Compositional Analysis of Complex Organic Mixtures by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry Zhigang Wu

Florida State University Libraries

Electronic Theses, Treatises and Dissertations The Graduate School

2004 Compositional Analysis of Complex Organic Mixtures by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry Zhigang Wu

Follow this and additional works at the FSU Digital Library. For more information, please contact [email protected]

THE FLORIDA STATE UNIVERSITY

COLLEGE OF ARTS AND SCIENCES

Compositional Analysis of Complex Organic Mixtures

by Electrospray Ionization Fourier Transform Ion

Cyclotron Resonance Mass Spectrometry

Zhigang Wu

A Dissertation submitted to the Department of Chemistry and Biochemistry in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Degree Awarded: Spring Semester, 2004

The members of the Committee approve the Dissertation of

Zhigang Wu

Defended on April 7, 2004.

Alan G. Marshall Professor Directing Dissertation

William M. Landing Outside Committee Member

William T. Cooper Committee Member

Timothy A. Cross Committee Member

Ryan P. Rodgers Committee Member

The Office of Graduate Studies has verified and approved the above named committee members.

To my parents, Xiaohua Wu and Shiping Cao;

iii

ACKNOWLEDGMENTS

First, I need to thank Dr. Alan G. Marshall, my academic graduate advisor and supporter. His deep and broad academic knowledge and great personality are invaluable throughout my graduate career. His wise and forecasting advices has helped me directly cut to the “point” and inspired me to discover more underneath. I also wish to acknowledge Dr. Ryan P. Rodgers, my supervisor, for his expertise, direction, and support. His initiation and development in petroleum applications have led me to a broader and better understanding of this area. His immediate and inspiring direction made an easy and fun research for me. Our numerous discussions have strengthened every aspect of my research. I also would like to thank Dr. Christopher L. Hendrickson for his expertise in FT-ICR’s instrumental optimization. He was always at the first place when I needed instrumental help. His helpful direction and discussions have helped me a better understanding and utilization of the FT-ICR instrument. The author would also like to thank John P. Quinn and Daniel McIntosh for their technical and mechanical expertise, respectively. I would like to express my gratitude to the past (from Fall 1999) and current (to end of Spring 2004) members of the ICR Program at Florida State University for their professional, personal, and moral support. In particular: Chrisi Hughey, Kicki Håkansson, Ying Xiong, Melinda McFarland, Michael Chalmers, Jinmei Fu, Greg Blakney, and Mark Emmett. Work supported by NSF (CHE-99-09502), NIH (GM-31683, AI-44626), Florida State University, and the National High Magnetic Field Laboratory in Tallahassee, FL.

TABLE OF CONTENTS

LIST OF TABLES ...... x LIST OF FIGURES...... xi ABSTRACT ...... xx Chapter 1. Introduction to Mass Spectrometry and FT-ICR MS ...... 1

History of Mass Spectrometry...... 1

FT-ICR MS ...... 5

Summary ...... 10

Chapter 2. Resolution of 10,000 Compositionally Distinct Components in Polar Coal Extracts by Negative-Ion Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry ...... 13 Introduction ...... 13

Experimental Methods...... 16

Sample Preparation...... 16

Mass Analysis...... 16

Mass Calibration and Data Reduction ...... 17

Results and Discussion ...... 19

Kendrick Mass Analysis ...... 24

Heteroatomic Classes...... 25

Alkylation Patterns ...... 30

Type Analysis (Rings + Double Bonds)...... 30

Conclusion...... 34

v Chapter 3. Detailed Compositional Analysis at Different Stages of Coal Liquefaction by Negative-Ion Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry ...... 37 Introduction ...... 37

Experimental Methods...... 39

Sample Preparation...... 40

Results and Discussion ...... 40

Kendrick Plot ...... 40

Heteroatomic Classes...... 46

Alkylation Patterns ...... 46

Type Analysis (Rings + Double Bonds)...... 47

Elemental Analysis ...... 47

Conclusion...... 50

Chapter 4. Compositional Determination of Acidic Species in Illinois #6 Coal Extracts by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry...... 54 Introduction ...... 54

Experimental Methods...... 55

Isolation of Acid Fractions...... 55

Sample Preparation...... 56

Mass Analysis...... 56

Results and Discussion ...... 57

Homologous Series and Compound Class...... 57

Comparison of Acid Fractions and Pyridine Extract...... 61

Compound Type and Carbon Distribution ...... 61

Conclusion...... 68

vi Chapter 5. Two and Three Dimensional van Krevelen Diagrams: A Graphical Analysis Complementary to the Kendrick Mass Plot for Sorting Elemental Compositions of Complex Organic Mixtures Based on Ultrahigh-Resolution Broadband FT-ICR Mass Measurements ...... 69 Introduction ...... 69

Experimental Methods...... 72

Sample Preparation...... 72

Mass Calibration and Data Reduction ...... 72

Results and Discussion ...... 72

Elemental Compositions from a Plot of Kendrick Mass Defect

vs. Kendrick Nominal Mass ...... 72

Sorting of Compound Classes by Use of a van Krevelen

Diagram...... 73

Three Dimensional van Krevelen Diagram ...... 77

Visual Comparisons between Various Fossil Fuels ...... 82

Conclusion ...... 83

Chapter 6. Comparative Compositional Analysis of Untreated and Hydrotreated Oil by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry ...... 88 Introduction ...... 88

Experimental Methods...... 90

Sample Preparation...... 90

Mass Analysis...... 90

Kendrick Mass and Data Reduction ...... 90

Results and Discussion ...... 90

Untreated Fuel...... 90 vii Species Removed or Generated by Hydrotreatment...... 91

Effect of Single-Stage Hydrotreatment on Aromaticity and

Carbon Distribution...... 92

Effect of Two-Stage Hydrotreatment on Aromaticity and

Carbon Distribution...... 99

Conclusion...... 100

Chapter 7. Composition of Explosives by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry ...... 104 Introduction ...... 104

Experimental Methods...... 106

Sample Preparation...... 106

Mass Analysis...... 107

Results and Discussion ...... 108

TNT, RDX, and HMX...... 108

Military TNT...... 109

Smokeless Powder ...... 110

Powermite...... 114

Military C4 Explosive...... 114

Conclusion...... 118

Chapter 8. Characterization of Vegetable Oils: Detailed Compositional Fingerprints Derived from Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry ...... 121 Introduction ...... 121

Experimental Methods...... 123

Sample Preparation...... 123

Mass Analysis...... 124 viii Mass Calibration and Data Reduction ...... 124

Vertical Scaling of Mass Spectra...... 124

Results and Discussion...... 125

Negative-Ion ESI FT-ICR MS...... 125

Elemental Compositions ...... 125

Acidic Heteroatomic Classes ...... 125

Fatty Acids ...... 128

Tocopherols...... 130

Positive-Ion ESI FT-ICR MS...... 130

Elemental Compositions ...... 130

Triacylglycerols & Diacylglycerols ...... 130

Detection of Intentional Adulteration...... 131

Conclusion ...... 139

Chapter 9. Limitations and Conclusion...... 141

REFERENCES ...... 143

BIOGRAPHICAL SKETCH...... 162

ix LIST OF TABLES

Table 2.1. Bulk elemental composition of Pocahontas #3 and Illinois #6 coals. The values are based on the organic components in the coal, and are reported relative to 100 carbon atoms per "formula weight." ...... 36

Table 2.2. Some of the identified peaks of Illinois #6 at 469 Da. The measured mass, theoretical mass and assigned elemental compositions are shown within a mass accuracy of ~0.5 ppm...... 36

Table 2.3. Some of the identified peaks of Pocahontas #3 at 469 Da. The measured mass, theoretical mass and assigned elemental compositions are shown within a mass accuracy of ~0.5 ppm...... 36

Table 7.1. Matched ions observed by ESI FT-ICR MS, before and after explosion of military TNT. The series of the sulfur-containing species appear to constitute a unique marker for this explosive product...... 119

Table 7.2. Matched ions observed by ESI FT-ICR MS, before and after explosion of military smokeless powder...... 119

Table 7.3. Matched ions observed by ESI FT-ICR MS, before and after explosion of military Powermite...... 120

