- Email: [email protected]
Characterisation of coke from FCC refinery catalysts by quantitative solid state 13C NMR
~
'~9 ~
AP PA LL IEYDSS C AT I A: GENERAL
ELSEVIER
Applied Catalysis A: General 129 (1995) 125-132
Characterisation of coke from FCC refinery catalysts by quantitative solid state 13C NMR C.E. Shape
a,*,
B.J. M c G h e e a, J.M. Andresen C.L. Koon b, G. Hutchings c
a, R.
Hughes
b,
~'UniversiO, of Strathclyde, Dept. of Pure and Applied Chemistry, Glasgow G1 1XL, UK b Dept. of Chemical Engineering, University ofSalford, SalJbrd, M5 4WT, UK c Leverhulme Centre for Innovative Catalysis, Dept. of Chemistry, Liverpool University, Liverpool L69 3BX, UK Received 8 March 1995; revised 27 April 1995; accepted 27 April 1995
Abstract Coke has been concentrated from two deactivated FCC refinery catalysts via demineralisation to facilitate detailed characterisation by solid state ~3C NMR. The catalysts were obtained from runs with a residue feed (5% Conradson carbon) and a hydrotreated vacuum gas oil ( H V G O ) . As for solid fuels, the use of a low-field spectrometer in conjunction with the single pulse excitation (SPE or Bloch decay) technique has enabled quantitative carbon skeletal parameters to be obtained for the cokes. Internal standard measurements demonstrated that most of the carbon was observed by SPE and, therefore, NMR-invisible graphitic layers are not thought to be major structural features of the cokes. SPE gave much higher values for both the carbon aromaticities and the proportions of nonprotonated aromatic carbon than the less quantitatively reliable cross-polarisation (CP) technique. Differences in feedstock composition were reflected in the structure of the cokes with the aromatic nuclei being more highly condensed in the residue-derived coke and corresponding to 15-20 pericondensed aromatic rings. Keywords: Coke; FCC; Nuclear magnetic resonance
1. Introduction In view of the importance of fluid catalytic cracking (FCC) to petroleum refining, the deactivation of cracking catalysts via coke deposition has been the subject of considerable investigation over the past 50 years [ 1-3]. Coke deposition is clearly relevant to the more recent trends in FCC operations of using heavier feeds with higher overall conversions to gasoline [4]. As well as being formed via the actual * Corresponding author. 0926-860X/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDIO926-860X(95)O0104-2
126
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
cracking reactions associated with the strongly acidic catalytic sites, coke can arise in FCC units from (i) normal thermal reactions, (ii) dehydrogenation reactions promoted by metals - - N i and V in heavy feeds - - a n d (iii) entrained catalyst products, symptomatic of incomplete stripping in the regenerator. The contributions of these different mechanisms is clearly going to be dependent on the type of feedstock, together with the design and operation of particular FCC units. However, the lack of detailed basic knowledge on the chemical nature of coke is preventing progress towards a proper overall mechanistic understanding and a clear identification of the different coke-forming routes in FCC operations. Fundamental deactivation studies involving NMR on zeolites thus far have generally involved excessively high concentrations of coke in relation to normal FCC operation where catalysts are regenerated typically after only ca. 1% w / w of carbon has been deposited. The high concentrations have been necessary to achieve sufficient sensitivity for characterising the coke on the deactivated catalysts by solid state 13C NMR. For example, Groten et al. [5] have investigated coke formation on zeolite USHY with 1-hexene as the feed with coke levels of ca. 5% w/w. Lange and coworkers [6] used 13C-enriched ethene to follow the formation of polyaromatic structures on H-mordenite. Deactivated y-alumina-supported hydroprocessing catalysts have also been characterised by solid state 13C NMR [7] but, again due to the relatively low sensitivity, the spectra have been obtained by cross polarisation (CP) which fails to observe all the carbon in coals and oil shales and usually discriminates heavily against aromatic carbon [ 8]. Further problems are posed by high magnetic field strengths where, to eliminate spinning sidebands, special pulse sequences or extremely rapid magic-angle spinning (MAS, > 10 kHz) are needed. There is a general consensus that the use of low field strengths with Bloch decay or single pulse excitation (SPE) offers the best compromise for quantitative 13C NMR analysis of coals and solid fuels [ 8-11 ], albeit with a considerable sacrifice in sensitivity since long recycle times (with 90 ° pulses, 5 times the 13C thermal relaxation times - - T l s ) are required to ensure that the 13C magnetisation fully regains equilibrium. In principle, the only carbon not observed is that in the vicinity of paramagnetic centres which obviously includes graphite. The only way this methodology can be applied successfully to deactivated catalysts is by demineralising the aluminosilicate matrices to concentrate the coke. In this communication, this approach has been used for the first time to characterise two coke concentrates isolated from deactivated FCC catalysts containing only ca. 1% w / w carbon. The samples chosen for investigation were from refinery runs with an atmospheric residue and a hydrogenated vacuum gas oil (HVGO) in order to ascertain whether the use of these two vastly different feedstocks had a significant effect on coke composition. The degree of condensation of the aromatic structure has been assessed from the proportions of non-protonated aromatic carbon derived by the SPE 13C NMR technique.
