4009
THERMAL DECOMPOSITION OF LiA1H4 This could very well imply that in the high-temperature range nitronium perchlorate decomposes by the evaporation of an ion pair followed by gas-phase decomposition. The sublimate from the decomposition of nitronium perchlorate under vacuum has been shown to be pure nitronium perchlorate.6 The possibility of an ion-pair evaporation decomposition mechanism under a helium atmosphere is therefore quite possible in view of the success of the analytical treatment of the data. The slight decrease observed in the final pressure for the high-temperature decompositions is due to a small amount of sublimation even under an inert atmosphere of helium. This again is to be expected for a mechanism based on an ion-pair evaporation process. It is proposed that the thermal decomposition of nitronium perchlorate under an inert atmosphere in-
volves three different chemical steps
+ '/~OZ NOzC104 + Clog + NO2 NO2 + + ClOz
NOzC104 +NOClOi 2NOC1O4 -.--t
NOzC104 -+-
0 2
(1) (2)
(3)
and that nitronium perchlorate undergoes a crystalphase change a t 156". The exact temperature of the phase change is obviously a function of the heating rate of the sample.
Acknowledgments. This work was supported by NASA under Jet Propulsion Laboratory Contract No. NAS7-562. We wish,to thank C. S. Gorzynski, Jr., and L. L. Rouch, Jr., for their valuable assistance in the experimental part of this study.
Thermal Decomposition of LiAIH, by M. McCarty, Jr., J. N. Maycock, and V. R. Pai Verneker Research Institute for Advanced Studies, Martin Marietta Corporation, Baltimore, Maryland (Received April 60,1008)
616.97
The thermal decomposition of LiAlH4 has been investigated using isothermal kinetics, differential thermal analysis (dta), and thermogravimetric analysis (tga). The dta confirmed that the decomposition occurs in four stages. The first stage accounted for only a small percentage of the evolved Hz and on occasion exhibited properties of an explosion. Contrary to prior work, the kinetics of the second stage were shown to follow a modified Prout-Tompkins relation. The isothermally determined activation energy of 23.8 kcal/mol for this stage was the same a t temperatures both above and below a phase change seen in the dta; however, the preexponential factors differed in the two regions. The third stage, corresponding to the stoichiometric equation LiAIHz -+ LiH A1 l/zHz,was found to follow first-order kinetics, with a n activation energy of 46.9 kcal/mol.
+ +
Introduction The thermal decomposition of LiAlH4 has been the subject of three prior investigation^.'-^ From the work of Garner and Haycock' who studied the lowtemperature isothermal decomposition and the diff erential scanning colorimetry work of Block and Gray2 and the differential thermal analysis work of Mikheyeva, Selivokhina, and Kryukova,8 the dehydrogenation is postulated to take place in four stages following the stoichiometric equations LiA1H4 (surface)
-+-
LiAlH4 -+- LiAlH2
Hz
+?
+ Hz
+ A1 + '/%HZ LiH +Li + l/zHz
LiAIHz -+- LiH
The first reaction accounts for only a few per cent of the total evolved Hz, and it has been suggested that it results from either a surface reaction' or impurities.2 The empirical basis of eq 2-4 rests with the observation that reaction 2 liberates roughly 50% of the total hydrogen and that reactions 3 and 4 each liberate about 25% of the available hydrogen. This work reports on a reexamination of the isothermal kinetics of reactions 1 and 2 and on an examination of the isothermal kinetics of reaction 3 and the kinetics of the entire system using simultaneous dif-
(1)
(2) (3)
(4)
(1) W.E. Garner and E. W. Haycock, Proc. Roy. Soc., A211, 336 (1952). (2) J. Block and A. P. Gray, Inorg. Chem., 4, 304 (1966). (3) V. I. Mikheyeva, M. S. Selivokhina, and 0. N. Kryukova, Dokl. Akad. Nauk SSSR, 109, 641 (1966). Volume 76,Number 18 November 1008
4010
At. MCCARTY, JR.,J. K. MAYCOCK, AND V. R. PAIVERNEKER
ferential thermal analysis (dta) and thermal gravimetric analysis (tga). Evidence will be given which shows that several of the conclusions reached in prior investigations are invalid.
