC- desorption steam
stage (mol −1
kgflow)). rate
More h kg-sorbent
(kgresults
−1 −1 ), DD-
are depicted in
Figure
amount S2ofin water
the Supplementary Information.
adsorbed during adsorption stage (mol kg−1)). More results are depicted in Figure
S2 in the Supplementary Information.
Processes 2019, 7, 503 12 of 23

Table 2. The results of the parameter estimation for the adsorption/desorption model and the
constants used.

Parameter Value
Adsorption stage mass transfer
Dry case
kCO2 ,ads (s−1 ) 2.21 × 10−5
qeq,CO2 ,ads (mol kg−1 ) 1.55 × 100
Humid case
kCO2 ,ads (s−1 ) 1.31 × 10−5
qeq,CO2 ,ads (mol kg−1 ) 2.80 × 100
kH2 O,ads (s−1 ) 6.74 × 10−5
qeq,H2 O,ads (mol kg−1 ) 8.30 × 100
Desorption stage mass transfer
CO2 mass transfer kinetics
Ea,CO2 (J mol−1 ) 1.44 × 105
k1 (s−1 ) 1.06 × 100
k2 1.03 × 100
k3 (kg-steam h−1 kg-sorbent−1 ) 7.56 × 10−1
k4 (kPa−1 ) −2.29 × 1013
k5 6.66 × 1014
Freundlich isobar for H2 O
K0 (mol kg−1 kPa−1/n0 ) 1.30 × 107
α 1.51 × 101
A0 (j mol−1 ) 1.63 × 103
H2 O mass transfer kinetics
Ea,H2 O (J mol−1 ) 4.05 × 103
k0,H2 O (s−1 ) 1.90 × 103
Desorption stage heat transfer
Ulab (W m−2 ◦ C−1 ) 1.70 × 101
Lab scale sorbent bed dimensions
Alab (m2 ) 3.14 × 10−3
msorbent,lab (kg) 3.45 × 10−3
Thermodynamic properties and constants
Cp,sorbent (J kg1 ◦ C−1 ) 2.00 × 103
Cp,CO2 (J kg1 ◦ C−1 ) 9.00 × 102
Cp,steam (J kg1 ◦ C−1 ) 2.00 × 103
Cp,air (J kg1 ◦ C−1 ) 1.00 × 103
Hads,H2 O (J kg−1 ) 2.61 × 106
Hads,CO2 (J kg−1 ) 2.27 × 106

4. Results of Multi-Objective Optimization

The plots of the Pareto non-dominated fronts from the results of the MOO are presented in
Figures 7 and 8 for the dry and the humid cases. From Figure 7, it is apparent that the dry case allows
for a lower cost of capture than the humid case for all capture rates. Moreover, it can be seen that the
cost of capture increases as the capture rate is increased for both cases. From Figure 8, it is apparent
.
that as the capture rate is increased, Ncontactors , msteam and qads are also increased. Additionally, Pd is
reduced for the humid case. Negligible changes are seen in the other parameters.
The set of variables which yielded the lowest cost scenarios for the humid and the dry cases are
listed in Table 3, along with the resulting cycle times and energy requirements of the process.
Processes 2019, 7, 503 13 of 23
Processes 2019, 6, x FOR PEER REVIEW 14 of 26

Figure 7. Pareto
Figure
Processes plots
7.6,Pareto
2019, x FOR forREVIEW
plots
PEER thethe
for objective
objectivefunctions
functions of
of the Multi-ObjectiveOptimization
the Multi-Objective Optimization (MOO)
(MOO) for
15 the
for the of 26
dry dry
andand
the the
humid cases.
humid cases.

Figure 8. Pareto plots for the variables plotted against the capture rate, for the humid and the dry case.
Figure 8. Pareto plots for the variables plotted against the capture rate, for the humid and the dry
The limits for the y-axes are the bounds for each parameter used in the MOO. Td was not plotted as it
case. The limits for the y-axes are the bounds for each parameter used in the MOO. 𝑇 was not plotted
was constant at 100 ◦ C
as it was constant at across
100 °C all the all
across datathepoints.
data points.
Processes 2019, 6, x FOR PEER REVIEW 16 of 26

Table 3. The process conditions which yield the lowest cost scenarios and the resulting cycle times
and energy requirements of the process.
Processes 2019, 7, 503 14 of 23
Parameter Dry Humid
Variables
Table 3. The process conditions which yield the lowest cost scenarios and the resulting cycle times and
Adsorption air flow rate to a single contactor
energy requirements3of−1the process. −1 𝑉 5.23 5.81
(m h kg-sorbent )
Total no ofParameter
contactors 𝑁 14
Dry 15
Humid
Desorption temperature (°C) Variables 𝑇 100 100
Desorption
Adsorption air flow rate to pressure (kPa abs)(m h kg-sorbent )
a single contactor 3 −1 −1 𝑃 .
V air 18.33
5.23 16.85
5.81
Desorption steam Total flow no rateof(kg h−1 kg-sorbent−1)
contactors 𝑚 Ncontactors 0.11
14 0.11
15
Extent Desorption
of adsorption (mol kg(−1◦)C)
temperature 𝑞 Td 100
0.79 100
1.72
Desorption pressure
Extent of desorption (mol kg ) (kPa abs)
−1 𝑞 . Pd 18.33
0.54 16.85
1.55
Desorption steam flow rate (kg h−1 kg-sorbent−1 ) msteam 0.11 0.11
Cycle times
Extent of adsorption (mol kg−1 ) qads 0.79 1.72
Adsorption (h)
Extent of desorption (mol kg−1 )
𝑡 q 6.87
0.54 19.53
1.55
des
Desorption (h) Cycle times 𝑡 0.52 1.39
Cooling(h)
Adsorption (h) 𝑡 tads 0.11
6.87 0.11
19.53
FullDesorption
cycle (h) (h) 𝑡 tdes 0.52
7.51 1.39
21.03
Cooling(h) tcool 0.11 0.11
Energy requirement
Full cycle (h) t 7.51 21.03
Electrical energy for fan (GJ tonne 𝐸 cycle
Energy)requirement 1.35 1.71
−1

