be 1.O6y0total, with 0.0707, as COO and the remainder as the cobalt-promoted MooBcomplex. With a surface area of 275 sq. meters per gram and a bulk packing density of 0.67. such a catalyst would be expected to give 947, desulfurization a t 400’ C. Although a catalyst with these properties was not tested in this program, the existence of this optimum composition has been demonstrated (4.9, 72).
T h e correlation between intrinsic activity and active Co : Mo ratio is independent of method of preparation. This universal relationship is a valuable aid in understanding the relationship between catalytic activity and physical properties. Extension of these techniques to catalysts under various treatments such as sulfiding. regeneration. etc., should add to our knowledge of desulfurization processes.
Conclusions
literature Cited
This investigation develops three main conclusions regarding desulfurization cobalt-molybdena-alumina catalysts. First, the catalyst may- be divided into inactive components CoA1204, COO,and CoMoO,, sulfiding under reaction conditions to give CoA41204.Cog&. and Moss. T h e active component consists of some MOOBcomplex promoted with nonreducible cobalt. T h e exact‘nature of‘this complex is as yet unknown. but presumably it sulfides to a cobalt-molybdenum sulfide. \vhich is the true catalyst for the reaction. Second, catalysts with initial Co:Mo ratios less than 0.3 yield the same composition catalysts when heat treated in the range 538’ to 650’ C. No cobalt aluminate is present in these preparations. However: above initial Co: Mo values of 0.3. the final composition is dependent on both heat treatment and cobalt content. This accounts for many of the differences’ found in catalysts with the same initial conditions and for optimum values varying from 0.2 to 1.O. Finally. a maximum in activity occurs a t a n initial C o : M o ratio of 0.20 for all heat treatments and 0.54 for a 650’ C. heat treatment. both preparations resulting in an active C o : M o ratio of 0.18.
(1) Abeledo, C. R., Selwood, P. li.Bull. . Am. Phys. Soc. 6 , 353 (1 961). (2) Badger, E. H.M.. Griffith. R. H., Kewling, h i . B. S.. Proc. Roy. SOC.,4197, 184 (1949). (3) Beuther, H.. Flinn, R . A . . McKinley, J. B.. Ind. Eng. Chem. 51. 1349 (1959). (4) kngel. -.producing a perfect cleavage. Hardness in the cry-stal. caused by low temperature, impurities, or working effects. causrs the cleavage crack to run more freely. T h e dislocation density can be varied by the way in which the cleavage crack is passed through the crystal. COSTACT O F BALLSWITH SURFACE.i f a small ball is pressed lightly into contact \vith the surface of a LiF crystal. dislocations are nucleated a t the area of contact ( 7 ) . T h e maximum shear stress in the area of contact, calculated from the theoretical maximum pressure under the ball. compares favorably with the yield stress for the LiF crystal. .4hall \vas suspended from a string as a pendulum, with a crystal a t the zero deflection point, and allowed to strike and rebound from the cr>-stal surface. If a speck of a hard substance lies kvithin the area of contact, the hard particle acts as a stress concentrator and produces a characteristic array of dislocations which is called a "rosette." T h e stress caused by the impact of the ball \vas limited to a relatively small area. and the dislocation nucleation was localized. A very large number of impacts Ivould be required to produce a dislocation density increase equivalent to that obtained by macrostressing the crystal. c 2 U E N C H I N G F R O M HIGH T E M P E R A T C R E S . By cleaving a crystal into tlvo parts and then annraling one part. it has been shown ( 7 ) that LiF crystals can be annealed ivithout changing their dislocation density. Heating and cooling rates of less than 5" C. per hour are necessary. If a crystal is cooled more rapidly from 600' C.. the dislocation density is increased as much as 2 orders of magnitude. Intermediate cooling ratrs produce intermediate numbers of neiv dislocations. .4large fraction of these new dislocations consists of tiny closed loops (inside the crystal) ; therefore. the effect of quenching on the surface is variable. Temperature is also a variable in the quenching treatment. since only a small number of these new dislocations are found in crystals cooled from 400' C. Effects of Foreign Heterogeneities. Gilrnan (72) has sho\vn in his studies on plastic deformation of LiF crystals that groitm-in sources d o not play a dominant role in nucleating dislocations. The function of thr groxvn-in netxvork. if it is not completely pinned by impurities. is to provide moving dislocation loops Lvhich multiply as they move. (filman ha5 concluded that this situation probably prevails in all crystals. €:ram these observations hr \vas led to pursur the r f f e c t s of impuritie? on strrss dislocation nucleation in a series of interest160
