Академический Документы
Профессиональный Документы
Культура Документы
Abstract
An apparatus for the measurement of vapor–liquid equilibrium (VLE, P–T–x) data and density of saturated-liquid and high-pressure liquid
phase of compressed gas-organic solvent systems was designed, built and tested. A synthetic method, with recirculation of the liquid phase
through a high-pressure densimeter by a magnetic driven centrifugal pump, was used. The thermodynamic equilibrium was reached in a visual
cell with variable volume ranging between 30 and 60 cm3 . The compositions of the phases were evaluated by external gravimetric method.
The estimated accuracy of the measured data was ±0.02 K for temperature, ±0.05 MPa for pressure, ±0.005 in mole fraction for liquid
compositions and ±0.05 g/L for liquid density. Vapor–liquid equilibrium data and saturated-liquid density were obtained for the following
systems: carbon dioxide–ethanol and carbon dioxide–acetone at 291.15, 303.15, 313.15 and 323.15 K and carbon dioxide–dichloromethane
at 291.15, 303.15 and 311.15 K. For the latter system, high-pressure liquid density data were measured up to 18 MPa. The experimental data
were correlated by both Peng–Robinson and SAFT equation of state. SAFT EOS gives better correlations of vapor–liquid data and accurate
predictions of high-pressure density of the liquid phase.
© 2004 Elsevier B.V. All rights reserved.
Keywords: Experimental data; Vapor–liquid equilibrium; Density; Carbon dioxide; Ethanol; Acetone; Dichloromethane; SAFT
0896-8446/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.supflu.2004.04.003
8 M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14
Fig. 1. Schematic diagram of the apparatus. Broken line contains the thermostated apparatus components.
M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14 9
electronic transducer. The overall accuracy of the tempera- 2.4. Method of VLE measurements
ture measurement was estimated to be within ±0.02 K.
Pressure was measured using a membrane transducer The samples were loaded after the cleaning procedure.
(Endress-Hauser, Rheinh, Germany) with a precision of The recirculation loop and the magnetic driven pump were
0.5% FS. The pressure transducer was calibrated using evacuated for 12 h. The VLE cell, that cannot withstand with
dead-weight gauge. The overall accuracy of the pressure vacuum, was purged with 10 cm3 /min CO2 flow rate for 12 h.
measurement was estimated to be within ±0.08 MPa includ- The entire apparatus was filled and discharged for five times
ing errors due to pressure gauge and pressure fluctuation with CO2 at CO2 -saturated pressure at ambient condition.
during VLE measurements. The samples were added to the cell in a desiderated amount
Density was measured using a vibrating tube densime- in order to obtain completely liquid-filled apparatus reducing
ter. The difference between the temperature recorded in the the volume of the VLE cell.
densimeter and that of the VLE cell was lower than 1 K. Expanding the volume of the liquid-filled cell, pressure,
The overall accuracy of the density measurement including temperature, window displacement (by a calibrated device
errors due to temperature fluctuation was estimated to be with 0.1 mm precision) and density were reordered. The
within ±0.05 g/L. density and window displacement versus pressure varia-
Video acquisition system allows to visualize on external tion were analyzed. Because vapor and liquid phases have
video the phase transition in the VLE cell. Data acquisition different compressibilities, the slope of curves should sig-
system (IRS, Padova, Italy) record on personal computer nificantly changes at the VLE equilibrium point. The equi-
pressure, temperature and video signal. librium point can be evaluated intersecting two gas-like and
liquid-like correlations obtained for either density–pressure
2.3. Gravimetric composition measurements or window displacement-pressure profile reported. These
results are confirmed by visual observation of small vapor
Liquid sample were prepared in glass syringe (Vetrotec- bubble formation in the VLE cell at equilibrium conditions.
nica, Padova, Italy) and they were weighted with a standard In order to evaluate the accuracy of the method, the mea-
analytical balance of 110 g capacity and 0.0001 g resolu- surements were repeated decreasing and increasing VLE
tion (Sartorius, Padova, Italy). Gas samples were prepared cell volume; same results were obtained. However, the
in 75 cm3 stainless steel cylinder with a low-weight me- method that requires to expand the liquid phase guarantee
tering valve. The overall tare weight of the cylinder, valve a better homogeneity and mixing in the VLE apparatus.
and connectors was 398 g and the maximum pressure In this work, to speed up the measurement, the equi-
was 12 MPa. This characteristics enable sample prepara- librium conditions were obtained expanding liquid-filled
tion by weighing with a standard analytical balance of VLE cell and using both density–pressure profile anal-
500 g capacity and 0.001 g resolution (Gibertini, Padova, ysis and visual observations. Temperature, pressure and
Italy). saturated-liquid density were recorded at equilibrium con-
The liquid samples were first added into the VLE cell pre- dition. High-pressure liquid density was measure increasing
viously filled with CO2 at room pressure and temperature. the pressure of the system by reducing the VLE cell volume
The weight was recorder and the evaluated instrumental un- at pressure over mixture bubble point.
certainty in liquid sample added to the cell was ±0.0003 g.