LIST OF FIGURES

Figure 1.1. Ion cyclotron motion. The magnetic field is directed perpendicular to the plane of the paper. The ion is moving into a circle by the inward-directed Lorentz magnetic force...... 6

Figure 1.2. Plot of cyclotron frequency as a function of m/z in different magnetic field strength. Typical ion mass-to-charge ratios (~15 ≤ m/z ≤ ~104) correspond to cyclotron frequencies from a few kHz to a few MHz for the given magnetic field strength...... 7

Figure 1.3. 9.4 T (passively shielded) FT-ICR mass spectrometer installed at the National High Magnetic Field Laboratory. Pressure in the cell is typically at 10-9 to 10-10 Torr by 4 stages of vacuum pumping system...... 8

Figure 1.4. Open-ended cylindrical cell with (right) and without (left) capacitive rf coupling between the three sectors. Ions are trapped in the z-direction by the trap plates, T, and excited, detected in the xy direction by the magnetic field. There are two excitation (E) and two detection (D) plates...... 10

Figure 2.1. Broadband electrospray ionization FT-ICR mass spectra of Illinois #6 coal (top) and Pochahontas #3 coal (bottom. All species are singly charged, as evidenced by the ~1 Da spacing between each monoisotopic species and its corresponding nuclide containing one 13C in place of 12C...... 20

Figure 2.2. Mass scale expanded mass spectra, 360 < m/z < 460, and 395 < m/z < 400, of the ESI FT-ICR mass spectrum of Illinois #6 coal, showing multiple peaks at every nominal mass. Asterisks denote members of the same alkylation series, separated by 14.0157 Da (CH2) bottom). For such a homologous series, the Kendrick mass scale simplifies assignment of chemical formulas. Peaks separated by

xi 2.0157 Da (upper right) correspond to successive losses of H2 (i.e., addition of a ring or double bond)...... 22

Figure 2.3. Top: Mass scale-expanded segments for ESI FT-ICR mass spectra of Illinois #6 coal at m/z 469 (top) and Pochahontas #3 coal (bottom). Of the 33 (or 19) peaks resolved in Illinois #6 (or Pochahontas #3) coal, 29 (or 15) can be matched to elemental composition masses to within 1.0 ppm, and sorted into 17 (or 11) different classes...... 23

Figure 2.4. Plot of Kendrick mass defect vs. nominal Kendrick mass for odd-mass ions in the ESI FT-ICR mass spectrum of Pochahontas #3 coal. Three classes are identified, each of which exhibits an alkylation series (horizontal rows) and type series (vertically spaced horizontal rows--see text)...... 26

Figure 2.5. Kendrick plot as in Fig. 2.4, but for even-mass ions from Pochahontas #3 coal...... 27

Figure 2.6. Kendrick plot as in Fig. 2.4, but for odd-mass ions from Illinois #6 coal...... 28

Figure 2.7. Kendrick plot as in Fig. 2.6, but for even-mass ions from Illinois #6 coal...... 29

Figure 2.8. Relative abundances of various ion classes of Illinois #6 (top) and Pochahontas #3 (bottom) coal, including species containing O, N, and S heteroatoms. O-containing classes such as O2, O3 and O4 are dominant for Illinois #6 coal, whereas the neutral nitrogen class (N) comprises almost 40% of the total for Pochahontas #3 coal...... 31

Figure 2.9. Alkyl carbon distributions for ions containing 2 oxygens and 19 rings plus double bonds in the ESI FT-ICR mass spectra of Illinois #6 and Pocahontas #3 coals...... 32

Figure 2.10. Distributions of rings plus double bonds for ions containing two oxygen atoms, in Illinois #6 and Pocahontas #3 coals. Note the much higher aromatic content for Pochahontas #3 coal, in accord with its higher rank...... 33

Figure 3.1. Broadband electrospray ionization FT-ICR mass spectra of early stage (resid, top) and final stage (liquid, bottom) of liquefaction of subbituminous coal from the Black Thunder

xii Mine (Wyoming). The resid sample contains ~7,000 resolved peaks and the liquid ~4,000 peaks...... 42

Figure 3.2. Plot of Kendrick mass defect vs. nominal Kendrick mass for all peaks in the ESI FT-ICR mass spectrum of coal resid (Fig.3.1, top)...... 43

Figure 3.3. Plot of Kendrick mass defect vs. nominal Kendrick mass for all peaks in the ESI FT-ICR mass spectrum of coal liquid (Fig. 3.1, bottom)...... 44

Figure 3.4. Relative abundance of different heteroatom-containing classes in resid (11 classes) and liquid (6 classes) from the data in Figure 3.1...... 45

Figure 3.5. Carbon number distribution for members of the O2 class of coal resid and liquid, each with 8 rings plus double bonds. Note the much wider carbon distribution for resid compared to liquid...... 48

Figure 3.6. Distribution of rings plus double bonds comparison for the O class of coal resid and liquid. The coal liquid is much more saturated than the coal resid...... 49

Figure 3.7. Mass scale-expanded segments for coal resid and liquid near m/z 381. Odd mass requires that all species contain an even number (in this case, zero) of nitrogen atoms. All species are singly charged, and parentheses denote the number of rings plus double bonds. The O2 class with 1 ring or double bond dominates the coal liquid mass spectrum...... 52

Figure 3.8. Mass scale-expanded segments for coal resid and liquid near m/z 430. Notation is as for Fig. 3.7. Even mass requires that all species contain an odd number (in this case, one) of nitrogen atoms; thus, the remaining On species have no nitrogens but contain one 13C in place of 12C...... 53

Figure 4.1. Broadband negative-ion ESI FT-ICR mass spectra of Illinois #6 coal samples. Top: Coal acids fraction; Middle: Acidic asphaltene fraction; Bottom: Coal pyridine extract...... 58

Figure 4.2. High-mass segment of the negative-ion ESI FT-ICR mass spectrum of saturated naphthenic acids from the coal acids fraction. Homologous series of compounds differing in

xiii degree of alkylation are evident from the characteristic mass spacings of 14.0157 Da mass (CH2)...... 59

Figure 4.3. Mass scale expansion of the negative-ion ESI FT-ICR mass spectrum for species of nominal mass, 619 Da. The fully saturated naphthenic acid is the most abundant species...... 60

Figure 4.4. Mass scale expansion of the negative-ion ESI FT-ICR mass spectrum for species of nominal mass, 437 Da. Note the higher compositional complexity for the pyridine extract relative to the two acid fractions...... 63

Figure 4.5. Mass scale expansion of the negative-ion ESI FT-ICR mass spectra for species of nominal mass, 407 Da coal acids, acidic asphaltenes and pyridine extract of Illinois #6 coal. Some acid species are detectable only in the two acid fractions, and the combined acid fractions contain more acid compounds than the pyridine extract alone...... 64

Figure 4.6. Type distributions (rings plus double bonds, or double bond equivalents) for the coal acids fraction and pyridine extract of Illinois #6 coal. The O3 class is dominant in the pyridine extract whereas the O2 class dominates in the coal acid fraction. Compounds in the pyridine extract are more unsaturated than in the acid extract...... 65

Figure 4.7. Type distributions for sulfur-containing classes in coal acids and acidic asphaltenes from Illinois #6 coal. The acidic asphaltenes fraction is more aromatic and more compositionally complex...... 66

Figure 4.8. Carbon number distributions for O2 species with 19 rings and double bonds, for acidic asphaltenes and the pyridine extract of Illinois #6 coal. The carbon number distribution for acidic asphaltenes is shifted to higher mass relative to that for the pyridine extract...... 67

Figure 5.1. Kendrick mass defect vs. Kendrick nominal mass for pyridine-extracted Pocahontas #3 coal. This plot can visually sort up to thousands of compounds horizontally according to number of CH2 groups, and vertically according to class (heteroatom composition) and type (rings plus double bonds), thereby extending assignment of unique elemental composition from accurate mass measurement to 900 Da or higher. Both odd and even masses are shown. DBE denotes xiv double bond equivalence: i.e., number of rings plus double bonds...... 74

