S
t V
V Y
V? US
* us0
W
slope of slip velocity correlation time, hr. or min. VF Equation 6 superficial velocity of continuous phase, ft./ hr. superficial velocity of continuous phase effluent (= qyUt/A),ft./hr. superficial velocity of dispersed phase, ft./ hr. superficial velocity of dispersed phase influent, ft. / hr. slip velocity, ft./hr. vso+ ts, slip velocity correlation intercept of slip velocity correlation, ft./hr. interstage mixing coefficient,ft./hr.
literature Cited
Bell, R . L., Ph.D. dissertation, University of Washington, Seattle, Wash., 1964. Beyaert, B. O., Lapidus, C., Elgin, J. C., A.I.Ch.E. J. 7,46 (1961). Biery, J. C., Boylan, D. R., Ind. Eng Chem. Fundeamentals 2,44 (1963). Di Liddo, B. A., Ph.D. dissertation, Case Institute of Technology, Cleveland, Ohio, 1960. Foster, H. R., Jr., Ph.D. dissertation, University of Washington, Seattle, Wash., 1967. Jones, S. C., Ph.D. dissertation, University of Michigan, Ann Arbor, Mich., 1963. Miller, R. S., Ralph, J. L., Curl, R. L., Towell, G. D., A . I.Ch.E. J . 9, 196 (1963). Tadamasa, S., Sagihara, K., Tamyama, S., Kagaku KGaku (abridged ed.) 2, 30 (1964). Thornton, J. D. Trans. Inst. Chem. Engrs. (London) 35, 316 (1957).
eUt,
GREEKLETTERS = dispersed phase fractional holdup 0 = time constant, hr.
RECEIVED for review April 22, 1969 ACCEPTED January 28, 1970
SUBSCRIPTS x = continuous phase y = dispersed phase
Work supported by the U S . Atomic Energy Commission, Division of Research, through Contracts AT(45-1)-1053 and AT(45-1)-2057.
MODEL FOR OXIDATIVE THERMAL DECOMPOSITION OF STARCH IN A FLUIDIZED REACTOR S .
1.
J A R I W A L A '
A N D
School of Engineering, University of
E. H O E L S C H E R Pittsburgh, Pittsburgh, Pa. 15213 H .
The oxidative thermal decomposition of starch was studied in a tapered fluidized bed reactor a t pressures from 1 to approximately 0.5 atm and at molar oxygen concentrations from 0 to 21 Yo in nitrogen as the fluidizing gas. The decomposition of starch proceeds very rapidly after the "critical" temperature is reached and is accompanied by the liberation of a pulse of heat. With air as the fluidizing medium, the critical temperature was about 240OC. The concept of starch particles of varying reactivity has been used in analyzing the results. The reaction rate was proportional to the oxygen concentration in the fluidizing gas. Yields of levoglucosan were low.
THEnonoxidative thermal decomposition of starch yields glucosans and acidic products of potential commercial value, but this reaction has not been developed commercially because of difficulties involved in heat transfer and product recovery. These difficulties arise largely from the poor heat-conducting properties of starch, which lead to localized overheating, charring, and decomposition of the products formed within the reactor. Surprisingly, the nonoxidative thermal decomposition of starch to glucosans could not be demonstrated in our fluidized bed reactor, although this reaction has been
' Present address, Heyden Division, Tenneco Chemicals, Inc., Garfield, N. J. 07026 278
Ind. Eng. Chem. Process Des. Develop., Vol. 9, No. 2, 1970
demonstrated repeatedly in batch reactors in the absence of air. Decomposition products of the type encountered with the oxidative decomposition were not isolated when nitrogen was used as the fluidizing medium. I t is probable that the desired reaction proceeds a t a rate slow enough to be virtually unmeasurable on the reactor configuration used in this study. When oxygen was introduced with the nitrogen, an explosive-type reaction was observed and only 1% yields of levoglucosan were produced. The objectives of this investigation were, therefore, changed to study the nature of the oxidative decomposition reactions. Explosions of starch dusts in air have wrecked starch factories, but no specific study of the reaction kinetics has ever been reported.
