The Kinetics of Fluorination of Nickel Oxide by Chlorine Trifluoride

The Kinetics of Fluorination of Nickel Oxide by Chlorine Trifluoride. R. Lynn Farrar Jr., and Hilton A. Smith. J. Phys. Chem. , 1955, 59 (8), pp 763â€...
0 downloads 0 Views 941KB Size
Aiigiist, 1965

FLUORINATION OF NICKEL OXIDEBT CHLORINE TRIFLUORIDE

763

THE KINETICS OF FLUORINATION OF NICKEL OXIDE BY CHLORINE TRIFLUORIDEl BY R. LYNNFARRAR, JR.,AND HILTON A. SMITH Joint Conlribulion from the Department of Chemistry, University of Tennessee, and Carbide and Carbon Chemicals Company, Oak Ridge, Tennessee Received M a r c h 8, 1055

A study of the kinetics of fluorination of nickel oxide powders to nickel fluoride with gaseous chlorine trifluoride has been made over the temperature range 25 to B O " , and a mechanism has been proposed by which this conversion process takes place. Pressure variations of 0.3 to 1.0 atmosphere produced a slight effect on the kinetics of conversion which could be explained by the difference in surface concentration of reactant gas at the various pressures. The activity energy was determined to be 19.6 kcal./mole when corrected for decreasing surface concentration of reactant gas a t elevated temperatures. The conversion proceeded by a linear rate which meant that the fluorinated material did not form a continuous voating over the unreacted material, in spite of the fact that the molar volume of the product is larger by a factor of t w o than that of the nickel ouide. Evidence for this mechanism is supported by electron photomicrographs.

Introduction This problem was undertaken t o provide an insight into the mechanism of the reaction by which very fine nickel oxide powders were converted to nickel fluoride by gaseous chlorine trifluoride. A related study has been carried out by Valensi2 in which larger nickel powders (diameter 100 to 500 p ) were oxidized a t temperatures in excess of 800". The oxide film was reported to be impervious and the kinetics obeyed the parabolic law of solid-gas reactions. Another similar study has recently been reported by Dunoyer3who studied the reduction of molybdenum trioxide t o molybdenum dioxide with hydrogen. Experimental Nickel Oxide.-All nickel oxide powders were prepared by decomposing basic nickel carbonate of high surface area in air at 200 to 250'. This material was jet black in color with a nitrogen surface area of 150 sq. m. per gram. Further sintering4under vacuum at elevated temperatures gave the desired particle sizes (from 3 to 30 sq. m. per gram) for the various experiments. The color of the oxide powders ranged from black for the high area powders (sintering temperature ea. 400') through gray to light green for the low area powders (sintering temperatures ca. 1000"). There was an increase in nickel analysis of the powders with increasing sintering temperature. All samples showed X-ray powder patterns of the typical nickel oside structure. One is drawn to the conclusion that there is an excess of oxygen in nickel oxide, as other workers have reported5 and that the excess is reduced to smaller proportions as the sintering temperature is increased. Chlorine Trifluoride.-The chlorine trifluoride was purified from technical grade material manufactured by the Harshaw Chemical Company. Hydrogen fluoride was removed by sorption on sodium fluoride and distillation was used to separate chlorine and chlorine oxides. Only the middle fraction was used and it analyzed, by the method of fractional melting in a calorimeter,6 99.8 mole chlorine trifluoride. Pretreatment of the Powder.-Very early in the course of this study i t became apparent that unless the amount of chlorine trifluoride which was brought in contact with the initial nickel oxide surface was very carefully controlled, the powders would burn in the first few seconds of the reaction. The heat of this reaction is approximately 113 (1) This document is based on work performed for the Government by Union Carbide and Carbon Corporation at Oak Ridge, Tennessee. (2) G. Valensi, Compt. rend., 202, 309 (1936). (3) J. M. Dunoyer, J . chim. phys.. 47,290 (1950). (4) The term ainter as used here has the same meaning as that used in S. J. Gregg, "The Surface Chemistry of Solids," Reinhold Publ. Corp., New York, N. Y.,1951,p. 67. (5) J. S. Anderson, Ann. Rept., Chem. Soc., 49, 104 (1946). ( 8 ) J. W.Grieard, H. A. Bernhardt and G. D . Oliver, J . Am. Chem. Soc., 73, 5725 (1951).