I&EC
FUNDAMENTALS
ing studies. the results of ivhich can be briefly summarized as follows (72. 77) : I N c r CSIOSS. As-gro\vn LiF usually contains some large foreign inclusions Xvhich can be seen niicroscopically--. IVhen a stress is applied. starlike patterns of dislocations appear at these inclusions. PRECIPITATES. Small inclusions of approximately 0.1 micron d o not scatter light. but they can often be seen as black pits lipon etching. Stress pulses on such cry-stals produce glide bands springing from placrs lvhere impuritirs are known to be present. i R R A D I A T I O S - ~ S D U C E DP R E C I P I T A T E S . Proloneed irradiation with any type of ionizing radiation hardens LiF crystals 176). This effect becomes noticeable only when defects that are larger than single-point defects are produced during the irradiation. The irradiated crystals exhibit profuse dislocation nucleation when they are subjected to stress pulses. An unusually pure crystal \\'as irradiated and did not exhibit the nucleation effect upon stressing. although it did exhibit the other expected properties. Thus impurities were again sho\\-n to be essential in dislocation multiplication. PRECIPITATES INDUCED BY : ~ V N E A L I S G . Grown-in dislocation5 in as-received LiF crystals are normally immobile, because they are pinned by impurities. The Cottrell atmoipherrs responsible for this "pinning" lvere shown by Gilman to be dispersed by annealing (200' C . or higher) folloxved by rapid cooling (greater than 5' C . per hour). Subsequent stresq pu1~e.q\vi11 move these dislocations. causing profuse glidr band formation. This dislocation nucleation increaxs rvith annealing tcmperatures (400"to 600" C.) and increased cooling rates ( 7 2 ) . If. however. a cry-stal is annealed and cooled slowly (1 ' C. per h o u r ) . it increases in hardness by a factor of 5 (77) and much higher shear stresses are needed to generate neiv didocations. In fact. the dislocations are pinned by impurities even more strongly than in the as-received cry-stal. DIRT PARTICI.ES. Microscopic dirt particles are needed to produce dislocations on surfaces bl- pressing with glass spheres for \\.hen carefully cleaned glass spheres are pressed into contact with dislocation-free regions of a LiF crystal. < h e x s t r e w < as much as 100 times greater than the normal yield stresc of the crystal ma)- be attained without homogeneous dislocation nuclearion. Yet lrss carefully cleaned ball5 produce starlike patters of dislocations on LiF surfaces with only moderate stress.
Thus it can be concluded that small foreign heterogeneities cause most of the dislocation nucleation in LiF crystals and probably in all real cry-stals. However, the impurity levels need not be high. Crystals used in this research and in Gilman's \\or:< contained on the order of 1 p.p,m. of impurities. Although foreign heterogrneiries are needed for dislocation nucleation upon stressing. they must be dispersed in such a manner that they can cause profuse nuclration Xvith moderatr stresses. Such conditions can be assured by annraling a t elevated temperatures. follolved by rapid cooling. Experimental Details
Reaction a n d Reactant. Dehydrogenation of the rthanol to acetaldehyde and hydrogen was chosen as the reaction for stud>-. Although deh>-dration is also known to occur. the determination of the reaction products is not as accurate as in dehydrogenation and a relation betiveen dislocations and activity Lvorild be difficult to observe. T h i s is particiilarly true because the Ivater from the dehydration rapidly deactivates the LiF for further dehvdration. Lvhilr cho\ving little effect on the dehydrogenation
CSi Chrmical C o . L-SP. pure. absolute. reagmt grade rthanol \vas used for these tests. Because of its hyyroccopic nature. there \vas a small and unknown quantity of ivater in the ethanol. The amount of water \vas determined hy chromarographic inrthodc for rise in preparing staqdard calibration mixtures of ethanol. acetaldehyde. and \\ate?. Catalyst Preparation. The lithium fluoride crystals usrd in the preparation of the catalyst charges \yere ext1wnirl)- p111-c.
