You are on page 1of 10

Energy &

Environmental Science
View Article Online
PAPER View Journal | View Issue

Improving the regeneration of CO2-binding organic


Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

liquids with a polarity change†


Cite this: Energy Environ. Sci., 2013, 6,
2233
Paul. M. Mathias,a Kash Afshar,a Feng Zheng,b Mark D. Bearden,b Charles J. Freeman,b
Tamer Andrea,c Phillip K. Koech,b Igor Kutnyakov,b Andy Zwoster,b Arnold R. Smith,a
Philip G. Jessop,c Omid Ghaffari Nikc and David J. Heldebrant*b

This paper describes a solvent regeneration method unique to CO2-binding organic liquids (CO2BOLs) and
other switchable ionic liquids: utilizing changes in polarity to shift the free energy of the system. The
degree of CO2 loading in CO2BOLs is known to control the polarity of the solvent; conversely, polarity
can be exploited as a means to control CO2 loading. In this process, a chemically inert nonpolar
“antisolvent” (AS) such as hexadecane (C16) is added to aid in de-complexing CO2 from a CO2-rich
CO2BOL. The addition of this polarity assist reduces the temperature required for regeneration of our
most recent CO2BOL, 1-((1,3-dimethylimidazolidin-2-ylidene)amino)propan-2-ol by as much as 73  C. The
lower regeneration temperatures realized with this polarity change allow reduced solvent attrition and
thermal degradation. Furthermore, the polarity assist shows considerable promise for reducing the
regeneration energy of CO2BOL solvents, and separation of the CO2BOL from the AS is as simple as a
cooling the mixture to promote phase separation. Based on vapor–liquid and liquid–liquid equilibrium
measurements of a candidate CO2BOL with CO2, with and without an AS, we present the evidence and
impacts of a polarity change on a CO2BOL. Equilibrium thermodynamic models and analysis of the
Received 25th March 2013
Accepted 30th May 2013
system were constructed using Aspen Plus, and forecasts of preliminary process configurations and
feasibility are also presented. Lastly, projections of solvent performance for removing CO2 from a
DOI: 10.1039/c3ee41016a
subcritical coal-fired power plant (total net power and parasitic load) are presented with and without
www.rsc.org/ees this polarity assist and compared to the U.S. Department of Energy's Case 10 monoethanolamine baseline.

Broader context
To combat climate change, new technologies must be developed to efficiently and economically remove CO2 from ue gas. Traditional water-based amine
systems are the only proven technology to date and will likely represent the rst generation of CO2 capture technologies, but the energy cost of CO2 removal is
very high. We present in this experimental and modeling study the application of our anhydrous CO2BOLs solvent technology in a way that exploits a polarity
change. There are two benets from CO2BOLs technology: (1) the low water level in the circulating solvent vastly reduces the energy cost due to vaporizing water;
and (2) the use of an antisolvent signicantly reduces the stripper reboiler temperature, which opens up the possibility of novel stripper technology
enhancements, namely reduced thermal degradation of CO2BOL, lower fugitive emissions, and potential for higher net power produced. Our analysis suggests
that CO2BOLs solvent technology has potential to be an efficient CO2 separating solvent system with estimated parasitic loads one third less than conventional
aqueous amine processes. While further solvent development is necessary to make the technology viable, the initial projected energy savings indicate
considerable promise for this second-generation of carbon capture solvents.

Introduction power plant ue gas with a solution of aqueous amines, and
subsequent regeneration of the amine solution is commercially
Considerable effort has been applied to removal of CO2 from available and effective.1–3 However, this method consumes a
coal-red power plant ue gas, which is a primary source of substantial amount of steam from the power plant steam cycle
anthropogenic CO2. Currently, absorption of CO2 from the and can reduce the net power from the plant by as much
as 29%.4
a
Fluor Corporation, 3 Polaris Way, Aliso Viejo, CA, 92628, USA. E-mail: Paul.M. The primary advantage of using all-organic solvents to sepa-
Mathias@Fluor.com; Tel: +1-949-349-3595 rate CO2 from power plant ue gas is to exploit the potentially
b
Battelle Pacic Northwest Division, Richland, WA, 99352, USA. E-mail: David. lower heat required for regeneration compared to water-based
Heldebrant@PNNL.gov; Fax: +1-509-375-2186; Tel: +1-509-372-6359
c
solvents. Numerous classes of “next generation” water-free
Queens University, 90 Bader Lane, Kingston, Ontario, K7L 4N3 Canada
solvents are gaining attention for CO2 separations, assuming
† Electronic supplementary information (ESI) available: NMR spectra of LLE
measurements and let-down turbine conguration. See DOI: 10.1039/c3ee41016a
water tolerance can be proven. Current advances in water-free

This journal is ª The Royal Society of Chemistry 2013 Energy Environ. Sci., 2013, 6, 2233–2242 | 2233
View Article Online

Energy & Environmental Science Paper

systems include conventional ionic liquids (ILs),5 amines in ILs,6


task-specic ILs (TSILs),5,7–9 nanoparticle organic hybrid mate-
rials (NOHMs),10 phase changing solvents,11 and alkanolamines
and hindered amines in organic co-solvents.12–14 Keeping with
this trend, a class of switchable ILs, originally developed by Jes-
sop et al.,15 was specically modied for CO2 capture.16 This
particular class, referred to as CO2-binding organic liquids
(CO2BOLs) acts like aqueous amines, but uses alcohols in place
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

