Piperazine/N-methylpiperazine/ N,N'-dimethylpiperazine as an Aqueous Solvent for Carbon Dioxide Capture

— Me´lange pipe´razine/N-me´thylpipe´razine/N,N’-dime´thylpipe´razine en solution aqueuse pour le captage du CO 2 — Un me´lange pipe´razine (PZ), N-me´thylpipe´razine (MPZ) et N,N’-dime´thylpipe´razine (DMPZ) en solution aqueuse est de´crit pour le captage du CO 2 par absorption-re´ge´ne´ration. Par rapport a` la pipe´razine concentre´e, ce me´lange permet un gain sur la solubilite´ des produits solides et sur la chaleur d’absorption. Aucune insolubilite´ n’a e´te´ observe´e a` fort taux de charge en CO 2 contrairement aux solvants a` base de pipe´razine. Ce me´lange a montre´ des performances e´quivalentes a` la pipe´razine concentre´e en termes de capacite´ et de cine´tique d’absorption du CO 2 , qui sont plus que double´es par rapport a` la traditionnelle MonoEthanolAmine (MEA) a` 7 M Abstract — Piperazine/N-methylpiperazine/N,N’-dimethylpiperazine as an Aqueous Solvent for Carbon Dioxide Capture — A blend of piperazine (PZ), N-methylpiperazine (MPZ) and N,N’-dimethylpiperazine (DMPZ) is described as a novel CO 2 capture solvent for aqueous absorp-tion-stripping. This blend provides improved solid solubility and heat of absorption compared to concentrated PZ. No insolubility was observed for regions of high CO 2 loading, unlike PZ solvents. The blend performed like concentrated PZ in terms of CO 2 capacity and CO 2 absorption rate, both of which were more than double that of a traditional 7 molal (30 wt%) MonoEthanolAmine (MEA). Thermal equilibrium was established between the three constituent amines that increases the thermal stability compared to traditional blended solvents. The primary drawback of this novel solvent system is enhanced amine volatility at absorber conditions compared with both concentrated PZ and MEA.


INTRODUCTION
Amine-based absorption-stripping is the most promising solution for post-combustion carbon dioxide (CO 2 ) capture from coal-fired power plant flue gas. Advanced amine solvents such as concentrated, aqueous piperazine (PZ) are being proposed to replace traditional alkanolamines such as MonoEthanolAmine (MEA) due to advantageous CO 2 absorption rates and overall energy performance [1,2]. A new solvent concept proposed here is the use of amines that demonstrate noticeable levels of overall thermal degradation but degrade to useful products that can react with CO 2 and maintain alkalinity of the solution. One example of this solvent concept presented here is the use of a blend of PZ, N-methylpiperazine (MPZ), and N,N'-dimethylpiperazine (DMPZ) [3].
Thermal degradation and oxidation are the two primary routes of amine loss during absorption-stripping along with volatilization of the amine in the absorber. Thermal degradation and oxidation of MEA and other alkanolamines have been investigated extensively previously and the impact of degradation was determined to be an important factor in solvent selection [4][5][6][7][8][9][10][11][12]. The thermal degradation of MEA generated products that have none or only limited CO 2 reactivity, effectively reducing the CO 2 capacity of the solvent. Other classes of amines have been found to generate CO 2 -carrying molecules as products that mitigate the overall effect of thermal degradation. For example, concentrated PZ solutions has been found to produce high concentrations of N-formylpiperazine (FPZ), N-(2-aminoethyl)piperazine (AEP) and ethylenediamine (EDA) during thermal degradation which can react with CO 2 directly (AEP and EDA), or be easily reacted to create PZ (FPZ and other amides) [13]. The same has been seen for N-methyldiethanolamine (MDEA)/PZ which degrades to diethanolamine (DEA) and substituted PZ molecules (1-methylpiperazine, N-(2-hydroxyethyl) piperazine and others) which can react and carry CO 2 during absorption-stripping [14,15].
The blend of PZ, MPZ and DMPZ offers a solvent that takes advantage of the mixture of multiple amines with CO 2 -carrying capacity that can interconvert to become degradation products of each other. The concept is to develop the solvent with a composition stable to thermal degradation that maximizes CO 2 capacity and absorption rate. In developing this solvent, the pertinent physical and thermodynamic properties of the blend that affect its expected performance in a CO 2 capture unit were investigated. Solid solubility, density and viscosity were evaluated over the expected range of CO 2 loading. CO 2 solubility, CO 2 capacity, heat of absorption, amine volatility and CO 2 absorption rate were studied over the pertinent range of lean and rich conditions expected. Finally, the thermal stability and thermal equilibrium of the amine blend were evaluated to determine the degradation potential of the solvent. In each case, the properties of the blend are compared to 8 molal (m) PZ, 7 m MEA, or both solvents to determine the advantages and disadvantages of the solvent compared to current baseline solvents.

