Understanding the Mechanisms of Decalin ... - ACS Publications

Article pubs.acs.org/IECR

Understanding the Mechanisms of Decalin Hydroprocessing Using Comprehensive Two-Dimensional Chromatography Elodie Blanco,† Luca Di Felice,†,§ Nelly Catherin,† Laurent Piccolo,*,† Dorothée Laurenti,† Chantal Lorentz,† Christophe Geantet,† and Vincenzo Calemma‡ †

Institut de Recherches sur la Catalyse et l’Environnement de Lyon (IRCELYON), UMR5256 CNRS-Université Claude Bernard Lyon 1, 2 Avenue Albert Einstein, 69626 Villeurbanne Cedex, France ‡ Eni S.p.A., R&M Division, Via F. Maritano 26, 20097 San Donato Milanese, Italy S Supporting Information *

ABSTRACT: Comprehensive two-dimensional gas chromatography (GC × GC) is a powerful technique for analyzing mixtures of hundreds of hydrocarbons. In the context of fuel upgrading through selective ring opening, we propose a methodology for GC × GC analysis of complex mixtures resulting from the hydroprocessing of a single model gas oil compound, decalin, over two different types of bifunctional catalysts based on a transition-metal sulfide (NiWS on amorphous silica−alumina) or a noble metal (Ir on La,Na− Y zeolite). The reactions lead to several products families, the dominant ones being ring-opening products (ROPs) and skeletal-isomerization products (SkIPs). First, it is shown that the ROP distribution can be characterized in terms of isomerization degree by using the cumulative distribution function of the GC × GC (GC Image) software. Second, in a more quantitative approach, the products families have been subdivided into chemical groups, according to the isomerization degrees of the individual compounds, which were almost all tentatively identified by two-dimensional gas chromatography coupled with mass spectrometry (GC × GC-MS) through the use of literature data. This allows us to thoroughly analyze the influence of the catalyst nature and the presence of H2S in the reactant feed on the products distribution, and thereby gain insight into the mechanism of decalin hydroconversion over bifunctional catalysts. In particular, it is shown that metal sulfidation suppresses the metal-catalyzed C−C hydrogenolysis pathway at the benefit of undesirable acid-catalyzed isomerization steps. The methodologic work presented here for decalin is believed to be applicable to other bicyclic (naphthenic or aromatic) compounds.

T

reaction on bifunctional catalysts, because of the variety of reaction steps occurring on metal and/or acid sites. In this context, full separation and identification of the molecules are difficult tasks. One-dimensional gas chromatography (GC) using a 150 m capillary column, coupled with mass spectrometry (MS), permitted one to tentatively identify ∼200 compounds and classify the products into several families.23 However, possible coelution, along with incomplete MS databases, lead to identification uncertainties, making the use of comprehensive two-dimensional gas chromatography (GC × GC) indispensable.14,27,28 In a recent work,12 we have investigated the SRO of decalin under high pressure of hydrogen and in the presence of H2S (a classical contaminant of petroleum cuts) over NiWS catalysts supported on amorphous silica−alumina (ASA). It was found that 1-ring-opening products (1ROP) and skeletal-isomer-

he foreseen shortage in fossil resources, together with the increasing demand in diesel fuel, lead the refiners to valorize nonconventional petroleum cuts, such as the light cycle oil (LCO), which is produced from fluidized catalytic cracking (FCC). The LCO contains large amounts of polyaromatic hydrocarbons, and thus presents a low cetane number (CN), i.e., poor combustion efficiency. The saturation of aromatic compounds into naphthenic ones through catalytic hydrotreatment increases the CN, but selective ring opening (SRO) of naphthenes into paraffins is necessary to reach acceptable CN values.1,2 Research on SRO has long been carried out on model hydrocarbons consisting of single-ring naphthenic compounds, such as methylcyclopentane,3,4 cyclohexane,5−7 and methylcyclohexane.8−11 The advantage of using these molecules is the small number of products, which facilitates the interpretation of reaction mechanisms. However, these compounds are poor representatives of the LCO composition. To account for this limitation, 2-ring probe molecules such as decalin and tetralin have been employed.8,12−26 However, even by starting from a single molecule (or two stereoisomers in the case of decalin), a complex mixture of ca. 200 hydrocarbons is formed upon © 2016 American Chemical Society

Received: Revised: Accepted: Published: 12516

September 7, 2016 November 5, 2016 November 14, 2016 November 14, 2016 DOI: 10.1021/acs.iecr.6b03472 Ind. Eng. Chem. Res. 2016, 55, 12516−12523

