E ﬀ ect of Various Parameters on the Thermal Stability and Corrosion of CO 2 -Loaded Tertiary Amine Blends

: In this study, the thermal stability and corrosivity of various CO 2 -loaded tertiary amine blends in both aqueous and non-aqueous form in stainless steel cylinders were studied for combined acid gas and water removal. The thermal stability was measured from the remaining amine concentration and the corrosivity was measured from the amount of various metals in blends using titration and inductively coupled plasma mass spectrometry (ICP-MS), respectively. The experimental data were used to calculate the rate constants of amine group loss. The developed model represented the experimental data very well. Solvent change from H 2 O to triethylene glycol (TEG) in blends decreased the thermal stability and vice versa for corrosivity. The amine stability was increased when contact with stainless steel was reduced. An increase in the amine concentration or CO 2 loading at constant temperature decreased the thermal stability and vice versa for corrosivity.


Introduction
Acid gas impurities such as carbon dioxide (CO 2 ) and hydrogen sulfide (H 2 S) are typically present in natural gas. Energy production from natural gas also produces large quantities of CO 2 during combustion [1]. H 2 S is naturally present with methane, water vapor, CO 2 , and other impurities in many natural gas reservoirs [2]. Water vapor can cause corrosion in the pipelines and can also form hydrates in the presence of methane at suitable temperatures and pressures. These hydrates can plug manifolds, pipelines, and various accessories, such as valves and fittings [3]. Absorption-based acid gas capture processes are the most versatile processes for the removal of CO 2 and H 2 S from natural gas [4]. Amine solutions are generally used for this type of process, and glycols such as monoethylene glycol (MEG) and triethylene glycol (TEG) are typically used to remove H 2 O and inhibit hydrate formation [5].
A process that can remove acid gases and water simultaneously from the natural gas stream can be beneficial and using this process for subsea applications will reduce the environmental footprint. Simultaneous acid gas and water removal was discussed as early as 1939 by Hutchinson [6] and later by McCartney [7] and Chapin [8]. An ethanolamine (MEA) and diethylene glycol blend was the first used blend, but it was discontinued due to higher amine degradation and severe corrosivity [4]. MEA is mostly used for CO 2 absorption processes, but tertiary amines are alternatives for CO 2 absorption due to their higher theoretical CO 2 loading capacity and low energy requirement during the regeneration process [9]. Additionally, tertiary amine increases the thermal stability and reduces corrosivity [10]. our other previous work, for 30 wt.% MDEA blends, the metal concentrations were in the order of MDEA.TEG > MDEA.H 2 O > MDEA.MEG [10].
The stability and corrosivity of amine are important factors to consider while selecting a blend for a combined acid gas and water removal process. In our previous work [43], we identified various potential blends for combined H 2 S and water removal. Among others, DEEA, 3DMA-1P, and 3DEA-1P showed good acid gas absorption, while MDEA is currently used as an industrial standard amine for H 2 S removal. As mentioned previously, most investigations of thermal stability in amine solutions have been conducted for aqueous and water lean solutions. There are very little data available for the thermal stability of amine blends in glycols without any water present in blends. Similarly, the effects of the initial amine concentration and CO 2 loading were previously only investigated for the thermal stability of amine, and no data are available for their effects on corrosion. Hence, in the current work, four tertiary amine blends in water, MEG, and TEG were studied to investigate their thermal stability and corrosivity. The effects of solvent type, amine concentration, and CO 2 loading were investigated. Finally, the impact of metal on MDEA blends' thermal stability was investigated by running thermal degradation experiments in glass tubes and stainless steel cylinders. The chemical structures of all the chemicals used in this study are shown in Figure 1.
Energies 2020, 13, x FOR PEER REVIEW 3 of 18 1P showed good acid gas absorption, while MDEA is currently used as an industrial standard amine for H2S removal. As mentioned previously, most investigations of thermal stability in amine solutions have been conducted for aqueous and water lean solutions. There are very little data available for the thermal stability of amine blends in glycols without any water present in blends. Similarly, the effects of the initial amine concentration and CO2 loading were previously only investigated for the thermal stability of amine, and no data are available for their effects on corrosion. Hence, in the current work, four tertiary amine blends in water, MEG, and TEG were studied to investigate their thermal stability and corrosivity. The effects of solvent type, amine concentration, and CO2 loading were investigated. Finally, the impact of metal on MDEA blends' thermal stability was investigated by running thermal degradation experiments in glass tubes and stainless steel cylinders. The chemical structures of all the chemicals used in this study are shown in Figure 1.

