Bench-Scale Testing and Process Performance Projections of CO2

Jan 12, 2016 - This manuscript provides a detailed analysis of a continuous-flow, bench-scale study of the CO2-binding organic liquid (CO2BOL) solvent...
0 downloads 11 Views 1MB Size
Article pubs.acs.org/EF

Bench-Scale Testing and Process Performance Projections of CO2 Capture by CO2−Binding Organic Liquids (CO2BOLs) with and without Polarity-Swing-Assisted Regeneration Feng Zheng,† David J. Heldebrant,*,† Paul M. Mathias,‡ Phillip Koech,† Mukund Bhakta,‡ Charles J. Freeman,† Mark D. Bearden,† and Andy Zwoster† †

Pacific Northwest National Laboratory, 902 Battelle Boulevard, Richland, Washington 99352, United States Fluor Corporation, 3 Polaris Way, Aliso Viejo, California 92698, United States



S Supporting Information *

ABSTRACT: This manuscript provides a detailed analysis of a continuous-flow, bench-scale study of the CO2-binding organic liquid (CO2BOL) solvent platform with and without its polarity-swing-assisted regeneration (PSAR). This study encompassed four months of continuous-flow testing of a candidate CO2BOL with a thermal regeneration and PSAR regeneration using a decane antisolvent. In both regeneration schemes, steady-state capture of >90% CO2 was achieved using simulated flue gas at reasonable liquid/gas (L/G) ratios. Aspen Plus modeling was performed to assess process performance, compared to previous equilibrium performance projections. This paper also includes net power projections, and comparisons to DOE’s Case 10 amine baseline, and comments on the viability of the CO2BOL solvent class for post-combustion CO2 capture.

1. INTRODUCTION Ever since Bottoms patented the first amine scrubbing technology for CO2 in the 1930s, there have been many improvements and advancements in engineering design and solvent chemistry.1−5 There are many aqueous formulations of amines that operate on similar principles; however, there are many new advanced organic formulations that are being developed primarily at the laboratory scale.6−13 The goal of water-lean or concentrated solvent system is to reduce the amount of water carried by the solvent in order to minimize the duty of boiling and condensing water in the process. Reducing water loads may enable coal-fired power plants to exploit the lower specific heats of organics as a means to decrease reboiler duty. These anhydrous/water-lean formulations have yet to be proven on any meaningful scale; thus, they cannot be directly compared to commercialized first-generation and piloted second-generation solvent systems. CO2-binding organic liquids (CO2BOLs) are water-lean solvent systems that fall under the United States Department of Energy (DOE) National Technology Laboratory (NETL)’s category of “transformational” solvent systems.14 CO2BOLs operate on Jessop’s “switchable solvents” chemistry that chemically fixate CO2 as an alkylcarbonate ionic liquid (see Figure 1).15−17 Operating a 100% concentrated CO2BOL solvent enables the use of polarity-swing-assisted regeneration (PSAR), which exploits the solvent’s fundamental polarity change as a function of CO2 loading (Figure 2). Here, the degree of CO2 loading can be controlled by the polarity of the solvent (adjusted by metered injection of a chemically inert nonpolar “antisolvent”) at a given temperature and pressure.18 Introducing a polarity swing to regenerate CO2BOL solvents has the potential to reduce reboiler temperatures from conventional thermal stripping (155 °C) by as much as 70 °C (down to 85 °C), © XXXX American Chemical Society

Figure 1. Reversible uptake of CO2 by IPADM-2BOL.

Figure 2. Conceptual “switch” by CO2 loading (polarity scale of Nile Red indicator dye, μM).

which may translate into improved efficiency gains from the plant’s steam cycle.18 The most advanced CO2BOL formulation that has been studied extensively is 1-((1,3-dimethylimidazolidin-2-ylidene)amino-propan-2-ol), an alkanolguanidine (IPADM-2BOL; see Figure 1).19 IPADM-2BOL is nonvolatile, and it is the least viscous CO2BOL derivative to date, having a heat of solution Received: October 16, 2015 Revised: January 12, 2016

A

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Figure 3. Conceptual CO2BOLs absorption and PSAR process. (Reproduced from ref 18 with permission from the Royal Society of Chemistry, London, 2013).

between −80 kJ/mol to −90 kJ/mol with CO2 release performed by thermal heating (Figure 1). Our laboratoryscale equilibrium and kinetic studies (assuming that a 20 cP viscosity target was achievable) suggested that IPADM-2BOL had the potential to be more energy efficient than DOE’s Case 10 amine baseline, primarily through faster-than-expected CO2 flux and the addition of PSAR, enabling large increases in net power.17 However, these values had yet to be validated in a continuous-flow system. Bench-scale continuous flow testing was needed before largescale projections of solvent performance and equipment sizing and costing could be made. Without these projections, anhydrous solvent performance could not be accurately compared to aqueous amine solvent technology. Bench-scale testing of anhydrous solvents has been performed, notably GE’s aminosilicones,11 RTI International’s nonaqueous solvents,13 and Ion Engineering’s Imidazoles, although there are currently no peer-reviewed published results from those studies.20 With any new solvent system, there remain uncertainties on equipment performance and compatibility. The CO2BOL solvent platform specifically had uncertainties with high viscosity and continuous mixing and separation of the antisolvent in the PSAR. To answer these uncertainties, and provide a thorough analysis of solvent performance, we set out to perform bench-scale testing to validate the solvent and PSAR. Here, we present the results of a four-month study of bench-scale testing of a representative CO2BOL with the goal of assessing feasibility of this solvent to capture 90% of the CO2 in a post-combustion flue gas stream (15% CO2 concentration), and confirming the viability of the solvent’s distinctive PSAR via continuous flow. The bench-scale results enabled a more accurate Aspen Plus model and improved net power projections.

2. RESULTS AND DISCUSSION 2.1. Cart Configuration for Testing. A conceptual diagram of the CO2BOLs/PSAR process was previously described in our prior work (Figure 3) and was used as the basis of the bench cart for testing.18 The process is proposed to run similarly as aqueous amine solvent, albeit with the addition of the polarity-swing infrastructure. First, incoming flue gas is cooled with a direct-contact cooler (DCC) to remove any particulate matter, but also to reduce the water content of the gas via condensation. Next, CO2 is absorbed at about 40 °C in a conventional absorber column with packing. The cold-rich solvent leaving the absorber is pumped into a static mixer, where it is mixed with a nonpolar and chemically inert “antisolvent.” The biphasic fluid mixture is then passed through the cross exchanger, where the increase in temperature increases the miscibility between the two liquids and they begin to mix. The fluid mixture is then delivered to the stripper column, where the reduced polarity of the solvent mixture facilitates thermal CO2 release via the PSAR effect. The now hot-lean fluid mixture is then pumped back through the cross exchanger, and then into a coalescing unit to cool. Further cooling by heat exchange with water promotes the phase separation of the CO2BOL and antisolvent into two distinct liquid layers. The denser CO2BOL is pumped to the absorber while the lessdense antisolvent is decanted and circulated back to the static mixer completing the cycle. 2.2. Bench-Scale Batch-Wise Absorption Testing and PTx Validation of Column Performance Using IPADM2BOL. IPADM-2BOL was synthesized and characterized and purified by our earlier methods.19 The absorption performance of the columns was then tested by comparing CO2 loading profiles of IPADM-2BOL in the cart to our previously measured vapor−liquid equilibria (VLE) data measured using a PTx cell.18 Two and a half liters (2.5 L) of IPADM-2BOL were pumped into the absorber side of the bench cart under an N2 atmosphere. Batch-wise absorption measurements were B

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

not heat-traced; thus, the solvent flowing through the system was cooler than what had exited the absorber (40 °C, ±5 °C). Viscosity data are reported at the temperature at which the data were measured, to ensure that accurate CO2 loadings could be calculated from the data. Under initial testing conditions, IPADM-2BOL was able to capture over 90% of the CO2 (Figure 6) and the CO2

performed using the absorber side in isolation at 40 °C and at 60 °C. The measured CO2 absorption data from the cart (based on mass balance calculations) are provided in Figures 4 (40 °C)

Figure 4. Measured VLE data of the bench cart (red line) for IPADM2BOL plotted against VLE data measured from PTx measurements (40 °C).18

and 5 (60 °C), where they are compared to VLE data.18 As seen in Figures 4 and 5, the measured batch data (red line) correspond with measured equilibrium data from the PTx measurements.

