Acknowledgment
T h e author acknowledges the contribution of J. A. Wildsmith in writing the computer programs used in the preparation of this paper and also the value of discussion held with P. P. King. Acknowledgment is also given to Imperial Chemical Industries, Ltd., for permission to publish this paper. literature Cited
Everett, D. H., Landsman, D. A., Trans. Faradav SOC.50, 1221 11954).
Fauyiol;,. C. J., Chim. Phys. 22, 1 (1925). Harned, H. S., Davis, R., J.Am. Chem. SOC. 65, 2030 (1943).
Harned, H. S., Robinson, R. A., Trans. Faraday Sod. 36, 977 (1940). Harned, H. S., Scholes, S. R., J . Am. Chem. SOC. 63, 1706 (1941). Hatch, T. F., Pigford, R. L., Ind. Eng. Chem. Fundamentals 1, 209 11962). Higson, G. I., Chem. Znd. (London) 1951, p. 750. MacMullins, R. B., Webber, M.. Trans. Am. Inst. Chem. Eners. . 31 409 (1935). Neumann, B. Z., 2. Angew. Chem. 34, 445 (August 1921). Parsons, R., "Handbook of Electrochemical Constants," Butterworths, London, 1959. Sherwood, T. K., Pigford, R. L., "Absorption and Extraction," 2nd ed., p. 11, McGraw-Hill, New York, 1952. RECEIVED for review March 10, 1967 ACCEPTED November 20, 1967 Division of Fertilizer and Soil Chemistry, 152nd Meeting, ACS, New York, N. Y . , September 1967.
KINETICS OF HYDROGEN REDUCTION OF MANGANESE DIOXIDE H. E. B A R N E R ' A N D C. L.
M A N T E L L
Newark College of Engineering, Newark, .?I. J .
Single porous pellets and small beds of particles of synthetic pyrolusite were reduced in hydrogen at various partial pressures in the temperature range of 200" to 500" C. The reaction kinetics were followed by recovering the water product, and the intermediate reduction products were identified b y x-ray diffraction analysis. Reduction proceeded topochemically through the sequence MnOz -+ Mn2Oa -3 Mn304+ MnO. Below 250" C. reduction subsided with the formation of MnaO4, and the process was controlled by a gas-solid chemical reaction step. Above 250" C.severe diffusional resistances were encountered, and further reduction to MnO became appreciable. Above 325" C. the over-all reduction process was again controlled by a gas-solid chemical reaction step. Variation of the reduction rate with temperature and with HZ and H 2 0 partial pressures i s consistent with the concept that the gas-solid reactions in the low (below 250" C.) and high (above 325" C.) temperature regimes involve adsorption of Hz, surface reaction, and desorption of H20, the surface rearrangement being rate-controlling.
only commercial source of pure manganese is based on electrolysis of manganese sulfate solutions. One important step is the reduction of oxidized ores to make them soluble in the recycled anolyte, usually accomplished by roasting the ore in a reducing atmosphere at approximately 1650" to 1700" F. T h e manganese content can be readily reduced in this manner to the divalent form. Very little, however, is known about the kinetics of the reduction reactions or the nature of the intermediate reduction products. T h e purpose of the present work was to study the reduction of manganese dioxide, with respect to both the kinetics and the identification and stability of the reduction products. Very few data on the reduction of MnO2 have been published. T h e only serious kinetic study heretofore made is the Russian work of Chufarov and coworkers (Chufarov et al., 1952; Tatievskaya et al., 1948, 1949). T h e first (Tatievskaya et a / . , 1948) publication of this series discusses the reduction of MnOz to M n O with hydrogen in the range of 300" to 500' C. T h e reduction rate was found to be approximately first order with respect to hydrogen partial pressure, and the apparent HE
T the
1
Present address, The M. W. Kellogg Co., New Market, N. J.
activation energy was determined to be 24 kcal. per mole. T h e two later publications dealt with carbon monoxide as the reducing agent. I n a much more restricted study (Cismaru and Vass, 1962) samples of MnO2 were reduced at a hydrogen pressure of 1 atm. in the range of 350" to 490" C. Contrary to the Russian workers, Cismaru and Vass claim that the reduction terminates with the formation of MnaOa in this temperature range. Mn-H-0 Reduction Equilibria
A thermodynamic analysis of the oxide phases which are in equilibrium with gas mixtures of H f and HzO (Barner, 1967) indicates that M n O is the equilibrium reduction product throughout the realm of practical temperatures. More specifically, reduction of MnOn to M n O is thermodynamically feasible (room temperature to 1000" C.) as long as the ratio P H Z ~ / PisHless 1 than approximately 1 X 105. Conversion of the monoxide to manganese metal cannot proceed to any appreciable extent except at very high temperatures. Even at 1600" K . (above the melting point of Mn) in the ratio PBzo/PHzmust be maintained below 3 X VOL. 7
NO. 2 A P R I L 1 9 6 8
285
order to make the reduction to metal thermodynamically possible. Hence, equilibrium considerations predict that the reduction of MnOz with H ? terminates with the monoxide.
Complete details on the experimental apparatus are available (Barner, 1967).
