Most petroleum upstream applications for GC–MS are for analyzing whole oils, while most downstream applications are for analyzing distillation fractions, re- fining streams, and products in the marketplace. In the downstream applications, knowledge of the composition of petroleum and its fractions allows the refiner to optimize conversion of raw petroleum into high-value products. It is important to understand the chemical transformations that occur in a refinery, and the termi- nology of petroleum products.
Initially, petroleum is distilled into fractions up to 350°C (650°F) under atmospheric pressure into gas, naphtha, middle distillates, gas oils, and residua.
The atmospheric residua can be further distilled up to 350°C under vacuum to produce vacuum gas oils, lubricant oils, and vacuum residua. A modern refinery is no longer just a big distillation column that sells various boiling fractions to different consumer markets. While distillation is usually the first step, it is impor- tant to understand that a refinery processes 100,000 to 500,000 barrels of oil a day, and must turn every barrel of that oil into something that can be sold econom- ically in the marketplace. For this reason, many of the crude oil fractions (‘‘streams’’) undergo catalytic transformation into more valuable streams and then are blended to produce petroleum products or petrochemicals that provide the highest value to the final end use.
Modern refineries use a sophisticated combination of heat, catalyst, and hydrogen. The gas oils and residua from vacuum distillation are further processed by cracking, either thermally or catalytically (such as fluid catalytic cracking, or FCC), and coking to break large molecules into lighter compounds, and produce highly olefinic streams [12]. This is in sharp contrast to crude oil, which has almost no olefinic content. Hydrotreating catalysts reduce olefins and aromatics and remove heteroatom-containing compounds to produce environmentally ac- ceptable products and to avoid poisoning catalysts downstream. Isomerization and reforming catalysts rearrange molecules to those having higher value, e.g., gasolines of high octane number and high energy content. These refinery streams
Petroleum Industry Applications 69
are blended into products, such as gasoline, diesel and jet fuels, lubricant oils, etc. Additives may be added into these products to improve their stability and performance. All together, the catalysts that turn less valuable streams in refiner- ies into more valuable streams in refineries are a $2.1 billion/year business world- wide [49].
For gaseous mixtures, GC alone can sometimes provide sufficient informa- tion on composition. Gas chromatography–mass spectrometry is therefore nor- mally applied to semi-volatile liquid streams and products: in the range of naphtha to vacuum oils in refinery streams and gasolines to lubricant oils in the petroleum products. For heavier and higher-boiling fractions than vacuum gas oils, pyrolysis GC–MS can be used for analysis. Most refinery streams and products are largely free of heteroatoms or metals, to usually the tens of ppm total level, which means individual heteroatom-containing molecules are generally present below GC–MS detection limits. To analyze these molecules, enrichment through sophisticated chromatographic procedures is usually needed.
4.1. Fuel Analysis
The simplest liquid mixtures are naphtha or gasolines. Even for these fractions, accurate analysis by GC alone can be difficult due to coelution of components even when a very high-resolution GC method is used. Gasoline, being one of the most important high-value products and the most thoroughly analyzed, is mostly a synthetic product with very strict compositional parameters. Gas chro- matography–mass spectrometry has been used to determine compound type, in- cluding paraffins, isoparaffins, olefins, naphthenes, and aromatics (PIONA), and carbon number distributions of a gasoline sample [50]. Gas chromatography–
mass spectrometry mapping is used to resolve coeluting compounds with differ- ent mass spectral patterns. The coeluting compounds are then identified by the
‘‘difference spectra.’’ Gas chromatography–mass spectrometry has also been used to determine the efficiency of a GC column for separating naphtha compo- nents [51].