Electrical energy forenergy

Electrical vacuum forpump
fan (GJ(GJ tonnetonne
−1 ) −1) 𝐸 E f an 0.61
1.35 0.65
1.71
Thermal
Electricalenergy
energy for steam (GJ
for vacuum pumptonne −1 )
(GJ tonne −1 ) 𝐸 Evacuum 7.05
0.61 6.26
0.65
ThermalThermal
energy to energy
heatfor steam (GJ tonne
sorbent tonne−1−1)) 𝐸 Esteam 7.05
6.14 6.26
9.45
Thermal energy to heat sorbent (GJ tonne −1 ) Ebed 6.14 9.45
Total electrical (GJ tonne ) −1 −1 𝐸 1.96 2.36
Total electrical (GJ tonne ) Eelectrical 1.96 2.36
Total thermal (GJ tonne ) −1 −1 𝐸 13.18 15.71
Total thermal (GJ tonne ) Ethermal 13.18 15.71
TotalTotal
(GJ tonne −1)
(GJ tonne−1 ) 𝐸 E 15.14
15.14 18.08
18.08
total
Cost
Costand andcapture rate
capture rate
Cost ofCostcapture (USD
of capture (USDtonnetonne−1) −1 ) 𝐶 CCO2 612
612 657
657
−1 )
Annual capture rate
Annual capture rate (tonne yr ) (tonne yr
−1 𝐶𝑅 CR 1521
1521 1682
1682

The
The temperature
temperatureandandRHRH at
at which
which the
the air
air could
could enter
enter the
the process
process for
for each
each of
of the
the points
points in
in the
the
Pareto
Pareto plots,
plots, as
as calculated
calculated with
with Equations
Equations (10)
(10) and
and (11),
(11), are
are depicted
depicted in
in Figure
Figure 9.
9.

Figure 9. The temperature and RH of the inlet air to the process for each of the points in the Pareto plots.
Figure 9. The temperature and RH of the inlet air to the process for each of the points in the Pareto
plots.
The breakdown of the cost of capture for the preferred conditions is shown in Figure 10. It was
observed that both the dry and humid case showed a similar cost breakdown. The biggest contributors
were identified as the cost of thermal energy and the contactors.
Processes 2019, 6, x FOR PEER REVIEW 17 of 26
Processes 2019, 6, x FOR PEER REVIEW 17 of 26
The breakdown of the cost of capture for the preferred conditions is shown in Figure 10. It was
observedbreakdown
The that both of
thethedry
costand
of capture
humid for theshowed
case preferreda conditions is shown
similar cost in Figure
breakdown. The10.biggest
It was
observed that both the dry and humid case showed a similar cost
contributors were identified as the cost of thermal energy and the contactors. breakdown. The biggest
Processes 2019, 7,were
contributors 503 identified as the cost of thermal energy and the contactors. 15 of 23

Figure 10. The contributions of various components to the cost of capture for the lowest cost case for
Figure
the dry10.
andThe contributions
humid case. The of various
value components
for the toisthe
humid case costinside
given of capture for the
brackets lowest
in the datacost case for
labels.
the dry and humid case. The value for the humid case is given inside brackets in the data labels.
The results
results of the sensitivity
sensitivity analysis
analysis on
on the
the model
model values
values is
is shown
shown in
in Figure
Figure 11.
11. As the two cases
had aThe results
similar ofbreakdown,
cost the sensitivity
theanalysis on the
sensitivities model
were values
almost is shown
identical.
the sensitivities were almost identical. in Figure 11. As the two cases
had a similar cost breakdown, the sensitivities were almost identical.

Figure 11. Tornado plot of the sensitivity of the cost of capture of the lowest cost scenarios for ± 10%
Figure 11. Tornado plot of the sensitivity of the cost of capture of the lowest cost scenarios for ± 10%
changes in the model values.
Figure
changes11.
inTornado plot
the model of the sensitivity of the cost of capture of the lowest cost scenarios for ± 10%
values.
changes in the model values.
5. Discussion
5. Discussion
5. Discussion
5.1. Influence of Process Conditions
5.1. Influence
It can beofnoted
Processthat
Conditions
the cost of capturing under humid conditions is higher than that under
5.1. Influence of Process Conditions
dry conditions
It can be noted that thecapture
for all the cost of rates in theunder
capturing Pareto frontsconditions
humid (Figure 7). The presence
is higher than thatofunder
moisture
dry
allowsIt can be noted
the process
conditions that the
to capture
for all the cost of
take advantage capturing
rates in of
thethe under
improved
Pareto humid
frontsCO conditions
2 adsorption
(Figure is higher than that
capacitiesofofmoisture
7). The presence under
the sorbent [16]dry
allows as
the
conditions
indicated by forthe
all higher
the capture rates
qads and in in
qdes thethe
Pareto
humidfronts (Figure
case. 7). The
However, presence
this of moisture
also leads allows
to a longer the
tads and
tdes , which increases tcycle . This reduces the annual capture rate, and so increases the cost of capture
(see Equations (12) and (13)).
Processes 2019, 7, 503 16 of 23