Materials
All chemicals used in this study were purchased from Sigma-Aldrich (Oslo, Norway), except for 3DEA1P and CO2. 3DEA1P was bought from TCI Europe (Zwijndrecht, Belgium) and CO2 was obtained from AGA AB (Oslo, Norway). All the chemicals used are listed in Table 1 and were used without further purification.

Materials
All chemicals used in this study were purchased from Sigma-Aldrich (Oslo, Norway), except for 3DEA1P and CO 2 . 3DEA1P was bought from TCI Europe (Zwijndrecht, Belgium) and CO 2 was obtained from AGA AB (Oslo, Norway). All the chemicals used are listed in Table 1 and were used without further purification. Amine solutions were prepared gravimetrically in flasks using the Mettler Toledo scale, model MS6002S/01 (Oslo, Norway) (±0.01 g). Solutions were loaded with CO 2 by using a continuously weighted gas wash bottle in an isolated environment to achieve the required CO 2 concentration. All solutions were titrated for amine and CO 2 concentration to verify the CO 2 loading prior to the start of the experiment [10,44,45]. The unloaded amine wt.% and CO 2 loading at the beginning of the experiment for all solutions are given in Table 2.

Methodology
Both stirred cell reactors [32,33,36,39] and metal cylinders [10,21,28,37] have been used to study amine solutions' thermal stability. In this work, metal cylinders were selected, as various amine solutions at different concentrations could be tested in multiple cylinders at the same time, consequently reducing the overall experimental duration. The corrosivity of the blends can be studied by analyzing the solutions for the total Fe, Cr, Ni, and Mo. The metal concentrations in the solution can be used as an indicator of corrosivity [42]. Therefore, multiple cylinders were made by using 10 cm long and 1.3 cm diameter 316 stainless steel tubes, with both ends closed with Swagelok ® end caps (Stavanger, Norway). Pre-loaded CO 2 amine solutions were added in cylinders, leaving around 1 cm gaps between the solution surface and cap. Additionally, glass tubes were used inside metal cylinders to study the effect of metal on the thermal stability of amine solutions for 30 wt.% MDEA solutions (both aqueous and non-aqueous). A replica of each cylinder was made to attain more precise results. Cylinders were stored vertically in a forced convection oven at 135 • C, in order to enhance the degradation and shorten the experimental time. Since several studies using these conditions have been performed previously, the current data can be easily compared with literature data [10,28,35,37]. The experiment was run for seven weeks, and one cylinder, along with a replica of each solution, were taken out from the oven after 1, 3, 5, and 7 weeks at the same time of day. Cylinders were cooled down to room temperature before opening in the fume cabinet and were stored in the fridge to prevent any further decomposition. Cylinders were weighed before and after the experiment to identify any leakages. All cylinders, except for aqueous MDEA solutions in glass tubes for week 3 and week 5, showed less than a 2% difference in weight. The cylinders that leaked, including aqueous MDEA solutions in glass tubes for week 3 and week 5, were not used in the calculations of results.
All solutions were titrated with 0.1 M H 2 SO 4 to measure the amine group in the solutions [37,44]. For all samples, the average difference in amine concentrations was less than ±2% between the solution and its replica in all non-leaking cylinders. In this paper, the average concentrations between the replicates are reported. At the end of the experiment, cation ionized chromatography (IC) was also used to determine the amine concentration of 30 wt.% solutions (α = 0.1) and the difference between the amine group concentration calculated by titration and cation IC was less than ±2.5 percentage points. The average of the total amine group concentration measured by titration was used in Equation (2) to calculate the experimental values of the remaining amine group (%). In this equation, C i is the amine group concentration at time = t (days), and C o is the amine group concentration in fresh solution at the start of the experiment. Metal concentrations (Fe, Cr, Ni, and Mo) were measured using inductively coupled plasma mass spectrometry (ICP-MS) for only week five solutions and used as an indicator of the relative corrosivity caused by amine solutions on the stainless steel [10,28,37,42]. The primary focus of the current work was to study the effect of TEG/MEG in the absence of water, amine concentration, and CO 2 loading on the stability and corrosivity of amine. Therefore, by-products produced due to amine loss were not studied.