kcal./mole of chlorine trifluoride, and because this heat cannot be dissipated there is no control of the temperature, amount of conversion, or surface area of the powder. Therefore the init.ia1 gas was diluted with nitrogen, so that the reaction was controlled by limiting the availability of chlorine trifluoride to the surface, and also by dissipating the heat of the reaction. After 2 to 3% of the particle (1000 1. average particle diameter) was converted, it became sufficiently stable so that 1 atmosphere of pure chlorine t.rifluoride did not cause t,he powder to burn. Apparatus for Kinetic Measurements .-This solid-gas r e a d o n differed from the usual type (such as the oxidation of metals) in that a gaseous roduct was evolved. Therefore it was necessary to circugte the gas in order to prevent a diffusion block by the gaseous products. Since the variation of reaction rate with pressure was small, the chlorine trifluoride pressure was maintained sufficiently constant by frequently evacuating the system and refilling with pure reactant whenever its partial pressurc was reduced t.o 90% of its initial value. The data obtained showed no discontinuities due to renewal of the reactant gas. The reactor was a small nickel vessel equipped with metal filter and valves. The vessel could be weighed on an analytical balance. This was loaded with about 5 g. of nickel oxide powder and placed in an all metal system. Facilities were provided in the system for introducing reactant gas, removing the gas phase, determining t.he pressure on the Booth-Cromer pressure transmitter' and recording the pressure automatically once each five minutes. The reactor, ballast volume and pump were enclosed in a box which could be thermostated a t any desired temperaThe circulating ture from 25 to 200' to better than 1 0 . 1 pump was similar t,o that described by Rosen.8 After the powder was pretreated as described previously, the reactor containing the sample was placed on the syatem and allowed to react under the desired conditions. At frequent time intervals the sample was removed, cooled, purged for 20 minutes with dry nitrogen, and weighed. By this method, the fraction converted was known to 1 0 . 0 5 % . Errors due to reaction with the walls of the reactor were negligible since its effective surface area was only about one per cent. of that of the sample.

.

Results The equations which are useful in discussing and handling the results are highly idealized but offer the best approximation to the true situation from the standpoint of utility and simplicity. The follo~vingassumptions are made. 1. The true area of the oxide powder is that given by the surface area method of Brunauer, Emmett and Teller (BET)gwith nitrogen gas as the adsorbent. (7) 8. Cromer, "The Electronic Pressure Transmitter and Self Balancing Relay," Atomic Energy Commission. declassified March 20, 1947, (MDDC-803). (8) F. D. Rosen, Rev. 8 c i . Inst., 24, 1061 (1953). (9) 9. Brunauer, P. H. Emmett and S. Teller, J . A m . Chem. Soc., 60, 309 (1938).

R. LYNNFARRAR, JR., AND HILTON A. SMITH

764

2. The fluoride film and the oxide powder have the same density as the corresponding bulk materials. 3. The film is uniform in thickness. 4. The particles are treated as spheres and hence r

=

3/Ad

thickness of the uniform fluoride film, x, is given by the expression x = rg - TO = ri[(l c F ) ' / 3 - (1 - FP"1 (5) Another useful quantity is the thickness of nickel oxide which has reacted, x', a much simpler expression, which is

+

(1)

where T is the radius, A is the specific surface area and d is the density of the solid. Consider a cross-section of a heterogeneous spherical particle containing a core of unreact,ed nickel oxide and completely surrounded by a layer of reaction product, nickel fluoride. One can see that as the reaction progresses the radius of the core of nickel oxide must decrease with time and the thickness of the fluoride layer must increase with time. The following quantities may be defined on this basis: ri = radius of initial particle of nickel oxide, obtained by eq. 1 ro = radius of unreacted oxide core a t any time t rg = radius of gross particle, oxide fluoride layer, at

+

any time t F = the fraction of the particle which has been converted to fluoride.