Figure 5.2. Two-dimensional van Krevelen diagram for compounds of class O3 in pyridine-extracted Pocahontas #3 coal. The compounds in homologous series corresponding to varying degrees of alkylation appear along lines that intersect at an atomic ratio of 2 on the H/C axis (see text). Similarly, a vertical line connects homologous series differing in degree of unsaturation...... 75

Figure 5.3. Two-dimensional van Krevelen diagram showing members of three classes: O, NO2, and O3 from pyridine-extracted Wyoming subbituminous coal. The three classes are separated graphically based on their different O/C ratios. Note the different slopes for alkylation series of the same type (10 rings plus double bonds)...... 78

Figure 5.4. Two-dimensional van Krevelen diagram showing members of classes O2 and NO2 from pyridine-extracted Wyoming subbituminous coal. Because these two classes have the same number of oxygen atoms, they have identical O/C ratios, but may still be distinguished by their different H/C ratios...... 79

Figure 5.5. Two-dimensional van Krevelen diagram for pyridine- extracted Wyoming subbituminous coal, but this time plotted with N/C ratio instead of O/C ratio as the abscissa. The N and N2 classes are thereby graphically separated due to their different N/C ratios...... 80

Figure 5.6. Three-dimensional van Krevelen diagram for members of the classes, N, NO, and NO2, from pyridine-extracted Pocahontas #3 coal. Each class differs by at least one heteroatom (by definition), and is thus shifted to a different plane. Different classes are thus completely separated in the three- dimensional display ...... 81

Figure 5.7. Three-dimensional van Krevelen diagram for the same classes (N, NO, and NO2) for two different fossil fuels: coal (blue) and crude oil (red). Because the coal components are more aromatic than are the constituents of crude oil, the two fuels are readily distinguished graphically in the diagram...... 85

xv Figure 5.8. Three-dimensional van Krevelen diagram for compounds of three classes (NO2, NO3, and O2) from two coals of different rank: Illinois #6 and Pocahontas #3. The higher rank Pocahontas #3 is more aromatic; thus, members of all of its shared classes have smaller H/C ratios and are displaced in the diagram from those of the lower rank and more saturated Illinois #6...... 86

Figure 5.9. Three-dimensional van Krevelen diagram for members of the three most abundant classes (NO, NO3, and O) from pyridine-extracted Wyoming (Thunder mine) coal at early and final stages of liquefaction. The number of nitrogen- containing species clearly decreases dramatically after liquefaction. Moreover, members of all classes become more saturated. The three-dimensional van Krevelen diagram provides simple graphical evidence for heteroatom removal and hydrodenitrogenation resulting from liquefaction of coal...... 87

Figure 6.1. Broadband negative-ion (top) and positive-ion (bottom) ultrahigh-resolution mass spectra of an untreated1:1 (v/v) mixture of light cycle oil and refined chemical oil. Non-basic compounds are detected as negative ions (top) and basic compounds as positive ions (bottom)...... 93

Figure 6.2. Relative abundances of nine major basic heteroatomic classes (i.e., detected as positive ions) in the untreated fuel sample of Figure 6.1.)...... 94

Figure 6.3. Mass scale expansion at nominal mass, 373 Da, for untreated (top) and after each of three different catalytic hydrotreatments (bottom). Note the different selective removal of various species by each the three catalytic processes (see text). EI-41 removes all heteroatomic compounds whereas EI-43 has high abundant species left...... 95

Figure 6.4. Mass scale expansion at nominal mass, 413 Da, for untreated (top) and after each of three different catalytic hydrotreatments (bottom). Note that the EI-43 process generates new species not originally detected (see text)...... 96

xvi Figure 6.5. Mass scale expansion at nominal mass, 466 Da, for untreated (top) and after each of three different catalytic hydrotreatments (bottom)...... 97

Figure 6.6. Rings plus double bonds distributions before and after EI-43 catalytic hydrotreatment of the sample of Figure 6.1, for each of its three most abundant heteroatomic classes. Highly aromatic compounds are partially reduced to yield increased relative abundance of less aromatic species (see text)...... 98

Figure 6.7. Carbon distributions for members of the N3 class class with 14, 17 and 20 rings plus double bonds, before and after one- stage EI-43 hydrotreatment...... 101

Figure 6.8. Rings plus double bonds distributions for members of the N2O class, before and after two-stage EI-42 hydrotreatment...... 102

Figure 6.9. Carbon distributions for members of the N2O class with 13, 25 and 30 rings and double bonds before and after two-stage E-42 hydrotreatment...... 103

Figure 7.1. ESI FT-ICR mass spectrum of RDX explosive. The mass difference between the two major peaks is CH2 (14.0157 Da), not 14N (14.0031)...... 111

Figure 7.2. ESI FT-ICR mass spectrum of HMX explosive. The two major - - peaks are [HMX+HCO2] and [HMX+CH3CO2] ...... 112

Figure 7.3. Low-mass (200-300 Da, left) and high-mass (300-500 Da, right) segments of the ESI FT-ICR mass spectra of species found in military TNT before (top, plotted normally) and after (bottom, plotted upside down) detonation. As for pure TNT, the most abundant species is the deprotonated molecular ion, [TNT-H]-. Elemental compositions for various other species recovered after the explosion are listed in Table 7.1...... 113

Figure 7.4. ESI FT-ICR mass spectra of species found in military smokeless powder, before and after combustion, plotted as in Figure 7.3. Elemental compositions for various other recovered species are listed in Table 7.2. The matched species arise from both the active agent, observed as - [nitroglycerin+NO3] , and the same series of sulfur-containing species found in military TNT (Fig. 7.3)...... 115 xvii

Figure 7.5. ESI FT-ICR mass spectra of species found in military Powermite, before and after explosion, plotted as in Figure 7.3. Elemental compositions for various other species recovered after the explosion are listed in Table 7.3. Major - species include the series, [(NaNO3)nNO3] ...... 116

Figure 7.6. ESI FT-ICR mass spectra of species found in military C4, before and after explosion, plotted as in Figure 7.3. The only species found in both spectra are

[RDX+NO3]-, [RDX+H2CO+HCO2]-, and [RDX+H2CO+CH3CO2]- (see text)...... 117

Figure 8.1. Broadband electrospray ionization FT-ICR negative-ion mass spectra of acidic components of canola oil (top), olive oil (middle), and soybean oil (bottom). In each spectrum, peak heights are scaled relative to the highest-magnitude peak...... 126

Figure 8.2. Relative abundances of various ion heteroatomic classes for the three vegetable oils. All three oils contain the same major classes, but soybean oil differs from the other two by its much higher relative abundance of class O2...... 127

Figure 8.3. Mass scale-expanded segments from Figure 8.1, showing relative abundances of various C18 fatty acids in three vegetable oils. The compositional differences readily distinguish soybean oil from canola or olive oils...... 129

Figure 8.4. Mass scale-expanded segment of the negative-ion ESI FT-ICR mass spectrum of soybean oil, showing relative abundances of three tocopherols...... 132

Figure 8.5. Mass scale-expanded segments from Figure 8.1, showing relative abundance of ,-tocopherol in three vegetable oils: olive oil can readily be distinguished from canola and soybean oils. Note the need for ultrahigh mass resolution, to separate the ,-tocopherol signal from other interferant components of the same nominal mass...... 133

Figure 8.6. Positive-ion ESI FT-ICR mass spectral segment for non- acidic components of soybean oil. Note the resolution of multiple elemental compositions at the same nominal mass...... 134

xviii

Figure 8.7. Mass scale-expanded segments of the positive-ion ESI FT- ICR mass spectrum of three vegetable oils, showing relative abundances of various triacylglycerols. The three oils are readily distinguished according to their triacylglycerol relative abundance patterns...... 135

Figure 8.8. Mass spectral segments as in Figure 8.7, but for diacylglycerols, providing yet another fingerprint to distinguish three vegetable oils...... 136

Figure 8.9. Fatty acid distributions for mixtures of olive oil and soybean oil. The presence and extent of soybean oil as an adulterant may be determined from these patterns (see text)...... 137