of water and a non-nucleophilic base in place of nucleophilic Fig. 1 Proposed reaction between CO2 and 1-((1,3-dimethylimidazolidin-2-yli-
primary and secondary amines. The CO2BOL base does not dene)amino)propan-2-ol, (A).
directly react with CO2, as do amines that form carbamates.
Rather, the alcohol group in the CO2BOL reacts with CO2 to rst
form an alkylcarbonic acid, which then protonates the base to undergo minor polarity changes, which are not estimated to be
form a liquid alkylcarbonate.16 Some alkylcarbonate salts have signicant enough to reverse CO2 loading.
CO2 capacities as high as 20 wt% at 1 atm CO2.15 The resulting The rst example of a polarity swing used to decomplex
polar ILs can revert to the less-polar nonionic solvent by thermal chemically bound gases from a capture solvent was demon-
removal of CO2.15–17 strated with a solvent developed for hydrogen sulde capture
Recent advances in switchable ILs for CO2 capture include (H2SBOL). Here, adding a nonpolar chemically inert hexane
different chemistries and improvements in material perfor- “antisolvent” (AS) to the H2S-loaded solvent drove the removal of
mance. Other derivatives of switchable ILs were developed by 97% of the bound H2S from the H2SBOL at 25  C.27 The degree of
Eckert and Liotta et al. using liquid carbamate salts from the H2S release was dependent on the AS molar loading as well as its
reaction of CO2 with siloxated amines.18,19 Other advances in miscibility with the BOL. CO2BOLs exhibit behavior similar to
switchable alkylcarbonate ILs, notably reducing the vapor H2SBOLs but require slightly higher temperature for full CO2
pressure of superbases and alcohols (to reduce evaporative release due to the higher binding strength of CO2 than H2S.
losses), have been achieved by conjoining the alcohol and base Polarity-swing assisted regeneration (PSAR) primarily shis
constituents into one molecule, a strategy employed by us20 and the free energy of the reaction but also may impact the kinetics
by Park and Kim.21 Dai et al. has shown two strategies to reduce of CO2 release. Conceptually the PSAR can be considered “self-
volatility of alkylcarbonate-based systems by making them actuating;” mild heating induces CO2 release from the CO2BOL,
completely ionic (thus removing the switch between ionic and making it less polar, but the temperature increase also
nonionic) by pairing superbases with alcohol-functionalized ILs increases the miscibility of the AS in the CO2BOL, which in turn
or by pairing highly acidic alcohols with superbases to form forces more CO2 release. One would expect the nonpolar AS to
protic ILs (PILs).22,23 destabilize the ionic form of a CO2BOL (the alkylcarbonate) and
The rst-generation CO2BOLs combined an individual base shi the equilibrium to the le (Fig. 2), in turn releasing CO2.
and an alcohol. A drawback of this dual-component system is While AS is present, the CO2 release is still equilibrium
the relatively high volatility of the individual constituents, controlled at a given temperature, but the equilibrium favors
which severely hinders the industrial applicability because of lower CO2 loading than it would without AS. Thus, the AS aids in
solvent losses in the absorber distillate. Similar to alkanol- CO2 release from the CO2BOL, allowing regeneration at
amines, the second-generation CO2BOL candidates utilized temperatures as much as 73  C lower than when using a
alkanolamidine and alkanolguanidine bases, which reacted conventional thermal swing.
with CO2 to form zwitterionic liquids upon carboxylation.20 Fig. 3 shows a high-level schematic of a process envisioned
Unfortunately, such molecules based on amidine cores such as for CO2BOLs with the added PSAR. The incoming ue gas is
(diazabicyclo[5.4.0]-undec-7-ene) or guanidine cores such as cooled to promote condensation of water to reduce the water
1,1,3,3-tetramethylguanidine, while nonvolatile, exhibited
limited CO2 uptake compared to their theoretical maximums
due to their high viscosity.24 Aer a number of development
variations, a low-viscosity alkanolguanidine (A), based on a
cyclic guanidine derived from 1,3-dimethylimidazolidine, was
discovered.25 This second-generation CO2BOL (A) reacts like the
rst-generation version and can be thermally reversed, but as a
zwitterionic guanidinium alkylcarbonate IL (Fig. 1).
One unique property of CO2BOLs and other switchable ILs is
the fundamental polarity change when CO2 is chemically
bound.15,17,26 Based on this, we hypothesized that CO2 loading
could be inuenced by an external polarity change. Note that
aqueous amines do not exhibit polarity shis with CO2 loading, Fig. 2 Conceptual “switching” of (A) polarity as a function of CO2 loading using
nor do amines dissolved in organic solvents. No reported a solvatochromatic scale (lmax, nm) of nile-red dye. Increasing polarity is nonpolar
polarity changes are reported for conventional TSILs, and PILs22 (green, left) to polar (pink, right).26,28,29

2234 | Energy Environ. Sci., 2013, 6, 2233–2242 This journal is ª The Royal Society of Chemistry 2013
View Article Online

Paper Energy & Environmental Science


Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

Fig. 3 Conceptual absorption and polarity-swing-assisted regeneration process.


Fig. 5 Gas–liquid isotherms for (A) and CO2 with one molar equivalent of water.
(A–D corresponds to run number).