Experimental Section Commercial LiA1H4 (K & K Laboratories) used was purified following a technique suggested by Davis, Mason, and Stegeman.4 A diethyl ether solution of LiAIH4 was filtered through either a "fine" fritted glass filter or a Seitz filter. Benzene was added to the filtrate, resulting in the slow precipitation of LiAlH4 crystals. The benzene and ether used had been purified by distillation of reagent grade materials to which had been added some LiAlH4. The crystals were heated in vucuo at about 70" for several hours, The LiAlH4 crystals were needle shaped and were about equally divided into Tyler screen scale mesh sizes 100-200 and 200-325. LiA1H4 was stored in a weighing bottle in a drybox under a Nz atmosphere. The isothermal decomposition studies were carried out in a conventional high-vacuum system with an effective volume of about 6.51. Samples of LiA1H4 of 5-20 mg were evacuated to ca. 10-5 torr and were introduced by the use of a rotatable l'spoon'' into a tube which was maintained a t the desired decomposition temperature. During a kinetics run, the pressure rise was monitored continuously by a liquid nitrogen trapped NRC Equipment Corp. Alphatron gauge whose output drove a Honeywell Electronic 19 strip chart recorder. The dta-tga work was done on a Mettler vacuum recording thermoanalyzer. Platinum crucibles (diameter of 3 mm and height of 4 mm) were used to hold the 5-mg samples which had been vibrated. In the dta the platinum cup which held the sample crucible served as one junction of the differential thermocouple; the other junction was the platinum cup which held the crucible with reference material. The oven temperature was measured at a platinum ring which encircled the crucibles. All runs were made in helium which was flowing past the sample a t 10 l./hr. With a standard drying line (HzS04, KOH pellets, and P&) it was found that the samples increased in weight during a run, presumably because of the presence of oxygen in the helium. This problem was eliminated by passing the helium through a tube filled with either copper turnings at 440" or powdered titanium at 800" and then through a liquid nitrogen trap. On the tga runs, vibrated 25-mg samples in a platinum crucible (diameter of 8 mm and height of 19 mm) were used. Debye-Schemer powder X-ray diffractograms of LiAlH, were taken. The over-all pattern was similar to that reported by Amendola15except that a number of lines were missing. The missing lines are explainable in terms of the preferred orientation taken by the elongated LiA1H4crystals in the narrow quartz capillary The Journal of Physical Chemistry
I
I
\
\
I
Figure 1. A dta trace of a 5 mg of LiAIHl sample and a superimposed tga trace of a 25-mg sample. The heating rate was Bo/min.
holder. A series of very weak lines was also found. These lines correlated very well with the spacings found for formed from the bombardment of A1 with 12.5MeV deuterom6 I n one instance when a mixed diethyl ether-tetrahydrofuran solvent was used in the purification procedure, well-formed cubic crystals were produced. These crystals were not used in any kinetics investigations.
Results Dtu and Tga. Simultaneous dta-tga traces were made using heating rates in the range 25-2"Imin. I n these runs the tga traces were only of qualitative value, since the total weight loss was only 10% and the samples were limited to 5 mg in order to prevent the sample from flowing out of the crucible (see below). Several tga runs were made using 25-mg samples. Figure 1 is a composite of a dta trace from a dta-tga run and a tga trace from a tga run. The time-temperature traces in the two runs were virtually identical. The results of this work qualitatively agree with the differential scanning calorimetry results of Block and Gray2 (see Table I) who found the same five peaks. I n an earlier the exotherms (designated as I and I11 in Figure 1) were not observed. I n several of the runs there was evidence of an exotherm between peaks I and 11. This new peak most likely is related to the low-temperature part of reaction 2 and is generally obscured by the large endotherm 11. It was found that, to within experimental error, the weight losses associated with the dta peaks IV and V corresponded to the loss of one hydrogen atom per LiAlH4 molecule initially present. The weight loss in the region of peaks 1-111 was about 45% greater than the weight loss predicted on the basis of the loss of two hydrogen atoms per LiA1H4 molecule. To determine the source of the excess mass loss, an isothermal run at 143" was made under conditions where the evolved (4) W. D. Davis, L. S. Mason, and G. Stegeman, J . Amer. Chem. SOC.,71, 2778 (1949). (6) A. Amendola, Index (Inorganic) to the Powder Diffraction File 1965, American Society for Testing and Materials, Philadelphia, Pa., 1965, p 469, card 12-473. (6) A. Appel and J. P. Frankel, J . Chem. Phys., 42, 3984 (1965).
