Ind. Eng. Chem. Res. 1988,27, 2127-2135
2127
Distillation Efficiencies for Methanol/l -Propanol/Water Michael W. Biddulph* Department of Chemical Engineering, University of Nottingham, University Park, Nottingham NG7 2RD, U.K.
Mohammad A. Kalbassi British Steel Corporation, Welsh Laboratory and Strip Mill Products, Corporate Research Organization, Port Talbot, W . Glam SA13 2NG, U.K.
An experimental study of the distillation of the nonideal system methanol/ 1-propanol/water is reported. The study involved the use of two different types of sieve tray column. These were (a) a rectangular column with a flow path of about 1 m and (b) a small Oldershaw column, specially modified to avoid “wall effects”. The rectangular column was designed to attempt to represent a slice across a conventional tray. The tray efficiencies observed on the rectangular tray were high and showed different values for each component. The point efficiencies deduced from them were in some cases equal for all components and in other cases different. Experimental point efficiencies from the Oldershaw column and various predicted point efficiencies were compared with them. There has recently been considerable interest in improving our knowledge of the characteristics of distillation of multicomponent mixtures on large chordal-weir crossflow trays. This is because it has become clear that we do not know how to predict satisfactorily individual component efficiencies, which can be different from each other, and furthermore that these large trays are operating much less efficiently than they should. This latter effect results from stagnant zones and flow nonuniformities which are now known to exist on conventional large circular trays. Thus, the effort to improve the design of these columns incorporates two aspects. Firstly, it is important to understand more about component efficiencies in multicomponent systems, particularly ways of easily measuring or predicting these. Secondly, it is essential to develop new trays to avoid the detrimental effects of flow nonuniformities and thus enable commercial-scale columns to achieve the very high efficiencies which are now known to be possible. This paper forms part of a continuing effort on the first of these aspects. Later publications will be concerned with the second. Multicomponent systems are broadly divided into two categories. Firstly, there are the thermodynamically ideal systems being composed of components having similar molecular structures. Close members of a homologous series of components with the same order of polarity fall into this category. Secondly, there are the thermodynamically nonideal systems in which the components are of different molecular structure and polarity. Toor (1957) showed theoretically that, for thermodynamically nonideal systems, there should be marked differences between the binary and ternary mass transfer, these differences arising from interactions between the diffusing species. This was confirmed by the early investigations into multicomponent efficiencies by Nord (1943) and Qureshi and Smith (1958), where different component efficiencies were reported. In a nonideal system, the individual components have different diffusion coefficients, and in addition diffusional interactions between components appear to play an important role. Figure 1 shows a comparison of the diffusivities of the binary alcohol/alcohol and alcohol/water pairs of interest in this investigation. It can be seen that there are large differences between these pairs, so it may be anticipated (Toor, 1957) that significant interaction effects may exist in a multicomponent mixture of alcohols and water and that this may result in the individual com-
* Author t o whom
correspondence should be addressed.
0888-5885/88/2627-2127$01.50/0
Table I. Tray Details for the Rectangular Column tray thickness 2 mm weir length 83 mm liquid flow path 991 mm 154 mm tray spacing 1.8 mm hole diameter 6 mm hole pitch 12.7 mm outlet weir height 70 free area 8% 4.8 mm inlet weir height
ponents exhibiting different point efficiencies. It has been shown previously theoretically (Biddulph, 1975) and experimentally (Dribika, 1986; Biddulph, 1975) that, for ideal systems exhibiting equal point efficiencies, the component tray efficiencies are different from each other when the tray has a significant flow path length. This results from the form of the definition of tray efficiency and the differing component volatilities, giving rise to differing concentration gradients. This experimental work was carried out on the system methanol/ethanol/ 1propanol. In the case of a nonideal system, such as that under investigation here, the picture is more complicated, as described above. In this paper, the results obtained from a study of the nonideal system methanol/ 1-propanol/water are reported, using both a rectangular crossflow tray with a path length of about 1 m (Biddulph and Dribika, 1986) and an 01dershaw column of special design (Kalbassi and Biddulph, 1987). The intention is to improve our understanding of the behavior of these systems and to assess the feasibility of using point efficiencies measured in a laboratory-scale column in the design of larger trays. The middle components in such systems are known to exhibit maxima in concentration (Cilianu et al., 1974; Dribika, 1986), and this introduces difficulties which are considered here. Lockett (1986), in his recent review, has emphasized the need to test multicomponentefficiency prediction methods (Diener and Gerster, 1968; Krishna et al., 1977; Medina et al., 1979) against experimental data, and this is also attempted here.
Rectangular Column Study The equipment used has been described in detail previously (Biddulph and Dribika, 1986), and this will now be outlined briefly. The steam heated, stainless steel reboiler, R in Figure 2, has a capacity of 450 L. The rectangular section distillation column, D, has three trays with dimensions 991 mm X 83 mm and a tray spacing of 154 mm. The sieve tray material is typical of that used cur0 1988 American Chemical Society
2128 Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 Table 11. Binarv Wilson Parameters system parameters ref MeOH/ 421.821, 245.905 Dribika and Biddulph, 1986 I-PrOH MeOH/water 216.851, 468.601 Gmehling and Onken, 1977 I-PrOH/water 906.526, 1396.6398 Gmehling and Onken, 1977 ~
t 005
00
01
02
04
03
0.6
05
OB
07
Q9
1.0
~~
measurements. The parameters used are given in Table 11, and these resulted in mole fraction predictions within about 0.11 mol %. Bubble point temperatures were predicted within about 0.8 O C . The samples taken from the column were analyzed by gas chromatography, with an accuracy of f0.53 mol %.
