Coking Kinetics of Asphaltenes as a Function of Chemical Structure

Dicho S. Stratiev , Cristopher A. Russell , Ron Sharpe , Ivelina K. Shishkova , Rosen K. Dinkov , Ivaylo M. ... Studies in Coking of Arabian Mix Vacuu...
0 downloads 0 Views 180KB Size
Download PDF
1048

Energy & Fuels 2003, 17, 1048-1056

Coking Kinetics of Asphaltenes as a Function of Chemical Structure Samina Rahmani,† William C. McCaffrey,† Heather D. Dettman,‡ and Murray R. Gray*,† Department of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta, T6G 2G6, Canada, and National Centre for Upgrading Technology, 1 Oil Patch Drive, Devon, Alberta, Canada T9G 1A8 Received January 9, 2003. Revised Manuscript Received May 15, 2003

The dependence of coking kinetics on the chemical structure of asphaltenes was examined by reacting five different asphaltenes in 1-methylnaphthalene and tetralin at 430 °C and ca. 9.8 MPa. The selected heptane-insoluble asphaltenes were Athabasca asphaltenes from Canada, Arabian Light and Arabian Heavy from Saudi Arabia, Maya from Mexico, and Gudao from China. The 13C NMR aromaticity of the asphaltenes ranged from 0.40 to 0.61, and the sulfur contents ranged from 4.44 wt % to 7.47 wt %. The cracking kinetics of the asphaltenes were consistent with a modified kinetic model for coke formation, incorporating phase separation and hydrogen transfer to the asphaltenes. The rate of cracking of asphaltenes in 1-methylnaphthalene correlated with the content of aliphatic sulfur, and the yield coefficient for coke correlated with the aromaticity. These correlations allowed prediction of coking kinetics for Iranian Light and Khafji asphaltenes on the basis of average structural properties of asphaltenes. Hydrogen transfer to the different asphaltenes did not correlate with any single average structural property.

Introduction Primary upgrading of bitumen and petroleum residues to distillate products is achieved by either coking or catalytic hydroconversion. During coking processes, the objective is to maximize the yield of cracked products, while also producing a coke material of the desired quality. During hydroconversion processes, the process objective is maximum yield of cracked product without formation of coke deposits. In both types of processes, formation of coke limits the possible conversion to distillable liquid products. Improvements in the operability and efficiency of these units requires control or management of coke formation, which can only be achieved through a better understanding of the fundamental mechanisms involved. Coke is defined operationally in petroleum refining as a carbonaceous material that is insoluble in an aromatic solvent such as benzene or toluene. The yield of coke in thermal and catalytic processes increases with the concentration of petroleum asphaltenes in the feed, defined as the pentane- or heptane-insoluble and toluenesoluble organic fraction of bitumen or vacuum residue.1,2 Heavy oil and residues can contain significant amounts of asphaltenes, in the range of 10-30 wt %; therefore, an understanding of asphaltene reaction pathways will help to develop an understanding of the mechanisms of * Author to whom correspondence should be addressed. Tel: (780) 492-7965. Fax: (780) 492-2881. E-mail: [email protected]. † University of Alberta. ‡ National Centre for Upgrading Technology. (1) Speight, J. G. The Desulfurization of Heavy Oils and Residua; Marcel Dekker Inc.: New York, 1981. (2) Speight, J. G. In Catalysis on the Energy Scene; Kaliaguine, S., Mahay, A., Eds.; Elsevier: Amsterdam, 1984.

coke formation. Asphaltene reaction pathways are affected by the chemistry of the feed material, the processing parameters of temperature and pressure, solvent properties, and the supply of reactive hydrogen.1 Prior treatments, such as distillation and blending of recycle material, will affect the feed composition in residue conversion processes. Any change in the feed composition may influence the rate and extent of coke formation, and also the product composition and yield. Despite the importance of feed composition, few studies have considered the relationship between thermal reactivity and chemical structure.3,4 The a priori prediction of coke formation from the chemical structure of petroleum and bitumen residues during thermal conversion is not yet possible. The objective of this work was to examine the behavior of a range of heptane-insoluble asphaltenes in the liquid phase, to predict coking kinetics from feed properties. Feed asphaltenes were selected to give a range of average chemical properties. Based on the available data from the literature,5-7 Athabasca asphaltenes from Canada, Arabian Light and Arabian Heavy from Saudi Arabia, Maya from Mexico, and Gudao from China were selected. These asphaltenes (3) Liu, C.; Zhu, C.; Jin, L.; Shen, R.; Liang, W. Fuel Process. Technol. 1999, 59, 51-67. (4) Gray, M. R.; Jokuty, P.; Yeniova, L.; Nazarewycz, L.; Wanke, S. E.; Achia, U.; Krzywicki, A.; Sanford, E. C.; Sy, O. K. Y. Can. J. Chem. Eng. 1991, 69, 833-843. (5) Zhou, J.; Mu, B.; Que, G. Prepr. Pap.sAm. Chem. Soc., Div. Pet. Chem. 1994, 39 (3), 426-429. (6) Trauth, D. M.; Yasar, M.; Neurock, M.; Nigam, A.; Klein, M.; Kukes, S. G. Fuel Sci. Technol. Int. 1992, 10, 1161-1179. (7) Ali, M. F.; Saleem, M. Arab. J. Sci. Eng. 1994, 19 (2B), 319333.