4011
THERMAL DECOMPOSITION OF LiAlH4
Table I Heating
rate,
Peak
1 2 3 4 5 a
Thermicity
Exo Endo Exo Endo Endo Endo Endo
--Range
Ref 2
of transition, OC-? This work
148-158 160-177 187-218
160-165 170-185 198-217 236-272
228-282 368-409 370483
degj
min
10 10 10 14 20 15 20
Activation energy,a kcal/ mol
~ 2 4 m42
Calculated according to M. E. Kissinger, Anal. Chem., 29,
1702 (1957).
gases could be sampled mass spectrometrically. The mass peak, 29, which was gated on initially showed a growth during the initial stages of the decomposition. The growth of mass peak 29 under the experimental conditions showed that at least one component of the decomposition products readily forms CzHS+ in the mass spectrometer. Later a scan from mass 32 through 80 was made. Besides the background peaks, the strongest peaks correlated with the expected fragmentation pattern of diethyl ether. The ether in the sample most likely existed as a stable etherate. Peak I1 was found to be only partially reversible, in that repeated scans of the peak on a single sample resulted in progressively smaller peaks. This partial reversibility has been observed in other systems when samples simultaneously undergo decomposition and a phase change. Associated with the exotherm was an apparent partial melting of the crystals. This melting was also observed in a hot-stage microscope. However, when the heating rate was sufficiently low so that the temperature was well defined, the melting was not observed, since decomposition of the sample had taken place to too great an extent. I n the experiment on thermal cycling, the minimum in the endotherms occurred a t about 179" and the maximum in the exotherms occurred at about 163" at heating rates of 25"/min. The first exotherm, peak I, like the initial small reaction attributed to reaction 1 in the isothermal studies was found to be irreproducible. The ratio peak height: width a t half height was found to have considerable variation from sample to sample at fixed heating rates. This ratio has been found to be a useful empirical test for detonations in work on Pb(Na)i.7 I n LiA1H4 this ratio, on several occasions, was large enough to indicate that a small explosion may have occurred. Further support for the hypothesis that explosi'ons can occur comes from the observation of what must have been an explosion during an isothermal run. The material used here was a single lump of LiA1H4 purified by an ether evaporation technique. During the first 93 sec
of the run, the pressure rise was typical of other runs; at 93 sec the sample appeared to explode. Isothermal Kinetics. The isothermal decomposition of LiAlH4 in the range 127-159' gave sigmoidal pressure us. time plots once the generally small and irreproducible contributions from reaction 1 were subtracted out. Garner and Haycock' analyzed their analogous data by plotting (P - C)'I8us. t, where C was chosen t o give the best straight-line plots. Over the same limited ranges in the fractional decomposition where the plots of Garner and Haycock were linear, the data of this work were also linear. However, it was found that range of linearity could be improved on a given run if an exponent in the range 0.4-0.2 were employed rather than 0.33. A more satisfactory basis for the analysis of the data was found to be based on the mechanism LiAIH4
+ D -%-
2D
+ (E) + Hz
(5)
Integration of the rate equation implied by eq 5 yields (t
- &?)(A0+ Do)k'
=
In [(AD - A kn(t
- to)
=
In (Pm/'C)
+ Do)Ao/ADoI (sa)
+ In [ ( P + C)/(Pm - P ) ]
(6b)
where the 0 subscript indicates initial values, A represents the LiA1H4concentration, A . - A is proportional to the pressure, P, of the evolved H,. A discussion of the identity of species E and D will be deferred to the Discussion section. Equation 6b depends on the fact that A,, which is just the concentration of pure LiA1H4, is a constant independent of the sample weight and that for all runs it was found that A. >> Do, thereby permitting the approximation k'(Ao DO)= ~ ' A =o k ~ . I n Figure 2 the superiority of eq 6b over the (P - C)'la vs. t mechanism is clear. Since the pressure during a single experiment was never observed completely to level off and since the values of P , which gave the best agreement with eq 6b were as much as 15% less than the highest observed pressure, it appears that reaction 3 is the main contributor to the Hz evolution rate at large times. Figure 3 shows an Arrhenius plot of k in the temperature range 127-159'. The values of k were determined from the pressure-time data using eq 6b, with values of P , or C chosen to give the best fit to the equation. The magnitudes of the rate constants so determined were not strong functions of either P , or C. From a leastsquares analysis of the data, the activation energy was determined to be 23.8 kcal/mol, with 75?4 confidence limits of *0.5 kcal/mol. The values of C / P , for the data presented were in the range of (4-2) X In an earlier set of experiments the, C / P , ratio appeared
