The Stereochemistry of Ethylene-L ,2-Dz Epoxidation Over Silver Catalysts

JOURNAL OF CATALYSIS 65, 297-310 (1980)

The Stereochemistry of Ethylene-l ,2-dz Epoxidation over Silver Catalysts

MAKOTO EGASHIRA,"~' ROBERT L. KUCZKOWSKI," ANDNOEL W.CANT~

* Deportment c?fChemistry, University qf Michigtrn. Ann Arbor, Michignn 48109, rrnd t School qf Chemistry. Mncqucrrie University, North Ryde, Net%’ South Wtrles 2 113, Austrdict

Received November I, 1979; revised March 19, 1980

The stereochemistry of the catalytic epoxidation of cis-ethylene-1 ,211, to ethylene oxide over four unsupported and two supported silver catalysts as well as over A&O and AgO was studied. Variations of the stereochemistry with reaction composition including addition of C,H,CI, to moderate the catalyst and the use of N,O as the oxidant were also explored. Equilibration in the product ranged between 57 and 99%. A correlation in the equilibration with extent of oxidation of the surface was proposed. Several models were discussed which could rationalize the observed partial equilibration and the sensitivity of the randomization kinetics to certain reaction variables such as oxygen concentration or addition of C,H,CI, or N,O but not others such as temperature and catalyst selectivity.

INTRODUCTION silver catalysts using similar, controlled experimental techniques during the reaction There have been three previous reports runs and product analyses. These catalysts directed at the stereochemistry of the oxi- varied in their activity and selectivity char- dation of ethylene- 1,2-dz (ET) to ethylene acteristics so that any correlation with ran- oxide (ETO) over unsupported silver cata- domization might be observed. Expeti- lysts. The earliest study indicated 60-64% ments were conducted over about a 140” equilibration in the epoxide (I). In two later temperature range and for various contact reports, Cant and Hall (2) obtained 91.6- times and flow rates. Pretreatments of the 93.2% equilibration, while Larrabee and catalysts and the reaction mixture composi- Kuczkowski (3) reported 84-87% equili- tions were also varied. The latter variation bration for a low-activity high-selectivity included the addition of 1,Zdichloroethane catalyst and 90-9 1% equilibration for a to the reactant gas, since it is often used to high-activity low-selectivity catalyst. These improve the selectivity toward ET0 (4). last authors also reported an absence of Several experiments were conducted in temperature dependence on the equilibra- which nitrous oxide was employed as the tion results and some preliminary experi- source of oxygen. Nitrous oxide has been ments suggesting a weak dependence on shown to produce predominantly mo- catalyst mass and flow rate. noatomic oxygen species on adsorption (5), It appeared valuable to direct additional and it was of interest to determine what stereochemical experiments at understand- correlation this might have with the ran- ing the differences between these studies in domization process. The stereochemistry order to deduce further mechanistic details. of epoxidation was also studied during the Therefore, experiments were performed reduction of silver oxides (Ag,O and AgO) with two supported and four unsupported with ethylene in the absence of any gaseous oxygen. ’ Permanent address: Department of Materials Sci- EXPERIMENTAL ence and Engineering, Faculty of Engineering, Naga- saki University, Nagasaki 852, Japan. Reagents. Four unsupported silver cata- 297 0021-9517/80/100297-14$02.00/O Copyright @ 1980 by Academic Press. Inc. All right? of reproduction in any form reserved. 298 EGASHIRA, KUCZKOWSKI, AND CANT lysts were employed. Ag(1) and Ag(2) were conditions on the randomization, the flow the silver powders previously described in rate of the gas mixture was varied over the Ref. (.3), while the Ag-sponge was similar in range of 15 to 90 cm3 min-‘. For most type to that reported in Ref. (2). Their studies the typical flow rate was 60 cm3 selectivities/activities as well as surface min-‘, where it was confirmed that the area were redetermined as necessary in the reaction was not diffusion controlled. The new system employed here and some differ- typical composition of the reactant gas was ences were noted from the previous re- 3% C2H4, 10.5% 02, and 86.5% N,. A ports. The Ag(3) catalyst was obtained by series of experiments were carried out in reducing Ag,O with C2H, at 180°Cfor more which the concentration of ET was varied than 3 hr. Two supported silver catalysts from 3 to 90% at a constant concentration were 1.5% Ag/a-AlpOa which was prepared of oxygen (10%). In another series, the by Engelhard Industries and 5% Ag/SiC& concentration of oxygen was varied from of the type used by Wu and Harriott (6). 0.2 to 10.5% with 3% C,H,. In both series, Ag,O was prepared by precipitation from a N, gas made up the balance of the mixture. silver nitrate solution with sodium hydrox- In some experiments, 1,2-C,H,CI, was also ide, followed by washing 20 times and included in the feed gas. This was achieved drying at 85-88°C for 2 days (7). AgO was by flowing the oxygen over a small amount obtained by the reaction of silver nitrate of the material condensed in a trap at -63 with potassium peroxydisulfate, followed or -78°C. by thorough washing and drying in air for 3 The oxidation of ethylene by N20 was days (8). conducted over the Ag-sponge catalyst at Matheson CP-grade C,H, and NzO, ex- 280°C by using the reactant composition of tradry Nz, and prepurified O2 were em- 2% CzH4, 2-13% N20, and 85-96% NB. The ployed. cis- and truns-ETd, were obtained reductions of Ag,O and AgO with ethylene from Merck Sharp and Dohme, and their were carried out at 150 and 140°C respec- isomeric purity was the same as specified tively, with the reducing gas comprised of before (3). 2% C2H, and 98% N2. Apparatus and procedure. A single-pass The amount of C2H2D2 reacted in a ste- flow reactor which could be evacuated was reochemistry experiment was usually 0.5- employed. The catalyst (0.15-10.0 g) was 1.0 mmol. In these runs, the N2 stream was held in a 6-, 10.5-, or 12.5-mm-i.d. Pyrex diverted over small amounts of C2H,D2 tube reactor. condensed in a trap at - 145to - 125°Cafter Pretreatment of the Ag catalysts nor- the C2H, flow had been stopped. mally consisted of reduction in flowing H2 Reaction products were sampled by a at 3 10°C for several hours except for Ag(3) syringe inserted into the effluent for gc which was reduced with C,H,. Reaction analysis or collected in a sample loop at runs were then conducted at a desired - 160°Cfor spectroscopic analysis. The lat- temperature (140-300°C) for l-3 hr until the ter samples were trap to trap distilled to catalyst stabilized. It was noticed that when concentrate either the ET or the ETO. A a catalyst was exposed to room air its Varian 920 gas chromatograph was em- activity markedly decreased unless it was ployed to analyze reactant and product reduced again. This led to several experi- gases. Three columns were used: 10% Car- ments where room-temperature air was bowax 1000 on Chromosorb W-HP (2.5 m, passed over the catalyst for l-3 days before 45°C) for the analysis of ETO, Porapak Q conducting an oxidation run. These runs of (2.5 m, 25 or 60°C) for COz, NzO, and ET, unreduced catalyst are labeled “air cata- and Molecular Sieve 5A ( 1.5 m, 60°C) for lyst.” 0, and Nz. The conversion/selectivity data To study the effect of diffusion-controlled for the oxidation of ET-d, were also ob- ETHYLENE- 1,2-& EPOXIDATION 299 tained. It was reliable only for the ratio RESULTS of CO, and ET0 formation, that is, the selectivity of reaction, because the reac- Reactor Variables tion time for ET-d, was usually less than The previous report (3) suggested a slight 10 min. increase in the percentage of retention of A Hewlett-Packard 8460A microwave the original stereochemical configuration spectrometer was employed to determine upon decreasing the catalyst mass or in- cisltrans ratios in the product epoxide creasing the flow rate through the reactor. &H,D,O. This analytical technique has The flow rate employed was 14 cm3 min-’ been discussed extensively and the pre- where breakdown of plug flow or diffusion vious report can be consulted for further effects might become noticeable. In order details (3). In most cases three to five pairs to test whether such conditions might have of rotational transitions were compared to an effect on the randomization kinetics, obtain estimates of precision and accuracy. experiments were carried out over catalyst The deviation of each percentage cis-ETO- Ag(2) while the flow rate was varied in the d, value from the average was less than range 15-90 cm3 min-’ at a constant contact -cO.5%. In this paper, data from only one time, W/F = 5.0 g ’ set . cm-” (W = cata- transition ( 111 c OuO)which was closest to lyst weight, F = flow rate). The results at the average are presented. The absolute 190-250°C show effects at flow rates less accuracy in the intensity ratios is believed than 40 cm” min’ : at 210°C for example, to be better than & 1.O% including consider- total conversion of ET at 15 cm” min-’ was ation of a possible systematic error due 23.2% while 33.4% at 60 cm3 min-‘. Also, to the lack of vibrational partition func- an increase in selectivity is observed in tions for cis- and trans-C,H2D,0 (3). The Table I (and Table 2) while conversions factor R, converting the microwave inten- were also rising contrary to expectations sity ratio I,,/Z,,,, into the ratio of the for the usual triangular kinetic scheme number of molecules of the two species, where CO, is formed in reactions parallel is 1.18 (3). and consecutive to ETO. No ready expla-