10.1021/ef030007c CCC: $25.00 © 2003 American Chemical Society Published on Web 06/17/2003

Coking Kinetics of Asphaltenes and Chemical Structure

span a range of sulfur and nitrogen contents, and a range of aromatic carbon content. The solvent 1-methylnaphthalene was used in pressurized batch reactors in order to control the liquid environment during cracking. Comparison with tetralin solvent allowed exploration of the role of hydrogen transfer in a liquid medium of similar solubility parameter.8 The phase separation model proposed by Wiehe9 and modified by Rahmani et al.8 was used as a kinetic framework to analyze the coke yields. Theory The coke formation model proposed by Wiehe9 was extended by Rahmani et al.8 to incorporate hydrogen transfer from solvents. During cracking, the asphaltene fraction was assumed to consist of A+, the unreacted asphaltenes with attached side groups, and two types of aromatic cores formed by cracking of the initial asphaltenes, each with different hydrogen-accepting capability. Cores that can accept enough hydrogen to change their solubility characteristics were designated A*A, and cores that cannot accept hydrogen were designated A*NA. At any time, A*A + A*NA ) A*, the total concentration of aromatic cores. The cores that accepted sufficient hydrogen from a donor such as tetralin (TN) were assumed to be converted to heptanesoluble material. The reaction model was as follows for asphaltenes in aromatic solvents: kA

A+ 98 (c - d)A*NA + dA*A + (1 - c)(H* + V) (1) k′[TN]

A*A 98 H*

(2)

where H* is the fraction of product heptane solubles, kA is the first-order reaction rate constant for the thermal cracking of reactant asphaltene and d and c are stoichiometric coefficients. The apparent first-order rate constant for hydrogen transfer from solvent was the product of the rate constant (k′) and the tetralin concentration, [TN]. In the case of a closed reactor, the cracked distillate products (V) are retained and remain mainly in the liquid phase. The solubility limit (SL) in a closed reactor defines how much reacted asphaltene / ) can remain in solution in the liquid phase: (Amax

A/max ) SL(S + H* + V)

(3)

where S ) solvent (1-methylnaphthalene or tetralin). The asphaltenes in excess of the solubility limit (A/ex) then react to form toluene-insolubles, or coke.

A/ex ) A*NA + A*A - A/max kc

A/ex 98 TI

(4) (5)

where TI ) toluene-insoluble coke and kc is the firstorder rate constant for formation of insolubles. Experimental Methods Materials. Asphaltenes from five different sources were used to study the effect of asphaltene structure on the kinetics. Athabasca, Gudao, Arab Heavy and Arab Light asphaltenes (8) Rahmani, S.; McCaffrey, W.; Gray, M. R. Energy Fuels 2002, 16, 148-154. (9) Wiehe, I. A. Ind. Eng. Chem. Res. 1993, 44, 2447-2457.

Energy & Fuels, Vol. 17, No. 4, 2003 1049 were separated from the respective vacuum residues, and Maya asphaltenes were separated from the crude oil. The toluene-insoluble solids were removed from Athabasca by dissolving vacuum residue in 40 parts of toluene for overnight and then filtering it on 0.22 µm filter paper. Toluene was evaporated by rotary evaporation. The other feeds had insignificant concentrations of solids; therefore, this step was omitted. Heptane-insoluble asphaltene was separated by blending oil 1:1 with toluene, then adding 40 parts of heptane and mixing overnight. The asphaltene precipitate was recovered by filtration on a 0.22 µm Millipore filter. The average molecular weights were obtained using vapor pressure osmometry (Westcan Corona model 232A) at the Micro Analytical Laboratory at the University of Alberta. The samples were dissolved at a single concentration of ca. 2 mg/ mL in o-dichlorobenzene and analyzed at a temperature of 130 °C. Elemental analysis was performed on a Carlo Erba Elemental Analyzer 1108 at the Micro Analytical Laboratory, University of Alberta. Thermal Cracking Experiments. Experiments were carried out batchwise in a 15 mL microreactor made from Swagelock fittings and tubing. The reactor was loaded with 3 g of reactant and then pressure tested with nitrogen at 4 MPa. The gas was then vented, and the reactor was pressurized twice more with nitrogen and vented to purge residual oxygen. The reactor was then filled again with nitrogen to a pressure of 4 MPa. The reactor was heated in a fluidized sand bath and agitated at ca. 1 Hz for the duration of the reaction interval, then quenched by plunging the reactor into cold water. The contents of the reactor reached the final temperature within 5 min, and an initial nitrogen pressure of ca. 9.8 MPa. All of the reactions were carried out at 430 °C. Approximately 85% of the solvents were estimated to be in the liquid phase at 430 °C, using the Peng-Robinson equation of state (ASPEN Plus software, Aspen Technology, Cambridge, MA). Separation of Products. Liquid product and coke were washed out of the reactor with 40 parts of toluene, then kept overnight at 70 °C to ensure the extraction of liquid products from the solid coke. Coke was then removed from the toluene solution by filtration on a 0.22 µm Millipore filter. Coke yield was determined by weighing the filter after drying in a vacuum at 70 °C for 12 h. Toluene was removed from the filtrate by rotary evaporation, the oil was then blended with 40 parts of n-heptane to precipitate asphaltenes. Solid asphaltenes were recovered by filtration and vacuum-dried at 70 °C for 12 h to give the asphaltene yield. Heptane was removed from the heptane-soluble fraction by rotary evaporation and then by vacuum-drying at 70 °C. GC Analysis of Solvents. The heptane-soluble fractions of the product were analyzed by GC in order to estimate the hydrogen transfer from tetralin to asphaltenes. Under thermal reaction conditions, tetralin can dehydrogenate to produce 1,2dihydronaphthalene and naphthalene in succession, isomerize to form 1-methyl indan, or crack to form alkyl benzenes. The quantity of hydrogen transferred from tetralin was calculated from the amount of naphthalene formed, given that for every mole of naphthalene produced, 4 atoms of hydrogen were transferred from tetralin. Formation of 1,2-dihydronaphthalene was not detected. A Hewlett-Packard 5890 GC using a HP-1 cross-linked methyl silicone gum column (25m, 0.32 mm i.d., 0.17 µm) equipped with a flame ionization detector (FID) and a computer for storing the chromatograms. The temperature program for the GC was as follows: initial oven temperature of 50 °C, then heating to 300 °C at a rate of 5 °C/min and then held at constant temperature for 15 min. Nuclear Magnetic Resonance (NMR) Spectroscopy. Samples were prepared for 1H NMR by mixing approximately 20 mg of the sample with 700 µL of deuteriochloroform (CDCl3). Samples for 13C NMR samples were prepared by dissolving approximately 50 mg of material in 700 µL of CDCl3.

