Sulfur poisoning and regeneration of palladium based catalysts

The physicochemical and catalytic properties of palladium deposited on alumina and silica-alumina have been studied after poisoning by hydrogen sulfide as well as at various stages of regeneration under hydrogen. The amount of free palladium surface atoms was mainly deduced from infrared measurements of adsorbed carbon monoxide. The catalytic activity was measured in the cyclohexane dehydrogenation. As far as the steady-state activities are concerned, a regeneration under hydrogen between 673 and 773 K leads to samples having either the same activity (alumina support) or a higher activity (silica-alumina support) compared to that of the unpoisoned catalysts. Nevertheless, in both cases, the initial metallic surface was not recovered. Pulse experi­ ments have been used for the determination of the initial activity. They showed that the deactivation by carbon­ aceous deposits is hindered when sulfur is present at the surface of palladium particles. Such deactivation is favoured when palladium is supported on an acidic carrier.

The physicochemical and catalytic properties of palladium deposited on alumina and silica-alumina have been studied after poisoning by hydrogen sulfide as well as at various stages of regeneration under hydrogen. The amount of free palladium surface atoms was mainly deduced from infrared measurements of adsorbed carbon monoxide. The catalytic activity was measured in the cyclohexane dehydrogenation. As far as the steady-state activities are concerned, a regeneration under hydrogen between 673 and 773 K leads to samples having either the same activity (alumina support) or a higher activity (silica-alumina support) compared to that of the unpoisoned catalysts. Nevertheless, in both cases, the initial metallic surface was not recovered. Pulse experi ments have been used for the determination of the initial activity. They showed that the deactivation by carbon aceous deposits is hindered when sulfur is present at the surface of palladium particles. Such deactivation is favoured when palladium is supported on an acidic carrier.
In a previous paper, 1 the dehydrogenation of cyclohexane was studied on palladium supported over alumina and silica alumina in the presence or absence of sulfur-containing com pounds, i.e. thiophene and hydrogen sulfide. Decon tamination by the feed free from sulfur as well as regener ations under hydrogen were performed. The poisoning effect was found to be of the same order of magnitude when using thiophene or hydrogen sulfide. In the case of alumina supported noble metals the thiotolerance as well as the decontamination were smaller for palladium than for plati num. Nevertheless, the regeneration under hydrogen was more effective for Pd than for Pt. When Pd was supported on silica-alumina instead of alumina, the thiotolerance was strongly improved. Moreover after the decontamination step or regeneration under flowing hydrogen, the activity was higher than the activity measured on the fresh catalyst.
In the present paper we examine more precisely the state of the palladium supported solids after poisoning by hydrogen sulfide and regenerations under hydrogen at various tem peratures. The free metal surface was determined either by the catalytic activity towards cyclohexane dehydrogenation using a feed free from sulfur or by the absorbance of the infrared bands of carbon monoxide adsorbed onto the avail able palladium surface atoms. After the various treatments the sulfur coverage of the palladium phase was determined after an oxidative treatment leading to the transformation of sulfur bonded to palladium into sulfate groups linked to the alumina support. The amount of sulfate species was deduced from infrared spectroscopy measurements.

IR Spectroscopy
The samples were pressed in order to obtain thin discs of known weight. The discs were placed in a sample holder made of quartz and introduced into a cell that allowed in situ treatments. IR spectra were recorded at room temperature on a Fourier-transform spectrometer (1.F.S. 110 from Bruker) with a resolution of 4 cm -1• Every spectrum was the sum of 100 scans. The plot function was set to absorbance.
Two kinds of spectra were obtained : (i) the IR spectra of CO adsorbed at 298 K on the freshly reduced solids and (ii) the IR spectra of CO adsorbed on the solids poisoned by After evacuation, the absorbance of the band at 1375 cm-1, due to sulfate groups bonded to alumina, allows the determi nation of the coverage of palladium by sulfur.