The fluidized technique of this study may be used to carry out in situ reactions of starch or the decomposition products. The introduction of other gaseous reactants such as ammonia, halogens, and sulfur dioxide along with or as the fluidizing medium might be used to prepare derivatives of starch by direct reaction in situ. This technique might also be used for reactions of cellulose and other natural products. Reaction Products
The thermal decomposition of starch a t elevated temperatures results in a carbonaceous residue and gaseous products, part of which can be readily condensed. Levoglucosan and water constitute the major proportion of the condensables. Pictet and Sarasin (1918) prepared levoglucosan by the vacuum distillation of starch. Dimler et al. (1946) isolated another anhydride of D-glUCOSenamely, D-glucosan (1,4)B(l,B)-from the products of the vacuum pyrolysis of starch. The decomposition products of starch thus include, in addition to the carbonaceous residue, a complex mixture of glucosans, water, carbon dioxide, and other gaseous compounds. The formation of levoglucosan from starch involves the scission of @-1,4linkages of starch and the formation of a 8-1,6 oxygen bridge. Wolff et al. (1965) suggest that a t higher temperature, a cleavage of the a-1,4 linkages of starch results in a free radical or “activated” form of glucose which is considered to be an intermediate in the reaction. This degradation of starch to the intermediate followed by the migration of a hydroxylic hydrogen and ring closure would lead to the stable levoglucosan molecule.
The mechanism of the formation of levoglucosan postulated above is not supported by direct experimental evidence. Until further experimental evidence is obtained confirming this mechanism, it must be regarded as a stimulus to further research. Experimental investigation
Apparatus. A fluidized bed reactor was chosen for the present study because it offers: Good heat transfer to the relatively nonconducting starch particles, permitting a higher heat flux than a reaction flask and improved heat utilization. Turbulence within the fluidized mass to ensure a uniform temperature throughout the reactor and minimize local overheating and charring of starch. Rapid product removal in the fluidizing medium, hopefully minimizing decomposition of glucosans within the reactor. Use of a tapered or a conical fluidization column. I n such a column, the velocity profile is constant throughout the reactor, minimizing the variance of the residence time distributions. Again, in a tapered column, the bed expansion is small even a t velocities greater than that required for fluidization. This and the reduced gas velocity in the free space above the bed result in reduced carry-over of solids.
The complete experimental setup is shown in Figure 1.
The reactor was a tapered stainless steel (Type 316) column, 5 feet in height; the internal diameter varied from 3 inches a t the bottom t o 4 inches a t the top, providing a > L o angle of taper. The method of Sutherland (1961) was used to calculate the taper. A “micromesh” stainless steel wire screen, welded to the bottom flange connection, served as both a support for the bed of starch particles and a distributor for the fluidizing gas, thus ensuring a uniform distribution of the fluidizing gas throughout the cross section of the column, and minimizing channeling. Two l/,-inch outlets were attached to the screen for removal of the unreacted solids from the reactor. The flange connection a t the top of the column had two openings, one to introduce the feed and one to permit exit of gaseous reaction products and the fluidizing medium t o the cyclone separator. The bed temperature was measured by an iron-constantan thermocouple introduced through the feed opening. The fluidizing gas used was nitrogen with varying amounts of oxygen. Pressure on nitrogen and oxygen cylinders was regulated by Moore pressure regulators and the gas flow rates were measured by rotameters. An air compressor was used when the fluidizing medium was air. The gas passed through an oil filter and next through a Moore pressure regulator, which reduced the pressure to the desired value. The flow rate was controlled by a needle valve and measured by an orifice meter. A Hoskins FH-305 heavy-duty electric furnace, operating a t 20 volts, 50 amperes, with a nickel-chromium alloy heating coil was used as the heating unit. The fluidizing gas was heated by passing it through a helical coil, 174 inches in diameter, inserted in the furnace. The coil was ‘/,-inch o.d. stainless steel tubing with a stretched length of 20 feet. A rheostat in the primary circuit permitted control of the furnace temperature and, hence, the fluidizing gas temperature. The furnace temperature was measured by a Chromel-Alumel thermocouple. The temperature of the heated fluidizing gas before it entered the reactor was measured by an iron-constantan thermocouple. The gaseous reaction products and the fluidizing medium from the fluidization column were passed through a cyclone separator to remove solids carried over from the column and then through a water-cooled condenser. The condenn@
