T h e 26 " region was scanned semicontinuously while Y - L - 0 3 \cas being reduced with hydrogen a t 450" C. Disappearance of the ? - C o s was followed by the diminution of the intensity of the strongest U 0 3 line, a t 25.8". Hexagonal c 3 0 8 was observed, as before, by means of the (110) line at 26.4'. The transition to orthorhombic L T 3 0 8 - was monitored by the (1 10) and (200) lines a t 26.1 " and 26.5". Typical data are shown in Figure 6. T h e UOZline disappears prior to the observation of the orthorhombic doublet. Thus. the correct reaction sequence for reduction of y-cO3 is: 7-L-03 (orthorhombic) U 3 0 8 (hexagonal) 4 L-308(orthorhombic) UOS+ (cubic) A UOs (cubic). This sequence provides a physical basis for the observed three-step kinetics of y U 0 3 reduction (9). Each time a structural change occurs, as at steps 1>2, and 3, there is also a change in the reaction rate. .An additional source of differentiation between these reactions is that l and 3 are heterogeneous in the solid phase, involving two different solid phases simultaneously, \vhile reactions 2 and 4 are homogeneous transitions \vithin monophasic fields. This reaction sequence provides a plausible explanation for the two-step reaction postulated by DeMarco and Mendel (5) for high-surface-area uo3 on the basis of kinetic data. They observed the reduction to proceed as UOa + LTO?.j6, followed by LT02,56L-02. They did not observe a rate change at the L T 3 0 s composition corresponding to the intermediate formation of hexagonal U 3 0 8 . Since their high sur-
A
face U 0 3 \vas probably hexagonal ~u-L-03,the actual reaction could have been CY-UO~ (hexagonal) L-308 (hexagonal) A C308(orthorhombic) cubic phase(s). Thus, a rate change Ivould not be expected at the end of reaction 1: since both structures are hexagonal. literature Cited
(1) Aronson, S.,U. S. At. Energy Comm. WAPD-TM-44, 15 (March 15: 1957). (2) J. C.. J . Znorq. 'Vuclear Chem. 7, 384 . , Aronson,. S.,. Clavton, . (1958). (3) Blackburn, P. E., J . Phys. Chem. 62, 897 (1958). (4) Burdick, M. D., Parker. H. S., J . A m . Ceram. SOL. 39, 181 (1956). (5) DeMarco, R. E., Mendel, M. G., J . Phys. Chem. 64, 132 (1960). (6) Gronvold, F., J . Inorg. .Yuclear Chem. 1, 370 (1955). (7) Hoekstra, H. R., Siegel, S., Fuchs, L. H., Katz, J. J., J . Phys. Chem. 59, 136 (1955). (8) Kempter, C. P., Elliot, R. O . , J . Chem. Phys. 30, 1524 (1959). (9) Notz, K. J., Mendel, M. G., J . Znorg. .Vuclear Chem. 14, 55 fl9hO\.
(lo'i !&el, S.,.4cta Cryst. 8, 617 (1955). (11) \Yeissbart. J.. Blackburn, P. E., Gulbransen. E. A., C. s. At Energy Comm. AECU-3729 (June 6, 1957). 112) LVilson. \Y. B.. U. S.At. Energy Comm. APEX-187 (Jan. 31, ' lb55) ; APEX-202 (June 1955) (uclassified)
RECEIVED for review June 9, 1961 .ACCEPTED November 29, 1961
--f.
Division of Industrial and Engineering Chemistry, 139th Meeting, ACS, St. Louis, Mo., March 1961. \Vork supported by the U. s. Atomic Energy Commission under contract .AT (30-1)-1156.
LOW TEMPERATURE GASEOUS REDUCTION OF IRON ORE IN T H E PRESENCE OF ALKALI A R T H U R M C G E O R G E , J R . , l A .
N O R M A N H I X S O N , A N D
K . A.
K R I E G E R
The School of Chemical Engineering, C-ninioersitj of Pennsylvania. Philadelphia, Pa.
Rates of reduction by hydrogen of pure Fe203 and selected iron ores were measured from 250" to 500" C., with and without the addition of N a 2 0 as Na2C03, to test the reported beneficial effect of alkali metal compounds. The reduction rate, a t 50% reduction, was greater for pure Fen03 than for either ore in the absence of added alkali. Although the high surface area lateritic ore reduced rapidly a t first, its rate was slower than that of the low surface area magnetic ore a t reduction. Addition of N a 2 0 decreased the reduction rate with pure FelOs but increased it with the ores, indicating that N a 2 0 functions b y freeing iron oxides from impurities rather than by promotion in the catalytic sense. The suggested reaction mechanism involves reduction by weakly adsorbed hydrogen, inhibited by the strongly adsorbed product, water.