Figure8.10. Triacylglyerol distributions for mixtures of olive oil and soybean oil. The presence and extent of soybean oil as an adulterant may be determined from these patterns (see text)...... 138

xix

ABSTRACT

Fourier transform ion cyclotron resonance mass spectrometry (FT-

ICR MS) has ultrahigh mass resolving power (m/∆m50% >300,000) and high mass accuracy (<1 ppm), which enables separation and identification of elemental compositions of complicated mixtures. Electrospray ionization (ESI) provides selective ionization of polar heteroatomic compounds without isolation from the complex mixtures. In

addition, the Kendrick mass scale (where CH2 is 14.00000 rather than 14.01565) is introduced in petrochemical analysis for an easy and fast data reduction. Petrochemicals like coal and crude oil are among the most chemically complex natural mixtures in the world. Heteroatomic compounds contribute to the instability in storage and environmental

contamination by releasing relevant acid precursor gases such as SOx

and NOx upon combustion. To meet stringent environmental regulations and to produce a quality product, it is necessary to remove those heteroatoms through clean fuel technology. Thus it is essential to monitor the fates of the heteroatom-containing compounds through those processes. We first apply ESI FT-ICR mass spectrometry to the analysis of Illinois #6 and Pocahontas #3 pyridine coal extracts. With the aid of Kendrick mass scaling, the elemental compositions can be sorted into homologous series according to compound “class” and “type”. Both the compositional and aromaticity data correlate well with known compositional and geochemical information. Furthermore, the same technique is extended into study of coal liquefaction and coal fractionation. In the coal liquefaction study, a distillation resid and a xx further processed liquid product were examined by ESI FT-ICR MS. The resid sample contains more heteroatomic compounds whereas the liquid sample is lower in average mass and also much more saturated. The coal fractionation technique facilitates sorting of coal components by comparing a standard pyridine extract with two alternative fractions designed to concentrate acidic components: coal acids and acidic asphaltenes. The resulting detailed compositional analysis of coal acids provides detailed distributions of heteroatomic classes, aromaticity, and alkylation of coal. That kind of information establishes a fundamental basis for assessing the role of those acids in coal processing. Moreover, we utilized two- and three- dimensional van Krevelen diagrams that allow visual resolution of complex mixtures. The van Krevelen plot not only graphically exaggerates the difference between different classes, but more importantly affords a simple graphical basis for exposing compositional differences between samples of different nature, origin, and processing. We also extended this work to other complex mixtures, such as hydrotreated fuels, military explosives and vegetable oils. In the study of fuel hydrotreatment, detailed elemental composition comparisons of different hydrotreating conditions provide a new and rational basis for optimizing parameters for hydrotreatment of commercial oils. And in explosive analysis, we are able to identify both active and non-active components in post-blast explosive residues. Thus it provides the forensic basis to trace the origin of an explosive. In vegetable oil analysis, we resolve and identify literally thousands of distinct chemical components of commercial canola, olive and soybean oils, without extraction or other wet chemical separation pretreatment. We suggest that adulteration of vegetable oils can be detected by detailed elemental compositional fingerprints.

xxi

CHAPTER 1. INTRODUCTION TO MASS SPECTROMETRY AND FT- ICR MS

HISTORY OF MASS SPECTROMETRY

In 1897, J.J Thomson at Cambridge’s Cavendish Laboratory constructed the first mass spectrometer, or a “parabola spectrograph” to determine mass to charge ratio of ions produced from electrical discharges through gases. The ions are first generated in discharge tubes then passed into electric and magnetic fields. The ions are forced to move through parabolic trajectories and then detected on a photographic plate or fluorescent screen. Thomason received the Nobel Prize in physics in 1906 for his initiation of mass spectrometry. At the end of World War I, Francis W. Aston (who helped design Thomson’s equipment) in Cambridge and Arthur J. Dempster at University of Chicago made the clear demonstration of mass spectrometry respectively. Aston used both electrostatic and magnetic fields to focus ions on a photographic plate whereas Dempster developed a magnetic deflection instrument with direction focusing. Dempster also developed the first electron impact source, which is still widely used in modern mass spectrometers today. Their continued work along with contribution of other researchers led to Aston’s 1922 Nobel Prize in Chemistry for isotope studies carried out with this type of instrument. In early 1931 E.O. Lawrence introduced the cyclotron principle, the core of Fourier transform ion cyclotron resonance mass spectrometry, and in 1936 F. M. Penning added an electrostatic confinement perpendicular to the magnetic field later known as the “Penning trap”.

1 In the 1940s the first commercial mass spectrometer entered the market based on Dempster’s single-focusing design. These early commercial mass spectrometers were primarily used in the petroleum industry for quantitative analysis of organic gas mixtures. In time of flight mass spectrometry (TOF MS), first designed and constructed in the late 1940s and mid-1950s, ions are separated by differences in their velocities as they move in a straight path toward a collector in the order of increasing mass-to-charge ratio. In the early 1950s Wolfgang Paul and co-workers invented the quadrupole mass filter and the quadrupole ion trap, later called the “Paul ion trap”. The concept of using crossed ratio- frequency and electrostatic fields to trap ions led to the first commercial ion trap system in 1983 as a GC detector. This eventuated in Paul’s Nobel Prize (in physics) in 1989. Ion cyclotron resonance MS (ICR MS) instrumentation was greatly improved in the middle to late 1960s. In 1965, the first commercial ICR MS was introduced by Varian Associates (Palo Alto, Calif.). In 1971 R.T. McIver took a major step in ICR MS by introducing combined trapping and measurement in one cell called “trapped ion analyzer cell”. In 1974, M. B. Comisarow and Alan. G. Marshall revolutionized ICR by developing Fourier transform ICR mass spectrometry (FT-ICR MS). The major advantage of FT-ICR MS is that it allows many different ions to be determined at once instead of one at a time. Since then Marshall and others have continuously improved FT-ICR MS, increasing the mass range, mass resolving power and sensitivity of this technique. In the recent years, hybrid mass spectrometers were also constructed such as the quadrupole-time-of-flight (Q-TOF) and TOF-TOF techniques. Tandem mass spectrometry (MS/MS) techniques have been proved effective in structural characterization and identification. In a MS/MS experiment, a precursor ion is first mass-selected and then fragmented by different fragmentation techniques, followed by mass analysis of the product ions. There are several means to produce 2 fragments, among which the most commonly used one is collision- induced dissociation (CID). Electron-capture dissociation (ECD), a newly invented dissociation method, has gained much attention for its unique sequence ability in peptide and small proteins analysis. The improvement of ionization techniques also plays an important role in mass spectrometry development. In field desorption, first observed by Muller in 1951, positive ions are desorbed from a surface in a strong electrostatic field gradient. Field desorption techniques were thus coupled with mass spectrometry to study non-volatile and thermally unstable molecules. In the mid-1960s Field and Muson introduced chemical ionization (CI), in which a reagent gas ionized by electrical discharge and the reagent gas then reacts with and ionizes the target molecules. Compared with electron impact ionization (EI), chemical ionization is a more gentle ionization method that usually produces molecular ions without fragments. During the 1960s and the following decades, secondary-ion mass spectrometry (SIMS) was introduced and developed. An ion beam with ion energies in the keV range is used to desorb and ionize molecules from a sample deposited on a surface, and was successfully applied in materials science and other fields. In the 1970s plasma desorption mass spectrometry (PDMS) was developed by Macfarlane and coworkers at Texas A&M university. PDMS uses very high-energy ions to desorb and ionize molecules in solid-film samples and it was the first ionization method that successfully ionizes high-molecular weight molecules such as peptides and proteins. In 1978 M. A. Posthumus, P. G. Kistemaker and L.C. Meuzelaar at the FOM Institute for Atomic and Molecular Physics in Amsterdam developed laser desorption mass spectrometry (LDMS), in which a photon beam is used to desorb sample molecules. In 1981 Barber and coworkers at the University of Manchester Institute of Science & Technology in England developed fast atom 3 bombardment mass spectrometry (FAB MS). It is also known as “liquid SIMS”, in which beams of neutral atoms ionize compounds of interest gently from the surface of a liquid matrix. The two most recent developed ionization techniques brought major impact on mass spectrometry in biological sample analysis: matrix assisted laser desorption/ionization (MALDI) and electrospray ionization (ESI). MALDI MS was initiated by F. Hillenkamp and M. Karas at the University of Frankfurt, Germany. In MALDI, a laser is used to desorb sample molecules from a solid or liquid matrix containing a highly UV- absorbing substance. ESI MS was first conceived in 1960s by M. Dole of Northwestern University and put into practice in the early 1980s by J. B Fenn and coworkers at Yale University. In ESI MS, highly charged droplets dispersed from a capillary in an electric field are evaporated and produced ions are drawn into the mass analyzer. The dramatic advantage of ESI MS and MALDI MS is their ability to ionize large biomolecules such as peptides and proteins, which made MS especially useful for sophisticated biomedical analysis. Moreover, due to the ability to generate multiple charged ions, ESI coupled with FT-ICR mass spectrometry is particularly powerful. Fenn and Tanaka shared the Nobel Prize (chemistry) in 2002 for their contribution in mass spectrometry of biomedical analysis. ESI and MALDI have and will continue to expand the application of mass spectrometry with molecular characterization of proteins, DNA and other large molecules.