FURNACE
U OXYQEN
1 1 9
4
FLUIDILATION COWMN
*
TO EXHAUST HOOD
Figure 1. Experimental setup
Ind. Eng. Chem. Process Des. Develop., Vol. 9,No. 2, 1970
279
sate was collected in a calibrated receiver, so that the rate of condensate collection could be metered. The low boiling products were condensed and collected in subsequent dry ice traps. An Ascarite column was used after the dry ice traps to remove the carbon dioxide formed during the run. A vacuum pump connected to the outlet of the system maintained subatmospheric pressures within the column when reduced pressure runs were made. The entire reactor, cyclone separator, and the interconnecting tubing up to the entrance to the condenser were wrapped with Chrome1 A wire (20-gage, asbestosinsulated) and heavily insulated to ensure that all the prodccts of reaction were maintained in the vapor phase until they entered the condenser. The gas chromatographic technique developed by Sawardeker et al. (1965) was used for the analysis of the pyrolysis products. Safety Considerations. Special precautions were taken to prevent the buildup of lethal static charges within the system. To reduce the danger of static discharges within the reactor, the entire reaction unit, including the flanges supporting the stainless steel wire screen, were electrically grounded. The cyclone separator and the lines leading to it from the reactor were also grounded. Furthermore, only a small quantity of starch was used in each run, to reduce hazards associated with violent oxidation. Experimental Procedure. Cornstarch has very poor flow characteristics and good fluidization is difficult to achieve. I t was necessary to reduce its moisture content to less than 3%. I n addition, fine sand (120-mesh) was mixed in a 1 to 1 weight ratio to improve the flow characteristics of starch and t-o scrub the reactor walls. Under these conditions, a good uniform fluidization of the starch could be achieved in the column. Channeling, starch blowover, and the formation of carbonaceous residues on the column walls were greatly reduced. Amylose used consisted of particles varying from 28- to 200-mesh and could be fluidized without admixing with sand. A measured quantity of the charge was introduced into the reactor. The fluidizing gas was started a t the desired flow rate and pressure. The mass flow rate was maintained just above that required for minimum fluidization to have a particulately fluidized bed as indicated by pressure drop. The furnace and the auxiliary heaters around the column were then switched on to heat the fluidizing gases. The decomposition of starch results in the evolution of gaseous reaction products. The higher boiling components, including glucosans and water, were condensed in the water-cooled condenser and collected in the graduated receiver. The condensate was a dark brown, slightly viscous, liquid. The volume of the condensate collected and the bed temperature were recorded at 1-minute intervals after the start of the reaction. The low boiling products were condensed and collected in dry ice traps. After the reaction proceeded to completion, the heating system was switched off and the reactor allowed to cool. The following quantities were observed a t the end of a run: Volume and weight of sirup collected in the graduated receiver. Weight of liquid condensate in dry ice traps. Weight of charred residue remaining within the reactor. Weight of unreacted starch carried over from the fluidized bed into the cyclone separator. 280
Ind. Eng. Chern. Process Des. Develop., Vol. 9,No. 2, 1970
Weight increase in the Ascarite column. A small sample of the sirup (2 ml) was evaporated to dryness; the levoglucosan content of the residue was then found by analysis on the gas chromatograph. Experimental Results and Discussion
In each run, the behavior of the system was studied with respect to the temperature variation of the bed of starch particles and the rate of collection of the liquid product. Figures 2 and 3 show typical temperature-time ( T us. t ) and volume of product collected-time ( V us. t ) curves. The data contributing to the arguments presented in this paper are shown in Table I. During the initial heating period, the temperature of the fluidized bed of starch particles increased with the temperature of the fluidizing gas. No gaseous products evolved during this stage. This may be regarded as the “prereaction zone,” in which the bed temperature follows the fluidizing gas temperature. Once a certain “critical” temperature was reached, very rapid changes were observed in the system. At this temperature, the degradation of starch started sharply, accompanied by the evolution of gaseous decomposition products. The “reaction zone” is considered the region of time during which the temperature pulse is observed.