Remaining amine group
Modeling Rate constants of amine group loss can be calculated with a linear regression model by using Equations (3) and (4) while assuming first-order reactions concerning amine and integrated over time "t" [37,46].
Plotting ln (C o /C i ) as a function of time from Equation (4) provided the first-order rate constants (k) as the slope in linear regression, while the correlation coefficient (R 2 ) gave the accuracy of the model. Rate constants were used in Equation (5) to calculate (C i /C o ), which were subsequently used to predict the remaining amine group (%) from Equation (2). First-order rate constants (k), the experimental duration (time "t"), and R 2 are given in Table 2.

Thermal Stability
The validation of degradation results is not simple. In the laboratory, the degradation is accelerated to shorten the experimental time and small differences in solvent compositions or experimental conditions can give large differences in the results. Of the solutions studied in this work, the thermal degradation of aqueous DEEA and MDEA has previously been studied and there is adequate agreement found between the current study and published data, as shown in Figure 2. For aqueous DEEA solutions, Gao et al. [39] (35 wt.% DEEA with α = 0.5) showed 4 percentage points and 7 percentage points higher amine loss after 12.5 and 20 days, respectively, probably due to the higher amine concentration and higher CO 2 loading. Eide-Haugmo [35] (30 wt.% DEEA with α = 0.5) showed 11 percentage points higher amine loss after five weeks at 135 • C in comparison to our current data, which can be partly explained by the higher loading. Finally, the new data are in good agreement with our previous data, showing a good reproducibility [10]. It can be concluded that there is a good overall agreement between the published literature data and degradation data reported in this paper.
Energies 2020, 13, x FOR PEER REVIEW 6 of 18 The validation of degradation results is not simple. In the laboratory, the degradation is accelerated to shorten the experimental time and small differences in solvent compositions or experimental conditions can give large differences in the results. Of the solutions studied in this work, the thermal degradation of aqueous DEEA and MDEA has previously been studied and there is adequate agreement found between the current study and published data, as shown in Figure 2. For aqueous DEEA solutions, Gao et al. [39] (35 wt.% DEEA with α = 0.5) showed 4 percentage points and 7 percentage points higher amine loss after 12.5 and 20 days, respectively, probably due to the higher amine concentration and higher CO2 loading. Eide-Haugmo [35] (30 wt.% DEEA with α = 0.5) showed 11 percentage points higher amine loss after five weeks at 135 °C in comparison to our current data, which can be partly explained by the higher loading. Finally, the new data are in good agreement with our previous data, showing a good reproducibility [10].

Corrosion
The corrosion results are in agreement with previously published data from various authors, as shown in Figure 3. Shoukat et al. [10] also found similar total metal concentrations and a similar order for MDEA blends as the current work. Furthermore, the new aqueous MDEA and DEEA data are also in acceptable agreement with Eide-Haugmo [35]. Individual metal ion concentrations display very small quantities (<67 mg/L), both in our current data set and previously published data sets. It can be concluded that the current data set is in good agreement with the literature data.

Corrosion
The corrosion results are in agreement with previously published data from various authors, as shown in Figure 3. Shoukat et al. [10] also found similar total metal concentrations and a similar order for MDEA blends as the current work. Furthermore, the new aqueous MDEA and DEEA data are also in acceptable agreement with Eide-Haugmo [35]. Individual metal ion concentrations display very small quantities (<67 mg/L), both in our current data set and previously published data sets. It can be concluded that the current data set is in good agreement with the literature data.