C.E. Snape et al. / Applied Catalysis A: General 129 (1995) 125-132
127
2. Experimental 2.1. Catalysts and their demineralisation The FCC catalysts were typical commercial formulations and the deactivated samples were obtained from units processing (i) a heavy feedstock containing ca. 1.5% sulphur, a Conradson carbon content of 5.0% w / w and Ni and V contents of 5 ppm each and (ii) a HVGO containing only 0.1% sulphur and below 2 ppm of Ni and V. The catalysts 'as received' both had carbon contents close to 1%. They were first refluxed in chloroform for 3 hours to remove any trapped molecular species and then vacuum-dried prior to demineralisation. This initial treatment reduced the carbon content of the catalyst deactivated with the atmospheric residue from 1.08 to only 1.00% (estimated error of +_0.05%), indicating that molecular species only account for an extremely small proportion of the coke. This observation is consistent with the fact that both catalysts were removed from the FCC units after stripping (collected at the base of the stripper). The cokes were concentrated by applying the standard demineralisation procedure for solid fuels [12], [ 13] to the chloroform-extracted catalysts. This involved successive extraction with 2 M hydrochloric acid (stirring overnight at 60°C) and 40% hydrofluoric acid (HF), the HCl-extracted sample being stirred at room temperature for 4 hours with 20 cm 3 of HF being used per gram of sample. The coke concentrates were finally washed with dilute hydrochloric acid to remove any remaining inorganic paramagnetics prior to collection in plastic filtration equipment. The vacuum-dried coke concentrates recovered from the catalysts deactivated by the residue and HVGO feedstocks had carbon contents of 55 and 39%, respectively.
13C NMR The CP and SPE 13C NMR measurements on the coke concentrates were carried out as previously for coals [9-11] at 25 MHz on a Bruker MSL100 spectrometer with MAS at 4.5-5.0 kHz to give spectra in which the sideband intensities are only ca. 3% of the central aromatic bands. Known weights of tetrakis (trimethylsilyl) silane (TKS) were added to the samples as the internal reference and to facilitate estimation of the fraction of the total carbon observed by SPE. Approximately ca. 150 mg of sample was packed into the zirconia rotors. The 1H decoupling and spinlock field was ca. 60 kHz and, for SPE, the 90 ° 13C pulse width was 3.4/xs. A recycle delay of 50 s was employed between successive 90 ° pulses in SPE since the ~3C T~ for the non-protonated aromatic carbons were 10 s as measured by Torchia's CP-based method [ 14] for both coke concentrates 9(Table 1). Normal CP spectra of the residue-derived coke were obtained using a range of contact times between 0.05 and 8 ms for the residue-derived coke to facilitate determination of the time constants for CP (Tc~) and ~H rotating-frame relaxation (Tip). The CP spectrum of the HVGO-derived coke was obtained using a contact time of 1 ms.
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125-132
128
Table 1 NMR and other structural parameters for the coke concentrates Parameter ~3C T~/s for non-protonated aromatic carbon CP-derived parameters obtained from variable contact time experiment, TcH/l~s, aromatic C b CH2 CH 3
TJms, aromatic CH CH2 CH 3
Residue feedstock 10.0 330 40 70 8 6 5
HVGO feedstock 11.5 N.D. N.D. N.D. N.D, N.D, N.D,
Carbon aromaticity,fa, ( _+0.01 )a CP (1 ms) SPE
0.91 0.97
0.85 0.96 N.D.
0.55 0.67 0.63 0.75 70
N.D 0.56 0.51 0.50 90
Fraction of aromatic carbon that is non-protonated from dipolar dephasing experiments ( + 0.02) a, CP ( 1 ms) SPE
Fraction of aromatic carbon that is bridgehead, CBR/CAR(SPE5: 0.03) a Fraction of aliphatic carbon that is CH3 ( 10-24 ppm range, SPE + 0.1 a) % of carbon observed by SPE
N.D. = not determined; SPE = single pulse excitation; CP = cross-polarisation.
Estimated errors, b Average value for protonated and non-protonated carbons.
Dipolar dephasing (DD) was combined with both SPE and CP to estimate the proportions of protonated and non-protonated aromatic carbon, and at least 8 separate dephasing periods in the range 5-500/xs were used (500 scans each), In order to check that the tuning had remained virtually constant throughout the duration of the DD experiments, the delays were arranged in a random order and between 500 and 1000 scans were accumulated for each delay. The spectra were processed using exponential line broadening factors of either 30 or 50 Hz. No background signal was evident in the SPE spectra from the Kel-F rotor caps. The measurement of aromatic and aliphatic peak areas manually was found to be generally more precise than using the integrals generated by the spectrometer software.
