DISTRIBUTIOK OF SURFACE ER'ERGY IS CHEMISORPTION
March, 1963
545
TABLE V COMPARISON OF EXPERIMENTAL A N D CALCULATED TERXARY DIFFUSION COEFFICIENTS Concn.
Dija
c1
MM
Exptl.
I . 039 0.032 - .023 .875
MB BM BB c2
MM MB
1.040
1,069 0.087 - ,029
-
,025 ,874
BB
,721
Mhl MB BM BB
,765 ,027 ,039 ,621
,794 ,042 ,051 ,628
-
-
&/IM
,909 ,030 ,009
1.505 0.211 -0,004
MB B R/I BB
Calcd. eq. 8
0,084
.920 ,042 - ,011 ,730
BM c3
Calcd. eq. 7
-
.887 .944 .044
-
-
1.550 0.199 -0.011
,012 .743 ,823 ,044 ,057 ,039
1.584 0.207 -0.012 1.133
1.383 1.414 A11 'iralues of Dij are in cm.2/sec. ( X lo6).
the binary diffusion coefficients may be taken as linear in w D i i = a)piowi
+
a)ijowj
+
a)ip0wp,
i,j = M,B (sa) i # j
These equations wereused with thevalues of the dilute binary diffusion coefficients in Table I1 to compute the ternary diffusion coefficients a t the interior points of the concentration field. The results are given in Table IT.The differences between the computed and measured values are in all cases within experimental error. It is interesting to note that the diffusion coefficients in this constant mass density system are also linear functions of volume fraction. I n the toluene-chlorobenzene-bromobenzene system the main diffusion coefficients were reported as ljnear functions of mole fraction,2v12but since this was a constant molar density system these coefficients are also linear in volume fraction. Thus, both systems are described by eq. sal3 if w is taken as volume fraction. This gratifying behavior which allows for estimation of the ternary coefficients from the dilute binary coefficients alone is not to be expected in less thermodynamically ideal systems. Acknowledgment.-The authors are grateful to the National Science Foundation for its financial support of this mark. (12) The precision of measurement of the cross diffusion coefficients was not great enough to determine the concentration dependence of the cross diffusion coefficients in that system. (13) The T-C-B system is described by a degenerate form of eq. 8a for in this system Pij"G%j0 so there is only "one" main diffusion coefficient.
RIICROC14LORIMETRIC STUDIES OF THE DISTRIBUTION OF SURFACE ENERGY IN CHEMISORPTION BY
Tr.
KEVOIZKIAN AKD
R. 0. STEINER
Esso Research & Engineering Co., Process Research Division, Linden, N . J . Received July 6, 1962 Surface energy distributions of a series of oxide catalysts were measured using NHs adsorbate in a differential isothermal microcalorimeter a t 50'. The catalysts studied were A1203 calcined a t 1100 and 1600'F., SiOz, KOH-A12O3, PtrAlaOa, and SiOz-MgO. It was found that A1203 (1100'F.) has a heterogeneous distribution of surface energy which is not destroyed by high temperature calcination. Si02, on the other hand, has a fairly homogeneous surface energy distribhtion. Alumina's energy distribution may be altered by impregnation, e.g., with KOH or Pt. When a co-gel of MgO and SiOz is formed, its surface is more energetic and its surface energy distribution more heterogeneous than that of Si02 alone.
Adsorption on a catalyst surface occurs on so-called adsorption sites. These sites vary in affinity for a given adsorbate, and the surface of the adsorbent is therefore said to be heterogeneous. When gas molecules are adsorbed on a clean catalyst surface, the highest energy sites tend to be covered first, the lower energy sites last. Consequently, the heat of interaction released generally falls off as successive doses of gas are adsorbed. hdicrocalorimetry is one of the best methods known for measuring accurately the differential heats of adsorption and so of characterizing a catalyst by the energy distribution of its surface sites. The major portion, by far, of microcalorimetric investigations has been devoted to physical adsorption (e.y., ref. l), and to chemisorption on metals (e.g., ref. 2 ) . (1) R. A. Beebe, B. Millard, and J. Cynarski, J . A m . Chem. Soc., 75, 839 (1953). (2) R. A. Fisher, Jr., "Adsorption Properties of Hydrogen and Oxygen on Platinum Black and Carbon Supported Platinum from 20 to 300 Degrees Kelvin," Ph.D. Thesis, Department of Chemistry, The Pennsylvania State University. 1961.