Thermal Stability
The thermal stability of amines is presented by the remaining amine group (%) (total remaining alkalinity) as a function of experimental days. The remaining experimental amine group % for each solution is shown as markers, while the solid lines are the values calculated from linear regression. All the experimental data of amine group concentrations is given in Table A1. Rate constants were computed for all solutions until 49 days, except MDEA.H2O (Glass) solutions, where the solutions leaked after week one. The week one results of 50 wt.% DEEA.H2O were also not included in linear regression. Unloaded amine wt.% and CO2 loading at the start of an experiment for all solutions, along with the experimental duration, rate constants (k), and correlation coefficients (R 2 ) for all blends, are given in Table 2. The rate constant increased with the increase in amine concentration and CO2 loading in aqueous solutions. The change of solvent had a different effect on each amine solution rate constant. The correlation factor (R 2 ) value was higher than 0.9 in all solutions, showing the good accuracy of the model, and is visible in the thermal stability results (Figures 4-6 & 10-12). Figure 4 shows the thermal stability results of 30 wt.% aqueous amine solutions with CO2 loading of 0.1. After seven weeks, the maximum thermal stability was observed in 3DMA-1P. Overall, the thermal stability was quite high, as less than 2 percentage points of amine were lost during the whole set of experiments. No significant difference in thermal stability was observed when the methyl group was swapped with an ethyl group in solutions, i.e., 3DMA-1P to 3DEA-1P.

