Catalyst Characterization by Adsorption of Petroleum Asphaltenes in

---

Ind. Eng. Chem. Prod. Res. Dev. 1980, 79, 71-75

71

Catalyst Characterization by Adsorption of Petroleum Asphaltenes in Soht ion Jacques Saint-Just Gulf Research & Development Company, Pittsburgh, Pennsylvania 15230

Dilute toluene solutions of Kuwait atmospheric residue have been contacted at ambient conditions with a series of hydrotreating catalysts. A fraction of the residual molecules is preferentially adsorbed by the catalyst. The process which has been followed by visible spectrometry is controlled by diffusion, catalyst pore mouth plugging, and adsorption equilibrium. An empirical catalyst characterization method is presented which, in view of its sensitivity to catalyst outer morphology, supplements the data obtained by nitrogen adsorption. I n particular the existence of a “skin” on a catalyst extrudate can be readily detected.

Introduction The development of catalysts for hydrocarbon processes proceeds along chemical lines as well as physical aspects of utilization and efficiency. The heavier the feed, the heavier becomes the emphasis on catalyst structural characteristics in order to control the diffusional resistances. In residue HDS, the combination of optimum pore size and particle size compatible with the other constraints of the process is essential to the design of a successful catalyst. The reactant diffusion rate is determined by the properties of the diffusing molecules and by the porous structure of the catalyst. Nitrogen adsorption isotherms are used to obtain a pore size distribution from which, in principle, relative diffusion rates could be estimated. However, the technique is subjective and inaccurate. It may even be misleading; e.g., the presence of a “skin” on a catalyst particle (which may have developed during the extrusion process) may not be revealed by the adsorption isotherms while it may have a strong influence on catalyst behavior. The same conclusions also extend, to a lesser extent, to mercury porosimetry. It is therefore worth investigating whether relative diffusion rates a t HDS reaction conditions could be estimated from measurement of diffusion rates a t ambient conditions. Prasher and Ma (1977) concluded that the effective molecular diffusivity is strongly influenced both by the ratio of the diffusing molecules to the catalyst pore size and by the adsorption coefficient. Their conclusions originate from diffusion and adsorption experiments with several low molecular weight solute solvent systems (e.g., hexane in benzene), and extension to complex systems involving asphaltenic molecules is questionable. Alpert et aI. (1971) presented evidence that catalyst performance in residue hydrotreating could be satisfactorily predicted by a parameter measuring the penetration into catalysts under ambient conditions of residue molecules dissolved in benzene. They reported correlations between the “penetration number” and catalyst efficiency and life. The “penetration number” was estimated by microscopic inspection of the cross-sectioned catalyst. In view of the potential of this method and the intriguing question of why the pattern over months of a chemical reaction run a t high pressure and high temperature could be predicted by such a simple test, we decided to further investigate the diffusion of asphaltenes in so-

* Gaz de France, Direction des Etudes et Techniques Nouvelles, B.P. 33, 93212 La Plaine Saint-Denis, France. 0196-4321/80/1219-007 1$01 .OO/O