X p o i e fPaCt.m mope v o l a t i l e :omPcnent
Figure 1. Binary diffusion coefficients.
c
D F R
h CONDENSERS
FLOWMETER HEBOILER
WWINDoWS S
AJt
DISTILATION COLUMN
WATER
Results The experimental composition profiles along the center line of the rectangular tray were matched with those predicted by using the previously published model (Biddulph, 1975), based on the eddy diffusion concept. The equations are as follows: Zi = dxi/dW
IN
(1)
SAMPLERlINT
n
t
81
I
where
b S T E A M IN
I R 1 Figure 2. General view of the rectangular column.
rently in the low-temperature air distillation industry, 2-mm-thick aluminum with 1.8-mm holes. Details of the tray are given in Table I. Double-glazed portholes are provided above the middle test tray to enable easy observation of the biphase. The glass condensers, C, provide a condensing surface area of 5.3 m2,and all the hot surfaces are insulated with 50-mm-thick Fiberglass and aluminum cladding. Eleven sample points are provided, S1-S11, including six, S3-S8, spaced at equal intervals along the center line of the test tray. The liquid samples are drawn out through narrow, insulated tubing set into the floor of the tray. These enable the concentration profiles to be established, and each of these points also has a thermocouple for temperature measurement. The temperatures are displayed on an instrument board for easy observation. It is hoped that the tray represents a "slice" across a correctly operating circular tray between the inlet and outlet weirs. All the experiments were carried out at atmospheric pressure and total reflux, with a vapor I: factor of about 0.5 (m/s) (kg/m3)0.5.The velocity used in this factor is the superficial value based on the bubbling area. This is similar to the rates used in some air separation columns, although much lower than is common in commercial-scale columns in other applications. The vapor/liquid equilibrium data used were obtained by Ochi and Kojima (1969) and compiled by Gmehling and Onken (1977). The Wilson model, incorporating binary parameters for the pairs constituting the ternary system, was used to test the predictions against the reported
The equations can be used in conjunction with the component point efficiencies to simulate the tray behavior and hence deduce the values of point efficiency. Thus, two of the three point efficiencies were successively varied until a satisfactory match with all three composition profiles was achieved. The third point efficiency is fixed by mass balance considerations. In principle, it does not matter which two components are chosen as those assumed, but in practice the middle component was avoided since this efficiency can take values over a wide range in certain composition ranges. An essential piece of data is the extent of backmixing on the tray, and this comes from earlier tracer studies of this tray using sodium nitrate in steam/water (Biddulph and Dribika, 1986). These studies indicated a value of the eddy diffusion coefficient of about 0.0014 m2/s, resulting in a Peclet number of about 39. The vapor and liquid enthalpy values were obtained from standard steam tables and Kern (1950). The distribution coefficients, K values, were calculated from VLE data computations,taking into account the nonidealities in both phases (Prausnitz et al., 1976). Table I11 summarizes the measured data from eight runs. It can be seen that the tray efficiencies of the individual components, calculated from the inlet and outlet downcomer compositions, were significantly different from each other. This is in accord with earlier theoretical predictions (Biddulph, 1975). The measured composition and temperature profiles are shown as the graph points on Figures 3 and 4. A comparison of the measured temperatures with the calculated bubblepoint values is shown in Figure 5. The measured temperatures tended to be somewhat higher than the bubble-point temperatures, as would be expected. In runs WA and WB, water was transferring weakly from the liquid to the vapor phase, and the system was
Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 2129
R U N :W A
MODEL
RUN:WB
~
0.6
06 DIENER
0.5 0.4
RUN: WC
MeOH
~
WATER
4
05 -
MEDINA ET A L
__
KRISHNA ETAL
-----.
0.5
X
0.4 0.3
X
0.3
0.2
02
0.1
INLET
2
I
5
3
6
-
a2
t
OOTLKT
POBITION ON T A E T R A Y
0.0
0.01
3
2
1
INLWI
4
6 om~m
5
INLET
,
,
,
,
,
,
1
2
3
4
5
6
POBITION ON THE TRAY
1
J
OWL=
PWITION ON TEE TRAT
RUN:WD
0.7
(16
0.5
X
0.5 .
0.2
X
-
0.01 INLIT
'
1
*
.
.
I
2
3
4
-
03 .
1
.
,
a4
0.3
6 ow~mr
5
POBlTlON O N TUG TRAT
0.7
0.7 I
0.6
0.6
0.5
0.5
I
0.4 X
0.3
0.2
t
O.I 0.01 INLET
, 1
, 2
.
.
3
4
,
5
.
1
J
6 om~m
POBITION ON THE T R A I
0.01 INLCT
1
1
,
,
,
2
3
4
,
5
.
