J . Phys. Chem. 1986, 90, 4894-4900
4894
Fluid Catalytic Cracker Catalyst Design for Nitrogen Tolerancet G . W. Young W. R. Grace and Co., Davison Chemical Division, Columbia, Maryland 21044 (Received: March 13, 1986)
Pulse reaction experiments in which mixtures of 1-4 wt % quinoline or pyridine in hexadecane were reacted over a commercial cracking catalyst (Davison’s Super DX) demonstrated the reversible equilibrium process of poison adsorption. The use of a low-activity (for cracking hexadecane) silica-alumina amorphous matrix component, mixed with the Super DX, resulted in a higher conversion of a hexadecane-quinoline mixture compared to Super DX alone. Gas oil cracking studies with various levels of preblended quinoline were performed using catalysts in which both zeolite and active matrix content were varied. The activity and selectivity results can be used to develop design criteria for cracking catalysts that are to be used in the processing of feedstocks containing unusually high organic nitrogen levels. The general applicability of the quinoline studies was verified by similar studies using a naturally occurring high nitrogen content gas oil and blends of it with a shale oil.
Introduction
Over the past 30 years many papers dealing with the effects of nitrogen compounds on the activity of cracking catalysts have appeared. The emphases of these published works have usually been on (a) how specific classes of compounds affect catalyst activity or selectivity; (b) how practical steps can be taken to minimize the operational impact of processing high nitrogen feeds; and (c) attempts to explain or rationalize the observed effects. No work has specifically focused on how to design a cracking catalyst for use with high nitrogen feeds and it is this aspect with which this current work is concerned. Occurrence and Classification of Nitrogen Compounds
The presence of nitrogen compounds in crude oils and fluid catalytic cracker (FCC) charge stocks has long been recognized’ and periodically, reviews of crude oil analyses have appeared (ref 2-5 and references therein, and ref 7). However, until the work of Poth et a1.,6 it was generally believed that the majority of nitrogen compounds were basic. While Poth showed that the major proportion of nitrogen compounds were nonbasic, they did not report specific structures nor define “basic”. Richter2 provided a meaningful, quantitative description of “basic” nitrogen by showing that the favored analytical procedure for determining basic nitrogen, Le., titration by perchloric acid in glacial acetic acid, was effective for compounds with a pK, > 2. Using this procedure, Richter went on to examine a number of crude oils and found the ratio of basic nitrogen to total nitrogen (NB/NT) to be 0.3 f 0.05 despite a range for N T of 0.3-2.3 wt % of feed. With the possible exception of coker gasolines8 and hydrotreated feeds5 this empirical feedstock ratio value has been reported by many other^.^,^.^ RichterZand Helm,3 who reported on the work arising out of API project 52, examined the nitrogen content (and NB/NT) as a function of feed boiling range. They showed that while the total nitrogen increased with boiling point, with the residuum fraction containing the majority of the nitrogen, the ratio NB/NT peaked 0.5). at - 3 1 5 “C boiling range ( N B / N T The types of nitrogen compounds present in feedstocks have been investigated by numerous worker^.^^^*^^'^ The basic nitrogen is largely associated with heterocyclics like pyridines, quinolines, ‘It is with sadness that we in the Davison Chemica! Division of W. R. Grace dedicate this paper to the memory of our long time friend and advisor, Professor Paul Emmett. Dr. Emmett was the W. R. Grace Research Professor at the Johns Hopkins University, and after his retirement, he continued his association with Grace by providing valuable consultation services to research at the Davison Chemical Division. His bimonthly visits to our Columbia, Maryland, research facilities continued through 1984 until he became too ill for travel. Professor Emmett maintained an active interest in our FCC catalytic research and provided critical reviews of much of the work that was performed. It is, therefore, appropriate that this contribution on the role of nitrogen poisons in catalytic cracking appear in this commemorative issue.