ZONE
’ W
360
a
3
‘4
320 280
W
I-
40
I
1
50
60
I
70
TIME (MINUTES)
Figure 2. Relation of bed temperature to time
6
> o
l 0 50
60
70
80
90
TIME ( M I N U T E S )
Figure 3. Effect of time on volume of product collected
Table I . Reactor Data and Calculated Values of n, In kl Composit ion of (‘oiumn Presh u re, PSI A 14.7 14.7 14.7 14.7 7.34 9.8 14.7
Fluidizing
ch.5
i r 02
L N2
21.0 4.2 11.0 0.0 21.0 21 .o 21.0
79.0 95.8 89.0 100.0 79.0 79.0 79.0
m t > Minutes
Beyond the reaction zone, there is no further heat generation within the reactor and the fluidizing gases remove the heat produced in the reaction zone. As a result, the reactor temperature decreases continuously. This is the “cooling zone.” When pure nitrogen was used as the fluidizing medium, the temperature of the bed of starch particles increased; however, no decomposition of starch was obtained. The temperature-time curve for the “noreaction” case is, therefore, obtained when nitrogen is used as the fluidizing medium. On the other hand, when oxygen was present in the fluidizing gas, there was a sharp increase in temperature due t o reaction as described earlier. The area between the two temperature curves (with and without reaction) is a measure of the amount of heat liberated during the reaction. A typical plot of A T ( = TWlth reacr1 to, the rate of product collection is high and decreases continuously to zero. A skewed sigmoid, with a short initial section, would, therefore, represent the volume-time behavior. However, there is
/
4c
0
10
20
30
COLUMN PRESSURE ( I N S . HO )
Figure 7. Relation of number of reaction stages to column pressure
uncertainty in the initial section due to the extremely small time interval over which the rate of product collection increases from zero to the observed initial value. A linear approximation has, therefore, been used on the initial section of the volume-time curve, giving the shape shown in Figure 3. The form of the V us. t curve supports the earlier observation that the reaction occurs as a pulse through the system. An examination of this curve for a typical run indicates that most of the product is collected in a short interval after the product collection has started. At higher values of t , the rate of product collection decreases and finally becomes zero. T h e rate of product collection could be represented in analytical form if the kinetics of the reaction and the reactor dynamics were known. Such an analysis must await further kinetic data. The volume-time data a t this point have only been empirically correlated t o within 1% by an expression of the form:
V = vl[l- e+ (1 + a t ) ] Values of 01, a, VI, and the initial rates of product collection, V;, are shown in Table 11. ~1 and a change with the fluidizing conditions, indicating that they represent system variables influenced in some manner by the column pressure and the oxygen concentration. A typical plot of the rate of product collection, d w d t , as a function of time (Figure l o ) , shows a high initial rate, indicating a rapid initial reaction rate beyond the critical temperature. Figure 11 shows the relation between the initial rate of product collection, V:, and the oxygen concentration in the fluidizing gas. The initial rate, a t a constant column pressure, is proportional to the oxygen concentration in the fluidizing gas.
Table 11. Values of a , a , VI, and V :
X OXYGEN
Figure 8. Relation of oxygen concentration to In
KI
3c
20
i
Run No.
a
a
VI
V:
VI1 IX XI XI11 XIV XV XVI
0.2682 0.2068 0.2444
0.0007 -0.0002 0.2445
0.2642 0.4346 0.1693
0.0962 0.1099 0.1081
61.0 12.0 47.6 0.0 43.4 60.0 86.0
15.0 3.0 7.0 0.0 7.0 12.0 6.67
...