TABLE 1 The Effect of Flow Rate on the Stereochemistry of Epoxidation of Ethylene over Ag(2) at a Constant Contact Time W/F = 5.0 g sec. cmJ (C,H, = 3%, 0, = 10.5%, N, = 86.6%)

Reaction Flow rate Oxidation of C,H, Oxidation of cis-&H,D, temp. F - (“C) (cm” . min-‘) x o s mb &I 0 LYlL,,, 9%cis -ETO” (ii) (%) (%)

210 15 23.2 33.1 47.0 1.440 54.7 60 33.4 36.5 48.5 1.393 54.1 250 15 55.1 33.4 37.1 I.389 54.0 60 72.9 37.5 44.0 1.392 54.1

” X,, = conversion of ET = 100 (nETO+ 0.5nTOz)/nE.l.(in,t,~,~. h SETo= selectivity to ET0 = lOOn,,,/(n,,, + 0.5n,,,). p Observed cis/rrons intensity ratio for MW transitions. d Percentage cis + percentage trans = 100. To estimate the randomization, a correction should be made for the small amount of fruns-ET-& present in the cis starting materials (3). These corrections will raise the randomization values slightly (from 0. I to 0.9%) from those estimated by ignoring them for the data in all tables except Table 10. These corrections were normally not made in the text except for Table IO, since the precise amount of trans-ET was not known. The text estimates are therefore lower bounds. 300 EGASHIRA, KUCZKOWSKI, AND CANT

TABLE 2 The Effect of Contact Time on the Stereochemistry of Epoxidation over Ag(2) (CzH4 = 3%. 0, = 10.5%. NZ = 86.5%, F = 60 cm3. min-*)

Reaction WIF Oxidation of C,H, Oxidation of cis-C2H2D2 temp. (g set . cmm3) (“C) &T (%I SE,,, (%I SE,,, (%‘o) LILz~ % cis-ET0