The largest current application of GC–MS to refinery streams and products is due to recent government regulations. For example, the U.S. Environmental Protection Agency (EPA) specified in theFederal Registerin 1994 that GC–MS must be used to determine the total aromatics in gasoline [52]. The method was later implemented as American Society for Testing and Materials (ASTM) method D5769, which uses deuterated surrogates and 23-component QC mix with accuracy of each component in the QC mix to be within 5% relative to the true, weighted concentration [53]. As part of the new rules for reformulated gaso- line (RFG), the U.S. government has mandated strict limits on various composi- tional parameters of gasoline to be used in parts of the United States with air quality problems. These parameters are to be determined by specified methods,
70 Hsu and Drinkwater
and the method specified for total aromatics is GC–MS. There are 50 or more significant aromatic species in gasoline, and one of the best ways to quantify all of them individually is by GC–MS. The prior method for total aromatics was by open column chromatography with yardstick measurement of colored bands, which suffered from serious precision problems and did not provide quantitation for individual species. Figure 6 shows a portion of the D5769 chromatogram of a gasoline sample between 6 and 6.5 minutes. The distribution of alkanes is dis- played by the 57-Da ion current trace, while the distribution of C-3 benzenes is displayed by the 120-Da ion current trace. The selectivity of aromatic compounds
Figure 6 Partial selected ion chromatogram of ASTM D5769 data (see text for refer- ence) for a reformulated gasoline. Solid line is m/z 120 ion current for various C-3 alkyl- benzenes. Dotted line is m/z 57 ion current, base peak for saturate gasoline components (unidentified), showing the selectivity provided by GC–MS for badly coeluting gasoline components. Cycloparafins and olefins are also present but not displayed. Conditions: gas chromatograph: Varian 3400 (Varian, Palo Alto, CA) running 20 m⫻0.10 mm ID DB- 1 phase (0.01-àm film thickness) at 1000 : 1 split of 0.1 ul injection of gasoline dosed with mixture of isotopically labeled aromatic standards as per method. Mass spectrometer:
Finnigan (Finnigan, San Jose, CA) TSQ710 in 70-eV EI, full-scan mode, 20 scans/sec.
Petroleum Industry Applications 71
using fast chromatography is demonstrated in this example in which the analysis of whole sample is completed in 9 minutes. Every batch of gasoline sold as RFG (approximately 40% of the U.S. gasoline market) must be analyzed by this method. Briefly, the sample is first dosed by weight with a mixture of deuterated aromatic compounds as internal standards (IS) for quantitation. The compounds of interest are quantified separately versus a relevant deuterated IS component.
This method has the added advantage of being able to quantify individual target species such as benzene and methyltert-butyl ether (MTBE), among others [53].
Retention indices are very useful for compound identification by GC–MS, especially for isomers that yield similar or identical mass spectra. Lai and Song [54] have determined retention indices of over 150 compounds with temperature programming of moderately polar and slightly polar GC columns for the compari- son of jet fuels JP-8 derived from petroleum and coal. They found that aliphatic compounds give nearly constant retention indices at different heating rates, while aromatic compounds show relatively large temperature dependence. With differ- ent polarity of the GC columns, the difference in retention indices is relatively small with aliphatic compounds but becomes larger with polycyclic aromatic and polar compounds. Petroleum-derived jet fuel saturates are mainlyn-alkanes and isoalkanes. Coal-derived jet fuel saturates are mainly cycloparaffins with 1 to 3 rings, with somen-alkanes. Similarly, petroleum-derived jet fuel aromatics are dominated by alkyl benzenes and alkyl naphthalenes, while coal-derived jet fuel aromatics are rich in hydroaromatics [54].
4.2. Lubricant and Mineral Oil Base Stocks
Vacuum gas oils (VGOs) are typically used for lubricant and mineral oil base stock. These materials vary in carbon number from 12 to 30 or more, and gener- ally present a totally unresolved ‘‘hump’’ for the total ion chromatogram, and almost every nominal mass is present at almost every retention time. In this case, the GC–MS analyst is limited to either target compound analysis of higher con- centration components, or more selective analysis using high-resolution GC–
high-resolution MS or high-resolution GC–MS–MS. Both are discussed below in section 6 on special techniques providing extra selectivity for specific compound classes. Important exceptions to this generalization about lubricant base stocks are: (1) waxy oils that contain primarilyn-alkanes that are well resolved chro- matographically, and (2) synthetic and ‘‘semisynthetic’’ base stocks that are oils of the same boiling range that have been replaced with or doped with poly-alpha- olefin (PAO) or polyether petrochemical streams to increase various lubricant performance properties. Such samples may be analyzed either by conventional GC–MS using high-temperature GC columns or by pyrolysis GC–MS where the sample is flash-heated to between 600 and 1000°C prior to the injection port, and the ensuing olefinic pyrolysis fragments are cryotrapped on the GC column.