Thermal Stability
The thermal stability of amines is presented by the remaining amine group (%) (total remaining alkalinity) as a function of experimental days. The remaining experimental amine group % for each solution is shown as markers, while the solid lines are the values calculated from linear regression. All the experimental data of amine group concentrations is given in Table A1. Rate constants were computed for all solutions until 49 days, except MDEA.H 2 O (Glass) solutions, where the solutions leaked after week one. The week one results of 50 wt.% DEEA.H 2 O were also not included in linear regression. Unloaded amine wt.% and CO 2 loading at the start of an experiment for all solutions, along with the experimental duration, rate constants (k), and correlation coefficients (R 2 ) for all blends, are given in Table 2. The rate constant increased with the increase in amine concentration and CO 2 loading in aqueous solutions. The change of solvent had a different effect on each amine solution rate constant. The correlation factor (R 2 ) value was higher than 0.9 in all solutions, showing the good accuracy of the model, and is visible in the thermal stability results (Figures 4-6 & 10-12). Figure 4 shows the thermal stability results of 30 wt.% aqueous amine solutions with CO 2 loading of 0.1. After seven weeks, the maximum thermal stability was observed in 3DMA-1P. Overall, the thermal stability was quite high, as less than 2 percentage points of amine were lost during the whole set of experiments. No significant difference in thermal stability was observed when the methyl group was swapped with an ethyl group in solutions, i.e., 3DMA-1P to 3DEA-1P.  Figures 5 and 6 represent the 30 wt.% amine blends in MEG and TEG with CO2 loading ≈ 0.1, respectively. In both cases, MDEA blends showed the least thermal stability compared to others. Installing a glass tube in between the metal cylinder and blend increased the thermal stability of MDEA blends. In both MEG and TEG blends of MDEA, the glass tube increased the thermal stability by 2 percentage points after seven weeks, which might be due to the absence of metal ions. An increase in the alkyl group in amine from 3DMA-1P to 3DEA-1P decreased the thermal stability in amine-MEG blends and the opposite trend was seen with amine-TEG blends. Eide-Haugmo [35] and Shoukat et al. [37] also found that an increase in the carbon number or alkyl group length reduced the thermal stability in aqueous amine solutions, and vice versa, for amine-water-glycol solutions. 3DMA-1P and DEEA showed the highest thermal stability among MEG and TEG blends, respectively, after 49 days.   Figures 5 and 6 represent the 30 wt.% amine blends in MEG and TEG with CO 2 loading ≈ 0.1, respectively. In both cases, MDEA blends showed the least thermal stability compared to others. Installing a glass tube in between the metal cylinder and blend increased the thermal stability of MDEA blends. In both MEG and TEG blends of MDEA, the glass tube increased the thermal stability by 2 percentage points after seven weeks, which might be due to the absence of metal ions. An increase in the alkyl group in amine from 3DMA-1P to 3DEA-1P decreased the thermal stability in amine-MEG blends and the opposite trend was seen with amine-TEG blends. Eide-Haugmo [35] and Shoukat et al. [37] also found that an increase in the carbon number or alkyl group length reduced the thermal stability in aqueous amine solutions, and vice versa, for amine-water-glycol solutions. 3DMA-1P and DEEA showed the highest thermal stability among MEG and TEG blends, respectively, after 49 days.   Figures 5 and 6 represent the 30 wt.% amine blends in MEG and TEG with CO2 loading ≈ 0.1, respectively. In both cases, MDEA blends showed the least thermal stability compared to others. Installing a glass tube in between the metal cylinder and blend increased the thermal stability of MDEA blends. In both MEG and TEG blends of MDEA, the glass tube increased the thermal stability by 2 percentage points after seven weeks, which might be due to the absence of metal ions. An increase in the alkyl group in amine from 3DMA-1P to 3DEA-1P decreased the thermal stability in amine-MEG blends and the opposite trend was seen with amine-TEG blends. Eide-Haugmo [35] and Shoukat et al. [37] also found that an increase in the carbon number or alkyl group length reduced the thermal stability in aqueous amine solutions, and vice versa, for amine-water-glycol solutions. 3DMA-1P and DEEA showed the highest thermal stability among MEG and TEG blends, respectively, after 49 days.   Figure 7 presents the effect of a change of solvent from water to MEG on the thermal stability of amine for 30 wt.% amine blends at 0.1 CO2 loadings. MEG usage instead of water replacement decreased the thermal stability of all solutions, except 3DMA-1P. The thermal stability of MDEA was decreased by 8 percentage points, which was more than three times that of other blends. Similarly, Figure 8 shows the results for the change of solvent from water to TEG. TEG decreased the thermal stability of all blends. Again, like the effect of MEG, the maximum thermal stability decrease was observed in the MDEA.TEG blend. Figure 9 represents the results of changing the solvent from MEG to TEG. Apart from 3DMA-1P, the thermal stability was increased for all blends. The maximum increase (3 percentage points) in thermal stability was observed in the MDEA (Glass) week five blend.  Figure 7 presents the effect of a change of solvent from water to MEG on the thermal stability of amine for 30 wt.% amine blends at 0.1 CO 2 loadings. MEG usage instead of water replacement decreased the thermal stability of all solutions, except 3DMA-1P. The thermal stability of MDEA was decreased by 8 percentage points, which was more than three times that of other blends. Similarly, Figure 8 shows the results for the change of solvent from water to TEG. TEG decreased the thermal stability of all blends. Again, like the effect of MEG, the maximum thermal stability decrease was observed in the MDEA.TEG blend. Figure 9 represents the results of changing the solvent from MEG to TEG. Apart from 3DMA-1P, the thermal stability was increased for all blends. The maximum increase (3 percentage points) in thermal stability was observed in the MDEA (Glass) week five blend.  Figure 7 presents the effect of a change of solvent from water to MEG on the thermal stability of amine for 30 wt.% amine blends at 0.1 CO2 loadings. MEG usage instead of water replacement decreased the thermal stability of all solutions, except 3DMA-1P. The thermal stability of MDEA was decreased by 8 percentage points, which was more than three times that of other blends. Similarly, Figure 8 shows the results for the change of solvent from water to TEG. TEG decreased the thermal stability of all blends. Again, like the effect of MEG, the maximum thermal stability decrease was observed in the MDEA.TEG blend. Figure 9 represents the results of changing the solvent from MEG to TEG. Apart from 3DMA-1P, the thermal stability was increased for all blends. The maximum increase (3 percentage points) in thermal stability was observed in the MDEA (Glass) week five blend.    [32] found that an increase in the initial MDEA concentration increased the amine loss rates due to the higher solubility of CO2 in the solutions achieving constant CO2 loading at higher amine concentrations. Similarly, Gao et al. [39] concluded that an increase in the initial concentration of DEEA significantly decreased the stability of amine due to the availability of a more reactive DEEA molecule for CO2 absorption.    [32] found that an increase in the initial MDEA concentration increased the amine loss rates due to the higher solubility of CO2 in the solutions achieving constant CO2 loading at higher amine concentrations. Similarly, Gao et al. [39] concluded that an increase in the initial concentration of DEEA significantly decreased the stability of amine due to the availability of a more reactive DEEA molecule for CO2 absorption.  [32] found that an increase in the initial MDEA concentration increased the amine loss rates due to the higher solubility of CO 2 in the solutions achieving constant CO 2 loading at higher amine concentrations. Similarly, Gao et al. [39] concluded that an increase in the initial concentration of DEEA significantly decreased the stability of amine due to the availability of a more reactive DEEA molecule for CO 2 absorption.