In the humid case, the process is restricted from using smaller qads and qdes , due to the large
amount of water adsorbed. As the water adsorption/desorption is significantly faster than that of CO2
(see Figure 3), at lower qads and qdes (shorter tads and tdes ) values, there is a much higher proportion of
H2 O adsorbed in comparison to CO2 . This in turn translates to a proportionately higher cost of thermal
energy consumed for desorbing the H2 O. And as the thermal energy accounts for a large fraction of
the cost, this has a significant impact on the economics of the process. These results indicate that using
a sorbent with a reduced affinity for water may improve the economics of the process.
When comparing the process conditions which yielded the lowest cost scenario, both the humid
.
and the dry case had a similar adsorption air flow rate (V air ), desorption temperature (Td ) and
.
desorption pressure (Pd ). At high V air , the rate at which CO2 can be captured is increased, which can
reduce the cycle time. However, this is done at the compromise of higher pressure drops across the
contactors resulting in an increased electrical energy consumption.
The preferred Td was identified as 100 ◦ C which is the upper limit used for the optimization.
This is due to the faster desorption kinetics presented at higher temperatures [17], which leads to
shorter cycle times. The upper limit for this study was set at 100 ◦ C, as solid supported amines have
been reported to undergo degradation at high temperatures [32]. However, the sorbent in this study
was previously reported [17] to be stable at temperatures up to 100 ◦ C, under the desorption conditions
evaluated here. In the case of Pd , the preferred value is a balance between benefitting from the faster
desorption rates (shorter cycle times) offered by the lower pressures [17] and the increased expenses
for electrical energy.
.
The identification of the preferred steam flow rate, msteam , is similar to that of Pd . The preferred
value is a balance between benefitting from the faster desorption rates (shorter tdes ) offered by the faster
steam flow rates [17] at the compromise of increased cost of thermal energy. In both cases, the steam
flow rate with the lowest cost was the close to its lower bounds, indicating that benefit of shorter tdes is
not justified by the higher energy cost. This can be attributed to the fact that thermal energy demand
has the one of the biggest impacts on the economics of the process (see Figure 10).
When comparing the energy requirements, the electrical energy consumption was similar for
both cases. However, the humid case needed more thermal energy to heat up the sorbent during
the desorption. This is because the energy consumed by the water being desorbed from the sorbent,
cools it down [17]. Therefore, more heat needs to be provided in the humid case, to maintain the
desired desorption temperature.
From Figure 11, it can be seen that both cases display similar sensitivities to the values studied.
The final result was seen to be relatively insensitive to the values studied, as the cost was seen to vary
by <±4% for a variation of ±10% in the parameters. The cost of thermal energy and the contactor
were identified to be the variables which most affected the cost of capture. This is consistent with
the fact that thermal energy and the cost of the contactor are the two largest contributions to the cost
(see Figure 10). The discount factor was also seen to have a large impact on the cost of capture and a
value of 10% is conservative for the current economic conditions.

5.2. Pareto Non-Dominated Fronts

In both cases, the cost of capture increases as the capture rate increases. As seen in the Pareto
plots of the variables (Figure 8), the increase in the capture rate is achieved mainly by using a higher
Ncontactors which increases the amount of CO2 captured. However, to increase Ncontactors , either tads needs
to be increased or tdes needs to be decreased to satisfy the inequality constraint in Equation (14). This is
.
fulfilled by increasing msteam , which shortens tdes by improving the CO2 desorption kinetics. Pd is also
decreased for the humid case, which can be attributed to the comparatively longer tdes , which demands
a greater increase in the kinetics to satisfy the inequality constraint in Equation (14). Furthermore,
qads is increased to increase the tads and qdes is kept relatively constant which shortens tdes when used
.
in combination with higher msteam , and lower Pd . These changes to the process conditions (except
qdes ) increase the annual capture rate while incurring additional expenses as explained in Section 5.1.
Processes 2019, 7, 503 17 of 23

The increases in the thermal energy demand and the contactors are particularly significant as they are
the two biggest contributors to the economics of the process.
.
V air , can be seen to be relatively constant for most of the Pareto front. This can be expected since it
is selected based on the compromise between the positive effect of a shorter tads and the negative of a
higher energy requirement for high flow rates. It is likely that the balance between the two does not
significantly vary for the different points in the Pareto plot. In the dry case, there is a steep increase in
.
V at the higher capture rate end which coincides with a steep increase in the cost. This highlights that
.
while slightly higher capture rates can be achieved with higher V air , it is accompanied by a significant
increase in the cost.
The results of the sensitivity analysis carried out on the MOO options is included in Section S6 of
the Supplementary Information, and it is evident that the Pareto fronts are insensitive to the MOO
options used.

5.3. Temperature and RH of the Incoming Air to the Process

For the lowest cost cases, the incoming air temperature, Tair , was calculated to be 25 ◦ C for the
dry case and 22 ◦ C for the humid case (see Figure 9). The RH of the humid case was calculated to be
39%. It is noted that, while the dry case yielded a lower cost, it is unrealistic to expect completely dry
conditions anywhere in the world. However, the conditions calculated for the humid case are far more
realistic and can be expected of regions with sub-tropical climates.
Tair can be seen to reduce with increases in the capture rate (see Figure 9). This can be attributed
to the higher steam flow rates available at the higher capture rate/higher cost end of the Pareto fronts.
It is also apparent that the humid case allows for significantly lower temperatures. This is owing to the
water which gets desorbed from the sorbent which condenses, releasing additional thermal energy.
These results indicate that the process could operate in colder climates with higher humidity, at the
compromise of a higher cost of capture. It is also noteworthy that when operating at a higher moisture
level than 1% mol-H2 O, lower Tair may be achieved due to the increased adsorption of water, although
this would also likely lead to a higher cost of capture.