This study, therefore, turned to the important area of supported metal catalysts and common catalyst supports. It is concerned with microcalorimetric studies of the chemisorption of ammonia on various catalyst surfaces. Experimental To measure differential heats of adsorption, or surface energy distributions, a small amount of catalyst, between I and 2 g., was loaded into a calorimeter which then was inserted into a high vacuum system. A thermocouple was attached to the calorimeter to nieastire the heat release, for adsorption is always exothermic. The calorimeter and thermocouple reference junctions were contained in the same constant temperature bath. The catalyst was outgassed overnight a t a temperature of 450" until a pressure of 5 X 10-6 mm. wm achieved a t 450". The calorimeter then was cooled and immersed in the constant temperature bath a t 50' under a pressure of 10-6 mm. About 0.7 mm. of helium was added to the calorimeter to facilitate the heat measurements. A small measured dose of NHI was adsorbed on the catalyst, liberating about 1 cal. of heat which generated an e.m.f. of several microvolts in the thermocouple (Pt/Pt, 13% Rh). This signal was amplified and sent to a recorder. The
V, KEVORKI.~". ASD R. 0 . STEIKER
546 Heater Wire
1
--+ A 4
Thermocouule
41 4
55 mm.
4-
Tray
Fig. 1.-Microcalorimeter.
I
25
5oOc.I
t
51-
OL 0
4
I
8
I I I I I 12 16 20 24 28 % Of S u r f a c e Covered
I
32
I
36
40
Fig. 2.-Surface energy distribution of alumina (1100°F.). recorder pen followed a straight vertical line when the catalyst and thermocouple reference junction were a t the same temperature. The heat of adsorption caused the pen t o be deflected sharply away from this equilibrium position. As the catalyst slowly lost the adsorption heat t o its surroundings, a cooling curve was traced out by the recorder pen. After thermal equilibrium was re-established, a calibration was made by sending a known amount of electrical energy through a heater wire in the calorimeter, generatin.z another heatipg-cooling curve. The area under the calibration curve was measured with a planimeter and compared to that under the adsorption curve to determine quantitatively the adsorption heat release. The amount of NH, which adsorbed, and caused this heat release, was determined by measurement of the equilibrium pressure, This process then was repeated numerous times by adsorbin, more small doses until an appreciable part of the surface was covered. I n this way, data were obtained which are necessary t o calculate the surface energy distribution of the catalyst. The adsorption system associated with the calorimeter has two hish vacuum manifolds which operate over a pressure range of 1 0 - 6 t o 1 mm. One manifold is used to prepare the adsorbate and the other to outgas the catalyst. A high pressure manifold (10-1 t o 10s mm.) is used for measuring BET surface areas of the catalysts.
Vol. 67
The constant temperature bath is doubly thermostated and designed to operate over a temperature range of 30-200 f 0.01 '. Sensitive temperature control was achieved in the central bath by use of a helix which contained glycerol. The glycerol either expanded against, or contracted from, a mercury column t o achieve the temperature control. Three centrif uzal pumps (Eastern Model DH-11) circulated the water of the inner and outer baths t o eliminate temperature gradients. The large bath and two electric furnaces (Hevi-Duty) were mounted on a hydraulic platform which moved on wheels and rails for correct positioning. Special shielding and an isolation transformer were installed t o eliminate stray electrical charges which might affect the e.m.f. measurements. A tray-type calorimeter