Continuous Sampling System for
Determining Vapor-Liqui W. L. KRANICH, R . E. WAGNER, D. W. SUNDSTROM1, AND HERBERT SLOTNICK2 F'orcester Polytechnic I n s t i t u t e , F'orcester, Mass.
rlNY methods are available for the experimental determination of vapor-liquid equilibrium data. These have been critically reviewed by Robinson and Gilliland (3). A weakness common to almost all methods is the necessity for removing samples from the equilibrium unit for analysis. This often results in the introduction of errors, particularly when the equilibrium is being measured a t high pressure. The bubble point-dew point method does not require the removal of samples, but it is capable of analyzing only a two-component system and it does not give vapor-liquid equilibrium composition data directly. I t was the object of the study to develop a method which could be applied to a wide variety of multicomponent systems over a considerable range of pressures m-ithout the removal of samples. The method chosen was spectrophotometry applied t o the analysis of gases. The specific objective of this paper was to demonstrate ivith relatively simple equipment and with a minimum of experinientation t h a t spectrophotometry can be directly applied to the determination of vapor-liquid equilibrium. This initial study was therefore confined to analysis with an ultraviolet spectrophotometer of a two-component system a t approximately atmospheric pressure. Further experimental work n i t h threecomponent systems at elevated pressures is in progress. THEORY
Spectrophotometric analysis has been widely applied to the determination of composition of both liquids and gases The range of wave lengths of radiation chosen depends on the absoiption properties of the mateiials analyzed. Because of the limitations of available equipment, this study was confined to the ultraviolet range, This limits the choice of compounds suitable for analysis; the principles, however, can be directly applied to infrared spectrophotometiy, which would allow much n ider application of the method. The absorption of radiatmn of a particular wave length by gas or liquid solutions made up of a single absorbing solute in a nonabsoibing medium is a function of the concentration of the absorbing solute. Thus a spectrophotometer which measuies the transmittance of incident radiation through such a solution can be directly calibrated in terms of concentration of solute. The calibration with knonn solutions must, of course, be cairied out under the same conditions as x-hen the equipment is to be used for analysis. I n many cases in which the solvent does not influence the absorption characteristics of the solute, complete calibration is unnecessary, as the data can be represented by the Beer-Lambert i an; log (Zo/I)= log ( l / t ) = D
=
EcL
1 2
=
E,c,L
+ E ~ c L+L E,c,L +
,
,
,
, ,
.
co =
xaLI-/T7
(4)
For ideal mixtures of perfect gases, by use of the perfect gas lam, Equations 3 and 4 may be modified to:
D
=
(L/V,)(P/P,)(3',/T)(E,r, f Ebxb f Eexo
+
For a mixture of n components, the additional relationship is also available: ((5)
Thus, for a two-component mixture it is necessary to nieasni.e absorbance a t only a single wave length in order to analyze the mixture, substituting the data in Equation 5 . The temperature and pressure must' be known, as ne11 as the individual extinction coefficients a t t h a t n-ave length. For a three-component niistuye, measurement of absorbance must be made at' ti$-o wave lengths so chosen that the individual values of extinction coefficient are widely different relat,ive to one another a t these wave lengths. This naturally limits the types of multicomponent systems which can be analyzed by the method. The above derivation for perfect gases in ideal mixtures assumes t h a t t,he extinction coefficient is independent of pi'essure and temperature. Slthough this is often true over a considerable range of temperatures and pressures, the probability of exror is reduced by experimentally measuring absorbances for the pure components in the apparatus and under the conditions used for the analysis of mixtures. If this is done, the working equation becomes:
D =
XaD,O
fX b D ;
+
XcD:
+ ... ...
(7)
where
D,O = absorbance of pure component under same conditions as observation of mixture Under these conditions, the only major departures from the Beer-Lambert Iaw arise from the assumption of ideal solutions in which the components do not influence one another, rather than from the assumption that the pure components are ideal gases. SJ7here the lams do not hold, experimental calibration of the apparatus under the desired conditions with Itnon-n mixtures makes analysis possible. SELECTIOX OF SYSTEM FOR STUDY
After preliminary investigation of a variety of types of coinpounds, the system acetone-benzene was selected for study. This system has the following advantages.