Figure 6. CO2 capture efficiency of IPADM-2BOL, as a function of lean solvent viscosity (data presented in terms of cP @ 31.5 °C).

absorption rate was found to be independent of lean solvent viscosity (see Figure 7, as well as data entry rows 1−6 in Table 1). Each data point corresponds to a run with at least 1 h at steady state.

Figure 5. Measured VLE data of the bench cart (red line) for IPADM2BOL plotted against VLE data measured from PTx measurements (60 °C).18 Figure 7. CO2 absorption rate by IPADM-2BOL, as a function of lean solvent viscosity (cP @ 31.5 °C).

Once absorber and material performance was validated, the absorber side of the cart was opened to the stripper side and IPADM-2BOL was then subjected to continuous flow testing under a variety of conditions. Testing conditions included 15% and 10% CO2/N2 gas concentrations, absorption at 40 °C (± 5 °C) absorption at atmospheric pressure. Testing conditions varied stripper temperature, gas flow and lean solvent viscosity. The stripper heating jackets were constrained to 100 °C due to the heating baths. 100 °C was not hot enough to enable thermal stripping of CO2. Thus, the stripper was configured to use a small N2 flow as a means to reduce the temperature required to strip CO2 during testing. NOTE: The stripping temperature and N2 are used only due to the limitations of the bench cart inf rastructure and are not required for CO2 stripping under real-world applications. The stripper N2 flow is taken into account in the Aspen Plus modeling of process performance vide inf ra. Attention was focused on measuring the viscosity and temperature in real time, as temperature and CO2 loading directly impact CO2BOL viscosity. Lean solvent viscosity measurements were controlled as a means to set a lean solvent loading for the absorber and in turn the CO2 absorption capacity of the BOL. The continuous-flow viscometer used in testing had an internal temperature probe that yielded the viscosity and temperature at which the viscosity was measured ensuring an accurate measure of lean solvent loading could be monitored in real time. Viscosities were measured near 31.5 °C, because the tubing lines to the viscometer were insulated but

The high CO2 capture observed in Figures 6 and 7 led us to believe that equilibrium loading was achieved during testing. The high surface area packing inside the absorber, compounded with a viscous fluid, resulted in a visually observed slow drainage of fluid out of the absorber column, suggesting the residence time of IPADM-2BOL was too high. Initial attempts to decrease residence time were futile, because of the increase in fluid viscosity with CO2 loading. The high residence time was a vicious cycle, where the solvent would load with CO2 and increase in viscosity, increasing residence time, which would then allow the fluid to pick up more CO2, resulting in yet a higher viscosity and longer residence time. This cycle continued until equilibrium-loading conditions were always reached. A continuous test of equilibrium absorption over 48 h showed >98% CO2 capture with no observed loss in activity or selectivity for 2 slm inlet gas flow at 14.9% CO2, absorber temperature of TA = 46.4 °C, L/G = 4 (mol/mol), stripper temperature of 80.7 °C with a 3.45 slm N2 flow. [Throughout this paper, the unit “slm” denotes standard liters per minute.] Note that even though equilibrium loading was reached in each initial case, steady-state operation was observed, even with rich solvent viscosities as high as 900 cP (40 °C) draining out of the absorber column. The cart was designed and built with an absorber column twice the length of the stripper on the assumption that C

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Table 1. Bench-Scale Data for the Thermal Release of IPADM-2BOL, without PSARa yCO2,in [mole fraction]

TA [°C]

0.0997 0.0997 0.1490 0.1492 0.1488

36.3 36.4 36.5 38.0 36.4

0.1491 0.1502 0.1495 0.1495 0.1496

43.2 44.4 46.0 45.1 49.3

TS [°C]

QS,N2 [slmb at 21.1 °C]

ηcapture [%]

L/G, abs [mol/mol]

μ [cP]

ρ [g/mL]

Tvisc [°C]

3.95 × 10−5 × 10−5 3.94 × 10−5 3.95 × 10−5 3.95 × 10−5 3.94 No PSAR after Column Reversal × 10−5 3.96 × 10−5 2.99 × 10−5 1.98 × 10−5 1.98 × 10−5 1.32

34.89 30.72 52.54 77.96 44.31

0.928 0.923 0.939 0.953 0.938

31.1 31.0 31.4 32.6 31.2

8.30 10.38 17.66 12.24 17.67

1.094 1.058 0.928 0.917 0.878

73.6 73.5 72.2 72.7 72.0

NCO2,abs [mol/m2/s]

81.5 3.059 99.4 3.83 80.9 5.654 99.4 3.83 80.5 3.561 99.5 5.72 75.1 2.244 99.1 5.70 80.8 3.459 99.5 5.72 Bench-Scale Data for Thermal Release of IPADM-2BOL, 75.4 5.063 98.6 1.34 76.4 5.061 97.3 1.76 75.3 5.061 83.7 2.27 75.9 7.060 87.4 2.37 76.8 6.065 80.7 3.28

a

Legend: yCO2,in, CO2 concentration at the absorber inlet; TA, average absorber temperature; TS, average stripper temperature; QS, stripper feed gas; η, percentage of CO2 captured; L/G, liquid/gas ratio; μ, viscosity; ρ, density; and Tvisc, temperature of viscosity and density measurements (on the lean solvent stream before column reversal; on the rich solvent stream after column reversal). bNote that, throughout the paper, the unit “slm” denotes standard liters per minute.