accumulation into the CO2BOL solvent. The CO2 is chemically


reacted with the CO2BOL in the absorber, aer which the CO2- Previous studies conrmed that (A) had high water tolerance
rich CO2BOL solvent is pumped to a regeneration column. A with respect to viscosity (<10% increase) and the resulting
non-polar AS is added to the regeneration column to decrease solution was a single phase and there is no evidence of bicar-
the bulk polarity of the CO2BOL aiding in decomplexation of the bonate precipitation.25 However, VLE data was needed to
CO2 (with mild heating). Aer the CO2 is released, it is delivered conrm water's impact on CO2 uptake. Therefore, VLE data was
for compression, while the AS and CO2BOL are separated by a also collected with one molar equivalent of water (10 wt%)
phase separation promoted by cooling in a coalescing tank. added to the solvent, as a worst-case scenario (Fig. 5) (at load-
Aer separation, both the AS and CO2BOL are recirculated and ings of >10 wt% water the CO2BOL becomes the solute and not
the process continues with the CO2BOL pumped back to the the solvent, and thus the CO2BOL would act like any aqueous
absorber and the AS delivered back to the regeneration column. amine system and no PSAR could occur). Here, the higher total
We describe here in detail the science and engineering required pressure is due to the high volatility of water (i.e. relatively high
to achieve this envisioned conguration. vapor pressure). The loading values in Fig. 5 were based on the
total amount of CO2 injected. Actual solvent loading was slightly
Results and discussion lower, as some injected CO2 remained in the gas phase. The
drop in total pressure as CO2 is loaded into the CO2BOL
VLE/VLLE behavior (PTx method)
corresponds to the consumption of water to form either a
To evaluate the PSAR impacts on CO2BOL performance, the bicarbonate salt (similar to aqueous amine systems) or a water-
vapor–liquid/vapor–liquid–liquid equilibrium (VLE/VLLE) solvated alkylcarbonate; however, no solid precipitation or
behavior of a representative CO2BOL had to rst be quantied. phase separation was observed in the liquid phase.25
All VLE measurements for a representative CO2BOL (A) were Isotherm measurements were also performed with hexadecane
collected using the total-pressure method (PTx) and CO2 load- (C16) AS to determine the effect of the PSAR on equilibrium
ings are reported in moles CO2 per mole BOL. Fig. 4 shows the loading of CO2 (Fig. 6). C16 was chosen due to its low cost and low
VLE results for the tested 1-((1,3-dimethylimidazolidin-2-yli- vapor pressure. The data in Fig. 6 shows little to no PSAR effect at
dene)amino)propan-2-ol (A) only (no water or AS additions). 40 or 60  C, which is consistent with the immiscibility of the
Here the total headspace pressure corresponds to the equilib- CO2BOL with the C16 phase as observed in the VLLE measure-
rium partial pressure of CO2 (P) because of the low vapor ments. This is advantageous because as the solvent is continu-
pressure contribution from the solvent. The VLE curves were ously circulated between absorption and desorption, any AS that
run in duplicate at 40, 60, 80 and 100  C. (A) shows markedly would carry over back into the absorber (CO2 absorption column)
greater CO2 uptake compared to previous CO2BOLs, indicating would have minimal impact on absorber performance. There was,
a stronger reaction with CO2. The heat of solution calculated however, a noticeable increase in CO2 partial pressure over the
from the data is on the order of 80 kJ mol1 of CO2.

Fig. 4 Gas–liquid isotherms for (A) and CO2 without water or AS additions (A–D Fig. 6 Gas–liquid isotherms for (A) and CO2 plus 1 molar equivalent of C16 AS.
corresponds to run number). (A–D corresponds to run number).

This journal is ª The Royal Society of Chemistry 2013 Energy Environ. Sci., 2013, 6, 2233–2242 | 2235
View Article Online

Energy & Environmental Science Paper

CO2BOL at 80  C, where the phases became miscible. It should be


noted that the CO2 loadings in the PTx measurements were lower
than those in the miscibility tests indicating that the AS was more
miscible at lower CO2 loadings. In all measurements, cooling to 40

C promoted phase separation between the C16 and the CO2-rich
BOL, supporting a straightforward approach of AS separation
from the CO2BOL in a full-scale process.
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

Separation of CO2BOL and antisolvent


In order to prevent biphasic conditions inside the absorber
column, AS carryover to the absorber needs to be minimized.
Currently two strategies are envisioned for removing the AS Fig. 7 Phase diagram for CO2-lean (A)/decane (O), CO2-lean (A)/hexadecane
(B), CO2-rich (A)/decane (:), partially carbonated (A) (20% mol CO2)/hex-
from the lean CO2BOL (A) aer it leaves the reboiler (regener-
adecane (C) and CO2-free (A)/hexadecane () as a function of temperature.
ation column): liquid–liquid partitioning and liquid–solid par-
titioning. Simply carrying over the AS into the absorber is
another option. upper C16 phase, but only slightly decreases the carryover of AS
Miscibility tests were performed on multiple hydrocarbon- in the recovered (A) stream (6 wt% C16 instead of 8 wt%).
based ASs (Table 1). Here, the miscibility temperature for Process simulations take into consideration an estimated
mixtures of 1 : 1 mole ratio of lean (A) and solvent (heptane, carryover of 0.3 wt% of AS into the absorber.
decane (C10), dodecane or C16), was found to be linked to the Similar tests were performed with C10 as the AS and lean (A)
chain length of AS—the upper critical solution temperature (Fig. 7, blue O). CO2-lean (A) and C10 are miscible in all
decreased with shorter chains. This nding allows candidate proportions above 40  C, but at 25  C or below they split into a
ASs to be selected that will promote phase separation upon lower liquid phase ((A)-rich phase: 81% (A) and 19% C10) and
cooling to projected operating temperatures. For the initial an upper liquid phase (AS-rich phase: 16% (A) and 84% C10).
vapor–liquid–liquid equilibrium (VLLE) and liquid–liquid Therefore, choosing C10 as the AS causes more AS carryover in
equilibrium (LLE) measurements, C16 was chosen as the most the recovered (A).
viable AS due to its low vapor pressure. However, C16 carryover When CO2 is added to the (A)–AS mixture, the phase
does not help lower the viscosity of (A) at high CO2 loadings. behavior changes dramatically. Partial carbonation of the
Therefore, further testing is required to include viscosity attri- compound (20 mol% which we dene as “partially carbonated”,
butes in AS evaluations and selection. It should be noted that Fig. 7, red C) dropped the amount of C16 in the lower liquid
hydrocarbon-based ASs were tested as a proof of concept of the phase at 40  C from 8 wt% to less than 1.5 wt% as determined
PSAR, however environmentally benign or “greener” AS are a by gas chromatography with thermal conductivity detection
focus of current study. (GC-TCD), and increased the amount of C16 in the upper liquid
Fig. 7 shows the phase diagram for C16 or C10 and “lean” or phase from 78 wt% to 91 wt%. CO2-saturated “rich” (A) showed
“rich” (A). In this experiment, “lean” (A) contains 4 mol% of the very low miscibility with C10 (Fig. 7, blue :); two phases are
carbonated species and 96 mol% of the CO2-free species as observed even at 90  C and the CO2-rich (A) did not completely
determined by proton nuclear magnetic resonance (1H-NMR) dissolve in C10 in the temperature range that our controlled
(Fig. S1†). Complete exclusion of CO2 by high vacuum and temperature bath allows (<95  C). At lower temperatures, the
heating for more than 4 h did not show signicant change in addition of CO2 to the (A)–C10 mixture expands the two-phase
the phase behavior (Fig. 7, red 7). Fig. 7 red B indicates region of the phase diagram, so that the C10 carryover in the (A)-
miscibility of the C16 and lean-(A) above 62  C, and sponta- rich phase is as low as 4%. The high viscosity and the two-phase
neous separation of the mixture into two liquid phases below behavior at high temperatures are only observed at CO2 satu-
that temperature. For example, if the lean-(A)–C16 mixture is ration (estimated stripper conditions). Note that some AS
cooled to 40  C, the lower liquid phase (le side of diagram) will carryover in the recovered (A) could help lower the solvent
contain 92 wt% (A) and 8 wt% C16. The upper liquid phase viscosity.
(right in the diagram) will contain 22 wt% (A) and 78 wt% C16 C16 was the preferred AS for this study due to its low vapor
AS. Cooling to 20  C ensures that more (A) is recovered from the pressure compared to C10, however C16 will be unlikely to be
used industrially due to environmental concerns. Preferred ASs
would be biodegradable, non-toxic, and cheap and have low
Table 1 Temperatures of miscibility for 50 mol% mixtures of various hydro- ammability. The search for more environmentally benign ASs
carbon-based ASs in CO2-lean (A) is currently a focus of study in our laboratory.