(1)
For two or more absoibing components, a, b, c, etc., which do not interfere with each other's absorption, the law niay be written:
D =D,+Db+De+
D
The concent,ration, e, may he expressed in terms of mole fraction, x, the number of moles in the anaIysis cell, and the volume of the cell, V:
It has been reported in the literature by Othmer ( 2 ) and can thus be used for comparison purposes. The materials are readily available and easily purified. The transmittance curves are sufficiently dissimilar in the ultraviolet to make possible an analysis by absorption spectrophotometry.
(2)
Present address, University of Michigan, Ann Arbor, RIich. Present address, P r a t t & Whitney .41roraft, West Hartford, Conn.
956
INDUSTRIAL AND E N G I N E E R I N G CHEMISTRY
May 1956
=
r
TO
BAROSTAT
I
41
CONDENSER
0
. . . , . . .. . ,--). ,I I.
.II.
951
The particular wave lengths chosen were 2400 and 2850 A. At the former, under the experimental conditions acetone is highly transparent while benzene absorbs strongly, and a t the latter the reverse is true. This meets the necessary conditions for successful analysis of a three-component system in which the third component is completely transparent a t both wave lengths. For analysis of the two-component system, only one wave length need be used in the procedure outlined. APPARATUS
T h e basic apparatus chosen for establishing vapor-liquid equilibrium was the Othmer equilibrium still (a), a device which has wide ABSORPTION acceptance. I n this still, a relatively small CELLS amount of the liquid in the still is vaporized, removed from the boiling flask, condensed, and returned. I n the normal application, a portion of the condensed vapor is trapped and removed as a sample for analysis. SimultaneF ously, a liquid sample is withdrawn from the boiling flask. The still was modified so that the vapor was superheated t o a constant temperature and passed through a quartz absorption cell before reaching the condenser. I n addition, a small stream of liquid was continuously nrithdrawn by gravity from the bottom of the boiling flask and passed through an electrically heated flash boiler, where it was totally vaporized. The flashed liquid was then superheated t o the same temperature as the vapor sample and passed through a second quartz iF THERMOCOUPLE analysis cell before going t o the condenser j CHAMBER and returning t o the flask. T h e equilibrium . . . . HEATING W I R E still is shown in Figure 1. It was decided t o flash the liquid sample and analyze as a vapor, rather than to attempt t o analyze the liquid spectrophotometrically, because the low transmittance of Figure 1. Diagram of equilibrium still ultraviolet light by many organic liquids would require a cell path so short as t o be impractical in a simple continuous system. The relative extinction coefficients of the two materials vary Furthermore, analysis of both samples a t the same temperawidely with wave length. This system can therefore be used to ture in the vapor phase reduces error by eliminating calibrashow the applicability of the method to analysis of a three-comtion of a second type of cell with different extinction coeffiponent system by analyzing the two-component mixture at two cients. I n addition, mixtures of superheated vapors are more different wave lengths-i.e., it may be assumed that a third component such as ethyl alcohol is present, which is completely likely to be ideal than are liquid solutions; hence, the Beertransparent at the wave lengths observed. If the binary analysis Lambert law is more likely to apply. is consistent at two wave lengths where the relative extinction Superheating of the vapors was necessary to prevent condencoefficients are widely different, then the data taken would have sation on the walls of the analysis cells and in the lines. An been adequate for the complete analysis of the hypothetical three-component system. indication of the temperature of the vapors was obtained by thermocouples at the analysis cells. This temperature was held I n selection of proper wave lengths i t is desirable to avoid regions where the extinction coefficient changes rapidly with wave length, as slight errors in the mechanical selection of the wave length might lead to large errors in analysis. It is also desirable Table I. Calibration with Known Mixture to avoid regions where one component is practically completely (For data on pure compq.nents, see Table 11. Mole fraction acetone absorbing in the existing cell path length, because the absorbance determined by weighing = 0.612) is difficult to measure accurately and may not show a linear variaOptical Analysis Mo1.e Wave TransFraction tion with concentration. Optimum conditions would be replength mittance Absorbance Acetone resented by a maximum or minimum in the transmittance curve 2400 0.2865 0.543 0.608 of one component associated with a widely different transmit0.612 2850 0.452 0.345 tance for the other component a t the same wave length.
/I
INDUSTRIAL A N D E N G I N E E R I N G C H E M I S T R Y
958
Vol. 48, No. 5
Table 11. Data and Results Observed TPmp.,
Absorbing Medium
Equil. liquid
c.