1050

Energy & Fuels, Vol. 17, No. 4, 2003

Table 1. Chemical Shifts of Proton Spectral Regions

Rahmani et al. Table 3. Group Classifications of Chemical Species Reported for Asphaltene Samples

region

chemical shifts (ppm)

structural type

carbon chemical species

spectral region

HA1 HA2 HO1 HO2 HO3 HP1 HP2 HP3 HP4

10.7 to 7.4 7.4 to 6.2 6.2 to 5.1 5.1 to 4.8 4.8 to 4.3 4.3 to 2.4 2.4 to 2.0 2.0 to 1.09 1.09 to -0.5

polyaromatic monoaromatic olefinic CH olefinic CH2 olefinic CH2 R-to-aromatic CH2 R-to-aromatic CH3 paraffinic CH2 paraffinic CH3

aromatic CH (CHar) methyl of an aromatic ethyl (E-CH3) terminal methyl of a paraffinic chain paraffin-substituted quaternary (Qar-S)

CA3 CP8 CP9 HP1 + HP2 converted to C wt % CA2 - Qar-S HP2 converted to C wt % CP1 + CP2/2 CP4 - CP8 - CP5/2 CP5

Table 2. Chemical Shifts of Carbon Spectral Regions region

chemical shifts (ppm)

CA1 CA2 CA3 CA4 CA5 CP1 CP2 CP3 CP4

190 to 170 170 to 129 129 to 115.5 115.5 to 113.5 113.5 to 100 70 to 45 45 to 32.7 32.7 to 30.8 30.8 to 28.5

CP5 CP6

28.5 to 25 25 to 21.9

CP7 CP8 CP9 CP10

21.9 to 17.6 17.6 to 14.7 14.7 to 12.3 12.3 to 0

structural type oxygenated quaternary aromatic aromatic CH olefinic CH2 olefinic CH2 paraffinic CH paraffinic CH & CH2 chain γ-CH2 β-to-aromatic CH2 chain δ-CH2 R-to-aromatic naphthenes, aromatic-attached ethyl CH2 cycloparaffin CH2 chain β-CH2, R to aromatic or isobutyl CH3 R-to-ring CH3 aromatic-attached ethyl CH3 chain R-CH3 branched-chain CH3

The NMR experiments were performed at room temperature (20 ( 1 °C) on a Varian XL-300 NMR spectrometer, operating at 299.943 MHz for proton and 75.429 MHz for carbon. The proton spectra were collected with an acquisition time of 2.1 s, a sweepwidth of 7000 Hz, a pulse flip angle of 30.8° (8.2 µs), and a 1-s recycle delay. These pulse recycle conditions permit the collection of quantitative spectra for all protonated molecular species in the petroleum samples where the maximum spin lattice relaxation time (T1) is expected to be less than 20 s. The spectra, resulting from 128 scans and using 0.3-Hz line broadening, were referenced to the residual chloroform resonance at 7.24 ppm. The quantitative carbon spectra were acquired using an acquisition time of 0.9 s and a sweepwidth of 16 500 Hz. A flip angle of 31.9° (5.7 µs) and a 4-s delay were used. These parameters are quantitative for carbons with spin lattice relaxation times (T1) of the order of 30 s. Reverse-gated waltz proton decoupling was used to avoid nuclear Overhauser effect enhancements of the carbon signals. The spectra were the result of 15 000 scans. Line broadening at 10 Hz was used to improve the signal-to-noise ratio of the spectra. All spectra were referenced to the CDCl3 resonance at 77.0 ppm. The chemical shift assignments for the 1H and 13C NMR are shown in Table 1 and Table 2, respectively. These assignments were based on model compound assignments,10,11 2-D NMR spectroscopic techniques such as HETCOR (heteronuclear chemical shift correlation), and COSY (homonuclear correlation spectroscopy)12 as well as 1-dimensional techniques such as DEPT (distortionless enhanced polarization transfer).13,14 The final calculations performed to obtain the quantities of the different carbon types are shown in Table 3.15 The aromatic CH’s, terminal methyls, and the methyls of aromatic (10) Snape, C. E.; Ladner, W. R. Anal. Chem. 1979, 51, 2189-2198. (11) Thiel, J.; Gray, M. R. AOSTRA J. 1988, 4, 63-73. (12) Sarpal, A. S.; Kapur, G. S.; Chopra, A. Fuel 1996, 75, 483490. (13) Netzel, D. A. Anal. Chem. 1987, 59, 1775-1779. (14) Kotlyar, L. S.; Morat, C.; Ripmeester, J. A. Fuel 1991, 70, 9094.