Article

Industrial & Engineering Chemistry Research

of the reactor with a cryostat kept at 0 °C. This analytical system has been described in detail in ref 30. A 30-m-long Phenomenex ZB1MS column (inner diameter (ID) = 0.250 mm, film thickness = 1 μm) was connected in series with a 3-mlong Varian VF17MS column (ID = 0.100 mm, film thickness = 0.2 μm) by means of a cryogenic two-stage thermal modulator (Zoex Corporation, modulation period = 11.63 s, N2 as cryogenic fluid, 300 ms hot air jet). The temperature of the first GC column was programmed from 60 °C (1 min hold) to 150 °C (0 min hold) at a rate of 0.6 °C/min, and then to 300 °C at a rate of 5 °C/min (0 min hold). For the second column, the temperature was programmed from 60 °C (1 min hold) to 300 °C at a rate of 5 °C/min (0 min hold). Then, 0.8 μL of pure sample was injected with a split ratio of 250. The carrier gas was high-purity helium flowed at a constant rate of 2 mL/min. Data treatment and matching of GC × GC-MS and GC × GC-FID were performed with GC Image 2.3 software (GC × GC Edition).31 Computer Language for Identifying Chemicals (CLIC) expressions were used for identification of compound families,32 and the NIST MS database (version 2.0) was used for the identification of 25 C10 molecules. The database was complemented with 148 mass spectra reported in the literature (see discussion below).33−45

ization products (SkIP) are the most representative compound families, accounting for 75%−80% of the entire product distribution. In this paper, we report on a comparative GC × GC investigation of the products formed on NiWS/ASA and a reference Ir/zeolite catalyst (with or without H2S in the reactant feed) with the objective of gaining insight into the catalyst and conditions-dependent decalin hydroconversion pathways. The choice of catalysts was undertaken to extend the peak attribution to a large amount of products, which can possibly be obtained from the hydroconversion of decalin and, beyond, two-ring hydrocarbons over a broad range of catalysts exhibiting various types and strengths of metallic and acidic functions. In particular, using two-dimensional gas chromatography coupled with mass spectrometry (GC × GC-MS) or flame ionization detection (GC × GC-FID), we focus on the identification and quantification of the individual molecules and related chemical groups inside each family of compounds.

■

EXPERIMENTAL SECTION Materials. A NiWS/F-ASA catalyst was prepared as detailed in a previous work.12 The ASA powder (Sasol SIRAL40) was modified by incipient wetness impregnation (IWI) of ammonium fluorine (NH4F), then dried overnight and calcined under air at 500 °C. The modified support was then impregnated (IWI) with Ni and W salts, and finally sulfided by exposure to a 10% H2S/H2 gas mixture at 400 °C. A 3 wt % Ir/LaNaY catalyst was prepared by IWI of a La-modified Naexchanged Y zeolite (Zeolyst CVB 760 PQ, SiO2:Al2O3 molar ratio = 56) with an aqueous solution of H2IrCl6:HCl:CH3COOH, followed by drying at 80 °C and calcination at 360 °C.29 The La:Al and Na:Al molar ratios were 0.24 and 0.26, respectively, as determined by elemental analysis. The Ir dispersion (number of surface Ir atoms divided by total number of Ir atoms) was 69%, as determined by H 2 chemisorption, corresponding to particles ca. 1.5 nm in size. Catalytic Testing. The catalyst performances for decalin hydroconversion were evaluated in a flow fixed-bed reactor at 5.0 MPa of H2, with a gas-phase decalin partial pressure of 16 kPa, a reaction temperature ranging from 270 to 380 °C and the possibility to feed H2S at relatively high concentration (0.8 vol % was employed in this work). The decalin cis−trans mixture (40−60%) was purchased from Sigma-Aldrich (reagent grade, 98%). The total weight hourly space velocity (mass of decalin fed/mass of catalyst) was 0.8 h−1 for standard tests. The gas mixture was analyzed online using a HP 5890 Series II gas chromatograph containing a HP-1 column with a CP-Sil 5 CB stationary phase (25 m × 0.2 mm × 0.50 μm, N2 as the carrier gas). The sampling loop volume was 250 μL. Technical details of the setup and the related GC analysis were provided elsewhere.12 Decalin conversion was calculated as follows: Xdec (%) =

Dec(bypass) − Dec(out) Dec(bypass)

■

RESULTS AND DISCUSSION Methodology for Analyzing GC × GC Chromatograms of Decalin Derivatives. In a previous work, five main product families were distinguished from the GC × GC-MS chromatograms of decalin hydroconversion products. These are listed below and are displayed in the chromatogram shown in Figure 1:12

Figure 1. Chemical families identified on GC × GC-MS chromatograms. Catalyst and specific conditions: NiWS/F-ASA, 0.8% H2S, 380 °C, Xdec = 80%. Retention time ranges: 50−120 min (first dimension) and 1.8−11.3 s (second dimension).