Soy0
HE PRODUCTIOS O F SPONGE IRON by low temperature reTduction of iron oxides with hydrogen or carbon monoxide has been studied by numerous investigators. The extensive literature has been surveyed in three articles ( 7 , 2. 22). Among other factors, it has been reported that the reduction rate is increased by adding to iron ores small quantities of alkali metal compounds. LVilliams and Ragatz (25) found that
Present address, E. I. du Pont de Nemours & Co., Inc., Wilmington, Del.
potassium and sodium salts increased the reduction rate of a magnetite ore, the potassium salts being superior. Most effective were K2C03: KHC03. K?C?H302, KZCr?O,, and KOH. KC1 and K 3 P 0 4were not very effective. Barrett (2) also noticed a promotional effect of alkali on the reduction of a magnetite concentrate in a fluidized bed. One of the authors of this article had observed pilot plant tests on the reduction of a lateritic hematite ore in which small amounts of alkali had a pronounced beneficial effect. The mechanism of this promotional effpct is discussed here. VOL. 1
NO. 3
JULY 1 9 6 2
217
Plan of Investigation
Hydrogen was chosen as the reducing gas for several reasons. The chief impurity, oxygen, could be readily removed by passage over a heated nickel catalyst and freezing out the resulting water. I t also avoided the problem of carbon deposition that might result with carbon monoxide. Further, the use of hydrogen at low total pressures should minimize the effect of gas film resistance to molecular diffusion which might obscure the true mechanism of the reaction. The temperature range, 250' to 500' C., was selected to stay well below conditions for the formation of FeO as an intermediate in the reduction process. The equilibrium data (8, 7 7 , 78) show that wustite is unstable below 570" C. I t was felt that the formation of alkali ferrites on particle surfaces might be involved in the promotion of reduction, so some studies were undertaken with C.P. FesOs, eliminating the complication of the presence of silica and alumina. Further, the effect of such variables as total pressure, hydrogen flow rate, temperature, and particle size could be determined under readily reproducible conditions. Two very different types of ores were selected for testing. One was a crystalline, dense magnetite which was concentrated, after grinding, by flotation and magnetic separation. T h e other was a lateritic hematite that occurs naturally as a very fine material of high specific surface area. I n this ore an appreciable amount of the iron is chemically combined with impurities and little concentration can be accomplished by physical methods. Apparatus
The apparatus had to fulfill two special qualifications. I t was considered essential to follow the course of the reduction continuously. Further, operation a t low, controlled total pres-
sure was desired. A loss-in-weight method described by Barrett (3)was adapted. The result was essentially a McBain-Bakr sorption balance converted to follow the reduction rate. A small platinum bucket was suspended by a fine platinum wire attached to the bottom of a quartz coil spring. This was enclosed in a borosilicate glass assembly. The quartz spring hung in a water-jacketed (thermostated) case. The platinum bucket hung down into a reaction tube equipped with gas connections and a thermocouple well. An electric resistance furnace was fitted around the reaction tube and gas inlet line, which served as a preheater. The gas entered the bottom of the reaction tube, flowed up around the bucket, and out a t the top. Temperature was measured by a Chromel-Alumel thermocouple in the well located just below the bucket. The control was within + I ' C. The quartz spring obeyed Hooke's law below a load of 1.5 grams; the extension was 1.64 mm. per 0.01 gram of applied load a t 25' C. Spring movements could be measured reproducibly by a cathetometer to the nearest 0.003 mm., equivalent to a weight loss of 0.00002 gram or about 0.1% in these experiments. Hydrogen and/or nitrogen flow from stock cylinders was controlled by reducing and needle valves, in series. and measured by a calibrated orifice. I t passed over a heated nickel catalyst to convert the small amount of oxygen in the hydrogen to water which was then frozen out in a dry iceacetone trap. After passing through the reactor, any water formed in the reduction was removed by a second dry iceacetone trap. The gas left the system through a Greiner manostat and a Welch Duo-Seal vacuum pump equipped with an oil diffusion pump. The pressure was controlled to i1 mm. Hg. The over-all assembly is shown in Figure 1.
)
McLEOD GAGE MANOMETER
ASSEMBLY
VACUUM PUMP
ORIFICE METER
TE MPE RAT U RE CON TROLL E R H 2 AND N2 TANKS Figure 1 . 218
l & E C P R O C E S S DES G N A N D D E V E L O P M E N T
Apparatus diagram
I n one series of experiments water vapor was added to the hydrogen stream by bubbling the hydrogen through water maintained at a controlled temperature. The first low-temperature trap was converted to a bubbler for this purpose. The assumption that the hydrogen was saturated was checked in blank runs by freezing out the water in the trap following the reduction vessel. The check was within 5y0 which was considered adequate. Surface area measurements were made in a n apparatus described by Krieger (74) using the Brunauer-Emmett-Teller ( 6 ) method. Materials Used
The sodium h>-droxideand sodium bicarbonate were reagent grade. The pure iron oxide was C.P. grade of Fe2O3 which contained 0.062SaZO. Hydrogen was used from standard cylinders. .4dditions of water were made with distilled water. The ores used were supplied from standard samples by a n industrial laboratory. T h e lateritic hematite was from Cuba, the magnetite \\’as from Cornwall, Pa. The magnetite was beneficiated and t\vo samples were used, one a concentrate: the other a middle cut. T h e following analyses are on a dr>basis : Cornwall Magnetite MiddIe Concentrate cut 89.6 81.0 0.6 1.3 4.7 9.5 1.7 2.0 2.4 4.1 1.1 2.0 _ 0_0_ _ 0_0 100 1 99 9 0
Lateritic Hematite
I’e203 NO COO Cr03
77.5 1.2 3.2 A1203 11.7 Si02 4.5 MnO 0.9 NalO 0 1 CaO 0 1 MgO 0 7 Combined H 2 0 __ 0 0 Tot a1 99.9
+
Fe304
S
Si02 A1203
MgO CaO
Combined H20 Total
SurEace areas determined for various samples are given in detail later. .%yerage values in square meters per gram were 1.0 for the magnetite. 10 for the F e 2 0 3 ,and 120 for the lateritic hematite.
