FIXED-BED, LIQUID-PHASE DRYING WITH MOLECULAR SIEVE ADSORBENT HENRY M. GEHRHARDT'AND B. G. K Y L E Department of Chemical Engineering, Kansas State University, Manhattan, Kan. The drying of benzyl alcohol by fixed beds of SA Molecular Sieve adsorbent was studied experimentally. Breakthrough curves were obtained for various bed heights, feed rates, and feed concentrations. The results were well correlated in terms of the mass transfer zone (MTZ) model and suggest that an internal solid-phase mechanism controls the adsorption rate.
I
of widespread commercial application, liquid-phase adsorption processes have received little attention. While the literature contains many significant contributions in the area of formulation and solution of models for fixed-bed adsorption, it lacks sufficient data, especially for liquid-phase adsorption, to test these models. Thus there is no sound basis for the design of fixed-bed adsorbers taking liquid feeds. The drying of liquids by synthetic zeolites represents a major application of liquid-phase adsorption. The effectiveness of the natural zeolites as drying agents is due to a sieving effect as well as a strong affinity for water. These adsorbents are also attractive for experimental studies, because their crystallinity imparts a well defined internal structure and hence reproducible adsorption behavior. The performance of fixed beds of molecular sieve adsorbent has been reported for the drying of air (Nutter and Burnet, 1963, 1966) the vapor-phase separation of n-hexane from benzene (Kehat and Rosenkranz, 1965), and the purification of hydrogen (Kidnay and Hiza, 1966). Here we report the performance of fixed beds of molecular sieve adsorbent for the drying of an organic liquid. N SPITE
f Figure 1 . Schematic diagram of apparatus A.
E. C. D.
Two-stage pressure regulator Low pressure regulator Pressurized feed tank Variable area flowmeter
E. F. G.
H. V.
Experimental
The system 5A Molecular Sieve-benzyl alcohol-water was chosen for study. Benzyl alcohol was selected as the nonadsorbable component because a favorable refractive indexcomposition relationship allowed small water concentrations to be determined simply and accurately. The equilibrium data have been reported (Jain et al., 1965) and were found, by a least squares procedure, to be well represented by the following Langmuir isotherm equation:
cs ~0.2239
1192 c
1
+ 1192 C
(1)
A schematic diagram of the apparatus used in obtaining breakthrough curves is shown in Figure 1. The test column had an inside diameter of 0.861 inch and was packed with '/16-inch pellets of adsorbent with liquid feed flowing downward. The feed rate was monitored by means of a flowmeter, but was determined by collecting and measuring the column effluent. T o eliminate problems associated with gas bubbles the column was prefilled with dry benzyl alcohol and a variable-height discharge tube used to ensure that the packing was completely covered. This procedure necessitated correction of the effluent volume for the construction of the breakthrough curves. After leaving the column, the effluent passed through a Waters Milan miniature liquid analyzer. The signal from this analyzer was used to
' Present address, Department of Chemical Engineering, University of New Hampshire, Durham, N.H.
Milan analyzer Collecting vessel Air filter Air heater Variac
monitor the effluent composition so that samples could be taken at the proper time for refractive index determination. I t was found unnecessary to provide for heat removal during operation of the column, as only a slight temperature change was observed. All runs were made at room temperature. The column was operated at various feed rates for three column heights and two feed concentrations. The adsorbent was regenerated in situ using external heaters and a flow of heated air. During regeneration the column temperature was monitored by thermocouples and was maintained a t 650' F. for 24 hours. No noticeable variation in adsorptive capacity was noted over the duration of the runs. Mechanism Studies
The common starting point for all mechanistic, fixed-bed models is a differential material balance. The various models are then characterized by specifying an adsorption rate expression and an isotherm equation. Usually the complexity of these rate and equilibrium expressions is limited by considerations of mathematical tractability. I n this study three mechanistic models were tested. A more detailed account is presented elsewhere (Gehrhardt, 1965). The breakthrough curves calculated from these three models are compared with a typical experimental breakthrough curve in Figure 2. Langmuir Kinetic Model. Thomas (1948) pictured the adsorption process as a reversible chemical reaction of the form VOL. 6
NO. 3 J U L Y 1 9 6 7
265
k1
A
+ sorbent *k?A
sor‘bent
The conventional rate expression for this reaction leads to a Langmuir isotherm. Although the breakthrough curves for this model match the experimental breakthrough curves reasonably well at concentrations up to C/C, = 0.5, the model must be judged inadequate because the model curve rises too steeply beyond this concentration. Linear Isotherm Model. This model, advanced by Hougen and Marshall (1947), was one of the first to be proposed and is widely used. Although originally developed with a fluid-film-based rate expression, Vermeulen (1958) and more recently Needham et al. (1966) have shown that the model is also applicable to a particle diffusion mechanism and a combined fluid film and particle diffusion mechanism. This model produces breakthrough curves similar to the Langmuir kinetic model, although the fit is somewhat better. This is to be expected because this model contains an additional parameter (slope of the linear isotherm). I n spite of this increased flexibility, the model does not fit well a t large values of C/C, and is therefore regarded as unsatisfactory. Langmuir Isotherm, Film-Diffusion Model. This model was developed as a part of this study and uses a Langmuir isotherm with a fluid-film diffusion rate expression. An analytical solution for this model was not possible and the method of characteristics as described by Lapidus (1962) was used with a digital computer to obtain a numerical solution.