Antisolvent Chain length Tmiscibility ( C)


Thermodynamic model
Heptane 7 > 30
Decane 10 38 A thermodynamic model was developed based on consideration
Dodecane 12 39.3 of the solution chemistry and used the four types of PTx data
Hexadecane 16 62 (herein BOL is dened as CO2-free, and CO2BOL is CO2-rich):

2236 | Energy Environ. Sci., 2013, 6, 2233–2242 This journal is ª The Royal Society of Chemistry 2013
View Article Online

Paper Energy & Environmental Science

 absorption of CO2 on the neat BOL.


 effect of water on the absorption of CO2 on BOL.
 effect of the AS on the absorption of CO2 on BOL. It should
be noted that the BOL-AS system has a region of liquid–
liquid immiscibility that is affected by the addition of CO2.
 quaternary data with all four principal components (BOL,
CO2, water and AS) present.
Special attention was devoted to ensuring that the thermo-
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

dynamic model captures the key effects of the solvent system


(BOL, water and AS). The rst effect is the complexation of CO2
and BOL, which is enhanced by addition of water and dimin-
ished when AS is added. Due to the thermodynamic nonideality
Fig. 8 Estimated CO2 partial pressures in BOL–H2O–C16 at 100  C. The relative
of the system, the ElecNRTL-RK model in Aspen Plus was
solvent concentrations are equimolar.
chosen, together with the chemistry capability. The chemistry
capability enables the effect of chemical reactions (chemical
complexation) to be captured in the thermodynamic model. The
ElecNRTL-RK model uses Chen's electrolyte activity-coefficient
model and the Redlich–Kwong equation of state to describe the
vapor phase. Further details are available in documentation
from Aspen Technology Inc.
The chemistry model for the CO2BOL-PSAR system is based
on the understanding of the fundamental chemistry, as well as
practical experience with electrolyte process simulation. When
tting the data, it was found that a zwitterionic guanidinium
alkylcarbonate (B) (Fig. 1) with a 1 : 1 stoichiometry of BOL : CO2
t the data more closely than a conventional 2 : 1 ionic complex
with a separated guanidinium cation and alkylcarbonate anion.
Thus, for the model, two complexes are postulated that result in
the following two chemical-equilibrium equations: Fig. 9 Fraction of CO2 in BOL–H2O–C16 at 100  C that is complexed. The relative
solvent concentrations are equimolar.

2CO2 + 2BOL 4 BOLCO2+ + BOLCO2 (1)


The enthalpy of solution of CO2 in the solvent is extremely
+
H2O + BOL + CO2 4 BOLH + HCO3 (2) important because the paramount goal of CO2 capture processes
is to reduce the energy cost of CO2 capture and this energy cost is
Breaking out separate charges for the zwitterionic CO2BOL- directly related to the enthalpy of solution. For reasons discussed
CO2 ionic species (BOLCO2 alkylcarbonate anion and in a later section, the enthalpy of solution of most interest is that
BOLCO2+ guanidinium cation) was needed for Aspen Plus to with BOL–H2O mixtures as the solvent, and these are presented
enable the Born term, which accounts for the effect of the ionic next. Fig. 10 shows the enthalpy of solution of CO2 in BOL as a
strength and the solvent dielectric constant.30 This term is vital function of temperature and CO2 loading. At 40  C (typical
in predicting the effect of the low-dielectric-constant AS to absorber temperature), the enthalpy of solution is about 82 kJ
reduce the tendency of the mixed solvent to complex CO2. mol1 at low CO2 loadings and then decreases in magnitude with
The resulting model quantitatively describes the data and CO2 loading. The model accounts for expected behavior; as the
captures the relative solubility of CO2 in the system as the
composition of the BOL–H2O–C16 solvent changes. Fig. 8
presents the CO2 calculated partial pressure at 100  C for four
representative solvent compositions. The composition on the x-
axis is the CO2 loading with respect to the BOL. Fig. 8 shows that
adding 1 mole of H2O to 1 mole of BOL signicantly reduces the
CO2 partial pressure, while adding 1 mole of C16 to 1 mole of
BOL sharply increases the CO2 partial pressure. The 1 : 1 : 1
BOL–H2O–C16 solvent has an absorption strength between
those of neat BOL and 1 : 1H2O : BOL. Fig. 9, which presents the
fraction of CO2 that is complexed, provides an explanation of
the results in Fig. 8. The reduction of the CO2 partial pressure is
caused by increased complexation of the CO2, and conversely,
the increase of the partial pressure is due to decomplexation of
CO2 from the BOL. Fig. 10 Calculated enthalpy of solution of CO2 in BOL.