lution, in relation to catalyst features. Another incentive to start this investigation was the observation of a decrease in the intensity of the color of the solution after catalyst contacting and the consequent possibility of following easily and accurately the penetration by visible spectrometry. Experimental Section The change in color of the residue solution is followed by spectrometry. The spectrometer (Bausch & Lomb, Spectronic 710) operates a t a wavelength of 550 nm. An amount of 50 cm3 of a 3 w t % Kuwait long residue (53% reduced crude) solution in toluene is contacted a t room temperature with 10 g of catalyst contained in a 15-mm i.d. glass tube. A recirculation system, described in Figure 1,is used to maximize mass transfer and catalyst wetting while minimizing catalyst attrition. A spectrometer flow cell which is inserted within the recirculation line allows the concentration of the recirculated solution to be followed continuously. The spectrometer flow cell (Hellma cell no. 170QS)has a 1-mm light path. The experimental setup is built and operated (recycling rate of 210 cm3/min) so as to minimize the response lag which exists between events occurring in the catalyst bed and spectrometer reading. Under typical conditions, the response lag is of the order of a few seconds. As a consequence of the forced flow of the toluene solution through the constrictions formed by the catalyst bed and the supporting glass wool, toluene bubbles are formed. These bubbles were causing an erratic reading of the absorbance by the spectrometer. This problem was solved by inserting a small glass reservoir after the reactor to trap the bubbles. The choice of using 50 cm3of solution for 10 g of catalyst is the best compromise between high accuracy and sound experimental procedure. As the ratio of the amount of solution to amount of catalyst decreases, the change in solution color increases and the relative error on the measurements decreases; but in bringing this ratio down, one is limited by the fact that the catalyst particles must always be surrounded by a liquid phase to ensure good flow recycling and identical particle wetting. An amount of 50 cm3 of solution was found to be a safe compromise to ensure that these conditions were always satisfied even in the case of very absorbent materials. Since the observed phenomena are of transient nature, the following operating procedure must be respected in order to obtain reproducible results: the recirculation system is first flushed with pure toluene and subsequently 0 1980 American

Chemical Society

72

Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 1, 1980

AI

VlTON TUBING

SPECTROPHOTOMETER

Figure 1. Recirculating reactor. ABSORBANCE

r-----l

IT

1 k -

SOLUTION BEFORE CONTACTING

2 0.8

I

I 5

I

10

TIME

UNITS

Figure 3. Relationship between solution absorbance and vanadium content.

I 15

I

I

20

MINUTES

Figure 2. Decrease in absorbance of a solution of residue after contacting with a catalyst.

evacuated, 50 cm3 of the 3% residue solution is introduced and recirculated, 10 g of catalyst is then poured in the reactor. A typical recorder trace, shown in Figure 2, shows the respective absorbance readings in these successive steps. The absorbance a t 550 nm of the residue solution has been correlated with its vanadium content. The vanadium content has been independently determined by X-ray fluorescence for solutions whose initial absorbance was modified by recirculation over different catalysts. Solutions of Ceuta long residue of high vanadium content have been used for this purpose, in order to obtain solutions of measurable concentration, i.e., above 1 ppm of vanadium.

Discussion Significance of the Absorbance at 550 nm. In general, the light absorbance of a residue a t 550 nm cannot be quantitatively related to any specific characteristic of the residue. However, it has been observed that the vanadium content of the demetallized fractions of a given residue could be determined from the absorbance in the 550-nm region for both solvent deasphalted residues (Funk and Gomez, 1977) and hydrotreated residues (Saint-Just, 1978). Since we were observing a continuous decrease in the absorbance of the residue solution at 550 nm upon recirculation over a catalyst, we measured independently by X-ray fluorescence the vanadium content of the initial and recirculated solutions and found that indeed demetallization was occurring. Similarly to what has been observed for soIvent deasphalted and hydrotreated residues, the vanadium content can be determined from the absorbance a t 550 nm (Figure 3). The vanadium in petroleum residues is mostly contained in the resin and asphaltene fractions. However, the amount of resins and asphaltenes removed cocurrently with the vanadium during the adsorption process has not been independently determined. Consequently, the ad-

E

Q U A R T Z 2 0 . 3 0 MESH A 1 16

-1

H 6 12MESH

’

i r 115A

i0

H 2 0 - 3 0 AND 3 0 . 4 0 MESH

5

10

15

20

TIME, M I N U T E S

Figure 4. Influence of catalyst particle size and pore radius on solution absorbance.