Vol. 59

x' = ri - ro = r i [ l - ( I

- I")'/s]

(6)

From these equations kinetic expressions for the correlation of the results are given below. The parabolic and linear laws have been used since they appear to be applicable to this work. The linear law states in effect that the amount of reaction which has previously occurred is of no consequence in the rate of propagation of the reaction. The controlling factor is the amount of surface which is exposed to the reactant gas. Accordingly (7)

where dN/dt is the number of moles of oxide reacting per particle per unit time and a. the instantaneous surface area of a single average oxide particle a t time t. Introducing the equations (8)

= 4mQ

The fraction converted is obtained experimentally by the equation

and

where

(no is the molar volume of nickel oxide) and integrating from F = 0 a t t = 0 to F = F a t t = t, one obtains

Si

Xt M F

weight of the initial nickel oxide sample wt. of the sample a t time t and Aso = mol. wt. of fluoride and oxide, respectively = =

The initial volume, vi, of a single average particle is

and after the reaction has proceeded, the volume of unreacted oxide, vo, is vo = (1

- F)vi

=

1 ~ r 0 3

then by combining these equations, one obtains terms of ri,

T O in

I'o

= r , ( l - F)'/s

(3)

To obtain rg in terms of r,, the volume of the gross particle, ug, is vg = vo

+

4 3

Z'F = -

7rry3

where uo has been defined and OF is the volume of nickel fluoride in the gross particle. But VF

=

(bi

-

v0)p

and

1

- (1 - F ) %

klQ0

= -t

Ti

=f(A)

(10)

If the linear law is obeyed, a plot of ! ( A ) us. t is a straight line and the rate of conversion of powders of different size is inversely proportional to the particle radius. The parabolic law is given by the equation where x is the film thickness, t is the time and I c p is the parabolic rate constant. Defining x, the thickness of the fluoride film, by equation 5 one obtains an equation whose integration is difficult. However, instead of the equation for x,one can use as an approximation the equation for the thickness of nickel oxide which has reacted, x', equation 6. This thickness of oxide is almost proportional to the thickness of fluoride which has been formed. Equation G is differentiated and introduced into equation 11. Then integrating from F = 0 st t = 0 to F = Fat t = t, one obtains

+ '/%(l-

F)*/3

- (1 - F)'/d = kri2 2.t

=f ( B )

(12)

If the parabolic law governs the reaction, a plot of f ( B ) us. t should be a straight line and the rates of where 6 = ( d o M ~ ) / ( M o d ~the ) , coefficient of ex- conversion of powders of different particle size pansion on conversion (do and d~ are density of should be inversely proportional to the square of the radius. oxide and fluoride, respectively) and The experimental data for three runs are prer, = ri[l - (1 - p ) F ] ' / 3 = r i ( l f c F ) ' / B (4) sented in Table I along with a summary of the condiFor t8he nickel oxide-nickel fluoride case under tions under which the runs were made. The time, consideration p = 2.083 or c = -1,083 and the t, is indicated in hours and the total fraction vg = V i [ l

- (1 - p ) ] F

SUMMARY OF DATA ox

THE

TABLE I CONVERSION OF NICKELOXIDETO NICKELFLUORIDE

Run no. Temp., "C. Pressure, atni. Til

A.

l/Fm Linear slope, f ( A ) / t Linear rate constant, k&, (+k./hr.) Parabolic slope, . f ( B ) / t Parabolic rate constant] kz, (A.Z/hr.)

1

0 17.0 20.1 33.6 38.8 54.8 74.4 79.6 95.9 137.1 151.9 178.4 199.0 222

765

FLUORINATION OF NICKELOXIDEBY CHLORINE TRIFLUORIDE

August, 1955

3 122

'/Z

1/3

1/3

515 0.873 0.00270 1.39 0.00122 323

620 0.806 0.000383 0.24 0.000136 52

205 0.826 0.00113 0.23 0.000455 19

FR

t

0.591 ,670 . 088 745 .779 ,840 ,911 .930 .986 1,099 1,123 1,134 1.140 1.145

0,510 ,585 ,601 ,658 .680 .733 ,795 .812 ,861 ,959 ,980 .990 ,995 1,000

0.0 3.0 109.7 197.4 318.5 418.4 528.4 690.7 808.8 896.9 1012 I1 1206.5 1463.2 1725.9 2217.0

4

0.296 ,300 ,402 ,492 ,607 ,704 .799 ,907 ,971 1,010 1.054 1.105 1.163 206 240

conversion, FT,has been calculated by equation 2. Inspection of the values of FT in this table shows that FT exceeds 1.0 before it reaches its limiting value. As will be seen later, the value FT contains a term Fs which is due to adsorbed and absorbed material, as well as the term FR due to chemical reaction. This may be written F R = FT