210 0.7 8.8 28.9 38.6 1.433 54.8 1.25 14.5 33.3 41.1 1.408 54.4 5.0 33.4 36.5 48.5 1.393 54. I 250 0.7 28.0 30.4 43.3 1.408 54.4 5.0 77.9 37.5 44.0 1.392 54.1 7.5 86.0 37.8 45.2 1.361 53.5 nation is apparent for the increase in selectivity. However, these effects were not Stereochemistry over Six Silver CutaIysts found to noticeably affect the randomiza- In order to determine whether the ran- tion kinetics of epoxidation as shown in domization process depends upon the prop- Table 1. erties of the silver catalyst, stereochemical Table 2 shows the effect of contact time experiments were conducted over four un- or catalyst mass at the constant flow rate of supported and two supported catalysts 60 cm3 min-’ on the stereochemistry over whose activities and selectivities varied Ag(2). A slight increase in the randomiza- considerably. The results over the reduced tion was again observed with increasing catalysts are listed in Table 3 with their contact time or conversely an increase in surface area, total oxidation rate, and selec- the conversion level of ET. The slight de- tivity to ETO. Table 3 also includes the crease in retention at longer contact times temperature-dependence experiments. could arise from increased randomization In comparing the catalytic properties of in the feed ET. It was previously shown (2) Ag(l), Ag(2), and Ag-sponge to the earlier (and confirmed in this study) that 2-4% of reports (2, 3), the following aspects were the reactant ET-u’, has randomized. This is noted. The catalytic behavior of Ag(1) was not enough to account for the much greater the most disparate from the earlier study. randomization in the ET0 but could ex- The catalyst was noticeably more active plain the weak dependence of the cis/trans with total conversions ranging from 1.5 to 2 ratios with contact time. times greater. The selectivity to ET0 was In some high-temperature olefin oxida- slightly lower and remained effectively un- tions (e.g., propylene over bismuth molyb- changed at about 68% in the temperature date at 400°C (9)) epoxides are formed by a range 180 to 25O”C, above which the con- surface-initiated homogeneous reaction in a secutive oxidation to CO, became notice- postcatalytic zone. If such a reaction was able. The difference is attributable partly to occurring in the ET oxidation, its effect on the different flow rate as mentioned before the stereochemistry data would need and also to the catalyst pretreatment since clarification. However, no significant catalyst performance closer to the previous changes in the reaction kinetics were ob- study could be obtained if H, reduction was served when the postcatalytic volume was not undertaken. The activity of Ag(2) was varied from 1.0 to 12.5 cm3, indicating the not greatly different from the previous absence of any homogeneous reaction ef- study. Selectivity was again low and in- fects in this case. creased from 32 to 38% in the temperature ETHYLENE- 1J-d, EPOXIDATION 301

TABLE 3 Comparison of the Stereochemistry of Epoxidation over Some Silver Catalysts (C,H, = 3%, 0, = IO.%%, N, = 86.5%, F = 60 cm:’ min-‘)

Catalyst Reaction WIF Oxidation of C&L Oxidation of cis- C,H,D, temp. (g set cm-“) (“C) x,r (55) .~,.T,, CR) Oxidation rate SET,, (%‘F) L,JL,,ts % h-ET0 (pm01 mine’ g-l)

.&(I) 180 1.5 7.7 67.6 0.74 72.6 1.364 53.6 s = 0.14 210 7.5 21.7 68.0 73.2 1.366 53.6 my g-* 250 2.0 23.2 67.1 72.4 1.396 54.2 7.5 54.3 68.0 13.2 1.382 53.9 2.0 30.0 67.8 73.9 I.420 54.6 7.5 16.6 62.3 69.8 1.397 54.2 Ad?) 180 5.0 II.0 34.5 1.58 53.6 1.409 54.4 s = 0.19 210 5.0 33.4 36.5 48.5 I.393 54. I m2 g-1 250 5.0 77.9 37.5 45.2 1.392 54.1

Ag-sponge 180 2.0 6. I 72.1 2.29 78.2 1.399 54.2 s = 0.43 210 2.0 20.9 73.7 78.8 I.41 I 54.4 m2 g-l 250 2.0 58.9 67.6 - I.410 54.4 Ag(3) I40 0.93 5.4 46.6 4.12 55.9 1.443 55.0 s = 0.71 I50 0.93 10.3 44.3 7.w 55.4 1.424 54.7 m’ g-l I80 0.93 69.8 39. I 46.0 52.5 I .442 55.0 k&On I80 10.0 1.8 41.0 0.13 - 1.359 53.5 (1.5%) 210 10.0 3.5 40.4 50.7 1.367 53.6 250 10.0 8.4 32.3 41.8 1.378 53.9 Ag-SO, 180 0.5 a.2 s4.1 II.7 46.5 1.410 54.4 (1.5%) 210 0.5 25.8 29.5 45.1 I.441 54.9 250 0.5 43.7 27.8 - I.448 55. I - range 170-250°C while total conversion of The selectivity to ET0 from cis-C2H2D2 ET increased from 7 to 78%. The Ag- is always larger than that from CzH4, an sponge behaved similarly to the promoted observation also seen in most data in other Ag-sponge in Ref. (2), having nearly con- tables. This isotopic effect can be explained stant selectivity of about 72% from 170 to in the manner similar to Cant and Hall (2) 240°C. The Ag( 1) and Ag-sponge catalysts who indicated that the enhancement in se- in Table 3 show selectivities somewhat lectivity for CzD4 resulted from a reduction greater than the 40-50% typical of rela- in the rate of complete oxidation. tively pure silver, The catalysts were not Regarding the stereochemistry, it is clear analyzed for the content of promoters such that the variability in randomization at a as Cl. It appears, however, from the in- given temperature is no larger than 1.5% creased selectivity that some accidental when all catalysts are given the same H, promotion had occurred. reduction treatment and even results for The Ag(3) catalyst was the most active of Ag(3) which was reduced differently fall in the unsupported catalysts examined and this range. This relative constancy for the the stereochemistry runs were made at 140 percentage &-ET0 in the range 53.5-55. I and 150°C. The two supported catalysts is the most striking result except to further were not as thoroughly studied over as observe that there is no systematic correla- many temperatures. It was clear that the tion between the degree of randomization Ag/A1,OR was relatively inactive while con- and the catalytic properties or ac- versely the Ag/SiO, had a quite high activ- tivity/selectivity data. It should be rec- ity. Selectivity for both catalysts was in the ognized that this constancy does not repre- range of 28-41%. sent equilibration. It was shown previously 302 EGASHIRA, KUCZKOWSKI, AND CANT