Apparatus
Synthetic pyrolusite was derived from 50% aqueous manganese nitrate solutions. The bulk of the water was first removed by evaporating the solution on a water bath; the concentrated solution was then pyrolyzed in a forced air circulation oven a t 200' C . for 24 hours. The product was removed from the oven, crushed, and screened into various particle-size fractions. The fractions were heated for a second 24-hour period a t 200' C. Pyrolysis of the nitrate has been reported (Gattow and Glemser, 1961; McMurdie and Golovato, 1948) to yield the pyrolusite form of MnOz. This was verified by comparing the x-ray diffraction pattern of the product with that reported (Gattow and Glemser, 1961 ; McMurdie and Golovato, 1948; Moore et al., 1950; Sorem and Cameron, 1960) for pyrolusite.
T h e majority of reactions were conducted in a vertical-tube differential-flow reactor. Relatively small samples of manganese dioxide (usually about 1.5 grams) were used and reducing gas flow rates were high (3 to 13 cu. feet per hour). I n the quantitative dry hydrogen kinetic runs the water vapor content in the gases leaving the reactor was always less than 0.2%; generally, only trace amounts of water vapor were present. Product water was recovered over known intervals of time, and these data formed the basis for following the kinetics of the reduction. The essential features of the equipment are represented schematically in Figure 1. T h e apparatus was made primarily of borosilicate glass with Tygon tubing used for the flexible connections and l/r-inch copper tubing for the long lines. Cylinder hydrogen, metered through a purification column, was directed through either the dryer or water saturator, depending on whether a dry or moist stream was desired. T h e reducing gas then flowed through a heated reaction tube in which the oxide specimen was reduced. For the dry hydrogen runs, product gas was directed through absorption tubes (packed with Anhydrone), where the water vapor formed by reaction was recovered. For the moist feed runs, conversion could not be based on water recovery because the water added to the feed (in the saturator) exceeded the small amount of water formed in the reaction by several orders of magnitude. I n these cases it was necessary to base the conversion on oxide weight loss alone. T h e vertical-tube reactor consisted of a borosilicate glass tube around which a three-zone electrical resistance furnace was constructed. Oxide specimens were supported in baskets fabricated from 250-mesh nickel wire cloth. I n most cases the apparent sample temperature could be controlled within =k 1' C. of the desired temperature. A few high temperature reductions were carried out in a Vycor tube inserted horizontally into a conventional hingedtop combustion furnace. Oxide specimens were supported in Alundum boats in this reactor. Apparent sample temperatures could be controlled within +5' to 10' C. of the desired temperature.
{
DRYER
PURIFIER
I{-)-'
~ R ~ T A K T E R
MANOMETER
80
80 ARGON CYLINDER A8SORPTlON
TUBES
REACTION
P L A S H BACK CONTROL
Tunc
I$
VALVES
N N G I:;R;:;;;
FLAME GAS ICE BATH
Figure 1 .
T h e available oxygen of the oxide was determined to be 18.4110.07%, as compared to a calculated value of 18.40% for stoichiometric MnOz. T h e manganese content analyzed at 63.18+0.08%, compared to a calculated value of 63.19% for MnOl. The composition of the synthetic pyrolusite was therefore very close to stoichiometric manganese dioxide. Both single compressed pellets and small particles of pyrolusite were used. Two batches of the oxide, prepared in seemingly identical fashion, exhibited slightly different pore structures. The initial BET surface areas, as determined on a modified Nelson and Eggertsen (1958) continuous-flow apparatus, were all relatively low, as shown in Table I. T h e surface area did not vary significantly with particle size (batch A). Obviously, the granules were porous and subdivision of the larger particles did not generate appreciable new surface. Cumulative micropore volume distributions calculated according to the method of Cranston and Inkley (1957) are given in Table I for two particle-size fractions. T h e porosities associated with micropores are relatively low. Furthermore, the surface area and pore volume data on two particlesize fractions (batch A) indicate that there is virtually no difference in the micropore characteristics between the two particle sizes. T h e micropore porosity of batch B is, however, significantly less than that of batch A ; likewise the BET surface area of batch A is somewhat greater.
-ROTAMETER
MANOMETER
Raw Materials
Flowsheet of reduction equipment
1
Nominal 1.5-gram pellets were prepared from a minus-200mesh fraction of batch A and a minus-170-mesh fraction of batch B. The pellets were disk-shaped with diameters of 13 mm. and thicknesses of 3 mm. Pellets with sufficient mechanical strength for handling purposes could be obtained by compressing the fine powder in a stainless steel die a t 9000 p.s.i. A single drop of water was added to the charged powder in the die as a lubricant; after compression the pellets were thoroughly dried a t 200' C. Pellets formed from batch A and B had virtually identical surface areas and micropore volume distributions (Table I). T h e total porosity of the pellets was estimated as 22%. This figure is based on a total pore volume determined by boiling
Physical Characteristics of MnOz Used in Reduction Experiments Pore Size Radius, A . Surface Area, 50 100 200 Particle Size, Mm: Sq. M . / G . Cumulative Micropore Volume, Cc./G. Table 1.
Batch A
B
A
B
286
0 ,15-0 ,21
0.07-0.10
0.07 0 .15-0 .21
Compressed pellet Compressed pellet
0.40 0.52 0.48 0.26 1.4 1.3
I & E C PROCESS D E S I G N A N D D E V E L O P M E N T
0.0048
500
0.0040
0.0063 0,0065
0,0080 0.0088
0.0095 0.0113
0 .0003 0.0085 0.0085
0 ,0008 0.0155 0.0136
0.0012 0.0198 0.0170
0.0221
...
0.0250
0
8
4
TIME
Figure 2.
20
16
I2
isEconos x
32
28
24
10-31
Reduction-time data a t various temperatures in pure Hz
, 1.4
1.5
Figure 3.