Solid-Liquid Equilibrium Measurements
Solid-liquid equilibrium of blended amine solutions was measured through observation of the conditions under which precipitation occurred. For the amine blend of interest, solutions containing a range of CO 2 loading were prepared by combining unloaded and a maximum rich loaded solution. The maximum rich loaded condition was the highest CO 2 loading found to remain soluble at controlled temperature at or below room temperature. The transition temperature between solid and liquid phases was observed by slowly heating a solution at a given CO 2 loading containing precipitation and observing the phase change, as previously described [16]. All solutions were observed for any precipitation at temperatures between 0°C and the melting point of the solid phase of the amine solution.

Equilibrium Cell for Volatility
Amine volatility was measured in a stirred 1 L equilibrium cell coupled with a hot gas DX-4000 Fourier Transform Infrared Spectroscopy (FTIR) analyzer (Gasmet Technologies Inc., La Prairie, Canada) as shown in Figure 1 and described previously [19]. The insulated glass reactor was agitated at 350 rpm using an overhead stirrer and temperature in the reactor was controlled by recirculating dimethylsilicone oil. Vapor from the headspace of the reactor was pumped at 5 to 10 L/minute through a heated Teflon Ò line to the FTIR. Both the sample line and analyzer were maintained at 180°C to prevent liquid condensation or adsorption of amine. The FTIR quantified amine, CO 2 , and water concentration in the vapor phase. After the gas was analyzed in the FTIR, it was returned to the reactor through a second heated line maintained 55°C hotter than the reactor temperature. The 55°C difference was sufficient to ensure that the returned gas did not upset the liquid-vapor equilibrium inside the reactor and to prevent potential heat loss at the bottom of the reactor.

High Temperature VLE Apparatus for CO 2 Solubility
A high temperature vapor-liquid equilibrium apparatus was used to measure CO 2 solubility in aqueous amine solutions. A 500 mL stainless steel autoclave with an agitator and a heating jacket was used as the equilibrium cell. The method details have been described previously [20]. Total pressure of the CO 2 loaded blend at 100-160°C was measured with ± 2.4 kPa accuracy. CO 2 partial pressure was derived by subtracting partial pressure of N 2 and water from the equilibrium total pressure. Data for the partial pressure of water were taken from the DIPPR database and was assumed to follow Raoult's Law [21]. The partial pressure of nitrogen at the experimental condition was calculated using a low temperature pressure measurement and the assumption that nitrogen behaves as an ideal gas. Liquid samples were collected before and after each experiment at room temperature, analyzed for amine concentration by acid titration and total CO 2 by TIC analysis, and corrected for the gas phase CO 2 at each condition to get CO 2 loading at high temperature.

Wetted Wall Column for Absorption Rates and CO 2 Solubility
The wetted wall column used in this study to characterize the mass transfer coefficient of amine solution (k' g ) is the same apparatus that had been described previously by Cullinane et al. and Dugas et al. [22,23]. In a typical experiment, a gas mixture of nitrogen and CO 2 counter currently contacted the amine solution under study. The flux of CO 2 (N CO 2 ) between gas and liquid phases for any given condition was calculated from the difference between inlet and outlet CO 2 concentrations quantified using an IR analyzer. By varying the inlet CO 2 partial pressure, both absorption and desorption were observed. The equilibrium CO 2 partial pressure of the amine solution, P* CO 2 ;l , was calculated using absorption and desorption data to estimate the pressure where zero flux occurs. The overall mass transfer coefficient (K G ) was calculated as the ratio of the N CO 2 and the gas phase partial pressure driving force between the inlet CO 2 partial pressure, P CO 2 ;g and P* CO 2 ;l , as shown in Equation (1) : The series resistance correlation for gas absorption into liquid is shown in Equation (2). The liquid film mass transfer coefficient based on a partial pressure driving force, k g ', correlates with the experimental conditions and geometry of the apparatus as established above (2): The values for k g were estimated from a previous correlation [1]. Data for the kinetic absorption rate and CO 2 solubility were obtained simultaneously with this technique.

