1552
TAYLOR B. JOYNER
The Thermal Decomposition of Solid Hexaamminecobalt(111) Azide. Kinetics of the Cobalt Nitride Reaction
by Taylor B. Joyner Research Department, Naval Weapons Center, China Lake, California 95665
(Received March 14, 1069)
The isothermal rates of the decomposition of solid hexaamminecobalt(II1) azide to CONhave been measured in the temperature range of 95-150’. Samples ranging from microcrystalline powders to large single crystals all show induction periods. Both the final reaction and the induction periods in the absence of ammonia yield apparent activation energies of about 25 kcal/mol. Small quantities of ammonia prolong the induction period and give higher apparent activation energies. The final reaction is independent of ammonia. The induction period can be shortened by dusling the reactant with CON or freshly prepared metallic cobalt. These observations and the similar kinetic parameters found for substituted cobalt(II1) ammine aaides suggest a rate-controlling step involving an interaction of the ionic, lattice azide with the solid product.
Introduction Solid cobalt(II1) ammine azides can decompose to either cobalt(I1) complexes or CON. The rather complicated relationship of the two systems has been discussed qualiLatively for hexaamminecoba,lt(111) azide,l and the kinetics of the cobalt(I1) reaction have been analyzed.2 Quantitative studies of the substituted azidopentaamminecobalt (111)azide3 and cis- and trans-diazidotetraamminecobalt(II1) azides415revealed analogous reactions and indicated that topography is particularly important in the C o x decomposition. This reaction [Co(NH3)6](93)3
CON
+ 6XH3 + 4x2
(1)
has a well-defined induction period and a relatively fast, final reaction. Study of both periods permits some understanding and separation of the chemical and topographic aspects of the decomposition. This paper considers the former. These depend only on a quite general analysis and form the basis for the comparison with the substituted compounds upon which the subsequent discussion of mechanism will rest. The experiments of this paper were designed to avoid the cobalt(I1) reaction
It is, however, a potential competitor, and a full discussion should account for both it and the unusual form of the CONreaction. iiconsideration of reaction topography (second paper in series6) can explain the existence of an induction period, its relationship to the observable CONreaction, and the complex interplay of the cobalt(I1) and CON reactions. The model is applicable to all four compounds. A final paper’ discusses mechanisms for the series. T h e Journal of Physical Chemistry
Experimental Section Preparation. Preparation and purification have been described. The usual recrystallization from warm water gives lathlike crystals ranging in size from about 1 X 0.3 x 0.1 mm to 0.2 X 0.1 X 0.02 mm. The kinetic study made extensive use of two such samples labeled “Crystals(A)” and “Crystals(B).” The latter was used in probing, in particular, the induction period, and detailed presentation of the data will be deferred until the following paper.6 To study the effects of particle size, “Powder” with crystal sizes in the range of to mm was prepared by rapid precipitation of an aqueous solution with excess alcohol. (This particular sample was also the “Preparation €3” in the study of the cobalt(I1) reaction.2) Slower addition of the alcohol gave a sample, labeled LLSmall-crystals,” with crystals up to ca. 0.01 mm in length but also containing smaller particles comparable to “Powder.” Dark, room-temperature evaporation of an aqueous solution gave “Large-crystals” up to 13 mg in weight and roughly 3 X 3 X 1 mm in size. All identities were checked by powder patterm8 During preparation particular attention was given to avoiding conditions which might lead to contamination with the double Storage was over P205 salt [Co(l\’H3)6](T\T3)3’3XaN3.* in an opaque vacuum desiccator. Procedure. The apparatus and procedure have been d e s ~ r i b e d . ~ ~Again, S “nornial” runs began with the (1) T. B. Joyner and F. H. Verhoek, Inorg. Chem., 2 , 334 (1963). (2) T. B. Joyner, J.P h y s . Chem., 72,4386 (1968). (3) T. B. Joyner, ibid., 69, 1723 (1965). (4) T. B. Joyner, ibid.,71, 3431 (1967). (5) T. B. Joyner, ibid., 72,703 (1968). (6) T. B . Joyner, ibid., 74,1558 (1970). (7) J. B. Joyner, ibid., 74, 1563 (1970). (8) T. B. Joyner, D. S. Stewart, and L . A . Burkardt, Anal. Chem., 30, 194 (1958).