16
1.7
1.8 X 103)
1.9
PO
2.1
Arrhenius plot for granules of MnOz
a weighed pellet in water to expel gases, superficially drying and weighing the pellet, and dividing the weight increase by the density of water.
All gases were purchased from T h e Matheson Co. The prepurified grade of hydrogen was used, and argon was employed as the purge gas. Mixtures of hydrogen and helium were purchased for the experiments in which the effect of hydrogen partial pressure was investigated. l o w Temperature Regime
A detailed description of the experimental procedure and complete tables of the experimental d a t a are available (Barner, 1967). T h e relative importance of the various resistances which enter the reduction process (boundary layer mass transfer, pore diffusion, and chemical reaction) was investigated in pure H:! (BOO mm. of H g pressure) at 250' C. A series of runs was conducted using granules of MnOz (batches A and B) in which flow rate, particle size, and oxide bed depth were varied. With the exception that the initial reduction rates of batch A samples were slightly greater, oxide specimens
from both batches reacted similarly. Neither a fourfold increase in flow rate nor a threefold increase in particle size significantly affected the reduction rate, thus demonstrating that boundary layer mass transfer and pore diffusion resistances were relatively unimportant a t these conditions. Likewise, oxide bed depths less than 5 mm. had no effect on the observed rate, indicating that neither hydrogen depletion nor product water vapor accumklation played a n important role under these conditions. From the results of these data operating conditions were selected a t temperatures below 250' C., which ensured measurement of the rate of chemical conversion alone. Typical conversion curves for granules of MnO:! are plotted in Figure 2. At temperatures of 250' C. and below the reduction process essentially subsides a t a conversion of approximately 65 to 70% (based on conversion to M n O ) . Reduction rates obtained by differentiating the conversiontime data correlate linearly on an Arrhenius plot in the temperature range of 200' to 250' C. (Figure 3). T h e fact that the 10 and 35y0 reduction lines have practically identical slopes indicates that diffusion resistance through a reduction VOL. 7
NO. 2
APRIL
1968
287
~~~~~
Table II.
~
~
Phases Formed by Reduction with Hydrogen According to X-Ray Investigations
L7, ReducI"
a
Run No. 39 20 22 16 42 37
Reducttn Temp., C: 250 250 250 250 225 315
tion to MnO 11 15 32 66 68 45
53
350
21
46
350
78
106
450
60
Barely detectable.
Comments
Reducing Reducing Reducing Reducing Reducing Reducing
MnOz MnOl MnOz Mn304 Mn304 MnOz
Reducing agent, 24.9% Hz, balance He Reducing agent, 24.970 Hz, balance He Reducing agent, 9 .73y0 Hz, balance He
MnOz MnO
Phases Observed Semimajor
Minor
MnsO4 MnO, MntOd, MniOs MnOz
MnO, MnOz
Very weak MnO lines.
-
I&EC
Major
agent Hz agent Hz agent Hz agent H z agent Hz agent Hz
product layer is not significant below approximately 250' C. By equating the slope of the linear portion of the curves to EO/2.3R, an apparent activation energy of 22 kcal. per mole is obtained. A value of this magnitude is characteristic of chemical reaction control. T h e sharp curvature in the Arrhenius plot beginning a t about 250' C. is indicative of a drastic change in kinetics; further discussion of the data in this region is postponed until discussion of the high temperature data. X-ray powder diffraction patterns were obtained on a number of partially reduced specimens (Table 11). I n all of the runs conducted at 250' and 225' C., an oxygen material balance between water product recovered and oxide weight loss agreed within 2 or 3%, indicating that reoxidation of the reduction products upon removal from the reactor was negligible. T h e x-ray pattern of the product of run 16, which was reduced to 66y0 completion at 250' C., showed predominantly the pattern of M n a 0 4 (hausmannite), although the strongest lines of the MnZO3 (partridgeite) pattern were also observed. An almost identical pattern was obtained for the product of run 42 which was reduced to 68% completion at 225' C. T h e practically horizontal portions of the reduction curves (at and below 250' C.) in Figure 2 correspond then to products which are predominantly M n 3 0 4 . If the reduction were to proceed stoichiometrically from M n O l to M1-1304, the extent of reduction (based on complete reduction to MnO) would be 66.67%. Inasmuch as the low temperature conversion curves all tend to level out at conversions in the range of 65 to 70%, both the x-ray and the kinetic data are consistent with the conclusion that the low temperature reduction product is chiefly MnrO4. This should not be interpreted to mean that additional reduction to M n O will not occur; further reaction proceeds only a t a relatively much sloiver rate, however. An attempt was made to determine the composition of a specimen reduced to 68% conversion at 250' C. If it is assumed that the monoxide is not present (it was not detected in the x-ray pattern), the relative amounts of MnOz, MnnOa, and M n 3 0 4 can be determined thermogravimetrically (Grasselly, 1955, 1956; Grasselly and Klivenyi, 1956). I n this manner the reduction product was found to contain 91% MnaOd, 7% MnOz, and 2% MnzO3. The M n 0 2 percentage in this analysis (and hence the hfnaO3 percentage which is obtained by difference) is subject to large error, in that a trace of water in the original sample would have a very large effect on the calculated M n 0 2 content. With the exception that MnOz was not detected in the x-ray diagram a t this level of reduction, the 288