3. Results and discussion Fig. 1 and Fig. 2 compare the CP and SPE 13C spectra of the highly aromatic coke concentrates and Fig. 3 presents the decays of the aromatic peak intensities in the CP and SPE-DD experiments on the residue-derived coke. The carbon skeletal parameters obtained from the SPE and CP 13C NMR experiments for both samples are summarised in Table 1. The first coke concentrate obtained from the catalyst deactivated with the residue feedstock gave a broad aromatic band, possibly due to the incomplete removal of the rare-earth metal during the HC1 wash. Indeed, it was found that, after the final HC1 wash, much narrower spectral bands were obtained
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125-132
129
Aromatic carbon
TKS
Aliphatic carbon SPE
. . . .
200t . . . .
1501 . . . .
1001 ~ ,
.
, i0 . . . .
0t . . . .
l
PPH
Fig. 1. CP ( 1 ms contact time) and SPE t3C NMR spectra of the coke concentrate from FCC catalyst deactivated using the residue feedstock.
J
SPE
~ .
.
.
.
L
200
.
.
.
.
l
TSO
.
.
.
.
~ L
.
100 PPH
.
.
, .
i
50
~ .
.
.
.
CP I 0
.
.
i
,
Fig. 2. CP ( 1 ms contact time) and SPE 13C NMR spectra of the coke concentrate from FCC catalyst deactivated using the HVGO feedstock.
130
C.E. Snape et al./Applied Catalysis A: General 129 (1995) 125 132 3-
~e
[] m
•
•
,;0
2;0
[]
CP
•
SPE
£
eL
1 ¸
,;0
,;0
~;0
Tau/~s Fig. 3. Decay of the natural logarithm of the aromatic peak intensity in the CP and SPE dipolar dephasing experiments on the coke concentrate from the FCC catalyst deactivated using the residue feedstock ( initial intensity assigned value of 10).
and the quality of the spectra shown are comparable to those obtained for lowvolatile coals and anthracites (with similar aromaticities as the cokes, see following). The sensitivity is obviously superior in the CP spectra (7-10 000 scans for CP compared to 1000-2000 for SPE, Fig. 1 and Fig. 2) but, as found for many coals [9-11 ], CP significantly underestimates the carbon aromaticity using relatively short contact times (Table 1, CP values for 1 ms contact, estimated error _+0.01 ). This arises from the non-protonated aromatic carbons, in particular, cross polarising much more slowly than the aliphatic carbons (both C H 2 and CH3 in the residue coke have much shorter TcHs than the aromatic carbon, Table 1 ). At longer contact times, the discrimination against aromatic carbon is not quite as acute, but the aromaticity values obtained are still lower than that by SPE (Fig. 4). This is indicative that some aromatic carbon, presumed to be in the vicinity of free radicals, is not polarised at all (~H T~ps are too short - - < c a . 0.2 ms, the values in Table 1 are only for hydrogens adjacent to the carbons being polarised). The fact that 70 and 90% of the carbon in the residue and HVGO-derived coke concentrates, respectively, has been observed by SPE demonstrates that the procedure is reasonably quantitative and that graphitic layers are probably not present in significant amounts. XRD is being used to check the possibility that the cokes may contain some extended graphitic structure. However, if it was present in significant amounts, the resultant paramagnetism would have detuned the probe resulting in little of the carbon being detectable. As for the total aromaticity, CP also grossly underestimates the fraction of nonprotonated aromatic carbon determined by dipolar dephasing (Table 1, estimated error + 0.02). This can be seen in the decay of the aromatic peak intensity (Fig. 3 ) ; the faster relaxing Gaussian component for the protonated aromatic carbon has virtually decayed completely after ca. 60/xs with the slower relaxing exponential component from the non-protonated carbon having a time constant of over 500/~s.
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
131
1,00-
From SPE
0.98
0.96
0.94
f
m
....__._.~__._------
0.92
0.90
0.88
0.86
0.84
j
j
2
4
6
i 8
i 10
Contact time/~s
Fig. 4. Variation of aromaticity as function of contact time in CP for the coke concentrate from FCC catalyst deactivated using the residue feedstock.