- Fs

125

Experimental data, t in hour8 FT FR Run 3

FT

Run 2

4

2 150

t

0,239 .212 .324 ,397 ,489 .507 , 644 .731 ,783 ,814 ,850 ,891 ,937 .972 999

0.0 17.8 21.7 39.4 00.5 84.6 129,I 176.5 222.8 245,7 298.9 340.3 385.9 450.3 500.0 038.7 783.4

FT

Run 4

0.180 ,357 ,384 ,543 ,523 ,592 ,606 899 ,885 ,926 1.003 1,047 1.088 1.135 1.161 1.191 1.206

FR

0.149 .29G ,318 ,374 .432 .488

,575 .660 ,731 .765 ,828 .865 .899 ,938 ,958 .984 ,997

good for the linear law, but from the parabolic plots variation by a factor of more than 2.5 is found. One is therefore led to the conclusion that the conversion proceeds by the linear rate of attack, a t least over this temperature range.

(13)

The quantity FRis the term which should be used to calculate f(A) and f(B). The term FR due to chemical reaction, can vary from 0, when only nickel oxide is present, to 1 when the limiting value is reached and oiily nickel fluoride is present. As a first approximation one can say that F s is directly proportional to FR,and FRis unity when FT equals its final maximum value, F,, which is obtained at complete conversion of the oxide to the fluoride. Experimental plots of FR vs. t were made and values of t taken at 10% intervals of conversion. Plots of f(A) us. and f ( B ) us. t are shown in Fig. 1. The slope of the best straight line portion is tabulated in Table I and used to calculate the rate constants for the linear and parabolic mechanism. On the basis of these plots, there is generally better agreement with the linear than the parabolic law. The parabolic plots generally have an upward curvature meaning that experimentally the conversion proceeds faster than the parabolic law would allow. The more sensitive test is shown in Table I where the rate constants have been calculated, for powders of different particle size (runs 3 and 4). The agreement between the rate constants for equivalent temperature runs is very

TIME, HOURS.

Fig. 1.-Typical

correlation plots.

The results a t lower temperatures, less complete because of the considerably longer time required for conversion, and those a t 180" are correlated by the linear law. The effect of pressure with other variables constant, is illustrated in Fig. 2. The influence of pressure will also be considered with that of temperature. The temperature variation was treated by the usual Arrhenius equation. Table I1 is a summary of the linear rate constants at the various temperatures] and the plot of log lcl!i& os. 1/T is shown in Fig. 3. There is definite curvature in this figure which will be discussed later. Discussion of Results The Temperature and Pressure Effect.-The curvature in Arrhenius plot (Fig. 3) is duc to two

A. SMITH R. LYNNFARRAR, JR., A.ND HILTON

766 1.4

VOl. 59

5 00

1.2

LO

b .a

IO0

q *- .6

0 50

.4

0

2-

.2

-

0

0.4 0.6 .OB PRESSURE, atmospheres.

0.2

0

ID

0 10 0 05

0.6

I

,6 1 0 01

a5

34

i

001

30 26 I I T X IO'

22

UNCORRECTED.

-4 -

0.4

Fig. 3.--hctivation

of chlorine trifluoride per unit area of surface. An attempt has been made to correct all rate constants t o some constant value of surface concentration by means of a proportionality constant, y , where

0.3

0.2

Y = 0.1

energy.