+
(7) V. R. Pai Verneker and J. N. Maycock, Anal. Chew., 40, 1325 (1968).
Volume 78, Number 18 November 1968
A I . MCCARTY, JR.,J. N. MAYCOCK, AND V. R. PAIVERNEKER
4012
TIME ( m i n )
Figure 2. Line B is a plot of the pressure us. time for a single run according to the relation used by Garner and Haycock to analyze their data. Line A applies to the same run plotted according to eq 6b.
Figure 3. An Arrhenius plot of
~ I in I
the range 127-159",
composition. In one run, a mirror was formed on the surface of the sample tube. Water and dilute hydrochloric acid plus shaking removed the rest of the decomposition products but did not disturb the mirror. The mirror, which did rapidly dissolve in concentrated sodium hydroxide, was almost certainly aluminum. The amount of aluminum necessary to form the mirror was less than 10 mg, assuming a minimum thickness of 100 monolayers. Two runs with initial over pressures of hydrogen to the extent of 2.4 and 10 times the hydrogen generated by the sample were made at 154". There was no discernable effect of the hydrogen on the rates, indicating that a t the temperatures and pressures employed the reverse of reaction 2 was unimportant. Experiments performed in the temperature range 170-186" exhibited irreproducibility and pressure-time curves which did not prove to be amenable to analysis. 8ince there appears to be a phase change which occurs in this temperature range, lack of reproducibility more than likely results from contributions to the H2 evolution rate from LiAlH4 in two different phases. To circumvent this problem and to investigate reaction 3, a series of decompositions was carried out in the temperature range 217-247". I n this range, the pressure us. time curves were similar in appearance to the lowtemperature curves; however, a plot of the reaction rate us. time clearly showed a break at pressures near two-thirds of the highest observed pressures. In analyzing these runs, it was found that the contribution from reactions 2 and 3 could be separated. Over the first half of the reaction (ie., P < one half the final pressure) eq 6b was found applicable. Since at longer times, the pressure did level off to a value, Pt, it was possible to determine the rate constant for the initial reaction by using the relation based on eq 6b ~ I = I
t o be a monotonic increasing function of time changing from about 5 X to 2 X loe2 over a 3-week period. The activation energy for this set of experiments was the same as for the set reported here; however, the preexponential factor differed slightly. The strongest set of lines visible in a Debye-Scherrer X-ray pattern of material which had been decomposed at 147.6" was unequivocally due to metallic A1.8 Because of the close similarity of the patterns of LiHg and Al, it was not possible to determine whether LiH was also present. A weaker set of lines which did not correlate with any of the patterns in the ASTM Index was also present. The dominance of the A1 pattern does not indicate that A1 is the dominant species present. The weakness of the other lines could result either from low scattering factors or polycrystallinity. It was not clear whether the A1 was formed in part by photochemical action of the X-rays. Some A1 is known to be formed during the low-temperature deThe Journal of Physical Chemistry
In [3/(t1/, - tl/Jl
(7)
where trin is the time for the pressure to reach 2Pr/3n. An Arrhenius plot for the k's so determined is shown in Figure 4. The line drawn in Figure 4 corresponds to the same activation energy found at lower temperatures. Except for three points the quality of the fit of the line to the points is reasonably good, indicating that the activation energy for the higher temperature process is close to that for the lower temperature process. However, the preexponential factor in the range 217-247" is about 7.4 times less than at the value below 160". On the basis of the granular nature of the decomposed material and visual observations of samples during dta-tga runs, the bulk of this reaction is felt to have taken place when the sample was not fluid. I n the last quarter of the reaction, it was found that a plot of the logarithum of the rate of Hz generation US. (8) Reference 5,p 409, card 4-0787. (9) Reference 5,p 471, card 9-340.
