I n d . E n g . Chem. R e s . 1987, 26, 1312-1323
1312
Supplementary Material Available: Tables containing the F and D matrices for different sensitivity-typemixtures and predictions of constant mass expansion experiment for mixtures 1, 2 , 4 , and 6, using three of the experimental data points for tuning (7 pages). Ordering information is given on any current masthead page.
Nomenclature a = SRK parameter defined in eq 1 b = SRK parameter defined in eq 1 DIk= element i k of the D matrix F,, = element i j of the F matrix k,, = binary interaction coefficient K , = equilibrium constant for component i m = SRK parameter defined in eq 8 p , = property (see eq 12 and 13) P = pressure R = gas constant T = temperature u = volume y = mole fraction of component i 2 = compressibility factor
Literature Cited Coates, K. H.; Smart, G. T. “Application of a Regression Based EOS PVT Program to Laboratory Data”, 57th Annual Fall Technical Conference of the Society of Petroleum Engineers of AIME, New Orleans, 1982; SPE 11197. Gani, R.; Fredenslund, Aa. Fluid Phase Equzlibr. 1986, 29, 575. Pedersen, K. S.; Thomassen, P.; Fredenslund, Aa. Ind. Eng. Chem. Process Des. Deu. 1984, 23, 566. Pedersen, K. S.; Thomassen, P.; Fredenslund, Aa. Ind. Eng. Chem. Process Des. Deu. 1985a, 24, 948. Pedersen, K. S.; Thomassen, P.; Fredenslund, Aa. On the Dangers of Tuning Equation of State Parameters; Instituttet for Kemiteknik: DTH, Denmark, 198513; SEP 8501, to be published. Peneloux, A,; Freze, R. Fluid Phase Equilibr. 1982, 8 , 7. Reid. R. C.; Prausnitz, J. M.; Sherwood, T. K. T h e Properties of Gases and Liquids; McGraw-Hill: New York, 1977. Soave, G. Chem. Eng. Sci. 1972, 27, 1197.
Greek Symbols a = parameter defined in eq 9
a, = fugacity coefficient of component i
W,
= acentric factor of component i
Subscripts i, j , k = component index, property index
c = critical property Superscripts
Receiued for review December 9, 1985 Revised manuscript received March 20, 1987 Accepted April 10, 1987
1 = liquid phase v = vapor phase
An Investigation of the Deactivation Phenomena Associated with the Use of Commercial HDS Catalysts J a m a l M. Ammus, George
P.Androutsopoulos,* a n d Athena H. Tsetsekou
Department of Chemical Engineering, National Technical University of A t h e n s , G R 106 82 A t h e n s , Greece
Catalyst deactivation data, obtained from the HDS of a Greek oil residue on a spinning basket laboratory scale reactor, are reported in this paper. Two commercial grade hydrodesulfurization (HDS) catalysts, viz., Harshaw HT-400 E and ICI-41-6, were used in the investigation. Initial activities and modes of activity loss vs. run time are compared (HT-400 E possessed an initial activity 25% higher than that of ICI-41-6; activity was reduced to 55-65% for the former and 70% for the latter catalyst after 100 h of operation). Coke and metal deposits built up vs. run time are also reported here (75% of the total amount was deposited within the initial 30 h of catalyst operation). C/H ratios and N2 content of coke deposits increased proportionally with operating time. Pore volume and surface areas (5% of initial value after 30 h of operation) respectively reduced for HT-400 E to 43-57 and 40-45 and for ICI-41-6 to 60-70 and 40-45. Pore structure recovery parameters following reactivation ranged between 80% and 95% of their respective initial values. The role of catalyst particle size upon the rate and the extent of activity reduction is also quantified. Experimental data showing the dependence of the catalyst pellet coke content vs. the nominal coke deposit thickness were fitted by using a random corrugated pore model, and the effect of particle size was satisfactorily quantified. Literature Survey General Aspects. It has been reported in recent articles (Hirotsugu et al., 1980; Kodama et al., 1980) that deactivation of oil hydrotreating catalysts follows a characteristic mode. The average catalyst bed temperature should be raised to compensate for catalyst activity losses. Temperature-time curves typical to several hydroprocessing applications appear to have an “S”shape form, being independent of the physical and chemical properties of both catalyst and oil feedstock, although the elementary deactivation mechanism may not always be the same. Hirotsugu et al. (1980) suggested that catalyst life is drastically affected by the physicochemical characteristics
of the catalyst and the oil feedstock. The analysis of used HDS catalysts led to the widely believed view that activity loss can be attributed to carbonaceous and metal deposits, the chemical changes of catalyst active surface, and the modification of the physical structure properties of the employed catalysts. The contribution of each of these factors depends on the reaction conditions, the origin of the oil feedstock, the desired degree of desulfurization, and the catalyst physicochemical features. Coke Deposition and Properties. Coke deposition in the catalyst structure may impose intraparticle diffusion limitations to reactant molecules, either via a partial or
0888-5885/87/2626-1312$01.50/0 0 1987 American Chemical Society
Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987 1313 a complete pore plugging mechanism. A corrugated (series) random pore model reported by Androutsopoulos and Mann (1978) was used to develop general mathematical expressions for foulant deposits profiles, foulant pellet content, and accessible surface area vs. foulant layer thickness. The main parameters in the latter mathematical model are the pore segment number distribution function with respect to pore size and the length of the corrugated pore (number of cylindrical segments, of distributed size, per pore). Furthermore, coke containing N- and 0-basic compounds may affect catalyst activity by poisoning acidic active sites (Furimsky, 1979a). It is also accepted by a number of authors, e.g., Furimsky (1979b) and Ohutsuka (1977), that coke develops from reactive, unstable, hydrocarbon species (Coke precursors) with a small or large molecular size, e.g., resins, asphaltenes, and heterocyclic compounds. Coke formation is the result of precursors adsorption on the catalyst surface, followed by an extensive dehydrogenation, aromatization, and further depolymerization (Beuther and Schmidt, 1963; Furimsky, 1979a,b; Ohutsuka, 1977; Furimsky, 1978). Ocampo et al. (1978) reported that coke deposited on the surface of Co- or Ni-Mo/y-A1203 during HDS of liquified coal caused a rapid catalyst deactivation within a few hours of liquid processing. The same authors showed that the extent of pore structure change and activity loss depended on the catalyst physical structure and chemical composition. They also concluded that these catalysts, deactivated after processing for short periods, could be fully regenerated to nearly their virgin active state. Johnathon and Howard (1982) and Inogouchi (1976) reported that coke deposits in the relatively narrow pores in the range of pore diameters d, < 70 A. Ternan et al. (1979) found out that coke deposited during the initial stages of operation has not caused an appreciable activity decline of a Co-Mo/y-A1203 catalyst during HDS of Athabasca Bitumen (a feedstock free of metallic compounds). Metal Deposits and Their Properties. The role of coke deposits is limited to a reversible alteration of the physical properties of the used catalyst, while metal deposits are considered as the most critical factors threatening catalyst life itself, because they are irreversibly adsorbed on the catalyst surface. A controlred oxidation of deactivated catalysts can remove carbonaceous materials and also transform sulfided metallic deposits into their oxidic form, i.e., NiS, V2S3 NiO, V205. The most important metallic elements that affect catalyst activity that are frequently found in petroleum crudes are vanadium and nickel. It is believed that metallic sulfides deposited on the catalyst surface originate from organometallic compounds (metallic precursors), e.g., porphyrins, asphaltenes, and resins. The metallic deposits consist mainly of V,S, and NiS. The adsorption and reaction of the metallic precursors and the subsequent deposition of the metallic sulfides are the basic steps of catalyst deactivation by partial or complete pore plugging (Pazos et al., 1983; Hirotsugu et al., 1980, 1981; Cir and Cirova, 1979; Newson, 1970). Chemical Behavior of Catalysts during HDS. Two arguments seem to support the view that the really stable active HDS catalyst species are the sulfided metals of the catalyst, i.e., Mo,S, and Co,S,, as it is not possible to get reliable composition data on industrial catalysts. The first argument is that the chemistry of HDS is abnormal when oxidic precursors are introduced and progressively changes as sulfidation proceeds. The second argument is that carefully sulfided unsupported catalysts demonstrate a
-
behavior resembling that of industrial catalysts under normal operating conditions (de Grange, 1980). The Nature of the Sulfide Phase. The physicochemical catalyst characterization usually precedes the determination of sulfur and metals deposited, during use, on the active catalyst phase. The sulfur content of the catalyst depends both on the conditions applied during its preparation and activation (reduction - sulfidation) treatment. Taking into account these facts, it is reasonable to assume that under steady operating conditions, the active catalyst phase is deficient in sulfur despite the fact that H2S adsorbed on A1203is accounted for. This is probably due to the fact that a fraction of molybdenum is not completely sulfided, being transformed into many sulfided forms and the cobalt is trapped in the chemical form CoA1204(De Beer et al., 1976). According to assumptions made by Schuit and Gates (1973), molybdenum oxide (MOO,) is partially sulfided to such an extent that the maximum possible atomic ratio was S/Mo = 1 and the cobalt species was thought to be inaccessible to sulfur. De Beer et al. (1976) showed that for commercial Co-Mo/yA1203 catalysts (400 “C, H2S/ HJ, the degree of sulfidation was 95% (S/Mo atomic ratio = 2.44, S/Co = 0.63). Ripperger and Saw (1977) state that Co-Mo catalysts incorporate only 70-80% of the theoretically possible sulfur content which would correspond to the complete transformation of cobalt and molybdenum into Cogs8and MoS2, respectively, and argue in favor of oxysulfides.