I 100

L

0

I 200 TIME,

Fig. 2.-The

J 300

hour$.

effect of pressure, 150".

effects: (1) the variation of surface concentration with temperature and (2) a change in mechanism of attack as the initial thin film becomes thicker. TABLE I1 EFFECT OF TEMPERATURE

26 26 Gl 65 98 100 122 125 148 148 148 148

33.4 0.90 0.020 0.74 0.015 33.4 ,019 0.74 ,014 .90 .024 29.9 .85 ,024 1.01 29.6 .85 ,024 1.01 ,024 26.9 .85 ,068 1.11 ,075 26.8 .058 1.11 ,064 .85 25.3 .30 .24 1.25 .30 25.1 .23 1.25 .29 .30 23.8 .08 .7G 1.G4 1.2 1.35 1.4 23.8 .40 1.0 1.2 0.72 1.64 23.8 .08 1.7 23.8 .80 1.3 1.30 150 1.8 23.6 .50 1.4 1.33 1.59 4.6 180 22.1 2.9 .50 A term to correct rate constants for the variation in surface concentration. See equation 14. (I

At the low temperatures, 26 and 65", the pressure was of the order of 0.8 atmosphere while at higher temperatures the pressure was lower. Adsorption experiments'o have shown a decrease in surface concentration with increasing temperature even though the surface coverage was in all cases approximately one monolayer. At the higher temperatures the monolayer contained fewer moles (10) R. L. Farrar, Jr., and H. A. Smith, J. A m . Chenz. Soc., 77, Sept. (1955).

GJGz

.

(14)

Here G1 = 20.0 mg. C1F8 per g. NiFz for a sample of surface area 30 sq.m./g. is the arbitrarily chosen value of the surface concentration to which all rate constants are corrected. Gz is the surface concentration at the particular temperature and pressure of the experiments, and has been taken from adsorption experiments. Effectively, bhe assumption is made that a t a given temperature the rate constant is directly proportional to the surface concentration. The logarithm of the new rate constant, 7c100u,is plotted against the reciprocal of the absolute temperature in Fig. 3. This correction tends to improve the curvature but does not remove it. The activation energy, determined on the straight line portion, is 19.G kcal./mole. Curvature in the low temperature portion of the plot is best explained by a change in mechanism. The last point or points on the graphs obtained with the linear rate constants do not follow the straight line. At this temperature, the rates of conversion, while already very slow and near limit of determination with the particular experimental set up, probably would continue to decrease with time giving a smaller value of the rate constant. In this range of conversion (the order of 1 to 3y0)at this temperabure, reaction is progressing by a parabolic mechanism with a relatively high rate of attack. The powder continues t o build up a film by the parabolic mechanism until it is thick enough for cracking and recrystallization to take place, whereupon the linear mechanism takes over. This, however, as evidenced by back extrapolation of the straight line in Fig. 3 to the temperature in question, would yield a rate constant so small that it would be undetectable in this experiment. Electron Photomicrographs, Electron Dsraction and X-Ray Diffraction.-Considerable information of a supportingnature has been obtained from the

August, 1956

7G7

FLUORINATION OF NICKELOXIDEBY CHLORINE TRIFLUORIDE

Fig. 4.--Elect,ron photomicrographs: upper left, nickel oxide, G.5 ni.2/g. (27,000 X); upper right, nickel Huoride from 6.5 m.?/g., nickel oxide (27,000 X); lower left, nickel oxide, 3.7 rn.z/g. (140,000 X); lower right, nickel oxide, (3.7 m.”g.) converted 10% to nickel fluoride (128,000 X ).

use of electron photomicrograph, and electron and X-ray diffraction patterns of the powders. Typical photomicrographs of the powders are presented for the initial and final material in Fig. 4. From this photograph a t 27,000X magnification the particle size distribution of the oxide powder and the fluoride powder have been obtained, and that for the fluoride powder is presented in Fig. 5 . The particle size distribution curves give the normal probability distribution curve when plotted on a logarithmic scale, and a badly skewed curve when plotted on a linear scale. This is probably due to the sintering process by which the particle size was governed. The particle size by nitrogen surface area for each of these powders is indicated for comparison. Good agreement is obtained for the oxides where no nitrogen surface area anomaly was observed, but radical disagreement is noted for the fluoride powder. This supports the contention that the oxide particles are single crystals while the fluoride particles are porous. The broadening of the lines of the X-ray and electron diffraction patterns is also in agreement with this idea.