The gas samples were added into the cell connecting the
cylinder with the apparatus inline. The amount of the gas 3. Modelling
added in the apparatus was measured by weighting the gas
cylinder before and after the loading operation. The gas, The binary vapor–liquid equilibria was correlated by a
remained between the valve of the cylinder and the valve simple cubic Peng–Robinson (PR) equation of state (EOS)
of the apparatus inline, was volumetrically measured in a and using a more physical based statistical associating fluid
1 cm3 resolution graduate glass cylinder filled with water. theory (SAFT). In particular, a modification of original
The final amount of gas added in the VLE cell was ob- SAFT by Huang and Radosz [14–16] presented by Pfohl
tained as difference between the gas introduced in the ap-
paratus and the one remained in the inline section. The
amount of gas introduced in the VLE cell was varied by Table 1
controlling the difference between the temperature of the SAFT EOS pure-component segment parameters
VLE cell and the gas cylinder and the charging time. The CO2 Ethanol Acetone DCM
evaluated uncertainty in gas sample added to the cell was
u0 /k (K) 216.08 213.87 209.50 218.13
±0.005 g. v00 (mL/mol) 13.578 12.449 12.320 13.732
The estimated accuracy of mixture composition data is m 1.417 2.438 2.282 2.572
usually better then ±0.001 in mole fraction. In the worst ǫA /k (K) – 2791.1 – –
cases, depending on the composition the mixture, accuracy 102 kA – 2.5819 – –
of mixture composition data is ±0.005. k is the Boltzmann’s constant.
10 M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14
and Brunner [17] was used. Pure component parameters for Table 3
organic solvents and CO2 used in this work are reported in P–T–x and saturated-liquid density experimental data for the carbon
dioxide–ethanol system
Table 1 for SAFT EOS only.
For our systems two interaction parameters, kij and λij , T (K) xCO2 P (MPa) ρ (g/cm3)
were used for both equation of states. Quadratic mixing rule 291.15 0.241 3.30 0.82852
was used for interaction parameter and linear mixing rule 0.364 4.18 0.84377
was used for volume parameter [17]. The binary interaction 0.486 4.73 0.85556
parameters were obtained by fitting both EOSs to our exper- 303.15 0.241 3.60 0.81139
imental data by minimizing the relative pressure deviation 0.364 4.67 0.82152
objective function: 0.486 5.62 0.82870
Np 313.15 0.241 4.38 0.80012
P 2
0.364 5.89 0.81010
obf = δ (1)
P 0.486 6.96 0.81721
i=1
323.15 0.241 4.93 0.78921
where Np is the number of isothermal data points considered 0.364 6.79 0.79832
and the residuals on pressure δ(P/P) are given by: 0.486 8.11 0.80404
P Pcalculated
δ = − 1. (2)
P Pexperimental
Table 4
The values of saturated-liquid density and high-pressure P–T–x and saturated-liquid density experimental data for the carbon
liquid density are predicted using the interaction parameters dioxide–acetone system
obtained from the VLE fitting. All calculation reported in T (K) xCO2 P (MPa) ρ (g/cm3)
this work were performed using a “Phase Equilibria” soft-
291.15 0.120 0.70 0.80441
ware (Technical University of Hamburg-Harbug, Germany).
– 0.212 1.12 0.82077
– 0.364 1.86 0.83888
0.530 2.78 0.85815
4. Results and discussion
303.15 0.120 0.73 0.79203
– 0.212 1.28 0.80536
Table 2 reports the summary of statistical errors and inter- – 0.364 2.25 0.82437
action parameter values used for both vapor–liquid equilibria 0.530 3.39 0.84011
and saturated-liquid density calculation performed by SAFT 313.15 0.120 0.78 0.77853
EOS. The PR EOS does not give satisfactory description of – 0.212 1.48 0.79412
both volumetric and equilibrium properties, especially for – 0.364 2.64 0.80872
associating molecules such as ethanol. – 0.530 3.85 0.82365
Tables 3–5 show the P–T–x and saturated-liquid den- 323.15 0.120 1.06 0.76956
sity experimental values for carbon dioxide–ethanol, car- – 0.212 1.88 0.74760
– 0.364 3.19 0.79540
bon dioxide–acetone and carbon dioxide–dichloromethane.