was specially desi.xned for this study so that it would have a minimum of heat transfer, mass transfer, and diffusional limitations. One already has been described8 for use in an adiabatic calorimeter system used for the measurement of specific heats of high polymers. A cut-away side view of the one used in this study is shown in Fig. 1. Seven shallow trays are gold-soldered t o a hollow, heavy-walled, central shaft. The entire calorimeter is made of platinum-rhodidm (3.595) alloy. Inside the shaft is the nichrome heater wire (39.2 ohms rmistance), and spot-welded t o its side is the thermocouple. The very broad area of contact between the trays and the shaft should result in good heat transfer to the thermocouple. Since the catalyst beds on the trays were only about 2 mm. of the catalyst durin.g out.pssin2 did not occur. a1 limitations were reduced t o a minimum and should not have affected the measured energy distributions because of non-selective adsorption. Current for the calibration was supplied by a 6-volt storage battery. A standard electric timer, which could be read to dzO.05 see., was used t o measure the length of the calibration period. Voltaxe and current were read with a d.c. voltmeter (Weston Model 1) and a d.c. ammeter (Weston Rlodell). The thermocouple e.m.f. was amplified by a Beckman d.c. breaker amplifier (Model 14) and recorded on a 0-10 mv. Brown recorder. Preparation of Adsorbates.-Helium of 99.997, min. purity was obtained from the Matheson Co. Before use, it was passed over activated carbon maintained at liquid nitroZen temperatures. Anhydrous ammonia of 99.9970 min. purity also was obtained from the Riatheson Co. It was admitted t o an evacuated system, liquefied a t -78', and distilled into a 3-1. storage flask. Prepurified nitrogen was obtained from the Matheson Co. with a purity of 99.996(r, min. and was med "as is." Preparation of Adsorbents. Alumina.-Adsorptive aluminas were prepared by ca1cinin.T p-alumina trihydrate in an atmosphere of dry nitrogen at 1100°F. for 4 hr.-AlpOl (110O0F.)-and a t 1600°F. for 8 hr.-AlzOs (1600°F.). These catalysts, like all others investigated, were ground to a small particle size. Potassium Hydroxide-Alumina.-P-Alumina trihydrate was treated with a solution containing KOH to provide 5 moles of KOH/100 moles of alumina and calcined at llOO'F.-KOH-AhO,. Platinum-Alumina.-A1203 (1100°F.) was wet with e n o q h of a solution of "P salt" (Pt(NH3)2(N0p)?)t o deposit 0.3 wt. 7, P t . This wet mixture was dried a t 250'F. overnight and finally calcined l hr. a t 1100°F. illthough Debye and Chu4have shown the deposited platinum to be highly dispersed, its exact nature (atomic or ionic) is unknown. Silica.-The silica gel was a pure sample supplied by the Esso Research Laboratories of Baton Rouge, La. It was heated for 3 hr. a t 850'F. before shipment. Silica-Magnesia.-This catalyst also was supplied by the Esso Research Laboratories of Baton Rouge, La. Sodium silicate wae mixed with sulfuric acid solution maintained below 60°F. Powdered Westvaco Co. magnesia then was sprinkled in with rapid stirrin.5 t o form a stiff co-gel. It was broken up and water washed. The catalyst then was calcined overnight at 1000°F. and for 3 hr. a t 1250'F. The resultin.3 catalyst consisted of 70y0silica and 30% magnesia.
Results The catalysts studied, and their BET surface areas, measured with nitrogen adsorbate at liquid nitrogen temperatures, are listed in Table I. (3) 4 . E. Worthington, P. C. Marx, and hI. Dole, Rev. Sei. Insty. 26, 698 (1966). (4) P. Debye and B. Chu. J . P h w . Chem., 66, 1021 (1962).
DISTRIBUTION OF SURFACE ENERGY ISCHEMISORPTIOS
March, 1963
547
TABLE I Catalyst
Surface area, m.*/g.