studies confirm that IPADM-2BOL can meet the DOE requirements of 90% capture of CO2 from post-combustion gas streams at reasonable L/G ratios. 2.3. Continuous Bench-Scale Testing of IPADM-2BOL with an Added PSAR Regeneration Cycle. After the thermal regeneration testing, the bench cart was retrofitted to introduce the PSAR modules (coalescing tank, static mixer, and antisolvent circulation pump). The viscometer was moved back to its original location at the bottom of the absorber to measure the cold−rich solvent (∼40 °C). Once the components were installed, 5 L of decane (antisolvent) was pumped into the coalescing tank. Decane was chosen as the antisolvent for this testing rather than hexadecane, because of latter’s tendency to gel upon CO2 loading, and its miscibility temperature with IPADM-2BOL was higher than the 100 °C that the bench cart heaters on the stripper column could achieve.18 PSAR testing (with decade addition) was performed at three different L/G ratios, varied stripping gas flows, solvent circulation rates, and gas inlet concentration of 15% CO2. In all PSAR testing, the circulation rate of decane was matched at one molar equivalent of IPADM-2BOL (∼80 mL/min). The stripper temperature was kept close to 75 °C (±5 °C) and the nitrogen strip was varied between 4.5 slm and 6 slm to maintain a 25 cP viscosity (lean loading of 0.1 mol CO2/mol BOL) of the anhydrous lean solvent. Each data point corresponds to a run with at least 1 h at steady state. As previously stated, there was no observed foaming or biphasic behavior in the absorber. Visual inspections of the coalescing tank during operation did reveal small microemulsions in the middle of the decane/ IPADM-2BOL fluid mixture. A clear and colorless decane layer was observed on top, with a clear yellowish-orange IPADM2BOL at the bottom layer. The microemulsions were not visible in the decane or IPADM-2BOL outlets, indicating they had separated prior to leaving the coalescing tank. The estimated residence time for the antisolvent in the coalescing tank was estimated to be 3 min, based on circulation rate and holdup volume. Antisolvent carryover in the fluids entering the absorber was quantified using 13C NMR. Liquid aliquots (1 mL) of lean solvent in all PSAR testing were sampled and subjected to 13C NMR analysis. NMR analysis of the neat samples was measured at 40 °C to ensure spectroscopic analysis was being performed under absorber conditions. The integrations of decane and IPADM-2BOL were quantified, and compared to previously

CO2BOLs would have a slower CO2 uptake than aqueous solvents due to a higher viscosity. This assumption was prior to our kinetic observations that CO2BOLs have comparable liquid-film mass-transfer coefficients (kg′) values to aqueous solvents. The absorber and stripper columns were switched, effectively cutting the absorber length in half. All hardware and other configurations were not modified, and key infrastructure such as the viscometer were kept in their original location. The viscometer was now measuring the hot−lean solvent coming out of the stripper and all data from those testing conditions show a higher temperature (∼75 °C) and far lower viscosities than the previously measured cold−rich solvent out of the absorber. Once the columns were swapped, the residence time of the IPADM-2BOL was reduced by half, and CO2 absorption followed kinetically limiting loading profiles. Under this new configuration, continuous-flow absorption experiments were performed at varied L/G ratios, stripping gas flows, solvent circulation rates, at a CO2 inlet gas concentration of 15%. The stripper temperature was kept between 75 °C and 82 °C, and the N2 strip was varied between 2 slm and 6 slm to keep the viscosity of the cold−lean solvent to 10% water. In the next batch of testing, water was added to the inlet gas stream to measure the effect of water on solvent performance. Gas humidifiers were installed upstream of the stripper and absorber sides of the cart. The humidifiers were filled with deionized water and saturated the gas streams with water at the respective temperatures of absorption (40 °C) and stripping (75 °C). Prior to this next batch of testing, 5 wt % deionized water was added to the IPADM-2BOL in the cart. This water loading was previously estimated as the thermodynamically optimal water load.18 Again, the decane circulation loop was set to 1 mol equiv (∼80 mL/min), and PSAR testing with water was run under similar PSAR conditions, with the exception of the 5 wt % water loaded to IPADM-2BOL and gas delivery to the absorber and stripper columns being saturated with water. PSAR testing with added water was performed at L/G ratios between 3 and 4 (mol/mol), at a stripping gas flow rate of 6 slm with CO2 inlet concentrations of 5.6%−15% with a decane flow rate of 80 mL/min (corresponding to 1 mol decane/mol of BOL). In these tests, the viscosity of the cold−lean, waterladen solvent was held near 25 cP (corresponding to a lean loading of 0.04 mol CO2/mol BOL with 0.5 equiv water). Figure 10 shows the PSAR/water data, which is also tabulated

measured liquid−liquid equilibria data of decane and IPADM2BOL.18 The NMR data matched that of the previously measured data, indicating the coalescing tank was efficiently separating the decane from IPADM-2BOL continuously.18 Note that a precise measured delivery rate of IPADM-2BOL to the absorber could not be directly measured: it was estimated based on mass balance data. For safety, parametric runs of PSAR were kept at steady state for no more than 1 h, because of the high evaporative losses of decane out of the absorber and stripper columns. Frequent draining of the liquid condensers on the absorber and stripper and recharging of decane in the coalescing tank was needed to maintain adequate liquid levels for decantation of decane using a diptube. Less-volatile antisolvents would prevent this limitation in future testing and any practical application, assuming that an achievable miscibility temperature with IPADM-2BOL is feasible. Examining the data, it is clear that PSAR addition does not negatively impact CO2 absorption as 90% CO2 capture was achieved at reasonable L/G ratios, and was slightly higher than the thermal regeneration cases. The measured PSAR data are plotted in Figure 9 and presented in

Figure 9. CO2 capture efficiency with PSAR, plotted against L/G for IPADM-2BOL at 15% CO2.

Table 2. Note that the decane molar flow was not included in the calculation of the L/G ratio, because there were trace amounts of decane entering the absorber column. CO2 capture with IPADM-2BOL, using a PSAR regeneration cycle, showed increasing CO2 capture with increases in L/G ratio. At a CO2 absorber inlet concentration of 15%, 95% CO2 capture was achieved at an L/G ratio of 4.8 (mol/mol) and 90% CO2 capture was estimated to require an L/G ratio of 4 (mol/mol). The PSAR did not impact absorber behavior, because of the immiscibility between the antisolvent and IPADM-2BOL at 40 °C, indicating that the PSAR occurs only in the elevated stripper temperatures where IPADM-2BOL and decane are miscible. 2.4. Continuous Bench-Scale Testing of IPADM-2BOL with PSAR Regeneration Cycle with 5 wt % Water Loading. Practical application of CO2 capture requires testing

Figure 10. CO2 capture efficiency with PSAR and 5% water, plotted against L/G ratio for IPADM-2BOL at a CO2 feed concentration in the range of 5.6%−15%.

in Table 3. Each data point corresponds to 1 h at steady state. Parametric runs were kept at steady state no longer than 1 h to keep evaporative losses of decane to a minimal level. Visual inspections of the absorber and coalescing tank showed no observable foaming or biphasic behavior. PSAR regeneration was found to be feasible in the presence of water. Our previous studies projected that water would make IPADM-2BOL a better solvent, because of bicarbonate formation or stabilization of the CO2-rich species by hydrogen bonding.18 The data in Figure 10 and Table 3 clearly shows stronger CO2 uptake than the PSAR data in the absence of

Table 2. Bench-Scale Data for PSAR Release of IPADM-2BOLa yCO2,in [mole fraction]

TA [°C]

TS [°C]

QS,N2 [slm @ 21.1 °C]

ηcapture (%)

0.1491 0.1493 0.1493 0.1489 0.1490 0.1491

40.4 41.8 42.6 42.8 40.1 40.1

74.1 73.3 72.5 74.0 74.4 74.2

4.664 6.067 6.061 5.057 6.066 5.058

98.6 95.1 83.8 84.5 96.0 95.6

NCO2,abs [mol/m2/s] 4.99 1.29 1.71 1.72 1.31 1.30

× × × × × ×

10−5 10−4 10−4 10−4 10−4 10−4

L/G, abs [mol/mol]

μ [cP]

ρ [g/mL]