Theory Quantitative Determination of Coke Deposited on HDS Catalysts. The deposition of coke and ash producing deposits on catalysts is a major problem in most organic reaction systems and expecially in the hydrotreating of heavy oil fraction processes. The evaluation of the real effect of the carbonaceous deposits upon both the pore structure and the intrinsic chemical activity of the reactive phase (in situ) requires a standard method of coke and metal determination. Up to now, there is no standard method for quantitative coke determination (Furimsky, 1979a,b; Ammus, 1985). HDS catalysts on discharging may contain the following species: unreacted oil reactants, oil products, intermediate products, metal deposits (mainly V and Ni), and finally coke. In the science of porous catalyst deactivation, activity and selectivity reduction effects can be attributed to coke and metal deposition. The reliable determination of coke requires (1)deoiling of the “as discharged” used catalyst to remove the soluble oily substances and (2) an accurate determination of the metal deposit level. It was stated by Furimsky (1979a,b) that the most effective method of deoiling is benzene extraction carried out on a Soxhlet apparatus. The catalyst deoiling and oxidation precede any coke deposit determination. The amount of coke deposited is usually evaluated by subtracting the weight of the dried catalyst, following benzene extraction, from the weight of the fully oxidized samples. For more accurate evaluations, coke deposits should be expressed per mass cf sulfided catalyst. As mentioned earlier, the active chemical species of the catalyst during use appear in their sulfided form. This coke evaluation based on a simple subtraction between catalyst weights of the used (sulfided) and fully oxidized forms is erroneous. An improved estimation of coke deposits should involve an “oxidation factor” that expresses the conversion of oxidic forms (present in the oxidized phase) to the sulfided form in the used phase. It is proposed in this article that coke quantitative determinations should be eventually ex-
1314 Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987
T = 350 "C; hydrogen partial pressure, P H z = 50 bar; gas volumetric flow rate, fIH, = 60 L/h; rate of catalyst rotation, w = 2900 rpm; duration of individual runs, ti = 10 h; liquid feedstock, Greek atmospheric oil residues; and bp, +315 "C. The reactor was operated in a semibatch mode. Hydrogen was flowing continuously through the reactor carrying with it hydrogen sulfide and hydrocarbon vapors. The oil was charged a t the start of a run and discharged after about 10 h of reactor operation, when the degree of sulfur compounds conversion had reached the desired level. Catalyst Treatment and Activation. The estimated amount of the selected HDS catalyst type and particle size was dried in a furnace at 400 "C for 24 h to remove traces of moisture. The dried sample was cooled in a desiccator, weighed, and subsequently placed in the reactor baskets. Glass wool was placed a t the top and bottom sides of the cubic baskets, made of a stainless wire mesh, to prevent the escape of catalyst particles from the basket. After the catalyst basket was fixed on the rotating disk, the autoclave was ready for operation. The careful inspection of the apparatus for probable hydrogen leaks was followed by the catalyst activation procedure, which included the following steps: (i) air and moisture removal from the reactor space under vacuum (-20 torr), at 300 "C for -30 min; (ii) catalyst reduction by passing a hydrogen stream, 6 L of H2/h, for 3 h at 250 "C and 8-bar pressure; (iii) catalyst sulfidation carried out in situ a t normal experimental conditions which included the charge of the appropriate amount of oil, a hydrogen flow rate of 60 L/h, a temperature T = 350 "C, and a rate of catalyst rotation w = 2900 rpm. The duration of this rum was 10 h. Initial HDS rates, under standard experimental conditions, were found to be for (i) catalyst HT-400 E, RHDs = 2.60 X lo4 gmol of S/(cm3of cata1yst.s) and (ii) catalyst ICI-41-6, RHDs = 2.05 X lo4 gmol of S/(cm3of catalystas). Raw Materials. The following raw materials were used in this investigation. Apart from hydrogen, a commercial grade (H, + O299.8% and air 0.2%) was supplied by Air Liquide Hellas and the liquid feedstock was Greek oil atmospheric residue (bp +315 "C) from Prinos Wells (Thasos). Properties of the oil residue are displayed in Table I. Two commercial grades of HDS catalyst differing in their physical structure and chemical composition were used in this study. Overall catalyst physical properties and chemical compositions are presented in Tables I1 and 111. Catalyst HT-400 E (Harshaw Chemical Co.) was used either in extrudate form (1/16 and '/* in.) or pulverized form (d, = 0.34 mm). Similarly catalyst ICI-41-6 (Imperial Chemical Industries, P.L.C., England) was used both in extrudate form d, = l/ls in. and pulverized form (d, = 0.34 mm). Analytical Instrumentation. (i) A nitrogen adsorption apparatus (Sorptomatic Model 1800, Carlo Erba Strumentazione, Milano, Italy) was used to carry out adsorption-desorption experiments on samples of fresh, used, and regenerated catalysts. Pore diameters above >400 A could not be detected by the adsorption technique.
Table I. Atmospheric Residuum Feedstock Characteristic Properties (Crude of Prinos, Thasos) urouertv Thasos atm residue 1.007 specific gravity, 15/4 "C, g/cm3 9 "API 5.6 S content, w t % 335 kinematic viscosity, 122 O F , cSt 14.3 C residue (Conradson), w t % 9" v, PPm 11" Ni, ppm +315 boiling range, "C 51 yield of crude, wt % Papayannakos and Marangozis, 1984.
pressed as grams of coke/grams of sulfided catalyst (CoMo/yAl,O,) + sulfided metal deposition (V3S5and NiS). The suggested form of coke evaluation bears a physical significance and is based on the assumption that both complete sulfidation and oxidation of the respective catalyst phases has occurred. A reduced error is anticipated to be achieved by applying the proposed technique (Ammus, 1985). Catalyst Activity Evaluation Procedure. To calculate the activity reduction of the crushed catalyst pellets, use is made of the two expressions (i) specific reaction rate constant
k = $&,/(1
+ KIPH,)
(1)
(ii) integrated form of mass balance over reactor volume
( l / S R ~ t-f )(l/S!~d))= ( n - l)kp~lWca,t/(3200"-'m,ilpcat) (2)
(Ammus and Androutsopoulos, 1987). The quantity tikv can be expressed as CikV = 320On-'~,i~hpCa~(l + KiPH,)/60P~,W~,,(n- l)p$l (3) [ &,]
= ~ m ~ ~ - s - ~ . b a r - ~ . g m o of l ~ catalyst)-3 -~(cm
The degree of activity loss is expressed by (tok, - €ikV)/(q+") = 1 - (ti/€()) = 9
(4)
where eo = initial intrinsic relative activity 1 and t i = mean intrinsic relative activity a t time t (min) corresponding to the ith experimental run.