After the initial decay of the protonated aromatic carbon, the intensity of the remaining non-protonated carbon is modulated by the rotation of the rotor at 5 kHz, this effect being particularly evident in Fig. 3 for the CP experiment (note that this modulation is not encountered in variable contact time CP measurements due to the much longer timescale). From the fractions of non-protonated aromatic carbon derived by SPE (Table 1 ), it is estimated that bridgehead aromatic carbons (CBR/ CAR) aCCOUnt for ca. 67 and 56% of the total aromatic carbon in the residue and VGO-derived cokes, respectively. The only assumptions needed are that (i) each aliphatic carbon is bound to one aromatic carbon, which is not unreasonable in light of the distribution of aliphatic carbon (see following) and (ii) the concentrations of heteroatoms in the cokes are relatively small (total concentration corresponding to less than ca. 2 mole % carbon). If peri-condensed aromatic structures are drawn to fit the CBR/CARvalues (Table 1 ), 15-20 rings are required for the residue feedstock coke compared to only 8-12 tings for the HVGO sample. This represents a significant difference in aromatic structure which is considered to arise primarily from the major differences in feedstock composition and, along with other factors, particularly the Ni and V concentrations, could well affect combustion behaviour in the regenerator. Although the cokes are clearly highly aromatic in character, some information on the distribution of aliphatic groups can also be obtained from the 13C NMR spectra. Intuitively, one would expect virtually all the aliphatic carbon to be adjacent to aromatic tings in either arylmethyl or diarylmethylene groups. The fraction of aliphatic carbon present as methyl has been estimated from the intensity of the 1024 ppm chemical shift range (Table 1, a fairly clear separation between CH2 and CH3 chemical shifts for the aliphatic structures in high-rank coals occurs at 23-24 ppm). Fig. 1 indicates that CP overestimates the fraction of CH3 for the residuederived coke; CH3 cross polarises at a slower rate than C H 2 which is reflected in
132
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
characteristic time constants, TCH(Table 1 ). The SPE spectrum (Fig. 1 ) suggests that arylmethyl groups account for ca. 75% of the aliphatic carbon in the residuederived coke (Table 1 ). Although the signal to noise levels of the aliphatic bands are not good, this fraction appears to be much higher than for the HVGO-derived coke where CH3 accounts for approximately only half of the aliphatic carbon (Fig. 2 and Table 1). The larger proportion of CH2 (and CH if present) in the HVGO-defived coke is again consistent with the differences in composition between the two feedstocks. By definition, the more aliphatic HVGO contains higher concentrations of both long chain alkyl and naphthenic moieties than the atmospheric residue. The demineralisation-quantitative 13C NMR methodology successively demonstrated here for refinery catalysts is now being extended to cokes at the 1% w / w level prepared using n-hexadecane as the feed in a laboratory-scale fluidised-bed reactor to help elucidate the influence of different compound classes on coke structure.
Acknowledgements The authors thank the Engineering and Physical Sciences Research Council (EPSRC) for financial support of this work (Grant No. GR/H/24990 including a studentship for B.J. McGhee) and Dr. N. Gudde of BP Oil International Ltd. (Oil Technology Centre, Sunbury-on-Thames, Middlesex TW 16 7LN, UK) for supplying the deactivated FCC catalysts.
References [ 1] [ 2] [3] [4] [5] [6] [7] [8] [9] [ 10] [ 11 ] [ 12] [ 13 ] [ 14]
J.B. Butt, Adv. Chem. Ser., 109 (1972) 259 and refs. therein. R. Hughes, Deactivation of Catalysts, Academic Press, London, 1984. E.H. Wolf and A. Alfani, Cat. Rev. Sci. Eng., 24 (1982) 329 and refs. therein. J.R. Kittrell, P.S. Tam and J.W. Eldridge, Hydrocarbon Process., 64 (8) (1985) 63. W.A. Groten, B.W. Wojciechowski and B.K. Hunter, J. Catal., 125 (1990) 311. J.P. Lange, A. Gutsze, J. Allgeier and H.G. Karge, Appl. Catal., 45 (1988) 345. N.O. Egiebor, M.R. Gray and N. Cyr, Appl. Catal., 55 (1989) 81. C.E. Snape, D.E. Axelson, R.E. Botto, J.J. Delpuech, P. Tekely, B.C. Gerstein, M. Pruski, G.E. Maciel and M.A. Wilson, Fuel, 68 (1989) 547 and refs. therein. J.A. Franz, R. Garcia, J.C. Linehan, G.D. Love and C.E. Snape, Energy and Fuels, 6 (1992) 598. G.D. Love, R.V. Law and C.E. Snape, Energy and Fuels, 7 (1993) 639. M.M. Maroto-Valer, G.D. Love and C.E. Snape, Fuel, 73 (1994) 1926. J.D. Saxby, Chem. Geol., 6 (1970) 173; J.D. Saxby, in T.F. Yen and G.V. Chilingarian (Editors), Oil Shale, Elsevier, 1976, p. 103. B. Durand and G. Nicaise, in B. Durand (Editor), Kerogen-insoluble organic matter from sedimentary rocks, Editions Technip, Paris, 1980, p. 35. D.A. Torchia, J. Magn. Reson., 30 (1978) 613.