~~
PROCESS DESIGN A N D DEVELOPMENT
gravimetric analysis is consistent with the x-ray and kinetic data. Consequently, it is concluded that the low temperature reduction experiments (at the 65 to 70% conversion level) produced Mn304 products of at least 90% purity. The x-ray patterns obtained on the products of runs 39, 20, and 22, all reduced to various extents at 250' C., show that MnsOd is formed early in the reduction and that its concentration increases as the reduction proceeds. In all cases, however, minor amounts of Mnz03 were found. This suggests that the reduction proceeds topochemically through the sequence Mn02+Mn~03+Mn304. the Mnz03 forming an intermediate phase between unreacted MnOz and reduced MnpOl. Single 1.5-gram porous pellets of M n O ? were also reduced. Preliminary experiments in pure H2 showed that neither a fourfold increase in flow rate nor the sectioning of a pellet into four more or less equal pie-shaped aggregates had any effect on the reduction rate a t 226' C. Thus, at these conditions the reduction of a single pellet proceeded without any appreciable diffusion resistance. I t is interesting to compare the reduction of a single pellet to that of a bed of porous granules (Figure 4). T h e initial rate is much greater for the pellet than for the bed of particles. Furthermore, the autocatalytic effect is considerably less pronounced for the pellet. These variations are probably caused by structural differences in the two oxide specimens. A pellet similar to the one used in run 44 had an initial surface area (per gram) approximately five times that of the particles used in run 42. The pore volume distribution data (Table I) indicated that the micropore volume of a pellet was also much greater than for the particles (batch B was used in runs 42 and 44). These data suggest the existence of some sealed-off pores in the granules. According to this viewpoint, the reduction rate must increase during the initial stages as more and more of the originally sealed-off pores become exposed to hydrogen by virtue of the formation of the more porous loiver oxides. Eventually all pore volume becomes exposed and the surface area, and hence the reduction rate, reaches a maximum. Further reduction can then only decrease the amount of surface available for reaction and, therefore, from this point on, the rate subsides. The fact that the maximum reduction rate always occurred at the same level of conversion, regardless of temperature, also suggests that the autocatalytic behavior resulted from the existence of originally sealed-off pores. T h e above argument does not preclude the existence of some degree of intrinsic autocatalytic behavior. Indeed, when one solid reacts in a gas-solid reaction to form a second solid phase,
80
--
100% HYDROGEN PRESSURE = 800 m m Hg
? 60
w-
I
I
/A
rt
I
0
z 0
%
40
w b-
0 Y
2
0 44
20
226
-I70 MESH PARTICLES COMPRESSED INTO DISK-SHAPED PELLET
-
0
0
8
4
Figure
4.
12
16 T I M E . SECONDS
15 PICTURED AS AN AGGLOMERATION OF NUMEROUS DENSE c u EACH OF EDGE I p .
m.
Figure 5.
Reduction model
Left. Macroparticle of MnO? Right. Partially reduced microparticle
24
28
32
10-3
Reduction of particles vs. single pellet
the reaction can occur only a t the boundary between the two solid phases (Langmuir, 1916), and if the growth of the interface begins at relatively few points on the surface, the reduction rate will increase until the interface covers the entire surface. Nevertheless, it appears that the primary cause for the autocatalysis observed here lies in the structural characteristics of the oxide and not in the intrinsic kinetic behavior. T h e reduction d a t a for the pellets were correlated in terms of the rate model depicted in Figure 5. A4porous pellet of MnOz is pictured as an agglomeration (Figure 5, left) of numerous dense cubes, each of identical edge x,. Under conditions where both boundary-layer transport and porediffusion resistances are negligible, it suffices to consider the reduction of a single (hypothetical) dense cube (Figure 5, right) exposed to the known bulk gas phase hydrogen concentration. I n view of the x-ray results, it is reasonable to consider the reduction process to occur topochemically, with MnOz in the core, followed by a dense layer of MneOa, and finally a porous layer of Mn3Oa. Since the volumes of oxide associated with 1 gram of manganese are 0.315, 0.319, and 0.287 cc. for M n O l , Mnz03, and MnaOl, respectively, one might expect Mnz03 to form a dense reduction-product layer on MnOz, and a porous layer of MnsOa to form as the reduction product of Mnz03. According to this hypothesis the gas-solid reduction occurs only a t the Mn203/Mn30.j boundary, while the reduction of M n 0 2 to MnnO3 would presumably proceed by a solidstate process. Furthermore, since Mn203 is present in relatively small amounts, the thickness of this layer may be neglected. T h e original dense cube edge is designated x,, and the cube edge of the unreacted core of MnOz is taken as x .