1
6 om~m
POBITION ON TEE T R A l
Figure 3. Composition profiles of the ternary MeOH/1-PrOH/H20 system across the tray.
surface tension negative according to the usual classification. In runs WC and WD, the water had reached its maximum in concentration, and the system was effectively surface tension neutral. In runs WE to WI, the system was
surface tension positive. As would be expected from this, the froth heights for runs WA to WD were lower. The procedure described above was used to infer the values of point efficiency for the components for each run,
2130 Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 Table IV. Inferred Point Efficiencies on the Rectangular Tray and Modeled Tray Efficiencies point efficiency, % tray efficiency, % run Eoe, Eoe? Eoe, Emv, Emvl WA 70 92 248 128 133 WB 70 86 13 111 110 75 70 60 105 127 WC 84 WD 73 46 117 117 33 70 70 140 WE 70 75 151 86 WF 75 75 2 83 83 83 154 WG WI 223 17 77 70 145
Emvl 173 173 124 118 88 105 111 89
85
80
i75 0
23 C m
70 Md)H/nPrOn/H20 INLET I
P
POSTION ON
3
I
0
8
.
I
OUTLET
+
T H E TRAY
Figure 4. Temperature profiles.
Figure 6. Modified Oldershaw column. Table V. Details of the Oldershaw Column tray diameter 3.8 cm column diameter above tray 6.4 cm no. of holes 46 1.1 mm hole diameter weir height 2.0 mm % free area 8% hole pitch 3.8 mm
6L
6G
68
72
76
80
84
BUBBLE P O I N T TEMPERATURE
88
92
‘C
Figure 5. Comparison of measured and bubble temperatures.
and these are shown in Table IV. These individual component point efficiencies can be seen to have been significantly different from each other for runs WA, WB, WC, WD, and WI, whereas in runs WE, WF, and WG, essentially equal point efficiencies appeared to be operating. A comparison of Tables I11 and IV indicates that the measured and predicted tray efficiencies for components not showing maxima in composition are very similar, as they should be.
Oldershaw Column Study The second column used to establish experimental efficiencies in this study was a small laboratory “Oldershaw” column of somewhat modified design and is illustrated in
Figure 6. The modifications have been described in detail elsewhere (Kalbassi and Biddulph, 1987), and briefly they are as follows. The column has external downcomere and has an increased diameter above the plate to prevent the problems of “wall-supportedfroth”, which are common in more conventional designs. It has been shown previously (Kalbassi and Biddulph, 1987) that this design provides much more realistic values of point efficiency in the case of surface tension positive systems than does the conventional Oldershaw design. Table V gives details of the column. The experiments were all carried out at the same F factor as used in the larger column, that is 0.5 (m/s) (kg/m2)0.5. Results. The Murphree point efficiencies for the ternary system studied here are shown in Table VI as a function of composition and second and third component K values. The biphase was observed closely for all the runs. The froth height varied between 1.5 and 2.5 cm. It was at its lowest, and seemed to be less bubbly, for runs 202-206. An examination of these runs reveals that at these concentrations the system would have been slightly neg-
Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 2131 Table VI. Modified Oldershaw Column Point Efficiencies System: MeOH/l-PrOH/H,O
08
run
Xl
x2
K3
K2
Eog,
Eogz
E%,
202 203 204 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 235 236 237
0.2707 0.2231 0.1911 0.1431 0.1699 0.1511 0.1322 0.1143 0.0977 0.0818 0.063 0.0522 0.0985 0.0867 0.0692 0.0572 0.0443 0.0491 0.1040 0.1850 0.2837 0.3360 0.4501 0.5271 0.5725 0.6322 0.6891 0.7359 0.7542 0.7862 0.7722 0.7864 0.8424 0.8651
0.5911 0.5454 0.5139 0.4784 0.4357 0.4292 0.4145 0.4100 0.3919 0.3912 0.3986 0.4032 0.3598 0.3626 0.3627 0.3403 0.2877 0.3226 0.2750 0.2233 0.1688 0.1499 0.1011 0.0723 0.0587 0.041 0.0256 0.0156 0.0110 0.0062 0.0221 0.0403 0.027 0.0211
1.23 1.14 1.08 1.03 0.92 0.91 0.89 0.89 0.88 0.88 0.90 0.91 0.84 0.85 0.86 0.84 0.79 0.82 0.75 0.68 0.61 0.58 0.52 0.40 0.47 0.45 0.43 0.43 0.42 0.42 0.44 0.43 0.44 0.44
0.56 0.61 0.65 0.17 0.75 0.78 0.83 0.87 0.92 0.95 0.97 0.99 0.97 0.99 1.03 1.11 1.30 1.18 1.16 1.08 0.98 0.90 0.74 0.67 0.64 0.56 0.50 0.46 0.9 0.42 0.42 0.39 0.36 0.35
0.60 0.56 0.56 0.57 0.57 0.57 0.58 0.60 0.65 0.71 0.78 0.77 0.73 0.76 0.78 0.80 0.82 0.84 0.78 0.78 0.76 0.71 0.74 0.66 0.72 0.73 0.75 0.78 0.69 0.78 0.73 0.87 0.79 0.79
0.66 0.66 0.68 0.72 0.71 0.72 0.76 0.80 0.82 0.90 0.80 1.21 1.40 2.31 0.70 0.71 0.63 0.75 0.59 0.08 3.56 0.86 0.86 0.43 0.62 0.68 0.73 0.83 0.84 0.91 0.27 0.86 0.81 0.80
1.26 1.56 2.58 -2.81 0.08 0.20 0.29 0.42 0.56 0.65 0.78 0.73 0.65 0.71 0.77 0.77 0.71 0.80 0.72 0.71 0.71 0.73 0.73 0.68 0.73 0.76 0.76 0.78 0.73 0.77 0.76 0.87 0.79 0.79
.