Experimental Section Apparatus and Conditions of Operation. The experimental apparatus used in the deactivation tests was a spinning basket catalytic reactor of laboratory size (working volume V , = 888 cm3) equipped with four symmetrically arranged catalyst baskets of a total capacity of 33 cm3. A detailed description of the reactor design and the experimental setup as a whole is reported elsewhere (Ammus and Androutsopoulos, 1987). Deactivation experiments were carried out under the following standard experimental conditions: temperature,
Table 11. Catalyst Type, Chemical Composition, and Physical Properties
catalyst" HT-400 E-H08 HT-400 E-H109 ICI-41-6
chem compos., wt 70 COO MOO, Al,O, 3.0 15.0 82.0 82.0 3.0 15.0 3.3 14.0 82.7
S,, m2/g of catalyst ~
230' 230 250
cm3/g of catalyst
Vpore, ~
215' 218 253
133d 0.50' 123 0.50 190 0.56
0.44' 0.44 0.57
0.36d 0.34 0.50
'116
6.0
l1.e
6.0 5.5
Pobsd,eatr
g/cm3 1.30 1.25 1.27
Commercial designation. bProvided by the supplier. 'Determined by the N, adsorption method. = - Pu,bed/Pw.cat.
etbed
d,, in.
mean mechan., strength, kg
l/16
Pobsd,bed,
gIcm3 0.77 0.74 0.67
ebede
0.412 0.409 0.472
Determined by mercury porosimetry.
Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987 1315 Table 111. Most Probable Pore Diameter, Pore Region Detected, Pore Volume, and Pore Surface Area Distributions of Fresh Catalysts catalyst type HT-400 E ICI-41-6 applied technique physical properties most probable pore diameter, A pore region detected, A most probable pore diameter, A
Nz adsorption-desorption mercury penetration
pore vol distribution
surface area distribution
84 48-170 108
74 42-322 99
d,, A 75 75-100 100-160 160-150000
%
d,, A 75 75-100 100-160 160-150000
%
v, 46 48 6
v, 30 68 2
d,, A 75 75-100 100-160 160-150000
%
d,, A 75 75-100 100-160 160-150000
%
v, 56 33 10
v, 62 33 5
Table IV. HDS Catalyst Deactivation Experiments Performed on Samples of Greek Atmospheric Oil Residue" catalvst HT-400 E for groups 1-111 ICI-41-6 for groups I and II I I1 I11 I I1 pulverized extrudates pulverized extrudates set j 1st 2nd 3rd
-dp = 0.34n L, = 0.00567
3, = '/16* L, = 0.0326
L, = 0.0589
dp E 0.34" Lp = 0.00567
3, = ' / M b L, = 0.0294
3d
4 7 11
3 7 14
3 7 15
2 5 9
7 20
ndpin millimeters. b d , in inches.
cepin centimeters.
-d, =
'/:
dNumber of experimental runs.
(ii) A mercury porosimeter (Model 200 Carlo Erba Strumentazione, Milano, Italy) was used to carry out supplementary pore structure measurements. With a maximum applied pressure of 2000 bar, pore diameters d, > 75 A were measured. Mercury penetration-retraction cycles could be performed. (iii) Determinations of sulfur concentration were done according to the standard method IP-336178 by means of an X-ray fluorescence facility (Telsec Lab. X-100 Analex Company Ltd., England). (iv) A thermal balance (Simultane Thermal Analyse Model 429, NETZSCH Geraten, GmbH) was used in coke deposit determinations by continuously recording the weight of sample, while the combustion was in progress. (v) Metallic deposit determinations were carried out by means of an atomic absorption spectrophotometer (Varian, AA-775 series). Vanadium and nickel were measured on catalyst samples used for varying periods of time. (vi) An elemental analyzer (Hewlett-Packard Model 185) was used in this analysis of C, N, and H on deactivated catalyst samples. Deactivation Experiments. Catalyst deactivation experiments are classified into five groups, each one characterized by the catalyst type and particle size used to form the catalyst basket bed. Each group involved a number of sets of experiments. HDS experiments belonging to a particular set were performed on the same catalyst charge but on different oil batches. A single experiment lasted for about 10 h. The complete picture of the HDS experiments is shown in Table IV.
Results and Discussion Catalyst Activity Decline. It is evident from the results of standard deactivation experiments that both HDS catalysts used in their pulverized form follow similar patterns of chemical activity loss. During the initial stages of operation (t < 50 h), catalyst HT-400 E indicated a slight (8 = 5%) activity loss, at a
I0'5r I
0
0,4
0.6
/
0
m
0 0 (b)
20
40
80
60
Time
100
120
140
160
180
200
(h)
Figure 1. Activity-time curves: (a) HT-400 E and (b) ICI-41-6 catalysts; (0) pulverized, (W) extrudates 1/1,+, (A)extrudates */*-in.