Equil. vapor
Pressure, Inches, Hg
Quartz disks
Equil. Vapor Wave Temp., C. Length, .4. (29.92 Inches Hg) Experiment 1 2400
80.2
80.0
29.72
80.2
Cell a n d sample
Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid
105.2 102.5 97.4 95.2 7.07 6.97 97.3 94.9 75.8 73.8 25.9 25.4 105.1 102.3 97.4
2400
Vapor Liquid
2850
Vapor
12.3 8.59 71.3 83.6 18.0 10,97 57.9 73.7 28.1 17.3 45.3 57.8 104.9 101.9 97.2 94.8
11.7 8.43 73.4 88.2 17.1 10.75 59.6 77.8 26.8 17.0 46.6 61 .O
0.232 0.091 0.235 0,095 0.392 0.196 0.394 0,196 0.582 0.387 0.581 0.376
41.1 28.4 37.2 43.3 57,l 47.8 30.0 32.0 64 9 58.6 28.5 28.8 104.8 101.6 97. I 94.5
39.2 27.9 38.3 45.7 54.5 47.0 a1 . 5 33.9
0.742 0.598 0.730 0.598 0.881 0.817 0.88.5 0,827 0.934 0,905 0.933 0.903
2400 2850
Pure acetone
56.6
56.4
29.72
56.6
2400 2850 2400
Quartz disks
2850 Equilibrium Run 1
Run 2
75.6
72.2
75.6
72.1
29.51
29.51
76.0
72.5
2400 2850
Run 3
66.5
66.4
29.51
66.8
2400 28.50 2400
Quartz disks (check) Equilibrium Run 4
2850 61.8
61.7
29.38
62.2
3400 2850
Run 5
58.0
58.0
29.38
58.5
2400 2850
Run 6
56.8
56.7
29.38
57.2
2400 2850
Quartz disks (check) a
2400 2860
Measured at 0.60-mm. slit width and referred t o standard quartz disks a s 100%.
constant a t 110' C. by varying the current in a resistance winding around the inlet lines t o the cells. Because i t was difficult to prevent leakage a t the analysis cell, no attempt was made t o operate a t a controlled pressure other than the existing barometric pressure. The analysis cell consisted of a glass cylinder about 14 mm. in length and 1.0 inch in diameter, divided in the middle by a glass partition. Quartz windows were placed on the ground ends of the cylinder to complete the cell and t o allow the transmittance of ultraviolet radiation. The end windows were held by clamps. An attempt to find a suitable sealing compound was abandoned, as all compounds tried formed a film on the quartz windows. For calibration purposes a reference cell containing two similar quartz windows was mounted beside the measurement cell. The cell assembly was built t o permit ready insertion into a Beckman DU quartz spectrophotometer equipped with hydrogen lamp. The instrument was mounted on tracks equipped with stops, so t h a t it could readily be moved to reproducible positions, sighting through either measurement cell or the rcference cell. Because the apparatus was designed primarily t o test the applicability of a combination of principles of analysis rather than to serve as a working tool for routine analyses, it is not described in greater detail.
Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor
Liquid Vapor Liquid Vapor Liquid
T'apor Liquid Vapor Liquid Vapor
Liquid Vapor Liquid Vapor Liquid Vapor Liquid
Sam& alone
Mole Fraction Acetone
Cell
2850
Pure benzene
-~ TransmittanceaLyo_
6.77
0,000
99.9 72.1
1,000
26.7
92.0
Fl,Q
07.7 29.4 30.5
(Coniinued o n pnge 959)
EXPERIWERTAL PROCEDURE
Purification of Materials. Both the acetone and the benzmc. were Merch reagent ACS grade, further purified by the methods suggested by Bramley ( 1 ) . Acetone was treated with potassium permanganate for several days, dried over calcium chloride, and fractionally distilled through a 5-foot column packed with glass helices. Benzene was treated with several portions of sulfuric acid, washed with water, dried first over calcium chloride and then over sodium, and finally fractionated through the same column used for acetone. Transmittance of Pure Components. Benzene or acetone, purified as above, was added to the boiling flask containing a boiling chip and heat was supplied through an electrical resistance winding. The upper portion of the flask was insulated by a temporary covering. The vapor was superheated by a separately controlled resistance winding on the vapor lines. The transmittance of the pure vapor was measured a t various wave lengths from 2100 t o 3200 A., using a slit width of 0.6 mm. in order to permit selection of wave lengths suitable for analysis. These were chosen a t 2300 and 2850 A. I n order to check the operation of the system, a liquid sample was also continuously withdrawn from the bottom of the flask, flashed, superheated by means of a separately controlled resistance winding to the same temperature as the vapor sample, and passed to the analysis cell. If the transmittance of the flashed
May 1956
INDUSTRIAL AND ENGINEERING CHEMISTRY
959
Table 11. Data and Results (Continued) Absorbing Medium
Observed Temp., Equil. Equil. liquid vapor
Pressure, Inches Hg
Equil. Vapor
Wave Length, A. Experiment 2