polyaromatic quaternary (Qar-P) R-to-aromatic ring methyl (Ar-CH3) paraffinic CH (CH) paraffin CH2 in > C5 chains (CHAIN) cycloparaffin CH2 (NAPH)

ethyls were determined directly from the carbon spectra. These groups were based on minimal assumptions and may be considered to be the most accurate. The other groups listed below were calculated using data from both proton and carbon spectra or had to be estimated due to spectral overlap and are therefore subject to larger margins of error. The substituted quaternary carbon (Qar-S) content was determined from the proton spectra by converting the R-to-aromatic CH2 and CH3 protons to their respective amounts of carbon. The remaining amount of quaternary carbon was considered to be bridgehead carbon (Qar-P). The methyl carbons attached to aromatic rings (Ar-CH3) were determined using the proton spectral region HP2. The remaining R-to-ring CH3’s were assumed to be attached to cycloparaffinic rings (Cy-CH3). The content of olefinic carbon would have been obtained from the proton spectra; however, olefinic resonances were not detected in the asphaltene samples. Half of region CP2 was estimated to be aliphatic CH’s, and the total aliphatic CH’s were calculated as the sum of region CP1 and one-half of region CP2. The CH2 carbons in paraffinic chains longer than 5 carbons were calculated from region CP4 after the contributions of CH2 carbon from aromatic-attached ethyls and R-to-aromatic cycloparaffinic CH2 were removed. The amount of cycloparaffinic CH2’s R to aromatic rings were estimated as half of the cycloparaffinic region CP5. Finally, the area of region CP5 was used as an indicator of cycloparaffinic CH2 content. Sulfide Analysis. Aliphatic sulfides were measured by selectively oxidizing them to sulfoxides followed by infrared spectroscopy (IR) determination, as described by Green et al.16 About 0.2 g of sample was dissolved in 25 mL of toluene plus 5 mL of methanol, and then about 0.2 g of tetrabutylammonium periodate was added as oxidant. Tetrabutylammonium periodate was obtained from Aldrich Chemical Co., Inc. The resulting mixture was magnetically stirred and refluxed for 30 min in a 100 mL round-bottom flask connected with a short condenser. The oxidized mixture was then cooled and extracted in a separatory funnel three times, with about 100 mL of high purity water per extraction. The organic phase was dried in a rotary evaporator, followed by an additional 2 h drying in a vacuum oven at about 70 °C. A measured quantity of the dried product, at about 0.2 g, was dissolved in dicloromethane to make a solution of 4 mL, and its sulfoxide content was measured by IR absorbance. All IR measurements were performed with a FTS 6000 infrared spectrometer with a removable cell containing KBr windows and a 0.015 mm Teflon spacer. Background was subtracted from the spectrum of the sample solution. Sulfoxides were measured via their maximum absorbance near 1025 cm-1. The molar absorptivity of 245 L mol-1 cm-1 was used to calculate sulfoxide concentration in all the asphaltenes.16 (15) Japanwala, S.; Chung, K. H.; Dettman, H. D.; Gray, M. R. Energy Fuels 2002, 16, 477-484. (16) Green, J. B.; Yu, S. K. T.; Pearson, C. D.; Reynolds J. W. Energy Fuels 1993, 7, 119-126.

Coking Kinetics of Asphaltenes and Chemical Structure

Energy & Fuels, Vol. 17, No. 4, 2003 1051

Table 4. Chemical Properties of the Feed Asphaltenes feed asphaltene

H/C ratio

sulfur (wt %)

sulfide S (wt %)

nitrogen (wt %)

molecular weight

Arab Heavy Arab Light Athabasca Gudao Maya

1.16 1.07 1.19 1.41 1.17

7.15 5.67 7.47 4.44 6.61

3.27 1.79 3.78 2.28 3.86

0.92 0.83 1.28 1.18 1.19

2600 1800 2900 2800 2500

Results and Discussion Properties of Feed Asphaltenes. Molecular weights of the feed asphaltenes are presented in Table 4. The data from VPO may suffer from the self-association of asphaltenes even after using a strong solvent,17 but on a relative basis Arab Light had the lowest molecular weight of the five samples. The other molecular weights were not significantly different from each other considering the error involved in the measured molecular weight, which is typically (10% of the molecular weight. Table 4 also lists the elemental analyses of the feed asphaltenes. Arab Light and Arab Heavy asphaltenes were low in nitrogen (0.83 and 0.91 wt %), while the other asphaltenes had similar nitrogen content (1.17∼ 1.27 wt %). The asphaltene from Gudao vacuum residue was highest in H/C atomic ratio (1.41) and Arab Light asphaltene was the lowest (1.07), while the others were similar (1.16∼1.18). Athabasca and Arab Heavy asphaltenes had the highest sulfur content (7.47 and 7.15 wt %), whereas Gudao and Arab Light had the lowest sulfur contents (4.44 and 5.67 wt %). Sulfide contents ranged from 32 to 58% of the total sulfur. NMR Analysis. The data of Table 5 list the concentrations of carbon groups in the asphaltenes. Arab Light had the highest percentage of quaternary aromatic carbon (Qar) and Gudao had the lowest. All other asphaltenes had comparable quantities of quaternary carbon. A similar trend was observed for protonated aromatic carbon. Arab Light had the highest total aromaticity and Gudao the lowest. NMR data were also used to calculate aromatic cluster size, i.e., the number of carbons in an aromatic cluster. For example, naphthalene has an aromatic cluster size of 10. This parameter was calculated using the aromatic functional group contents and the method described by Solumn et al.18 The Arab Light asphaltenes had the largest cluster size (21), followed by Athabasca and Maya (19), Arab Heavy (17), and lowest in Gudao asphaltenes (15). The asphaltenes also differed in the concentrations of aliphatic carbon groups (Table 5). Paraffin CH, indicating branched chains or substituted cycloparaffins, were highest in Arab Heavy and Gudao asphaltenes and lowest in Arab Light. Chain paraffin CH2 groups that are at least six carbons in length were highest in Gudao asphaltenes. All other asphaltenes had similar values. Cycloparaffinic (i.e., naphthenic) CH2, indicating carbons that are at least β to an aromatic ring or not attached to an aromatic ring, were highest in Gudao asphaltenes followed by Athabasca. CH2 groups that are R or β either to alkyl attachment sites in rings or chains, or to terminal methyls in chains, were lowest in Arab (17) Altgelt, K. H.; Boduszynski, M. M. Composition and Analysis of Heavy Petroleum Fractions; Marcel Dekker: New York, 1994; Chapter 4. (18) Solum, M. S.; Pugmire, R. J.; Grant, D. M. Energy Fuels 1989, 3, 187-193.