gradually increasing rate, as Figure l a depicts. A high, though, continuously decreasing rate of deactivation was observed after the 50-h period, resulting in an almost 29 = 40% activity drop. The final deactivation period is characterized by a nearly constant rate and an increasing degree of activity reduction. The general deactivation mode of the pulverized ICI41-6 catalyst resembled that of catalyst HT-400 E and appeared to differ only in the duration of the respective time periods; i.e., the initial deactivation period lasted for
1316 Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987
only 20 h, as Figure l b shows. The evaluation of the degree of catalyst activity reduction of both catalysts used in the form of cylindrical extrudates can be deduced by applying an indirect procedure based on the use of the generalized Thiele modulus expressed by cp = L,(((l
+ n)ktCs"-1)/2Deff)'/2
(5)
where cp = generalized Thiele modulus, t,= characteristic length defined as the ratio of particle volume upon the external particle surface area, t = relative catalyst activity, n = intrinsic reaction order, k = specific reaction rate constant k = kJ'HJ(1 + K ~ P H J (6)
Deff= sulfur compound effective diffusivity (cm2/s),6s =
L
T
mean sulfur concentration (gmol of S/cm3 of oil). The observed HDS reaction rate expression is (RHDS)obsd
= kobsdCSnobd = qtkCSn
(7)
Thus Ttk = kobsdCSnob"-n
(8)
then vt can be calculated if the intrinsic reaction constant ( k ) , the observed reaction rate constant (kobsd), and the mean observed and the intrinsic reaction order (iiobsd and n, respectively) are known. Effective diffusivities and effectiveness factors for the employed catalyst samples based on the fresh catalyst have been reported elsewhere (Ammus and Androutsopoulos, 1987). Thus, if the values L i, Deff+and di are introduced into eq 5, the quantity can be evaluated while the product qt is calculated from eq 8. Values of q, t, and d for each experimental run can be determined by using the known quantities qc, and the Thiele plot, Le., effectiveness factor vs. the generalized Thiele modulus (Froment and Bischoff, 1979). Deactivationcurves of the HT-400 E catalyst (cylindrical extrudates, t = 0.0326 cm) follow the general deactivation pattern whick is also valid for its pulverized form (Figure la), differing only in the duration of the respective sections of the deactivation curves and the degrees of activity loss associated with each particular period of operation. In a similar manner, the catalyst cylindrical extrudates of the HT-400 E-H109 (Ep= 0.0589 cm) showed that the deactivation curve (Figure la) can be sectioned into the following three regions. The first region extends over the initial 20 h of catalyst operation, and a low, though increasing, rate of catalyst activity reduction was observed, while the total activity reduction over this period was rather limited ( 8 = 12%). The second region is characterized by a steep increase of both the rate and the ultimate degree of catalytic activity loss (8 = 35%) and extends over the time period between 20 and 50 h of operation. In the third region the initially high deactivation rate tends to become nearly constant as the operating time is progressing. Another observation is that the cylindrical extrudates show a higher rate and a final degree of deactivation during the first 60 h as compared to crushed pellets, although they are superior in the sense that they keep their activities for longer times compared to crushed catalyst pellets. The effect of catalyst particle size upon the degree of activity loss has been reported by Shah and Paraskos (1975) and Tamm et al. (1981). The results presented in this work are in reasonably good agreement with those reported by Shah and Paraskos but not entirely with those
1 5 1
Lr1
I
30
60
90
120
150
180
210
T i m e (h)
Figure 2. Coke deposit-time curves: (a) HT-400 E and (b) ICI-41-6 catalysts; ( 0 )pulverized, (m)extrudates 1/16-in.,and (A)extrudates '/a-in. Table V. Coke Deposits (wt %) on Catalysts Used in HDS Experiments on Greek Oil Atmospheric Residue, as a Function of Operating Time time of coke wt ?& in operation, sample code h a b C HT-400 E-1-1 30 26.93 24.79 30.36 -2 70 30.00 26.17 34.67 -3 200 34.69 32.00 36.74 HT-400 E-11-1 40 19.71 18.15 23.43 -2 20.50 18.67 24.71 70 110 22.73 18.51 -3 26.96 HT-400 E-111-1 30 16.73 15.41 20.55 -2 70 18.76 16.58 21.24 -3 140 23.10 18.61 26.15 ICI-41-6-1-1 28.26 26.03 31.94 30 -2 29.26 24.46 32.04 70 -3 140 31.56 26.83 35.30 ICI-4 1-6-11-1 15 18.54 17.15 21.74 -2 50 22.48 20.35 26.26 -3 90 24.03 23.83 27.19 g of organic comounds/g of sulfided catalyst (MoS2-Co9S8/y A1203).b g of organic compounds/g of sulfided catalyst and sulfided deposits (i.e., MoS2, CoeS8,y-Al2O3,V& NiS). Weight loss due to oxidation/weight of oxidated catalyst.
reported by Tamm et al. Deactivation curves valid for cylindrical extrudates of the ICI-41-6 catalyst showed a slightly different behavior in comparison with those obtained from tests on samples of crushed ICI-41-6 pellets. Kinetics of Coke and Metal Deposition. Coke and metal deposition kinetic studies made clear that coke is formed rapidly during the initial stages of catalyst operation, Figure 2. A further increase in the operating time gave rise to a minor increase of the amount of carbonaceous material deposited, independently of catalyst and particle
Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987 1317 Table VI. Vanadium and Nickel Depositions on Used Catalysts, a s a Function of Run Time on a HDS ExDerimental %actor g of Ni/g g of V/g of of regen. regent. catal., catal., wt time of sample code wt% YO operation, h HT-400 E-1-1 0.532 0.178 30 -2 0.986 0.222 70 -3 2.028 0.884 200 HT-400 E-11-1 0.144 0.059 40 -2 0.214 0.074 70 110 -3 0.316 0.103 HT-400 E-111-1 0.136 0.058 30 -2 0.176 0.088 70 -3 0.212 0.154 140 ICI-41-6-1-1 0.240 0.060 30 -2 0.712 0.286 70 -3 1.250 0.750 140 ICI-41-6-11-1 0.066 0.044 15 -2 0.148 0.108 50 -3 0.266 0.142 90 Table VII. Elemental Chemical Composition of Coke Depositsa sample code HT-400 E-1-2 -1-3 -11-3 -111-2 -111-3 ICI-41-61-3 -11-3
T i m e (hl
Figure 3. Vanadium deposition vs. operating time: (a) HT-400 E and (b) ICI-41-6 catalysts; (0)pulverized, (M) extrudates 1/16-in.,and (A)extrudates l/s-in.