(2, 3) that use of truns-ET would result in Pretreatment of Catalyst about 53.5-S% trans-ETO. (See also Ta- ble 8.) The minor variation in percentage The difference between the present and &-ET0 among the six catalysts may be the earlier result on Ag(1) suggests that the attributable to different extents of ran- pretreatment procedure for the catalysts domization, although small, in the recov- plays an important role in the randomiza- ered ET-d, due to the different contact tion process. The effect of air treatment time as mentioned before, or to the differ- was systematically examined over Ag( I) ent surface state of the catalyst which and Ag(2). As shown in Table 4, air treat- will be discussed later. ment caused a noticeable increase in the The effect of temperature is hardly well retention of the ET0 conformation. In this defined, and most randomization variations way the earlier stereochemistry results on with temperature are within the experimen- Ag(l) could be nearly reproduced. Also, tal precision. Hence, the data indicate that the activity markedly decreased and the there is essentially no temperature depen- selectivity increased somewhat to approach dence on the randomization kinetics in the values more similar to the earlier study. range 140-280°C as also reported in the This indicates a significant correlation be- previous study (3). tween the surface s&ate of silver and the The stereochemistry results for reduced randomization process. Ag(2) and Ag-sponge are identical to the previous studies (2, 3) within experimental Composition of Reactant error. For reduced Ag(l), the results differ The stereochemistry results at low con- for percentage c&ET0 and are 3-4% lower centrations of O2 with a fixed concentration than the earlier results where values of of ET are listed in Table 5. The selectivity 57.2-58.0% were observed and where air to ET0 was markedly decreased by lower- was passed over the catalysts at 200- ing the concentration of 02, which is likely 250°C for several hours prior to reaction caused by the predominant formation of runs. atomic species O- or 02- on the surface

TABLE 4 The Effect of Catalyst Pretreatment on the Stereochemistry (C1H4 = 3%, 0, = 10.5%, N2 = 86.5%, T = 25OC)

Catalyst Pretreatment W/F Oxidation of CzH, Oxidation of cis-CzH,D, (g set . cm”) XET (%) SET,,(%) S,,, (%) I,,,/I,,,., % cis-ET0

.Ml) H, reduction, 3 10°C 7.5 54.3 68.0 73.2 1.382 53.9 Room-temperature air flow for 16 hr 7.5 8.5 74.8 17.6 1.508 56.1 Room-temperature air flow for 64 hr 7.5 5.1 75.1 72.6 1.520 56.3 A&‘) H, reduction, 310°C 5.0 77.9 37.5 44.0 1.392 54.1 As stored in air 6.5 10.3 64.6 68.0 1.528 56.4 Room temperature air flow for 72 hr 6.5 2.4 54.5 53.5 1.590 57.4 ETHYLENE- 1,2-d? EPOXIDATION 303

TABLE 5 The Effect of 0, Concentration on the Stereochemistry (C,H, = 3%, O2 = 0.22-15.5%, N2 = 86.5-%.8%, T = 25o”C, W/F = 7.5 for Ag(l) and 5.0 g. sec. crnm3for Ag(2))

Catalyst Concentration Oxidation of C,H, Oxidation of ck-C2H2D2 of 0, (%) X02” YET s ET” s ETO 4is14rons % cis-ET0 (%) (%) (%) (%)

Ml) 10.3 20.5 54.5 67.9 73.2 1.382 53.9 0.51 96.6 7.3 32.5 51.7 1.312 52.6 0.22 -100 3.0 31.4 35.8 1.309 52.6 Ag(2) 10.5 44.4 77.9 37.5 44.0 1.392 54.1 1.16 98.0 13.9 13.0 25.8 1.342 53.2

a X0. = conversion of 0, = 100 (0.5nET, + 1.5n,,,)/n,, (initial).

(5, 10, 11). The corresponding percentage nificance of adsorption sites for ET in the &-ET0 values also decreased by about l- randomization process. Since organic hal- 1.5%, a small but still significant decline. ogen compounds are well known to mod- Table 6 shows the effect of ET concentra- ify the surface state and produce higher tion on the stereochemistry. The purpose of selectivity, some experiments were car- this experiment was to repeat the earliest ried out on a catalyst heavily moderated study (I), which employed a reactant mix- with chlorine (12, 13). Results are listed ture of 80% ET and 20% 0, and reported in Table 7 where 1,2-C2H4Cl, was added about 70% c&ET0 from &-ET-d,. How- to the basic feed gas at concentrations of ever, we found that the percentage c&ET0 0.004 and 0.01%. The poisoning of the was almost independent of the concentra- catalysts with C2H4C12is seen to result in tion. This indicates no significant, direct an increase in the retention of the original contribution from the gaseous ethylene olefin conformation as in the case of the molecule to the stereochemistry data. The air catalyst (Table 4). A pronounced de- slight decrease at higher ET concentrations crease in activity and relative increase in may be due to the lowered effective con- selectivity are also observed. The lower centration of oxygen which is indicated by selectivity for cis-CpHzD, compared to its high conversion level. CzH4 for the last entry in Table 7 appears These results may suggest the sig- anomalous and has no readily apparent

TABLE 6 The Effect of ET Concentration on the Stereochemistry over Ag(l) (C,H, = 3.0-90.2%, 0, = IO%, NZ = O-87%, T = 250°C)