Effect of Initial CO2 Loading
The effect of CO2 loading on the thermal stability in the initial 30 wt.% MDEA and DEEA aqueous solutions is presented in Figure 12. After 49 experimental days, the thermal stability was reduced with an increase in CO2 loading from 0.1 to 0.4 for both amines. Aqueous DEEA solution exhibited a greater reduction in the thermal stability compared to aqueous MDEA. Again, our data agree with the published literature. Chakma and Meisen [32] and Gao et al. [39] found that an increase in the CO2 concentration in solutions decreased the thermal stability of aqueous MDEA and DEEA solutions, respectively. Lepaumier et al. [33] and Eide-Haugmo [35] also concluded that the addition of CO2 reduced the thermal stability of amines.

Effect of Initial CO2 Loading
The effect of CO2 loading on the thermal stability in the initial 30 wt.% MDEA and DEEA aqueous solutions is presented in Figure 12. After 49 experimental days, the thermal stability was reduced with an increase in CO2 loading from 0.1 to 0.4 for both amines. Aqueous DEEA solution exhibited a greater reduction in the thermal stability compared to aqueous MDEA. Again, our data agree with the published literature. Chakma and Meisen [32] and Gao et al. [39] found that an increase in the CO2 concentration in solutions decreased the thermal stability of aqueous MDEA and DEEA solutions, respectively. Lepaumier et al. [33] and Eide-Haugmo [35] also concluded that the addition of CO2 reduced the thermal stability of amines.

Effect of Initial CO 2 Loading
The effect of CO 2 loading on the thermal stability in the initial 30 wt.% MDEA and DEEA aqueous solutions is presented in Figure 12. After 49 experimental days, the thermal stability was reduced with an increase in CO 2 loading from 0.1 to 0.4 for both amines. Aqueous DEEA solution exhibited a greater reduction in the thermal stability compared to aqueous MDEA. Again, our data agree with the published literature. Chakma and Meisen [32] and Gao et al. [39] found that an increase in the CO 2 concentration in solutions decreased the thermal stability of aqueous MDEA and DEEA solutions, respectively. Lepaumier et al. [33] and Eide-Haugmo [35] also concluded that the addition of CO 2 reduced the thermal stability of amines.

Corrosion
Week 5 blends were analyzed for individual metal components (iron (Fe), nickel (Ni), chromium (Cr), and molybdenum (Mo)) using ICP-MS, and the results as a sum of all of these metals for 30 wt.% amine blends with initial 0.1 CO2 loadings are given in Figure 13. A higher corrosivity of a blend was indicated by a higher metal concentration. Each solution and its replica were analyzed and the average was used for result calculations. The average deviation in the replicas was ±0.1 mg/L. Various metal concentrations (mg/L) in all solutions after five weeks are given in Table A2. Iron contributed most to the total metal concentration in all amine blends in MEG and TEG. In aqueous amine solutions, Mo showed the highest contribution, except for aqueous MDEA solutions, where Cr contributed the most. In all amine blends except the DEEA blend, the total metal concentration was highest in amine-TEG blends, followed by amine-H2O blends. The lowest metal concentrations were found in amine-MEG blends. The total metal concentration in DEEA blends was in the order of DEEA.H2O < DEEA.MEG < DEEA.TEG.