This journal is ª The Royal Society of Chemistry 2013 Energy Environ. Sci., 2013, 6, 2233–2242 | 2237
View Article Online

Energy & Environmental Science Paper


Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

Fig. 11 Calculated enthalpy of solution of CO2 in 1 : 1H2O : BOL.

temperature increases, the magnitude of the enthalpy of solution Fig. 13 Flow diagram of the CO2BOL/PSAR process showing heat and material
decreases, as does the CO2 loading at which the enthalpy of flows to a subsection containing the stripper, cross exchanger and coalescer. The
solution drops. Fig. 11 presents the calculated CO2 enthalpy of rich stream (SR) flows into the section, and the lean solvent (SL) and CO2 product flow
out of the section. The heat flows QR and QC are the reboiler and condenser duties.
solution in the 1 : 1H2O : BOL solvent. The difference here is that
the enthalpy of solution rst increases with CO2 loading, and
may decrease if the temperature is sufficiently high (say, $80  C).
Fig. 13 is a simplied representation of the material and heat
ows to/from a owsheet subsection (within the dashed square)
Process concept
containing the stripper, cross exchanger and coalescer. The
Fig. 12 shows a high-level process ow diagram of the CO2BOL representation is simplied because it neglects any make-up
process using PSAR. The envisioned process would operate solvents, and assumes that the coalescer removes all AS from
similarly to a conventional aqueous amine system, with the the rich solvent leaving the stripper. Note that this gure also
main difference being the introduction of an AS to aid desorp- does not include equipment such as a lean cooler to control
tion of CO2 from the rich solvent. As with other solvents that coalescer temperature. The purpose of the simplied analysis is
work on chemical absorption, the incoming ue gas is cooled simply to understand the dominant energy effects of the
and conditioned in a direct contact cooler using process cooling process. A more detailed process analysis will be performed
water. The CO2 is absorbed at about 40  C in the lean CO2BOL later to determine a suitable process conguration.
solvent, where “lean” refers to the low CO2 content of the stream Fig. 13 indicates that adding antisolvent does not directly
leaving the stripper. Aer exiting the absorber, the rich CO2BOL affect the reboiler duty, QR, since the AS is entirely within the
solvent (“rich” refers to the low CO2 content of the stream envelope chosen. eqn (3) presents the energy balance for the
exiting the absorber) is mixed with the AS, heated in the cross subsystem.
exchanger to improve energy efficiency, and then pumped to a
regenerator (e.g., a stripper) column. The AS changes the HSR  HSL  HCO2 + QR  QC ¼ 0 (3)
polarity of the rich CO2BOL solution, which destabilizes the
chemically bound CO2 so that release can occur at lower QR ¼ (HSL + HCO2  HSR) + QC (4)
temperatures. The hot rich solution exchanges heat with the
lean solution to recover heat, and then is separated into two Note that both QC and QR are positive in the present analysis
liquid phases in a coalescer. Similar analysis using the Kirk- since their ow direction has been taken into account. The rst
wood–Onsager approach will be the focus of future studies. term on the right side of eqn (4) is the integral heat of solution
from adding CO2 to the lean solvent as well as sensible heat
effects since the lean solvent is warmer than the rich solvent,
and the temperature difference is the temperature approach in
the cross exchanger.
Any condenser duty increases the reboiler duty, and, as
noted above, QC is strongly related to the condensation of water;
reduced water ows in the reboiler will reduce QC and conse-
quently reduce QR.
With this analysis as background, we can now propose ways
in which the AS can reduce the heat rate of the process (QR)
relative to the process without the AS.
The AS releases CO2 from the rich solvent, and hence the
lean loading of CO2 may be reduced. For the same rich loading,
the BOL ow rate could be reduced.
The present analysis does not directly consider the effect of
Fig. 12 Block flow diagram for the CO2BOL/PSAR process. water. It is known from the measured absorption data that

2238 | Energy Environ. Sci., 2013, 6, 2233–2242 This journal is ª The Royal Society of Chemistry 2013
View Article Online

Paper Energy & Environmental Science

water makes BOL a stronger solvent for CO2. Since the use of an any signicant impact on solution polarity. As more AS is
AS reduces the lean CO2 loading, it is possible that for a spec- added, its mole fraction is high enough to decrease the solution
ied BOL rate, the water level in the process may be reduced. polarity and reduce the reboiler temperature, but the effect of
Water is the main contributor to QC, so a lower water ow rate the specic heat still appears to overcome any net reduction in
will reduce QR at a given BOL ow rate. heat rate. In summary, the ow of AS has a relatively small effect
Another way of viewing the AS effect is that it drops the on the calculated heat rate (10–15%), but a signicant effect on
required reboiler temperature. Given the present analysis, a reducing the reboiler temperature (up to 73  C).
lower reboiler temperature does not directly reduce QR, but a One reason for the small heat rate reduction with PSAR has
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