FT-ICR MS

THEORY

An ion is subject to a Lorentz force (FL) when traveling in a spatially uniform magnetic field (B) v v v FL = qν × B (1.1) in which q and ν are the charge and velocity of the ion, respectively. The scalar form of Eq. (1.1) is then

F = qν xy B0 (1.2)

2 2 in which νxy (= ν x + ν y ) is the ion velocity in the plane perpendicular to

the magnetic field (z-axis is defined as the direction opposite to B) as

2 described in Figure 1.1. Because angular acceleration dν dt = ν xy r ,

2 F= mν xy r (1.3) where r is the radius of the circular path. Then Eq. (1.2) equals (1.3) to get

2 mν xy qν B = (1.4) xy 0 r and simplified to

qB ν xy 0 = . (1.5) m r The angular velocity, ω (in radians/second), is defined by

ν xy = ω . (1.6) r Substitute Eq. (1.5) to (1.6) to get

5 qB ω = 0 (S.I. units) (1.7a) m or

ω .1 535611×107 B ν = c = 0 (1.7b) c 2π m z

in which νc is in Hz; B0 is in Tesla; m is in unified atomic mass units, and z is in multiples of elementary charge. A unique feature of equation 1.7 is that ions with a given m/z value have the same ICR frequency, independent of their velocity. It is especially useful for mass spectrometry because frequency is more precise than translational energy focusing in determination of m/z. Furthermore, ICR frequencies for ions formed from typical molecules range from a few kilohertz to a few megahertz (shown in Figure 1.2), a very convenient range for commercially available electronics.

v + v v - q v × B q v × B

Figure 1.1. Ion cyclotron motion. The magnetic field is directed perpendicular to the plane of the paper. The ion is moving into a circle by the inward-directed Lorentz magnetic force. [1]

Figure 1.2. Plot of cyclotron frequency as a function of m/z in different magnetic field strength. Typical ion’s mass-to-charge ratios (~15 ≤ m/z ≤ ~104) correspond to cyclotron frequencies from a few kHz to a few MHz for the given magnetic field strength. [1]

In order to resolve and identify components in complex mixture, a high magnetic field is required. From eq. (1.7a), taking the first derivative with respect to m, yields:

dωc − qB = 0 (1.8) dm m2

m qB = 0 (1.9) ∆m m∆ωc where ∆m stands for the peak width at some fixed fraction of the peak height (usually 50%). Therefore, mass resolving power (m/∆m) increases in proportion with increasing magnetic field (B). Hence, increase of high magnetic field increases mass resolving power (and thus mass accuracy) resulting in the improved separation and decrease of peak coalescence. 7 Second, the maximum number of trapped ions increases as the square of the magnetic field. Thus, the higher the magnetic field is, the more ions can be trapped in the cell. This facilitates simultaneous identification of thousands of components in a complex mixture by raising the number of low abundant ions to a detectable level.

Instrumentation

The homebuilt 9.4T FT-ICR mass spectrometer (Figure 1.3) consists of five major components: the ion source and guide, the vacuum system, the magnet, the ICR cell/ion trap, and the data acquisition.

Storage Open Trap Octopoles Transfer Electron Gun Octopole ESI Source CO2 Laser

9.4 T Magnet

Quadrupole Mass Filter Figure 1.3. 9.4 T (passively shielded) FT-ICR mass spectrometer installed at the National High Magnetic Field Laboratory. Pressure in the cell is typically at 10-9 to 10-10 Torr by 4 stages of vacuum pumping system. [2]

Ion source (Electrospray Ionization)/guide

The ion source and guide generates ions and transfers them to the ICR cell, respectively. Our instrument is implemented with electrospray ionization and all of the data from this dissertation were obtained from this 9.4 T electrospray (ESI) FT-ICR mass spectrometer. In ESI, a large

8 positive or negative potential (1000~5000 V) is applied to a capillary needle through which a sample solution is moving. For negative mode ESI, negative ions are enriched at the surface of the liquid at the capillary tip whereas the positive ions are moving to the needle electrode. The repulsion of negative ions at the surface and the pull of the electric field on the counter electrode on the other side overcome the surface tension of the liquid to form a so-called “Taylor cone”. As the droplets pass through a heated capillary in the mass spectrometer, the solvent evaporates and they break into smaller droplets, finally yield gas-phase ions. The generated ions are externally accumulated [3] in n-poles and are transferred to the ICR cell for excitation and detection.

The vacuum system

A low vacuum system is required in any types of mass spectrometers to present the absence of collision. The pressure inside the cell must be so low that mean free path exceeds the size of cell. The ions are generated from ESI in atmospheric pressure whereas the cell inside is vacuum. There are 4 stages of pumping systems in our instrument and the pressure in the cell is typically 10-9~10-10 Torr.

The magnet

A spatially homogeneous high magnet is crucial to FT-ICR experiments. Our instrument uses a 9.4 T superconducting magnet for its stronger magnetic field, higher homogeneity and lower maintenance cost.

The ICR cell/trap

The ICR cell is where the ions are excited and detected. Various forms of ICR cell have been investigated. Basically, they all consist of three major components: the excitation, detection, and end cap (trapping) plates. In a capacitive-coupled open-ended cylindrical cell

9 (Figure 1.4) dipolar excitation is more linear and “z-ejection” is eliminated.

Data acquisition

A high performance data system for FT-ICR mass spectrometry termed MIDAS [53] (Modular ICR Data Acquisition System) has been developed in house. The time-domain signal (containing information on all ions detected) is acquired and then Fourier transformed into a frequency spectrum. After calibration terms, the frequency spectrum can be further transformed into a magnitude spectrum as a function of m/z. The software was developed with Lab Windows/CVI programming and runs on a PC.

SUMMARY

This Chapter presents some history and fundamentals of mass spectrometry and details of Fourier transform ion cyclotron resonance mass spectrometry. FT-ICR-MS, with its ultrahigh resolving power and mass accuracy, allows for the resolution of multiple peaks at one nominal mass and assignments of elemental compositions for most of the peaks in the mass spectrum. Electrospray ionization (ESI), introduced in

10 the late 1980’s but applied only recently to petrochemicals and its derivatives can selectively generate singly-charged molecular ions from NSO-containing compounds in petroleum, without prior extraction or chromatographic separation. Chapter 2 introduces ESI FT-ICR mass spectrometry for the analysis of coal pyridine extracts. The advantages of Kendrick mass scale are further discussed for mass analysis and data reduction. Chapter 3 continues the success of the instrument in identifying polar heteroatom containing compounds through coal liquefaction process. A fractionation technique is discussed in Chapter 4 that isolates and concentrates coal into two acidic components: coal acid and acidic asphaltene. The combined acid fractions provide detailed distributions of heteroatomic classes, aromaticity, and alkylation of coal. Chapter 5 presents a novel graphic method- van Krevelen diagram to visually distinguish different classes in the same sample and samples of different nature, origin, and processing. Chapter 6 studies hytrotreatment conditions of fuel on removal efficiency of heteroatomic compounds. Identification of species that increase or decrease in abundance, and (especially) which new species are produced, should greatly improve understanding of differential catalytic efficiency for removal and/or conversion of potentially all of the chemical constituents of an oil feedstock. Chapter 7 describes the first application of ESI FT-ICR mass spectrometry to military explosives. Both active and non-active components were recovered in the post-blast residues. Since non-active components present characteristic fingerprints (indigenous or artificial additives) for specific explosives, their presence and elemental composition can potentially identify the source of the product. Finally, chapter 8 extends application of ESI FT-ICR mass spectrometry to characterization of vegetable oils. FT-ICR MS resolves and identifies literally thousands of distinct chemical components of commercial canola, olive and soybean oils, without extraction or other 11 wet chemical separation pretreatment. A "fingerprint" provides unprecendented detail for correlating vegetable oil origin and/or adulteration with inherent chemical composition. Relative abundances within each of several chemical families (e.g., fatty acids, di- and triacylglycerols, tocopherols, etc.) offer multiple independent bases for comparisons.