Tvisc [°C]

6.55 4.71 3.16 3.16 4.71 4.71

16.73 15.86 23.84 23.79 22.78 23.78

0.903 0.909 0.912 0.918 0.907 0.908

44.8 45.0 43.2 43.9 41.1 41.1

Legend: yCO2,in, CO2 concentration at absorber inlet; TA, average absorber temperature; TS, average stripper temperature; QS, stripper feed gas; η, percentage CO2 captured; L/G, liquid/gas ratio; μ, lean solvent viscosity; ρ, lean solvent density; and Tvisc, temperature of viscosity and density measurements. a

E

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Table 3. Bench-Scale Data for PSAR Release of IPADM-2BOL in the Presence of 5 wt % Watera yCO2,in [mole fraction]

TA [°C]

TS [°C]

QS,N2 [slm @ 21.1 °C]

ηcapture (%)

0.0562 0.0809 0.1051 0.1275 0.1489

40.2 40.5 40.9 41.8 42.8

76.1 76.0 75.5 75.0 74.5

6.062 6.062 6.057 6.055 6.056

99.6 99.7 99.4 95.5 90.8

NCO2,Abs [mol/m2/s] 4.61 6.84 9.07 1.09 1.24

× × × × ×

L/G, abs [mol/mol]

μ [cP]

ρ [g/mL]

Tvisc [°C]

3.98 3.35 3.28 3.19 3.11

24.89 24.10 25.47 26.27 26.26

9.909 0.912 0.912 0.912 0.911

43.0 42.3 41.9 41.9 43.3

10−5 10−5 10−5 10−5 10−5

yCO2,in, CO2 concentration at absorber inlet; TA, average absorber temperature; TS, average stripper temperature; QS, stripper feed gas; η, % CO2 captured; L/G, liquid/gas ratio; μ, lean solvent viscosity; ρ, lean solvent density; and Tvisc, temperature of viscosity and density measurements. a

2.6. Finalized Process Model Based on Cart Data. The process simulations in this paper (both the modeling of cart data and the modeling of industrial-scale processes) have been performed using rate-based distillation (RateSep) and other unit-operation models in Aspen Plus. This requires accurate and reliable models to describe the equilibrium (thermodynamic) and kinetic phenomena occurring in the process. Our previous publication18 presented the thermodynamic model, which used the ElecNRTL model in Aspen Plus, together with a chemistry model to describe the complexes that form upon chemical absorption of CO2. This model enables quantitative description of the VLLE (vapor−liquid−liquid equilibrium) phase behavior as well as the heat of solution for CO2 absorption. Our recent publication21 used wetted-wall data to develop the model for rate-based processes that occur in the system, and, in particular, these include the kinetics and mass transfer rates as a function of system properties (e.g., viscosity). In principle, the model developed in our two previous publications18,21 can be used to simulate the cart data and also industrial-scale plants. In practice, however, one additional empirical parameter (the interfacial area factor, or IAF) is adjusted to improve agreement with absorption-column data. These analyses are presented in this section. The focus has been on the absorption data, which has been modeled as follows: • The absorber dimensions are D = 0.072 m and H = 0.316 m. • The interfacial area of the packing has been specified as 12.336 cm2/cm3. • The thermodynamic model, and kinetic and transportproperty correlations are the same as used previously.16,19

water. Water addition during testing allowed for >90% capture to be achieved at an L/G ratio of 3, instead of an L/G ratio of 4 in the absence of water. 2.5. Impact of PSAR on CO2BOL Absorption/Regeneration Based on Bench Data. A detailed analysis of the 15% CO2 inlet concentration data for the thermal and PSAR cases was performed to quantify the impact of PSAR on absorber performance. Absorber performance was assessed by plotting the percentage of CO2 removal from the inlet gas against the L/G for the thermal and PSAR test cases. The data are plotted in Figure 11 and are tabulated in Table 4. The

Figure 11. Bench-scale run data for CO2BOLs, with and without PSAR.

PSAR shows minimal impact at L/G ratios at 2.7 and 3.8 (mol/ mol), but has a small dip in capture efficiency at an L/G ratio of 1.8 (mol/mol). This dip is believed to be due to experimental error or incorrect assumptions of antisolvent carryover into the absorber that would influence the rate of BOL delivery to the absorber. Ultimately, we conclude comparable absorber performance with and without PSAR.

Table 4. Summary of Bench-Scale Run Data for CO2BOLs, with and without PSAR molar concentration of CO2 in inlet gas

capture efficiency (%)

rich solvent loading (mol CO2/mol BOL)

0.1491 0.1502 0.1495 0.1495 0.1496 0.1490 0.1494

98.6 97.3 88.1 87.4 80.7 94.5 89.2

0.11 0.13 0.15 0.15 0.18 0.16 0.19

0.0769 0.1493 0.1493 0.1489 0.1490

98.9 79.2 80.0 95.1 94.6

0.10 0.16 0.16 0.14 0.14

lean solvent loading, no antisolvent (mol CO2/mol BOL) Without PSAR 0.07 0.09 0.08 0.08 0.09 0.13 0.14 With PSAR 0.08 0.11 0.11 0.10 0.10 F

L/G ratio in the absorber

average absorber temperature (°C)

average stripper temperature (°C)

3.96 2.99 1.99 1.98 1.32 3.92 2.63

43.2 44.4 46.0 45.1 49.3 45.2 46.7

75.4 76.4 75.9 75.9 76.8 76.1 76.0

3.77 1.82 1.82 2.71 2.71

40.4 42.6 42.8 40.1 40.1

74.1 72.5 74.0 74.4 74.2

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels • In addition, the calculated interfacial area has been adjusted by changing the IAF in RateSep. • The gas and liquid flow rates for each point are as measured in the experiment. The bench-scale absorption runs test the combined effect of several independent variables: molar concentration of CO2 in the inlet gas, lean solvent loading, relative solvent flow rate or L/G ratio, and absorption temperature. There is a complex interplay between these independent variables, and modeling has a useful role in understanding this interdependence. Figure 12−Figure 14 study the calculated fraction of CO2 capture as a

Figure 14. Model calculations for the percentage of CO2 captured as a function of inlet CO2 concentration and L/G ratio for a lean ratio of 0.16 (αlean = 0.16). The full lines represent a rate-based calculation, while the dashed lines are for an equilibrium calculation. The black and red lines represent inlet CO2 mole fractions of 10% and 15%, respectively.