ET concen- WIF Oxidation of C,H, Oxidation of cis- C,H,D, tration (g set cme3) (%o) X02 XET s ET0 L51LYln % cis-ET0 (%) (Go) (%o)

90.2 20.0 95.7 6.4 55.4 1.326 52.9 20.3 20.0 74.4 24.7 63.4 1.352 53.4 3.0 7.5 20.5 54.5 67.9 1.382 53.9 304 EGASHIRA, KUCZKOWSKI, AND CANT

TABLE 7 The Stereochemistry of Epoxidation over Ag-Sponge Heavily Moderated by I,ZC,H,CI, (C2H4 = 3%, 0, = 10.5%, N2 = 86.5%. 2’ = 280°C)

Concentration W/F Oxidation of of C,H,CI, (g set cmm3) CA Oxldatlon of cis-C,H,D, (So) &T s ET0 s ETO 4,,/L,,,, % cis- ET0 (%) (%I (So)

0.0 0.15 40.0 55.0 58.9 1.340 53.2 0.004 3.0 9.1 78.5 80.1 1.639 58.1 0.01 3.0 8.4 78.5 73.5 1.622 57.9 explanation. However, only one run at concentration of N,O. The amount of iso- this concentration of C,H,Cl, was made. merization in reactant ET also seemed somewhat larger (-4-5%) for these runs Oxidation by N20 compared to a normal oxidation with OZ. Table 8 compares the results of NzO It was also noted that NzO caused a versus O2 as the oxidant over Ag-sponge at pronounced decrease not only in conver- 280°C. It was found that the oxidation of sion levels of ET but also in selectivity to cis- or trans-ET-d, by NzO resulted in more ET0 particularly at the lower concentra- randomization than in the normal oxidation tions of N20. The decrease in activity and by 0,. The randomization increased selectivity can be ascribed to the slow slightly with the decreasing concentration activation step of NzO and to the predomi- of N,O. However, complete randomization nant formation of atomic oxygen species on was not quite reached even at the lowest the surface, respectively (5).

TABLE 8 Comparison of the Stereochemistry of Epoxidation by N,O and by 0, over Ag-Sponge at 280°C (C,H, = 1.9%, W/F = 3.0 g. sec. cmm3 for oxidation by N,O, and CzHl = 1.5%, W/F = 0.15 g sec. cmm3 for oxidation by 0,)

Reaction Concentration Oxidation of of C&L Oxidation of ciS-C2HZDZ oxidant (%I X s ET0 s ET0 LislIlmns % cis-ET0 (?g (So) (%o)

ck-C2H2DZ 13.3 18.1 23.5 36.2 1.275 51.9 N,O + or 7.5 15.2 18.7 28.1 1.265 51.7 CA 2.1 6.4 12.8 19.3 1.261 51.6 rrans-C,HpD, 13.5 20.4 23.2 35.5 1.126 48.8 N,O + or 7.5 15.1 19.2 27.9 1.132 48.9 CA 2.0 6.8 12.2 19.9 1.139 49.1 ci.K2H2DZ 10.2 42.2 51.2 58.9 1.340 53.2 02 + or 7.1 45.2 55.0 62.4 1.347 53.3 U-L 3.6 33.1 50.0 58.6 1.329 53.0 1.0 16.9 36.7 48.0 1.302 52.4 ETHYLENE- 1,2-d, EPOXIDATION 305

100 ET0 obtained is very close to the results 6 80 5 over the air-treated Ag(1) and Ag(2) cata- a k lysts. Ag,O itself seems to be inactive for 60+ z epoxide formation, because the selectivity kf 40 B was very low at the beginning and gradually E increased up to 40-45.%, a level character- 20 5 ? WY istic of a plain silver catalyst. Therefore, 0 0 50 100 150 200 the formation of ET0 should not be as- REDUCTION TIME, lmln) cribed to catalysis by Ag,O but rather by FIG. I. Reduction of Ag,O ( 1.0 g) with C,H, at 150°C Ag itself. Thus, one could reasonably ex- (C,H, = 2%, NZ = 98’S, F = 60 cm3 min’). pect a good coincidence between the stereochemistry for this reduction experiment and the usual oxidation run over the air Reduction of Silver Oxides with ET catalyst in Table 4. It has been reported (14) that the reac- A similar study over AgO at 140°C is tions of some silver oxides with ET pro- shown in Fig. 2. The behavior of AgO was duced ET0 even in the absence of gaseous very different from Ag,O, again in agree- oxygen. The stereochemistry during these ment with the earlier report (14). It is reactions was also investigated since it apparent from Fig. 2 that AgO itself has a could provide insights on the effect of the good catalytic activity for epoxidation with oxidation state of the catalyst on the ran- a selectivity of 40%. The selectivity gradu- domization process. Results of the reduc- ally decreased with the progress of reduction of Ag,O with C2H, at 150°C are shown tion and became closer to the initial value in Fig. 1, where oxidation products from for Ag,O, which is not reduced very fast at CPH4are plotted versus the reduction time. this temperature. Thus, the reduction of Figure 1 also shows the percentage reduc- AgO in Fig. 2 is due to the change of AgO tion of bulk oxide, calculated from the rate into Ag,O with 50% reduction correspond- of incorporation of oxygen into products. ing to the completion of this change. The One hundred percent reduction corre- stereochemistry results obtained at two dif- sponds to the complete reduction of Ag,O ferent degrees of reduction are listed in into metallic Ag. These results are in good Table 10. The results were markedly differ- agreement with the earlier study (14). The ent from the usual oxidation runs in Table stereochemistry results which were con- 3. The percentage cis-ET0 from c&-ET-d, ducted at four stages of the reduction cycle was about 70%, which implies 43 2 3% are listed in Table 9. The percentage cis- retention of original olefin configuration (for an ET composition of 96 + 2% cis- and 4 k 27~ truns-d,). TABLE 9 The Stereochemistry of Expoxidation during the Reduction of Ag,O (1.0 g) with &-ET-d2 at 150°C