T
e
0.8-
N, wt % 0.39 0.58 0.38 0.18
0.37 0.39 0.39
C, w t 90 23.56 28.66 20.45 14.31 15.97 22.75 20.65
H, w t 9i 1.90 2.36 1.82 1.64 1.48 2.22 1.83
Z(C+N+H), w t 90 25.85 31.60 22.65 16.13 17.81 25.36 22.97
metallic depositions (gram of metal/gram of regenerated catalyst) are proportional to catalyst operating time for all catalyst samples (Figures 3 and 4). A reasonable conclusion is that the hydrodemetalization reaction rate is effectively controlled by the intraparticle diffusion of organometallicreactions. It can also be stated that the reactivity of vanadium organometallic compounds was higher than that of nickel compounds. The ratio between deposited vanadium and nickel, i.e., (V, w t %)/(Ni, wt %), was found to be in the range 3-4 (Table VI). Figures 3a and 4a show that vanadium removal is more sensitive to smaller particle size variations than nickel removal. Chemical Composition of Carbonaceous Deposits. Quantitative chemical analysis of depositions showed that extensive catalyst operating times resulted in an increase of the C/H atomic ratio and the N2 content values (Table VII). A comparison between the sum of wt % (N + C H) obtained by chemical analysis and the coke w t % [(gram of coke)/(gram of sulfided catalyst)] verified a good agreement between the two values, determined experimentally, although oxygen and sulfur were not taken into consideration. Pore Structure Changes. I t is evident from Table VI11 that catalyst pore structure parameters underwent a rapid decline during the initial 20-30 h of operation. An increase of the operating time caused a further loss of the catalyst structure parameters. Thus, the most probable pore diameter changed from 82-86 to 36 A and from 74 to 36-38 A for catalysts HT-400 E (Figure 5) and ICI-41-6 (Figure 6), respectively, during the initial 20-30 h of operation. Pore volume, specific surface area, and pore size distribution changes during catalyst operation were found to be dependent on the catalyst type and particle size. Thus, crushed pellets of the HT-400 E catalyst indicated a 48% and a 90% loss in specific surface area after 70 and
+
(bl
t/P a , , 60 Time
,
120 lhl
,
,
,
180
Figure 4. Nickel deposition vs. operating time: (a) HT-400 E and (b) ICI-41-6 catalysts; ( 0 )pulverized, (M) extrudates 1/16-in.,and (A) extrudates 'Is-in.
size (Table V). The calculations of coke deposited within the pore structure include a correction factor, so that coke is expressed as gram of coke/gram of sulfided catalyst. The corrected coke level (coke wt %) showed a strong dependence on particle type and size (Table V). On the contrary, metals deposition kinetic studies indicated that
1318 Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987
5
20
40
10
20
60
40
23
60
60
pore r a d i u s I,?!
Figure 5. Catalyst HT-400 E, pore volume distributions. (f) Catalyst before use. (a) Pulverized catalyst pellets: (1)after 30, (2) 70, and (3) 200 h of operation. (b) Extrudates 1/16-in.: (1) after 40, (2) (1) 70, and (3) 110 h of catalyst operation. (e) Extrudates l/&: after 30, (2) 70, and (3) 140 h of operation.
Table VIII. Pore Structure Parameters of Used (Deactivated) Catalyst, as a Function of Operating Time total most catalyst S,, m2/g , 'V /g of probable of em operating sample code time, h catalyst catalyst dpore,A HT-400 E-1-1 30 146 0.219 36 -2 70 111 0.188 36 -3 200 25 0.047 36 HT-400 E-11-1 40 109 0.199 36 -2 70 109 0.182 36 -3 110 94 0.174 38 30 123 0.223 36 HT-400 E-111-1 70 116 0.209 36 -2 -3 140 109 0.180 36 ICI-41-6-1-1 30 178 0.272 36 -2 70 107 0.159 40 -3 140 94 0.152 38 ICI-41-5-11-1 15 147 0.229 38 -2 50 136 0.222 38 -3 90 93 0.174 36 Table IX. Specific Surface Area (S,) and Pore Volume Reduction ( V - J of HDS Catalyst after 30-h ODeration ~
catalyst t w e HT-400 E HT-400 E HT-400 E ICI-41-6 ICI-41-6
t
?
t
a All
70 Vpore, cm3/g of catalyst 52 58 58 52 52
% S,, m2/g
particle size, in. crushed
of catalyst 30 50 50 30 42
l/160
l/8"
crushed 1/16u
Extrudates. f
pore r a d i d s ( A I
Figure 6. Catalyst ICI-41-6, pore volume distributions. (0Catalyst before use. (a) Pulverized particles: (1)after 30, (2) 70, and (3) 140 h of operation. (b) Extrudates (1)after 15, (2) 50, and (3) 90 h of operation.