sorbing molecules will be referred to as “vanadium-containing molecules” in the remainder of the discussion, rather than asphaltenes or resins. The Adsorption Phenomenon. Drushel (1972) observed that the molecular weight of asphaltenes desorbed from a catalyst which had been contacted with a benzene solution of asphaltenes was lower than the molecular weight of the asphaltenes in the solution. He concluded that the largest asphaltenes were excluded from the catalyst pores. On the basis of this conclusion and in view of the fact that a fraction of the vanadium is contained in the largest asphaltenes, we expected to observe an increase in the absorbance of the residue solution after contacting with a catalyst. One notices in Figure 2 that the absorbance of the solution indeed increases initially when the solution is contacted with the catalyst. However, this phenomenon could simply arise from the presence of fines carried away from the catalyst during the initial contacting; when pure toluene is contacted with the catalyst, there is no initial surge in solution absorbance. The presence of fines can, therefore, be ruled out. Selective initial toluene penetration appears as the explanation for the initial absorbance surge; the catalyst acts as a membrane allowing initially only the smaller toluene molecules to penetrate, which results in an increase in asphaltene concentration of the fraction of the solution which has not been absorbed.

Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 1, 1980

Table I. Vanadium Index of Penetration for Catalysts of Different Pore Radius and Particle Size

I

73

I

particle size 20-30 mesh VIP,

1/32 in. VIP,

cat.

‘3

%

A B C D E F

20.7

G

H

1/16 in. VIP, 9% 4.2

5.4 12.81 26.6 26.7 84.0 90.2

7.2 13.4

r, A

area, m’ig

41 38 12 45 50 76 91 115

215 206 289 17 1 165 151 151 108

-

pore vol, cm3/g 0.44 0.39 0.61

8

VI

t 3

u 6

z a

0.41 0.57 0.69 0.62

The fact that this behavior is more clearly in evidence (Figure 4) with catalysts which have a larger proportion of small pores supports this interpretation. However, after the initial absorbance surge, Figure 4 shows that even with small pore catalysts such as A, the absorbance of the solution decreases with time. A decrease in solution absorbance reflects the disappearance of vanadium-containing molecules from the solution. Preferential adsorption of these molecules must therefore occur on the external surface of the catalyst and on a fraction of the internal surface; the kinetic dependence of the phenomenon on pore radius (Table I) proves that vanadium-containing molecules do penetrate the catalyst pores and that the process is diffusion limited. No decrease in solution absorbance is observed on a poreless material such as quartz (Figure 4). The dependence on catalyst particle size is apparent in Table I. When catalysts of very large pore radii are used, the residue solution is very rapidly depleted of its vanadium suggesting that the adsorption step is very rapid and that the adsorption equilibrium between vanadium-containing molecules adsorbed and in solution is strongly shifted toward the adsorbed state. In their investigation of the restricted diffusion in binary liquid systems within fines pores, Satterfied et al. (1973) have observed that aromatic hydrocarbons tended to adsorb preferentially in a larger proportion as the number of aromatic rings in the molecule increased. The preference of vanadium-containing molecules to be in the adsorbed state in the presence of a strong solvent such as toluene would explain why a catalyst which has been used once in an asphaltene adsorption experiment could not be completely regenerated by toluene washing. A peculiar characteristic of the adsorption process of the vanadium-containing molecules suggests that multilayer adsorption occurs: when adsorbents of different particle size originating from the same catalyst which has been crushed to different mesh sizes are contacted for long periods of time (72 h), the quasi-equilibrium level which is reached is strongly dependent upon the size of the particle (Figure 5). Approximately the same equilibrium level is expected to be reached ultimately when adsorbents differing only by the particle size are contacted with the residue solution if diffusion only is interfering with adsorption. Since different equilibrium levels are obtained, an additional phenomenon must be present. Pore mouth plugging caused by multilayer asphaltene/resin adsorption is a possibility. Supplementary evidence for the occurrence of pore mouth plugging is the fact that very little increase in diffusion rate is observed when fresh solution is added to a used solution which has reached its quasi-equilibrium concentration, the catalyst being still in contact with the used solution. The pore mouth plugging occurs so fast that the influence of the total surface area is not felt, as practically only the outer pore volume and surface area of the

P U

n

I

L

E 20 -30 MESH

1

I

0

10

I 20

I

I

I

30

40

50

60

T I M E , HOURS

Figure 5. Influence of catalyst particle size on equilibrium absorbance. Table 11. Influence of t h e Metallic Impregnation o n t h e Vanadium Index of Penetration