or lifting effect of the gas stream was measured. The system was allowed to cool and, after evacuation: purged with nitrogen before admitting air. Results
Reduction began as soon as hydrogen was admitted to the system. During the unsteady state a t start-up the pressure and flow rate were always lower than the final conditions. The reduction in this period varied from 10 to 207,. Thereafter, the reduction rate was constant until 7 5 to 807, had been reduced. I t then decreased to approximately zero a t 97 to 99% reduction. .411 of the reduced samples proved to be pyrophoric and reoxidized rapidly a t room temperature, when air was admitted a t the end of a run. Typical curves showing the progress of reduction are shoum in Figure 2, which represent experiments to determine the effect of particle size using pure Fe?Os. Four samples were prepared: -65 100, -100 150, -200. and -100 mesh (labeled unsized), All \ v e x reduced under identical conditions. For comparison the reaction rates were calculated a t the point of 50% reduction (Table I). This point corresponded to the maximum rate with the pure oxide and for many of the ore samples. Further, it was at the approximate mid-point of the constant rate period. The effect of particle size on the rate \vas slight with only the largest size showing an appreciable decrease. A comparison a t 80% reduction shows a wider divergence than at the chosen 50yc point. Table I also shows surface area measurements made for three of the samples. The method is accurate to
+
+
8 0 --
Analyses
The standard ore samples were analyzed by the laborator) that supplied them. This laboratory also made the Dial0 analyses using a method developed under the supervision of one of the authors. It involved the Smith fusion method (23) co bring the sodium salts into solution and a Perkin-Elmer flame photometer to determine the sodium content of the resulting solutions.
c
vt L
Procedure
A sample benveen 0.1 and 0.15 gram was weighed into the platinum bucket. which was then suspended from the quartz spring and placed in the reaction tube. The cathetometer was focused on the bottom coil of the spring and the position noted. T h e entire system was evacuated, the vacuum broken Ivith nitrogen and re-evacuated to ensure removal of oxygen. T h e iemperature of the reaction tube was brought to 350’ C. for 30 minutes. The temperature was then raised or lowered to the desired level and the spring position again recorded. Hydrogen !vas admitted to the system until the desired pressure (100- or 200-mm. Hg) had been reached. T h e Greiner manostat was set and the flow rate adjusted to the selected value. Steady-state conditions were reached in 4 to 6 minutes. h record of the spring position was made at intervals of one minute or more (depending on the rate). O n completion of reduction, the hydrogen flow was stopped and the buoyant
-
A x .
60 -
20
-
-
I
I
Time
Figure 2. Fe20s
I
20
0
-
I
40
Mlnulss
Effect of particle size using chemically pure
W. 6 5 mesh
X. 200 mesh
A. 100 $- 150 mesh
0 . Unsized
VOL.
1
NO. 3 J U L Y
1962
219
Table I.
Reduction Rate vs. Particle Size
C.P. grade FezOa. Temperature, 350' C. Pressure, 100 mm. Hg. Hydro. gen flow rate, 0.069 g./min.
Run -VO
48 49 51 50
53 52 65
Reductton Rate, G. Lost/Mtn./G. Fe?Ov 0 00966 0 0137 0 0134 0 0115 0 0117 0 0135 0 0117
Table II.
Surjace Area,
Sq. .M./G.
Size, Mesh 100
- 65 -100 -100 - 200 - 200 - 100 -100
+ + 150 + 150
11 7 9 1
9 1 12 8 12 8
Reduction of Pure Fe2O3
Effects o f temperature, pressure, and hydrogen flow rate Surface area, 12.8 s q . m./g.
Run
Temp..
-YO
C. 325 350 375 400 325 350 375 325 375 350
80 65 78 91 86 63 81 84 85 64
Prrssuie. 'Mm. H g
100 100 100 100 100 100
100 200 200 100
Hjdrogen F h Ratc, G,/Min. 0.069 0.069 0.069 0,069 0.1335 0.1335 0.1335 0.069 0,069 0,0965
Rate of Reduction, G. Lost/Min./G. Fed:,
0.0064 0.0117 0.0192 0.0210
0.00795 0.0138 0.0204 0.00865 0.0199 0.0113 Time
about 10 to 157, in this range; thus the results are not decisive. It was concluded that grinding samples finer than - 100 mesh \vas not necessary. The next series of experiments (Table I1 and Figure 3) measured the variation of reduction rate of pure ferric oxide \vith changes in temperature, pressure, and hydrogen flow rate. I t is clear that at constant total pressure and hydrogen flow rate each increase in temperature caused a definite increase in reaction rate. The range covered was from 325' to 375' C. with one measurement at 400' C. The effect of doubling the total pressure a t the same hydrogen mass flow rate gave a rate increase at 325' C., but a t 375' C. the difference was slight. Similarly. increasing the hydrogen flow rate, with other conditions constant, gave an increased rate at 325' C.: but at the higher temperatures the difference was small. An additional experiment was made to investigate the lower temperature range using a new technique. iVhen the reduction rate was low: it \vas found that advantage could be taken of the long constant rate period to make measurements a t more than one temperature. I n run 82 reduction was started a t 246' C. it'hen measurements indicated that the constant rate period had been reached, the temperature was increased and the procedure repeated. Table I11 lists energies of activation over the temperature range 246' to 400' C. calculated from an .4rrhenius graph of the experimental data (Figure 4). The most complete data Lvere obtained at 100 mm. of Hg total pressure with a hydrogen flow rate of 0.069 gram per minute. Between 290' and 370' C. the energy of activation is 15,300 calories per gram-mole. Between 250' and 290' C. it is 27,200, and between 375' and 400' C. it is 2900. Both of these latter values are open to question because of the limited temperature range covered. Doubling the hydrogen flobv rate gave an increase in energy that is within experimental error. The decrease determined 220
I&EC PROCESS DESIGN A N D DEVELOPMENT
-
60 Minutes
Figure 3. Effects of temperature, pressure, and flow rate of chemically pure FezOs Temp.,
C.