– 0.530 4.02 0.80931
Table 6 shows high-pressure liquid density experimental
Table 2
Summary of statistical errors and interaction parameters obtained from the fitting of vapor–liquid experimental data and from the prediction of
saturated-liquid density using the SAFT EOS
T (K) Data number P/P (%) |P|P (%) ρ/ρ (%) |ρ|ρ (%) kij λij
Carbon dioxide–ethanol
291.15 3 −0.63 1.2 −0.29 0.5 0.081 0.048
303.15 3 −0.49 1.0 −0.57 1.1 0.070 0.060
313.15 3 −0.43 1.1 −0.65 1.3 0.079 0.058
323.15 3 −0.46 0.9 −0.61 1.2 0.078 0.059
Carbon dioxide–acetone
291.15 4 −0.12 1.5 0.49 0.8 0.041 0.024
303.15 4 −0.01 1.2 0.4 9.7 0.040 0.030
313.15 4 0.23 1.1 0.46 1.0 0.045 0.029
323.15 4 −0.31 0.8 0.82 1.7 0.042 0.029
Carbon dioxide–diclhoromethane
291.15 5 0.36 0.6 −0.57 1.1 0.041 0.024
303.15 5 0.08 0.4 −0.55 1.2 0.040 0.030
311.15 5 −0.01 0.5 −0.71 1.5 0.045 0.029
M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14 11
Table 5
P–T–x and saturated-liquid density experimental data for the carbon
dioxide–dicholoromethane system
T (K) xCO2 P (MPa) ρ (g/cm3)
measurements for the carbon dioxide–dicholoromethane perimental bubble pressure; statistical errors are reported in
system at 291.15, 303.15 and 311.15 K. Table 2.
Fig. 2 shows an example of comparison of P–T–x ex- Fig. 4 shows a comparison of saturated-liquid density
perimental data measured in this work with those reported measured in this work for carbon dioxide–ethanol with liter-
in the literature for carbon dioxide–ethanol system. Our re- ature data of two different investigator: Chang and coworkers
sults shows a good agreement with other experimental data [6,7]. Although the temperature are not the same, our results
at the higher temperature. However, for the lowest tempera- well agree with the density data presented by Kordikowski
ture data reported in the literature show singular behaviors. et al. [6]. On the other hand, the data from Chang et al. [7]
Although our data was measured at 291.15 and 303.15 K, shows a different and contradictory behaviour for both equi-
our results seem to properly reproduce the behavior of the librium and volumetric properties. The SAFT EOS gives an
data of Kordikowski et al. [6] accurate density prediction for both temperature: the density
Fig. 3 that shows a good agreement between the P–T–x calculation were performed using the interaction parameters
experimental data measured in this work and those reported fitted from bubble curve at same temperature. Statistical er-
in the literature for carbon dioxide–acetone system, con- rors for all calculation concerning the saturated-liquid den-
firming the reliability of the experimental setup and proce- sity are reported in Table 2.
dure proposed. The SAFT EOS accurately correlates the ex-
Fig. 2. Comparison of our experimental P–T–x measurements with liter- Fig. 4. Comparison of bubble points measured in this work for carbon
ature data for carbon dioxide–ethanol: (䊉) in this work at 291.15 K; (䊏) dioxide–acetone with literature data: (*) Giacobbe et al. [19] at 303.15 K;
in this work at 303.15 K; (䉱) in this work at 313.15 K; (䉬) in this work (䊊) Day et al. [20] at 313.13 K; (+) Adrian and Maurer [21] at 313.15 K;
at 323.15 K; (䊐) Chang et al. [7] at 298.15 K; (䉫) Kordikowski et al. (䊐) Bamberger et al. [22] at 323.2 K; (䊏) in this work at 303.15 K;
[6] at 298.16 K; (*) Suzuki et al. [8] at 313.4 K; (+) Jennings et al. [9] (䊉) in this work at 313.15 K; (䉬) in this work at 323.15 K. Lines
at 314.2 K; (䊊) Gonzalez et al. [5] at 314.45 K; (-) Jennings et al. [9] at are calculated by SAFT equation of state using interaction parameters
325.2 K. Lines are calculated by SAFT equation of state using interaction reported in Table 2. Lines are calculated by SAFT equation of state using
parameters reported in Table 2 at 291.15, 303.15, 313.15 and 323.15 K. interaction parameters reported in Table 2 at 303.15, 313.15 and 323.15 K.