• At lower L/G ratios, the extent of CO2 capture decreases with CO2 inlet concentration, and this is because a greater amount of CO needs to be captured, which requires more solvent. • When an equilibrium model is used, the extent of CO2 capture is only weakly dependent on L/G ratio until the solvent capacity is reached, and once this capacity is reached, the extent of CO2 capture drops sharply with decreasing solvent flow rate. • By comparison, the rate-based model shows a more gradual decrease in extent of CO2 capture with decreasing solvent flow rate. In the region where the calculated extent of CO2 capture varies with solvent flow rate, a higher CO2 inlet concentration (higher CO2 flow rate) requires a higher L/G ratio. The L/G ratios for 15 mol % CO2 in the inlet gas are shifted to higher L/G ratios compared to the 10 mol % CO2 in the inlet gas, and this is mainly due to the limiting kinetics and mass transfer. • The progression of results from Figures 12−14 demonstrates how increased lean solvent loading decreases the extent of CO2 capture. The list of bulleted results presented above identify the various effects contributing to CO2 capture by absorption, and also indicate that the model treats the underlying phenomena in a proper way. In order to show comparisons between the experimental data and the model predictions, the Data-Fit capability in Aspen Plus (see Aspen Plus V7.3 documentation)22 was used to adjust the IAF, which was used as an empirical factor to best fit the measured data. IAF is an additional factor that is applied to calculate the wetted area, and may be needed, since the wetted-area factor has been defaulted to unity. Figure 15 presents the Data-Fit in a form of a parity chart; CO2 recoveries calculated by the model are compared to the experimental values measured in the cart. In the case of the data without antisolvent (denoted as “NO AS”), the model calculations with IAF = 1 indicate higher recoveries than the measurements, while the Data-Fit value of IAF = 0.61 provides an adequate average description of the data. In the case of cart runs with antisolvent (denoted as “AS”), the value of IAF that best fits the data is slightly lower (i.e., IAF = 0.51). We consider the difference between the two fits to be within experimental uncertainty. In conclusion, an IAF value between 0.5 and 0.6 is

Figure 12. Model calculations for the percentage of CO2 captured as a function of inlet CO2 concentration and L/G ratio for a lean ratio of 0.12 (αlean = 0.12). The full lines represent a rate-based calculation, while the dashed lines are for an equilibrium calculation. The black and red lines represent inlet CO2 mole fractions of 10% and 15%, respectively.

Figure 13. Model calculations for the percentage of CO2 captured as a function of inlet CO2 concentration and L/G ratio for a lean ratio of 0.14 (αlean = 0.14). The full lines represent a rate-based calculation, while the dashed lines are for an equilibrium calculation. The black and red lines represent inlet CO2 mole fractions of 10% and 15%, respectively.

function of L/G ratio for two inlet gas CO2 concentrations: 10 mol % and 15 mol %. The calculations are done using two modeling assumptions: an equilibrium model that assumes that all rate processes are extremely fast and a rate-based model with kinetic and transport processes described previously21 All absorption calculations were effectively performed at an absorption temperature of 43 °C. The following effects are evident from the three figures: • The extent of CO2 capture increases monotonically with L/G ratio, and at high L/G ratios the equilibrium and rate-based models converge. • At high L/G ratios the extent of capture increases with CO2 inlet concentration, and this is because the CO2 driving force for absorption is greater. G

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

Little is known about the corrosion potential of the amidine and guanidine bases used in CO2BOLs. A recent study of the impacts on corrosion by diazabicyclo[5.4.0]undec-7-ene (DBU), which is an amidine base structurally similar to IPADM-2BOL, on steel coupons in steam was performed in 2007 by Nasrazadani et al.23 Their study focused on quantifying corrosion by measuring the weight loss of AISI 1018 steel coupons that have been exposed to steam (120 °C) in the presence of varied amine additives, as a function of time. DBU was found to retard the weight loss of steel in the coupons, when compared to coupons in the absence of amine additives. DBU showed the lowest corrosion rate when compared to other amine inhibitors, such morphaline and DMA. The conclusions by Nasrazadani et al.23 were that DBU acts as a corrosion inhibitor in the presence of steam, which closely resembles a stripper of a CCS unit. We predict that other CO2BOL bases, such as the alkanolamidines or alkanoguanidines, would exhibit similar retardation in the corrosion of steel as other amine additives. A visual inspection of the bench cart fittings, tubing and packing was performed after the four months of continuous testing was performed. There was no observable corrosion or metal pitting seen in any of the components. Note that there still may be potential metal leaching into the CO2BOL solvent, but ICMS testing was not performed on the solvent. More formal corrosion testing is the focus of current work in our laboratory. 2.8. Energetic Projections. The process performance of the CO2BOL/PSAR flow sheet in Figure 3 was modeled using Aspen Plus. Detailed process flow diagrams and stream tables can be found in Figures S5−S7 and Table S3, respectively, in the Supporting Information. The model projects the compatibility of IPADM-2BOL with the conventional CCS infrastructure. We propose that the process operates analogous to the NETL Case 10 amine baseline (30% MEA) process, albeit with additional hardware for the PSAR antisolvent circulation loop, pumps, and coalescing system, as shown in Figure 3.18 A detailed flow sheet can be found in Figure S5. Another key distinction between the CO2BOL/PSAR process and the DOE’s Case 10 baseline is the introduction of a 13 MW refrigeration unit that would be downstream of the DCC. This refrigeration unit was introduced in our previous studies to provide extra dehumidification of the incoming flue gas to limit a 5 wt % steady-state loading of water in the circulating

Figure 15. Parity plot relating the CO2 recoveries calculated by the model to the experimental measurements in the cart.

expected to provide a good description of absorption columns using the CO2BOL solvent. 2.7. Solvent Lifetime and Corrosion Potential of CO2BOL Formulation. There was no apparent chemical degradation of IPADM-2BOL after the four months of continuous testing. Routine sampling of rich and lean solvents was performed. One milliliter (1 mL) aliquots were taken from both of the cart’s solvent holdup tanks during testing and subjected to 13C NMR characterization. Each spectrum was compared to that of the freshly synthesized material. The absence of chemical or thermal degradation was not surprising, because no oxygen was present in the gas streams to promote oxidation, nor was the stripper hot enough to allow thermal degradation. Structurally speaking, the central carbon in the guanidine base core is susceptible to hydrolysis by water. In testing, there was no evidence of any hydrolysis of the central carbon in the guanidine base core during testing. The lack of hydrolysis is likely due to the low temperatures used in the stripper. Also, the presence of the decane antisolvent in the PSAR during stripping would greatly reduce the polarity of the IPADM2BOL, which, in turn, would disfavor polar transition states and intermediates involved in a hydrolysis reaction and retard it. Note that any hydrolysis would produce the precursor imadazolidinone and amino alcohol, which could be recovered and used to resynthesize more IPADM-2BOL. Nevertheless, hydrolysis cannot be ruled out in a strictly thermal regeneration in application, although it appears that the low polarity of the fluid mixture in a PSAR regeneration process may retard hydrolysis.

Table 5. Summarized Projections for NETL Case 10, Compared to Three CO2BOL/PSAR Cases

rich solvent loading (mol CO2/mol solvent) temperature required for regeneration (°C) estimated reboiler duty (BTU/lb CO2) amount of CO2 removed (%) increase in net electric power over the MEA base case (%) estimated increase in levelized cost of electricity (LCOE) (%)

MEA base case (recreated NETL Case 10)

CO2BOL/PSAR (current formulation, operated at 356 cP max viscosity)

CO2BOL/PSAR (current formulation, operated at 578 cP max viscosity)

CO2BOL/PSAR (if 20 cP max viscosity formulation could be achieved)

0.49

0.28

0.34

0.50

120

104

104

85

1520

1107

965

870

90 0

90 7

90 9

90 16

87

119

not estimated

71a

a

Based on early capital estimates, from initial techno-economic analysis. Results from this simulation have not been validated experimentally. Furthermore, this case is based on achieving equilibrium conditions. H