.
E"g
06
04
Figure 7. Point efficiencies of methanol in the ternary MeOH/lPrOH/H,O system.
ative, with water transferring from the liquid to the vapor phase (see the K values of water in Table VI). In the rest of the runs, the biphase was more bubbly, these being slightly positive. The biphase increased in height from 1.5 to 2.5 cm as more water was added to the reboiler. In Figure 7 the point efficiency of methanol is plotted against its composition on the tray. There is a decrease in the methanol point efficiency corresponding to the negative runs. This is probably partly due to the lower biphase height for these runs. Interestingly, the individual component point efficiencies are different from each other, as would be expected for a nonideal ternary system. The system shows maxima in composition for the middle components (water in the cases of runs 202-207), and the point efficiencies of these components were sometimes outside the range 0-1.0 as the volatility of the component passed through unity (i-e., its composition maximum). However, at high methanol compositions, the differences between individual point efficiencies were found to be small for some of the runs (e.g., 229, 235, 236, and 237). This supports the theory of interaction effects, as at high methanol compositions the content of the polar, that is
water, and large molecule, that is propanol, components are markedly reduced.
Interpretation of Results Prediction of Individual Component Point Efficiencies. These methods are based on interpretations of the Maxwell-Stefan equations for diffusion and their application to ternary distillation using binary data. The individual component point efficiencies are predicted by applying the equations of Diener and Gerster (1968), Krishna et al. (1977), Krishna (1977), and Medina et al. (1979) to the system methanol/l-propanol/water. A summary of these methods is given by Lockett (1986). The following assumptions are made: (i) equimolar mass transfer, (ii) no influence of finite mass-transfer rates on the mass-transfer coefficient, and (iii) thermodynamic correction factors neglected. It is also assumed that gas-phase resistance to mass transfer is controlling, although it is known that the liquid-phase resistance can be significant (Lockett, 1986). The following steps are taken to carry out the calculations: (a) Evaluation of Binary Overall, Liquid Phase, and Vapor Phase Number of Transfer Units. These can be obtained experimentally by carrying out experiments under similar hydrodynamic conditions. This is the approach taken here whereby the binary experimental data in the rectangular distillation unit are used. These binary data can also be calculated by using the standard procedures available (see Chan and Fair (1984)). The numbers of liquid-phase and vapor-phase transfer units were available from experiments using the rectangular column for the binaries constituting the system MeOH/l-PrOH/H20 from Dribika (1986) and Kalbassi et al. (1987) for the system 1-PrOH/H20. These experiments were all carried out at similar hydrodynamic conditions. Table VI1 summarizes the values of NLand NG.
2132 Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 Table VII. Values of Nn and Nr svstem N ", MeOH/H20 2.56 MeOH / 1-PrOH 1.61 1-PrOH/H20 1.88
NT
I
12.5 5.83 7.05
The vapor phase was controlling the mass transfer for all these binary systems. (b) Evaluation of Equivalent Ternary Transfer Units (Method of Diener and Gerster (1968)). Diener and Gerster (1968) suggested the following equations: (4) NTGll = NG13[~1NG23+ (1-YJNGI,I/S
-NGd/S NTGl2 = ~1NG23(NG13
(5)
NTGzl = ~2NG13(NG23- NG12) / S
(6)
NTG22 = N G , ~ [ Y ~ N G +~(1, - Y ~ ) N G ~ ~ I / (7) S where
s
= Y1NG23 + YPNG13 + Y3NGlP (8) Equations 4-8 are used to calculate the ternary equivalent liquid-phase number of transfer units (NTL) using binary liquid-phase transfer units (NLij),by substituting NTL for NTG and NL, for NGip It should be noted that no theory was developed to take into account the thermodynamic nonidealities in the liquid phase. This is the main reason why the vapor-phaseresistance to mass transfer is required from the binaries, if the above theory is to be used. In this context, Krishna (1980) questioned the work by Medina et al. (1979) regarding the effect of surface tension on mass transfer, as the multicomponent theory does not take this into account. The surface tension of the ternary MeOH/ l-PrOH/H,O system measured at boiling point was obtained by Kalbassi (1987) by using a glass tensiometer. Surface tension gradients are only significant at very low alcohol concentrations and will not have any significant effect at higher alcohol concentrations where the experiments are carried out. (c) Ternary Equivalent Slopes of the Equilibrium Line (mjj). These are calculated from theoretical tray column simulations taken from two adjacent trays, p + 1 and p :
m11 = m12 = m21 = m22 =