Pd/Al203
After hydrogen reduction at 673 K, the dispersion of the metallic phase is equal to 38% and corresponds to a mean particle diameter close to 3.0 nm. The dispersion values determined by H2 chemisorption or by H2-02 titrations agree well with the dispersion deduced from electron micros copy measurements. The reduction at 773 K leads to a dis persion of 35% (mean particle diameter of 3.  to 13% after reduction at 673 K and to 9% after reduction at 773 K (Table 1). Among the B bands associated to multi coordinated CO, the 1980 cm-1 band is clearly predominant (62% of the absorbance of the LF bands). The overall inte grated corrected absorbance which is normalized taking into account the weight of palladium is close to 0.82 (arbitrary units) after reduction at 673 K and 0.83 after reduction at 773 K (Table 1) IR Spectra of CO adsorbed on Sulfur-poisoned Pd/Al203 Samples The treatment by H2S at 373 K followed by a regeneration under hydrogen at various temperatures produces changes in the spectrum of adsorbed CO (Fig. 2).
The band corresponding to CO linearly bonded to palla dium is shifted towards higher wavenumbers. For instance, after a generation at 573 K, the v(CO) band is located at 2080 cm -1• After a regeneration at 673 K, two bands at 2080 and  It has already been reported that the M-S bond is partially ionic leading to an electron transfer from Pd atoms to S adatoms. 1 5 The position of the LF bands is strongly modified (Fig. 2) with a downwards frequency shift, from 1980 to 1925 cm-1 and from 1930 to 1855 cm-1 when going from the fresh solid ( Fig. 1) to the solid poisoned by H2S and regenerated at 573 K (Fig. 2). This phenomenon might be explained by a decrease in the CO coverage.

Catalytic Activities of Fresh and Poisoned Pd/Al203
The catalytic activities of the samples poisoned by H2S and  (Table  3). For the freshly reduced solid the decrease in activity observed between 1 min on stream and the steady state is equal to ca. 4-5%. It is attributed to carbon deposition. Fig. 3 gives the percentage of regeneration as a function of the temperature of hydrogen treatment. This percentage is expressed by the ratio of the steady-state activity of the regenerated solid to that of the freshly reduced one. After poi soning by H2S and treatment with H2 at 543 K, the regener ation increases from 12 to 25% between 1 and 180 min on stream. For the regenerations at higher temperatures the activity is stable after 1 min. After treatment with H2 at ca. 650 K the catalyst recovers its activity Ast in spite of the presence of remaining adsorbed sulfur and the fact that the initial metallic surface accessible to CO is not fully restored. The initial activity A0 is not re stored. The regenerations under flowing hydrogen after the  Pd/Si02-Al203 The palladium dispersion values determined by hydrogen chemisorption (Benson or Aben's methods) are ca. 13  After reduction at 673 K the sulfur, chlorine and carbon contents are 0.12, 0.21 and 0.16 wt.%, respectively.

IR Spectra of CO Adsorbed on the Freshly Reduced
Pd/Si02-Al203 The IR spectrum of CO irreversibly adsorbed at 298 K on Pd/Si02-Al203 reduced at 543 K [ Fig. 4(a)] is slightly differ ent from that observed for the alumina-supported palladium solid reduced at 673 or 773 K (Fig. 1). For instance, the HF band corresponding to linear adsorbed CO is broad, between used to achieve a low conversion at 543 K and to avoid diffu sion, the accuracy of the measurements is lower than in the case of the Pd/ Al203 solid. The activity measured at 1 min on stream is especially inaccurate; it is known to within 10% (reduction at 673 K) and 16% (reduction at 543 K) ( Table 3). is more efficient than the treatment by H2 alone. It is likely that the regeneration at 543 K would be very slow and conse quently not achieved after one night of hydrogen treatment.

Deactivation during the First Periods of Reaction with Sulfurjree Mixtures
The pulse experiments allow one to determine the extent of deactivation which is occurring during the first moments on stream with the sulfur-free mixtures. Carbon deposits are formed as soon as the solids are contacted with the cyclohexane-hydrogen mixture since the amount of benzene formed decreases strongly between the first pulse and the 240th pulse. This latter corresponds to the amount of cyclo hexane fed to the catalyst after 1 min on stream. For the following pulses, the activity decrease is less pronounced (Fig. 6).
The activities A0 determined after 1 min on stream differ strongly from the initial activities, especially for the fresh solids not contacted with H2S. For the fresh solid reduced at 543 K the ratio R of the activities measured at the 240th pulse and at the first pulse reaches 0.55. This ratio R becomes equal to 0.71 when the fresh solid is reduced at 673 K, i.e.
when the poisoning of the metal particles by the hydrogen sulfide arising from the reduction of the sulfate impurities of the support increases. The deactivation is still less pro nounced after treatment by H2S. The ratio R becomes equal to 0.91 for the solid reduced at 543 K, treated by H2S at 373 K and regenerated under hydrogen at 543 K. It is equal to 0.83 for the solid reduced at 673 K, poisoned and regenerated under hydrogen at 573 K.
The deactivation by carbon residues is less pronounced in the presence rather than the absence of preadsorbed sulfur.
Sulfur atoms remaining adsorbed on the metal particles inhibit the formation of carbonaceous residues since their for mation requires an ensemble of several surface atoms.18-2 1