5.4. Comparison to Results of Other Studies

The cost and energy data for the lowest cost case described in the current study is presented
alongside data found for other DAC processes [4–9,12,14,33] in Figure 12. The data are that of the
lowest cost case described in the respective studies.
It is evident that the cost of capture presented in the current study is in the upper end of the range
of costs reported for similar processes. However, it can be noted that the cost in the current study is
similar to that reported by Climeworks based on the actual performance of their first-generation DAC
system (DAC-1) [12]. In comparison, the other values depicted in Figure 12 are subjective to the scope
of the respective study and the assumptions that were used. For example, Kulkarni and Sholl [7] and
Zhang et al. [6] did not include the capital expenses of the equipment in their studies. Krekel et al. [5]
expanded on the work carried out by Zhang et al. [6], and reported that once the capital expenses
are included, the cost of capture is increased significantly. Moreover, the costs of capture depicted in
Figure 12 for Zhang et al. [6] and Krekel et al. [5] are based on scenarios where the energy is supplied
with fossil fuels, as opposed to the low carbon solar energy used here. Krekel et al. [5] reported that the
cost of capture increases from 792 to 1333 USD tonne−1 when the emissions from energy generation are
considered. They further reported that if wind energy is used, a capture cost of 824 USD tonne−1 can
be achieved. Zhang et al. [6] estimated an increase from 91 to 225 USD tonne−1 for a system utilizing
wind and nuclear energy.
Processes 2019, 7, 503 18 of 23
Processes 2019, 6, x FOR PEER REVIEW 20 of 26

Figure 12. Comparison of the cost and energy data from this study to that of previous studies. The data
Figure
are that12.
of Comparison of case
the lowest cost the cost and energy
described in thedata from this
respective studya Carbon
studies. to that of previous studies.
Engineering refers toThe
an
data are that of the lowest cost case described
aqueous hydroxide-based process for DAC. in the respective studies. a Carbon Engineering refers to

an aqueous hydroxide-based process for DAC.

The large difference between the results of this study and that of the Sinha et al. [4] is in part due
to theIt two
is evident that the cost
main differences. of capture
Firstly, the cash presented
flow wasinnot thediscounted
current study is insecond
[4]. The the upper
reason end of the
is due to
range of costs
the nature of thereported
sorbentfor similar
used. Sinha processes. However, aitmonolithic
et al. [4] evaluated can be noted that the
sorbent which costwasin the current
assumed to
study is similar to
be self-contained. that reported
In comparison, the by Climeworks
pelletized sorbentbasedin the on currentthe study
actualneedsperformance
to be containedof their in
first-generation
contactors which DAC system to
contribute (DAC-1)
about a[12]. third In ofcomparison,
the total cost.the other
This may values depicted
indicate thein Figure 12that
advantage are
subjective
monolithictosorbents the scope have ofover
the respective
pelletized ones. studyFurthermore,
and the assumptionsSinha et al. that[4]were
also didused.notFor example,
consider the
Kulkarni and Sholl [7] and Zhang et
CO2 emissions from energy generation in their calculations. al. [6] did not include the capital expenses of the equipment in
their Apart
studies. Krekel
from these, et al. [5] expanded
another major reason on thefor work thecarried
higherout costby Zhang et
reported inal.
the [6], and reported
current study isthat the
once the capital expenses are included, the cost of capture is
slower kinetics of the sorbent evaluated. It was previously reported [16] that the large loading ofincreased significantly. Moreover, the
costs
PEI inofthiscapturesorbentdepictedresultedin Figure 12 for
in a large CO Zhang et al. [6] and Krekel et al. [5] are based on scenarios
2 uptake capacity although at the sacrifice of significant
where the energy
mass transfer is supplied
limitations, whichwithresulted
fossil fuels,in a asslowopposed
uptaketorate theof low CO carbon solar energy used here.
2 . The effect of this is evident
Krekel
here, where long cycle times are required for the process (7.51 h for the dryUSD
et al. [5] reported that the cost of capture increases from 792 to 1333 case).tonne
−1 when the
In comparison,
emissions from energy
for other studies, generation
the reported cycleare considered.
times are in theThey range further
of 1.27reported
[4] to 4 hthat [7].ifThese
windlongenergy cycleis used,
times
aresult
capture cost of 824 USD tonne −1 can be achieved. Zhang et al. [6] estimated an increase from 91 to
in a lower annual capture rate and hence a larger cost. Furthermore, it should be noted that
225
the USD
subject tonne −1 for a system utilizing wind and nuclear energy.
sorbent was shown to provide better adsorption performance at the higher temperature
of 46The
◦ largethus,
C [16]; difference
it maybetween the results
be that further of this studyinand
improvements thethat
processof theeconomics
Sinha et al. [4] isbe
could in achieved
part due
to
by tailoring the design of sorbent material so that its optimum sorption performance occursisatdue
the two main differences. Firstly, the cash flow was not discounted [4]. The second reason the
to the nature
intended of the sorbent
adsorption used. Sinha et al. [4] evaluated a monolithic sorbent which was assumed
temperature.
to be The
self-contained.
slow kinetics In comparison,
also have a large the pelletized
effect on the sorbent
energyin the current study
requirement of theneeds to be As
process. contained
seen in
in
Table 3, the thermal energy requirement is dominated by the energy for steam generation, andthat
contactors which contribute to about a third of the total cost. This may indicate the advantage for
monolithic
longer desorption sorbentstimes, have larger
over pelletized
amounts of ones.
steamFurthermore,
are neededSinha to desorbet al.the[4] CO
also2 .did
Thisnot consider
would the
explain
CO emissions
the2larger thermalfromenergyenergyrequirement
generation in of their calculations.
the process in comparison to the other studies depicted in
Apart from these, another major
Figure 12. However, the electrical energy requirement reason for the higher cost reported
of the process in the
is in close current with
agreement studymost is the
of
slower
the values kinetics
reportedof the in sorbent
literature.evaluated. It was previously
While Wurzbacher et al. [9]reported
and Gebald [16]etthat thereported
al. [8] large loadingmuchof PEI
lower
in this sorbent
values, they only resulted
evaluated in a the
large CO2 uptake
desorption stage capacity
and didalthough
not include at the
thesacrifice
energy neededof significant
to pump mass
the
transfer limitations, which
air through contactors during adsorption. resulted in a slow uptake rate of CO 2 . The effect of this is evident here,
where It long cycle times
is difficult to make are required for the process
direct comparisons with(7.51 h for the
the costs dry case).
reported In comparison,
by Carbon Engineering for other
[14],
studies, the reported cycle times are in the range of 1.27 [4] to 4
as this corresponds to a completely different type of process. However, it is noted that the need forh [7]. These long cycle times result in
avery
lower
high annual capture (900
temperatures rate and
◦ hence
C) for the aregeneration
larger cost. of Furthermore,
the sorbent, it should
may makebethe noted thatmore
process the subject
reliant
sorbent was shown to provide better adsorption performance at the higher temperature of 46 °C [16];
thus, it may be that further improvements in the process economics could be achieved by tailoring
Processes 2019, 7, 503 19 of 23