Figure 1. Residual asphaltene yields from the thermal cracking of asphaltenes in 1-methylnaphthalene. Curves are drawn from the model eqs 1 and 3-5 using the parameters of Table 6. Initial concentration was 25 wt % asphaltene in solvent. Data for Arab Heavy were almost identical to Athabasca at 20 and 40 min.

Light asphaltenes. The sum of cycloparaffinic CH2 and these R- and β-CH2 groups, roughly representing the total cylcoparaffinic carbon, was higher in Gudao and Athabasca compared to the other asphaltenes. Overall, the five samples were very different in elemental compositions and structural parameters. This result suggests that the five samples used in this study were representative of a range of variations in chemical and structural compositions in petroleum asphaltenes. Thermal Cracking of Asphaltenes in 1-Methyl Naphthalene. All the asphaltenes were reacted in 1-methyl naphthalene (MN) for a series of reaction times. A concentration of 25 wt % initial asphaltene was used for all of the reactions. Figure 1 shows the product asphaltene yield from the thermal cracking of five different asphaltenes in 1-methyl naphthalene. Gudao and Arab Light cracked more slowly than the other three asphaltenes. These two asphaltenes had the lowest amount of sulfur among the five. Yields of residual asphaltenes after long reaction times were not significantly different between the samples. After long reaction times, the unreactive asphaltene cores should remain in the solution and contribute to the product asphaltenes. The similar yields of residual asphaltenes from different feeds suggests that their solubilities in 1-methyl naphthalene were similar, or in other words, structurally the asphaltene cores from different asphaltenes were not very different from each other after reaction. To test this hypothesis, NMR spectra of residual asphaltenes from highly paraffinic Gudao and moderately aromatic Athabasca asphaltenes were collected. The difference in aromaticity between the two asphaltenes decreased from 15% in the feed to only 4% in the product. The aromaticity (without olefin) of the product asphaltenes became 73% for Athabasca and 69% for Gudao. Aliphatic carbon decreased significantly in both cases. Although the starting materials were very different in chemical structure, after they went through the thermal cracking they became more similar in structure. This similarity was consistent with the yields of residual asphaltenes after 60 min in Figure 1.

1052

Energy & Fuels, Vol. 17, No. 4, 2003

Rahmani et al.

Table 5. NMR Data for Concentration of Carbon Types carbon types total aromatic (fa) Quaternary (Qar) polyaromatic Qar alkyl-substituted Qar aromatic CH paraffin CH paraffin CH2, > 5 C methyl of aromatic ethyl terminal paraffinic methyl cycloparaffinic CH2 CH2 R and β to chain ends, attachment sites aromatic CH3 cylcoparaffinic CH3

Arab Heavy (% Ctot)

Arab Light (% Ctot)

Athabasca (% Ctot)

Gudao (% Ctot)

Maya (% Ctot)

52 26.6 17.1 9.5 24.6 11.4 8.9 1.1 2.3 3.7 15.7

61 34.4 26.2 8.2 26.5 7.1 8.4 1.2 2.1 3.6 12.6

55 29 19.8 9.2 24.6 8.6 8.2 1.2 2.3 4.7 15.5

40 21.4 12.2 9.2 18.4 10.6 16.7 1.2 2.9 5.1 17.1

53 28.2 19.6 8.6 24.6 8.3 8.1 1.5 2.8 3.8 16.1

2.3 2.9

2.1 3.4

2.4 3.5

2.2 3.8

2.2 3.6

Table 6. Estimated Kinetic Parameters for Different Asphaltenes in 1-Methylnaphthalenea,b feed asphaltene

cracking rate constant, kA (min-1)

yield coefficient, c

Arab Heavy Arab Light Athabasca Gudao Maya

0.17 ( 0.04 0.11 ( 0.01 0.15 ( 0.03 0.10 ( 0.03 0.20 ( 0.09

0.52 ( 0.05 0.63 ( 0.03 0.54 ( 0.03 0.37 ( 0.02 0.60 ( 0.05

a Error estimates are for 95% confidence intervals. b The other kinetic constants in eqs 1 and 3-5 were SL ) 0.0295 g/g and kC ) 0.17 min-1.

Figure 2. Coke yields from the thermal cracking of asphaltenes in 1-methylnaphthalene. Curves are drawn from the model eqs 1 and 3-5 using the parameters of Table 6.

