Hydrogen sulfide solubility in 50 wt% and 70 wt% aqueous methyldiethanolamine at temperatures from 283 to 393 K and total pressures from 500 to 10000 kPa

23 The hydrogen sulfide (H 2 S) absorption capacity of wt.% aqueous 24 methyldiethanolamine (MDEA) solution was investigated in a static-analytic apparatus 25 at temperatures of 283, 353 and 393 K and pressures of 2000, 6000 and 10000 kPa in the 26 presence of methane. New experimental data were also produced for a 50.1 wt.% aqueous 27 MDEA at 323 K and pressures of 500 and 3000 kPa as part of the apparatus validation 28 procedure. A model based on electrolyte non-random two-liquid (eNRTL) activity 29 coefficient model to describe the liquid phase and Peng-Robinson Equation of State to 30 describe the vapor phase non-idealities was developed for the system H 2 S-MDEA-H 2 O, 31 which can potentially be used also for the system in the presence of methane at low 32 pressures. Vapor pressure measurements of pure MDEA were also performed in the range 33 of 405 – 435 K in an ebulliometer and parameters for the Antoine correlation were 34 proposed. 35

excluded during our thermodynamic modeling. During the evaluation, the partial 98 pressures for H2S from Kuranov et al. [17], Kamps et al. [18] and Sidi-Boumedine et al. 99 [19], who all report total pressures in the absence of makeup gases, were calculated by 100 subtracting the vapor pressure of the solvent calculated by Dalton's Law (Eq. 1). The MacGregor and Mather [14] (pressure, loading, composition) as well as by Zoghi and 126 Shokouhi [22] (pressure and composition). At lower loadings, significant deviations are 127  is similar to the one previously presented by [29] and its schematic is given in Figure 1.       (NaOH) in order to neutralize the system. At each temperature, a new experiment was 260 conducted using fresh solution. We aimed at having the same global loading at all temperatures, 261 however it was not practically possible to reach exactly the same loadings in every experiment.

262
The study at each temperature and global loading lasted approximately one week.

263
The analysis of vapor phase concentration was performed in a GC equipped with a Porapak- where is the molar density of the gas mixture, calculated using REFPPROP and # is the   The chemical equilibrium constants as well as Henry's constant for hydrogen sulfide are 312 described by Eq. 8, parametrized according to

318
The vapor pressure for hydrogen sulfide and water is estimated using the Riedel correlation 319 (Eq. 9) where T expressed in K and P sat in Pa. The parameters are presented in Table 6. MDEA 320 vapor pressure has been measured in this work and fitted to Antoine correlation. The Antoine 321 parameters used in this work can be found in Section 1.5.2.

324
A sensitivity analysis was performed to evaluate the significant numbers in the parameters 325 retrieved from the literature. In Table 5 and As a result, the number of parameters is reduced to 36. The temperature dependency of the 338 energy parameters is described by Eq. 10, where a $9 and b $9 were fitted to experimental data.

339
; $9 = a $9 + b $9 6 Eq. 10 The fixed non-randomness factors and fixed energy parameter values are presented in Table 7, 340 where m denotes molecule and c-a cation-anion (salt). The non-randomness factors were fixed  the partial pressure of H2S, PH2S, or the total pressure, Ptot.
Eq. 11  Table 8 and Table 9. As mentioned earlier,       The observation of increased H2S partial pressure upon increase in total pressure can be 411 made also for the 70 wt.% aqueous MDEA system for the temperatures of 283 K and 353 K.  Ebulliometer. The measurements conducted in the ebulliometer are shown in

529
In this section, we present first the results from the ebulliometer following by the modeling 530 results for the high-pressure VLE data, since the first ones are used in the model parametrization 531 for the H2S-MDEA-H2O equilibrium.

533
Ebulliometer. The Antoine correlation was fitted to available data from the literature (Table   534 3) as well as the newly obtained data of this work, covering a large range of temperatures and 535 pressures. In Table 10 Table 11. Parameters for the Antoine correlation a for pure MDEA vapor pressure.  Supporting Information.

564
The fixed parameters in Table 7 and the regressed parameters for the binary subsystem values of partial pressures in comparison with total ones, leading to higher relative numbers.

572
Therefore, we have decided to also perform the data regression excluding all data at loadings 573 lower than 0.05 mol H2S/mol MDEA (Case B). This indeed improved substantially the fitting 574 of the partial pressures, as one can see in the AARDs in Table 13, from approximately 30% to 575 18%. The parity plot for the predicted and experimental values is shown in Figure 8 while 576 Figure 9 shows the difference between predicted and experimental H2S partial pressure as a 577 function of the experimental value.

617
The differences in H2S partial pressure noticed in the literature data as well as in our data 618 obtained in the presence of methane for relatively low total pressure levels, are comparable to 619 the accuracy of the model. Therefore, since also the effect of methane in the liquid loading has 620 been found to be negligible for a 70 wt.% MDEA-H2O, we also fitted the model to data available 621 in the presence of methane. However, the code was not modified but, instead, the data for partial   The model parameters obtained from the data regression in each case studied are given in 641 Supplementary Information. Figure 11 shows experimental and modelled values for a 70 wt.% 642 aqueous MDEA system as a function of temperature in Case C while Table 13 contains 643 information about each regression in terms of Bias, AADs and AARDs. The performance of 644 the model for a 70 wt% MDEA-H2O system is good, especially considering the few data 645 available for this solvent concentration. In Table 13, it can be seen that the accuracy of the 646 model does not significantly change upon the addition of the experimental points with methane 647 in the regression. The overall AARD for the partial pressure is altered from 18% to 21%, which 648 is also the AARD calculated for the data published in this work. The data from MacGregor and  To illustrate the latter, we repeated the data regression for Case A. The resulted AARDs were 655 29.8% and 30.1%, using the exact same data and fixed parameters. As far as the ability of the 656 model to predict the total pressure is concerned, the accuracy has surprisingly improved. This 657 is merely a lucky coincidence due to the fitting of the experimental points for methane-included activity coefficient model for the liquid phase. The AARD for the partial pressure of H2S and 695 for the total system pressure was found to be 18% and 16% respectively. The effect of including 696 data in the presence of methane and maximum total pressure of 2000 kPa in the data regression 697 was studied and found minimal. However, for higher total pressure and different conditions 698 than the studied ones, the use of models taking into account the methane presence was    The traditional van der Waals one-fluid mixing rules were used for the estimation of the gas 830 mixture parameters from the pure components' properties.

833
The critical properties used in this work can be found in Supporting Information.