on fossil fuel-based energy. In comparison, the processes using SSA could more easily be powered
with relatively inexpensive low carbon energy sources like solar thermal hot water systems.

5.5. Case Studies

In this section, the effect of five different scenarios on the cost of capture is evaluated. The scenarios
were selected based on the main cost drivers and the limitations of the process identified in the previous
sections. For these evaluations, the MOO was repeated with the respective changes described below
for each scenario.
Case study A: Thermal energy from waste heat utilization—This scenario was chosen because
the main contributor to the cost of the capture was identified to be the thermal energy requirement.
Moreover, the low temperatures (≤100 ◦ C) used for the desorption process makes it possible for
integration with waste heat sources. It was assumed that this energy could be achieved free of charge.
Case study B: Hypothetical sorbent with faster kinetics—This scenario addresses the limitation
of slow mass transfer kinetics of the sorbent (longer cycle times). Here a hypothetical sorbent,
with improved mass transfer kinetics of CO2 is used. Potential means of achieving this include using
additives which improve CO2 diffusion in the amine phase [34–37] and using support materials with
high surface areas [38] and/or large pore sizes [39] to better disperse the amines and thus improve
their accessibility to the CO2 . To carry out the evaluation, kCO2 ,ads in Equation (1) and k1 in Equation (3)
were multiplied by a factor of 2, to emulate a sorbent with twice as faster CO2 mass transfer kinetics.
Case study C: Hypothetical sorbent with faster kinetics but lower CO2 uptake capacity—This
scenario considers an iteration of case study B. It is possible that the improvements made to the kinetics
would come at the expense of a smaller CO2 uptake capacity. For example, it has been reported in
literature that lower PEI loadings (% wt) on the sorbent results in lower uptake capacities and faster
kinetics [40,41]. For this scenario, a hypothetical sorbent where the kinetics are increased by a factor of
2, at the compromise of a halved equilibrium capacity, is evaluated. To carry out the evaluation, kCO2 ,ads
in Equation (1) and k1 in Equation (3) were multiplied by a factor of 2, and qeq,CO2 ,ads in Equation (1)
was multiplied by 0.5.
Case study D: Combination of case studies A and B, where a sorbent with superior kinetics is
used with waste heat utilization.
Case study E: Hypothetical sorbent with lower H2 O uptake capacity—As mentioned earlier,
while adsorbing from humid air enhances the CO2 uptake, a proportionately higher amount of water
is also adsorbed. This results in a higher energy requirement for desorption of water. It has been
reported [42] that the water uptake by PEI-silica sorbents could be reduced by using hydrophobic
silica as the support material. The study [42] further demonstrated that the benefit of enhanced CO2
capacities when adsorbing in the presence of moisture could be retained, in spite of the reduction of
the water uptake. Modifications to the amine such as reducing the proportion of primary amines [43]
and introduction of methyl groups [44] have also been reported to reduce the water uptake by these
sorbents. For this scenario, a hypothetical sorbent, with half the water uptake capacity is evaluated.
To carry out the evaluation, qeq,H2 O,ads in Equation (1) and K0 in Equation (5) were multiplied by a factor
of 0.5.
For case studies B–E, the current study does not attempt to identify exactly how the sorbent may
be developed to meet the specified criteria but simply aims to estimate the impact each modification
would have on the economics of the process. The results of this exercise are meant to be taken as a
guide, for identifying the direction future research into sorbent development may head towards.
The results of the case studies are depicted in Figure 13. From the results, it is evident that using
waste heat to provide the thermal energy yields the biggest reduction (~ 42%) in the cost of capture.
This reduction is an effect of two things. Firstly, the large contribution of the thermal energy to the
cost is eliminated. Secondly, the process takes advantage of the free thermal energy and utilizes
.
higher msteam to increase the capture rate. This is evident in the thermal energy demands which are
.
more than 3-fold higher than that of the base case. The higher msteam facilitates faster desorption
 Processes 2019, 7, 503 20 of 23

which reduces tcycle , which in turn lowers the cost of capture. While the reductions are attractive,
it should also be noted that the utilization of waste heat would restrict the locations where DAC
systems
Processes could
2019, be operated
6, x FOR to areas such as industrial parks. DAC systems with waste heat utilization
PEER REVIEW 23 of 26
would also likely be more practical for carbon capture and utilization projects rather than carbon
capture
captureandandsequestration
sequestrationprojects.
projects.This
Thisisisbecause
becausesequestration
sequestrationsites
siteswould
wouldlikely
likelybe
belocated
locatedfar
far
away from industrial sites, which would require the transportation of CO 2 , incurring additional
away from industrial sites, which would require the transportation of CO2 , incurring additional costs. costs.
InIncomparison,
comparison,the theutilization
utilizationcould
couldbebedone
doneatatthe
thesame
samesite
sitethe
thecapture
captureisisdone.
done.