KINETICS OF
THE
COBALTSITRIDE REACTION
system initially evacuated (to ca. Torr) and the usual 10-mg sample exposed to only its own products. The useful “trapped-normal” runs had a liquid nitrogen trap present during the induction period and absent during the final reaction (the technique has been discussed4)so that a single run gave quantitative data on both periods. With recognition of the sensitivity of the induction period to ammonia, there was concern that the trap-even assisted by periodic venting-might not adequately remove ammonia from the solid. Thus “pumped-normal” runs exposed the solid to continuous pumping for the first 90 k 5% of the induction period. The cold trap was then maintained until the initiation of the fast reaction. The results resemble trappednormal runs, and the data are considered together in the calculations. A few runs were pumped continuously with progress crudely checked by intermittently isolating the trap and measuring the accumulated ammonia. The rather sparse data appeared to agree with trapped-normal runs but were inadequate for analysis. A large number of runs “trapped” throughout were made on four independently prepared samples during early stages of the study. These yield quantitative confirmation of the induction period and qualitative (due to slow diffusion of ammonia to the trap) support for the final reaction. In “catalyzed” runs the surface of the hexaammine was dusted with CON, A thin-stemmed, glass funnel was used to add fresh hexaammine to the residue from a previous run. The solids were very gently mixed taking care to avoid damaging the crystals. The system was evacuated, generally for 15 min, and the subsequent run commenced under trapped-normal conditions. Usually the catalyst was used the day of preparation or after standing overnight in the vacuum system. Two runs used metallic cobalt-a commercial powder and a sample freshly prepared by heating CoNg-as the catalyst. The reaction has been studied isothermally between 95 and 150”. At 150’ it is fast and warm-up time is a problem. Thermocouples in loose contact with the sample cup took about 100 sec to reach bath temperature. This is of no concern to the final reaction since even at 150” there is a 600-sec induction period. It is also insignificant for the induction periods at lower temperatures. It is, however, obviously of concern at the high temperatures; hence, those runs with induction periods under 1000 sec were omitted from the calculation of the apparent activation energies. The warm-up time was otherwise ignored. The tables report total, uncorrected times. Products. Some additions to the early reportg are necessary. The usual solid product is CON. However, a few runs (at temperatures ranging from 95 to 140”) yielded products with relatively faint lines in the powder patterns indicating the presence of metallic cobalt as well. Careful observation of the solid upon
1553 I .o
.6 U
.4
.2
0
IO0
50
0
I50
MINUTES
Figure 1. Representative reaction curves for the decomposition of [ Co(NHE)B](Na)3 samples of various particle sizes; temp 130’: curve A, Powder; curve B, Small-crystals; curve C, Crystals(A); curve D, Large-crystals. The solid symbols indicate the presence of a liquid nitrogen trap in this and the following figures.
first exposure to air also detected the pyrophoric reaction observed with the substituted compounds4but not previously reported for the hexaammine product. Again, it involved only a small portion of the solid.