VIP, % bare alumina with metal oxide impregnation with metal chloride impregnation

50 70 25

catalyst are involved in the adsorption. It is likely that for some combination of small particle size and large average pore radius, diffusion limitations disappear and surface area alone becomes the limiting factor in the adsorption process. An example is given by adsorbent H (Figure 4); the same equilibrium absorbance level is reached for the 2G30 mesh and the 30-40 mesh particles. The diffusion limitations are not severe and the surface area is fully utilized before the occurrence of pore mouth plugging. Another feature of the adsorption process is its sensitivity to the chemical nature of the adsorbing surface. Table I1 shows that an alumina support impregnated with a metal oxide adsorbs faster than the bare support. This increased adsorption capability reflects an equilibrium shift toward the adsorbed state: as mentioned above, the diffusion limitations are not severe for adsorbent H and the equilibrium level is reached rapidly after contacting with the residue solution; when the solution is contacted with the same adsorbent H impregnated with a metal oxide, more vanadium-containing molecules are adsorbed a t equilibrium. The equilibrium absorbances are 0.075 and 0.035 for each case, respectively. The results should be interpreted cautiously; it is apparent in Table I1 that with a metal chloride impregnation, the amount of adsorbed vanadium-containing molecules is lower. However, it is not possible to decide whether adsorption has been chemically hindered in the presence of chloride, or penetration has been physically hampered by chloride deposits, or the porous structure of the support has been modified by the impregnation. Other analytical methods could provide an answer, but experiments with materials having no pore structure may also help choosing between the alternative explanations by allowing to uncouple chemical and structural effects. Finally, a few experimental observations on the influence of the asphaltene solvent have been made. Inspection of Table I11 shows that adsorption is higher in toluene than in cyclohexane while very little adsorption occurs in pyr-

74

Ind. Eng. Chem. Prod. Res. Dev., Vol. 19, No. 1, 1980

Table 111. Vanadium Index of Penetration for a n Alumina Support in Several Solvents solvent

VIP, %

cyclohexane toluene pyridine

51.5 80.2 23.8

~~

idine. It is likely that the solvent determines the values of the equilibrium between the amount of adsorbed molecules and the amount in solution. Also, the diffusion limitations may vary with the solvent as it is known that asphaltenic molecules in solution associate to an extent which depends on the solvent (Moschopedis et al., 1976; Schwager et al., 1977). The quantitative treatment of the data presented here is outside the scope of this paper. Satterfield et al. (1973) proposed an empirical correlation based on their observations on several binary systems among which are solutions of aromatic hydrocarbons in paraffins. Difficulties specific of the system discussed here include: the large number of types of diffusing molecules with the estimation of appropriate diffusion and adsorption coefficients; the presence of what appears as a pore mouth plugging phenomenon; and wide and uneven pore size distributions. Catalyst Characterization. In view of the many variables influencing the diffusion process of vanadiumcontaining molecules, it is clear that this test, used by itself, cannot predict with certainty catalyst performance. However, it may prove valuable when used in conjunction with other catalyst characterization techniques. In order to obtain a classification of catalysts based on their efficiency for the diffusion process, it has been found meaningful to use a dimensionless quantity, the Vanadium Index of Penetration (VIP) defined as

VIP = 1 -

VZO "0

where V , and Vo are the vanadium content of the solution after 20 min of contacting and a t time 0, respectively. Although a longer contacting time increased the accuracy with which the result of a single run could be obtained, repeatability was such that there was no point to extend that time beyond 20 min. Since there is a linear relationship between the solution absorbance and its vanadium content, the VIP can be directly obtained from the absorbances Azoand Ao, respectively

where 1 is the thickness of the spectrometer cell and tm and co are the solution extinction coefficients a t time 20 and 0, respectively. The absorbance a t time 0 is obtained by extrapolation of the adsorption curve (Figure 21, as the curve in the first minute of contacting is not experimentally well defined. At this point, only correlations with another characterization method, namely nitrogen adsorption, have been drawn. Table I shows the VIP for several catalysts and the corresponding main data computed from the nitrogen adsorption characteristics. Meaningful comparisons between catalysts can only be made between catalysts of the same particle size or between particles of different size originating from the same catalyst. Caution should be exercised if the chemical nature of the surface is different, since its influence on the VIP has not yet been systematically evaluated. Examples are given below which show that despite these limitations, the residue adsorption test