Figure 13. Results of the case studies for the dry and the humid case.
Figure 13. Results of the case studies for the dry and the humid case.
When comparing cases B to E, it can be seen that developing sorbents with improved CO2 mass
transfer kinetics would have the biggest impact on reducing the cost (~27% reduction). With faster
When comparing cases B to E, it can be seen that developing sorbents with improved CO2 mass
kinetics, shorter tcycle could be achieved, which reduces the cost of capture. The thermal energy demand
transfer kinetics would have the biggest impact on reducing the cost (~27% reduction). With faster
can also be seen to reduce as hypothesized. However, case study C shows that the improvements to
kinetics, shorter 𝑡 could be achieved, which reduces the cost of capture. The thermal energy
the kinetics of the sorbent should not be made at the expense of the CO2 uptake capacity. Case study
demand can also be seen to reduce as hypothesized. However, case study C shows that the
D shows that combining faster sorbents with waste heat utilization could lower the cost of capture
improvements to the kinetics of the sorbent should not be made at the expense of the CO2 uptake
by ~54%. Finally, case study E shows that an 8% reduction in cost can be achieved by using less
capacity. Case study D shows that combining faster sorbents with waste heat utilization could lower
hygroscopic sorbents, which can be attributed to the reduced thermal energy requirement. Although
the cost of capture by ~54%. Finally, case study E shows that an 8% reduction in cost can be achieved
the costs for these case studies are still quite high, these give an indication of the relative impact each
by using less hygroscopic sorbents, which can be attributed to the reduced thermal energy
modification could have on the economics of the process.
requirement. Although the costs for these case studies are still quite high, these give an indication of
the relative impact each modification could have on the economics of the process.
6. Conclusions
A process model was proposed, for a DAC process employing a S-TVSA cycle, and validated with
6. Conclusions
laboratory experimental results. To evaluate the technoeconomic performance of the process, the model
A process model was proposed, for a DAC process employing a S-TVSA cycle, and validated
was subjected to a MOO with the objectives of minimizing the cost of capture and maximizing the
with laboratory experimental results. To evaluate the technoeconomic−1 performance of the process,
amount of CO2 captured. A minimum cost of capture of 612 USD tonne was calculated for a process
the model was subjected to a MOO with the objectives of minimizing the cost of capture and
with air entering at 25 ◦ C under dry conditions, and a cost of 657 USD tonne−1 was calculated for air
maximizing the◦ amount of CO2 captured. A minimum cost of capture of 612 USD tonne−1 was
entering at 22 C and 39% RH. The humid case yielded higher costs than the dry case, as an effect of
calculated for a process with air entering at 25 °C under dry conditions, and a cost of 657 USD tonne−1
the additional energy required to desorb the water that gets co-adsorbed on the sorbent. While the dry
was calculated for air entering at 22 °C and 39% RH. The humid case yielded higher costs than the
case yielded the lower cost, it is noted that the humid case is more realistic for practical DAC systems,
dry case, as an effect of the additional energy required to desorb the water that gets co-adsorbed on
and that the inlet air conditions correspond to that typical of subtropical climates. It was observed
the sorbent. While the dry case yielded the lower cost, it is noted that the humid case is more realistic
that the capture rate of the process could be increased (at the expense of a higher cost of capture) by
for practical DAC systems, and that the inlet air conditions correspond to that typical of subtropical
increasing the number of contactors, using more aggressive desorption conditions and changing the
climates. It was observed that the capture rate of the process could be increased (at the expense of a
higher cost of capture) by increasing the number of contactors, using more aggressive desorption
conditions and changing the cycle times. The process variables which had the most effect on the
results were identified as the steam flow rate and the number of contactors. The relatively higher
costs calculated here, in comparison to the results of prior studies, were identified to be a result of
the differences in the assumptions and scope of the respective studies, and the slower kinetics of the
Processes 2019, 7, 503 21 of 23

cycle times. The process variables which had the most effect on the results were identified as the steam
flow rate and the number of contactors. The relatively higher costs calculated here, in comparison to
the results of prior studies, were identified to be a result of the differences in the assumptions and
scope of the respective studies, and the slower kinetics of the sorbent evaluated in the current study.
It was identified that using a sorbent with two-fold faster kinetics could reduce the cost by ~27% and
that the utilization of waste heat could produce a ~42% reduction in cost. A combination of both would
allow the cost to be reduced by ~54%.
In summary, the process evaluated in this study does not appear to be economically feasible in
the current state. However, the study identified several avenues which could lead to improvements.
Of these, the improvements in sorbent design suggested here could realistically be achieved even
with the current state of the technology. However, further research is needed for areas such as the
development of low-cost contactors. If a combination of these improvements is achieved, this DAC
process has the potential to be economically feasible and a valuable tool for combatting climate change
in the future.