Thermal Cylinders for Thermal Degradation
Thermal degradation of amine solutions was assessed using thermal cylinders as described previously [4,13,18]. Cylinders consisted of (1/2)-inch OD 316 stainless steel tubing with two stainless steel Swagelok Ò fittings and end caps (Swagelok Company, Solon, OH, USA). Cylinders were filled with 10 mL of the amine solution of interest, sealed according to Swagelok Ò specifications, and placed in forced convection ovens maintained at the experimental temperature. Individual cylinders were removed from the oven periodically and the liquid contents were analyzed for total alkalinity and CO 2 concentration, and analyzed using cation ion chromatography to quantify the parent amines and cationic degradation products.

Cation Ion Chromatography (IC)
Cation Ion Chromatography (cation IC) was used to quantify PZ, MPZ, DMPZ and other cation degradation products in solution. A Dionex ICS-2500 Ion Chromatography System with AS autosampler was used as described previously (Dionex Corporation, Sunnyvale, CA, USA) [4,13,18]. Separation occurred in an IonPac GC17 guard (4 9 50 mm) and CS17 analytical column (4 9 250 mm) with varying concentrations of methanesulfonic acid in analytical grade water and conductivity detector cell. The transition temperature of this blend at 0 to 0.39 mol CO 2 /mol alkalinity was observed in order to quantify the solubility envelope where an aqueous amine solution is known to exist. The transition temperature over this range of CO 2 loading for 3.75 m PZ/ 3.75 m MPZ/0.5 m DMPZ is shown in Figure 2. The transition temperatures of 8 m PZ are included in the figure for comparison purposes [18]. For the blend, the temperature at which any precipitation in the solution completely melted or the liquid precipitated into a solid was obtained and the transition point recorded.

RESULTS AND
The blended system has a larger solubility envelope and this solvent is predicted to have fewer solubility issues than a concentrated PZ system. In CO 2 -loaded, concentrated PZ, solid PZ hexahydrate (PZÁ6H 2 O) is formed where the loading is less than 0.22 mol CO 2 /mol alkalinity and the temperature is less than 25°C [13]. The blend contains less PZ and PZÁ6H 2 O, the expected solid formed in the blend in the same region of Figure 2, precipitates at lower temperature and CO 2 loading. This behavior expands the solubility envelope and usefulness of the blend.
At CO 2 loading greater than 0.43 mol CO 2 /mol alkalinity, 8 m PZ was found to form hydrated protonated PZ carbamate (H + PZCOO À ÁH 2 O) when precipitation occurred [16]. No solids precipitated in 3.75 m PZ/ 3.75 m MPZ/0.5 m DMPZ at any of the high CO 2 concentrations tested. It appears that at least 4 m PZ is required to form significant concentrations of H + PZCOO À , so this species is not present to precipitate in its hydrated form in the blended solvent. in Table 1 with CO 2 concentration reported in units of CO 2 loading and mol/kg. The density and viscosity data are compared to that of 8 m PZ at 40 and 60°C in Figure 3 and Figure 4, respectively. The data demonstrate the expected trends of increased density and viscosity with increasing CO 2 loading and decreasing temperature. In comparison with 8 m PZ, the blend demonstrates a lower density and higher viscosity at a given CO 2 loading at 40 and 60°C. The higher viscosity of the blend, approximately 25% higher across the board, is a slight disadvantage of this solvent as it is expected to decrease mass transfer and reduce heat transfer coefficients. The viscosity of 5 m PZ/2 m MPZ/1 m DMPZ is correlated well in terms of the viscosity of water obtained from DIPPR, l water , CO 2 concentration in mol/kg, C CO 2 and temperature in Kelvin, T [21]. The details of the regression and parameters are shown in Equations (3,4) and Table 2. The regression was able to fit all data  The CO 2 partial pressure, P CO 2 , of 3.75 m PZ/3.75 m MPZ/0.5 m DMPZ was measured from 40 to 160°C for a range of CO 2 loading. Data were obtained from the  equilibrium cell at 40°C, the wetted wall column for 40 to 100°C, and the high temperature VLE apparatus from 100 to 160°C. The total pressure, P total , was obtained from the high temperature measurements and the P CO 2 was estimated as described above. The CO 2 solubility values measured directly from the equilibrium cell and wetted wall column are compared in Table 3. The estimated P CO 2 and P total measurements are shown in Table 4. The CO 2 solubility measured from 40 to 160°C is shown in Figure 5 as the P CO 2 plotted against the CO 2 loading. Data from three separate experimental apparatuses, the wetted wall column, high temperature VLE apparatus and equilibrium cell, are shown and demonstrate strong agreement that provides confidence in the data. An empirical model for the prediction of P CO 2 was developed using the data from 40 to 160°C (excluding the single point from the equilibrium cell). The regressed form of the model is shown in Equation (5) where the natural log of P CO 2 in Pa is predicted as a function of CO 2 loading, a, and temperature in K, T.
The uncertainty of the regression is demonstrated by the error for each of the three regressed coefficients:

CO 2 Capacity
The intrinsic CO 2 capacity of 3.75 m PZ/3.75 m MPZ/0.5 m DMPZ was estimated assuming lean loading (0.23 mol CO 2 /mol alkalinity) and rich loading (0.33 mol CO 2 /mol alkalinity) that correspond to P CO 2 values at 40°C of 0.5 and 5 kPa, respectively. These two pressure values are estimates of the P CO 2 that would provide an appropriate driving force at the bottom or top of the absorber [1,24]. With the appropriate unit conversions, the capacity of the blend is 0.88 mol CO 2 /kg (amine + water). This compares to an intrinsic CO 2 capacity of 0.79 and 0.47 mol CO 2 /kg (amine + water) for 8 m PZ and 7 m MEA, respectively [25,26].

Estimation of Heat of Absorption
The heat of CO 2 absorption (DH abs ) in 3.75 m PZ/ 3.75 m MPZ/0.5 m DMPZ was estimated from the derivative of Equation (5) according to the Gibbs-Helmholtz Equation (Eq. 6) and is given by Equation (7). Over the expected CO 2 loading range, 0.23 to 0.33 mol CO 2 /mol

Amine Volatility
The volatility of amines in 3.75 m PZ/3.75 m MPZ/ 0.5 m DMPZ was measured at 40°C and a nominal lean loading of 0.23 mol CO 2 /mol alkalinity, the conditions expected at the top of the absorber. The conditions that exist at the top of the absorber are the most important for volatility comparisons in order to assess potential fugitive losses to the atmosphere. For this solvent, this condition resulted in a P CO 2 of 287 Pa and partial pressure of water, P H 2 O , or 5 875 Pa. The measured amine volatility of each of the three amine components in the blend is presented in Table 5 as the partial pressure of each amine (P PZ , P MPZ and P DMPZ ). The amine volatility for 8 m PZ and 7 m MEA are also shown for comparison purposes [19]. The CO 2 loading for these two baseline systems were also chosen to mimic the operating conditions expected at the top of the absorber. Based on the results, it is clear that methylation increases volatility. The volatility of MPZ and DMPZ are significantly higher than that of PZ in the blend and by itself, suggesting a significant impact of methylation on the behavior of the amino function. Even though DMPZ is present only in a small concentration, 0.5 m, its volatility is the highest and 65 times greater than PZ. The fact that this molecule has two methyl groups renders it fairly nonpolar in the polar environment consisting of water and electrolytes. Therefore, DMPZ has  unfavorable interactions with the surrounding species and is highly volatile. MPZ also suffers the effects of a nonpolar methyl group but to a lesser extent than DMPZ. This observation highlights the fact that degradation products can be more volatile than the parent amine itself, even at small concentrations, an important fact in solvent selection.