Yl,P+l
- Yl,P
Xl,p+l
- Xl,P
Yl,P+l
- Yl,P
X%p+l
-
Y2,p+l
- Y2,p
Xlp+1
- x1,p
Y2,p+1
- Y2,p
XZ,P+l
- J2,p
(9)
(10) X2,P
(11)
-NOG21
-NOG2,
Eog = 0.8482 system MeOH/ 1-PrOH
+ 0.101~
(14)
Eog = 0.6449
+ 0.166~
(15)
system 1-PrOH/H20 Eog = 1.0048 - 1 . 7 4 ~+ 4 . 4 ~ 3~ . 1 9 ~ ~ (16) Note that all the above binary measurements were carried out under the same running conditions as the ternary measurements. The composition of the vapor leaving the test tray was evaluated by the following equations, since total reflux was used: (17) Yp,l = Y*P,l + G11(xp,1 - Y*P,l) + G12(Xp,2 - Y*P,2) Yp,2
= Y*P,Z + G2l(Xp,l - Y*P,l) + G22(Xp,2 - Y*P,2) Yp,3
- 1 - Yp,l - Yp,2
(12)
(18) (19)
The individual component point efficiencies were then calculated by using the Murphree equation, for total reflux: (20) Eogn,, = (YpJ - xp,l)/(Y*p,l - X p , J
Method of Krishna et al. (1977). Equations 4-8 and 13 were used to compute the ternary equivalent gas-phase numbers of transfer units, using the NG values as tabulated in Table VII. The individual component point efficiencies were then calculated by using the following equations: Eog, = Eog11 + Eog12/r (21) Eog2 =
(d) Ternary Overall Gas-Phase Numbers of Transfer Units (NOGij). These are then calculated by combining the ternary gas- and liquid-phase transfer units in the same manner as for a binary system by using the two film theory. Diener and Gerster (1968) give further details of the equations used. (e) Evaluation of the Elements of the Matrix. exp[-NOG] = exp
The elements Gll, G12,G21,and Gzz are obtained by using Silvester's theorem and are given by Diener and Gerster (1968), with a slightly different form given by Krishna et al. (1977). Calculations of vapor compositions leaving the test tray, ylp, y2p,and y3p,follow from these calculations (Diener and Gerster, 1968). ( f ) Calculations of the Individual Component Point Efficiencies. The point efficiencies Eog,, Eog,, and E0g3 follow from the compositions predicted. Method of Medina et al. (1979). All the basic steps as indicated by eq 4-8 and 13 are taken to calculate the overall number of transfer units, replacing NOGii by NGik The point efficiencies for the binary systems of interest were measured by Dribika (1986) and for the system 1PrOH/H20 measured as reported by Kalbassi (1987). These data were correlated by a least-mean-squares polynomial method to give the following equations: system MeOH/H20
+ Eog21/r
(22)
where (23) Eog11 = 1 - GI1 Eog12 = 4 1 2 (24) Eog21 = 4 2 1 (25) E0g22 = 1 - G22 (26) r is the ratio of driving forces of components 1 and 2. Comparison with Experimental Results. The point efficiencies for the rectangular column were predicted by using the three methods described previously using the average conditions obtained on the tray. These point efficiencies are compared with the predictions from the model in Table VIII. Included as well are point efficiencies measured in the modified column. The results from those runs where the composition in the Oldershaw
Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988 2133 Table VIII. Comparison of Point Efficiencies. System: MeOH/1-PrOH/H2On Eog1 Eogz Eog3 run model Kri Medi Diener O.C. model Kri Medi Diener O.C. model Kri Medi Diener O.C. WA 0.70 0.84 0.72 0.63 0.57 0.92 0.81 0.73 0.71 0.72 2.48 0.62 0.78 1.27 -2.81 WB 0.70 0.84 0.76 0.69 0.86 0.81 0.71 0.70 0.13 0.94 0.92 0.64
wc
WD WE WF WG WI Eogi - Eogimdd
0.70 0.73 0.7 0.75 0.83 0.77
0.85 0.85 0.91 0.92 0.90 0.89 0.14
0.77 0.76 0.85 0.85 0.83 0.82 0.06
0.71 0.69
0.61
0.87
0.78
0.03
0.08
0.75 0.84 0.7 0.75 0.83 2.23
0.82 0.82 0.84 0.85 0.84 0.84 0.08
0.72 0.71 0.87 0.85 1.33 0.96 0.13
0.70 0.69
0.69
1.33
0.73
0.16
0.12
0.6 0.462 0.70 0.75 0.83 0.70
0.92 0.93 0.89 0.88 0.85 0.89 0.13
0.91 0.91 0.84 0.85 0.90 0.84 0.11
0.73 0.69
0.52
0.89
0.72
" Model, value used in eddy diffusion model; Kri, Krishna predicted value; Medi, Medina et al. predicted value; Diener, Diener predicted value; O.C., Oldershaw column measured value. Table IX. Comparison of T r a y Efficiencies. System: MeOH/l-PrOH/H20 Emvl Emvz Emv3 run measd model Medi Diener O.C. measd model Medi Diener O.C. m e a d model Medi Diener O.C. WA 1.22 1.28 1.27 1.05 0.90 1.16 1.33 0.89 0.95 0.96 1.73 1.42 0.20 0.76 1.07 WB 1.17 1.11 1.27 1.09 1.16 1.10 0.94 0.92 1.73 1.29 9.6 6.9 wc 1.19 1.05 1.22 1.07 0.87 1.27 1.01 0.98 0.93 0.90 1.24 1.24 2.25 1.67 0.75 WD 1.19 1.17 1.27 1.08 1.12 1.17 0.95 0.91 1.42 1.18 4.7 2.82 WE 1.41 1.40 1.88 0.59 0.33 0.34 1.02 0.88 1.13 WF 1.50 1.51 1.78 0.77 0.86 1.22 0.98 1.05 0.98 WG 1.49 1.54 1.53 1.66 1.40 0.43 0.02 0.22 0.12 -0.25 1.16 1.11 1.17 1.12 0.93 WI 1.5 1.45 1.59 0.37 0.17 0.56 1.11 0.89 1.14 (Emvi - Emv,Bp)/n 0.06 0.14 0.13 0.22 0.18 0.25 0.25 0.4 0.18 0.05 "Model, value used in eddy diffusion model: Kri, Krishna Dredicted value; Medi, Medina et al. predicted value; Diener, Diener predicted value; O.C., Oldershaw column measured value.