200 h of operation, respectively, while pore volume reductions by 58% and 90% were noted on the same samples over the respective operating times. Pore size distributions of partially deactivated crushed catalytic pellets indicated a progressive disappearance of pore sizes higher than d, = 50 A (Figures 5 and 6). The HT-400 E-H08(Lp = 0.0326 cm) cylindrical extrudates followed a different deactivation path. Thus, a specific surface area reduction by 50% occurred within the initial 40 h of operation, while an increase of the operating time to 70 h did not cause any appreciable surface area loss. A further increase of the operating time to 110 h caused a negligible decrease of the specific surface area (total loss after 110 h of operation, 56%). The relevant pore volume reductions reached the values 55%, 59%, and 61% for the respective run times of 40,70, and 110 h. The pore size distribution curves showed that over the operating period 0-110 h, the most probable pore diameter changed from 86 to 36 A and a continuous disappearance of pores fell in the range 50-100 A. Similar remarks can be made for the deactivation results on samples of HT-400 E-H109 (5, = 0.0589 cm) catalyst extrudates. Specific surface area of crushed pellets of the ICI-41-6 catalyst underwent reductions of the orders 30%, 57%, and 63% for 30, 70, and 140 h of operation, respectively. The respective pore volume reductions were
I
I
(b) 30
60
90
120
150
100
Catalyst Operating T i m e t ( h )
Figure 7. Specific pore volume vs. catalyst operating time: (a) pulverized, (0) extrudates HT-400 E and (b) ICI-41-6 catalysts; (0) 1/16-in.,and (A)extrudates '/8-in.
51%, 72%, and 74% for 30, 70, and 140 h of operation (Table VIII). Figures 7 and 8 correlate the catalyst operating time with the specific pore volume and surface area, respectively. Thus, for 140 h of operation, catalyst HT-400 E lost 62% and 53% of its original specific pore volume and surface area, respectively. Similarly, catalyst ICI-41-6 lost 75% and 62% of its original pore volume and surface area after 140 h of operation. Pore structure reduction properties after 30 h of operation are displayed in Table IX. Plots of catalyst relative activity vs. the specific pore volume and surface area are depicted in Figures 9 and 10,
Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987 1319
I
250 1
1 .a U >.
2
0.6
U i
2
:
0.6
4 3
c:
0.4
50
100
150
200
50
100
1JO
Specific Surface Area Loss (u’/sc)
Figure 10. Relative activity vs. specific surface area loss (m*/g): (a) HT-400 E and (b) ICI-41-6 catalysts: (0)pulverized, (m) extrudates 1/16-in.,(A)extrudates 1/8-in.
I
0
0
(h)
Time
i
Figure 8. Specific surface area vs. catalyst operating time: (a) pulverized, ( 0 )extrudates HT-400 E and (b) ICI-41-6 catalysts (0) 1/18-in.,and (A)extrudates ’/&.
-
30
60
120
90
150
180
C a t a l y s t Run T i m e (h)
Figure 11. Nominal coke deposit thickness vs. catalyst operating time: (0)HT-400 E and (e) ICI-41-6.
(a)
0
0.2
0.4
5,05
0,25
0,45
Specific Pore Jolune Loss (cm’/gc)
Figure 9. Relative activity vs. pore volume loss (cm3/g): (a) HT-400 E and (b) ICI-41-6 catalysts: (0) pulverized, (m) extrudates 1/16-in., (A)extrudates 1/8-in.
for both catalyst types. It is worth noting that in all curves despite the initial sharp decrease of pore volume and surface area, the relative activity remains high and rather stable, while a steep activity drop is noted within a narrow pore volume or surface area drop interval. These observations can be plausibly attributed to the fact that the texture of the coke deposit allows an effective contact of the reacting species with the active catalyst phase, existing either on the virgin catalyst surface or within the coke phase itself (active sites may be created by metal compounds deposited in coke layers (Soon-Jai and Mosby, 1986). Coke Foulant Profile Simulations. Pore plugging phenomena due to coke laydown can be interpreted in a realistic manner by using the corrugated random pore model as described by Androutsopoulos and Mann (1978). The simulation involved the following steps. (i) There was conversion of the coke deposit vs. catalyst run time experimental curves (Figure 2) into dimensionless pellet foulant vs. coke deposit thickness curves (Figure 12, dotted curves). The nominal coke deposit thickness ( h ) was correlated with the catalyst operating time ( t )(Figure ll),using the pore volume and surface area data (Figures
7 and 8), hi Vpore(tJ/Sg(Q. (ii) There was selection of a pore number distribution density function defined over the experimentally determined pore size range, its substitution into eq A-7 presented in the Appendix, and integration for various values of the parameter J, which represents the number of segments forming a corrugated pore. Parameter J is essentially a characteristic value of the pore length. (iii) There was plotting of the predicted coke profile over the relevant experimental data and determination of J values for which a satisfactory fit was achieved for each one of the catalyst particle sizes involved. If the fit is not satisfactory overall for a specified set of J values, the distribution needs to be modified and the calculations to be repeated until a satisfactory fit will result. (iv) There was transformation of the pore segment number distribution F N ( 0 ) or its truncated equivalent FNT(D)into the pore volume distribution Fv(D) according to
FV(o) = 0 2 F N T ( o ) / l D w D 2 FNT(D) DIU,
(9)
(v) There was comparison of the assumed pore volume distribution for which the best fit of the experimental points was achieved with the relevant experimental distribution deduced from porosimetry data. Parts a and b of Figure 12 depict the results of a simulation for catalysts HT-400 E and ICI-41-6, respectively. In both cases, a truncated normal distribution function (Figure 13) was substituted in eq A-7 in the Appendix, to
1320 Ind. Eng. Chem. Res., Vol. 26, No. 7, 1987 i.0
‘2 0
0.1;
9.4
C.2
la) 1
I
.o
3 .d
t
0.0
0.4
0.2
(b)
Y
I
10
30 40 30 0 Kominal Coke Deposit Thickness h l k ) 20
Figure 12. Corrugated random pore model fitted over coke foulant experimental profiles: (a) HT-400 E and (b) ICI-41-L catalysts; (0) pulverized, (m) extrudates 1/16-in,,(A)extrudates l/s-in. J = length of the random corrugated pore.