Percentage SET0(clo) Ll4~,,W 56 cis -ET0 reduction From From CJL C&D2

o-4 2-6 3-11 1.473 55.5 0 0 20 40 60 80 TOO 120 40-44 28-31 - 1.555 56.8 REDUCTION TIME, ho 1 73-78 42-43 49 1.569 57.1 92-% 41-44 47-50 1.545 56.7 FIG. 2. Reduction of AgO (0.5 g) with C,H, at 140°C (C,H, = Z’S, N, = 98S, F = 60 cm” min’). 306 EGASHIRA, KUCZKOWSKI, AND CANT

TABLE 10 tion occurring during the reduction of Ag,O The Stereochemistry of Epoxidation during the with ET in the absence of any gaseous Reduction of AgO with c&ET-d, at 140°C oxygen. However, a much higher retention (70% &-ET0 from &-ET-d,, i.e., -57% Percentage .%TO(%) I~is~l,,.a,,.~% k-ET0 equilibration) was observed in the reduc- reduction tion of AgO with ET. From From (vi) A catalyst heavily modified with 1,2- C,H, C&D, C,H,Cl, also resulted in a decrease of 9- O-5 40-30 36 2.711 69.7 10% in the equilibration compared to runs 11-17 21-15 12 2.734 69.8 without it in the feed mixture. (vii) The use of N,O in place of 0, as the oxidant increased the randomization al- DISCUSSION though complete randomization was not The important observations from this quite attained (-98-99% equilibration). study are summarized as follows. The stereochemistry results with Ag cat- (i) Four unsupported and two supported alysts summarized under (i) and (ii) are catalysts when given similar reduction pre- essentially the same as those of Cant and treatments displayed similar stereochemi- Hall (2) and of Larrabee and Kuczkowski cal behavior in the oxidation of ET-& to (3). However, the result of Richey (I), who ETO-d2, independent of their wide variabil- reported about 70% cis-ETO+& from cis- ity in catalytic properties. 53.5-55% of the ET-&, could not be reproduced in the product epoxide had the same configuration present study for any reaction conditions as the starting material, i.e., about 90-93% using Ag catalysts and will not be included equilibration occurred upon epoxidation. in the further discussion. The isomerization of ethylene-d, during re- The above findings show insensitivity of action runs was only a few percent, which the stereochemistry to temperature and cat- is not enough to account for the much alytic properties such as selectivity but a higher randomization in the product ETO. dependency on the catalyst pretreatment or (ii) Randomization was independent of reactant composition. These latter varia- temperature over the range 140-280°C. tions shoiv a qualitative correlation of the (iii) Mass transfer limitations did not af- stereochemistry with the state of oxidation feet the stereochemistry based on results of the surface. This is summarized in Table from flow rate variations although a slight 11. dependence on catalyst mass or contact In Table 11, the highest retention is ob- time was observed. (iv) No effect on the randomization was TABLE 11 observed upon varying the ET-& concen- A Correlation of the Stereochemistry of Epoxidation tration in the range of 3-90%. However, and the Surface State of Silver Catalysts decreasing the 0, concentration increased % c&ET0 Reaction Oxidation the randomization. For concentrations of from cis-ET-d, state of O2 I l%, when the 0, was essentially surface consumed, values of cis-ETO-& from cis- -70 Reductionof AgO by ET High ET& decreased to 52.4-52.6% (-95% -58 Oxidation over Ag modified with CIH,C& equilibration). 56.5-57.5 Reduction of A&O by ET (v) Air treatment of the catalysts at room Oxidation over air catalyst 53.5-55 Oxidation over reduced Ag temperature raised the percentage &-ET0 52.5-53 Oxidation at low concn from c&ET-& to 56.5-57.5% (85-87% of 0, over reduced Ag 51.5-52 Oxidation by N,O over equilibration). Similar stereochemi.cal reduced Ag Low results were also obtained for the epoxida- ETHYLENE-1,2-d, EPOXIDATION 307 served for the epoxidation occurring during than for the air-treated or C,H,Cl,-modified the reduction of AgO into Ag,O. AgO con- catalysts and should also depend upon the sists of Ag+, Ag7+, and O*- and can be partial pressure of oxygen under operating written as Ag+Ag?+(02-)2 (15). Altema- conditions. Thus, runs with low concentra- tively, Kagawa et al. (24) indicated by tions of oxygen or the employment of N,O electron diffraction that the activity of AgO as the oxidant, whose decomposition is so toward epoxidation arose from the forma- slow as to become the rate-determining tion of a superoxide AgO, with the NaCl step, would lead to a lower overall oxida- structure on the surface (16). A recent ESR tion state for the catalyst. study (17) also indicated the existence of These experimental findings and the con- the 0; species in AgO. Regardless of the straints they impose will be examined in oxidizing species, it is apparent that the terms of three general categories of reac- surface exists in a high oxidation state tion models that have received attention in during the reduction since the main product the literature: (a) Nonconcerted process- from AgO was not Ag but Ag,O. radical intermediate; (b) concerted process In the case of the reduction of Ag,O by -symmetric intermediate; (c) two-site or ET, the stereochemistry was quite analo- competitive reactions. gous to the standard oxidation over the air- One approach to the epoxidation process treated catalyst in spite of the formally has been to emphasize an initial, noncon- higher oxidation state of the former. How- certed adsorption process. Gerei et al. (22) ever, as discussed before, the formation of suggested from their ir study that ET ad- ET0 in the reduction of Ag,O is ascribable sorption occurred with rupture of the dou- to the catalytic function of metallic silver ble bond to form the peroxide I. Force and produced which is likely to have many Bell (23) also proposed a similar type of oxygen ions available at nearby sites. One can therefore reasonably expect a similar trend for the reduction of Ag,O and the oxidation over the air catalyst. The influence of l,ZC,H,Cl, is to modify or block surface sites by the adsorption of chlorine atoms presumably as Cl- (18). I II: In Consequently, the surface would become more oxidized and randomization results intermediate, II, but differing in the form of similar to air-oxidized catalysts could be the active oxygen species. These two inter- expected. mediates, with rotation permitted about the The surface must be considered as parti- C-C link, can correlate with the stereo- ally oxidized even for the reduced Ag cata- chemistry results which show extensive lysts. It has been reported (4, 10, 19) that equilibration usually between 90 and 93% catalysts reduced with H2 have oxygen and sometimes higher. Of course, the small species on their surfaces under reaction degree of retention requires that ring clo- conditions. Kagawaet al. (20) indicated the sure to ET0 can compete with full develop- formation of AgOz on the surface by elec- ment of rotation about the C-C bond in the tron diffraction whereas Sato and Seo (22) radical intermediate and that this competi- observed that the adsorption of oxygen on tion is temperature independent. The small silver resulted in the formation of Ag,O barrier leading to the partial equilibration with 0; adsorbed on it. In fact, the adsorp- and its variation with positive charge on the tion of ethylene is possible only on a Ag surface might be visualized through an in- surface having preadsorbed oxygen. The teraction depicted by III. oxidation state of the Ag should be lower One of the epoxidation mechanisms pro- 308 EGASHIRA, KUCZKOWSKI, AND CANT