MICROPARTICLES
20
x
If the fractional conversion based on Mn304 as the final product is designated Z', and if the chemical reaction step is assumed to control, it can be shown (Levenspiel, 1962) that
xu
where k', an apparent rate constant, is a function of temperature and hydrogen pressure. Data for four temperatures are plotted according to Equation 1 in Figure 6 . With the exception of the low conversion points, excellent linearity is observed. T h e low conversion data are characterized by an apparent autocatalytic effect \\hich, of course, has not been considered in formulating the model; for the same reason the straight lines d o not pass through the origin. Apparent rate constants, obtained by evaluating the slopes of the lines in Figure 6, correlate linearly in an Arrhenius plot (Figure 7). T h e apparent activation energy, 26.8 kcal. per mole, agrees well Lvith the 22 kcal. per mole value obtained for the reduction of granules of MnOn (Figure 3). The effect of hydrogen partial pressure upon the reduction of pellets was investigated at 226' C. Helium was used as the diluent, and the total pressure was maintained at 800 mm. of Hg. T h e results, shown in Figure 8 in terms of the variation of the apparent rate constant, are characteristic of a chemically controlled solid-gas reaction where the surface is only partially saturated \vith Hz and adsorption of product €I20 is negligible. T h e retarding effect of water vapor was also investigated at 226' C. Nominal 1 and 10 mole y6 levels of water vapor concentrations were used, and the total pressure was maintained a t 800 mm. of Hg. T h e extent of reduction was periodically determined by purging the system with argon, removing the sample from the reactor, and weighing the partially reduced oxide. This procedure was valid because reduction products at 226' C. Lvere stable upon exposure to the atmosphere. T h e results are shown in Figure 9. I n the first run of this series, a pellet placed in a hydrogen stream containing 9.9% water vapor reacted to only 0.8% reduction in 4000 seconds. Thus, the water vapor severely hindered the start of the reduction process. T h e sample was replaced in the furnace and made to react in dry hydrogen for a second 4000-second period. During this interval the oxide reacted in normal fashion to 18% reduction. After this: the sample was once again made to react in a 9.9% water vapor stream for several intervals of 4000 seconds. T h e reduction proceeded, but a t a highly retarded rate. VOL. 7
NO. 2
APRIL
1968
289
06 60
0.5
91'
P P
0.4
50
II
3-
z 0 V
0.3
-
n
w
I
-
40
z
-I
0
0.2 30
k!
20
0.I
10 n
A
-0
8
4
Figure 6.
16 20 TIME, SECONDS X IO"
12
24
32 "
28
Low temperature model plots for pellets
REDUCTION OF POROUS PELLETS OF M N O ~
e
400 600 eo0 HYDROGEN WRTIAL PRESSURE, m m Wg
200
- 0 I90
I95
2 00 (*K-'
2 05
e IO
X IO')
Figure 7. Low temperature Arrhenius plot of apparent rate constant
Data for a run with a hydrogen stream containing l.lyO water vapor are also plotted in Figure 9. The retardation of the initial rate was not nearly so severe as with the 9.9% water run. I n this case (1.1% H20) the retardation was equivalent to a 45% decrease in the apparent rate constant. T h e strong inhibiting effect of water vapor is apparently caused by adsorption of H20 on active reaction sites at the solid-gas reaction interface. I n summary, the data on synthetic MnOz indicate that below 250' C. the reduction by hydrogen is limited by the chemical reaction step. T h e reduction practically subsides with the formation of MnaOd in this region. T h e rate increases with hydrogen concentration a t least up to 800 mm. of H g pressure, and the apparent activation energy is approximately 22 to 27 kcal. per mole. Small concentrations of water vapor severely inhibit the reduction process. T h e kinetic and x-ray data are consistent with the hypothesis that the primary gas-solid reaction is controlled a t the MnaOsMnsOl interface; the mechanism probably consists of adsorption of hydrogen, surface reaction, and water vapor desorption, the slowest step being the surface rearrangement. 290
IC
2 I5
I&EC P R O C E S S D E S I G N A N D DEVELOPMENT
Figure 8. 2 2 6 ' C.
Effect of hydrogen partial pressure at
I
I
00
8
'-
Figure
I
1
I
I
16 24 32 TIME, SECONDS X IO-'
40
I
9.
Effect of water vapor a t 2 2 6 " C.
High Temperature Regime
T h e low temperature reduction rate a t the 35% reduction level correlates linearly on an Arrhenius plot only up to about 250' C . (Figure 3). Above this temperature extreme curvature appears, and above 275' C. the rate actually decreases. This behavior is characteristic of a drastic change in the kinetics of the reduction process. Contrary to the rates a t higher conversion, the initial reduction rate continues to increase with increasing temperature. Consequently, the Arrhenius plot for the rate a t the 10% reduction level (Figure 3) does not deviate greatly from linearity in the 200' to 300" C. temperature range. Above 315' C . the initial rate for the exothermic reduction was so high that temperature control could be achieved only by diluting the hydrogen stream \vith argon during the first few minutes of reduction. After this initial period no further temperature control difficulties \cere experienced. Because of the relatively poor heat transfer characteristics of large single pellets, the experimental study in the high temperature region was limited to the reduction of small granules of pyrolusite. X-ray powder patterns were obtained on a number of partially reduced oxides in the high temperature region (Table 11). I n runs 37 and 53, an oxygen material balance betlreen oxide weight loss and water product recovered agreed within 2 or 370, indicating that reoxidation of the reduction products upon atmospheric exposure was negligible. O n the other hand, in runs 46 and 106, the water recovered exceeded the water formed based on weight loss by 5yo (despite the fact that in run 106 the specimen was cooled to room temperature in argon before exposure to the atmosphere). I t is very likely that some of the M n O in the products of runs 46 and 106, kno\vn to have formed, was reoxidized as the samples \vere \vithdrawn from the furnace. M n O was not detected in any of the products reduced a t 250' or 225' C. O n the contrary, an appreciable amount of M n O (manganesite) was detected in a specimen reduced at 315" C. (run 37), even though the conversion was only 45%. T h e MnnOs (partridgeite) and Mn3Ol (hausmannite) patterns were well established in a sample reduced to 217, conversion at 350' C., but the M n O lines in the same sample were noticeably weaker. Inasmuch as the initial high reduction rate subsided a t approximately 17% conversion at this temperature, the early stage of the reaction must be associated with the formation of -hfnfOs and MnsOa. At a conversion of 78y0 (350' C.), minor quantities of MnnOa and MnsOl were again found. I t is somewhat surprising still to find an appreciable amount of MnOn at this point. Finally, a sample reduced to 6OY0 conversion at 450' C. contained both M n O and MnOz as major constituents, with once again minor quantities of MnlOa and h h s o l . T h e x-ray data indicate that the initial very high reduction rate results from a very rapid reduction of some MnOn KO MnlOs and MnsOl. Contrary to the low temperature results, however, the formation of MnsOc apparently subsides after a n appreciable amount has formed. This would be the case if the MnsOl were to become protective-that is, sufficiently dense to restrict severely the counterdiffusion of hydrogen and product water vapor through the layer of reduced oxide formed, T h e reduction process would then be sharply curtailed, or, in the limit of a perfectly dense layer, stopped altogether. I n any event, in this region the temperature is sufficiently high for the MnsOl itself to undergo reduction to the monoxide. At the higher temperatures Mn2Oa and MnsOl were found only in minor amounts, yet a t all levels of conversion. This sug-