0

20

40

60 PORE

80

100

150

RADIUS ,

Figure 6. Pore size distributions for catalysts B and C.

may provide information that would be difficult to obtain by other techniques. It appears that, in general, the VIP increases with the average pore radius as determined by nitrogen adsorption (Table I). No quantitative agreement is expected since what is measured by the two techniques is different. For reasons formulated above, the VIP is thought to be very dependent on external surface morphology. The data computed from nitrogen adsorption measurements are certainly less sensitive to outer catalyst morphology, because nitrogen adsorption is an equilibrium technique and also because our computer program does not take into account catalyst pore shape. In that respect, the data obtained for catalysts B and C can be reconciled. The VIP of C is 2.4 times that of B while the nitrogen average pore radii are similar (37.5 and 42 A, respectively). The discrepancy probably could not be explained by the existence of different pore size distributions although C has indeed a slightly higher concentration of larger pores (Figure 6). Pore shape considerations as expressed below for catalyst A are more likely to account for the discrepancy. It can therefore be inferred from comparison of the VIP that, despite similar average pore radii, B may be less efficient than A in diffusion-limited processes and more susceptible to pore mouth plugging. Residual HDS pilot plant data which were obtained subsequently confirmed this conclusion. When catalysts of different particle size are compared, not only the sensitivity of the residue adsorption process to the amount of catalyst external surface area appears, but anomalies may be revealed. An example is given by the comparison of catalysts A and E. The catalysts have been compared as received, i.e., as 1/16-in.extrudates and as 20-30 mesh crushed extrudates. The VIP of the E crushed extrudates is twice the VIP of the E 1/16-in.extrudates. In contrast, the A crushed extrudates are five times easier to penetrate than the A 1/16-in.extrudates. Before drawing any conclusions, the increase in amount of external (geometric) area which occurs during the crushing operation must be evaluated. This increase is about threefold and the proportional increase by itself in the number of open pores cannot therefore account for the difference in ease of penetration between the crushed and f ,6-in. A extrudates. A fivefold increase means that, as a result of the crushing operation, pores of larger diameter have become available. The A extrudates are, therefore, likely to have a skin. The existence of a skin appears as the only explanation compatible with the observations concerning the diffusion of vanadium-containing molecules

75

Ind. Eng. Chem. Prod. Res. Dev. 1980, 19, 75-77

and the data gathered by nitrogen adsorption. On the other hand, this anomaly does not appear to affect catalyst E. A VIP ratio of 2 for 20-30 mesh particles and 1/16-in. extrudates is between the ratio of 3 which indicates that possibly only the external surface is involved in the adsorption and the ratio of 1 which is characteristic of a non-diffusion-limited process on a porous catalyst of large internal surface area.

Drushel, H.V., Am. Chem. SOC.,Div. Pet. Chem. Prepr., 17,No. 4 (1972). Funk, E. W., Gomez, E., Anal. Chem., 49, 972 (1977). Moschopedis, S. E., Fryer, J . F., Speight, J . G., Fuel, 55, 227 (1976). Prasher, B. D., Ma, Y. H., AIChE J . , 23. 303 (1977). SaintJust, J.. Anal. Chem.. 50, 1647 (1978). Satterfield, C. N., Cotton, C. K.. Pitcher, W. H., Jr., AIChE J., 19, 629 (1973). Schwager, I., Lee, W. C., Yen, T. F., Anal. Chem., 49, 2363 (1977).

Literature Cited

Presented before the Division of Petroleum Chemistry, 176th National Meeting of the American Chemical Society, Miami, Fla., Sept 15, 1978.