posed by many authors (19, 20, 24) is the gaseous ethylene molecule and a diatomic symmetric Rideal-Eley model involving a oxygen species, O;(ad) or O,(ad):

?- 7 H2W.ti2 + 0 - H$*-,CH2 - H$-CHZ + 4 -. ,I & ‘0’ ‘0’ (1)

In the limit of a strictly concerted process, that the equilibration rates IV t, I (random) the epoxide would completely retain its in competition with IV --f ET0 (reten- starting configuration. Amendments to tion) have similar thermal dependencies. such a mechanism to incorporate the ob- Cant and Hall further speculated that I served partial randomization might include might also be the active species leading to (a) ascribing a finite lifetime to IV with CO, via an intermolecular isomerization to occasional ring opening/closing/random- form the species Ag-0-CH-CHs. This as- ization processes before rupture of the pect now appears less attractive in view of O-O bond to give ET0 and (b) subse- the relative insensitivity of the randomiza- quent steric isomerization in the epoxide. tion to wide variations in catalyst selectiv- This latter possibility has been ruled out ity. This presupposes that increasing loss by Cant and Hall (2). of I through total oxidation, as expected Process (a) was also discussed by them. for very active/low-selectivity catalysts, They proposed a facile equilibrium between would decrease the production of random- the peroxide IV and the open chain form I ized ET0 by decreasing the reverse rate for which would undergo rotation about the C- reaction (2). The data, however, strongly C bond before reforming the C-O bond and suggest that kinetic processes leading to desorption to ETO: CO, formation are uncoupled from those leading to partial randomization. “&,=,FH2 Ii+ - CHz ‘\ ,’ Larrabee and Kuczkowski (3) pointed 0 . 0’ out that the incomplete-randomization :, :, (2) results might also be rationalized by a two- h, & site mechanism with two competitive reactions occurring with randomization and re- nz I tention, respectively. One of the simplest To account for partial equilibration, this models would be a combination of the proposal would require that the equilibrium concerted and radical processes just dis- not always develop but that IV occasionally cussed, for example, involving intermedi- proceed directly to epoxide with retention ates II and IV. The randomization ratio or equivalently that species 1 not always should then vary with any change in the sterically equilibrate before epoxidation or number of -0-Ag and -O,-Ag sites leading before returning to IV. Variations in extent to intermediates II and IV. of randomization with surface charge could Greater randomization was indeed ob- be incorporated into the model through served when O2 concentrations decreased stronger interactions with a more negative or when N!,O was the oxidant which might oxygen species decreasing the forward rate be consistent with increases in atomic 0 of reaction (2) or with surface interactions species and the intermediate II. However, similar to III. It would also be necessary some variation in the randomization ratio ETHYLENE- I ,2-&, EPOXIDATION 309

might also be expected as the Ag catalysts ?C=C’ H. H ; IX ‘c-cl differ in surface area, selectivities, and ac- \ I’ H’ I H 4’ I tivities, but this was not observed. Also, an 4 effect on the randomization with ET pres- P I!u sure would normally be expected. The reaction is positive order in ethylene so cov- also be considered. This seems attractive erage increases with ET pressure. If two since (a) species such as V and VI are sites are available, then their relative popu- plausible chemical species for isolated ET lations will change unless isotherms are and Ag and could accommodate the varia- identical which seems unlikely. It also ap- bility in randomization with state of oxida- pears problematical to expect that these tion of the silver. (b) The observation of two processes would have similar activa- small amounts of randomization in the eth- tion energies since their transition states ylene is also easily correlated with such are likely to be quite dissimilar. Hence, the adsorption intermediates that do not in- plausibility of a proposal involving compe- volve strong interactions with oxygen in the tition between intermediates II and IV is initial stages of the reaction. (c) Infrared debatable and any two-site mechanism studies by Force and Bell (23) suggested at must be carefully formulated in view of the least two different Et-Ag precursors. One constraints imposed. set of bands strongly resembled that for ET The above proposals all contain oxygen coordinated to isolated silver ions (25) and in the adsorbed species accounting for the they postulated structure V. randomization. Intermediates involving an One possibility is described by reaction ET-Ag species prior to epoxidation might (3):

(3)