A. 325 X. 350 375 V. 325 0 . 375 w. 325 -. 375
.
Press., Mm.
Flow Rate, G./Min.
100 100
0.069 0.069 0.069 0.1335 0.1335 0.069 0.069
100 100
100 200 200
on doubling the pressure is significant; however, it is based on only nvo points. Sodium bicarbonate was added to the C . P . grade of Fez03 by two methods. In runs 30 and 34, the ore was moistened Lvith a bicarbonate solution, dried, and pulverized. I n runs 37 and 40, dry bicarbonate was sifted through a 150-mesh screen onto the powdered oxide and mixed by tumbling overnight. I n both cases the mixture was heated for 3 hours a t 350' C. to decompose the bicarbonate to the carbonate. Reductions were made a t 350' C.: 100-mm. Hg total pressure. and a hydrogen flow rate of 0.069 gram per minute. The results are given in Table I V and Figure 5 . The addition of the alkali caused a pronounced decrease in rate in all cases. R u n 65, which gave a reduction rate of 0,0117 gram lost per minute per gram with no alkali added, can be compared with run 30. which gave a rate of 0.00535 for a sample containing 0.68% S a ? O . Increasing the amount of alkali present caused a greater retarding effect. Apparently addition of the bicarbonate as a dry powder had less retarding effect than as a solution. The lateritic hematite ore was used in the next series of experiments. I t was ground so that 50% passed a 200-mesh screen and \vas run "as is" (run 46), with 0.687, S a 2 0 (runs 41, 43), and lvith 10% N a 2 0 (runs 42, 44). For runs 57 and 72, the ore \vas leached in boiling sodium hydroxide
for 30 minutes and washed 10 times in distilled Ivater. I t \\as dried a t 110' C. T h e analysis of the leached ore on a dry basis (%) \vas:
70
Ore
84.4
Fen03
4.6
.\I203
Si02 NiO COO MnO Cr,O, Mi0
3 6 1.5 0 9 3 2 0 8 1 0 0 0 100.0
+
~~
Nan0 Combined H,O Total
~
T h e resulrs are given in Table \. and Figure 6. 'The reduction rate of this ore was much sloiver than ivith C.P. grade of FenOs, although the surface area was approximately tenfold greater. T h e reduction curves are also of a different shape. T h e initial rate was the highest observed. even though during the start-up period conditions \vere not as favorable. T h e rate fell off continuously \vith time or per cent reduction and, depending on temperature and S a 1 0 content: the rate dropped to a negligible amount ar -Oyc to 90%, reduction. Addition of only 0.68% S a 2 0 seemed to have little effect a t the temperatures tested. but samples with 10% N a 2 0 shokved considerable improvement in both the reducticn rate and total amount reduced. T h e leached ore showed a small effect at 350" C . : but a pronounced one at 450' C. Energies of activation \\-ere 9400 calories per grammole with 0.68% NaZO, 15,500 calories per gram-mole with 10% S a Z 0 , and 19,000 calories per gram-mole for the leached material.
0001 14
18
16
20
10'/ToK
Figure 4. Arrhenius graph of reduction rate of chemically pure F e z 0 3 Press., Mm. 100 0 . 100 A. 2 0 0
x.
Flow Rate, G./Min. 0.069 0.1335 0.069
T h e samples of' magnetite were reduced at 350' C. lvitli and Ivithout the addition of sodium bicarbonate ['rable 1-1 and Figure 7). T h e added alkali again increased the reduction rate, although not as significantly as \\.ith the hematite ore. However, at 450' C., the effect would probably have been greater. T h e amount of alkali selected was approximately the stoichiometric amount to react with the alumina and silica present, as was the case with the 107, addition to the hematite ore. All of the magnetite samples shoived a definite induction period, after the introduction of hydrogen, before reduction became detectable. I'o test the effect of adding water vapor to the inlet hydrogen, further experiments lvere made with the C . P . grade of Fe?Oa. T h e technique of determining rates at several differenr teinperatures in one experiment was employed. Table \.I1 lists the results. Also shoivn is the drivins force. P&o-PIi20, that would be involved in removing rhe product ivater vapor from the reaction surface. Pigo is the equilibrium partial pressure of water vapor in a mixture lvith hydrogen in contact ivith iron and its osides at the specified temperature. 7'he effect of the addirion was a drastic loivering of the rate. S o reduction was detected below 400' C. at a partial pressure of water of 6.0-mm. Hg, Lvhile a t a parrial pressure of 9.2mm. Hg, no reduction \vas observed below 4'5' C. This means that the reduction rate \vas beloiv about 0.0002gram lost per minute per gram.