4013
THERMAL DECOMPOSITION OF LiA1H4 1.01,
--30 2'01
I
1.90 1 0 3 1 ~ PK)
\ * I
I 1.94
1
1.98
'1 2.02
'
1 2 ' .
2.06
>
1 0 3 1 ~( O K )
Figure 4. An Arrhenius plot of ~ I inI the range The line drawn through the data has the same slope as the line in Figure 3.
Figure 5. The Arrhenius plot of ~ I I in I the range 217-247'. The line drawn through the data was determined by least-squares analysis.
time was linear, indicating that reaction 3 obeys firstorder kinetics. The corresponding rate constant, ka, had an activation energy of 46.9 i 2.7 kcal/mol, the indicated uncertainty being the 75% confidence level. , 5, was The line given in the Arrhenius plot of ~ I I IFigure determined by least squares.
detonation quality exhibited on occasion by the first exotherm, the mechanism suggested by Block and Gray appears suspect. Most likely the source of the first exotherm is a small amount of material which is either introduced by the purification techniques or was present in the impure samples and not removed by the purification techniques employed. The identity of the material could be either an unstable organometallic compound or a species similar to one which can cause an explosion when ether solutions of LiA1H4which have been exposed to COz are evaporated.'O It would be tempting to attribute the first endotherm to the melting of LiAlH4, which by this time is a rather dark syrupy, bubbly single mass. At heating rates of 2"/min or less, no sign of the endotherm or melting was evident. The observations can equally well be explained by either the melting of LiA1H4 in the presence of its partial decomposition products or the melting of a reactive impurity present in small concentrations, e.g., lithium metal which in the free state melts at 186" and which would most likely undergo the exothermic reaction
217-247'.
Discussion The results of the dta-tga are consistent with the stoichiometric equation proposed by Block and Gray2 on the basis of their differential scanning calorimetry work which was coupled with monitoring the evolved H,. There are some questions as to the processes which result in the lowest temperature exotherm and endotherm. Block and Gray noted that the quantity of evolved Hz associated with the first exotherm increased with atmospheric exposure and indirectly inferred that the gas was indeed HOand not ether (elemental analysis of their LiA1H4 indicated a minimum purity of 98% LiAlH4 with less than 0.5% diethyl ether). They concluded that the exotherm was associated with sample impurity and suggested the possibility of reactions of the type >AlOH
+ HA1< +>AlOAl< + Hz
(8)
as the source of the exotherm. They also noted that the heat evolved per gram of evolved Hz was one order of magnitude greater for this exotherm than for the exotherm associated with reaction 3. The results of our work tend to confirm the experimental results of Block and Gray. There is little question that the initial Hz evolution observed in isothermal runs results from the mechanism which gives rise to the first exotherm. No direct attempt was made to check for a correlation between sample age or atmospheric exposure with the amount of initially evolved Hz; however, with the available data no clear-cut correlation was evident. However, in the light of the
Li
+ LiAIH4 * 2LiH + A1 3- */,HZ
(9)
The presence of free lithium could result from the reaction responsible for the first exotherm and could contribute to the observed formation of A1 in the lowtemperature reaction and could explain the observation of etching of the Pyrex sample tube in isothermal runs a t temperatures in excess of 170". The etching, however, could also have been done by LiAIOz. Following Kissinger,l' the activation energy for dta peaks IV and V (corresponding to reactions 3 and 4) were calculated from the slope of a plot of In (cp/Tm2) (10) G. Barbaras, G. D. Barbaras, A. E. Finholt, and H. I. Sohlesinger, J . Amer. Chem. SOC.,70, 877 (1948). (11) M. E. Kissinger, Anal. Chem., 2 9 , 1702 (1967).