'~9 ~
AP PA LL IEYDSS C AT I A: GENERAL
ELSEVIER
Applied Catalysis A: General 129 (1995) 125-132
Characterisation of coke from FCC refinery catalysts by quantitative solid state 13C NMR C.E. Shape
a,*,
B.J. M c G h e e a, J.M. Andresen C.L. Koon b, G. Hutchings c
a, R.
Hughes
b,
~'UniversiO, of Strathclyde, Dept. of Pure and Applied Chemistry, Glasgow G1 1XL, UK b Dept. of Chemical Engineering, University ofSalford, SalJbrd, M5 4WT, UK c Leverhulme Centre for Innovative Catalysis, Dept. of Chemistry, Liverpool University, Liverpool L69 3BX, UK Received 8 March 1995; revised 27 April 1995; accepted 27 April 1995
Abstract Coke has been concentrated from two deactivated FCC refinery catalysts via demineralisation to facilitate detailed characterisation by solid state ~3C NMR. The catalysts were obtained from runs with a residue feed (5% Conradson carbon) and a hydrotreated vacuum gas oil ( H V G O ) . As for solid fuels, the use of a low-field spectrometer in conjunction with the single pulse excitation (SPE or Bloch decay) technique has enabled quantitative carbon skeletal parameters to be obtained for the cokes. Internal standard measurements demonstrated that most of the carbon was observed by SPE and, therefore, NMR-invisible graphitic layers are not thought to be major structural features of the cokes. SPE gave much higher values for both the carbon aromaticities and the proportions of nonprotonated aromatic carbon than the less quantitatively reliable cross-polarisation (CP) technique. Differences in feedstock composition were reflected in the structure of the cokes with the aromatic nuclei being more highly condensed in the residue-derived coke and corresponding to 15-20 pericondensed aromatic rings. Keywords: Coke; FCC; Nuclear magnetic resonance
1. Introduction In view of the importance of fluid catalytic cracking (FCC) to petroleum refining, the deactivation of cracking catalysts via coke deposition has been the subject of considerable investigation over the past 50 years [ 1-3]. Coke deposition is clearly relevant to the more recent trends in FCC operations of using heavier feeds with higher overall conversions to gasoline [4]. As well as being formed via the actual * Corresponding author. 0926-860X/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDIO926-860X(95)O0104-2
126
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
cracking reactions associated with the strongly acidic catalytic sites, coke can arise in FCC units from (i) normal thermal reactions, (ii) dehydrogenation reactions promoted by metals - - N i and V in heavy feeds - - a n d (iii) entrained catalyst products, symptomatic of incomplete stripping in the regenerator. The contributions of these different mechanisms is clearly going to be dependent on the type of feedstock, together with the design and operation of particular FCC units. However, the lack of detailed basic knowledge on the chemical nature of coke is preventing progress towards a proper overall mechanistic understanding and a clear identification of the different coke-forming routes in FCC operations. Fundamental deactivation studies involving NMR on zeolites thus far have generally involved excessively high concentrations of coke in relation to normal FCC operation where catalysts are regenerated typically after only ca. 1% w / w of carbon has been deposited. The high concentrations have been necessary to achieve sufficient sensitivity for characterising the coke on the deactivated catalysts by solid state 13C NMR. For example, Groten et al. [5] have investigated coke formation on zeolite USHY with 1-hexene as the feed with coke levels of ca. 5% w/w. Lange and coworkers [6] used 13C-enriched ethene to follow the formation of polyaromatic structures on H-mordenite. Deactivated y-alumina-supported hydroprocessing catalysts have also been characterised by solid state 13C NMR [7] but, again due to the relatively low sensitivity, the spectra have been obtained by cross polarisation (CP) which fails to observe all the carbon in coals and oil shales and usually discriminates heavily against aromatic carbon [ 8]. Further problems are posed by high magnetic field strengths where, to eliminate spinning sidebands, special pulse sequences or extremely rapid magic-angle spinning (MAS, > 10 kHz) are needed. There is a general consensus that the use of low field strengths with Bloch decay or single pulse excitation (SPE) offers the best compromise for quantitative 13C NMR analysis of coals and solid fuels [ 8-11 ], albeit with a considerable sacrifice in sensitivity since long recycle times (with 90 ° pulses, 5 times the 13C thermal relaxation times - - T l s ) are required to ensure that the 13C magnetisation fully regains equilibrium. In principle, the only carbon not observed is that in the vicinity of paramagnetic centres which obviously includes graphite. The only way this methodology can be applied successfully to deactivated catalysts is by demineralising the aluminosilicate matrices to concentrate the coke. In this communication, this approach has been used for the first time to characterise two coke concentrates isolated from deactivated FCC catalysts containing only ca. 1% w / w carbon. The samples chosen for investigation were from refinery runs with an atmospheric residue and a hydrogenated vacuum gas oil (HVGO) in order to ascertain whether the use of these two vastly different feedstocks had a significant effect on coke composition. The degree of condensation of the aromatic structure has been assessed from the proportions of non-protonated aromatic carbon derived by the SPE 13C NMR technique.