0022-3654/86/2090-4894$01.50/0
and related molecules. The nonbasic molecules are also largely heterocyclics such as pyrroles, indoles, and carbazoles. Effect of Nitrogen Compounds on Cracking Catalysts
The recognition that catalytic cracking proceeded by carbonium mechanism and required acidic catalysts stemmed from the efforts of a number of workers as described recently by Voge.” It was therefore logical that catalyst characterization was attempted by titration of acidity using basic nitrogen compounds like butylamine, pyridine, and quinoline. The work of Mills et al.I3 gave a practical demonstration that organic nitrogen compounds severely affect the activity of cracking catalyst, specifically under realistic cracking conditions (Le., -500 “C). This group determined the influence of several compounds that could be present in feedstocks and demonstrated for cumene cracking that the severity of deactivation followed quinaldine (5.8) quinoline (4.9) > pyridine (5.2) > piperidine (1 1.1) >> decylamine ( 1 0.6) >> aniline (4.6) Although it was natural to assume an acid-base interaction between basic nitrogen compounds and acid cracking catlayst, the pK, values, in parentheses, indicate no direct correlation between severity of deactivation effect and basicity. Mills also reported ammonia chemisorption results and showed that organic nitrogen compounds were more strongly held at cracking temperatures than was ammonia. Similar results were reported about this time by Voge and co-w~rkers.’~ They reported the severity of poisoning for decalin cracking (500 “C) as follows acridine > quinoline >> carbazole > naphthylamine > indole > pyridine > dicyclohexylamine > pyrrole They reported no poisoning effect from methylamine, diamylamine, or ammonia. Thus, both of these early works demonstrated that basicity and poisoning effect did not necessarily go together. (1) Mayberry, C. F.; Wesson, L. G . J . A m . Chem. S o t . 1920, 42, 1014. (2) Richter, F. P. et al. Ind. Eng. Chem. 1952, 44, 2601. (3) Helm, R. V. et al. J . Chem. Eng. Datu 1957, 2, 95. (4) Venuto, P. B.; Habib, E. T. Fluid Catalytic Cracking With Zeolite Catalysts; Dekker: New York, 1979. ( 5 ) Fu, C. M.; Schaffer, A. Ind. Eng. Chem., Prod. Res. Deu. 1985, 24.
68. (6) Poth, E. J . et al. Ind. Eng. Chem. 1928, 20, 8 3 . (7) Lochte, H. L. Ind. Eng. Chem. 1952, 44, 2597. (8) Samuelson, G.J.; Woelflin, W. Pet. Ref. 1959, 38(2), 221. (9) Brandenburg, C. F., Latham, D. R., J. Chem Eng Data 13(3), 391 (1968). (10) Katzer, J. R., Sivasubramanian, R. Catal. Rev. Sci. Eng. 1979, 20(2), 155
( 1 1) Voge, H. H. Heterogeneous Catalysis; ACS Symp. Ser. 1983, 222(19). 235. (i2jTamele, M. W. Discuss. Faraday S o t . 1950, 8, 270. (13) Mills, G. A. et al. J . A m . Chem. SOC.1950, 72, 1554. (14) Voge, H. H. et al. Proc. 3rd World Pet. Cong., Section IV 1951, 129.
0 1986 American Chemical Society
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4895
Cracking Catalyst Design for Nitrogen Tolerance TABLE I: Properties of Feed Components
WTHGO gravity, OAPI sp gravity (I5OC)
aniline pt., OC
s,W t %
total nitrogen, wt % basic nitrogen, wt % Ni, ppm PPm Fe, ppm Cu, ppm Pb, ppm As, ppm
27.4 0.8905 93 0.38 0.07
"3
West Coast feed
shale oil
22.7 0.9176 67 1.18 0.36 0.10 0.2 0.4 5 0.1 0.4
15.9 0.9600
0
2.3 1.1 5 0.3 0.3
IO 20 30 40
50 60 70 80 90 95
fbP
UOP K factor
366 375 383 391 399 407 416 428 442 458 472 489 12.1
C 0 I
03 L
m C > 0
0
C
0
0
-
4-
.-0 a,
3-
3-
.-
E
X
21 -
0. 0
,
10
I
I
I
20
30
40
0
Zeolite
Zeolite W %
0
W%
10
20
Zeolite
1
I
30
40
W%
Figure 5. MAT study of quinoline-poisoned W T H G O feed; reaction conditions: temperature 500 O C ; 16 WHSV, 3 catalyst/oil ratio.