AlzOa (1lOO'F.) 9i02 AlzOa ( 1600" F.) KOH-A1208 Pt-Alz03 SiOz-MgO
296 572 75 250 220 536
The surface energy distributions are shown in Fig. 2-8. Heats of adsorption in kilocalories per gram-mole ammonia adsorbed are plotted against the degree of surface coverage by ammonia. Surface coverage was calculated assuming the surface of the outgasszd catalyst to be bare and a cross sectional area of 16 A.Z for the NH3 molecule. It was assumed further that multi-layer adsorption did not occur and that K1& did not displace any other adsorbed species. The surface coverage corresponding to a given heat measurement was taken as the mid-point of the incremental increase in surface coverage due to the dose. To determine the precision of the data, heat measurements mere made using three different samples of A1203 (1100°F.). The ma,ximum spread in the data was &8%. Discussion The surface energy distributions (Fig. 2-8) are plotted assuming that the catalyst surfaces were , start at completely cleaned by outgassing ( i ~ they zero surface coverage). However, since the calorimeter housing was m,ade of Pyrex, the maximum outgassing temperature was 450". Molecular water is removed by outgaseing at 150" but hydrogen atoms and hydroxyl groups remain chemisorbed to the surface. Weight loss experiments as a function of outgassing temperature showed that there are enough of these species remaining at 450" to cover about 36% of the surfaces. From Fig. 2, it is seen that when the surface energy distribution of A1203(1100°F.) is plotted as differential heats of adsorption us. surface coverage, the resulting curve has three distinct regions. The first region, which is due to adsorbate-adsorbent interaction, consists of high initial heah of about 20 kcal./g. mole. AH continuously drops off to a minimum value as surface coverage increases. The heats of the first region of the energy distribution curve may be due to either or both of two causes. First, the initial ammonia doses are adsorbed on the strongest available Lewis acid sites, forming strong bonds with the surface (due to electron transfer), and liberating large heats. It is also kiiowii from associated infrared spectroscopy studies of this and other laboratories that strongly hydrogen-bonded hydroxyl groups are present on this surface. Thus, the adsorbed ammonia might react with surface protons to form ammonium ions. This is a measure of the Bronsted, or protonic acidity of the surface. As Lewis sites of decreasing strength are covered by the succeeding doses, the curve drops to a minimum. At the minimum in the distxibution curve, all of the surface acidity, both Lewis and Bronsted, is probably neutralized by the basic ammonia. If the surface coverage due to surface protons and hydroxyl groups is included, it is seen that the minimum occurs at about 40y0 coverage.
i15
c % Of S u r f a c e Covered
Fig. 3.-Surface
energy distribution of Sios.
5,
I
I
I
I
2 21-
I
6l-
I0
I
3
I
6
Fig. 5.-Surface
I 9
I I I I 12 15 18 21 O f S u r f a c e Covered
I 24
I
27
I
energy distribution of KOH-Al2Oa.
The second distinct region of the distribution curve is the portion rising about 2-3 kcal. above the minimum, probably representing adsorbate-adsorbate interaction. This rise, which has been reproducibly measured (see also ref. 5 for another example of minima and maxima in surface energy distributions) is too great to be ascribed to simple van der Waals adsorbate interaction, which should cause a rise of no more than about 1 kcal. However, it could be due to hydrogen bonding of the XH3 to surface hydroxyl groups, the formation (5) A. 1'. Kiselev, Quart. Rev., 16, 99 (1961).
V. KEVORKIAX AND R. 0. STEINER
548
I
- _ - _ Outgassed (3 __ Outgassed (3
50°C.
I
I 36
40
450°C. 25°C.
4
I 8
I 4
0
Fig. 7.-Effect
I 12
I I I I 16 20 24 28 Of S u r f a c e Covered
I 32
of outgassing temperature on surface energy distribution of A1203 (1100°F.).
I
50°C.
I
20
$
16
Ei . j
12
rd
Y
4
II 0
3
I 6
Fig. 8.-Surface
I
I
9
12
I
I
I
15 18 21 % Of S u r f a c e Covered
I 24
I 27
30
energy distribution of SiOn-MgO.
of surface coordination complexes, or perhaps capillary condensation. Thus, if the rise is due to surface coordination complexes, ammonia molecules of succeeding doses after the miiiimum could be forming these complexes with ammonium ions and highly polarized ammonia molecules already on the surface. These complexes would result in coulombic forces which are much stronger than the relatively weak van dcr Waals interaction forces. Additional effort would be needed to determine which of the possible explanations, if any, for the rise is the correct oiie. Both the shape and
Yol. 67