C.E. Snape et al. / Applied Catalysis A: General 129 (1995) 125-132
127
2. Experimental 2.1. Catalysts and their demineralisation The FCC catalysts were typical commercial formulations and the deactivated samples were obtained from units processing (i) a heavy feedstock containing ca. 1.5% sulphur, a Conradson carbon content of 5.0% w / w and Ni and V contents of 5 ppm each and (ii) a HVGO containing only 0.1% sulphur and below 2 ppm of Ni and V. The catalysts 'as received' both had carbon contents close to 1%. They were first refluxed in chloroform for 3 hours to remove any trapped molecular species and then vacuum-dried prior to demineralisation. This initial treatment reduced the carbon content of the catalyst deactivated with the atmospheric residue from 1.08 to only 1.00% (estimated error of +_0.05%), indicating that molecular species only account for an extremely small proportion of the coke. This observation is consistent with the fact that both catalysts were removed from the FCC units after stripping (collected at the base of the stripper). The cokes were concentrated by applying the standard demineralisation procedure for solid fuels [12], [ 13] to the chloroform-extracted catalysts. This involved successive extraction with 2 M hydrochloric acid (stirring overnight at 60°C) and 40% hydrofluoric acid (HF), the HCl-extracted sample being stirred at room temperature for 4 hours with 20 cm 3 of HF being used per gram of sample. The coke concentrates were finally washed with dilute hydrochloric acid to remove any remaining inorganic paramagnetics prior to collection in plastic filtration equipment. The vacuum-dried coke concentrates recovered from the catalysts deactivated by the residue and HVGO feedstocks had carbon contents of 55 and 39%, respectively.
13C NMR The CP and SPE 13C NMR measurements on the coke concentrates were carried out as previously for coals [9-11] at 25 MHz on a Bruker MSL100 spectrometer with MAS at 4.5-5.0 kHz to give spectra in which the sideband intensities are only ca. 3% of the central aromatic bands. Known weights of tetrakis (trimethylsilyl) silane (TKS) were added to the samples as the internal reference and to facilitate estimation of the fraction of the total carbon observed by SPE. Approximately ca. 150 mg of sample was packed into the zirconia rotors. The 1H decoupling and spinlock field was ca. 60 kHz and, for SPE, the 90 ° 13C pulse width was 3.4/xs. A recycle delay of 50 s was employed between successive 90 ° pulses in SPE since the ~3C T~ for the non-protonated aromatic carbons were 10 s as measured by Torchia's CP-based method [ 14] for both coke concentrates 9(Table 1). Normal CP spectra of the residue-derived coke were obtained using a range of contact times between 0.05 and 8 ms for the residue-derived coke to facilitate determination of the time constants for CP (Tc~) and ~H rotating-frame relaxation (Tip). The CP spectrum of the HVGO-derived coke was obtained using a contact time of 1 ms.
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125-132
128
Table 1 NMR and other structural parameters for the coke concentrates Parameter ~3C T~/s for non-protonated aromatic carbon CP-derived parameters obtained from variable contact time experiment, TcH/l~s, aromatic C b CH2 CH 3
TJms, aromatic CH CH2 CH 3
Residue feedstock 10.0 330 40 70 8 6 5
HVGO feedstock 11.5 N.D. N.D. N.D. N.D, N.D, N.D,
Carbon aromaticity,fa, ( _+0.01 )a CP (1 ms) SPE
0.91 0.97
0.85 0.96 N.D.
0.55 0.67 0.63 0.75 70
N.D 0.56 0.51 0.50 90
Fraction of aromatic carbon that is non-protonated from dipolar dephasing experiments ( + 0.02) a, CP ( 1 ms) SPE
Fraction of aromatic carbon that is bridgehead, CBR/CAR(SPE5: 0.03) a Fraction of aliphatic carbon that is CH3 ( 10-24 ppm range, SPE + 0.1 a) % of carbon observed by SPE
N.D. = not determined; SPE = single pulse excitation; CP = cross-polarisation.
Estimated errors, b Average value for protonated and non-protonated carbons.
Dipolar dephasing (DD) was combined with both SPE and CP to estimate the proportions of protonated and non-protonated aromatic carbon, and at least 8 separate dephasing periods in the range 5-500/xs were used (500 scans each), In order to check that the tuning had remained virtually constant throughout the duration of the DD experiments, the delays were arranged in a random order and between 500 and 1000 scans were accumulated for each delay. The spectra were processed using exponential line broadening factors of either 30 or 50 Hz. No background signal was evident in the SPE spectra from the Kel-F rotor caps. The measurement of aromatic and aliphatic peak areas manually was found to be generally more precise than using the integrals generated by the spectrometer software.