Temp., C. (29.92Inahes Hgf
2400
Quartz disks
2850 Pure acetone
56.2
56.1
29.57
56.8
2400 2850
Equilibrium Run 1
56.6
60.5
29.57
5G.Q
2400 2850
Run 2
57.9
57.8
20.57
58.2
2400 2850
Run 3
G1.3
61.3
29.57
61.7
2400 2850 2400
Quartz disks (oheck)
Pure benzene
2850 79.7
79.6
29.45
80.1
2400 2850
Equilibrium Run 4
75.6
75.4
29.45
75.8
2400 2850
Run 5
70.2
70.1
29 45
70.5
2400 2850
Run 6
65.5
65.3
29.45
65.8
2400 2850
Quartz disks (final check)
a
2400 2850
Cell
Transmittance", % Cell and Sample sample alone
Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid
100.7 99.2 94.2 94.2 73.3 72.3 25.2 25.3
Vapor Liquid Vapor Liquid Vapor Liquid Vapqr Liquid Vapor Liquid Vapor Liquid Vapqr Liquid Vapor Liquid Vapor Liquid Vapor Liquid
03.5 57.9 27.5 28.7 54.7 46.5 29.4 32.7 29.1 35.2 41.6 100.5 98.6 94.1 93.5 6.67 6.52 93.9 93.4
Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid Vapor Liquid
11.8 8.11 67.8 82.5 19.9 11.9 51.6 67.9 28.7 18.9 42.1 52.8 100.3 98.1 94.0 93.3
40.3
72.8
Mole Fraction Acetone
1.000
26.8 63.1 58.4 29.2 30.5 54.4 47.0 31.3 34.8 40.0 29.5 37.4 44.5
6.04
0.938 0,908 0.940 0.903 0.878 0.818 0.884 0 823 0.752 0.624 0.748 0.615
0.000
99.8 11.8 8.24 72.1 88.2 19.8 12.1 54.9 72.7 28.6 19.3 44.8 56.0
0.243 0.095 0.250 0.098 0.459 0.254 0.457 0,243 0.611
0.449 0.613 0.434
Measured a t O.6O-mm. slit width and referred t o standard quartz disks as 100%.
liquid sample did not closely check that of the vapor sample, the usual difficulty was an accumulation of a film on the end plates of the cell. These plates were removed and cleaned and t h e transmittance was checked again until agreement was obtained. Calibration with Known Liquid. I n order t o have a direct check on the suitability of the Beer-Lambert law in analysis of benzene-acetone vapor mixtures, without the added complications brought on by vapor-liquid equilibrium, a known sample was analyzed. A known liquid mixture was prepared by carefully weighing the two components and was introduced a t room temperature into the cold boiling flask. Cold liquid was slowly withdrawn into the flash tube, totally vaporized, superheated to the same temperature used in other parts of this work, and analyzed in the cell. The results are shown in Table I. The agreement at the two wave lengths between actual and analyzed compositions justified application of the method. Analysis of Mixtures. Each run on a mixture was begun by the introduction of one pure component into the still. A check was made at one wave length on its absorbance in both vapor and vaporized liquid cells, to be certain that the system was clean and in good working order. The second component was then added in an amount required t o give the desired approximate liquid composition. After the system had boiled and recirculated for 30 to 45 minutes, transmittance readings were made a t two wave lengths over a period of 20 to 30 minutes. The equilibrium vapor
temperature in the still was measured by a calibrated thermocouple. Additional amounts of the second component were introduced and the analysis procedure was repeated. When several compositions had been analyzed, the still was drained and the transmittance of the empty cell checked. This transmittance was observed to decrease slightly as a film formed on the quartz disks. The still was flushed with the second pure component, and drained again. A fresh portion of the second pure component was added and the absorbance determined in both cells. A second series of analyses was made after addition of successive portions of the first component. All determinations were made a t the same cell temperature of 110' C., at approximately the same cell pressure, and using a fixed slit width of 0.6 mm. RESULTS
The data and results are summarized in Table I1 for two independent experiments, each analyzed a t two wave lengths and covering a wide range of compositions. The results averaged for the two wave lengths are plotted in Figure 2 as a vaporliquid equilibrium diagram and in Figure 3 as a boiling pointcomposition diagram corrected t o 1 atm. total pressure. The line plotted in Figure 2 is actually t h a t of Othmer (g). The mean deviation of the individual results (before averaging at the two wave lengths) from these previously published values
960
INDUSTRIAL AND ENGINEERING CHEMISTRY
Vol. 48, No. 5
1.0
K
2
0.8
80
0.6
70
0.4
60
a >
z2 0 IW
v
a
0 I-
o
a
W
IL
eEXPERIMENT 2
a
(r
0.2
50 - - - 0 T H M E R
W
50
1
,
I-
E
l
0.2
1
1 0.4
I 0.6
I
I
I
0.8
+,?-I-
1
I .o
XA, MOLE FRACTION ACETONE IN LIQUID Figure 2.