Mode I
The details of this solution are given by Gehrhardt (1965). This model produces breakthrough curves entirely different from the experimental curves. The model curve breaks more slowly and rises more steeply than the experimental curve. T h e inability of the models to predict the spreading of the breakthrough curve a t high concentrations suggests that dispersion effects, neglected in the model formulatibns, may be operating. The magnitude of the dispersion effect is easily estimated using the method advanced by Levenspiel (1962). Calculations based on this method resulted in a reactor dispersion number, D/uL, less than 0.01, which indicates that dispersion effects should be small for this system. While no definite conclusions regarding the rate-controlling mechanism may be drawn, the dispersion calculations together with the failure of the Langmuir isotherm, film-diffusion model appear to eliminate any purely fluid-phase resistance. The MTZ Model
Perhaps the simplest approach to correlating fixed-bed adsorption performance and the most useful for design purposes is the mass transfer zone model. This model, proposed by Michaels (1952) and rederived by Campbell et al. (1963), makes no assumption regarding the controlling adsorption rate mechanism or the isotherm equation. The model is based on the existence of a n exchange zone of constant length across which the fluid concentration changes from 0.05 C, to 0.95 C,. This zone is established at the top of the column and is assumed to move down the column with a constant velocity. Thus, according to the model, the height of the bed, if it is larger than the mass transfer zone length, Z M ,has no effect on the mass transfer process. The limitations listed by Michaels include dilute feed, favorable isotherm, and bed length large compared to the mass transfer zone length. The working equation for the model is
Kinetic Model
0.1
Langmuir Isotherm, Film- Diffusion Model
1
TIME, MINUTES
Figure 2. Experimental curves for a typical run
Table I.
Run 1A
and
1E
1F
5.00
200.7 200.7 200.7 200.7 200.7 200.7
1G
5.00
200.7 -~
6.0
300.0 300.0 300.0 150.0 150.0
6.0 3.3 7.9 5.3 8.3
1c 1D ~~
2A 2B 2c 3A 3B a
266
breakthrough
Experimental Variables and Calculated Parameters for MTZ Model Feed Flow &fenBed Weight, Velocity, /ration, Wt. % ’ G. Ml./Min. F Z M , Ft.
5.30 5.00 5.00 5.00 4.84
1B
model
5.12 4.74 4.84 2.64 2.43
~
4.8 5.2 3.5 3.7 6.70 4.0
0.39 0.38 0.43 0.40 0.45 0.34 0.37 0.36 0.44 0.43 0.40 0.40
2.4 2.2 1.2 1.2
T h e fraction of the mass transfer zone free of adsorbate, F, is obtained from a graphical integration utilizing the break, through curve. Equation 2 is then used to calculate 2 from the experimentally determined break and exhaustion times. Table I contains the experimental conditions and values of the model parameters determined for this system. The average value of F is 0.40, which is in agreement with the value of 0.408 reported by Nutter and Burnet for the drying air on 4A Molecular Sieves. (These authors actually reported F = 0.592, but based their calculations on the area under the breakthrough curve.) As required by the conditions of the model, 2, was found to be independent of bed height. Figure 3 shows that the variation of Z,v with feed rate is linear and indicates that ZM increases with increasing feed concentration. These findings are also in agreement with the observations of Nutter and Burnet. If parameters ZM and F can be related to the system variables, break times and exhaustion times can then be predicted using Equation 2 and the following material balance equation:
_2 . 2_
2.3 2.5 2.7 1.2 3.2 1.9
2.3
Considerable variation in velocity during run.
l&EC PROCESS DESIGN A N D DEVELOPMENT
(3) Break times calculated from Equation 3 with F = 0.40 and 2, determined from Figure 3 are in very good agreement with experimental values, as can be seen from Figure 4 . Treybal (1955) has extended the M T Z approach to the calculation of breakthrough curves. He utilized Michaels’
400
3
zM,Ft. 2
0
Ce
i
0
Ce
8
5.0 w t , % 2.5 w t . %
-
h
0
300
c
-
c W E
.L
W
a
x W
u
+P ML. PER MIN.