CO 2 Absorption Rate
The liquid film mass transfer coefficient (k g ') for 3.75 m PZ/3.75 m MPZ/0.5 m DMPZ was measured from 40 to 100°C at a range of loadings ( Fig. 6 and Tab. 3). The use of P CO 2 at 40°C as the abscissa allows direct comparison of rate data at different temperature as well as for different amine solvents on the same basis without influences from differences in CO 2 loading. The range of P CO 2 from 0.5 to 5 kPa, as used previously for capacity calculations, represents a reasonable driving force between gas and liquid at the top and bottom of an absorber expected during the treatment of flue gas from coal-fired power plant. At 0.5 kPa and 40°C, the expected conditions at the lean end of the absorber, the kg' of the blend is 2.0 9 10 À6 mol/s.Pa.m 2 compared to 2.1 9 10 À6 and 7.6 9 10 À7 mol/s.Pa.m 2 for PZ and MEA, respectively. The temperature range expected for the absorber is 40 to 60°C. In this temperature range, the rate for the blend stays relatively constant, approximately equal to that for 8 m PZ and over 2.5 times faster than 7 m MEA. Temperature above 60°C leads to a decrease in the mass transfer rate of the blend, as has been observed for other amine systems [24][25][26]. The blend performed as well as concentrated PZ and represents another solvent that can outperform the baseline MEA solvent.

Thermal Degradation of the PZ-Based Blends
The thermal behavior of PZ/MPZ/DMPZ was studied at 150°C. This temperature represents the expected temperature of the stripper and reboiler section of an absorber-stripper system utilizing a thermally stable solvent in order to maximize high temperature energy savings.
Six PZ-based solvent blends were investigated for their thermal stability. All blends are compared to 8 m PZ as the baseline case of a highly stable solvent. In the blends tested, the total alkalinity of the solvent was maintained at 8 m, while PZ was replaced with methylated PZ molecules (MPZ or DMPZ). Blend compositions are indicated as the concentration of PZ/MPZ/DMPZ, all in units of m. For example, one two-component blend of 4 m PZ/4 m MPZ, or 4/4/0, was investigated. Five three-component blends (5/2/1, 5/2.5/0.5, 5/1.5/1.5, 3.9/3.9/0.2, and 3.75/3.75/0.5) were also investigated.  The overall rate of thermal degradation was assessed using a first order rate constant of amine loss, k 1 , as described in detail previously [13,27]. For comparing the blends, the rate was determined based on the total amine loss as a sum of PZ, MPZ and DMPZ rather than the rate of loss of the individual amine species. This was done to allow an overall comparison of the blends on a similar basis. The k 1 values for thermal degradation at 150°C of the 8 m PZ and the blends are compared in Table 6. The k 1 values of the blends vary from 8.6 9 10 À9 to 7.5 9 10 À8 s À1 and are substantially less than the comparable value for 7 m MEA with 0.4 mol CO 2 /mol alkalinity degraded at 150°C of 8.3 9 10 À7 s À1 [4].
Blending of PZ with MPZ and DMPZ also reduces solvent stability as shown with the high k 1 values for all blends studied. The changes in the rate constant are small for most of the blends with the exception of the 5/1.5/1.5 solvent, so the slight increase in degradation may be balanced by the improvement in solid solubility.

Thermal Equilibrium of PZ, MPZ and DMPZ
A thermal equilibrium was found to establish itself in the presence of methylated PZ. In the degradation of 8 m MPZ, PZ and DMPZ were generated as products through a S N 2-type substitution or arm-switching reaction. Observations of this system indicate it tends toward an equilibrium set of concentrations as the species interconvert to produce a solvent that is thermally stable at its equilibrium concentrations. The thermal equilibrium established between PZ, MPZ, and DMPZ was studied in the thermal degradation experiments performed at 150°C described in the previous section. At this temperature, some degradation was expected to occur, but the interconversion of PZ, MPZ and DMPZ was assumed to occur more rapidly than degradation processes based on the slow degradation rates discussed in the previous section.
The expected equilibrium reaction of MPZ is as follows, where the brackets indicate the concentration of the given species in units of mmol/kg (Eq. 8). The equilibrium constant, K eq , of Equation (8) can then be calculated (Eq. 9): The goal of this K eq analysis is to determine the concentrations of the three amine species in equilibrium at 150°C. These concentrations would be what were expected to be stable in a system with a reboiler temperature of 150°C. The K eq values for the thermal equilibrium of PZ, MPZ, and DMPZ are shown in Figure 7. For each blend, the K eq was calculated at each time point according to Equation (9) and plotted against degradation time.
All of the experiments, including those with limited data, were found to tend toward a K eq value of 0.27-0.30, indicating the equilibrium concentrations of amine expected at 150°C. The experiments containing 3.9/3.9/0.2 and 3.75/3.75/0.5 demonstrate this tendency toward a K eq of 0.29. Experiments with initial amine concentrations far from equilibrium, such as the 5/1.5/1.5 blend, demonstrated a higher thermal degradation rate as the amines reacted toward equilibrium, as shown in Table 6.
Degradation of PZ alone will not yield these results as MPZ and DMPZ are only minor products [13]. However, when a blend begins with at least some N-methylated PZ (i.e., MPZ or DMPZ), these methyl groups can easily  undergo a straightforward nucleophilic attack and disproportionation, resulting in switching of the methyl groups between molecules. Overall, the data suggest that any system starting with MPZ will tend toward a K eq of 0.28 to 0.30 at 150°C.