extent of the second adsorptioii region are most probably dependent on the adsorbate being used to measure the energy distribution. The third adsorption region is the drop-off from the maximum. The heats probably would approach the heat of liquefaction of ammonia, 5-6 kcal., at higher surface coverages. The surface energy distribution of silica gel was measured and compared to that of alumina (1100°F.) to see whether the microcalorimeter is sensitive enough to differentiate between various surfaces (Fig. 3). Since the initial heats for silica are oiily about 60% of those for alumina, it is clear that the microcalorimeter differentiates between catalyst surface types. These microcalorimetric data are in agreement with results of infrared spectrometric studies of this Laboratory6 on the two catalysts where oiily oiie type of OH group was found on silica, compared to three on alumina (1100°F.). This would indicate a more homogeneous energy distribution as far as OH groups are concerned. It was also observed that all the ammonia could be very easily desorbed from the silica surface simply by evacuation at room temperature, implying that the bonds formed between adsorbed ammonia and the silica surface are weak. About 600" is necessary to desorb SH, under similar conditions from some of the aluminas. Substantiating the above, heats of adsorption for silica were found to be much lower (about 10 kcal.) than for alumina. Also, the silica energy distribution is considerably more uniform. This may indicate a correlation between surface energy distribution and OH surface structure. The surface energy distribution of A1203 (1600°F.) was measured and compared to that of A1203 (1100°F.) (Fig. 4). Although calcination at the higher temperature reduced the surface area by a factor of l/*, the heterogeneity of &03 (1100°F.) was not destroyed. Some of the high energy sites after the first minimum in the curve have been weakened by about 3 kcal. because of calcination at the higher temperature. However, the sites above lS% coverage seem to be appreciably stronger than those of alumina calcined at 1100°F. Calcination at different temperatures, therefore, does not merely affect the catalyst surface area, but also the surface energy distribution. Impregnation was investigated to learn whether it would be a good technique for altering the surface energy distribution (Fig. 5 ) . Impregnation with KOH at 1100°F. slightly weakened the strong sites of A1203 (1100°F.) a t initial coverages. The minimum and maximum of the energy distribution curve of untreated alumina also are almost eliminated by the impregnation. Between coverages of 5-22%, sites of the treated alumina are much weaker than those of A1203 (1100°F.). These data indicate that energy distributions may be conveniently altered by impregnation in a way which is quantitatively measurable in the microcalorimeter. Addition of Pt to the A1,0, (1100°F.) surface lowered both the minimum in the distribution curve by about 2 kcal. and the second adsorption region by several kcal. (Fig. 6). The Pt-A1203 surface between 17 and 35y0 coverage is more energetic by several kcal. than that of A1,03 (1100°F.). (6) R. 0. Steiner, unpublished results.
March, 1963
PHOTOSENSITIZED OXIDATIOX OF AQUEOUS PHESOL BY EONS
A sample of Alz03 (1100°F.) was outgassed at room temperature in order to determine the effect of adsorbed water on the surface energy distribution (Fig. '7). After outgassing at 25", this catalyst had sufficient water to form 1.5 monolayers, compared with 36% surface coverage when outgassed at 450". The heats are sharply reduced on the "wetter" sample. Actually, these heah, 13-15.5 kcal., represent the heat of solution of ammonia in water adsorbed on the alumina surface. They may be compared with the value for solution in water, 8.5 kcal.
549
The addition of AlgO to SiO:! causes a change in the surface energy distribution (Fig. 8). Si02-MgO has high initial heats of about 20 kcal. which drop off regularly to 10.5 kcal. a t 25% coverage. Si02 by itself has a more homogeneous surface, whose sites do not differ appreciably from each other in energy. Acknowledgment -The authors wish to acknowledge gratefully the helpful suggestiom of Dr. P. J. Lucchesi and the assistance of Messrs. W. Bracht and J. L. Carter, who performed the microcalorimetry and infrared experiments, respectively.
TRANSIEKT JIEASUREMENTS OF PHOTOCHEMICAL PROCESSES IN DYES. 11. THE MECH,A?JIS?II OF THE PHOTOSENSITIZED OXIDATIOX OF AQUEOUS PHENOL BY EOSIN1 BY E. F. ZWICKERAND L. I. GROSSWEIKER Department of Physics, Illinois Tnstitute of Technology, Chicago 16, Illinois Received July 16, 1962 The mechanism for the oxidation of aqueous phenol as photosensitized by eosin has been invwtigated by flashphotolytic and continuous-irradiation methods. Oxygen quenches triplet eosin and also reacts with it in a slower permanent oxidation. In the absence of oxygen, phenol reduces triplet eosin to a semiquinone, producing the neutral phenoxy radical. The radical products disappear primarily by fast back-reactions, with competing permanent decomposition from dismutation of semi-reduced eosin and bimolecular reaction of phenoxy radical. The basic form of semi-reduced eosin undergoes debromination. The dissociation constant of semi-reduced eosin was cletermined from the dependence of the rate of radical back-reaction on acidity. Numerical rate constants for the elementary reactions were calculated and the absorption spectra and extinction coefficients of the unstable species mere measured.