RATE,
i
7 100
FEED
1.
I
o L
0.01
REYNO’LDS
0.15
0.1
NUMBER
tb ( C a l c u l a t e d )
concept of a n infinitely long column operated with counter flows of fluid and adsorbent. By using a n adsorption rate expression based on the fluid-film driving force and assuming that the height of a transfer unit is independent of concentration he obtained the following expression for the calculation of the breakthrough curve:
t
- to - ta
-=
SCdc c ca
s,
- C’ dC
(4)
c T *
When applied to this system Equation 4 produced poorly fitting breakthrough curves similar to those of the Langmuir isotherm, film-diffusion model. When Treybal’s derivation is altered by replacing the fluidphase rate expression with an expression based on a solidphase driving force, the following expression results.
r“ dC (5) JC,
400
0.2
Figure 3. Variation of MTZ height with flow rate and Reynolds number
te
300
200
IO0
0
CS’ - cs
Equation 5 was found to yield breakthrough curves of a shape similar to the experimental curves. I t was not possible, however, to obtain a well fitting curve using Equation 5 and the values of to and t, calculated from Equations 2 and 3. Conclusions
One can conclude from the mechanism studies that the major resistance to mass transfer is not the fluid film. Some insight into the nature of the rate-controlling process is afforded by the M T Z model. The superiority of Equation 5 over Equation 4 for the calculation of breakthrough curves suggests that the rate-controlling resistance resides in the solid phase. Further, as pointed out by Michaels, the linear variation of Z, with feed rate also indicates a n internal mechanism. These observations are consonant with those of Nutter and Burnet, who found that a pore-diffusion mechanistic model well represented their breakthrough curves. T h e adsorption rate data are well correlated by the M T Z approach and the most important design variable, the break time, is predicted well by the model. T h e value of F = 0.40 appears to be well established for the adsorption of water on molecular sieves. T h e results of this study along with those
Figure 4. Experimental break times
and
calculated
of Nutter and Burnet should serve as a basis for preliminary design calculations for fixed-bed drying of gases and liquids with synthetic zeolites. Nomenclature
A C C,
= cross-sectional area
fluid-phase water content, mass fraction solid-phase water content, mass fraction D, diameter of sphere having volume of particle F = fractional ability of adsorbent in M T Z still to adsorb water t = time u = superficial fluid velocity V = volumetric flow rate Z = column length Z,M= length of M T Z pf = fluid density pB = bulk density of adsorbent fi = fluid viscosity = = =
SUBSCRIPTS b = break point where C/C, = 0.05 e = exhaustion point where C/C, = 0.95 o = feed SUPERSCRIPT * = equilibrium literature Cited
Campbell, J. M., Ashford, R. B., Needham, R. B., Reid, L. S., Petrol. Rejiner 42, No. 12, 89 (1963). Gehrhardt, H. M., Ph.D. dissertation, Kansas State University, 1965. Hougen, 0. A., Marshall, W. R., Chem. Eng. Progr. 43, 197 (1947 ). Jaini L.’ K., Gehrhardt, H. M., Kyle, B. G., J . Chem. Eng. Datu 10, 202 (1965). Kehat, E., Rosenkranz, Z., IND. ENG. CHEM.PROCESS DESIGN DEVELOP. 4,217 (1965). Kidnay, A. J., Hiza, M. J., A.Z.Ch.E. J . 12,58 (1966). LaDidus. L.. “Digital ComDutation for Chemical Engineers,” McGr’aw-Hill. flew York. i962. Levenspiel, 0.; “Chemidal Reaction Engineering,” Wiley, New York, 1962. Michaels, A. S., Ind. Eng. Chem. 44, 1922 (1952). Needham, R. B., Campbell, J. M., McLeod, H. P., IND. END.CHEM. PROCESS DESIGN DEVELOP. 5 , 122 (1966). Nutter, J. I., Burnet, G., A.I.Ch.E. J . 9, 202 (1963). Nutter, J. I., Burnet, G., IND. ENG. CHEM.PROCESS DESIGN DEVELOP. 5 , l(1966). Thomas, H. C., Ann. N . Y . Acad. Sci.49, 161 (1948). Treybal, R. E,, “Mass Transfer Operations,” McGraw-Hill, New York, 1955. Vermeulen, T., Advan. Chem. Eng. 2, 147 (1958). ’
RECEIVED for review August 29, 1966 ACCEPTED December 5, 1966 VOL.
6
NO. 3
JULY 1967
267