LEGEND Temperature Recorder No.1, Point N o . 2 , e t c . Tern pe ro t u re I n d Ico t In g Controller No. I
PRECISION FEED P U M P
Figure 2.
General flow arrangement for microreactor
(on the order of 1 p,p.m. of impurities) LiF crystals from the Harshaw Chemical Co.. Crystal-Solid State Division, Cleveland, Ohio. These crystals were obtained in the form of "random cuttings," measuring 5 to 15 mm.: and "random ingors," measuring 3/4 to 1' 2 inches, both produced during the cleavage of larger crysta in the preparation of windoivs for spectrorncters. T h e crystals came in t\vo grades: (1) ultraviolet quality and (2) infrared quality. T h e infrared quality showed absorption in the vacuum ultraviolet range and was not buitable for that application. Harshaw Chemical Co. (23) has assigned this absorption to a slightly higher impurity con tent. 'The crystal catalyst consisted of small LiF crysrals measuring approximately 1 y 2 X 2 mm. 'The crystals were prepared by cleaving. SO that all rxposed faces Liere the normal (100) cleavage planes. A wedge-type barber's razor honed to a n acute angle of about 30' was used. Cleavage steps were visible, but the faces at these steps were at right angles. so that only the (100) face \vas exposed. Invisible steps \ v e x undoubtedly present ( 3 J . 3 6 ) , but right angles would still typify the steps, exposing only the (100) face. T h e dislocation density in the cleaved crystal can be varied over a \vide range by the speed \vith which the cleavage crack is passed through the crystal ( 7 , 8.27). ..\ sharp tap on the chisel causes the crack to pass rapidly through the crystal. producing flat, smooth faces on the cleaved crystals. A softer blow causes the crack to pass more slowly, sometimes causing the cleaved parts to bend, and even allowing the crack to stop within the crystal. I n preparing the c a t a l y t charge a large crystal was cleaved into rods about 2 mm. thick. T h e cross-sectional dimensions of each rod werr measured with a measuring magnifier to +0.1 trim., and a number of thin wafers Lvere cleaved from rhe ends of the rod. T h e catalyst charge was iveighed on a microbalance, and the surface area was then easily calculated. Equipment. Catalytic conversion tests were made icith a rieu ly constructed automaric precision microreactor. T h e gr11c1~alarrangement of the equipment is shomsn in Figure 2 . Liquid feed \\?as metered to the system by a single-acting posi~ive-displacemerit pump. T h e construction and performance of this pump have been reported (35). T h e feed entervd a vaporizing tube \vhich was enclosed in the heated bampling box? and then passed through a '/'*-inch 0.d. stainless steel tubing bay-onet lvhich held the microreactor and extended into a n electrically heated fluidized sand bed. T h e
vapors reached the temperature of the constant temperature sand bed before entering the microreactor. T h e effluent from the microreactor flowed through a coil of ''*-inch 0.d. stainless tubing, which was enclosed in a heated box: and then to a vent. This box was separated from the fluidized sand bed by a sand disengaging space. Periodically the flow route was automatically changed to purge the contents of the coil into a chromatograph while the microreactor effluent vias bypassed directly to the vent. Heat for the sampling box was furnished by hot air flowing from the fluidized sand bed. 