Vapor-liquid equilibrium diagram for acetonebenzeri e
~
~
1
1
1
I 1 ~
'
I
MOLE F R A C T I O N A C E T O N E Figure
3.
Temperature-composition acetone- benzene
diagram
for
Solid line from Othrner (2)
is 0,55Tc (absolute value) with a maximum deviation of l.Si.ib. The boiling point data are all within 0.7" C. of those reported by Othmer, but, except for the pure components, fall consistently below his results. The two runs gave about the same deviations, and there appeared to be no significant difference in accuracy of measurement a t the two wave lengths. Agreement with the published results is sufficiently close to justify extension of this work into the more useful fields of threecomponent mixtures a t elevated pressures. The acetone-benzene-ethyl alcohol system could be studied, for example, with data taken only a t the two wave lengths studied in this work, since ethyl alcohol is essentially nonabsorbing. It is believed that greater accuracy can be obtained by closer control of tempcrature of the superheated vapor. This is being done in an apparatus now under construction for use a t elevated pressures. SUMMARY AND CONCLUSIONS
Spectrophotometric analysis has been successfully applied to the determination of vapor-liquid equilibrium data in a modified Othmer still. Both liquid and vapor compositions are determined in the superheated vapor phase without removal of samples from the equilibrium still. Although the equipment described is restricted t o use near 1-atm. pressure, modifications are possible for extending the usefulness of the method t o pressure or vacuum operation. The system acetone-benzene as analyzed spcctrophotometrically using the Beer-Lambert law compares with published data with a mean deviation of 0.55% and a maximum deviation of 1.8a/o (absolute value). The method is most simply applied to those systems which obey the perfect gas laws and which form ideal solutions with no mutual interference with absorption characteristics. However, by determination of extinction coefficients of pure components under the conditions of analysis, and by the use of empirical calibration with known mixtures, these limitations may be overcome. For successful analysis, the components present must have
significantly different absorption characteristics. With multicomponent systems, the relative absorption of light by the various components must vary significantly with wave length. Readings must be made a t a number of wave lengths equal to one less than the number of components. ACKNOW LEDGh? E S T
Acknowledgment is sincerely given to A. E. Parker for help with the optical phases of this work and to the Norton Co. for grinding the analysis cells. The Research Corp. provided a financial grant in support of this study, which is gratefully acltnowledged. A. J. Tasso contributed significantly t o the early stages of this work. NOMENCLATURE
= concentration of absorbing component D = absorbance of sample E = extinction coefficient c
zo
= intensity of radiation entering sample
z =
intensity of radiation leaving sample = length of light path through sample = number of moles of gas in absorption cell = pressure in cell = standard reference pressure t = transmittance of sample T = absolute temperature in cell T , = standard reference absolute temperature volume of absorption cell ,T' = molal volume of gases analyzed under standard rrfrrence conditions x = mole fraction in liquid phase Y = mole fraction in vapor phase
L .\* P P,
v =
LITERATURE CITED
(1) Braniley, A . J., J . Chem. Soc 109, 10 (1916). (2) Othmer, D. F.,IND. ESG. CHEM.35, 614 (1943); And. Chem. 20,
763 (1948). (3) Robinson, C. S.,Gilliland, E. R., "Elements of Fractional Distillation," 4th ed., McGraw-Hill, New York, 1950. RECEIVED f o r review August 10, 1953.
ACCEPTEDDecember 7 , 1955.