The data of Figure 2 show the coke yield from the same sets of experiments. Gudao had the lowest coke yield among all asphaltenes. The other four samples clustered together. The data showed that Arab Light reacted more slowly, but ultimately produced the highest amount of coke. Data from the reactions in 1-methylnaphthalene were used in the modified phase separation kinetic model.8 The solubility limit, SL, and firstorder rate constant for coke formation, kC, were taken to be the same for all the asphaltenes as for Athabasca asphaltenes, assuming that the asphaltene cores were not significantly different from each other. The asphaltene cracking rate constants, kA, and the stoichiometric coefficient, c, were estimated by minimizing the sum of squares of the error following the same procedure as described in the work by Rahmani et al.8 The parameter d, giving the split between cores that could and could not accept hydrogen, and the rate constant for hydrogen transfer (k′) were not needed in the absence of tetralin. The data of Figures 1 and 2 show the fit of the experimental data to the model. Estimated parameters, kA and c, are presented in Table 6. The error estimates represent the 95% confidence interval in the parameter estimation. The standard deviation for each estimated parameter was calculated by using the principal diagonal terms in the variance-covariance matrix.19 Arab (19) Himmelblau, D. M. Process Analysis by Statistical Methods; John Wiley & Sons: New York, 1970.

Light had a significantly lower cracking rate constant compared to Athabasca and Arab Heavy, while the stoichiometric coefficient, c, for Gudao was significantly lower than the other asphaltenes. Thermal Cracking of Asphaltenes in Tetralin. Product Yield. All the asphaltenes were reacted in the presence of tetralin for 40 min. From the study with Athabasca asphaltene8 it was found that coke formation was almost complete after 40 min; therefore, this reaction time was used to minimize the error involved due to the initial heating time versus overcracking of the products at long reaction times. A concentration of 25 wt % initial asphaltene was used for all of the reactions. Coke and product asphaltene yields were measured for each experiment. The data of Figure 3 show that the coke yield was suppressed by the addition of tetralin for all asphaltenes, in comparison to reaction in a solvent of equivalent solubility parameter (1-methylnaphthalene). Asphaltene from Gudao vacuum residue produced the lowest amount of coke, whereas coke yields from all other asphaltenes were not significantly different from each other. The residual asphaltene yield varied significantly between the different feeds, which suggested that hydrogen transfer to give asphaltene conversion to heptane-solubles did not occur to the same extent for all asphaltenes. Data from the reactions in tetralin can be interpreted by the modified phase separation kinetic model.8 For each asphaltene, the rate constants, kA, and stoichiometric coefficients, c, from reaction in 1-methylnaphthalene (Table 6) were assumed to apply to reaction in tetralin. The stoichiometric coefficient, d, was estimated from the difference in heptane-solubles plus volatiles (gas and light naphtha) formed when each asphaltene was reacted in tetralin and 1-methylnaphthalene.8 The rate constant for hydrogen transfer, k′, was assumed

Coking Kinetics of Asphaltenes and Chemical Structure

Energy & Fuels, Vol. 17, No. 4, 2003 1053

Figure 3. Coke and asphaltene yield from the thermal cracking of asphaltenes in tetralin for 40 min. A ) asphaltene. Model estimates are from eqs 1-5 using the parameters of Table 7. Table 7. Estimated Kinetic Parameters for Asphaltenes Reacting in Tetralina,b feed asphaltene

coking rate constant, kC (min-1)

fraction of hydrogenaccepting cores, d (g/g)

Arab Heavy Arab Light Athabasca Gudao Maya

0.0089 0.0125 0.028 ( 0.008 0.0108 0.0167

0.064 0.018 0.16 0.077 0.060

a

Error estimates are for 95% confidence intervals. b The value of k′ from Athabasca bitumen8 was 0.23 ( 0.09 wt fraction tetralin-1 min-1; the other kinetic constants in eqs 1-5 were SL ) 0.0295 g/g and values of kA and c from Table 6.

to be the same as Athabasca bitumen (k′ ) 0.23 ( 0.09 wt fraction tetralin-1 min-1). The product yields in tetralin were then represented in the kinetic model by adjusting the rate constant, kC. Estimated rate constants, kC are presented in Table 7. Data were too limited to estimate error bounds except for Athabasca. From the results presented in Table 7, the values of kC and d for Athabasca were significantly higher than those for the other asphaltenes. Analysis of the Heptane-Soluble Product. The heptane-soluble portions of the products were analyzed by GC for each experiment. Naphthalene was identified as the major reaction product from tetralin by comparing the retention time with pure naphthalene and by confirmation with GC-MS. A peak for 1-methyl indan was identified as the other major product by GC-MS. Trace quantities of butyl benzene, 1-methyl tetralin, and 1- and 2-methylnaphthalene were also observed. The major products of tetralin found in this study are the same as previous studies.20-22 Total recovery of (20) Curran, G. P.; Struck, R. T.; Gor, E. Ind. Eng. Chem. Process. Des. Dev. 1967, 6, 166-173. (21) Bockrath, B. C.; Schroeder, K. T. In New approaches in coal chemistry; Blaustein, B. D., Bockrath, B. C., Friedman, S., Eds.; ACS Symp. Ser. No. 196, American Chemical Society, Washington, DC, 1981; pp 191-200. (22) Collin, P. J.; Gilbert, T. D.; Rottendorf, H.; Wilson, M. A. Fuel 1985, 64, 1280-1285.