%%W A L WTRIX
INERT MTRIX
t
0 -0 0.00 0.05 0.10 0.15 0.20 0 . 0 0 0.05 0.10 0.15
% Basic Nitrogen
4UX SI/AL MTRIX
t
- 0 0 . 0 0 0.05 0.10 0.15 0.20 0.20
% Basic Nitrogen
96 B a s i c N i t r o g e n
Figure 6. Conversion contour map: effect of quinoline-poisoned W T H G O feed; conditions as in Figure 5 .
Catalysts used in the various studies were either commercially available fluid cracking catalyst or materials prepared in our laboratories with similar physical properties (e.g., particle size distribution). All fresh catalysts were hydrothermally treated (8 h, 2 bar steam pressure, 732 "C) to simulate a commercial equilibrium catalyst condition.
Results and Discussion Pulse Reactor Studies. Pulse experiments at 500 OC were performed using 50 mg of a commercial cracking catalyst (Davison Super DX) that had been steamed prior to testing. The conversion of CI6after each 1-pL pulse was calculated, and the results were plotted against the pulse sequence number. Two similar sets of experiments, one with 1 wt % quinoline poisoning, the other with 1 wt % pyridine poisoning, were performed (Figures 1 and 2). The poison-free catalyst deactivation (from coking?) was monitored by a series of 10 pulses with pure c 1 6 as feed. For both sets of experiments, initial conversion was high (>98%) and declined slowly during the first the 10 pulses. Immediately after introduction of the poisoned feed, conversion dropped abruptly, with
quinoline exhibiting a much more severe poisoning effect than pyridine (Figure 2). After the initial, abrupt activity loss, conversion declined slowly, as if from coke deactivation. Even after 20 pulses with the poisoned feed, it is clear that neither poison at 1 wt % concentration irreversibly accumulated nor completely deactivated the catalyst. After 20 pulses of poisoned feed, pulses of pure C,, were again used. Immediately, and in both cases, cataiyst activity raGdly increased. For pyridine, activity was completely restored after only one pulse, whereas with quinoline approximately 5 pulses were required. Following restoration of catalyst activity, c16 containing 4 wt % quinoline or pyridine poison was pulsed. For both cases this was a sufficient poison level to almost completely deactivate the catalyst (Le., conversion C 25%). The feed was again returned to pure C I 6and immediate restoration of catalyst activity was observed for pyridine poisoning and restoration after 4-6 pulses with quinoline. As would be expected from a competitive equilibrium process the extent of deactivation (poisoning) depends on the poison concentration in the feedstream and not on the catalyst time on
4898
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 0% QUINOLINE
1% QUINOLINE
0.20-
-
0.15
0.20-
0.15-
0. 5!\
' 2
I
0.00
'
0.05120%
'
'
MAT Conversion W %
'
INERT
:
's>
0.051
C
' 0.00 : INERT
o
2% SI/AL O'
WAL
/A
WAL
40%
/
/
./$
1 INERT
/
SI/AL
0.10:40%
120% W A L
2% QUINOLINE
0.20-
0.10-
0.05
Young
S
-
I
/
0.00
'
I
MAT Conversion W%
INEP,T (1% 0) INERT (0% Q)
/
4
I
/ .,
.
I
MAT Conversion W %
Figure 7. MAT study of quinoline-poisoned W T H G O : effect of increasing poison level on hydrogen yields: conditions as in Figure 5
1% QUINOLINE
0% QUI!!OLINE
I
4% SI/AL0 20% W A L A
'
INERT
4
2-
0
t 70
60
50
I
io
6I -
2% QUINOLINE
00
90
w%
MAT Conversion
MAT Conversion W %
MAT Conversion W%
Figure 8. M A T study of quinoline-poisoned WTHGO: effect of increasing poison level on coke yield; conditions as in Figure 5
0% I3UIPIOLItiE
1% QUIMLINE
r
16
r-
18-
14
-
14-
40% S ~ A L
'h 20% SI/&
INERT
12-
10
2% WIYOLINE
\ m SI/AL "20% W A L
\
-
SI/AL
1 : a\
SI/&
12-
N IE R T \!
10
'h
0
\
.