Results
,
It is necessary to analyze both the induction period (when little of a directly observable nature is occurring) and the final reaction. These differ in their response to conditions. In particular, the former is very sensitive to ammonia while the latter is not. Particle-size effects and the competition between eq 1 and 2 are related to the ammonia dependence.’ (Briefly, powders are prone to eq 2, and careful removal of selfgenerated ammonia is necessary for observation of the CON reaction; larger crystals favor the CON reaction.) The qualitative explanation of the rather complicated behavior1 as amended3 is believed to be essentially correct and will be elaborated on in the following paper.‘j Figure 1 illustrates the reaction. The fraction reacted, a, is plotted against time. Closed symbols indicate the presence of the cold trap, open symbols its absence. The induction period is obvious. The final reaction curves are generally deceleratory. There is a slight tendency towards sigmoidal curves at the lowest temperatures and with Large-crystals. The Smallcrystals yield irregular curves probably because of the distribution of particle sizes. The duration of the induction period, 7 ,is arbitrarily defined as the time to reach cr = 0.1. It is quite reproducible (generally within 10%) for trappednormal runs. In untrapped runs the extreme sensitivity to ammonia requires very careful control of condi(9) T.B.Joyner and F.H.Verhoek, J. Amer. Chem. Soc., 83, 1069 (1961). Volume 74, Number 7 April 8 , 1970
1554
TAYLOR B. JOYNER 10
Table I : Induction Periods ( 7 ) and Rate Constants ( k , ) for Various Samples of [Co(?SH,)e](N& 8
Temp O C
ku Conditions
r,
sec
x
104,
8ec-1
6
150 150 150 140 140 140 140 130 130 130 130 130 130 120 120 110 100 95 95 150 150 140 130 120 110
Cxystals(A) Trapped-normal Trapped-normal Pumped-normal Trapped-normal Trapped-normal Pumped-normal Pumped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Pumped-normal Pumped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Pumped-normal Normal Normal Normal Normal Normal Normal
1,149 1,106 1,080 1,153 2,151 2,178 2,040 2 ,355 2,364 2,191 6,090 5,169 13,020 24,615 60,300" 60 ,000 598 616 1,388 4,094 14,117 65,586
140 130 120 110 100 95
Powder Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal
860 1,302 2 ,826 6,900 17,400 29 ,640
150 140 130 120 110 100
Small-crystals Trapped-normal 445 Trapped-normal 705 1,017 Trapped-normal Trapped-normal 2 , 160 Trapped-normal 8,028 Trapped-normal 19,680
150 150 140 130 130 120 120
Large- crys t als Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal
645 587 610
567 573 1,200 2 ,640 2,211 7 ,800 7 ,500
123 131 114
4
55.6 53.8 56.6 53.5 24.6 23.3 26.4 22.6 24.0 26.4 10.0 11.2 5.41 2.48 1.46" 1.48 104 107 52.6 25.4 12.2 5.26 Explosion 215 124 36.4 20.8 13.8 352* 172 153 118 45.0 9.60 28.2 33.7 13.3 4.08 7.82 1.68 1.45
'
a Composite of two runs. The curves tend to be irregular. I n general the constants represent only the last 50% of the decomposition. The 100 and 120' plots are particularly unsatisfactory.
tions to realize a similar reproducibility. Table I reports 7's. The final reactions of trapped-normal and trapped runs have similar durations despite the obvious difThe Journal of Physical Chemistry
a
2
0 0
IO0
50
I50
MINUTES
Figure 2. The effect of ammonia on the decomposition of [Co(NH&] (N,),. Crystals(A) a t 130": curve A, trapped-normal run; curve B, normal run; curve C, run in the presence of 50 Torr of added ammonia. Crystals(B) a t 120'; curve D, trapped-normal run; curve E, an initially trapped run interrupted (arrow) and completed in the presence of its own products and 63 Torr of added ammonia.
ferences in ammonia pressures. This suggests an independence of ammonia. Confirmation was provided by a trapped run interrupted a t a = 0.34 (Figure 2, curve E). During the 5-min interruption the trapwhich contained sufficient added ammonia to raise the pressure by 63 Torr-was warmed. The experiment was then completed as a normal run. The excess ammonia had no effect on the rate of the final reaction. I n contrast, ammonia seriously alters t8he induction period. With "Powder,' the self-generated ammonia suffices for a conversion to the cobalt(I1) system.'B2 With "Crystals" it prolongs the induction period prior to a typical, final CON reaction (Table I and Figure 2, curve B). The CON reaction can also occur-again with Crystals-under 50 Torr of added ammonia (Figure 2, curve C) with similar final rates but even longer induction periods. At higher pressures runs with crystals become too erratic for analysis with very long induction periods (during which cobalt(I1) complexes can be observed) and eventual partial explosions or pressure bursts.' The consistency of growth patterns at various temperatures was probed with split-temperature runs a t 110 and 130". (The extremes were avoided due to rapidity or inconvenient slowness.) A 110", trappednormal run was carried to a = 0.57 (Figure 3, curve c) and cooled to room temperature for 19 min (represented by C') while the bath was raised to 130". Upon reinitiation the run took up a rate (Figure 3, curve C'') appropriate for 130" (Figure 2, curve A) indicating that the early decomposition had not imposed a growth pattern which would be abnormal for a 130" run. In a similar experiment, the induction period required 9000 sec a t 110" and 446 sec a t 130". The times correspond to about 69 and 21% of the usual
KINETICS OF
THE
COBALT NITRIDE REACTION
1555
Table 11: Interrupted Runs with Crystals(A) -----Initial
decompositiona----kU X 104,
Conditions
T,
Trapped-normal' Normal' Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped-normal Trapped Trapped
sec
see-1
2151 4094 2417 2248 2075 2335 2200 2487 2447
24.6 25.4 21.4 25.0 26.5 22.7 24.0 27.4 22.7
...