Time
- Minuter
Figure 5. Effect of addition of NoHC03 to chemically pure ~e203 Temperature, 3 5 0 ' C.;
pressure, 100 mm.;
X. Batch 1 .
0.68% as 0. Batch 2. 0.77% as H. Batch 3. 0.88% as A. Batch 4. 5.55% as C.P.FeaOs 0.06% as
+.
v 0 L.
1
flaw rate, 0.069 g./min Nan0 Nan0 Nan0 Na?O Na?O
NO. 3
JULY
1962
221
Figure 6.
b
Reduction of a h e m a t i t e o r e
Hydrogen flow rate,
I
I
0.069 g./min.; pressure, 100 mm.
% 0 s NaaOa
Temp., C. $. 4 5 0 0. 3 5 0 W. 4 5 0 X. 3 5 0 A. 4 5 0 w. 350
0.09 0.68 0.68 10.00 10.00 1 .oo leached in hot 50% N a O H
V. 4 5 0
I
.
1 .oo
c
L
,
60
Discussion c
The previously reported beneficial effect of adding alkali or alkali salts in the reduction of iron ores has been confirmed. However, the mechanism involved would not appear to be as a promoter, using this term as it is applied in catalysis. The results obtained \Yith the C.P. grade of Fe2O3 are significant. Contrary to what \cas expected, the addition of soda had a deleterious effect. Apparently, the alkali coated the oxide particles, making them more difficult to reduce. Sodium bicarbonate as a dry powder had less retarding effect than when added as a solution. The latter method might be more effective in coating the particles. A comparison of these results with those of the hematite ore is interesting. The specific surface area of the ore was about 120 square meters per gram compared with 10 for the pure oxide. Yet the pure oxide reduced a t a rate 10 times that of the ore under comparable conditions. -4dding sodium bicarbonate to pi\e a residual soda content of 0.68% seemed to have little effect, but the effect of 10.0% was marked. At 350' C. the rate increased from 0.00108 gram lost per minute per gram of Fe?Os to 0.00162, while a t 450' C. the increase \cas from 0.00411 to 0.00906. The sample that was preleached in sodium hydroxide showed little difference at 350" C., but a pronounced one at 450' C. The high surface area of the ore caused the initial reduction rate to be the highest noted. Apparently, 30% of the reduc-
0
0 c 0
W
a
8
40
20
0
P
I
-
110
-
I50
Minutes
Figure 7. Reduction of a m a g n e t i t e o r e
% NazO
0.67 X. 68 W. 69 A. 7 0
222
I&EC P R O C E S S D E S I G N A N l :
0.00 0.00 5.25 5.25
DEVELOPMENT
I
I
I20
Minutes
tion occurs without penetrating deeply deep11 into the pal particle. ticle .As reduction proceeds, unreduced oxide is less available atailable to the surface. I t is, apparently, in this regard that the alkali becomes a factor. By Bv reacting with impurities, chiefly chiefl1 alumina, foi silica, and chromia, chromia. it uncovers or frees the iron oxides for further reduction. reduction Preleaching the ore to remove remo\ e impurities
Temperature, 3 5 0 ' C.; pressure, 100 mm.; flow rate, Run No.
I
80
40 Time
70 Time
f
0.069 g./min. Type Mids. Con& Cone. Midr.
190
Table 111.
Energies of Activation in Reduction of Pure Fep03 Rate of Experimental TernbelaReduction. Enerev c f Activ"atioh, Run _7 _x_ To4 G. LostlMk. ture, ATQ. Cal./G.-Mole T , K. G. Fez03 c. Toto! pressure, 100 mm. Hg; hydrogen flow rate,
82 82 82 82 80 65 78 91
246 263 286 305 325 350 375 400
19.25 18.65 17.90
17.30 16 71 16.05 15 45 14.85
Total pressure, 100 mm. Hg;
86 63 81
325 350 375
16.71 16.05 15.45
Total pressure, 200 mm. Hg;
84 85
325 375
16.71 15.45
0.0006 0,00136 0.00298 0.00379 0.0064 0.0117 0.0192 0.0210
0.069 g./min.
27,200
Run 30 34 37 40
hydrogen flow rate,
0,00865 0.0199
Hydrogen flcw rate, 0.069
Pressure, 100 mm. Hg. g./min.
RPduction Rdr, G Lost/Mzn / G . Fen08 0.00535 0.00555 0,00765 0.0048
Air20 0.68 0.77
70
0.88
5.55
Surfa-e Area, Sq. M./G.
... ...
10.8
10.8
15,300 2,900
Reduction of lateritic Hematite
Table V.
hydrogen flow rate, 0 . 1 3 3 5 g./min.
0.00795 0.0138 0.0204
Reduction of Pure Fez03in Presence of NatO
Table IV.
Temperature. 3 5 0 ' C.
Pressure, 1 00-mm. Hg.
Hydrogen flow rate,
0.069 g./min.