lower temperature will increase the applicability of advances in to do with the large molecular weight of C16 as an AS (224 g
the use of steam to supply QR, and thus to increase process mol1) compared to the mass of the CO2BOL (171 g mol1). If
efficiency. The steam used to supply the reboiler duty QR now lighter ASs, such as C10, or solid ASs are utilized, the reboiler
may be expanded to lower pressure, thus generating more heat rate could be signicantly reduced. The initial AS studied
power before proceeding (still at a sufficient temperature) to (C16) also has a high upper critical solution temperature (62  C
drive the stripper reboiler. CO2 free, and 120  C CO2 loaded), requiring higher tempera-
As noted above, this simplied analysis is intended to be tures to make the C16 miscible with the CO2-rich CO2BOL. If
complementary to detailed Aspen Plus simulations. more miscible ASs such as C10 (critical solution temperature
38  C CO2 free and 92  C CO2 rich) are used, the regeneration
Full-scale projections of reboiler duty temperature is projected to decrease compared to the C16 case.
If full optimization of ASs was performed, a minimum regen-
Once the thermodynamic model was completed, process
eration temperature $65  C is expected, as indicated by critical
simulations were run to estimate the energy impacts of the
solution temperatures of ASs.
CO2BOL-PSAR system. The Aspen Plus simulations were aimed
Simulations with water added to the BOL showed signicant
at understanding the effect of AS concentration on the stripping
increases in reboiler heat rates (Fig. 14). However, this increase
conditions and duties. A key assumption in the simulations was
must be balanced with the energy required to remove the water
that the rich solvent viscosity was 20 cP – a development target
upstream, and requires further process optimization.
for the solvent. Further, mass transfer and kinetics were not
considered, i.e. equilibrium columns were used to model the
Absorber and Stripper. In these simulations, the molar ratio of Benet of lower regeneration temperatures available through
AS to BOL in the stripper rich-solvent feed was varied from no PSAR
AS to 2 moles of AS per mole of BOL. The results indicated a There are three primary benets to reducing the regeneration
wide variation of reboiler temperatures and heat rates temperature of the CO2BOL solvent: reduced thermal degrada-
depending on the specied molar ratio of AS to BOL. Fig. 14 tion of CO2BOL, lower fugitive emissions and potential for
shows a plot of the predicted heat rates and reboiler tempera- higher net power produced. The rst two benets are self
tures as a function of AS loading. Here, the predicted reboiler explanatory, while the third requires a more thorough discus-
temperatures are signicantly reduced with increasing AS sion (vide infra).
loading, as expected. At two molar equivalents the calculated The previous section outlines a number of potential
reboiler temperature is projected to decrease to 86  C from a improvements to the PSAR system that could lower the corre-
starting point of around 159  C. However, the corresponding sponding reboiler duty relative to aqueous amines. However,
heat rates, or reboiler duties, are not predicted to change even if the reboiler duty is unchanged with AS additions, the
signicantly. The initial increase in heat rate is due to the lower regeneration temperature offers signicant efficiency
additional mass and sensible heat of the AS prior to it having gains from the steam cycle of the power plant. A simple let-down
turbine could expand the reboiler steam prior to its condensing,
to produce more power from reboiler steam while it was cooled
to the lower operating temperatures.

Net electric power predictions of PSAR


Projections of the net power production from a full-scale power
plant were also estimated for the CO2BOLs-PSAR process using
the previously described ASPEN Plus models. The DOE refer-
ence Case 10 using monoethanolamine (550 MWe net electric
power, subcritical PC plant) was used as the basis for the
analyses.4 Case 10 is a conventional coal red 550 MWe net
electric power plant. The ue gas ow rate (3 213 000 kg h1)
and composition (13.5 vol% CO2 and 15.4 vol% H2O, balance
mostly N2) from Case 10 was xed for all of the CO2BOL-PSAR
Fig. 14 Heat rate and regeneration temperature as a function of AS (C16) cases evaluated.4 The ue gas ow rate and composition from
loading. Case 10 were xed for all of the CO2BOL-PSAR cases evaluated.

This journal is ª The Royal Society of Chemistry 2013 Energy Environ. Sci., 2013, 6, 2233–2242 | 2239
View Article Online

Energy & Environmental Science Paper

Table 2 Projected net electric power output for CO2BOL-PSAR as a function of duty is estimated to reduce the parasitic load of the CO2BOL/
AS (C16) loading PSAR process at 65  C to 17%, about half that of the DOE Case
10 baseline of 30%. This equates to as much as 100 MWe more
Antisolvent loading Regeneration Net electric power Parasitic
(molar equivalent) temperature ( C) produced (MWe) load net power production than the DOE Case 10 baseline, at an
equivalent coal feed rate.
0 159 594 25% This assessment of the CO2BOL/PSAR system is still new and
0.5 132 603 23% the results are preliminary as these results were derived from
1 109 621 21%
2 86 637 19%
equilibrium simulations of (A) with a representative AS. As we
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

TBDa 65 652 17% test other candidate ASs we will extend our simulations to
a
quantify their behavior in the PSAR process and compare it to
Based on projections of upper critical solution temperature.
incumbent systems.