column was very similar to the average composition in the rectangular column are included. These were runs 206, 246, and 222 matching runs WA, WC, and WG, respectively. The average deviation values of these point efficiencies inferred from the eddy diffusion model are also included in this table. Note that the statistical test did not include the efficiencies of the components passing through a maximum in concentration, as experimental errors are predominant in those cases. These point efficiencies from each prediction method were then incorporated into the eddy diffusion model, simulating the large column experimental runs to predict the composition profiles across the test tray. These composition profiles are included in Figure 3 and are compared with the experimental points. This comparison suggests that the methods of Diener and Gerster (1968) and Medina et al. (1977) are slightly better. The tray efficiencies inferred from the simulations are included in Table IX and compared with measured and model tray efficiencies, with the Diener and Gerster (1968) method showing the least deviation from the measurements. As expected, the individual components usually showed different point efficiencies due to the diffusional interactions arising from different molecular structure and polarity. However, there were some runs where equal component point efficiencies were operating on the tray, for both of the distillation columns used. This behavior could be the result of the interaction effects of reverse diffusion, diffusion barrier, or osmotic diffusion, due to large nonidealities and the very different diffusional characteristics of alcohol/alcohol and alcohol/water systems (see Figure 1). This is in agreement with the theory of Toor (19571, Krishna et al. (1977), and Krishna and Standart (1979). The intermediate components, H20 or 1-PrOH, were capable of transferring from vapor or liquid or could exhibit a concentration maximum in the composition profile across the tray. In most of the runs, the intermediate component showed the highest point efficiencies. The
I15 II6
20 0
/IO
I
Figure 8. Surface tension (mN/m) of MeOH/l-PrOH/H20 system at boiling point.
point efficiencies of the intermediate component were also found to go outside the interval (0-1.0) when concentration maxima occurred. This is the direct result of errors in evaluating the Murphree point efficiencies. However, these values will not have any effect on the prediction of the composition of that component (Medina et al., 1979). This system was capable of exhibiting both positive and negative surface tension behavior, although the surface tension driving force was fairly low for the composition ranges studied (see Figure 8). The biphase in the positive runs seemed more bubbly, whereas for the negative runs less bubbling but more spraying was observed. These observations are in agreement with works of Zuiderweg and Harmens (1958) and Bainbridge and Sawistowski (1964). However, the foamy biphase suggested by Zuiderweg and Harmens (1958) was never observed. This was because the modified column does not support foam or froth (Kalbassi
2134
Ind. Eng. Chem. Res., Vol. 27, No. 11, 1988
and Biddulph, 1987). The biphase in the negative runs was generally smaller in height than the positive system runs, which explains why lower point efficiencies were obtained. The eddy diffusion model (Biddulph, 1975) was found to represent the differences in component tray efficiencies, including the higher and the lower efficiency values. This model is flexible and requires N-1 component point efficiencies. This means that the component which shows a maximum in concentration can be left out, as its concentration is independent of the point efficiency, to carry out the simulation. One of the most important features of the work is the illustration of differences between component tray efficiencies, even when equal component point efficiencies were operating across the rectangular tray. This is a result of differing volatilities interacting with the limited backmixing on the tray. This influences the individual component composition gradient across the tray. Similar predictions were noted in a study of an air distillation column and an aromatics column, by Biddulph (1975) and Biddulph and Ashton (19771, respectively. These experimental findings confirm such predictions. The fact that component tray efficiencies can vary widely from one another in multicomponent systems, due to the thermodynamic nonidealities or the effect of liquid back-mixing, can obviously cast serious doubts on the validity of the normal design approach of using constant and equal component tray efficiencies or an overall column efficiency. Individual component point/tray efficiencies and composition profiles were predicted by using the methods of Krishna et al. (1977),Medina et al. (1979),and Diener and Gerster (1968), respectively. From these methods, the largest deviation from the measured values was obtained with the Krishna et al. (1977) method. This is due to using the number of gas-phase transfer units (NGJin the original computations