13,200
would be expected to produce a similar effect. Although the results a t 350' C. showed little effect, a t 450' C. preleaching was about as beneficial as the addition of 10% Na,O. Increasing the N a Z O content from 0.68'% to 10% increased the energy of activation from 9400 calories per gram-mole to 15,500, reflecting the greater effectiveness of the larger soda content a t high temperatures. This latter value is very close to that for the pure oxide. This increase i. different from that normally associated with promoters in catalytic reactions. Such promotion is generally attributed to compound formation between the catalyst and promoter, or to the formation and preservation of more active sites on the catalyst surface. This is frequently associated with a decrease in energy of activation. Both samples of magnetite contained smaller amounts of impurities than the hematite : however, the specific surface area was only 1 square meter per gram. At the point of 50% reduction at 350' C. the reduction rates were considerably higher than the hematite, but below pure ferric oxide. T h e rates a t the start of reduction, though, were very low. There was a period of about 5 minutes before any detectable reaction occurred in distinct contrast to the hematite which reduced rapidly a t this point. Without question, this difference in behavior was caused by their great difference in surface areas. The fact that the magnetite was reacting faster a t the 50%) point is a reflection not only of the lower content of impurities, but that a much lower fraction of the iron is chemically combined with the impurities. The addition of NalO to the magnetite showed a beneficial effect on the reduction rate. Without doubt, higher temperatures would have shoi+n a greater effect. Here again, it would appear that the alkali functioned by reacting with impurities to free F e 3 0 4for reduction. The energy of activation for the reduction of C.P. grade of Fe103 for the range 290' C. to 370' C. was 15,300 calories per gram-mole. Rostovtsev and coworkers (20-22) reported values for a three-stage reduction of Fez03 between 300' and 400' C. to be 17,700. 15,450, and 14.200 calories per grammole, respectively, for the Fen03 + FQO+ stage, for FeaO, -+ FeO, and for FeO + Fe. The average of these values compares well with the value calculated in this work. The existence
yo
Run No. 46 41
Temp., a C. 450 350
42 43
350 450
10.00 0.68
44 57a
450 350
10.00 1.00
a
Xa?O
0.10
0.68
7211 450 Caustic leached.
1.00
Reduction of Magnetic Ore
Table VI. Temperature, 3 5 0 ' C.
Table VII.
Surface area, 1 sq. m./g.
Pressure, 100 mm. Hg.
Flow Rate, G./Min. 0.069 0.069 0.069 0,069 0.1335
Run 'Vo. 67 Mids. 68 Conc. 69 Conc. 70 Mids. 71 Conc.
Rate of Reduction, G. Lmt/Min./ G. FeaOc 0.0033 0.0025 0.00383 0.00377 0,00358
yo A'az0 0.00 0.00
5.25 5.25 0.00
Effect of Addition of Water Vapor
Mm. Hg
(P;,oPgzO)
Temp.,
c.
PHz
PlzO
pH20
100
Run No. 7 4 . Total pressure,
350 400 435 461 468 468 470 470
94.5 94.3 92.6 93 8 93.8 92.5 94.6 94.9
5.5 5 7 7.4 6.2 6.2 7.5 5.4 5.1
Run No.
95.1 95.1 95.0 94.9
6.3 9.7 11.9 14.1 14.6 14.6 14.9 14.9
4.9 4.9 5.0 5.1
76. Total pressure, 00
370 90.8 41 3 90.8 455 90.8 48 1 90.8 90.7 508 0 Less than 0.0002 g.
mm. Hg.
Rate of Reduction, G. Lost/Min./G. Fcd~
H? flow ra!e,
0.8 4.0 4.5 7.9 8.4 7.1 9.5 9.8
0.069 g./min.
None obs.0 0.0004 0 00068 0.00414 0,0055 0,00313 0,00632 0.00811
00 mm. Hg. HZ flow rate, 0.1 3 4 g./min.
Run No. 75. Total pressure,
350 382 430 482
0.069 g./min.
Experimental Reduction Rate, G. Surface Energy of Lost/Min./ Area, Sq. Actination, G. .&f./G. Cal./G.-Mole 0,00411 130 0.00108 130 9,400 (41. 43) 0.00162 112 0,00411 130 15.500 (42, 44) 0.00906 112 0,00117 115 19.000 (57, 72) 0.0099 115
15,700
6.3 8.3 11.7 15.8
1.4 3.4 6.7 10.7
mm. Hg.
H? flow 1.8
9.2 7.4 10.1 9.2 13.6 9.2 15.6 9.2 18.0 9.3 lost/min./g.
VOL. 1
0.9
4.4 6.4 8.7
NO. 3
None obs.. None obs.0 0.00345 0.011 rate,
0.069 g./min.