CHAPTER 2. RESOLUTION OF 10,000 COMPOSITIONALLY DISTINCT COMPONENTS IN POLAR COAL EXTRACTS BY NEGATIVE-ION ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

Coal composition is affected by climate, water level and water chemistry during its formation. [4] The inorganic material in coal (mainly Al, Ca, Fe, K, Mg, Na and Si) results from deposition by wind or incorporation during plant decomposition. [5] All coals, regardless of their origin, age or type, can be arranged in an order of coal rank. Rank, as applied to coal, defines the degree or extent of maturation and is therefore a qualitative measure of carbon content. [4] Higher rank is accompanied by higher carbon and energy contents and lower moisture content. Coal is one of the world's most chemically complex natural mixtures. Size exclusion chromatography (SEC) is commonly used to characterize the molecular size distributions of complex coal mixtures. [6-9] Molecular size and structure of different coal-derived extracts have also been analyzed. [10-13] Winans et al. have characterized organics in Argonne Premium coals, including hydrocarbon composition, conversion behavior, and linkages between aromatics in different coal samples by pyrolysis fast-atom bombardment mass spectrometry. [14-17] NMR has been used to identify functional groups (e.g., carbonyls) in coal and coal- derived liquids. [18-22] Thin-layer chromatography has been applied for the analysis of hydrocarbon types [23] and to characterize polar compound distributions. [24] Metals in coal have been determined by X- ray absorption spectroscopy, [25] atomic absorption spectroscopy, and X- ray fluorescence spectroscopy. [26, 27]

13 It is desirable to process coal prior to combustion to improve its utilization efficiency and reduce the emissions of oxides of sulfur. For example, the many heteroatom-containing components of coal yield gaseous pollutants such as SOX and NOX, [28] whose levels are controlled by the Environmental Protection Agency (EPA): e.g. the 1990 Clean Air

Act Amendment limits the annual emission of SO2 to 8.9 million tons in the U.S.A. Liquefaction (including hydrodesulfurization and denitrogenation) of coal reduces the heteroatom content [29-31] and facilitates transport. [4] Thus, it is important to monitor the fate of sulfur- and nitrogen-containing compounds during the coal liquefaction process. Veloski and coworkers detected nitrogen-containing compounds in coal by ammonia chemical ionization coupled to a high-resolution double-focusing sector mass spectrometer. [32] Using gas chromatography/mass spectrometry, Damsté et al. [33] found sulfur- containing polycyclic aromatics with more than one sulfur atom per molecule. Oxygen- and nitrogen-containing compounds in benzene- methanol extracts from coal during the coalification process have been studied by GC, GC/MS, and pyrolysis fast atom bombardment mass spectrometry by Winans et al. at Argonne National Labroatory. [29, 34, 35] Nevertheless, relatively little is currently known about the detailed chemical composition of coals and how that composition varies among different coals. Historically, analysis of polar constituents of coal required prior isolation from a predominately hydrocarbon matrix. That isolation often involved time-consuming and difficult separation schemes. Even so, subsequent GC/MS analysis of those polar components typically lacked the chromatographic and/or mass resolution to distinguish and identify individual compounds. For distillates of geologically related petroleum crude oil, the first detailed elemental composition analysis came from electron ionization coupled with high-resolution Fourier transform ion cyclotron resonance 14 mass spectrometry (EI FT-ICR MS). Hundreds of molecular radical cations could be resolved and their elemental composition determined by accurate mass measurement of volatile hydrocarbons and heteroatom- containing aromatics. [36-41] Electrospray ionization (ESI), introduced in the late 1980’s but applied only recently to petroleum and its derivatives generates singly-charged molecular ions from NSO-containing compounds in petroleum, without prior extraction or chromatographic separation. [42] ESI coupled to high-field (9.4 T) FT-ICR MS has resolved and identified elemental compositions (CcHhNnOoSs) of up to literally thousands of NSO-containing compounds (observed as positive or negative molecular ions from basic or acidic species, respectively) in crude oils and their distillates. [43-49] Those chemical formulas in turn reveal not only the heteroatom content ("class") but also the compound "type" (number of rings plus double bonds) and alkylation distribution for each "class" and "type". FT-ICR MS routinely affords ultrahigh mass resolving power (m/∆m50% > 350,000, in which ∆m50% is the mass spectral peak full width at half-maximum peak height) and mass accuracy (<1 ppm) allowing for molecular formula assignment of singly charged heteroatomic ions up to ~1000 Da. [45] Even more recently, ESI FT-ICR MS has provided the first detailed elemental compositional analysis of humic substances as well. [50-52] In this work, we extend the ESI FT-ICR MS methods previously demonstrated for petroleum to the first detailed chemical compositional analysis of polar compounds from coal. We resolve and identify several thousand compounds in coals of two different geochemical origins, and examine their class, type, and alkylation distributions. This effort lays the groundwork for future connections between the chemical composition of coals and their properties (processibility, pollution potential, etc.).

EXPERIMENTAL METHODS

Sample preparation. Illinois #6 coal and Pocahontas #3 coal were obtained from Argonne National Laboratory and have previously been analyzed by nuclear magnetic resonance (NMR), laser desorption mass spectrometry and fast atom bombardment tandem mass spectrometry techniques. The contents of a 5 g ampoule of coal (100 mesh particle size) were added to 80 mL of pyridine in a glass flask and extracted in a sonicator waterbath for 45 min. The solid particles were separated from pyridine by vacuum filtration. The filter cake was rinsed with 70 mL of pyridine in small portions and washed with 1 mL of pyridine. (Pyridine is chosen because of its high efficiency for extraction of organics.) The sample was concentrated and dried with a rotary evaporator to give a final weight of about 1 gram. Prior to analysis by ESI-FT-ICR MS, a solution of 10 mg of sample was dissolved in 3 mL of pyridine and then diluted with 17 mL of methanol. One milliliter of the solution mixture was removed and spiked with 10 µL of pure (99.9%) ammonium hydroxide. All solvents used were HPLC grade (Fisher Scientific, Pittsburgh, PA). Mass analysis. Mass analysis was performed with a homebuilt FT- ICR mass spectrometer equipped with a 22 cm diameter horizontal bore 9.4 T magnet (Oxford Corp., Oxney Mead, England). [2] Data were collected and processed with a modular ICR data acquisition system (MIDAS). [53, 54] Negative ions were generated from a microelectrospray source equipped with a 50 µm i.d. fused silica micro ESI needle. Samples were infused at a flow rate of 400 nL/min. Typical ESI conditions were: needle voltage, -1.8 kV; tube lens, -390 V; and heated capillary current, 4 A. Ions were accumulated external to the magnet [3] in a linear octopole ion trap (14 cm long) for 45 s and transferred through rf-only multipoles to a 10 cm diameter, 30 cm long open

16 cylindrical Penning ion trap. Multipoles were operated at 1.7 MHz at a peak-to-peak rf amplitude of 170 V. Broadband frequency-sweep ("chirp") dipolar excitation [55] (70 kHz to 1.27 MHz at a sweep rate of 150 Hz/µs and peak-to-peak amplitude, 190 V) was followed by direct- mode image current detection that yielded 4 Mword time-domain data. The time-domain data were processed and Hanning-apodized, followed by a single zero-fill before fast Fourier transformation and magnitude calculation. Frequency was converted to mass-to-charge ratio (m/z) by the quadrupolar electric trapping potential approximation [56, 57] to generate a mass-to-charge ratio (m/z) spectrum.