Supplementary Materials: The following are available online at http://www.mdpi.com/2227-9717/7/8/503/s1,

Section S1: Calculations for the unit operations, Section S2: Calculation of energy consumption, Section S3: Capital
cost estimation, Section S4: Estimation of cost of sorbent, Section S5: NPV analysis, S6: Desorption Stage: Model
Validation contd., S7: Sensitivity of the results to the MOO options
Author Contributions: Conceptualization, R.P.W., G.P.K., H.Y., A.F.A.H. and A.L.C.; methodology, R.P.W., G.P.K.,
H.Y., A.F.A.H. and A.L.C.; software, R.P.W.; validation, R.P.W., G.P.K., H.Y., A.F.A.H. and A.L.C.; investigation,
R.P.W.; resources, G.P.K., H.Y., A.F.A.H. and A.L.C.; writing—original draft preparation, R.P.W.; writing—review
and editing, R.P.W., G.P.K., H.Y., A.F.A.H. and A.L.C.; visualization, R.P.W.; supervision, G.P.K., H.Y., A.F.A.H.
and A.L.C.; project administration, A.F.A.H. and A.L.C.; funding acquisition, A.F.A.H. and A.L.C.
Funding: This research received no external funding.
Acknowledgments: The authors acknowledge Antonio Benci and David Zuidema, Monash Instrumentation
Facility, for assistance in designing and programming some of the componentry in the experimental set up.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. IPCC. IPCC Special Report on Carbon Dioxide Capture and Storage; Prepared by Working Group III of the
Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, UK; New York, NY,
USA, 2005.
2. Koytsoumpa, E.I.; Bergins, C.; Kakaras, E. The CO2 economy: Review of CO2 capture and reuse technologies.
J. Supercrit. Fluids 2018, 132, 3–16. [CrossRef]
3. Sanz-Pérez, E.S.; Murdock, C.R.; Didas, S.A.; Jones, C.W. Direct Capture of CO2 from Ambient Air. Chem. Rev.
2016, 116, 11840–11876. [CrossRef] [PubMed]
4. Sinha, A.; Darunte, L.A.; Jones, C.W.; Realff, M.J.; Kawajiri, Y. Systems Design and Economic Analysis of
Direct Air Capture of CO2 through Temperature Vacuum Swing Adsorption Using MIL-101(Cr)-PEI-800 and
mmen-Mg2 (dobpdc) MOF Adsorbents. Ind. Eng. Chem. Res. 2017, 56, 750–764. [CrossRef]
5. Krekel, D.; Samsun, R.C.; Peters, R.; Stolten, D. The separation of CO2 from ambient air—A techno-economic
assessment. Appl. Energy 2018, 218, 361–381. [CrossRef]
6. Zhang, W.; Liu, H.; Sun, C.; Drage, T.C.; Snape, C.E. Capturing CO2 from ambient air using a polyethyleneimine–
silica adsorbent in fluidized beds. Chem. Eng. Sci. 2014, 116, 306–316. [CrossRef]
7. Kulkarni, A.R.; Sholl, D.S. Analysis of Equilibrium-Based TSA Processes for Direct Capture of CO2 from Air.
Ind. Eng. Chem. Res. 2012, 51, 8631–8645. [CrossRef]
8. Gebald, C.; Repond, N.; Wurzbacher, J.A. Steam Assisted Vacuum Desorption Process for Carbon Dioxide
Capture. U.S. Patent No. US20170203249A1, 20 July 2017.
9. Wurzbacher, J.A.; Gebald, C.; Piatkowski, N.; Steinfeld, A. Concurrent Separation of CO2 and H2 O from Air
by a Temperature-Vacuum Swing Adsorption/Desorption Cycle. Environ. Sci. Technol. 2012, 46, 9191–9198.
[CrossRef]
Processes 2019, 7, 503 22 of 23