Volume 72, Number 12 November 1968
4014
M. MCCARTY, JR.,J. N. MAYCOCK, AND V. R. PAIVERNEKER
l/Tm, where cp is the heating rate and T, is the temperature where the dta trace has a maximum or minimum. Kissinger's treatment assumes a reaction of the form (rate) = (1 - x)", where x is the fraction decomposed. Since reaction 3 was found to be first order ( i e . , n = l), the theory should apply to this reaction. However, the activation energy found from isothermal runs was 46.9 kcal/mol, whereas the value determined from dta was only 24 kcal/mol. The reason for the marked discrepancy is not clear; however, it does indicate that care should be exercised in using activation energies found from dta. I n the low-temperature decomposition of LiA1H4 (reaction 2), Garner and Haycock correlated their pressure-time data according to the equation P - C = k8(t - to)a,where to is the time at the end of an induction period and C is an empirically determined pressure which was always greater than the observed value of the pressure at to and was chosen to give the best fit of the data to their equation. They interpreted the cubic pressure-time relation in terms of a mechanism which required that at to a number of nuclei had been formed which did not grow in number with time but increased in mass as t3. They showed that the pressure-time curves did not follow the Prout-Tompkins relation,12 since plots of In [ P / ( P ,- P ) ] us. t gave curved lines. I n this work it has been shown that a modified ProutTompkins relation, eq 6b, fits the pressure-time data over a wider range in the fractional decomposition than the cubic expression, failing only in the region where reaction 1 dominates and where reaction 3 begins to interfere. The modified Prout-Tompkins expression was deus.
The Journal of Physical Chemistry
rived from reactioq 5. If it is presumed that this is indeed the rate-determining step in the decomposition, it becomes necessary to identify the species D and E; however, on the basis of the available information, it is not possible to do this unequivocally. It is not difficult to suggest possible models; for example, the reaction LiA1H4
+ LiAlH2 +2LiAlHz + Hz
(10)
would be consistent with the observed kinetics and stoichiometry. Because of the ease with which bridge hydrogen bonds can form, it is doubtful whether LiAIHa could exist in a crystal as an identifiable entity. Under reaction conditions, the hydrogen atom (ions) can be expected to be quite mobile so that the LiAIHz species could rapidly redistribute throughout the crystal. Further work using different techniques is necessary before the correct mechanism can be determined. The compound with the stoichiometric formula LiAIHz resulting from the initial decomposition of LiA1H4 has been found to decompose much more slowly than its parent compound. I n view of the greater thermal stability of LiAIHz over LiAlH4, an investigation of the chemistry of LiAlHz is indicated, since it may prove to be a useful reducing agent at temperatures where LiA1H4 decomposes too rapidly to be of use. Acknowledgnaents. We wish to acknowledge the assistance of C. S. Goraynski, Jr., and L. Rouch in the experimental work and of B. Huntington in the X-ray analysis. (12) E. G. Prout and F. C. Tompkins, Trans. Faraday Sac., 40, 488 (1944).