Mass calibration and data reduction. Mass spectra were calibrated from internal standards (n-pentadecanoic, n-heptadecanoic, n- nonedecanoic and n-henicosanoic fatty acids) introduced by a dual electrospray ion source. [58] The mass spectra were then recalibrated with respect to identified homologous series in each coal sample. Each recalibration included at least 15 peaks to yield an rms error less than 0.3 ppm. The mass values for singly-charged ions of 250-700 Da, with relative abundance greater than 3 times the standard deviation of the baseline noise, were imported into an Excel spreadsheet. Conversion of measured masses from the IUPAC mass scale (12C = 12.00000 Da) to the

Kendrick mass scale (CH2 = 14.0000 rather than 14.01565 Da) facilitates identification of homologous series. [45] Kendrick mass is obtained from IUPAC mass as shown in Eq. 1. [59] Kendrick Exact Mass = IUPAC Mass x (14.0000/14.0156)

The Kendrick mass scale provides the advantage that the members of a homologous series (namely, compounds with the same heteroatom composition and number of rings plus double bonds, but different

17 numbers of CH2 groups) have identical Kendrick mass defect (KMD), defined in Eq. 2.

KMD = (Nominal Kendrick Mass-Kendrick Exact Mass)*1000

Homologous series are thus readily selected from a list of all observed ion masses. Nominal Kendrick mass is obtained by rounding the Kendrick mass to the nearest integer. Homologous series are separated and grouped by sorting even and odd nominal Kendrick masses and KMD's as described elsewhere. [45, 60] Molecular formulas are assigned by use of a molecular formula calculator in the in-house MIDAS FT-ICR analysis software. The mass tolerance was set to ±1 ppm, namely, 3 times the standard deviation of the error (0.3 ppm) in assignment of the internal mass calibrants. Molecular formulas were limited to a maximum of 50 12C atoms, 100 1H, 2 13C, 5 14N, 10 16O, 2 32S and 1 34S. If two elemental compositions lie within the mass tolerance, one formula may usually be confirmed/eliminated unequivocally by the presence/absence of a peak corresponding to replacement of one 12C by 13C. Because members of a

homologous series differ only by integer multiples of CH2, unequivocal assignment of a single member of such a series suffices to identify all of the other members (see Kendrick mass discussion below). The composition of a heteroatom-containing hydrocarbon is often

expressed by the chemical formula, CcH2c+zX, in which c is the carbon number, Z is the hydrogen deficiency index with respect to a saturated hydrocarbon with the same number of carbons, and X denotes the constituent hetero-atoms (N, S, O). Each additional ring (not aromatic ring) or double bond eliminates two hydrogens, and thus makes Z more negative by 2. In this paper, we prefer to report the number of rings plus double bonds (R + DB) as a more direct measure of unsaturation.

18 R+DB= -Z/2+1 for species that do not contain nitrogen, whereas R+DB =(-Z+1)/2+1 for species containing nitrogen. For example, for benzene

+. radical cation, C6H6 , Z = -6 and R+DB = 4 = (6/2) + 1, whereas for

+. pyridine radical cation, C5H5N , Z = -5 and R+DB = 4 = (5+1)/2 + 1 For convenience, we shall abbreviate molecular formulas according to their R+DB ("type") and heteroatomic composition ("class"). Finally, all species in the present mass spectra are singly charged, as evidenced by the ~1 Da spacing between each monoisotopic species and its corresponding ion with one 12C replaced by 13C.

RESULTS AND DISCUSSION

Negative ion ESI FT-ICR MS resolves more than 10,000 compositionally distinct compounds (enabling assignment of chemical formulas (CcHhNnOoSs) to more than 7,500 species) in a pyridine extract of Illinois #6 coal, and resolves ~3,000 chemical species in Pocahontas #3 coal (Fig. 2.1). Because all detected ions are singly charged, we shall henceforth denote each ion by its mass in Da rather than its mass-to- charge ratio, m/z. For both coals, the overall molecular weight distribution is roughly Gaussian, centered at ~400 Da (Illinois #6 coal) or ~450 Da (Pochahontas #3 coal. Thus, coals do not appear to contain significant quantities of pyridine extractable heteroatom-containing species greater than ~800 Da--a finding similar to that previously reported for ESI FT-ICR MS of petroleum crude oil and asphaltenes. The big difference in the number and variety of elemental compositions of the two coals is consistent with their previously reported oxygen content than Pocahontas coal (Table 2.1 ). That higher heteroatom content is manifested in a corresponding increase in the number of observed neutral nitrogen compounds as well as petroleum acids, both of which are known to be selectively ionized by negative ion ESI. [49]

Figure 2.1 Broadband electrospray ionization FT-ICR mass spectra of Illinois #6 coal (top) and Pochahontas #3 coal (bottom. All species are singly charged, as evidenced by the ~1 Da spacing between each monoisotopic species and its corresponding nuclide containing one 13C in place of 12C.

20 Conversely, Pocahontas #3 coal has a higher carbon content, and thus a higher coal rank (i.e., higher degree of aromaticity) with fewer heteroatom compounds.

For Illinois #6 coal, the average mass resolving power, m/∆m50%, exceeds 190,000 from 225-750 Da. Mass scale expansion (Figure 2.2), 360 < m/z < 460, reveals peaks separated by multiples of 2.0157 Da,

corresponding to successive loss of H2 (i.e., an additional ring or double bond). Further mass scale-expansion over a 1 Da interval at m/z 469 Da of the ESI FT-ICR mass spectrum of Illinois #6 coal (Figure 2.3, top) resolves 33 compositionally distinct compounds, of which 29 (spanning 17 heteroatom classes, mostly containing 2-4 oxygen atoms) could be identified by matching to within 1.0 ppm of a putative elemental composition. Similarly, the same mass segment for Pochahontas #3 coal (Figure 2.3, bottom) resolves 19 compositionally distinct compounds, of which 15 (spanning 11 heteroatom classes) were identified. It is also worth noting that most of the assigned ions in these two spectra are different: all species found in Illinois #6 coal contain oxygen, whereas most species from Pocahontas #3 coal contain nitrogen (most notably "neutral nitrogen", containing just one N atom). Thus, it appears likely that it will be possible to distinguish geologically different coals according to their elemental composition patterns.

r R

400

* 460 460 399 *

398 2.0157 Da 2.0157

* 440 440 m/z *

397 ) )bottom). For such a homologous series, the series, a homologous such For ) )bottom). 460, and 395 < m/z < 400, of the ESI FT-IC the of < 400, < m/z 395 and 460, 2 < * al formulas. Peaks separated by 2.0157 Da (uppe by 2.0157 separated Peaks al formulas.

Da 2.0157 m/z m/z 396 aks at every nominal mass. Asterisks denote members denote mass. Asterisks aks at every nominal < 420 420 * * m/z m/z 395

addition(i.e., of a ring or double bond). 2

400 400 *

380 380

Da 14.0157

spectra, 360 mass expanded scale Mass

Coal #6 Illinois 360 360

MS T FT-ICR 9.4 ESI

mass spectrum of Illinois #6 coal, showing multiple pe multiple #6 coal, showing of Illinois mass spectrum 14.0157 Da (CH by separated series, alkylation of the same of H losses to successive right) correspond of chemic assignment simplifies scale mass Kendrick 2. 2. Figure

2 O

469.402 O 469.258 3 O #3) coal, 29 (or Pochahontas

NO 469.361 469.238 4 O

2 O N S

2 2 2 2 O O

(top) 469 m/z at coal #6 Illinois of spectra mass O 469.32 Pochahontas #3 Coal NO 469.217 5 4 (or 11) different into 17 within 1.0 ppm, and sorted O 19 Mass Cpds Nominal Same of O aks resolved in Illinois #6 aks resolved in Illinois #6 3