This process would assume identical initiat- The absence of any significant tempera- ing steps for epoxidation versus isomeriza- ture dependence on the stereochemistry tion with their relative rates determined by must be a constraint for such a process site configuration differences. These latter necessitating similar adsorption isotherms two processes must also have similar acti- and activation energies for the epoxidation vation energies. step. An invariance in numbers of sites with Alternatively, a two-site proposal could temperature is also necessary but this could involve different numbers of V and VI. If it be correlated with the oxygen data of Czan- is assumed that the two different adsorption derna (lo). He measured the 0, adsorption sites react with an oxygen species and that isobar at -77 to 351°C at an 0, pressure of the subsequent epoxidation process occurs 10 Torr. In the range of 140 to 280°C which with retention of configuration for species was employed in the present work, the V, then the stereochemistry results might surface coverage of oxygen was very close be reasonably explained. The extent of to monolayer coverage. The coverage was randomization or retention would depend nearly independent of the adsorption tem- upon the number of each site which is perature in this range, showing only a small principally dependent upon the oxidation variation in 8 between about I. I to I .O. If state of the surface and the rates at which the oxidation state of the surface is an the intermediates form and react. important variable affecting the degree of EGASHIRA, KUCZKOWSKI, AND CANT randomization by affecting the ratio of 3. Larrabee, A. L., and Kuczkowski, R. L., J. sites, then temperature changes may not Catal. 52, 72 (1978). affect this ratio since the overall site cov- 4. Hucknall, D. J., “Selective Oxidation of Hydro- carbons,” p. 8. Academic Press, London/New erage remains nearly constant. York, 1974. In summary, various reaction mecha- 5. Herzog, W.,Ber. Bunsenges. Phys. Chem. 74,216 nisms involving intermediates such as I-VI (1970). can be incorporated into the stereo- 6. Wu, J. C., and Harriott, P., J. Catal. 39, 395 chemistry results. It is obviously difficult to (1975). 7. Brauer, G., “Handbuch der Preparation infer the precise surface intermediates con- Anorganischen Chemie,” p. 771. Ferdinand Enke trolling the stereochemistry in isotopic ex- Verlag, Stuttgart, 1954. periments which do not directly examine 8. Bailar, J. C., Jr. (Ed.), “Inorganic Syntheses,” the surface species. In particular, interpre- Vol. 4, p. 12. McGraw-Hill, New York, 1953. tation of the stereochemistry is hampered 9. Daniel, C., and Keulks, G. W., J. Cafal. 24, 529 (1971). by lack of answers to the following ques- IO. Czandema, A. W., J. Phys. Chem. 68, 2765 tions. (i) Is the small amount of ethylene (1964). isomerization independent of randomiza- Il. Kilty, P. A., Rol, N. C., and Sachtler, W. M. H., tion in the epoxide or directly related to it? in “Proceedings, 5th International Congress on (ii) Does randomization take place before, Catalysis, Palm Beach, 1972” (J. W. Hightower, Ed.), Vol. 2, p. 928. North-Holland, Amsterdam, after, or during attachment of the oxygen? 1973. (iii) Is the temperature invariance of ran- 12. Janda, J., and Kluckovsky, P., Chem. Zvesti 20, domization brought about simply because 267 (1966); Chem Abstr. 65, 3817a. two independent processes happen to have 13. Metcalf, D. L., and Harriot, P., Ind. Eng. Chem. identical activation energies or as a conse- Process Des. Develop. 11, 478 (1972). 14. Kagawa, S., Kono, K., Futata, H., and Seiyama, quence of an unusual mechanism? Hence, T., Kogyo Kagaku Zasshi 74, 819 (1971). it is not easy to presently justify a strong 15. Scatturin, V., Bellon, P. L., and Salkind, A. J., J. preference among the various alternatives. Electrochem. Sot. 108, 819 (1961). Nevertheless, the randomization data for 16. Vol, Y. T., and Shishakov, N. A., Iz. Akad. ethylene are intriguing and provide infor- Nauk SSSR Otd. Khim. Nauk, 586 (1962); Chem. Abstr. 57, 4088e (1962). mation on the reaction mechanism, albeit 17. Tanaka, S., and Yamashina, T., J. Catal. 40, 140 shadowy, that should help experimentalists (1975). and theorists ultimately evolve a better 18. Voge, H. H., and Adams, C. R., Advan. Catal. 17, consensus on the details. 151 (1967). 19. Kilty, P. A., and Sachtler, W. M. H., Catal. Rev. ACKNOWLEDGMENTS Sci. Eng. 10, 1 (1974). The authors are grateful to the donors of the Petro- 20. Kagawa, S., Tokunga, H., and Seiyama, T., Ko- leum Research Fund, administered by the American gyo Kagaku Zasshi 71,175 (1968). Chemical Society, for support of the portion of this 21. Sato, N., and Seo, M., Denki Kagaku 38,649,768 work conducted at the University of Michigan. The (1970); 39, 623 (1971); J. Catal. 24, 224 (1972). National Science Foundation (Washington, D. C.) 22. Gerei, S. V., Kholyavenko, K. M., and Rubanik, provided funds to purchase and maintain the micro- M. Y ., Ukr. Khim. Zh. 31, 449 (1965). wave spectrometer used in this work. 23. Force, E. L., and Bell, A. T., J. Catal. 38, 440 (1975); 40, 356 (1975). 24. Kenson, R. E., and Lapkin, M., J. Phys. Chem. REFERENCES 74, 1493 (1970). 1. Richey, W. F., J. Phys. Chem. 76, 213 (1972). 25. Carter, J. L., Yates, D. J. C., Lucchesi, P. J., 2. Cant, N. W., and Hall, W. K., J. Catal. 52, 81 Elliott, J. J., and Kevorkian, V., J. Phys. Chem. (1978). 70, 1126 (1966).