to fulfil the assumption of no liquid-phase resistance to mass transfer. The other two methods are also based on the same assumption; however, since the overall number of transfer units (NOG,) takes into account the number of liquid-phase transfer units (NLLl), better predictions were obtained. An attempt has also been made to model the rectangular column distillation runs using Oldershaw column point efficiencies. These efficiencies were chosen from a number of runs made using this column with approximately similar compositions on the tray. The deviations of the tray efficienciesand the composition profiles obtained were similar to those using the prediction methods. However the work on this column indicated (Kalbassi, 1987) that better point efficiencies may be obtained by fitting this column with a 12.7-mm outlet weir. These efficiencies may be scaled up to values very close to the ones operating across the rectangular tray. This should allow the prediction of the point efficiencies for any system, either binary or multicomponent comprising three or more components. Kalbassi et al. (1987) demonstrated that hole size up to 6.4 mm has a relatively small effect on point efficiencies. This should allow the modified column to be used for the design of trays with different perforation sizes. Conclusion The point efficiencies of a highly nonideal ternary system were measured by using a special design of an 01dershaw column to remove wall effects and a rectangular column with a long flow path to avoid flow nonuniformities and stagnant zones. The individual component point efficiencies of this system were found to be different in the majority of the
runs, probably due to the interaction effects on individual component mass-transfer characteristics. The tray efficiencies of the rectangular column were found to be significantly different due to the differing volatilities interacting with limited liquid back-mixing on the tray. The modified column point efficiencies were successfully incorporated into the prediction of the efficiencies and composition profiles of the rectangular column. These profiles and efficiencieswere found to be compatible with measurements and predictions from the mass-transfer models. This should allow the measurement and prediction of the point efficiencies for the design or a large column for any system, including those comprising more than three components, where no prediction model is available. The individual component tray efficiencies were generally high, emphasizing the desirability of removing the stagnant zones and flow nonuniformities from conventional circular trays, and this is currently the subject of further study. Nomenclature D, = binary gas diffusivity, cm2/s Emv = tray efficiency Eog, point efficiency F = F factor, up,1/2 IGl = matrix elements h = liquid enthalpy, m H = vapor enthalpy, m L = liquid flow rate, kmol/(m s) m = Slope of the equilibrium line n = number of points considered N = number of components NOG = overall number of transfer units NTG = ternary equivalent number of gas-phase transfer units NTL = ternary equivalent number of liquid-phase transfer units NL = binary number of liquid-phase transfer units NG = binary number of gas-phase transfer units Pe = Peclet number T = temperature, "C u = vapor velocity, m/s V = vapor flow rate, kmol/(m2 s) W = dimensionless flow path length x = mole fraction liquid y = mole fraction vapor y* = equilibrium vapor mole fraction Zi= term defined by eq 1 ZI = liquid path length, m Greek Symbol pv = vapor density, kg/m3 Subscripts and Abbreviations I , 2, 3 = component number in order, lightest first b = bubble point BUB = bubble point i, j = component numbers p = trays p , p + 1, p - 1, numbering upwards w = at a point on the tray Medi = Medina M.C. = modified column Kri = Krishna Pred = predicted Exp = experimental Registry No. MeOH, 67-56-1; PrOH, 71-23-8.
Literature Cited Bainbridge, G. S.; Sawistowski, H. Chem. Eng. Sci. 1964, 19, 992. Biddulph, M. W. AZChE J. 1975, 21(3), 327. Biddulph, M. W.; Ashton, N. Chem. Eng. J. 1977, 14, 7.
Ind. Eng. Chem. Res. 1988,27, 2135-2139 Biddulph, M. W.; Dribika, M. M. AZChE J. 1986, 32, 8. Chan, H.; Fair, J. R. Znd. Eng. Chem. Process Des. Deu. 1984, 23, 820. Cilianu, S.;Brauch, V.; Schlunder, E. Verfuhrenstechnic 1974,8(3), 84. Diener, D. A.; Gerster, J. A. Ind. Eng. Chem. Process Des. Deu. 1968, 1(3), 339. Dribika, M. M. Ph.D. Thesis, University of Nottingham, U.K., 1986. Dribika, M. M.; Biddulph, M. W. AZChE J. 1986, 32(11), 1864. Gemhling, S.; Onken, U. Vapour-liquid Equilibrium Data; Dechema, Chemistry Series; Frankfurt, Germany, 1977; Vol. 1, p 24. Kalbassi, M. A. Ph.D. Thesis, University of Nottingham, U.K., 1987. Kalbassi, M. A.; Biddulph, M. W. Znd. Eng. Chem. Res. 1987, 26, 1127-1132. Kalbassi, M. A,; Dribika, M. M.; Biddulph, M. W.; Kler, A.; Lavin, J. T. Proceedings of the Institute of Chemical Engineers (London). International Symposium on Distillation, 1987; Znst. Chem. Eng. Symp. Ser. 1987, 104, A511. Kern, D. Q. Process Heat Transfer; McGraw-Hill: New York, 1950. Krishna, R. Chem. Eng. Sci. 1977, 32, 1197.