.[I Introduction The identification of the short-lived intermediahe species produced by the photosensitized oxidation of aqueous phenol in the presence of fluorescein or its halogenated derivatives was reported in earlier publications.2 Flash spectra obtained by the irradiation of phenol and dye solutions with visible light, absorbed only by the dye, showed that the primary act is the oxidation of phenol by triplet dye, to give phenoxy radical and semi-reduced dye. Flash bleaching showed that the dye triplet molecules decay in a fast bimolecular quenching reaction and that the radical products disappear in fast bimolecular processes. This paper includes additional flash photolysis and continuous-irradiation experiments, which further establish the reaction mechanism, and reports numerical values for the rate constants of the elementary processes. 11. Ekperimental Details A. Chemicals.-Commercial eosin Y (National Aniline, C.I. no. 768) was purified chromatographically following the method of Koch.3 The impure dye was dissolved in o.o3yOammonium hydroxide and adsorbed on a column consisting of activated alumina (1 part by wt.) and talc (3 parts by wt.). Elution with water separated four distinct bands. The eosin phase was recovered by breaking the column and dissolving the dye in ammonium hydroxide. Acidic eosin was precipitated with glacial acetic acid, filtered, washed, and dried. The maximum extinction coefficient obtained in lightly buffered aqueous solution a t pH 9 is 9.15 f 0.15 X 104 iM-l em.-' a t 516 m@. Previolls values of 8.0 X IO4 and 8.2 X lo4 were reported by Imumiira4 and Adelman and Oster,5 respectively, where chromatog(1) Supported by the U 8 Atomic Energy Commission. I. Grossneiner and E. F. Zwioker, J . Chem. P h y s . , SI, 1141 (1959); (b) 34, 1411 (1961). (3) L. Koch, J. Aesoc. Ofbc. Agr. Chemists, 39, 397 (1966). (4) hf. Imamura, J . Inst. PoZytech. Osaka City Univ., 6, 86 (1956). ( 5 ) A. H. Adelinsn and G . Oster, J. Am. Chem. Soc., 18, 3977 (1956). (2) (a) I,.
raphy had not been employed. A comparable 10% increase in the extinction coefficient of fluorescein was obtained by Lindqviste by means of chromatographic purification. The other dye phases on the column were id ed from their absorption spectra as tribromofluorescein yo), dibromofluorescein (O.O6y0), and fluorescein (O.O2%)-all wt.% organic matter. The determination of the dissociation constant of aqueous eosin by spectrometric titration of dilute solutions (10 p M ) was inconclusive. The absorption spectrum is unchanged from p H 12 to pH 4.4. Between pH 4.4 and 3.7 there is a gradual shift of the peak wave length from 516 to 520 mp and a corresponding 30y0 decrease in extinction. Near pH 3 a colloidal suspension could be detected, arid finally, a colored gelatinous precipitate was formed a t lower pH. Since the basic dye solution is di-anionic, the wave length shift observed is opposite t o the expected direction for the production of the mono-anion. Additional information was obtained by potentiometric titration of more concentrated (2 mM) eosin solutions. A single broad plateau occurred a t pH 4.3 and ti colored colloidal suspension was produced below p H 4. The observations can be explained by the influence of bromine substitution on the dissociation of aqueous fluorescein. Xoting that Lindqviste obtained p K 4.4 and 6.7 for neutral fluorescein and the mono-anion, respectively, the plateau obtained with eosin probably represents a composite of the two dissociations. The colored precipitate indicates that the eosin mono-anion is insoluble in water and that the hydroxyl group is partially ionized in the solid. This form may resemble the o-quinonoid zwitterion of fluorescein proposed by Nash.7 Recently Oster, Oster, and Karg* have reported a value of 3.6 for the eosin pK, as obtained from the optical method. The present authors found the results of the optical method to be ambiguous a t the low dye concentrations which had t o be employed, because of the difficulty in distinguishing spectral changes due to dissociation from those due t o the formation of a colloid. Therefore, we favor the potentiometric results a t higher eosin concentrations which gavg a pK of 4.3 for the monoanion. All photochemical measurements reported below were made with the di-anionic dye. (6) L. Lindqvist. Arkiu Kemi.. 16, 79 (1960). (7) T.Nash, J. Phys. Chem.. 6 2 , 1674 (1958). (8) G. Oster, G. K. Oster, and G . Karg, a'bid., 66, 2514 (1962).