10. Climeworks. Our Technology. Available online: https://www.climeworks.com/our-technology/ (accessed on

14 June 2019).
11. Gobal Thermostat. A Unique Capture Process. Available online: https://globalthermostat.com/a-unique-
capture-process/ (accessed on 14 June 2019).
12. Kramer, D. Can carbon capture from air shift the climate change equation? Phys. Today 2018, 71, 26–29.
[CrossRef]
13. Carbon Engineering. Carbon Engineering: CO2 Capture and the Synthesis of Clean Transportation Fuels.
Available online: https://carbonengineering.com/ (accessed on 16 July 2019).
14. Keith, D.W.; Holmes, G.; St. Angelo, D.; Heidel, K. A Process for Capturing CO2 from the Atmosphere. Joule
2018, 2, 1573–1594. [CrossRef]
15. Jackson, R.B.; Baker, J.S. Opportunities and Constraints for Forest Climate Mitigation. Bioscience 2010, 60,
698–707. [CrossRef]
16. Wijesiri, R.P.; Knowles, G.P.; Yeasmin, H.; Hoadley, A.F.A.; Chaffee, A.L. CO2 Capture from Air Using
Pelletised Polyethyleneimine Impregnated MCF Silica. Ind. Eng. Chem. Res. 2019, 58, 3293–3303. [CrossRef]
17. Wijesiri, R.P.; Knowles, G.P.; Yeasmin, H.; Hoadley, A.F.A.; Chaffee, A.L. Desorption Process for Capturing
CO2 from Air with Supported Amine Sorbent. Ind. Eng. Chem. Res. 2019. [CrossRef]
18. Glueckauf, E.; Coates, J.I. 241. Theory of chromatography. Part IV. The influence of incomplete equilibrium
on the front boundary of chromatograms and on the effectiveness of separation. J. Chem. Soc. 1947, 315–1321.
[CrossRef]
19. Sircar, S.; Hufton, J.R. Why Does the Linear Driving Force Model for Adsorption Kinetics Work? Adsorption
2000, 6, 137–147. [CrossRef]
20. Do, D.D. Adsorption Analysis: Equilibria and Kinetics; Imperial College Press: London, UK, 1998.
21. Knowles, G.P.; Chaffee, A.L. Shaped Silica-polyethyleneimine Composite Sorbents for CO2 Capture via
Adsorption. Energy Procedia 2017, 114, 2219–2227. [CrossRef]
22. Gebald, C.; Piatkowski, N.; Rüesch, T.; Wurzbacher, J.A. Low-Pressure Drop Structure of Particle Adsorbent
Bed for Adsorption Gas Separation Process. Patent No. WO2014170184A1, 23 October 2014.
23. Couper, J.R. Chemical Process Equipment: Selection and Design, 2nd ed.; Elsevier/Butterworth-Heinemann:
Amsterdam, The Netherlands, 2010.
24. Antoine, L.C. Tensions des vapeurs; nouvelle relation entre les tensions et les températures Comptes Rendus.
Comptes Rendus des Séances de l’Acad. des Sci. 1888, 107, 681–836.
25. Seider, W.D. Product and Process Design Principles: Synthesis, Analysis, and Evaluation, 3rd ed.; John Wiley:
Hoboken, NJ, USA, 2009.
26. Sinnott, R.K. Coulson & Richardson’s chemical engineering. In Chemical Engineering Design, 4th ed.; Elsevier
Butterworth-Heinemann: Oxford, UK, 2005; Volume 6.
27. Brennan, D. Process Industry Economics: An International Perspective; IChemE: Rugby, UK, 1998.
28. IEA Solar Heating & Cooling Programme. Solar Heat Worldwide 2018; International Energy Agency: Paris,
France, 2018.
29. IRENA. Renewable Power Generation Costs in 2017; International Renewable Energy Agency: Abu-Dhabi,
UAE, 2018.
30. South East Water Corporation. 2018-19 Pricing Handbook; South East Water Corporation: Melbourne,
Australia, 2018.
31. Deb, K. Multi-Objective Optimization Using Evolutionary Algorithms, 1st ed.; John Wiley & Sons: New York,
NY, USA, 2001.
32. Jahandar Lashaki, M.; Khiavi, S.; Sayari, A. Stability of amine-functionalized CO2 adsorbents: A multifaceted
puzzle. Chem. Soc. Rev. 2019, 48, 3320–3405. [CrossRef]
33. Bajamundi, C.J.E.; Koponen, J.; Ruuskanen, V.; Elfving, J.; Kosonen, A.; Kauppinen, J.; Ahola, J. Capturing
CO2 from air: Technical performance and process control improvement. J. CO2 Util. 2019, 30, 232–239.
[CrossRef]
34. Wang, J.; Huang, H.; Wang, M.; Yao, L.; Qiao, W.; Long, D.; Ling, L. Direct Capture of Low-Concentration
CO2 on Mesoporous Carbon-Supported Solid Amine Adsorbents at Ambient Temperature. Ind. Eng. Chem.
Res. 2015, 54, 5319–5327. [CrossRef]
35. Goeppert, A.; Meth, S.; Prakash, G.K.S.; Olah, G.A. Nanostructured silica as a support for regenerable
high-capacity organoamine-based CO2 sorbents. Energy Environ. Sci. 2010, 3, 1949–1960. [CrossRef]
Processes 2019, 7, 503 23 of 23

36. Sakwa-Novak, M.A.; Tan, S.; Jones, C.W. Role of Additives in Composite PEI/Oxide CO2 Adsorbents:
Enhancement in the Amine Efficiency of Supported PEI by PEG in CO2 Capture from Simulated Ambient
Air. ACS Appl. Mater. Interfaces 2015, 7, 24748–24759. [CrossRef]
37. Choi, S.; Gray, M.L.; Jones, C.W. Amine-Tethered Solid Adsorbents Coupling High Adsorption Capacity and
Regenerability for CO2 Capture from Ambient Air. ChemSusChem 2011, 4, 628–635. [CrossRef]
38. Lu, W.; Sculley, J.P.; Yuan, D.; Krishna, R.; Zhou, H.C. Carbon Dioxide Capture from Air Using Amine-Grafted
Porous Polymer Networks. J. Phys. Chem. C 2013, 117, 4057–4061. [CrossRef]
39. Chen, Z.; Deng, S.; Wei, H.; Wang, B.; Huang, J.; Yu, G. Polyethylenimine-Impregnated Resin for High CO2
Adsorption: An Efficient Adsorbent for CO2 Capture from Simulated Flue Gas and Ambient Air. ACS Appl.
Mater. Interfaces 2013, 5, 6937–6945. [CrossRef]
40. Darunte, L.A.; Oetomo, A.D.; Walton, K.S.; Sholl, D.S.; Jones, C.W. Direct Air Capture of CO2 Using Amine
Functionalized MIL-101(Cr). ACS Sustain. Chem. Eng. 2016, 4, 5761–5768. [CrossRef]
41. Goeppert, A.; Zhang, H.; Czaun, M.; May, R.B.; Prakash, G.K.S.; Olah, G.A.; Narayanan, S.R. Easily
Regenerable Solid Adsorbents Based on Polyamines for Carbon Dioxide Capture from the Air. ChemSusChem
2014, 7, 1386–1397. [CrossRef]
42. Zhang, H.; Goeppert, A.; Olah, G.A.; Prakash, G.K.S. Remarkable effect of moisture on the CO2 adsorption
of nano-silica supported linear and branched polyethylenimine. J. CO2 Util. 2017, 19, 91–99. [CrossRef]
43. Didas, S.A.; Kulkarni, A.R.; Sholl, D.S.; Jones, C.W. Role of Amine Structure on Carbon Dioxide Adsorption
from Ultradilute Gas Streams such as Ambient Air. ChemSusChem 2012, 5, 2058–2064. [CrossRef]
44. Sehaqui, H.; Gálvez, M.E.; Becatinni, V.; cheng Ng, Y.; Steinfeld, A.; Zimmermann, T.; Tingaut, P. Fast and
Reversible Direct CO2 Capture from Air onto All-Polymer Nanofibrillated Cellulose—Polyethylenimine
Foams. Environ. Sci. Technol. 2015, 49, 3167–3174. [CrossRef]

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).