-
gests that the reduction proceeds topochemically through the MnaOa + M n O , with the two interseries MnOz -+ Mn203 mediate oxides forming relatively small layers between the unreacted dioxide and the reduced monoxide. A series of reduction runs was carried out in pure hydrogen (800 mm. of H g pressure) at 375' C. in order to assess the relative importance of the various resistances in the reduction process. These data indicated that a twofold increase in either H:! flow rate or oxide particle size had no effect on the observed reduction rate, thus demonstrating that boundary layer mass transfer and pore diffusion resistances were relatively unimportant at these conditions. However, when particle bed depth exceeded 1 mm., sufficient product water vapor could accumulate in the bed to decrease the measured reduction rate significantly. Based on this analysis appropriate operating conditions were selected for runs conducted at other temperatures so as to minimize diffusional and water vapor accumulation resistances. Kinetic runs were carried out in pure H Z and in H2-He mixtures of nominal 10, 25, 50, and 75 mole % Hz concentration. Typical conversion curves for the pure hydrogen runs are plotted in Figure 2. Rates of reduction in pure hydrogen (obtained by differentiating the conversion-time data) are plotted in Figure 3 at constant levels of reduction. Inasmuch as the apparent activation energy in the range of 325' to 400' c. is practically the same at the 35 and 85% conversion levels, the M n O reaction product did not appreciably resist the counterdiffusion of H Z and H 2 0 . T h e magnitude of the apparent activation energy-namely, 28 to 30 kcal. per mole-is characteristic of chemical reaction control. T h e kinetic and x-ray data suggest that the transition region of 250' to 325' C. be envisioned as one in which the M1-1304 phase forms in a gradually more dense form as the temperature increases, thereby restricting passage of Hl and H20 to and from the Mnl03/Mn304 reaction surface. At the same time the rate at which MnsOa reduces to M n O becomes progressively more important with increasing temperature. Finally, above 325' C . the M n 3 0 1layer becomes essentially nonporous, and the gas-solid reaction may be envisioned to entail primarily the reduction of LhfnsOl to M n O . According to this point of view, reaction bet\reen hydrogen and oxide above 325' C . occurs only a t the MnsOd/MnO boundary, the other reductions presumably proceeding by solid-state processes. Since the volumes of oxide per gram of M n (based on roomtemperature densities) are 0.315, 0.319, 0.287, and 0.238 cc. for MnOl. Mn203, MnsOa, and M n O . respectively, one would at first not expect MnsOa to be able to form as a dense reduction product layer on MnsO3. However, theoretical arguments have been presented (Frank and van Der Merwe, 1949, 1950) which show that a dense, nonporous layer of one phase may form on a second phase if the extent of misfit in specific volume is less than about 14%. T h e high temperature (above 325' C.) reduction data were correlated in terms of a simple rate model similar to that used for porous pellets at low temperatures. A porous granule is first imagined to consist of an agglomeration of dense cubes (Figure 5 ) , each of edge x,. Neglecting boundary-layer and pore-diffusion resistances (as well as temperature gradients), it suffices to consider the reduction of a single hypothetical dense cube. A partially reduced cube is, in turn, pictured as a dense MnOz core surrounded by a porous layer of the monoxide phase; the intermediate layers of MnaOs and ?vlnaOc are neglected. If the fractional conversion based on M n O as a final product is designated Z", and the chemical reaction step is assumed to control, it can be shown that VOL. 7
NO. 2 A P R I L 1 9 6 8
291
where k”, the apparent rate constant, is a function of temperature, pressure, and gas composition. According to Equation 2, plots of 1 (1 - Z n ) l I sagainst time should be linear. Illustrative data are plotted in this manner in Figure 10. As expected, the model does not apply during the initial high rate period of reduction; for this reason, the straight lines d o not intersect a t the origin. Nevertheless, the model fits the experimental data very nicely in the range of 30 to 95% conversion. In some runs where the temperature stabilized more quickly and where data a t lower conversions could be taken, the model was valid a t conversions as low as 15 to 20%. The temperature variation of the apparent rate constant at various hydrogen partial pressures is shown in Figure 11. T h e apparent activation energies range from 26 to 29 kcal. per mole and are consistent with the hypothesis that the reduction is limited by chemical rcaction. Apparent rate constants are plotted in Figure 12 as a function of hydrogen partial pressure. T h e curves were generated from the equation
-
a = 8.81 X lo3 exp ( - 3 0 , 7 0 O / R T ) (mm. Hg-’) (sec.-l) b = 1.015 X exp (-2,300/RT) (mm. Hg-’)
Equation 3 correlated the data with an average deviation of 8.570. T h e empirical rate expression, Equation 3, may be interpreted in terms of a Langmuir-Hinshelwood model in which adsorption of Hz and H20 is postulated to occur without dissociation, surface coverage with adsorbed HzO is negligible, and the surface rearrangement is rate-controlling. T h e effect of water vapor on the rate of reduction was studied at 375’ C. Because of the relatively large quantities of water added to the hydrogen feed stream, conversion could not be followed by recovery of H20 product, and the extent of reduction was determined by oxide weight loss. The high initial reduction rate was found to decrease only slightly, if at all, on the addition of 10% water vapor to the hydrogen stream. On the other hand, the reduction rate at higher conversions was severely retarded by water vapor. Specimens were reduced for identical periods of time in hydrogen streams containing
06
05
04
03
02
01
Figure 10.