• 1ROPs: 1-ring-opening products, i.e., C10 alkyl-mononaphthenes; • AROPs: aromatic 1-ring-opening products, i.e., C10 alkyl-benzenes; • 2ROPs: 2-ring-opening products, i.e., C10 paraffins; • SkIPs: skeletal isomerization products, i.e., C10 alkyldinaphthenes; and • DHPs: dehydrogenation products, i.e., all C10 unsaturated products, except AROPs (tetralin, naphthalene, methyl-indans, etc.). Since, in the presence of H2S in the reactant feed and for any catalyst, SkIPs and 1ROPs together represent 60%−80% of the

× 100 (1)

where Dec(bypass) and Dec(out) are the sums of cis and trans decalin peak areas when decalin is bypassed off the reactor and at the reactor outlet, respectively. GC × GC Analysis. A two-dimensional GC × GC Agilent 6890N chromatograph equipped with either an Agilent 5975B mass detector (single quadrupole, mass/charge range: 50−300 amu, up to 22 scans/s) or a flame ionization detector (FID) was used to analyze the condensed phase collected at the outlet 12517

DOI: 10.1021/acs.iecr.6b03472 Ind. Eng. Chem. Res. 2016, 55, 12516−12523

Article


Figure 2. CLIC-filtered images of the 1ROP family. Catalyst and conditions: same as Figure 1. Corresponding chemical groups: (a) methyl-1ROP, (b) ethyl-1ROP, (c) propyl-1ROP, (d) butyl- and pentyl-1 ROP. See text for details.

chromatograms, using the GC Image software. The associated cumulative function is calculated as follows:

products, a methodology for identifying these products is desirable in order to appreciate the differences in the reaction mechanisms. Following the approach of Kubička et al.,46 a first tentative identification of chemical groups inside the 1ROP family has been carried out based on the dominant MS m/z fragment, which results from the corresponding alkyl-group loss (the molecular weight of the starting C10 molecule is 140): methyl1ROP (m/z = 125), ethyl-1ROP (m/z = 111), propyl-1ROP (m/z = 97), and butyl-1ROP (m/z = 83). Figure 2 shows 1ROP peaks based on specific criteria, which are taken into account in GC Image software’s CLIC expressions, along with typical molecule structures of each group. Figure 2a shows that the methyl-1ROP group (criterion: peak at m/z = 125 has >20% of the intensity of the main peak) has the tendency to elute first, i.e., is located at the left-hand side of the 1ROP region. The ethyl-1ROP (m/z = 111 (>50%), Figure 2b) and propyl-1ROP (m/z = 97 (>40%), Figure 2c) appear more scattered, and consistently contain more compounds, than the methyl-1ROP group. Finally, Figure 2d shows the result of the search for m/z = 83 (>50%), and corresponds to the last eluted compounds, i.e., butyl- and pentyl-1ROP groups. On the basis of normal configuration of the two GC columns (apolar × polar), it is expected that molecules with a higher degree of branching elute faster.28 For example, among a same ethyl-1ROP group, trimethylethylcyclopentanes elute faster than dimethylethylcyclohexanes, which themselves elute faster than diethylcyclohexanes. In addition, with the same number of substituents, the compounds with the biggest alkyl substituent elute faster, e.g., methyl-iso-propylcyclohexanes elute faster than diethylcyclohexanes. This approach evidences the general trend that more isomerized 1ROPs (with low cetane number) elute faster than linear ones (with higher cetane number), and this feature can be used to quickly characterize the quality of the obtained 1ROP mixture. The GC × GC-FID technique was used for calculating the so-called “cumulative distribution function” of 1ROP compounds, C1ROP. A uniform grid consisting of n = 100 meshes along the horizontal axis (see Figure S1 in the Supporting Information) was applied to the 1ROP region in the GC × GC

i=n

C1ROP(n) =

∑i = 1 V1ROP(i) Vtot

(2)

where V1ROP(i) and Vtot are the detected 1ROP peak volume at mesh i and the total 1ROP peak volume, respectively. Under the assumption that all C10 compounds have the same response in FID, it was assumed that the distribution is quantitative. Note that the chromatograms obtained from MS and FID for a same sample were very similar; only small differences in the relative intensities of the peaks were observed. Figure 3 shows cumulative 1ROP distributions for decalin hydroconversion over two types of catalysts (Ir/LaNaY vs

Figure 3. Cumulative distribution of 1ROPs at similar products yields (YROP = 12%) for reaction over Ir/LaNaY without (270 °C) or with (310 °C) H2S, and NiWS/F-ASA with H2S (380 °C).

NiWS/F-ASA) under two different conditions (with or without H2S). As previously mentioned, the more branched the product, the smaller the mesh number should be. Hence, in order to reach a high cetane number (linear products, with low isomerization degree), the distribution should be shifted toward higher mesh numbers (i.e., to the right-hand side of the graph), 12518

DOI: 10.1021/acs.iecr.6b03472 Ind. Eng. Chem. Res. 2016, 55, 12516−12523

Article


• dimethylbicyclo[2.2.2]octanes, 14 isomers (Figure 4VI)34 The following products, present in the NIST MS database, were also found in our gas mixtures: • spiro[5.6]decane (Figure 4-VII) • bicyclo[5.3.0]decane (Figure 4-VIII) • 1,1-bicyclopentyl (Figure 4-IX) A group of “unknown products” (UkP) was also created, due to uncertainty in MS identification. In all cases, the selectivity to UkPs was found to be