'The temperature of the box was held a t 120' C. by manually adjusting the flow of a n auxiliary air purge to the box. T h e microreactor design used in this study \vas essentially the same as that employed by Pozzi and Rase (18).T h e microreactor consisted of a length of '!,-inch 0.d. 0.180-inch i.d. Type 316 stainless steel tubing, with the volume defined by pieces of Refrasil cloth held in place by Swagelok fittings as illustrated in Figure 3. Reactor lengths varied from 13/32 to 4 inches, depending on the size of catalyst charged. T h e chromatograph was Perkin-Elmer Co., Model 154-D, equipped with an adsorption column inch in 0.d. by 4 meters long, packed with P-E Type W packing. conqisting of Teflon coated with Carbowax 1500 (polyethylene qlvcol) and operated a t 128' C. This appararus gave an analyGis of the microreactor effluent in about 5 minutes with a gas peak at 48 seconds, an acetaldehyde peak a t 84 seconds, an ethanol peak a t 150 seconds, and a water peak at 236 seconds. T h e instrument was adapted for this work by piping the helium carrier gas from the reference thermal conductivity cell to the automatic sampling system and returning this helium to the heated liquid-sample injection block. Standard liquid samples, for calibration, were injected before, during, and after each catalytic conversion test. Analyses were calculated from peak areas obtained from a mechanical integrator on the chromatograph recorder. T h e temperature of the microreactor apparatus was mraiured and recorded by a six-point Brown Electronik recording potentiometer. TVith the iron-constantan thermocouples used ' to 385' C , \vith an accuracy of the instrument span was 0 1.9' C . Iron-constantan thermocouples were mounted in Conax thermocouple glands at the bottom and top of the fluidized bed. T h e temperature of the microreactor sand brd was controlled by a Gardsman temperature controllrr. Thc
*
VOL. 3
NO. 2
M A Y
1964
161
the preliminary test and the catalytic test, samples of the feed being used a t the time were injected into the chromatograph with a syringe. Since the sensitivity of the chromatograph varied over long periods of time. samples of a standard mixture of acetaldehyde, ethanol, and water were also injected before each test.
b i O D 316 S S Tubing
d k$
Reducing U n i o n Refrosil
T h e microreactor to be used in the test: u i t h new Refrasil cloths installed, was placed in the sand bath which had been a t reaction temperature for several hours. This microreactor was purged with helium for a short time and then a four-way valve was changed so that the vaporized ethanol feed was admitted to the reactor. After the analysis became constant in the preliminary test, the four-way valve was turned to allow helium purge into the microreactor and the microreactor was removed from the sand bath. T h e catalyst charge was then placed in the microreactor without damaging the Refrasil cloths. M'hen the system had apparently reached steady state. the fourway valve was changed from bypass to reactor position, thereby allo\ving the feed to enter the microreactor. T h e effluent from the microreactor was then sampled every 5 minuter for about 1 hour. '4t the completion of the catalytic activity test. helium purge was admitted to the microreactor ba)onet. 'I'he catalvst charge was removed from the microreactor immediately, and reserved for determination of the dislocation denrity.
Reducing U n i o n
Figure 3.
Microreactor detail
Determination of Dislocation Density. T h e most satisfactory method for determining rhe dislocation density was found to be the visual comparison of photographs of the etched cr)sral surface \z ith photographs of known densities. 'l'hr etc.h procedure used in this study was that reported by (;ilman and Johnston ( 7 5 . 20).
span of the controller was 0' to ,315" C . , with an accuracy of abour + l . 6 ' C:. Recarisr of the l o ~ vconversions, thc variation of rraction ternperatirrr drie to heats o f reaction was less than the nieasrirable variations in the sand bath temperatiire.