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels solvent.18 Additional dehumidification in the model was considered, although it was concluded to be uneconomical, because of the extreme refrigeration cost and auxiliary draw to cool the flue gas any further. With the outlined process arrangement and infrastructure, energetic projections for several CO2BOL/PSAR cases were performed using Aspen Plus. Detailed power projections can be found in Table S4 in the Supporting Information. All modeled cases are summarized in Table 5, along with a recreation of DOE’s Case 10 500 MW subcritical power plant baseline. The recreation of the DOE baseline was performed as a means to standardize our model against a referenced standard. Our Case 10 recreation was found to be in close agreement with that in NETL’s baseline report.14 A description of the net power calculations can be found in the Supporting Information. The first two CO 2 BOL/PSAR cases correspond to projections based on kinetic models, while the third case is based on an equilibrium projection, assuming that any future CO2BOL formulations could achieve a 20 cP rich solvent viscosity at 40 °C. The 20 cP case is provided as a hypothetical solvent performance limit for this study. In all cases, the processes are configured to capture 90% CO2 off of Illinois No. 6 coal. The two kinetic cases are based on IPADM-2BOL where CO2 loading was limited to keep rich solvent viscosity from exceeding 356 and 578 cP, respectively. Figure 16 shows a

primarily due to the lower water content, which results in the lower condenser duty. Another component of the higher net power output is the lower reboiler temperatures available via the PSAR regeneration, where we previously had shown that intermediate-pressure steam can be run into a let-down turbine to provide extra power.18 This is clearly shown in Table 5, where lowering the reboiler temperature from 104 °C to 85 °C results in an increase net power output from the let-down turbine. In all cases, the energetic projections imply that CO2BOLs may be energy-efficient solvents for CO2 capture using conventional infrastructure. 2.9. LCOE Projections. Once the energetic performance was modeled, preliminary capital cost projections were made for the 20 cP equilibrium case and the 356 cP case using Fluor’s proprietary costs and vendor quotes of equipment sized from the Aspen Plus simulations. Solvent costing was projected at $35/kg, with an assumed makeup rate comparable to MEA, with respect to oxidative and thermal degradation and heatstable salt formation. A full description of the operational expenditures (OPEX) and capital expenditures (CAPEX) and the final levelized cost of electricity (LCOE) cost analysis are described in the Supporting Information, in Tables S5, S6, and S7, respectively. All equipment is off-the-shelf hardware, including the coalescing unit, which is a simple oil and water separator that is used in oil refining. The 578 cP case was not considered for cost analysis, because the authors concede that it would be more expensive for an LCOE analysis than the 356 cP case, and would not be needed. While we cannot provide the proprietary costs, we can confidently state that the estimated capital costs in the 356 cP case project to be more than twice that of the Case 10 amine baseline. The high capital cost is primarily due to high viscosity of the solvent, resulting in a larger absorber, larger pumps, and a larger cross exchanger. The final LCOE values are presented in Table 5. Using DOE’s NETL Case 10 as a guideline,14 we combined the energetic performance with the equipment costing, projecting an LCOE increase of 115% for the 356 cP case. Note that the capital costing was the key driver of the associated in the LCOE increase. If viscosity could be held to the desired 20 cP rich target, equipment costing would be comparable to the Case 10 baseline, and thus a 71% LCOE with currently available infrastructure, in this first pass assessment of the proposed process. The key conclusion from this comparison is that viscosity is the critical contributor for LCOE, primarily through capital costs. We emphasize that (1) equipment costing and energetic projections described here are preliminary and considered to be one of a kind, and (2) the solvent and process by no means are considered optimal. IPADM-2BOL was the first of potentially hundreds of molecules of this class of materials, and it is expected that other formulations will involve a lower rich viscosity and likely better performance and, in turn, lower capital costs. Furthermore, the proposed configuration used in this study may not be optimal. There are potential improvements in absorber design, packing choice, stripper configurations, and process conditions that have yet to be simulated or optimized. Process optimization (including examples such as the incorporation of a two-stage flash stripper that exploits higher stripper pressures via PSAR)24 have been considered, but were not performed at this time, because they are outside the scope of this initial feasibility study. Process optimization and testing alternative absorbers, strippers, and other unique configurations is the focus of future work in our laboratory. As

Figure 16. Effect of lean solvent loading on viscosity and reboiler temperature; the ratio of antisolvent to binding organic liquid (BOL) is 1:1.

breakdown of the impacts of lean solvent loading on the rich solvent viscosity and corresponding reboiler temperature to achieve that lean loading. The data clearly show that a viscosity restriction greatly limits the possible rich solvent CO2 loadings and thus increases the corresponding reboiler duty. A higher reboiler duty in the kinetic cases is expected, as a lower lean loading corresponds to higher reboiler temperatures and greater solvent circulation rates, as seen in Figure 16. The 356 cP case has a reboiler duty of 1107 BTU/lb CO2 (73% that of the amine baseline), while the more viscous 578 cP case has a higher CO2 loading, and a lower reboiler duty of 965 BTU/lb CO2 (63% that of the amine baseline). If a 20 cP target were achieved, the reboiler duties are projected to be as low as 870 BTU/lb CO2, which is 57% that of the amine baseline. In each case, it can also be seen that the projected reboiler duty is smaller than the Case 10 amine baseline, which translated to a higher net power output. The 356 cP case and 578 cP case are projected to have a significant 8% and 10% increases in net power, respectively, at an equivalent coal feed rate. If the 20 cP target could be achieved, a sizable 16% increase in net power could be theoretically achieved. The increases in net power for the CO2BOL/PSAR cases are I

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

factory calibrations for pure gases. Primary flow meter calibrators with 1% accuracy (Mesa Laboratories, Definer 220 series) were used to recalibrate mass flow controllers and to measure mixture gas flow rates directly. NMR analysis was performed using a Varian 300 MHz, 75 MHz 13C NMR spectrometer running Varian software. Stripper outlet gas concentrations were continuously measured with an infrared CO2 analyzer (Quantek Instruments, Model 902P). The absorber outlet gas was analyzed by a quadrupole mass spectrometer capable of atmospheric pressure sampling and corrosive gas handling (MKS Instruments, Cirrus 300 amu). The mass spectrometer was calibrated for CO2 and N2 prior to each experiment. IPADM-2BOL was loaded using a 4 L flask under a N2 atmosphere topped with a rubber septum. The septum was pierced with an 18 gauge needle connected with Tygon tubing to a N2 bubbler filled with silicon oil. Approximately 1.2 L of IPADM-2BOL was pumped into the cart via an integrated gear pump. 4.1. Custom-Built Solvent Cart. Continuous flow CO 2 absorption and stripping of IPADM-2BOL was tested on a custombuilt solvent cart made in house (see Figure S1 in the Supporting Information). A detailed description of the cart’s flow sheet (Figure S2) and its dimensions can be found in the Supporting Information. The solvent testing system is on two separate “carts.” The first cart is a utility cart that supplies feed and purge gases, electrical power, heat exchange fluids, and data acquisition functions. The second cart contains an absorber column and a stripper column, heat exchangers, coalescing unit, circulation pumps, and solvent reservoirs. In the batch absorption operation, a solvent is pumped via gear pump at a metered flow rate through the heat exchanger, then downward through the absorber column. At the same time, a given gas concentration at a metered flow rate was constantly fed upward into the absorber column while solvent flows down the column. The CO2 concentration at the gas inlet was controlled by feed gas mass controllers. The gas concentration and flow rate at the absorber column outlet was measured directly. The CO2 capture efficiency was calculated from the gas concentrations and gas flow rates by mass balance. Periodic sampling of the lean and rich liquid streams enabled 13 C NMR analysis to verify the liquid CO2 loading. The absorber and stripper loops had similar auxiliary equipment such as feed tanks, pumps, flow meters, level sensors, heat exchangers, and gas effluent conditioners. The column vessels were made of one or two identical sections of jacketed tubes (304 stainless steel, with an inside diameter of 2.875 in. and a length of 17 in.). In one configuration, the absorber had two sections and the stripper had one section. In another configuration, the absorber had one section and the stripper had two sections. Both columns contained a random packing (Cannon Instruments, 0.24 in. Pro-Pak, stainless steel 316). Liquid was introduced from the top through a 1/4-in.-inside-diameter tube above the packed bed center. The gas inlet at the bottom was a simple 90° bend, using tubing with an inside diameter of 1/2 in. The bottom of the packed bed was supported on a liquid redistributor tray. The column temperatures were controlled by heat-transfer fluids circulated through the column jackets and external baths. Temperature limitations on the heating bath for the stripper required the use of nitrogen stripping gas to enable stripping below 100 °C. The authors concede that such a configuration could not be used in a functioning plant, although the modeling and energetic projections were based on stripping performance using thermal means only, in the absence of a stripping gas. All viscosity (and some density) measurements were performed using an inline process viscometer with oscillating piston style sensors (Cambridge Viscosity, Inc., model VISCOpro 2000 with a flowthrough SPC-372 sensor head and density measurement software). This viscometer was chosen for the ability to flow sample through it continuously at a range of flows and temperatures while monitoring both the viscosity and density of the CO2BOL under absorption and desorption conditions. The viscometer was fitted with three interchangeable piston heads, to measure three ranges of viscosity. In some measurements, the viscometer was used to measure the hot− lean solvent, and in other measurements, it was used to measure the cold−rich solvent.