Table X. Pore Structure Parameters of Fresh Used and Regenerated Catalyst Samplesa most VF, S,, m2/g probable catalyst code cm /g of of diameter, run time, and oarticle size catalyst catalyst 8, h A: Catalyst HT-400 E I;, = 0.00567 cm 1-1 f 0.442 215 86 0.219 146 U 30 36 0.465 221 74 r 0.442 215 1-2 f 86 U 0.188 111 70 36 r 0.429 208 64 1-3 f 0.442 215 86 u 26 200 0.047 36 r 141 0.323 56 t,= 0.0326 cm 11-1 f 0.442 215 86 40 U 0.199 109 36 r 0.433 218 66 11-2 f 215 86 0.442 u 70 109 36 0.182 r 68 203 0.431 11-3 215 f 86 0.442 94 110 36 0.174 u 177 r 70 0.405 I;, = 0.0589 cm 111-1 f 0.435 218 82 36 123 U 30 0.223 198 r 98, 86 0.453 218 111-2 f 82 0.435 186 U 70 36 0.209 r 95 0.510 189 218 111-3 f 82 0.435 109 U 140 36 0.180 178 r 88, 74 0.382
B: Catalyst ICI-41-6
t,= 0.00567 cm I- 1
f u r f u r
0.557 0.272 0.537 0.557 0.159 0.530 0.557 0.152 0.465
253 178 232 253 107 229 253 44 211
74 36 68 74 40 62 74 38 64
f u r f u r f u r
0.557 0.229 0.538 0.557 0.222 0.509 0.557 0.174 0.518
253 147 230 253 136 253 253 93 220
74 38 68 74 38 66 74 36 72
f
u r
1-2 1-3
I;, = 0.0294 cm 11-1 0
io
40
Pore
Ga
8C
-00
123
140
Dfameter 8
Figure 13. Pore number truncated normal distribution function, FNT(D)(mean value of original F N ( D ) , = 50 A, and standard deX viation r~ = 30 A): FN@) = F~/0.933,F N ( D ) = exp[-0.5((D - p ) / r ~ ) ~ ] .
obtain the fit. From the results plotted in Figure 12, it is clear how the corrugated random pore can be used to quantify the effect of particle size on the amount of coke deposited and how the ultimate coke content for a particular catalyst particle size is reached. The corrugated pore model could be further exploited in the interpretation of catalyst activity reduction due to a pore plugging mechanism, being the result of pore structure-coke deposit interactions. It would be of particular significance to develop mathematical descriptions for evaluating the relative amounts of the “accessible”and “inaccessible” reactants to coke, at intermediate stages of the catalyst deactivation process, and also elucidate in physical terms the catalytic role of the “accessible” coke deposits. Figure 14 demonstrates the overlap between the experimental distribution functions and the theoretical one
11-2 11-3
30 70 140
15 50 90
“ f = fresh, u = used, r = regenerated.
used to obtain the best fit of the coke profiles. The comparison is considered as fairly satisfactory, if one bears in mind that the derivation of the experimental distribution from capillary condensation isotherms involves the use of a calculating procedure which includes a number of assumptions. Catalyst Regeneration. Catalyst pore structure parameter recovery upon regeneration proved to be dependent upon catalyst run time. Table X shows that catalyst deactivated over short time periods (30 h of operation) are nearly fully regenerated and the original pore structure parameters recovered. Higher operating times, e.g., HT-400 E for t = 200 h, are associated with appreciable differences in the physical parameters of the fresh and the respective regenerated catalyst samples. These
Ind. Eng. Chem. Res., Vol. 26, No. 7 , 1987 1321
c' Pore Diameter A Figure 14. Comparison of the pore volume distribution function F@), used to fit the coke foulant profile of catalysts (a) HT-400 E and (b) ICI-41-6 with the experimentally deduced distribution (0) for the respective catalyst.
I
*lMetal ( V + N i l deposition, wt 8
Figure 16. Catalyst activity vs. metal deposit: (a) HT-400 E and (b) ICI-41-6 catalysts; ( 0 )pulverized, (W) extrudates 1/16-in.,and (A) extrudates 'fs-in.
Figure 15. Catalyst activity vs. coke deposit: (a) HT-400 E and (b) ICI-41-6 catalysts; ( 0 )pulverized, (W) extrudates 1/16-in.,and (A) extrudates 'ls-in.
particular type of deposit cannot be easily distinguished, so that Figures 15 and 16 should not be considered separately. Inspection of Figures 15a and 16a, valid for catalyst HT-400 E, leads to the following observations. (i) For samples of pulverized catalyst, 25 w t % coke and 0.7 wt % metal deposits developed before a measurable activity loss was detected. (ii) Further increases of the deposits caused an almost linear activity loss. Cyclindrical extrudates proved to be more vulnerable to coke and metal deposits. A 17 wt 3' % coke content and 0.1 w t 70 metal deposit content for 1/16-in.pellets and a