2135
Krishna, R. Chem. Eng. Sci. 1980, 35, 2371. Krishna, R.; Standart, G. L. Chem. Eng. Commun. 1979, 3, 201. Krishna, R.; Martinez, H. F.; Sreedhar, R.; Standart, G. I. Trans. Inst. Chem. Eng. 1977,55, 178. Lockett, M. J. Distillation Tray Fundamentals; Cambridge University Press: Cambridge, U.K., 1986. Medina, A. G.; McDermott, C.; Ashton, N. Chem. Eng. Sci. 1979,34, 861. Nord, M. Trans. Inst. Chem. Eng. 1943,42, 863. Ochi, K., Kojima, K. Kagaku Kogaku 1969, 33, 352. Prausnitz, J. M.; Eckert, C. A.; Orye, R. V.; O'Connell, J. P. Computer Calculations for Multicomponent Vapoul-liquid Equilibria; Prentice-Hall: Englewood Cliffs, NJ, 1967. Qureshi, A. K.; Smith, W. J. Znst. Pet. 1958, 44, 413. Toor, H. L. AIChE J. 1957, 3(2), 198. Zuiderweg, F. S.; Harmens, A. Chem. Eng. Sci. 1958,9, 89.
Receiued for review October 23, 1987 Revised manuscript received June 23, 1988 Accepted July 25, 1988
Preparation of Oxygen-Enriched Air by the Selective Adsorption with Platinum Crystallites Supported on ?-Alumina Hyung C. Kim and Seong I. Woo* Department of Chemical Engineering, Korea Advanced Institute of Science & Technology (KAIST), P.O. Box 131, Cheongryang, Seoul, Korea
A new method for the separation of the oxygen from air via selective adsorption of oxygen on highly dispersed Pt crystallites supported on y-alumina was investigated. The oxygen in air was selectively adsorbed on Pt/y-alumina a t 500 "C. The oxygen adsorbed on Pt crystallites was desorbed in a carrier flow of air by raising the temperature to 640 "C a t a rate of 10 "C/min. During this temperature swing process, the flow of air enriched with oxygen up to 25% was obtained. However, the extent of oxygen enrichment decreased significantly after four adsorption-desorption cycles. The redispersion of sintered Pt crystallites in an atmosphere of air for 2 h a t 500 "C restored the 80% of oxygen-enrichment capability of fresh Pt/y-alumina. Many papers and patents have appeared dealing with the separation of oxygen or nitrogen from air. Rutheven (1984) and Yang (1986) recently published monographs on the same topic. Mitchell (1831) reported that various kinds of gases permeate through thin rubber film with different velocities. Graham (1866) showed that oxygen can be separated from air through this rubber membrane. Since then, the enrichment of air with oxygen using membranes was studied, and the commercial application was realized. However, power consumption in separating oxygen from air with the membrane process is not small (1.1kwh/m3 of 02). The initial investment for the plant and operating cost is quite large because of the expensive membrane materials and the cost of its fabrication. In spite of these disadvantages, membrane technology finds its way in medical applications where the amont of oxygen-enriched air needed is not large. For large-scale industrial processes, cryogenic pressure swing technology using zeolite as an adsorbent was commercializedto produce pure oxygen or nitrogen. This process requires a large consumption of power (1.2-1.4 kwh/m3 of Oz), because the separation factor of oxygen with respect to nitrogen is not large in zeolite at ambient temperatures, requiring the separation process to be operated at low temperatures. Hence, we became interested in developing a new method for the separation of oxygen or nitrogen from air using the selective adsorption of oxygen on highly dispersed
* To whom all the correspondence
should be addressed.
0888-5885/88/2627-2135$01.50/0
metal crystallites. This idea can be applied to separate Hzfrom gas mixtures of CH4/H2or H,/He. Bond (1962) reported that most metals except gold can adsorb oxygen much more strongly than nitrogen. Kim et al. (1988) tested various metals (Pt, Fe, Cu, Zn, and Co) to find the most economic metal for separation of oxygen from air. In order to be an economic process for the separation of oxygen from air, the selective adsorption of oxygen as well as easy desorption of adsorbed oxygen on supported metal crystallites should occur at ambient temperatures. The differences between adsorption and desorption temperatures should be small in order to have many adsorption-desorption cycles per unit time. In this research, platinum crystallites highly dispersed on y-alumina, which was shown to be the best candidate metal by Kim et al. (19881, were evaluated for the separation of oxygen from air.
Experimental Section 1. Preparation of Platinum Crystallites Supported on y-Alumina. Platinum black (Strem Chem. Inc.) was dissolved in aqua regia to prepare HzPtClg6Hz0. A known quantity of chloroplatinic acid diluted with water was impregnated into the y-alumina in the Rota-vapor to prepare 1wt % platinum supported on y-alumina. After the impregnation was completed by evaporating excess water, samples were further dried at 120 "C in a drying oven. Dried samples were further calcined at 500 "C for 4 h under 60 mL/min flow of air. Other metal crystallites supported on SiOz were prepared similarly. 0 1988 American Chemical Society