Overall Comparison of Blends
A comparison of the important parameters for CO 2 capture solvents is presented in Table 7 for 7 m MEA, 8 m PZ, and 3.75 m PZ/3.75 m MPZ/0.5 m DMPZ. The CO 2 loading range when the solvent is completely soluble at 20°C is shown based on solubility data. The improved solubility of both MEA and the blend is a strong advantage over 8 m PZ. The CO 2 capacity was calculated for each system as described in Section 2.2.2 based on CO 2 solubility at P CO 2 of 0.5 and 5 kPa. The blend and 8 m PZ have approximately double the CO 2 capacity as MEA, an important advantage. The average k' g , k' g avg , based on a log-mean flux and log-mean driving force across a typical absorber design, as described in detail previously, is compared at 40°C [26]. Both PZ and the blend have k g ' avg values double that of MEA. Empirical regressions for the DH abs are shown for each solvent based on CO 2 solubility regressions developed for MEA and PZ previously or for the blend as described in this manuscript [20]. The DH abs was calculated from the empirical regressions at a midpoint in the expected loading range for each amine system. The midpoint was determined to be at a P CO 2 of 1.5 kPa, which corresponded to CO 2 loadings of 0.49, 0.35, and 0.28 mol CO 2 /mol alkalinity for the MEA, PZ and blended systems, respectively. Error in the calculation of this value was assessed based on the confidence of each regressed coefficient. The larger DH abs for the blend over PZ is a crucial advantage that will reduce overall energy performance.
Volatility was reported as the overall amine volatility of each solvent at 40°C at the expected lean loading as this should correspond to the top of the absorber in a real system. For the blend, this is a sum of the partial pressure of PZ, MPZ and DMPZ at this condition. The volatility of the blend is significantly higher than the two baseline amines. Implementation of this amine system will require an adequately designed water wash system to avoid amine losses from the absorber.
The thermal stability of each solvent is expressed as the k 1 for thermal degradation at 150°C. Data for MEA, PZ and the blend were obtained at 0.4, 0.3 and 0.3 mol CO 2 /mol alkalinity, respectively. The reported value for the blend is based on total amine loss from PZ, MPZ and DMPZ. Both PZ and the blend are significantly more thermally stable than MEA as demonstrated by the large k 1 value of MEA.

CONCLUSIONS
A blend of PZ, MPZ and DMPZ is an example of a new solvent design based on the thermal equilibrium of multiple amines useful for CO 2 capture. Blends of this nature reach a thermal equilibrium at reboiler temperature where the overall thermal degradation rate is low while multiple species exist that are valuable to the CO 2 capture process. 3.75 m PZ/3.75 m MPZ/0.5 m DMPZ has been fully evaluated for this application.
In comparison with 8 m PZ, this blend was found to have advantageous solid solubility and DH abs , with approximately the same behavior in terms of CO 2 capacity, CO 2 absorption rate and thermal degradation as shown in Table 7. Disadvantages of the blend include mildly increased viscosity and substantially greater amine volatility as DMPZ. Both of these properties are important to the optimized design of a CO 2 capture system and would need to be addressed. A carefully designed water wash system would be needed to address the amine volatility. As with PZ, the blend was found to have significantly better CO 2 capacity, CO 2 absorption rates and thermal stability compared to 7 m MEA.