Catalytic-Test Procedure. T h e esper imrntal work in this stridy consisted essentially of treating I.iI.' (r)stals i n various w a ) s IO produce different dislocation densities, nicaviring the tliclocation densities. and determining the L'ate of drh>rlrogcnation of ethanol in a precision microreactor. ,411 phases of this itork werr standardized so that t h r mass transport and thermodvnamic effects were constant. \Vith such a standard procedure a relation brrween dislocation density and catal) tic activity would be meaningful. A small convrrsion was detected when there was no catal!st in the microrrac,tor; therefore? a preliminary test \vas made \vith the empty reactor just before each catalytic test. l h r i n g
w-
n
II 3
'I he Etch A devrloped in that work was used. which consisted of equal parts of concentrated hydrofluoric acid and glacial acetic plus 1 volume % of conrentrated f t F saturated with FrF3. An etch time uf 60 to 75 seconds was found most effective. ~l'heetching was followed by a rinse in absolute cthanol, and then in anhydrous ether. This etch produced a square-based pyramidal pit a t each dislocation that cut the (100) cleavage face of the I.iF cryctal. 'I'he base edges of the pits lie parallel to (110) directions (ser Figure 1 ) . Gilman ( 1 5 ) has discussed the esidencr \vhich confirms the correspondence of these etch pits and diqlocations Standard draxvings were prepared by placing ~ c a l e db1ac.k squares on a large piece of paper. and photographic ncgarivcs
-
TEMPERATURE
$12-
w n -I
.I1
2 W
.IO-
-
ut< z .09 .W 2 .08 za gk.07
AVERAGE CONVERSION
PRESSURE
:
FEED RATE
'
:
260°C.
Atmospheric 6 cc/ hr.
b
v)
zy.06-
gg.050
"
u t3
2 0
.03 .04
.02
-
1
0
-
0
0
.01 -
0
162
a
I
I
I
I
0 2 4
I
6
8
IO 12 14 16 18 2 0 22 24 26 28 3 0 32 3 4 36 38 4 0 4 2 4 4 46 4 8 5 0 52 5 4 56 58
I&EC FUNDAMENTALS
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
l
i
l
l
of the drawings made as standards of known densities. T h e negatives for the crystals and those for the drawings were then compared on a light table. Independent observations by different people were made and the most likely corresponding negative was chosen as the density for each photograph of a crystal. ,4n average of the densities for the crystals from a catalyst charge was taken as the dislocation density for that charge.
Reporting of Reaction Rates. T h e reaction rates for the optical crystals are reported on a unit surface area basis. so that the various catalyst charges could be compared. T h e average conversions used in calculating the rates were taken so that the unsteady initial period and the less accurate later period were excluded from the calculation. T h e choice of the conversion measiirements used for calculations of the average conversion is illustrated for run 32 in Figure 4. T h e first sample was usually higher in conversion than the
Table 1.
remainder of the samples in the test. This may have been because of a higher activity for a fresh catalyst, or unsteadystate operation. and for this reason it was excluded from the average. T h e thermal conversion \vas measured as a preliminary test for each run. and this was subtracted from the total conversion for the run to obtain the catalytic conversion. As the test proceeded the catalytic activity declined to a level close to the thermal conversion. I n this range the inaccuracies were such that the data were not considered usable. Therefore, these points were excluded from the calculation of the average conversion. Results
Catalytic conversion tests were performed using three forms of LiF: reagent grade powder, optical grade crystals, and granular material. T h e LiF powder was used to confirm its
Summary of Catalytic Conversion Tests Showing Catalyst Preparation Methods
Run 'Vo.
Feed Rote, Cc./Hr.
Surface Area, S9. Cm.
AU. Rate, Conrcrsion, Cc./(Sq. Cm.) (Hr.) Mole
Dislocation Density per Sq. Cm.
Catalyst Razc .Material
.Method of P r ~ p a ~ a f i o n
6
6
3.41
0.59
1,045
2 2
x
10'
Random cutting
1.36
2 5
x
10'
Random cutting
0.01
0.006
5.6 X lo6
5.39
0.07
0.025
7
106
Random ingot
4
14.18
0 .08
0,023
5 X lo6
Random ingot
15
4
10.5
0.10
0.039
9
x
106
Random ingot
18
4
12.06
Xi1
Nil