...
-
Y
1 z
IC
COLUMN PRESSURE ( INS. Hc )
Figure 9. Relation of column pressure to In
K1
2\ 0
60
70
80
TIME (MINUTES)
Figure 10. Effect of time on rate of product collection
Ind. Eng. Chem. Process Des. Develop., Vol. 9, No. 2, 1970 283
being accompanied by the liberation of a pulse of heat. The model correlates the available data and may guide the planning for further experimentation. The yields of levoglucosan were much lower than expected. On the basis of total starch converted to all products, they were less than 1%. On the basis of starch converted to recoverable liquid products, the yields were approximately a factor of 3 greater than this. Since the reaction time is very small compared to residence time of products within the reactor, it may be assumed that levoglucosan decomposed within the reactor.
i
Nomenclature
AT = temperature rise above no-reaction case, OC T/ = final bed temperature, C 0
10
20 X
30
OXYGEN
Figure 1 1 . Relation of oxygen concentration to initial rate
of product collection
The weight distribution of the pyrolysis products determined in runs was approximately: 20 to 30% of the dry weight of starch appeared as the condensate. 10 to 20% remained in the flask as charred residue. 4 to 10% was carried over unreacted from the fluidized bed and collected in the cyclone separator. 8 to 15% was collected in the dry ice traps. This material contained some oily droplets. 35 to 50% appeared as gaseous reaction products, mostly carbon dioxide absorbed in the Ascarite column, that could not be condensed and usually was allowed to vent from the system. Summary
The fluidized bed reactor system designed and constructed for the study was used to carry out the oxidative starch decomposition reactions. Decomposition of starch in the fluidized system proceeds very rapidly above a “critical” temperature and with oxygen in the fluidizing gas. No reaction was observed below this critical temperature or in the absence of oxygen. Over the range of variables studied, the reaction was proportional to oxygen concentration in the fluidizing medium. A model is proposed based on the postulate that starch granules with different “reactivity” react at different times, the reaction of each unit with a particular reactivity
284
Ind. Eng. Chern. Process Des. Develop., Vol. 9,No. 2, 1970
t = time at which A T is measured, minutes n = number of “reaction units” or size ranges, each with a particular reactivity, which react T = lag time between reaction of two successive size ranges K , = constant related to heat liberated due to reaction
(K:)“ t, = constant characterizing first reaction unit (size range) and assumed constant for entire mass. Physically, t, is time required by first unit, were it present alone, to reach maximum value of AT/ T/ V = volume of product collected (ml) in time, 1 Vi = total volume of product collected, ml V: = initial rate of product collection, ml/min cy, a = constants for each run Literature Cited
Carbery, J. J., Bretton, R. H., A.Z.Ch.E. J . 4, No. 3, 367 (1958). Dimler, R . J., Davis, H. A., Hilbert, G. E., J . Am. Chem. SOC.68, 1377 (1946). Gal-Or, B., Hoelscher, H. E., A.I.Ch.E. J . 12, No. 3, 499 (1966). Pictet, A., Sarasin J., Helu. Chim. Acta 1, 87 (1918). Sawardeker, J. S., Sloneker, J. H., Dimler, R. J., J . Chrornatog. 20, No. 2,260 (1965). Sutherland, K. S., Trans. Inst. Chem. Engrs. London 39, 188 (1961). Wolff, I. A., Olds, D. W., Hilbert, C. E., Starch and Dextrose Division, Northern Regional Research Laboratory, Peoria, Ill., personal communication, 1965. RECEIVED for review May 14, 1969 ACCEPTED November 17, 1969 Work supported by the Agricultural Research Service, United States Department of Agriculture, administered by the Northern Utilization Research and Development Division, Peoria, Ill.