such, we consider this study to be representative of the potential for this solvent class and suggests that continued refinement of the solvent chemistry and process optimization could yield a promising carbon capture solvent system.

3. CONCLUSIONS Continuous flow testing of the IPADM-2BOL solvent was performed on a custom-built bench-scale cart. Testing was performed using thermal regeneration and added PSAR regeneration strategies. The testing confirmed that >90% CO2 capture is achievable at reasonable liquid/gas (L/G) ratios in the absorber in both thermal and PSAR cases. A key test finding was that high rich solvent viscosities (up to 900 cP) could run at steady-state CO2 capture, as liquid film masstransfer coefficients of CO2 were not greatly hindered by the high viscosity. Another key finding was no visible evidence of foaming during bench-scale testing and minimal evaporative losses or measurable degradation over 4 months of testing, even with a water loading of 5 wt %. Continuous PSAR operation was achieved using a coalescing tank that consistently and flawlessly separated the CO2BOL and antisolvent layers. The minimal carryover of the antisolvent into the absorption column showed little or no adverse impact to the absorption system but had a noticeable positive impact on stripper performance. Testing with 5 wt % water loading did not compromise anhydrous performance behavior; instead, it showed enhancement of CO2 capture performance. Modeling the kinetic data suggested that the current formulation must be operated at a lower lean solvent loading than its optimal thermodynamic performance range in order to keep viscosity to a manageable level. In turn, the configuration translates into higher solvent circulation rate and higher reboiler temperatures. The model projects that CO2BOLs with PSAR have the promise of presenting 8%−10% increases in net power with the potential for 16% if a theoretical 20 cP rich solvent viscosity could be achieved. Aspen Plus modeling of cart data projected the CO2BOLs/ PSAR system is compatible with the off-the-shelf infrastructure and hardware used in aqueous solvents, reducing the need for custom-built equipment. While conventional equipment can be used, the equipment is larger in size and more costly, because of the high circulation rates and viscosity, making this current formulation too costly for commercialization. IPADM-2BOL was the first of its kind studied, out of potentially thousands of CO2BOL molecules. IPADM-2BOL is considered to be representative of the solvent class and has not yet been optimized. There is potential for less-viscous formulations of CO2BOLs that may achieve higher performance projections and less-expensive capital costs. In closing, we conclude that the CO2BOL solvent class with PSAR regeneration, while costly, has potential for significant efficiency gains and merits further study. 4. MATERIALS AND METHODS Decane was purchased from Aldrich and dried over 3A molecular sieves prior to use. Compound IPADM-2BOL was synthesized by previous methods with a purity of 95% (13C NMR).17 All solvents were handled under nitrogen until CO2 loading. CO2 gas (99.99%) and N2 gas (99.998%) were purchased from Oxarc and used without further purification. The feed gases used for testing were set to the desired CO2 concentration by blending the bottled CO2 and N2 using mass flow controllers. The mass flow controllers (Brooks Instruments, SLA5800 series) were accurate within ±0.9% of the set point with J

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels

simulant flow was initiated as the solvent was pumped from the stripper feed tank, through the absorber column, and collected in the absorber feed tank. The solvent that had been loaded to a known level with CO2 was then heated through the stripper preheater and flowed down the stripper column. As the CO2 was released, the lean solvent was returned to the stripper feed tank. Flow rate and concentration measurements were made periodically on the gas products from both columns. Flow rates of the two pumps were adjusted to the desired solvent circulation flow while maintaining the feed tank liquid levels. The system was allowed to reach steady-state operation under each set of operating conditions before moving to the next. Both lean and rich streams were sampled periodically for offline analysis. In PSAR testing, a third pump controlling the decane flow rate was set to provide a desired ratio of antisolvent to IPADM-2BOL. For testing in the presence of water, 5 wt% deionized water was added (via massed additions) to the IPADM-2BOL solvent. Gas streams to the columns were then prehumidified to saturation at the respective column temperatures.