...
...
..,
-Interrupted Time, see
ata
0.55
2700 2490 2280 2610 2520 2760 2790 1800d 2100d
0.56 0.55 0.56 0.61 0.55 0.59 N O ffO
-Interruption---Duration, min
Sample exposure
-Final
15 15 15 15 900 3780 900 15 930
Products Pumping Air Air Products Products Pumping Pumping Pumping
decomposition'----
x
ku crbsec
Conditionsb
0 0 1290 660 702 1500 990 714 835
Normal Normal Normal Trapped Normal Normal Normal Trapped Trapped
104,
sec-1
20.1 24.6 20.0 19.2 24.0 27.4 23.3 22.1 24.8
a All experiments were a t 130'. The second t is obtained under the indicated conditions. The final ku'sare for normal conditions. Check runs. Interrupted during the induction period.
0
0
so
100
200
100
300 0
25
50
150 A B 400 C 75 C"
MINUTES
Figure 3. Interrupted and split-temperature runs with crystals (A): curve A, 130' run interrupted and cooled for 15 min; curve B, 130" run interrupted and cooled for 63 hr; curve CC'C", split-temperature run commenced a t 110' (C), interrupted and cooled for 19 min (a period represented by C'), and completed a t 130' (C"). The time scales for C" and curves A and B are identical. The arrows indicate the points of interruption.
induction periods a t the respective temperatures with the total reasonably close to expectation. The final reaction was appropriate for 130". The absence of self-heating (a possibility in exothermic solid reactions) is indicated by the consistent analyses (following section) of both fast and very slow runs. As a further check, runs at 140 and 130" (Figure 3, curve A) were carried to cy = ca. 0.5, cooled for 15 min while the solid was either subjected to pumping or left in contact with the product gases, and reinitiated. In all cases the run took up the previous rate indicating an absence of excessive temperatures prior to the interruption. Exposure of the partially reacted solid to air or prolonged interruptions (with or without pumping) produced a second induction period. Figure 3, curve B
IO0
200
300CD
MINUTES
Figure 4. Comparison of runs catalyzed by dusting with CON (curves B and D) and conventional, trapped-normal runs (curves A and C ) ; temp: curves A and B, 140'; curves C and D, 110'; sample: Crystals (A).
is illustrative. Table I1 summarizes the experiments. The duration of the second induction period appears to be somewhat dependent on the length of the interruption. It is shorter (660 to 1500 sec) than the usual induction period under normal conditions (about 4100 sec). The two runs exposed to air indicate it can be shortened by trapping. Interruptions during the induction period seemed less serious although 930 min of pumping did cause some prolongation. The final reactions do not seem to be greatly altered by the various interruptions. These observations have not been pursued in detail. Since the interrupted sample consists of unreacted hexaammine (confirmed by powder patterns) and the solid products, it seemed worthwhile to determine the effect of CONon fresh material. These catalyzed runs were concurrent with the experiments of Table I; hence, the changes are not due to aging. The results are reported in Table I11 and illustrated by Figure 4. Dusting gives shortened induction periods, more sigmoidal and generally slower final reactions, and Volume 74, Number 7'
April ?231070
TAYLOR B. JOYNER
1556 irreproducibility in both r and the final reaction. The C O Sappears to maintain its catalytic ability after prolonged pumping or exposure to air although, perhaps, with some loss in efficiency. Shortening of T was also observed with freshly prepared metallic cobalt but not with the commercial powder.