3. Results and discussion Fig. 1 and Fig. 2 compare the CP and SPE 13C spectra of the highly aromatic coke concentrates and Fig. 3 presents the decays of the aromatic peak intensities in the CP and SPE-DD experiments on the residue-derived coke. The carbon skeletal parameters obtained from the SPE and CP 13C NMR experiments for both samples are summarised in Table 1. The first coke concentrate obtained from the catalyst deactivated with the residue feedstock gave a broad aromatic band, possibly due to the incomplete removal of the rare-earth metal during the HC1 wash. Indeed, it was found that, after the final HC1 wash, much narrower spectral bands were obtained
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125-132
129
Aromatic carbon
TKS
Aliphatic carbon SPE
. . . .
200t . . . .
1501 . . . .
1001 ~ ,
.
, i0 . . . .
0t . . . .
l
PPH
Fig. 1. CP ( 1 ms contact time) and SPE t3C NMR spectra of the coke concentrate from FCC catalyst deactivated using the residue feedstock.
J
SPE
~ .
.
.
.
L
200
.
.
.
.
l
TSO
.
.
.
.
~ L
.
100 PPH
.
.
, .
i
50
~ .
.
.
.
CP I 0
.
.
i
,
Fig. 2. CP ( 1 ms contact time) and SPE 13C NMR spectra of the coke concentrate from FCC catalyst deactivated using the HVGO feedstock.
130
C.E. Snape et al./Applied Catalysis A: General 129 (1995) 125 132 3-
~e
[] m
•
•
,;0
2;0
[]
CP
•
SPE
£
eL
1 ¸
,;0
,;0
~;0
Tau/~s Fig. 3. Decay of the natural logarithm of the aromatic peak intensity in the CP and SPE dipolar dephasing experiments on the coke concentrate from the FCC catalyst deactivated using the residue feedstock ( initial intensity assigned value of 10).
and the quality of the spectra shown are comparable to those obtained for lowvolatile coals and anthracites (with similar aromaticities as the cokes, see following). The sensitivity is obviously superior in the CP spectra (7-10 000 scans for CP compared to 1000-2000 for SPE, Fig. 1 and Fig. 2) but, as found for many coals [9-11 ], CP significantly underestimates the carbon aromaticity using relatively short contact times (Table 1, CP values for 1 ms contact, estimated error _+0.01 ). This arises from the non-protonated aromatic carbons, in particular, cross polarising much more slowly than the aliphatic carbons (both C H 2 and CH3 in the residue coke have much shorter TcHs than the aromatic carbon, Table 1 ). At longer contact times, the discrimination against aromatic carbon is not quite as acute, but the aromaticity values obtained are still lower than that by SPE (Fig. 4). This is indicative that some aromatic carbon, presumed to be in the vicinity of free radicals, is not polarised at all (~H T~ps are too short - - < c a . 0.2 ms, the values in Table 1 are only for hydrogens adjacent to the carbons being polarised). The fact that 70 and 90% of the carbon in the residue and HVGO-derived coke concentrates, respectively, has been observed by SPE demonstrates that the procedure is reasonably quantitative and that graphitic layers are probably not present in significant amounts. XRD is being used to check the possibility that the cokes may contain some extended graphitic structure. However, if it was present in significant amounts, the resultant paramagnetism would have detuned the probe resulting in little of the carbon being detectable. As for the total aromaticity, CP also grossly underestimates the fraction of nonprotonated aromatic carbon determined by dipolar dephasing (Table 1, estimated error + 0.02). This can be seen in the decay of the aromatic peak intensity (Fig. 3 ) ; the faster relaxing Gaussian component for the protonated aromatic carbon has virtually decayed completely after ca. 60/xs with the slower relaxing exponential component from the non-protonated carbon having a time constant of over 500/~s.
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
131
1,00-
From SPE
0.98
0.96
0.94
f
m
....__._.~__._------
0.92
0.90
0.88
0.86
0.84
j
j
2
4
6
i 8
i 10
Contact time/~s
Fig. 4. Variation of aromaticity as function of contact time in CP for the coke concentrate from FCC catalyst deactivated using the residue feedstock.