None obs.0 None obs.5 None 0bs.S 0.00218 0.0058
JULY 1962
223
of these stages was inferred from the existence of minima in curves showing per cent reduction LIS. time. The occurrence of such minima was not detected in the work presented here. The unsteady state a t the start of each run might obscure the FesO4 break, but the others did not occur. At the Fez03 temperatures involved. FeO is not stable and its appearance as a n intermediate stage is difficult to explain. Bitsianes and Joseph (5) did not find any evidence of FeO in partially reduced samples quenched from 500' C., which is in accord lvith the equilibrium data. Olmer (77) reduced FelO3 and F e 3 0 4in pure hydrogen increasing temperature linearly with time. Below 325' C. the FelO3 reduced to FeaOl and then to Fe. Above 325' C. both FelOs and FesOl reduced directly to metallic iron; the formation of FeO was not observed. The energy of activation between 250' and 290' C.? calculated to be 27,200 calories per gram-mole, can be compared with the value 30,000 calories per gram-mole reported by Roiter and others (19) for the range 200' to 300' C. I n the upper range of temperatures (375 to 400' C.) the energy of activation appeared to drop abruptly to a value of 2900 calories per gram-mole, although the exact value is admittedly doubtful. This might indicate a shift from a surface chemical reaction as the controlling factor to an increased importance of gaseous diffusion. I t might also indicate an alteration of the surface, perhaps caused by sintering. The sensitivity of the reduction rate of C.P. grade of F e 2 0 3 to the presence of water vapor is illustrated in Figure 8 by plotting the reaction rate us. the driving force, P*HzO-PHzO: that would be active in removing the water vapor from the reaction surface. P*HzOis the equilibrium value a t the reacting surface, PHlo the value in the main hydrogen stream. \Vater vapor was added to the hydrogen stream in runs 1 4 through 16, while in runs 65, 78, 80, etc., the hydrogen stream was "dry," Runs 74 and 16 were at the same total pressure and hydrogen flow rate as the dry runs and can be directly compared. For example, no water was added in run 80; therefore, the inlet partial pressure of water was
-
A P: P*H,O-PH,O
Figure 8. tion rate
m m.Hg.
Effect of driving force on reduc0 . Runs 65, 78, 80, 82, 91 X. Run 7 4 H. Run 7 6
+. Run 224
7575
l&EC PROCESS D E S I G N A N D DEVELOPMENT
0.002-mm. Hg. From the reaction the outlet partial pressure was approximately 0.1-mm. Hg. Under these ccnditions a reduction rate of 0.0065 gram lost per minute per gram v a s measured a t 325' C. I n run '4 the partial pressure of water was 5.4-mm. Hg, and it was necessar) to raise the temperature to 470' C. to obtain the same reduction rate. I n run 76 the water partial pressure was raised to 9.3-mm. Hg, and a t 508' C. the rate was only 0.00.33gram lost per minute per gram. Further. the graph sholvs that at the same values of the driving force, the dry runs gave much higher rates. In view of these results. an attempt was made a t correlation using the Langmuir (75) adsorption isotherm for a strong11 adsorbed gas assuming the adsorption of water vapor on iron oxide to be a good example.
where 5' = fraction of total number of sites occupied by HzO b = a constant-adsorption coefficient PHzO= partial pressure of HlO in mm. of Hg. The fraction of sites left bare is therefore:
Assuming that the reduction could be described by the mechanism for a weakly adsorbed gas, hydrogen, retarded by the presence of a strongly adsorbed product gas, water, the reaction rate is proportional to the amount of the reactant adsorbed and to the uncovered oxide surface. Also, for a weakly adsorbed gas the amount adsorbed is proportional to the partial pressure of the adsorbed gas. Therefore, the rate equation for a constant temperature is:
where R = rate of reduction. grams lost per minute per gram k = a constant PRz= partial pressure of hydrogen in mm. of Hg.
To evaluate the expression, the constants were calculated from experimental values of R, PH2.and P,,,. These were used to predict a rate for a similar run and compared with that actually obtained. The partial pressure of \\ater vapor was accurately knolvn in those cases where it was added to the hydrogen stream. Here the water produced by the reaction was insignificant and the partial pressure was essentially constant. In the dr! runs, the change in partial pressure caused by the reaction was important and a log mean of the inlet and outlet values was used. One run of each type !$as used to evaluate the constants. The predicted rates agreed with the observed values within 257,. However, because of the tremendous effect of small amounts of water vapor. it is necessary to know the water partial pressure very accurately. For example, the use of the arithmetic mean partial pressure in the dry runs was much less satisfactory than the log mean. To settle the issue, considerably more data would be necessary and the range between l and 3 mm. of water vapor should be accuratelycovered. This could not be done satisfactorily in the present experimental equipment. Chufarov and others (7) have studied the effect of gaseous reduction products on the reduction rate of iron oxides with hydrogen and carbon monoxide and report that data on the retarding effect agree with experimental data for the surface adsorption of reducing gases and reaction products on iron oxides. Other work in this field has been done (4,9, 70, 72, 73, 76,24).
Conclusions
A loss-in-weight method was used to determine the reduction characteristics of C.P. grade of Fez03 and samples of two very different types of ores. Being continuous and accurate, comparisons were readily made. The addition of Na20 in the form of N a H C 0 3 to (c.P. grade) F e ? 0 3 results in a decrease in the reduction rate with pure hydrogen as the reducing agent. .A similar addition of NaZO to a Cuban lateritic hematite containing alumina, silica, and chromia as impurities results in an increased reduction rate. An increase in rate upon the addition of S a 2 0 is also obtained for a dense, crystalline, Cornwall magnetite ore. The alkali apparently reacts with the impurities in the ores, thus freeing the iron oxide for reduction; i t does not appear to act as a "promoter" in the accepted sense of the word. 'The reduction rate of iron oxide is extremely sensitive to the presence of water vapor. There is some indication that the mechanism may be described as reduction by weakly adsorbed hydrogen, inhibited by a strongly adsorbed product, Ivater vapor. literature Cited
(1) Barrett, E. P., U. S. Bur. Mines, Bull. 519 (1954). (2) Barrett, E. P., U. S. Bur. Mines, Rep. Invest. 4402 (February 1949). (3) Bakett, E. P., IYood, C. E., IND. ENG. CHEM...ANAL. ED. 18, 285 (1946).