72 Hsu and Drinkwater
Figure 7 shows the GC–MS from a semisynthetic oil showing the PAO distribu- tion along with the mineral oil ‘‘hump’’ characteristic of conventional vacuum gas oils [55]. Pyrolysis GC–MS is also commonly applied to finished (additized) motor oils, which may contain polymeric additives with molar mass up to 500,000 Da.
4.3. Process Gas Chromatography–Mass Spectrometry
Traditionally, refineries have made great use of gas chromatography for process stream characterization. It is not unusual to walk through a refinery and see doz- ens of process GC instruments stuck to walls and bulkheads. They generally contain very long capillary columns (100 m⫻0.25 mm DB-1 columns are very common) and take automated quantitative flame ionization detection (FID) mea- surements of various constituents. The data are then fed to the computerized plant control room. The slow analysis time and limitation to only the largest GC peaks causes some amount of efficiency loss. These units are slowly being replaced by process GC–MS instrumentation using shorter, faster columns, as well as process near-infrared (NIR), nuclear magnetic resonance (NMR), and 2-dimensional GC (GC–GC) instrumentation in that order.
Typically, fuels products are blended under computer control. There are constraints against which the blend recipe is constantly checked. The constraints are typically product specifications (cetane, octane, cloud point, distillation points, etc.) and blend stock quality and availability. The quality of the on-going blend is monitored using key quality analyzers that feed back to the blend control computer the current quality status. The blend recipe is then adjusted to optimize the quality (specifications) using the currently available blending components.
A refinery-compatible GC–MS instrument has been developed recently for use in on-line blending and process control applications [55]. For rapid analysis, a 10-m microbore GC column coated with 0.17-àm film thickness of DB-5 (methyl silicone) is used. The column temperature is programmed for gasolines from 35 to 230°C and for diesel fuels from 110 to 295°C, both at 30°C/min. Typical results of this rapid on-line GC–MS analysis for gasolines, jet fuels, and diesel fuels are shown in Figure 8.
Partial least square (PLS) modeling is used to develop property predictions for the streams of interest [56]. The procedure requires the use of a training or reference sample set that has known values for the properties being modeled.
The reference samples are analyzed by GC–MS after which the data generated are treated by multivariate correlation methods. The resulting inferential models are then used with subsequent GC–MS analysis of an unknown mixture to pro- duce a predicted value for the property or groups of properties of interest. In the on-line case, the training set is developed from actual blender samples with on- line GC–MS data coupled with standard laboratory analyses over a long period
Petroleum Industry Applications 73
Figure 7 Gas chromatography–mass spectrometry total ion chromatogram (TIC) of a semisynthetic motor oil base stock. The peaks at lower retention time (RT) are poly-alpha- olefin oligomer clusters separated by three carbons. The wide ‘‘hump’’ at high RT is the mineral oil component, with insufficient chromatographic resolution to distinguish individual species without mass spectral separation. Conditions: Hewlett Packard (Palo Alto, CA) 5890 gas chromatograph with MSD detector in 70-eV EI mode, 30 m⫻0.25 mm ID DB-5MS phase (0.25-àm film thickness) with temperature ramp from 35 to 350°C at 2°C/min.
74 Hsu and Drinkwater
Figure 8 Typical results of on-line GC–MS for gasoline, jet fuel, and diesel fuel. Condi- tions: GC–MS: Hewlett-Packard (Palo Alto, CA) 5971/2A MSD, 10-m microbore GC column coated with DB-5 (0.17-àm film thickness), temperature ramp 35 to 230°C at 30°C/min for gasoline and 110 to 295°C at 30°C/min for jet fuel and diesel fuel, sample size 0.5àl injection with 250 : 1 split, mass scan range 10 to 300 Da for gasoline and 10 to 400 for distillate fuels.
of time to account for refinery variations, process unit shutdowns, and crude slate changes.
One of the advantages of GC–MS over an IR spectroscopic analyzer is the ability to measure distillation characteristics as well as predict other properties.
There are several other materials that can be directly measured and reported.
These include benzene, total aromatics, oxygenates, certain sulfur compounds and additives. The properties that can be predicted include (among others) cetane number and index, research and motor octanes, refractive index, distillation prop- erties, aniline point, cloud point, pour point, volatility, flash point, density, con- ductivity, and viscosity [57].