After the initial decay of the protonated aromatic carbon, the intensity of the remaining non-protonated carbon is modulated by the rotation of the rotor at 5 kHz, this effect being particularly evident in Fig. 3 for the CP experiment (note that this modulation is not encountered in variable contact time CP measurements due to the much longer timescale). From the fractions of non-protonated aromatic carbon derived by SPE (Table 1 ), it is estimated that bridgehead aromatic carbons (CBR/ CAR) aCCOUnt for ca. 67 and 56% of the total aromatic carbon in the residue and VGO-derived cokes, respectively. The only assumptions needed are that (i) each aliphatic carbon is bound to one aromatic carbon, which is not unreasonable in light of the distribution of aliphatic carbon (see following) and (ii) the concentrations of heteroatoms in the cokes are relatively small (total concentration corresponding to less than ca. 2 mole % carbon). If peri-condensed aromatic structures are drawn to fit the CBR/CARvalues (Table 1 ), 15-20 rings are required for the residue feedstock coke compared to only 8-12 tings for the HVGO sample. This represents a significant difference in aromatic structure which is considered to arise primarily from the major differences in feedstock composition and, along with other factors, particularly the Ni and V concentrations, could well affect combustion behaviour in the regenerator. Although the cokes are clearly highly aromatic in character, some information on the distribution of aliphatic groups can also be obtained from the 13C NMR spectra. Intuitively, one would expect virtually all the aliphatic carbon to be adjacent to aromatic tings in either arylmethyl or diarylmethylene groups. The fraction of aliphatic carbon present as methyl has been estimated from the intensity of the 1024 ppm chemical shift range (Table 1, a fairly clear separation between CH2 and CH3 chemical shifts for the aliphatic structures in high-rank coals occurs at 23-24 ppm). Fig. 1 indicates that CP overestimates the fraction of CH3 for the residuederived coke; CH3 cross polarises at a slower rate than C H 2 which is reflected in
132
C.E. Snape et al. /Applied Catalysis A: General 129 (1995) 125 132
characteristic time constants, TCH(Table 1 ). The SPE spectrum (Fig. 1 ) suggests that arylmethyl groups account for ca. 75% of the aliphatic carbon in the residuederived coke (Table 1 ). Although the signal to noise levels of the aliphatic bands are not good, this fraction appears to be much higher than for the HVGO-derived coke where CH3 accounts for approximately only half of the aliphatic carbon (Fig. 2 and Table 1). The larger proportion of CH2 (and CH if present) in the HVGO-defived coke is again consistent with the differences in composition between the two feedstocks. By definition, the more aliphatic HVGO contains higher concentrations of both long chain alkyl and naphthenic moieties than the atmospheric residue. The demineralisation-quantitative 13C NMR methodology successively demonstrated here for refinery catalysts is now being extended to cokes at the 1% w / w level prepared using n-hexadecane as the feed in a laboratory-scale fluidised-bed reactor to help elucidate the influence of different compound classes on coke structure.
Acknowledgements The authors thank the Engineering and Physical Sciences Research Council (EPSRC) for financial support of this work (Grant No. GR/H/24990 including a studentship for B.J. McGhee) and Dr. N. Gudde of BP Oil International Ltd. (Oil Technology Centre, Sunbury-on-Thames, Middlesex TW 16 7LN, UK) for supplying the deactivated FCC catalysts.
References [ 1] [ 2] [3] [4] [5] [6] [7] [8] [9] [ 10] [ 11 ] [ 12] [ 13 ] [ 14]
J.B. Butt, Adv. Chem. Ser., 109 (1972) 259 and refs. therein. R. Hughes, Deactivation of Catalysts, Academic Press, London, 1984. E.H. Wolf and A. Alfani, Cat. Rev. Sci. Eng., 24 (1982) 329 and refs. therein. J.R. Kittrell, P.S. Tam and J.W. Eldridge, Hydrocarbon Process., 64 (8) (1985) 63. W.A. Groten, B.W. Wojciechowski and B.K. Hunter, J. Catal., 125 (1990) 311. J.P. Lange, A. Gutsze, J. Allgeier and H.G. Karge, Appl. Catal., 45 (1988) 345. N.O. Egiebor, M.R. Gray and N. Cyr, Appl. Catal., 55 (1989) 81. C.E. Snape, D.E. Axelson, R.E. Botto, J.J. Delpuech, P. Tekely, B.C. Gerstein, M. Pruski, G.E. Maciel and M.A. Wilson, Fuel, 68 (1989) 547 and refs. therein. J.A. Franz, R. Garcia, J.C. Linehan, G.D. Love and C.E. Snape, Energy and Fuels, 6 (1992) 598. G.D. Love, R.V. Law and C.E. Snape, Energy and Fuels, 7 (1993) 639. M.M. Maroto-Valer, G.D. Love and C.E. Snape, Fuel, 73 (1994) 1926. J.D. Saxby, Chem. Geol., 6 (1970) 173; J.D. Saxby, in T.F. Yen and G.V. Chilingarian (Editors), Oil Shale, Elsevier, 1976, p. 103. B. Durand and G. Nicaise, in B. Durand (Editor), Kerogen-insoluble organic matter from sedimentary rocks, Editions Technip, Paris, 1980, p. 35. D.A. Torchia, J. Magn. Reson., 30 (1978) 613.