The crystallit: sizes in the fluoride powder are in the 50 to 100 A. range. 25

2o

2

15

2 p

lo

5

100

1,000

10,000

AVERAGE PARTICLE DIAMETER, angstrom units.

Fig. 5.-Particle

size distribution of NiF,.

768

R. LYNNFARRAR, JR., AND HILTONA. SMITH

The high resolution photomicrographs presented in Fig. 4 show the details of some of the cracks and crevices. This constihtes definite proof of the structure of the particles which has been postulated to explain the observed experimental facts. Gaseous Products.-When gaseous chlorine trifluoride reacts with nickel oxide, there are two possibilities which might reasonably be expected, namely

VOl. 59

During the evacuation of one of the samples, which had gained about 20% more weight than simple stoichiometry would predict, the degassed material was collected a t liquid nitrogen temperature. Mass spectrographic analyses showed this material to be chlorine trifluoride.

Conclusions It is possible now to describe the mechanism by which chlorine trifluoride converts nickel oxide t o 3/2Ni0 ClF3 +3/2NiFz + 1/2C12 3/402 ( A ) NiO ClFs +NiFz CIF 1/~0~ the fluoride. It has been shown that an unreacted (B) and perhaps a secondary reaction of O2 Clz, niokel oxide surface is extremely unstable with reOZ CIF, or Oz ClF, to yield chlorine oxides or spect to chemical attack by the fluorinating agent, but only a very thin film of fluoride is sufficient to chlorine oxyfluorides. A reaction attempt was made using a large sample prevent an uncontrolled reaction from taking place. of nickel oxide (20 g.) with chlorine trifluorideat room This film of fluoride increases in thickness by a diffutemperature. The system which had been filled sion process as evidenced by initial obedience to with helium was displaced with chlorine trifluoride parabolic kinetics. After the fluoride film reaches some critical thickand all products and unreacted chlorine trifluoride passed through a trap a t liquid nitrogen tempera- ness, a recrystalliaation process takes place allowing ture. This material was then transferred to an a major change in the mechanism of attack. analytical distillation column and the amount of Whereas previously the film was more or less material in each fraction between -183" ( p = continuous, after recrystallization a mosaic net1 atm.) and 11O was determined. The percent- work of crystallites is formed which opens grain boundary paths allowing the adsorbed reactant age results are shown in Table 111. gas to move closer to the unreacted oxide surface. TABLE I11 Conversion from this point on is by linear kinetics CHLORINE TR~FLUORIDE-NICKEL OXIDEREACTION consisting of migration of chlorine trifluoride molecules down the grain boundary paths until they PRODUCTS Temp., Mole enter the transition zone where chemical reaction OC. Component % (or the electron transfer process) takes place. - 183 0 2 1 The rate controlling step now is diffusion through - 100 CIF 1 the transition zone of relatively constant thickness - 80 cos 4 rather than migration of the chlorine trifluoride - 48 COClF 4 molecules down the grain boundary paths. If the 88 - 35 Clz latter were rate controlling, the parabolic law a - 6 3 would have been followed. 100 The completely converted nickel fluoride cona The identity of this fraction is unknown but by infrared tained excess chlorine trifluoride which could be aiialysisit i R not ClzO, ClOzor Freon-114. removed by evacuation. It was concluded that The oxygen was not condensed in the cold trap this material was sorbed in the inter-crystallite since its partial pressure in the gas mixture was boundaries. considerably below its vapor pressure." The oxyAcknowledgments.-The authors wish to exgen which was observed was probably due to physi- press their appreciation to Drs. H. A. Bernhardt cally adsorbed material The important thing to and E. J. Barber of the Carbide and Carbon note is the Cl?/CIF ratio. These results would Chemicals Company for their valuabIe suggestions favor reaction (A) above. At no time have mass and assistance during the course of this work. spectrographic analyses or infrared analyses of The authors also wish to thank Mr. W. W. Harris reaction products indicated that chlorine mono- and his electron microscopy group for the electron fluoride was any major fraction of the products and photomicrographs, electron diffraction patterns and always large amounts of chlorine have been detect,ed. other photographic work associated with this research. ( 1 1 ) Vapor pressure of oxygen at -195' is approximately 150 i n i n .

+

+

+

+

+

+ +

+

+

b