3c
Table I11 : Catalyzed Runs” with Crystals( A) I
Temp, “C
150 I40 140 130 130 130 120 110 100 95
1C”
seo
x
104,
seo-1
7,
414 660 653 I , 113 1,231 1,050 2,100 3,768 8,460 15,000
64.1 29.3 31.2 20.0 17.1 18.8
130b 13OC
I , 466 I , 515
16.7 11.2
130d 130’
1,545 2,131
26.2 21.4
7.77 3.51 1.37 0.640
The runs vere made under trapped-normal conditions. The catalyst was C o x except as noted below. The CoK was pumped The CON was exposed to air for 63 hr on f o r 46 hr prior t o use. prior to use. Freshly prepared cobalt(0) was used as the catalyst. e Commercial cobalt(0) powder was used as the catalyst.
’
O D
20
20 30 0
tion period and the final reaction can be analyzed. The latter is well described by the usual first-order (or “unimolecular”) decay law -In (1 - a ) = k,t
+ e,
(3) where t is time and IC, and e, are constants.lOall Figure 5 is illustrative; Table I reports rate constants. The induction period gives little direct evidence, either visual or by gas evolution, of the processes taking place. Fortunately, its duration is useful. Plots of both T and k, against 1/T give good straight lines. The similarities in the apparent activation energies, E,, and preexponentials, A , are apparent in Table IV. The treatment of T as a measure of rate has been d i s c u ~ s e d . ~It may be noted that the onset of the final reaction is so rapid that alternate definitions of T in the approximate range 0.05 < (Y < 0.2 would not seriously alter the analysis. The final reactions of normal runs give similar results; the induction periods, however, have an apparent activation energy approaching the 46 kcal/mol of the cobalt(I1) reaction under low arnmonias2 The Crystals(B) data and the early exploratory studies generally agree with the above. Thus the The Journal of Physical Chemistry
I
25 25
30
40
50 IO0
50
J
30 0 A C
D
MINUTES
Figure 5 . Illustrative kinetic treatments for samples of various particle sizes; temp, 130’: curve A, Powder; curve B, Small-crystals (the poor fit is apparent); curve C, Cryst,als(A); curve D, Large-crystals.
Table IV : Apparent Activation Energies ( E a )and Preexponentials ( A )for Various Samples of [Co(NHa)B](Na)r TreatSample
Conditions
Crystals(A)
Trappednormal” Kormal
Crystals(A)
ment 7
k, 7
k, Crystals(A)
Discussion Aizalysis. The trapped-normal runs with “Crystals(A)” provide the most detailed data. Both the induc-
I
0
Crystals(B)d Crystals(B) Powder
Catalyzed, trappednormalc Trapped Normal Trappednormal
7
k, 7
k, 7
k,
E,, kcal/mol
log A , aec-1
26.3b 24.6 40.4 24.1 21.0 25.2
10.75b 10.86 18.25 10.44 8.31 10.85
24.1 20.6 26.3 23.9
9.68 8.80 11.18 11.28
All treatments are least a Includes pumped-normal runs. squares. Catalyzed runs were carried out under trappednormal conditions. Data for Crystals(B) will be more fully presented in the following paper.8
results are not peculiar to a single batch of crystals. Some variations are observed (Table IV). These reflect, it is thought, the importance of topography. (The model will suggest explanations. For now, the number and distribution of early reaction centers may influence growth patterns and hence the adequacy of the kinetic analysis and definition of 7.) Table IV (10) P. W. >I. Jacobs and F. C. Tompkins in W. E. Garner, “Chemistry of the Solid State,” Butterworth and Co., Ltd., London, 1956, Chapter 7. (11) Equation 3 should be considered as descriptively useful rather than as evidence of a decomposition pattern. The reaction curves can also be quite adequately handled by a shrinking-cube treatment to yield a similar temperature dependence. This is often the case. Ideal shrinking-cube and firsborder decay curves do not differ greatly. In practice, moreover, inherent features such as particle-size variations and the nucleation process can cause deviations from ideal behavior. It will be seen (following paper)6 that the proposed model is a modified shrinking-volume decomposition.