This means that processes more efficient than Case 10 will be Experimental
projected to produce more net electricity than 550 MWe.
Solvent preparation and handling
Table 2 shows the estimates for net electric power as a
function of AS loading for CO2BOLs-PSAR cases using C16 AS. (A) was synthesized according to literature methods25 and was
As presented in Table 2, the predicted net power of the CO2BOL stored under N2. (A) was evacuated overnight before each use to
case with no AS addition is 44 MWe higher than that of Case 10. remove any N2 or water in the sample. (A) was then delivered via
This corresponds to a parasitic power of 25%, versus 30% for syringe for each use.
Case 10. When the PSAR is applied, the net power for the (A) was either used fresh or “regenerated” in situ by heating
CO2BOL process rises with increases in AS loading. The addi- to 60–80  C under vacuum until all water and CO2 were removed
tion of AS produces more predicted power because of the from the sample. Neat 13C NMR conrmed the integrity of the
described let-down turbine benet, with as much as 90 MWe net sample before each subsequent use.
electric power increase using 2 molar equivalents of C16. The
lowest theoretical reboiler temperature is estimated to be 65  C VLE measurements
with a best-case AS. Assuming this temperature, and no further Equilibrium PTx data were obtained using a custom-made
reductions in reboiler duty, the net electric power for the PSAR stainless steel cell of approximately 430 cm3 internal volume
at full strength is estimated to increase by over 100 MWe with a chamber diameter of 5.08 cm. Two opposing borosilicate
compared to the baseline Case 10, which corresponds to a sight glass windows provided for visual observation of the liquid
parasitic load of 17% (nearly half that of Case 10). level and phase behavior. The cell vessel was submerged in a
silicone oil bath to control temperature. The test sample in the
Conclusions cell chamber was mixed by a magnetic stir bar. Liquid solvent
sample was injected through a valve on the cell top ange. A gas
CO2BOLs (when paired with polarity-swing regeneration) are
injection device and a vacuum pump were connected to the cell
estimated to be efficient solvents for CO2 separations from ue
via another valve manifold. The injection device consisted of a
gas. Equilibrium measurements show CO2 loadings of CO2BOLs
gas-tight syringe with pressure transducers at the syringe and
are comparable to those of aqueous amines, and that our
the manifold. A third pressure transducer was installed on the
current candidate molecule (A) has high water tolerance.
cell main chamber.
CO2BOLs show a stronger temperature dependence on CO2
VLE measurements were done under isothermal conditions.
loading than aqueous amines but still require high regenera-
The cell was initially loaded with a known quantity of solvent
tion temperatures (159  C) for conventional thermal regenera-
sample between 20 and 50 mL. The cell was then evacuated to
tion. Addition of the polarity swing using C16 as a
remove any residual absorbed gas and volatile vapor. A series
representative AS was found to impact equilibrium loading
of gas injections between 0.1  103 and 10  103 mol was
through a reduction in the CO2BOL's polarity, allowing for
made using a supercritical grade CO2 from Praxair. The cell
similar CO2 stripping quantities at far lower stripping temper-
pressure and temperature were allowed to reach a new steady
atures (86  C) and a projected theoretical minimum regenera-
state prior to each injection. The equilibrium pressure and
tion temperature of 65  C. The stripping temperature was found
temperature were then recorded. The solvent CO2 loading was
to be dependent on the amount of AS added, indicating the
not directly measurable. Assuming ideal VLE behavior, gas
PSAR is tunable.
phase compositions were calculated from pure component
The lower regeneration temperatures available in the PSAR
vapor pressures using Raoult's law. The CO2 concentration in
allow two key process benets: reduced solvent attrition and
the liquid phase was then calculated from the mass balance
higher net power output. The low regeneration temperatures
within the cell.
reduce fugitive emissions and reduce thermal degradation of
CO2BOL compared to conventional thermal regeneration. The
lower regeneration temperature also allows for a simple let- LLE experiments
down turbine to be used to extract power before delivery to the n-C16 (99%) and 2-dimethylamino ethanol (>95.5%) were
reboiler. This extra power coupled with the low reboiler heat purchased from Sigma-Aldrich. n-C10 (99%) was purchased

2240 | Energy Environ. Sci., 2013, 6, 2233–2242 This journal is ª The Royal Society of Chemistry 2013
View Article Online

Paper Energy & Environmental Science

from Alfa Aesar. high performance liquid chromatography Acknowledgements


(HPLC) grade methanol was purchased from Fisher. CO2 (CD
3.0 grade, 99.9%) was purchased from Praxair. All chemicals The authors would like to acknowledge the Department of
were used as received without further purication. Energy's Office of Fossil Energy for funding, (award # DE-
A constant temperature bath (Cannon CT500 series II) was FE0007466), Mike Elliot for discussions on process congura-
used to get accurate temperature for phase behavior studies. A tions and Don Camaioni for thermodynamic discussions. PNNL
Shimadzu GC-17A gas chromatograph and silica column (CP- is proudly operated by Battelle for the United States Department
Volamine by Agilent J&W) were used to determine the compo- of Energy.
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

sition of the different phases.