High temperature model plots for particles
Effect of Water Vapor on Reduction a l 375’ C. (Partially reduced specimens cooled to room temperature in argon prior to removal from reactor. Total pressure = 800 mm. Hg) Reduction after 6200Second Reaction T i m e % Hz0 in Reducing G a s (by Weight Loss)
Table 111.
I
Y
40
L“
?!
GRANULES OF M N O i
x 20
*
0 1.1
DILUENT IS HELIUM
+
E
9.9 10
%
e
w
6
t
a
s 4 m z a
I I 4 0
I 45
I 50
I 5 5
1ooo
Figure 1 1. 292
(e%-
I 6 0
I 6 5
I70
)
High temperature Arrhenius plot for particles
I & E C PROCESS D E S I G N A N D D E V E L O P M E N T
94.00 77.1 52.2
97.970 by water recooery.
0.0, 1.1, and 9.9% water vapor (Table 111). In each case the reduction product was cooled to room temperature in argon before removal from the furnace in order to minimize reoxidation. I n terms of the apparent rate constant, these data indicate that 1% moisture in the hydrogen decreases the reduction rate approximately 45y0, while 10% moisture decreases the rate 75%. T h e strong inhibiting effect of water vapor is consistent with the premise that the reduction process is controlled by the
'-
I
I,
REDUCTION OF POROUS GRANULES (.08-.21 rnml OF MNOp
., . -..,
DILUENT IS HELIUM TOTAL PRESSURE = 800 m m Hp
I4
Z 8
CURVES ARE GENERATED FROM THE EOUATION
0 W
2
$ 4
a
h 4 0
I
I
I
0
200
400
I
600
1
I
1.000
800
1,200
1,400
HYDROGEN PARTIAL PRESSURE ( m m H q l
Figure 12. Effect of hydrogen partial pressure in the range of 325" to 425" C.
chemical reaction step. T h e retardation is apparently due to adsorption of water on active reaction sites a t the solid-gas reaction interface. I t was desirable to determine whether the kinetic data could he reasonably well extrapolated to higher temperatures which would be more applicable to industrial practice and where reduction times on the order of 1 or 2 minutes \vould be encountered. Since the vertical-tube furnace was constructed of borosilicate glass, the runs above 425' C . were conducted in a horizontal-tube furnace using a Vycor combustion tube. I n order to maintain temperature control during the initial high reduction rate period, the samples were prereduced to approximately 12% reduction at 325' C. T h e reactor was then purged with argon while the temperature was raised to the desired level. Hydrogen flow was again initiated after the furnace had leveled out a t the proper temperature. No attempt was made to determine the relative importance of nonisothermal effects or diffusional resistances in these runs. Several conversiontime curves were simply generated, and the reaction times required to achieve 95% reduction to M n O Lvere estimated. T h e reaction times are plotted on Figure 13. The verticaltube furnace data fall on a straight line, as would be predicted as long as k" follows an Arrhenius relationship. Extrapolation of the lo\ver-temperature data indicates that the reduction time at 500' C . should be somewhat less than 1.5 minutes. T h e horizontal-tube data. on the other hand, shoiv that the reduction time at this temperature is in excess of 3 minutes. T h e discrepancy is probably due to severe diffusional and water retardation resistances in the horizontal-tube reactor; in this arrangement the oxide (supported in an Alundum boat) is clearly restricted from free access to hydrogen. As the temperature is decreased, the data from the two furnaces appear to merge, as would be expected. I t is concluded that, for engineering purposes, extrapolation of the lower temperature data to temperatures as high as 500' C. does not seem unreasonable as long as other resistance limitations are taken into account-that is to say, it is unlikely that the intrinsic kinetics undergo a major transition between 400' and 500" C. (as occurs in the 250' to 325' C . range). T h e stability of the M n O reduction product deserves special comment. In a reduction experiment conducted at 600' C. (with no prereduction a t lower temperature), the sample decomposed almost completely to stoichiometric MnzOa during
0
'0
+
8
W n
.m I /
E $
,/
REDUCTION OF GRANULES OF MNOp
...