I
"
*
'
I
MAT Conversion W%
8'
'
'
" *
'
'
*
MAT Conversion W%
t - 8 50
60
70
80
90
MAT Conversion W%
Figure 9. MAT study of quinoline-poisoned WTHGO: effect of increasing poison level on light cycle oil yields. Conditions as in Figure 5
The Journal of Physical Chemistry, Vol. 90, No. 20, 1986 4899
Cracking Catalyst Design for Nitrogen Tolerance
10.0
TABLE 11: Ratio of Kinetic Conversion with Poison (K,)/Poison-Free ( K O )at 500 and 538 OC for Gas Oil Cracking (16 WHSV. 3 c/o) Catalyst Inert Matrix and -20% Zeolite
35% CW I N
temp, 'C quinoline level, wt 5%
500
538
0 1
1 .O 0.74
2
0.32
1.o 0.81 0.39
lNRW SI/^ MTRIX
8.0
8.0 6.0 4.0
3.0 C
.-0 m
2.0
L
0
> C
0
0 0 .c 0
1 .n
Y
0.8
.-C 0 '
100
200
300
400
500 0.6
X-Ray
Line Intensity. Arbitrary Units
Figure 10. Nitrogen tolerance of commercial ( 0 )and experimental (0) catlaysts. Ratio of kinetic conversion on high nitrogen feed (K,) and WTHGO (KO).Reaction conditions: temperature 500 'C, 16 WHSV, 3 c/o.
0.5 0.4
0.3 I
stream. For refractory molecules like quinoline or pyridine, secondary reactions of the adsorbed species are not a significant contribution to catalyst deactivation. Whether this is generally true for other nitrogen poisons that are commonly found in high nitrogen content FCC feedstocks remains to be determined. To determine the effect of temperature on quinoline poisoning, a similar set of experiments was performed at 538 "C (1000 O F ) . The results (Figure 3) qualitatively indicate that at the higher temperature, the reduction in C16 conversion by 1% quinoline is considerably less than at 500 "C, although at 4% quinoline complete deactivation is again observed. To separate the temperature effect on kinetics from reduction in the extent of poisoning, the ratio of observed rate constants with or without poison at the two temperatures can be calculated. Based on the kinetic analysis reported by NaceZ8the rate constant ratio for the 10th to 15th pulse for each temperature is -7.5 at 500 O C and -2 at 538 O C . Thus, for hexadecane cracking there is a significant effect of temperature in reducing the poisoning, but even at 538 'C (the upper end of the commercial range for FCC) poisoning by quinoline is still appreciable. To examine the influence of matrix activity on nitrogen poisoning, an amorphous silica--alumina microspherical catalyst was tested by itself and also mixed (20%) with the commercial FCC catalyst tested earlier. The activity of the amorphous catalyst for C,6 cracking was low (Figure 4). However, when the blended catalyst was tested, the effect of 1% quinoline on (216 conversion was significally lower than for the pure commercial zeolite catalyst. The amorphous component has presumably adsorbed enough quinoline to effectively reduce its partial pressure and thereby shift the equilibrium for the adsorption on the zeolite. These studies provide a clearer understanding of the poisoning mechanism (at least for simple nitrogen compounds like pyridine and quinoline) and suggest ways that can be used to develop catalysts that can be more effective against high nitrogen feeds. Effect of Catalyst Composition
To provide a more realistic framework for evaluating catalyst performance two additional studies were performed. The first involved testing a variety of catalysts by MAT using the low nitrogen WTHGO that could be blended with quinoline. Second, MAT studies were performed using the high nitrogen West Coast (28) Nace, D. M. Ind. Eng., Chem., Prod. Res. Deu. 1969, 8, 24
0
0.2
.
t
0.4
.
1
0.0
.
l
.
0.8
l
1.0
Total Feed Nitrogen, W%
Figure 11. Effect of matrix chemistry on catalyst activity as feed nitrogen increases (blending shale oil). Reaction conditions: temperature 500 OC, 16 WHSV, 3 c/o. Fully rare earth exchanged, calcined Y zeolite (CREY) catalysts.