(4) Bitsianes, G., Joseph, T. L., J . Metals 5, A I M E Trans. 197, 1641 (1953). (5) Ibid.. 6, A I M E Trans. 200, 150 (1954). (6) Brunauer. S., Emmett, P. H., Teller, E., J . Am. Chem. Soc. 60. 309 (1938). (7) C h u f a k . G. I., .4verbukh. B. D., Tatievskaya, E. P., Antono\, V. K., Zhur. Fzz. Khzm. 28, 490 (1954). (8) Darken. L. S.. Gurry, R. \V., J . A m . Chem. SOC.67, 1398
,-,
i l C)A\\ ,",.
(9) Edstrom, J. O., J . Iron Steel Znst. (London) 175, 289 (1953). (10) Edstrom, J. O., Bitsianes, G.? J . Metals 7, A I M E Trans. 203,
760 (1955). (11) Emmett, P. H., Schultz, J. F., J . A m . Chem. SOC. 55, 1376 (1933). (12) Franklin, .4. D.: Campbell, R. B., J . Phys. Chem. 59, 65 (1957). (13) Kondakov, V. V., Chukanov, Z. F., Doklady Akad. LVaui; S.S.S.R. 106, 697 (1956). (14) Krieger. K. A.. IND.ENG.CHEM...\NAL. ED. 16. 398 (1944). (15) Langmuir: I., 2. Am. Cheni. SOC. 38, 2263 (1916j. (16) Moskvicheva. A. G.. Chufarov, G. I., Doklady Akad. LVauX S.S.S.R. 105, 510 (1955). (17) Olmer. F.. J . Phys. Chem. 47, 317 (1943). (18) Ralston. 0. C.. U. S. Bur. .\.lines. Bull. 296 (1929). (19) Roiter. V. A , , Yuza. V. .4.,Kutznetzov. .4.N., Zhur. Fiz. ' k h i m . 25, 960 (1951). (20) Rostovtsev. S. T., Trudy SozNeshchaniya Inst. M e t . Ural. Filiala Akad. -Yauk S.S.S.R. i .Magnitogorsk. -5fet. Kombinat. 1955, p. 65. (21)- Rostovtsev. S. T., Fiz.-h-him. Osnovy Pritoodstva Stali. Akad. AVauk S.S.S.R.. Ins!. M e t . im. .4. A. Baikova, Trudy 3 4 Konf., ~Moscow,1955, 191. (22) Rostovtsev, S. T., Em, .4.P., Doklady Akad. Nauk S.S.S.R. 93, 131 (1953). (23) Smith, J. L.. -Inn. Chem. Pharm. 159, 82 (1871). (24) IVetherill, \l'. N.: Furnas, C. C., IND.ENG. CHEM.26, 983 (1934). (25) \ l X a m s , C. E., Ragatz, R. A , ! Ihid., 28, 130 (1936). I
,
RECEIVED for review August 18, 1960 ACCEPTED January 25, 1962
FRACTIONATION OF FISSION ELEMENTS BY ELECTRODIALYSIS G . J .
BUB, J.
D. V I E , ' AND W .
H. W E B B
Department of Chemical Engineerinf and Chemistry, C-nimsity of Missouri, School of .Mines and Metallurgy, Rolla, .\lo, Electrodialysis methods employing Permutit cation exchange membranes 3 142 and anion exchange membranes 31 48 were feasible for the fractionation of some of the radioactive fission products. The methods are based on the characteristics of anionic, cationic, and radiocolloidal materials formed in the presence of complexing agents with adjustment of acidity. Fractionations of radioactive cerium, cesium, promethium, strontium, and zirconium, in equilibrium with their daughter decay products, were made. Studies were made with three-cell and seven-cell units. Best results were obtained with ammonium acid fluoride solutions using a seven-cell unit with a flowing system.
of aqueous solutions of irradiated reactor fuel elements for the recovery of fissionable materials by solvent extraction leaves a radioactive waste solution containing the fission products. This solution also contains various acids ("03, HF? etc.) depending on the method of dissolution of the fuel elements. The acid is recovered, leaving a waste solution containing the fission products. The nuclei formed by fission have mass numbers from 72 to 158 (70). There have been many suggestions for the disposal of this waste solution-some include recovery techniques for certain fission elements. 'The volume is usually reduced and the conHE PROCESSING
' Present address, Mallinckrodt Chemical Works. \Veldon Spring, MO.
centrate stored until the radiological hazard is reduced. The presence of cesium-1 37 and strontium-90 prolongs this storage. Electrodial!sis has been proposed by several investigators as a means of volume reduction and acid or electrolyte recovery (5-8). The electrodialysis process is accompanied by electroendosmosis. \\ater transport. Zirconium can be sorbed on silicage1 and recovered with oxalic acid ( 3 ) . Cesium and strontium can be fixed on siliceous materials ( 7). Other fission elements can be recovered by ion exchange or solvent extraction (2. 3 ) . These are only a few of the methods employed to treat the fission product solution and recover some of the radioelements. Since some of the fission elements are in demand for radiochemical and medical research, a single operation for separation and recovery would be most desirable. VOL. 1
NO. 3
JULY 1962
225