//
4
100% HYDROGEN TOTAL PRESSURE APPROX 800 mm Hg
W
P I-
I
I
/
6
2-
-
I
/
/I 8
r12
I
13
15
14
1,000
17
16
(.n-l,
Figure 13. Time to achieve function of temperature
9570
reduction as a
the short period required to charge the reactor and stabilize the temperature. T h e M n O reduction product formed at this temperature was stable in air and retained its characteristic green appearance. Likewise, the reduction product formed at 700' C. was stable, in contrast to the products formed at temperatures of 500' C. and lower, where some reoxidation of the M n O was always experienced. For example, a t 425' C., the completely reduced green M n O product turned a very distinct brown-black color after the reaction tube was opened, and the final conversion after exposure to air (based on \+,eight loss) was only 93%. I t appears, then, that a reduction temperature on the order of at least 600' C . is required to produce an M n O product which is stable in air. Reduction of a Pyrolusite Ore
The reduction kinetics of a Belgian Congo pyrolusite ore was also briefly studied. T h e ore specimens reacted considerably faster than the synthetic pyrolusite samples a t all conditions VOL. 7
NO. 2
APRIL
1968
293
investigated. T h e ore data did, however, exhibit characteristics similar to those of the synthetic pyrolusite. Barner (1967) describes detailed results. Conclusions
T h e reduction of MnOz proceeds topochemically through the sequence MnOz -+ Mnz03 + MnsO4 + M n O . The most distinguishing feature of the kinetics of this reduction is the change in mechanism encountered in the range of 250’ to 325’ C. This characteristic is most clearly shown by the reversals of slope in an Arrhenius plot (Figure 3). We attribute this change to the formation of a progressively more protective (less porous) MnsO4 reduction product as the temperature is raised. Below 250’ C. the reduction process is controlled by the gassolid reaction step. Here, reduction subsides with the formation of M1-1304, and stable products of approximately 90% MnaOa can be obtained. Above 250’ C. severe diffusional resistances are encountered as the Mn304 product becomes progressively more protective. Further reduction to M n O becomes appreciable a t these temperatures. After an initial buildup of layers of Mnz03 and Mn304, the reduction process above 325’ C. is once again controlled by a gas-solid chemical reaction step. The data in both the low temperature (below 250’ C.) and high temperature (above 325’ C.) regimes are consistent with the concept that the gas-solid reactions involve adsorption of Hz, surface reaction, and desorption of HzO, the surface rearrangement being rate-controlling. Acknowledgment
T h e authors express their gratitude and appreciation to Edwin Keel and T h e M . W. Kellogg Co. for conducting the x-ray analyses; A. J. Haley and the Research and Development Division of Engelhard Industries, Inc., for conducting the surface area and pore volume measurements; and the Union Carbide Corp. (Metals Division) for supplying the Belgian Congo pyrolusite ore specimens. Nomenclature
E, = apparent activation energy k = apparent reaction rate constant
294
I & E C PROCESS D E S I G N A N D D E V E L O P M E N T
Pi = partial pressure of component i R T
= =
t
=
x,
=
x
=
Z =
gas constant temperature time original edge length of cube in shrinking-core reaction model cube edge length of unreacted core in shrinking-core reaction model fractional conversion of solid reactant
SUPERSCRIPTS
’ ’I
= low temperature regime = high temperature regime
literature Cited
Barner, H. E., “Kinetics of Hydrogen Reduction of Manganese Dioxide,” D. Sc. dissertation, Newark College of Engineering, Newark, N . J., 1967 (available from University Microfilms, Ann Arbor, Mich.). Chufarov, G. I., Averbukh, B. D., Tatievskaya, E. P., Zh. Fir. Khim. 26,834 (1952). Cismaru, I. D., Vass, M., Rev. Chim., Acad. Rep. Populaire Roumaine 7, No. 1, 101 (1962). Cranston, R. W., Inkley, F. A., Aduan. Catalysir 9, 143 (1957). Frank, F. C., van Der Merwe, J., Proc. Roy. Soc. (London), Ser. A, 198,203 (1949). Frank, F. C., van Der Merwe, J., Proc. Roy. Sot. (London), Ser. A, 200. 125 (19501. Gattow, G., Glemser, O., Z. Anorg. Allgem. Chem. 309, 121 (1961). Grasselly, Gy., Acta L’nir. Szeged., Acta Mineral.-Petrog. 8, 13 (1955). Grasselly, Gy., Acta Lrniv. Szeged., Acta Mineral.-Petrog. 9, 41 (1956). Grasselly, Gy., Klivenyi, E., Acta Uniu. Szeged., Acta Mineral.Petrog. 9, 15 (1956). Langmuir, I., J . Am. Chem. Soc. 38, 2263 (1916). Levenspiel, O., “Chemical Reaction Engineering,” Wiley, New York, 1962. McMurdie, H. F., Golovato, E., J . Res. ‘Vatl. Bur. Std. 41, 589 (1948). Moore, T. E., Ellis, M., Selwood, P. W., J . Am. Chem. Soc. 72, 856 (1950). Nelson, F. M., Eggertsen, F. T., Anal. Chem. 30, 1387 (1958). Sorem. R., Cameron. E.. Econ. Geol. 55. 278 11960). Tatievskaya, E. P., Antonov, V. K., Chufarov, G. I., Dokl. Akad. ‘Vauk SSSR 68,561 (1949). Tatievskaya, E. P., Chufarov, G. I., Antonov, V. K., Izv. Akad. Nauk SSSR, Otd. Tekhn. .Yauk 1948, p. 371. RECEIVED for review April 3, 1967 ACCEPTED August 30, 1967 Condensed from a dissertation submitted by H. E. Barner in partial fulfullment of the requirements for the degree of doctor of engineering science at Newark College of Engineering.