Two 1 L stainless steel reservoir tanks, one at the bottom of each column, served as the feed tanks to the solvent pumps. Liquid levels were measured using capacitive level sensors (Nivelco, Model CTA325-2) that were calibrated in the feed tanks prior to operation. The feed tanks were filled through a 3/4 in. fill port and were drained following testing at a low point drain valve. The absorber tank discharged into a Cole Parmer gear pump with a digital drive with a flow rate between 5 mL/min and 330 mL/min. The stripper tank was connected to the inlet of a Mahr gear pump with a variable frequency drive for speed control in the range of 22−220 mL/min. The solvent flow rates were measured using oval gear flow meters. Typical liquid flow rates were 150−200 mL/min. The hot−lean solvent stream and the cold−rich solvent stream passed through a cross heat exchanger (Exergy LLC, Model 00540-5) to recuperate heat and then was passed through separate trim heat exchangers (Exergy LLC, Models 00540-4 and 00540-1) to reach target test temperatures prior to entering the absorber and the stripper columns, respectively. The gas products from both the absorber and stripper were cooled to ∼4 °C before entering gas/liquid separators. Small slip streams were drawn from the outlet gas streams for online analysis and the rest was vented. The liquid condensates collected after the gas/liquid separators were returned to the feed tanks periodically via manual valves. Column operation was performed both manually and automatically. The control system was based on a National Instruments Compact RIO hardware platform and custom-built LabVIEW software. Batch operations with only one column were typically performed with manual control. Continuous operation of the absorber-stripper full loop was done with closed-loop temperature and liquid-level controls. The L/G ratio was manually adjusted to achieve the target CO2 capture efficiency. 4.2. Packing Selection. The cost of scale-up synthesis of the solvent was a major constraint to the scale of the CO2 absorption process testing in the laboratory. A smaller solvent inventoryand, thus, smaller columnshave a tendency to drive packing selection to high-efficiency laboratory packing, such as Sulzer EX. However, in our initial testing, the gauze-type structured packing showed reduced capacity for the CO2BOL solvents, which had about half of the surface tension as water. In the end, Pro-Pak, which is a random packing that is widely used for high-efficiency laboratory distillation, was selected for its robustness and lower cost. This choice should not be taken to project the suitable packing type for plant-scale processes. 4.3. Coalescer. The coalescer unit (Figure S4 in the Supporting Information) was modified from a small flow liquid/liquid coalescer with 316 stainless steel housing and PhaseSep cartridge (Pall Corporation, Part Nos. 1PH4F1F11 and LCS06H2AH). The mixture of lean CO2BOL and decane was fed by the lean solvent pump to the horizontally oriented coalescer and separated into two liquid phases. The heavier CO2BOL phase was collected in a small boot and discharged to the absorber column. The lighter decane phase was removed using a separate gear pump. The decane stream was combined in a static mixer (Cole Parmer, Model EW-04669-07) with the rich CO2BOL solvent stream from the absorber before entering the stripper column. The mole ratio of the decane antisolvent flow to the IPADM-2BOL solvent flow was set to 1:1 as identified from our prior modeling studies.15 Continuous operation of the coalescer unit showed facile separation of the CO2BOL and antisolvent (decane) phases. The calculated residence time for the CO2BOL and decane in the coalescing tank was 3 min. Visual observations of the two separate flows showed singlephase fluids from the coalescing tank, with the decane layer being clear, colorless, and nonviscous, with the CO2BOL layer being yellow/ orange and slightly viscous. Samples of the CO2BOL and decane layers were analyzed by 13C NMR, and their compositions were confirmed. 4.4. Representative Continuous-Flow Testing Procedure. In a representative continuous CO2 absorption test (in the absence of PSAR and water), a measured quantity of the solvent was placed into the stripper feed tank. Heat exchangers were allowed to reach their setpoint temperature prior to the initiation of solvent flow. The flue gas



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.energyfuels.5b02437. (PDF)



AUTHOR INFORMATION

Corresponding Author

*Tel.: (509)-372-6359. E-mail: [email protected]. Author Contributions

All authors contributed equally and have given approval to the final version of the manuscript. Funding

This work was funded by the U.S. Department of Energy’s Office of Fossil Energy (Award No. FWP-65872), and Award No. DE-0007466 managed by the National Energy Technology Laboratory. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors would like to dedicate this manuscript in memory of Jeri Bearden. The authors also acknowledge the contributions of Kriston Brooks and Greg Whyatt (PNNL) for discussions on the bench cart design, assembly, and retrofit and Maura Zimmerschied for technical editing. PNNL is proudly operated by Battelle for the U.S. Department of Energy.



REFERENCES

(1) Bottoms, R. R.Process for Separating Acidic Gases. U.S. Patent 1,783,901, Dec. 3, 1930. (2) Kohl, A. L.; Nielsen, R. B. Gas Purification; Gulf Publishing Company, Houston, TX, 1997. (3) Rochelle, G. T. Science 2009, 325, 1652. (4) Adams, D.; Davison, J.Capturing CO2; IEA Greenhouse Gas R&D Programme Report, 2007. (5) Cousins, A.; Wardhaugh, L. T.; Feron, P. M. H. Int. J. Greenhouse Gas Control 2011, 5, 605. (6) Bates, E. D.; Mayton, R. D.; Ntai, I.; Davis, J. J. Am. Chem. Soc. 2002, 124, 926. (7) Gurkan, B.; Goodrich, B. F.; Mindrup, E. M.; Ficke, E. L. E.; Massel, E.; Seo, S.; Senftle, T. P.; Wu, H.; Glaser, M. F.; Shah, J. K.; K

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX

Article

Energy & Fuels Maginn, E. J.; Brennecke, J. F.; Schneider, W. F. J. J. Phys. Chem. Lett. 2010, 1, 3494. (8) Qi, G.; Wang, Y.; Estevez, L.; Duan, X.; Anako, N.; Park, A. H.; Li, W.; Jones, C. W.; Giannelis, E. P. Energy Environ. Sci. 2011, 4, 444. (9) Wang, C.; Luo, H.; Jiang, D.; Li, H.; Dai, S. Angew. Chem., Int. Ed. 2010, 49, 5978. (10) Blasucci, V. M.; Hart, R.; Pollet, P.; Liotta, C. L.; Eckert, C. A. Fluid Phase Equilib. 2010, 294, 1. (11) Perry, R. J.; Davis, J. L. Energy Fuels 2012, 26, 2512. (12) Im, J.; Hong, S. Y.; Cheon, Y.; Lee, J.; Lee, J. S.; Kim, H. S.; Cheong, M.; Park, H. Energy Environ. Sci. 2011, 4, 4284. (13) Lail, M.; Tanthana, J.; Coleman, L. Energy Procedia 2014, 63, 580. (14) Cost and Performance Baseline for Fossil Energy Plants, Volume 1: Bituminous Coal and Natural Gas to Electricity, Revision 2; U.S. Department of Energy, Report No. DOE/NETL-2010/1397, 2010. (15) Jessop, P. G.; Heldebrant, D. J.; Li, X. W.; Eckert, C. A.; Liotta, C. L. Nature 2005, 436 (7054), 1102. (16) Heldebrant, D. J.; Yonker, C. R.; Jessop, P. G.; Phan, L. Energy Environ. Sci. 2008, 1, 487. (17) Privalova, E.; Nurmi, M.; Marañoń , M. S.; Murzina, E. V.; MäkiArvel, P.; Eränen, K.; Murzin, D. Yu.; Mikkola, J. P. Sep. Purif. Technol. 2012, 97, 42−50. (18) Mathias, P. M.; Afshar, K.; Zheng, F.; Bearden, M. D.; Freeman, C. J.; Andrea, T.; Koech, P. K.; Kutnyakov, I.; Zwoster, A.; Smith, A. R.; Jessop, P. G.; Nik, O. G.; Heldebrant, D. J. Energy Environ. Sci. 2013, 6, 2233. (19) Koech, P. K.; Zhang, J.; Kutnyakov, I.; Cosimbescu, L.; Lee, S. J.; Bowden, M. E.; Smurthwaite, T. D.; Heldebrant, D. J. RSC Adv. 2013, 3 (2), 566. (20) http://www.netl.doe.gov/research/coal/carbon-capture/postcombustion. (21) Mathias, P. M.; Zheng, F.; Heldebrant, D. J.; Zwoster, A.; Whyatt, G.; Freeman, C. M.; Bearden, M. D.; Koech, P. ChemSusChem, 2015, DOI: 10.1002/cssc.201500288. (22) ASPEN. (23) Nasrazadani, S.; Diaz, J.; Stevens, J.; Theimer, R. Corros. Sci. 2007, 49, 3024. (24) Madan, T.; Van Wagener, D. H.; Chen, E.; Rochelle, G. T. Energy Procedia 2013, 37, 386.

L

DOI: 10.1021/acs.energyfuels.5b02437 Energy Fuels XXXX, XXX, XXX−XXX