gas oil feed that could be further mixed with very high nitrogen content shale oil (feed properties shown in Table I). A series of catalysts of varying zeolite content and active matrix content were activity tested by using MAT and WTHGO that had been blended to various quinoline levels (0, 0.1, 0.2% basic nitrogen). All catalyst were steam-treated prior to activity testing. Some tests were performed at a reaction temperature of 538 O C but most were tested at 500 O C , 16 WHSV 3 c/o. The results, plotted as kinetic conversion against zeolite content (Figure 5), show the expected increase in activity with zeolite content at all nitrogen levels. These data are replotted (Figure 6) as conversion contours on a zeolite content-basic nitrogen surface. To maintain constant activity at a given nitrogen level, compared to the nitrogen free case, either zeolite content can be increased or the active matrix content can be increased. Thus, to maintain 75% conversion at 0.1% basic nitrogen, either the zeolite content can be increased by approximately 30% or the matrix content can be increased. These type of data can be used to design a catalyst for a specific application. In actual practice both zeolite and matrix activity may be increased to provide an adequate ratio of zeolite/matrix that maintains proper catalytic ~ e l e c t i v i t y . ~ ~ ~ ~ ~ The hydrogen and coke yields from these experiments are shown in Figures 7 and 8. Consistent with other studies on the effects of nitrogen compounds, an increase in hydrogen yield with increased nitrogen in the feed was observed. Also, as expected, increasing the matrix activity increased the hydrogen yield. The source of the increased hydrogen yield is not likely to be from specific reactions involving the adsorbed quinoline as it has shown to be quite refractory under typical cracking conditions.20 A more likely explanation is that the strongly adsorbed compound inhibits the hydrogen-transfer reaction within the zeolite. The coke yields (Figure 8) at constant conversion increase as the nitrogen level increases reflecting the increased adsorption (29) Ritter, R. E.; Young, G. W. Presented to NPRA Annual Meeting, San Antonio, March 25-27, 1984; paper AM-84-57 (30) Bremer, H . et al. A d a Phys. Chem. 1985, 31, (1-2), 369.
4900
The Journal of Physical Chemistry, Vol. 90, No. 20, 19616
capacity of the higher zeolite content required to maintain cracking conversion. Both these effects have important consequences for the commercial operation. Thus, the limiting factor in processing high nitrogen feeds may not be the inability to catlaytically attain or maintain a specific conversion, but rather the limitation caused by higher coke load (e.g., regenerator temperature) or higher hydrogen yield (compressor limit). Therefore, a desirable attribute for a catalyst used in a high nitrogen environment will be superior coke selectivity. Figure 9 shows the benefit of having some matrix activity to improve bottoms cracking, Le., the conversion of the high boiling range feed components (>482 "C) which results in increased yields of the light cycle oil (LCO) boiling range (216-338 "C) products. Even in the presence of nitrogen the active matrix retains its bottoms cracking activity. These data also demonstrate that for bottoms cracking a little matrix activity is as effective as a lot of matrix. Lower matrix activity also minimizes the hydrogen yields. For these gas oil cracking studies there was only a small effect of temperature on improved nitrogen tolerance, in agreement with the Mobil result^.^' These results (Table 11) compare the kinetic conversion ratio (poison/poison free) at 500 and 538OC for 1 % and 2% quinoline levels. Catalyst Testing with a High Nitrogen Feed
The performance of many commercial catalysts as well as a wide variety of experimental preps was evaluated by MAT with a high nitrogen West Coast feed (Table I). Zeolite content was estimated by using relative intensity of XRD peaks. The kinetic conversion of the high nitrogen feed (K,) (Table 11) was measured for each catalyst in the study at 482 "C, 16 WHSV, and 3 c/o. The kinetic conversicn of each catalyst was also measured with a low nitrogen gas oil (KO). The ratio K,,/K,, plotted against relative zeolite content (Figure lo), can be used to highlight catalysts of improved nitrogen tolerance. All the commercially available cracking catalysts that were tested fall within a relatively narrow band of K,,/Ko, with few commercial catalysts being able to exceed a K , / K o ratio of 0.5 even at very high zeolite contents. Many of the experimental catalysts, which involved alterations of the matrix chemistry, demonstrated high activity retention (>0.5) with a few exceeding 0.6. To furthzr highlight the importance of the matrix chemistry in the cracking catlayst two catalysts were made with the same
Young amount of a fully rare earth exchanged Y zeolite but with different matrices. Catalyst A was prepared with a catalytically inert, silica-clay matrix whereas catalyst B contained a highly active, high alumina content silica-alumina matrix. Both catalysts were tested on the gas oil feed, the high nitrogen feed, and blends of the high nitrogen feed with a shale oil. Total nitrogen content of the different feeds varied from