Corrosion
Week 5 blends were analyzed for individual metal components (iron (Fe), nickel (Ni), chromium (Cr), and molybdenum (Mo)) using ICP-MS, and the results as a sum of all of these metals for 30 wt.% amine blends with initial 0.1 CO 2 loadings are given in Figure 13. A higher corrosivity of a blend was indicated by a higher metal concentration. Each solution and its replica were analyzed and the average was used for result calculations. The average deviation in the replicas was ±0.1 mg/L. Various metal concentrations (mg/L) in all solutions after five weeks are given in Table A2. Iron contributed most to the total metal concentration in all amine blends in MEG and TEG. In aqueous amine solutions, Mo showed the highest contribution, except for aqueous MDEA solutions, where Cr contributed the most. In all amine blends except the DEEA blend, the total metal concentration was highest in amine-TEG blends, followed by amine-H 2 O blends. The lowest metal concentrations were found in amine-MEG blends. The total metal concentration in DEEA blends was in the order of DEEA.H 2 O < DEEA.MEG < DEEA.TEG.

Corrosion
Week 5 blends were analyzed for individual metal components (iron (Fe), nickel (Ni), chromium (Cr), and molybdenum (Mo)) using ICP-MS, and the results as a sum of all of these metals for 30 wt.% amine blends with initial 0.1 CO2 loadings are given in Figure 13. A higher corrosivity of a blend was indicated by a higher metal concentration. Each solution and its replica were analyzed and the average was used for result calculations. The average deviation in the replicas was ±0.1 mg/L. Various metal concentrations (mg/L) in all solutions after five weeks are given in Table A2. Iron contributed most to the total metal concentration in all amine blends in MEG and TEG. In aqueous amine solutions, Mo showed the highest contribution, except for aqueous MDEA solutions, where Cr contributed the most. In all amine blends except the DEEA blend, the total metal concentration was highest in amine-TEG blends, followed by amine-H2O blends. The lowest metal concentrations were found in amine-MEG blends. The total metal concentration in DEEA blends was in the order of DEEA.H2O < DEEA.MEG < DEEA.TEG.

Effect of the initial amine concentration
The results for the effects of the amine concentration on corrosivity at a constant initial CO 2 loading of 0.4 are presented in Figure 14. An increase in the amine concentration increased the corrosivity, mainly due to higher amine loss. The maximum metal concentration (534 mg/L) was found with 70 wt.% MDEA aqueous solution, whereas the minimum value (1 mg/L) was found with 10 wt.% DEEA aqueous solution. The total metal concentration in aqueous MDEA increased exponentially, whereas in aqueous DEEA it increased quite linearly.

Effect of the initial amine concentration
The results for the effects of the amine concentration on corrosivity at a constant initial CO2 loading of 0.4 are presented in Figure 14. An increase in the amine concentration increased the corrosivity, mainly due to higher amine loss. The maximum metal concentration (534 mg/L) was found with 70 wt.% MDEA aqueous solution, whereas the minimum value (1 mg/L) was found with 10 wt.% DEEA aqueous solution. The total metal concentration in aqueous MDEA increased exponentially, whereas in aqueous DEEA it increased quite linearly.  Figure 15. An increase in CO 2 loading increased the total metal concentration in both solutions. The aqueous MDEA solution with 0.4 CO 2 loading contained the highest amount of metals compared to other solutions. Eide-Haugmo [35] also found that CO 2 -loaded amine solutions corroded more in comparison to their unloaded counterpart.

Conclusions
In this work, the thermal stability and corrosivity of four tertiary amine blends in water, MEG, and TEG, loaded with CO2, were studied in stainless steel cylinders at 135 °C for seven weeks. The effects of direct metal contact with the blend, the initial amine concentration, and the initial CO2 loading were discussed. An increase in the initial amine concentration, CO2 loading, and direct

Conclusions
In this work, the thermal stability and corrosivity of four tertiary amine blends in water, MEG, and TEG, loaded with CO 2 , were studied in stainless steel cylinders at 135 • C for seven weeks. The effects of direct metal contact with the blend, the initial amine concentration, and the initial CO 2 loading were discussed. An increase in the initial amine concentration, CO 2 loading, and direct contact between the cylinder and blends decreased the thermal stability, and vice versa, for corrosivity. Amine blends with TEG displayed a higher thermal stability, except for 3DMA-1P, and higher corrosion, in comparison to MEG. 3DEA-1P blends showed a good combination of low corrosion and a higher thermal stability in comparison to other solutions.