Table 8. Hydrogen Transfer to Asphaltenes from Tetralin feed asphaltene

H transfer (mg-H/g-A)

Athabasca Arab heavy Arab Light Gudao Maya

13.27 8.68 7.72 7.73 8.73

tetralin, naphthalene, and 1-methylindan was approximately 85 wt % of the original tetralin fed in the reactor, and these data were used to calculate hydrogen transfer to asphaltenes. This approach assumes that the missing tetralin underwent addition reactions without donating hydrogen, as observed by Khorasheh and Gray.23 The data of Table 8 give the amount of hydrogen transfer to each asphaltene after 40 min reaction in tetralin. Hydrogen transfer to Athabasca asphaltene was significantly higher than to any other asphaltene. This observation seemed somewhat contradictory in that Athabasca asphaltene had the highest concentration of cycloparaffinic groups, which would be expected to donate hydrogen themselves. Relationships between Chemical Structure and the Kinetic Parameters. The aim of this series of experiments was to predict the behavior of asphaltenes (coke formation and asphaltene conversion) on the basis of their chemical structure. Coke formation in two different solvents was correlated by the kinetic model; therefore, the model can be used for prediction of coke formation at these reaction conditions if the parameters in the model can be estimated using some chemical structural property of the asphaltenes. Although general predictions for different feeds are difficult, it might be possible to find statistical correlations among the properties of asphaltenes and the estimated parameters in the model. Estimated Rate Constant, kA. There are few average structural properties in the feedstock that will have a (23) Khorasheh, F.; Gray, M. R. Energy Fuels 1993, 7, 960-967.

1054

Energy & Fuels, Vol. 17, No. 4, 2003

Rahmani et al.

Figure 4. Correlation between asphaltene cracking rate constant, kA, and sulfide concentration in feed.

Figure 5. Correlation between stoichiometric coefficient, c, and aromatic carbon fraction of the feed, fa.

strong influence on the cracking rate of asphaltenes. The most important is the sulfide component of the feed. The facile rupture of the sulfide bonds has been postulated as a major mechanism for cracking of the high-molecular-weight components of bitumen.24 The presence of sulfide as a free-radical initiator was found to increase the rate of reaction.8 Similar observations are expected for the cracking rate of asphaltenes. To check this hypothesis, the amount of sulfide in each asphaltene was determined by selectively oxidizing the sulfide component in the asphaltenes and then measuring the sulfoxide concentration by infrared spectroscopy (Table 4). The amount of sulfide in the various asphaltenes was linearly correlated with the cracking rate constant of asphaltenes (Figure 4). The correlation was statistically significant (to 99% confidence). The weighted-leastsquares method was used to do the regression and the slope was found to be significantly different from zero (to greater than 95% confidence). Incorporation of the influence of other structural properties of the feed may improve the correlation between kinetics and chemical structure. The rate of thermal cracking reactions is expected to increase with molar mass of asphaltenes.25 However, Gray et al.4 found a monotonic decrease in the thermal cracking rate with increasing molar mass when they hydroconverted different Alberta residues to determine the relationship between chemical structure and reactivity. According to these authors, when the residue molecules decompose into large fragments, a heavier feed molecule would require more scission reactions to become distillate. If the decomposition were dominated by the loss of side chains and groups, leaving a more aromatic residual core, then the residue conversion should be independent of molar mass. In this study, no relationship was found between the feed molecular weight and the reactivity of the asphaltenes, supporting the observation of Gray et al.4 The C-N bonds present in asphaltene molecules are thermally stable; however, the nitrogen can activate the C-C bonds adjacent to nitrogen-containing heterocyclic rings.26 Therefore, it seems possible that an increase in

the nitrogen content will enhance the cracking conversion of the asphaltene. Long chain paraffins can also have an impact on the rate constant, since they are more reactive compared to the other constituents present.24 However, only the amount of sulfide had a significant correlation with the cracking rate constant. None of the other structural properties had a statistically significant correlation with the value of kA, even at the 80% confidence level. These observations suggest that although the cracking rate was likely due to the combined effect of several different structural properties of the feed, the sulfide content was the dominant factor. Stoichiometric Coefficient, c. The stoichiometric ratio, c, determines what fraction of the asphaltenes will form aromatic cores and then react to form coke in the absence of significant hydrogen transfer. Although not all the aromatic carbons in each feed are necessarily in polynuclear aromatic clusters, which would be most likely to give coke, the correlation between stoichiometric coefficient, c, and the aromaticity of the feed, fa, was statistically significant in the experimental data from the coking reactions with 1-methylnaphthalene (Figure 5). The variation in the total aromatic carbon between the asphaltenes was due mainly to the polyaromatic quaternary aromatic carbon, with some contribution from changes in aromatic CH (Table 5). Increasing yields of coke with polyaromatic quaternary carbon content is consistent with the known chemistry of coke formation.1,9,27,28 Rate Constants in Tetralin, kc and d. Analysis of reactions of different feeds in the presence of tetralin suggested that the parameters kC and d varied significantly between the asphaltenes (Table 7). Parameter d indicated the fraction of cores that could accept hydrogen and convert to heptane-solubles, and was calculated from the difference in total yield of asphaltenes and coke between 1-methylnaphthalene and tetralin. This parameter correlated significantly with the hydrogen transferred to the asphaltenes (Table 8) with a r2 ) 0.84. The values of d did not correlate with elemental compositions (Table 4) or NMR data (Table 5). Similarly,

(24) Gray, M. R. Upgrading Petroleum Residues and Heavy Oils; Marcel Dekker: New York, 1994; Chapter 3. (25) Asaoka, S.; Nakata, S.; Shiroto, Y.; Takeuchi, C. Ind. Eng. Chem. Process Des. Dev. 1983, 22, 242-248.

(26) Speight, J. G., Prepr. Pap.sAm. Chem. Soc., Div. Pet. Chem. 1987, 32 (2), 413-418. (27) Wiehe, I. A. Energy Fuels 1994, 8, 536-544. (28) Gray, M. R.; McCaffrey, W. C. Energy Fuels 2002, 16, 756766.