12 M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14
Table 6
High-pressure liquid density experimental data for the carbon dioxide–dicholoromethane system
xCO2 P (MPa) ρ (g/cm3) P (MPa) ρ (g/cm3) P (MPa) ρ (g/cm3)
Interestingly, Tables 3–5 and Fig. 4 show that the of the CO2 in the saturated-liquid phase is favored, the
saturated-liquid density at constant temperature increases value of the saturated-liquid density shift towards the value
with carbon dioxide molar fraction up to a maximum for of the pure CO2 density. This value is generally lower than
the systems containing acetone and ethanol, whereas, it de- that of organic solvents. However, for all binary systems,
creases with carbon dioxide molar fraction for the system the expansion of the saturated-liquid phase by the carbon
containing dichloromethane. The saturated-liquid density dioxide solubilization predominates at high pressure.
behavior of the carbon dioxide–organic solvent binary mix- Fig. 5 shows new experimental data at temperature of
ture is governed by two contrasting phenomena: the liquid 291.15 and 303.15 K for carbon dioxide–dicholoromethane
compression due to the high pressure of the system and the system. These temperature conditions are close to those typ-
solubilization of the CO2 in the saturated-liquid phase. The ically used in the particle production by gas antisolvent pre-
first phenomena is dominant when the CO2 is not highly cipitation. Our results shows a good agreement with experi-
soluble in the organic solvent, and in this case, the density of mental data from Gonzalez et al. [5] at 311.15 K. The SAFT
the saturated-liquid phase raises by increasing the pressure give accurate prediction for all temperatures presented in
of the system. On the other hand, when the solubilization Fig. 5.
M. Stievano, N. Elvassore / J. of Supercritical Fluids 33 (2005) 7–14 13
References
[11] A. Striolo, A. Favaro, N. Elvassore, A. Bertucco, V. Di. Noto, [17] O. Pfohl, G. Brunner, Use of BACK to modify SAFT in or-
Investigation of Conformational Changes of Proteins exposed to der to enable density and phase equilibrium calculations con-
High-Pressure CO2 by FT-IR Spectroscopy, J. Supercrit. Fluids, nected to gas-extraction processes, Ind. Eng. Chem. Res. 37 (1998)
(2003). in press. 2966.
[12] J.O. Werling, P.G. Debenedetti, Numerical modeling of mass trans- [18] O. Pfohl, S. Petkov, G. Brunner, “PE” quickly makes available the
fer in the supercritical antisolvent process: miscible conditions, J. newest equations of state via the internet, Ind. Eng. Chem. Res. 39
Supercritic. Fluids 18 (2000) 11. (2000) 4439.
[13] J.O. Werling, P.G. Debenedetti, Numerical modeling of mass transfer [19] F.W. Giacobbe, Thermodynamic solubility behavior of carbon dioxide
in the supercritical antisolvent process, J. Supercritic. Fluids 16 in acetone, Fluid Phase Equilibr. 72 (1992) 277.
(1999) 167. [20] C.Y. Day, C.J. Chang, C.Y. Chen, Phase equilibrium of ethanol plus
[14] S.H. Huang, M. Radosz, Equation of state for small, large, polydis- CO2 and acetone plus CO2 at elevated pressures, J. Chem. Eng.
perse, and associating molecules, Ind. Eng. Chem. Res. 29 (1990) Data, 44 (1999) 365
2284. [21] T. Adrian, G. Maurer, Solubility of carbon dioxide in acetone and
[15] S.H. Huang, M. Radosz, Equation of state for small, large, polydis- propionic acid at temperatures between 298 K and 333 K, J. Chem.
perse, and associating molecules: extension to fluid mixtures, Ind. Eng. Data 42 (1997) 668.
Eng. Chem. Res. 30 (1991) 1994. [22] A. Bamberger, G. Maurer, High-pressure (vapour plus liquid) equi-
[16] S.H. Huang, M. Radosz, Additions and corrections, Ind. Eng. Chem. libria in (carbon dioxide plus acetone or 2-propanol) at temperatures
Res. 32 (1993) 762. from 293 K to 333 K, J. Chem. Thermodyn. 32 (2000) 685.