KINETICSOF
THE
COBALTNITRIDEREACTION
indicates an uncertainty greater than that suggested by any single series of experiments. For instance, trapped-normal runs with Crystals(A)-probably the best series-give Ea = 24.6 f 0.4 kcal/mol and log A = 10.75 f 0.20 sec-’, while the k,’s from catalyzed runs (discussed below) scatter rather badly but still yield E, = 25.2 + 0.9 kcal/mol and log A = 10.85 f 0.50 sec-’. Although one is inclined to weigh the Crystals(A) data rather heavily, Table IV would suggest a conservative E, = 24 += 3 kcal/mol as a generous margin of uncertainty. l 2 Catalyzed runs show considerable irreproducibility and a pronounced acceleratory period. Both are relevant to the model. The acceleratory period hinders the estimation of T and is, itself, too short and irreproducible for useful analysis. However, the deceleratory portion of the curve, about the last 75% is fit by eq 3 with the results discussed above. The slight decrease in Ea given by the induction periods (Table IV) is of doubtful significance in view of the difficulties with r . (The lower temperature d s , in particular, closely parallel the trapped-normal curve.) It thus seems likely that catalysis shortens the induction period but does not change the temperature dependence. The remaining samples support the Crystals results. Powder, with probably uniform particles, yields good agreement. Large-crystals show an expected dependence on the particular crystal, making detailed treatment of the rather limited data of little value. The 7’s and Ic,’s do approximately parallel the results from Crystals and Powder. The Small-crystals data scatter badly, probably due to the distribution of particle sizes. (The smallest crystals may commence the final reaction early, produce ammonia, and so influence the larger particles.) The results are of little quantitative value-showing, at best, only rough agreement with the other samples-but of some qualitative interest in indicating the borderline conditions for the C O S and cobalt(I1) reactions. Although the particles sizes are not greatly different, normal runs with Powder go to cobalt(II), while Small-crystals prefer the C O Sreaction.
Conclusions The results of immediate interest are summarized. (1) The similarity in the apparent activation energies for the induction period and final reaction suggests the same mechanism is responsible for both periods. (2) CON can shorten T indicating that the solid product is involved in the reaction mechanism of the induction
1557 period and, by (l),the final reaction as well. (3) The hexaamrriine reaction is quantitatively similar to those of the substituted compound^.^-^ To anticipate the following paper,e during the induction period the CON reaction is thought to grow a fixed distance on a microscopic scale prior to commencing the observable decomposition. Since the same reaction is responsible for both periods, similar activation energies are expected. Ammonia, added or self-generated, interferes with the microscopic CONreaction. The early growth is then by the cobalt(I1) reaction and the 7’5, as observed for normal runs, yield the apparent activation energy of that reaction. Apparently both CON and metallic cobalt can shorten r . Some uncertainty is caused by the possible presence of some cobalt in the CON catalyst and the conceivable survival of CON in the freshly prepared cobalt. The contamination, however, should be minor, and it thus seems likely that both CONand fresh cobalt can play a role in the reaction mechanism. (It is doubtful that a purely physical alternative-for instance, more efficient heat transfer to the reactantcould avoid altering the apparent activation energy.) Direct involvement of the product is also suggested by the interrupted runs. During a prolonged interruption intimate contact between the reactant and product may be ruptured by some slow process (crystal growth involving the agglomeration of COX from the reactant surface with the bulk of the product can be envisioned) so that upon reinitiation a second induction period is needed to reestablish the reactant-product interface. Presumably rupture is assisted by oxidation since 15min exposures to air produce the second induction period. An oxide coating might also explain the failure of the commercial cobalt to shorten the induction period. The quantitative similarities in the reactions of the series of cobalt(II1) ammine azides suggest a single mechanism. The compounds, particularly the cations, differ considerably but all have the common feature of at least one ionic azide. It seems reasonable to suggest a critical role for these azides and, specifically, to hypothesize a rate-controlling mechanism involving an interaction between the ionic azide and the solid product. This will be discussed in a summary paper.’ (12) I t may be noted that Table I V provides a very clear warning of the danger in attaching too much precision to results from a single batch of crystals when reaction patterns are not known with certainty.
Volume 7 4 , Number 7
April 0 , 1970