Coking Kinetics of Asphaltenes and Chemical Structure

Energy & Fuels, Vol. 17, No. 4, 2003 1055 Table 9. Prediction of Kinetic Parameters for Iranian Light and Khafji Asphaltenes in 1-Methylnaphthalenea experimental

predicted

feed

S, wt %

fa

kA

c

Iranian Light Khafji

5.9 7.6

0.61 0.56

0.129 0.157

0.649 0.587

a The other kinetic constants in eqs 1 and 3-5 were S ) 0.0295 L g/g and kC ) 0.17 min-1.

Figure 6. Change in yield of coke, between reactions in tetralin and 1-methylnaphthalene at 430 °C for 40 min, and content of cycloparaffinic CH2 in feed asphaltenes (as % of total C).

hydrogen transfer (Table 8) did not correlate with the chemical compositional data. This lack of correlation between hydrogen transfer and chemical structural data can be attributed to two factors. First, hydrogen transfer to asphaltenes during cracking will likely be due to two major reactions; hydrogen transfer to aromatic groups and hydrogen transfer to olefins formed by cracking of C-S and C-C bonds.28 When more than one reaction contributes to hydrogen transfer, we would not expect simple correlation with individual chemical properties. The second reason is that the ability of aromatics to accept hydrogen depends intimately on the ring structure and substitution.29 Such characteristics may not be reflected well by average concentrations from NMR. The parameter d, calculated from the change in total coke and asphaltene content, did not correlate with the chemical structural data. The change in coke yield alone between the two solvents was, however, related to average structural information from NMR. As illustrated in Figure 6, the change in the yield of coke between 1-methylnaphthalene and tetralin was significantly correlated with the concentration of cycloparaffinic CH2 in the feed asphaltenes. This observation was consistent with hydrogen transfer from the cycloparaffins in the asphaltenes to suppress formation of tolueneinsoluble material. Gudao, with the highest concentration of cycloparaffin CH2, gave the smallest change in coke yield when a donor solvent was added, relative to Arab Light which had the lowest concentration of this type of carbon (Table 5). Apart from the stoichiometric yields of coke and asphaltenes, hydrogen transfer will also affect the rate of coke formation from phase-separated asphaltenes and the rate of asphaltene conversion to heptane solubles. These processes are represented in the kinetic model by eqs 2 and 5, with rate constants kc and k′, respectively. Either kinetic constant could be influenced by the chemical structure of the asphaltenes, but due to limited data in this study the value of k′ was held constant while kc was allowed to vary between feeds (Table 7). The lack of any significant correlation between kc and the chemical structural data of Tables 4 and 5 (29) Wang, S.-L.; Curtis, C. W. Energy Fuels 1994, 8, 446-454.

Figure 7. Coke and asphaltene yield from the thermal cracking of Iranian Light asphaltene in 1-methylnaphthalene. The curves are from the model eqs 1 and 3-5 with parameters from Table 9.

could be due to this method of estimating the kinetic parameters. Similar to the parameter d, discussed above, the value of kc may also depend on more than one chemical reaction or on the specifics of aromatic structure and substitution. Further study is required to better understand the relationship between hydrogen transfer reactions and chemical structure in conversion of asphaltenes and residues. Prediction of Coke Formation in 1-Methylnaphthalene. Iranian Light and Khafji asphaltenes, which were not used in estimating kinetic parameters in this study, were used to study the predictive ability of the model. The kinetic constants for cracking of the other five asphaltenes in 1-methylnaphthalene were linearly correlated with the chemical structure of the asphaltenes (Figures 4 and 5). The rate constant for cracking of asphaltenes was correlated as follows:

kA ) 0.0377Sd + 0.0336

(6)

where Sd is the weight % of sulfide sulfur in the asphaltenes, as determined by oxidation analysis. The yield coefficient for coke formation was correlated as follows:

c ) 1.2434fa - 0.1091

(7)

where fa was the fraction of aromatic carbon in the asphaltenes as determined by 13C NMR. These relationships should allow the prediction of cracking of asphaltenes on the basis of simple chemical analysis. The average sulfide fraction of the sulfur in the other five asphaltenes was 0.47 ( 0.09. The sulfide contents of Iranian Light and Khafji asphaltenes were not determined experimentally, but the fraction of sulfide sulfur

1056

Energy & Fuels, Vol. 17, No. 4, 2003

Rahmani et al.

concentration and aromaticity, can give predictions of kinetics under controlled liquid-phase conditions. Conclusions The modified-phase-separation kinetic model8,9 was consistent with the data on coke formation from asphaltenes with different structural properties in the solvents 1-methylnaphthalene and tetralin. The two fitting parameters, asphaltene cracking rate constant (kA) and stoichiometric coefficient (c), were found to correlate with two chemical properties: sulfide content and aromaticity of the asphaltenes, respectively. The kinetic model was shown to predict coke formation in the solvent 1-methylnaphthalene, based on the chemical composition of Iranian Light and Khafji asphaltenes. Figure 8. Coke and asphaltene yield from the thermal cracking of Khafji asphaltene in 1-methylnaphthalene. The curves are from the model eqs 1 and 3-5 with parameters from Table 9.

of 0.43, within this range, was used to estimate values for kA to obtain good agreement between the model predictions and the data (Table 9 and Figures 7 and 8). The results from coking in 1-methylnaphthalene were very encouraging because they showed that simple models for asphaltene chemistry, based on sulfide

Acknowledgment. The research was supported by a Canadian International Development Agency scholarship (S. Rahmani), the Syncrude/NSERC Industrial Research Chair for Advanced Upgrading of Bitumen, and by a collaborative research and development project sponsored by the New Energy and Industrial Technology Development Organization (Japan). Discussions with R. Tanaka, Idemitsu Kosan Co., Ltd, S. Sato and T. Takanohashi, AIST Japan, and K. Chung, Syncrude Canada, and were most helpful. EF030007C