Alpert, S. B., Wolk, R . H.. Maruhnic, P., Chervenac, M. C. (to Hydrocarbon Research), US. Patent 3630888 (Dec 28, 1971).

Received f o r review February 28, 1979 Accepted October 11, 1979

Liquid-Phase Oxidation of Ethylbenzene over Cobalt Complexes Supported by Polymeric Materials Ayaz

A. Efendiev,' Togrul N. Shakhtakhtinsky, Leila F. Mustafaeva, and Harry L. Shick

Institute of Theoretical Problems of Chemical Technology of the Academy of Sciences of the Azerbaijan SSR, 29, Narimanov Prospect, 370 143, Baku, USSR

The oxidation reaction of ethylbenzene proceeding in the liquid phase in the presence of complexes of cobalt with a specially arranged for cobalt sorption copolymer of diethyl ester of vinylphosphonic acid with acrylic acid cross-linked by methylenediacrybmidewas studied in a gasometric unit. The experiments were performed at atmospheric pressure in a temperature range of 70-135 O C in the absence of solvent. The catalyst amount ranged from 1.5 to 6 g/L, which corresponds to a cobalt content of 6.45 to 25.8 mequiv/L. The catalyst can be used repeatedly without losing its catalytic activity. The influence of the amount of catalyst and the temperature on the reaction has been studied. The activation energy of the oxidation reaction appears to be 10 kcal/mol.

Introduction The last few years have witnessed a growing interest in the preparation and use of catalysts on the base of transition metal complexes with cross-linked polymers having complex forming groups. Such catalytic systems are of particular interest because they can combine the advantages of heterogeneous catalysts, such as simplicity of separation from the reaction mixture and stability. as well as homogeneous catalysts, such as high selectivity and possibility of handling more exact information about active centers structure. At the Institute of Theoretical Problems of Chemical Technology of the Academy of Sciences of Azerbaijan SSR (USSR) we have developed a method of preparation of transition metal complexes with specially arranged polymers. The general principle of our method involves interaction of linear polymer and ions to be sorbed in solution, Le., under conditions when macromolecules' segments are still mobile enough, subsequent fixation of optimal for the ion uptake conformation of macromolecules by cross-linking of metal-polymer complexes, and removal of ions from cross-linked sorbent. As a result of such treatment, we managed to improve essential sorption characteristics of cross-linked sorbent, namely its capacity, selectivity, and sorption kinetics. As an object for such treatment we used a copolymer of' diethyl ester of vinylphosphonic acid and acrylic acid 0196-4321/80/1219-0075$01.00/0

(Efendiev et al., 1977; Kabanov et al., 1974) containing 9.2% by weight of phosphorus having molecular mass of 160000. Cross-linked sorbents on the base of that copolymer could form complexes with ions of copper, cobalt, and nickel. Experimental Section Diethyl ester of vinylphosphonic acid was prepared in accordance with the method of Kolesnikov et al. (1959), and after double distillation the product with bp 62 "C (2 torr), nmD1.4300 (lit.bp 68-70 "C (3 torr), nmD1.4300) was isolated. Conventional acrylic acid was distilled twice in vacuo (bp 39 "C (10 torr) before use. Bulk copolymerization of the monomers was carried out by photochemical initiation of the mixture of monomers with 1 wt % cumene hydroperoxide in a sealed test tube of quartz glass which was evacuated at lo4 torr. UV light was supplied from a high-pressure mercury lamp of 300 W. The resulting copolymer was dissolved in ethanol and then precipitated dropwise in an excess amount of diethyl ether. After three reprecipitations the copolymer was dried in a vacuum desiccator a t only a slightly elevated temperature. A fairly dilute ethanol solution of copolymer (0.8 g/100 cm3) was mixed with a double amount of 0.075 M solution of cobalt chloride salt adjusted to pH 1.1with HC1. The resulting mixture was then slowly titrated by ammonia solution until a pH value of 5.20 was achieved. The rate of titration was so chosen that the p H increase was not 1980 American Chemical Society