Notes and references
Liquid phase behavior of (A)/antisolvent
1 D. Aaron and C. Tsouris, Sep. Sci. Technol., 2005, 40, 321.
(A) (650 mg, 3.8  103 mol) and n-C16 (860 mg, 3.8  103 mol) 2 A. N. M. Peeters, A. P. C. Faaij and W. C. Turkenburg, Int. J.
or n-C10 (508.1 mg, 3.8  103 mol) were added to a 10 mL glass Greenhouse Gas Control, 2007, 1, 396.
vial equipped with a Teon magnetic stir bar. The vial was 3 G. T. Rochelle, Science, 2009, 325, 1652.
closed with a rubber septum, which was then pierced with a 4 NETL Cost and Performance Baseline for Fossil Energy Plants,
venting needle to prevent pressure buildup. The vial was Department of Energy, 2010.
inserted into a constant temperature water bath and stirred for 5 M. Ramdin, T. de Loos and T. Vlugt, Ind. Eng. Chem. Res.,
1 h at a specic temperature. The stirring was then stopped and 2012, 51, 8149.
the phases were allowed to separate for at least 2 h. A sample 6 D. Camper, J. Bara, D. Gin and R. Noble, Ind. Eng. Chem. Res.,
was then carefully taken from one of the two phases. The 2008, 47, 8496.
remaining liquids were then stirred again for 1 h, aer which 7 E. D. Bates, R. D. Mayton, I. Ntai and J. H. Davis, J. Am. Chem.
they were allowed to resettle for 1 h before a sample was taken Soc., 2002, 124, 926.
from the other phase. Each phase was dissolved in 100 mL 8 B. E. Gurkan, J. C. de la Fuente, E. M. Mindrup, L. E. Ficke,
methanol solution containing 30 mL of 2-dimethylamino B. F. Goodrich, E. A. Price, W. F. Schneider and
ethanol as internal standard. The diluted solutions were injec- J. F. Brennecke, J. Am. Chem. Soc., 2010, 132, 2116.
ted into the GC, and the AS weight fraction was determined 9 P. Sharma, S. Park, I. Baek, K. Park, Y. Yoon and S. Jeong,
from the AS/internal standard ratio using a calibration curve Fuel Process. Technol., 2012, 100, 55.
obtained for each AS system. 10 K. Lin and A. Park, Environ. Sci. Technol., 2011, 45, 6633.
11 R. Perry, B. Wood, S. Genovese, M. O'Brien, T. Westendorf,
Liquid phase behavior of (A)/antisolvent/CO2 M. Meketa, R. Farnum, J. McDermott, I. Sultanova,
The phase behavior of ((A) + CO2) was performed in a similar T. Perry, R. Vipperla, L. Wichmann, R. Enick, L. Hong and
way: (A) (650 mg, 3.8  103 mol) and n-C16 (860 mg, 3.8  103 D. Tapriyal, Energy Fuels, 2012, 26, 2528.
mol) or n-C10 were added to a 10 mL glass vial equipped with a 12 F. Barzagli, M. Di Vaira, F. Mani and M. Peruzzini,
Teon magnetic stir bar. The vial was closed with a rubber ChemSusChem, 2012, 5, 1724.
septum and pierced with a needle to prevent pressure buildup. 13 Y. Cheon, Y. Jung, J. Lee, H. Kim, J. Im, M. Cheong, H. Kim
The vial was placed in a constant temperature water bath and and H. Park, ChemPhysChem, 2012, 13, 3365.
stirred for 1 h. Then CO2 was bubbled into the solution at a 14 J. Tan, H. W. Shao, J. H. Xu, L. Du and G. S. Luo, Ind. Eng.
specic temperature via a second needle for 1 h. The stirring Chem. Res., 2011, 50, 3966.
was stopped and the phases allowed to separate at the specic 15 P. G. Jessop, D. J. Heldebrant, X. W. Li, C. A. Eckert and
temperature for 2 h under CO2-saturated atmosphere. The AS C. L. Liotta, Nature, 2005, 436, 1102.
weight fraction was determined for each phase by GC using the 16 D. J. Heldebrant, C. R. Yonker, P. G. Jessop and L. Phan,
same method as before. Energy Environ. Sci., 2008, 1, 487.
17 L. Phan, D. Chiu, D. J. Heldebrant, H. Huttenhower, E. John,
X. W. Li, P. Pollet, R. Y. Wang, C. A. Eckert, C. L. Liotta and
Liquid phase behavior of (A)/C16/20 mol% CO2 P. G. Jessop, Ind. Eng. Chem. Res., 2008, 47, 539.
Partially carbonated (A) was prepared by weighing the required 18 V. Blasucca, C. Dilek, H. Huttenhower, E. John, V. Llopis-
amount of lean (A) (600.0 mg, 3.50  103 mol) into a 10 mL Mestre, P. Pollet, C. A. Eckert and C. L. Liotta, Chem.
glass vial equipped with Teon magnetic bar. CO2 bone-dry Commun., 2009, 116.
grade (30.8 mg) was then bubbled directly into it while the 19 V. Blasucci, R. Hart, V. Llopis Mestre, D. J. Hahne,
weight increase was monitored with an analytical balance. The M. Burlager, H. Huttenhower, B. Joo, R. Thio, P. Pollet,
vial was then sealed with a rubber septum and C16 added via a C. L. Liotta and C. A. Eckert, Fuel, 2010, 89, 1315.
syringe. The mixture was heated in the sealed vial at the 20 D. J. Heldebrant, P. K. Koech, J. Rainbolt, C. R. Yonker,
required temperature and stirred for 1 h. Stirring was stopped T. Ang, C. Liang and P. G. Jessop, Green Chem., 2010, 12, 713.
and the phases allowed to separate at the specic temperature 21 M. Kim and J. W. Park, Chem. Commun., 2010, 2507.
for 2 h. GC determined the C16 weight fraction in each phase as 22 C. Wang, H. Luo, D.-E. Jiang, H. Li and S. Dai, Angew. Chem.,
specied before. Int. Ed., 2010, 49, 5978.

This journal is ª The Royal Society of Chemistry 2013 Energy Environ. Sci., 2013, 6, 2233–2242 | 2241
View Article Online

Energy & Environmental Science Paper

23 C. Wang, S. M. Mahurin, H. Luo, G. A. Baker, H. Li and 26 P. G. Jessop, D. A. Jessop, D. Fu and L. Phan, Green Chem.,
S. Dai, Green Chem., 2010, 12, 870. 2012, 14, 1245.
24 D. J. Heldebrant, P. K. Koech, J. E. Rainbolt, F. Zheng, 27 P. K. Koech, J. E. Rainbolt, M. D. Bearden, F. Zheng and
T. Smurthwaite, M. Oss and I. Leito, Chem. Eng. J., 2010, D. J. Heldebrant, Energy Environ. Sci., 2011, 4, 1385.
171, 794. 28 J. F. Deye, T. A. Berger and A. G. Anderson, Anal. Chem., 1990,
25 P. K. Koech, J. Zhang, I. V. Kutnyakov, L. Cosimbescu, S. Lee, 615.
M. E. Bowden, T. D. Smurthwaite and D. J. Heldebrant, RSC 29 C. Reichardt, Chem. Rev., 1994, 94, 2319.
Adv., 2013, 3, 566. 30 M. Born, Zeitschri für Physik, 1920, 1, 45.
Published on 31 May 2013. Downloaded by University of Illinois at Chicago on 28/10/2014 10:00:09.

2242 | Energy Environ. Sci., 2013, 6, 2233–2242 This journal is ª The Royal Society of Chemistry 2013

You might also like