Coal #6 Illinois N Not Not ID’d 3 S 3 3 2 O 3 469.197 O O O 469.279 m/z 2

3 S 33 Cpds of Same Nominal Mass Nominal Cpds Same 33 of 3 N NO O 2 O N 4 N

O 2 S 2 469.176 N 3 4

2 O 469.238 O O 2

NO 2 N 5 O 4 O O 2 3 O 4 N O 469.156

2 N S N 3 3 469.197 NO O O 5 2 6 3

O O O 2 3 O O Top: Mass scale-expanded segments for ESI FT-ICR FT-ICR for ESI segments scale-expanded Top: Mass 2 N 469.135 2 N 4 N N O

469.156 (or 15) can be matched to elemental composition masses to composition to elemental (or 15) can be matched classes. pe 19) #3 coal (bottom). Of the 33 (or and Pochahontas Figure 2.3. Figure

23 Kendrick mass analysis. Different heteroatomic classes (e.g., O vs. O2) as well as any species differing in number of rings plus double bonds have different Kendrick mass defects. Moreover, members of a given class and type but different degree of alkylation are separated by

multiples of 14.0157 Da (CH2) in mass (Fig. 2.2, bottom), but all share the same Kendrick mass defect. Thus, a particularly convenient graphical method for organizing all of the above information a plot of Kendrick mass defect vs. nominal Kendrick mass. [45] It is convenient to construct separate plots of odd-mass (even number of nitrogens, Figures 2.4 and 2.6) and even-mass (odd number of nitrogens, Figures 2.5 and 2.7) ions. For example, consider all ions of the same class (i.e., containing one oxygen atom) in the Kendrick plot of Figure 2.4. Each horizontal row represents an alkylation series, in which each successive data point corresponds to an additional CH2 group. Each additional ring or double bond moves that row upward by the Kendrick mass defect of two hydrogens. Thus, for a given class and type, it is easy to recognize members of a given alkylation series. The importance of such a display is not simply visual. At low mass (<300 Da or so), it is possible to assign a unique elemental composition, based on mass resolving power of 300,000. At higher mass, assignment based on mass measurement accuracy alone is no longer unique--however, starting from one unique mass assignment at low mass, the Kendrick plot quickly identifies the other members of a homologous alkylation series, so that mass assignments can extend with confidence to ~900 Da.

A second advantage of the Kendrick plot is that it quickly exposes anomalous peaks, because they do not fit in any of the class, type, and/or alkylation series. Thus, one might suspect that the high- magnitude peaks at ~620 Da in the mass spectrum of Fig. 2.1 (top) could represent some sort of non-coal impurity. However, those spectral data fall in standard alkylation series in the Kendrick plots of Figs. 2.5 and 24 2.7), and thus appear to be genuine constituents of coal. We infer that the unusually high magnitude for those components results from unusually high abundance of those classes and types, and/or unusually high ionization efficiency for those species. Whatever the source, those ions have elemental compositions consistent with other components of coal. Finally, the Kendrick plots for Illinois #6 coal (Figs. 2.6 and 2.7) exhibit a significantly wider range of mass defects than those for Pochahontas #3 coal (Figs. 2.4 and 2.5). The reason is that Pochahontas #3 coal contains more nitrogen, and therefore exhibits typically smaller mass defects, because 14N has a smaller absolute mass defect (0.003 Da) than 16O (0.005 Da), and components of Illinois #6 coal typically contain more oxygens than there are nitrogens in Pochahontas #3 coal. Heteroatomic classes. The distribution of heteroatomic classes for each coal was determined by dividing the sum of the relative abundances of all species of a given class by the sum of all the relative abundances of all identified species in the mass spectrum (Figure 2.8). These distributions obviously provide a convenient detailed fingerprint for heteroatom content of coal. However, one must be careful not to equate relative abundances of ions and the relative abundances of the corresponding neutrals in the original mixture. For example, the bulk abundances of nitrogen and sulfur in Illinois #6 approximately are the same (Table 2.1), the mass spectrum exhibits significally higher relative abundances for nitrogen-containing over sulfur-containing ions. One reason is that sulfur-containing molecules are typically less polar (e.g., thiophenes) and therefore not as easily ionized by ESI compared with polar nitrogen-containing compounds.

2 2 O N O

ions in the odd-mass

coal. Three classes are identified, each o are identified, coal. Three classes nal Kendrick mass for

Nominal KendrickMass

#3 Pocahontas Odd-Mass Ions

Plot of Kendrick mass defect vs. nomi mass Plot of Kendrick

250 300 350 400 450 500 550 600 which exhibits an alkylation series (horizontal rows) and type series (vertically spaced (vertically type series and rows) (horizontal an alkylation series which exhibits rows--see text). horizontal ESI FT-ICR mass spectrum of Pochahontas #3 #3 of Pochahontas mass spectrum ESI FT-ICR 450 300 250 400 350 200 2.4. Figure Mass Kendrick Defect

2 3

N NO NO NO

r even-mass ions from Pochahontas #3 coal. ions from Pochahontas even-mass r

Mass Kendrick Nominal

#3 Pocahontas Ions Even-Mass

as in Fig.2.4, but fo plot Kendrick

250 300 350 400 450 500 550 600 650

2.5. Figure

450 400 350 300 250 200

Mass Kendrick Defect

S S 4 2 3 2 3 O O O O O

4 O 78 H

40 C

Nominal KendrickMass

Kendrickplot as in Fig. 2.4, but forodd-mass ions from Illinois #6 coal.

Coal #6 Illinois Odd-Mass Ions

2.6. Figure

200 300 400 500 600 700 800 0 Kendrick Mass Kendrick Defect 50 500 450 400 350 300 250 200 150 100

4 5 2 3

NO NO NO NO

650 700 750

600

for even-mass ions from Illinois #6 coal. #6 coal. ions from Illinois even-mass for

450 500 550

Mass Kendrick Nominal

350 400

Kendrick plot as in Fig. 2.6, but plot Kendrick

Illinois #6 Coal: Even-Mass Ions Even-Mass Coal: #6 Illinois

2.7. Figure

250 300

Mass Kendrick Defect 400 300 100 500 450 350 250 200 150

29 Moreover, Pocahontas #3 coal contains only slightly less bulk nitrogen than Illinois #6 coal, but nevertheless exhibits mainly N- containing ions (almost 40% of all heteroatomic species) rather than O- containing ions. Evidently most of the nitrogen-containing ions from Illinois #6 coal originate from compounds that also contain oxygen or sulfur and are thus more acidic and thus exhibit higher efficiency of ionization to yield negative ions. We infer that most of the oxygen- containing species in Pocahontas #3 coal do not occur as carboxylic acids. Alkylation patterns. Figure 2.9 shows the distribution in number of alkyl carbons for each of the two coals, for species containing two oxygen atoms and 19 rings plus double bonds. Illinois #6 coal clearly exhibits a wider alkyl carbon distribution as well as a higher average number of alkyl carbons. Thus, Illinois #6 coal differs from Pochahontas #3 coal not only in its proportion of carboxylic acid groups, but also in its degree and range of alkylation.

Type analysis (rings + double bonds). The number of rings + double bonds is a direct measure of unsaturation. For petroleum and coal, a higher degree of unsaturation typically reflects increased aromaticity. For example, the number of rings + double bonds for the most abundant class (O3) in Illinois #6 coal ranges from 5-31 (mainly 12- 23), whereas the most abundant class (N) of Pocahontas #3 ranges from 15-36 (mainly 20-29). Alternatively, Figure 2.10 shows the rings + double bonds distributions for the same class (O2) in the two coals, for which the rings + double bonds range from 3-23 (Illinois #6) and 14-30 (Pochahontas #3). The mass spectral data are therefore consistent with the higher rank (higher aromaticity) of Pochahontas #3 coal. However, the mass spectral results provide a much more detailed picture, by providing the full distribution of unsaturation, rather than just an overall carbon content. 30

are

ied

ied if if

t t

and O and

4 4 en en 3

O O

, O Unid Unid

N N 2

Unidentified

O O

2 N N

S S 5 5

O O

S S

4 4 O

O O

S S 3 3

O O

NS S S 2 2

O 6

6 3

NO NO (bottom) coal, #3 Pochahontas (top) and linois #6

NO 3

O O

roatoms. O-containing classes such as O such classes O-containing roatoms. ONO

N 2 2

2 N

Illinois #6 Coal

NO 2

NO O O

#3 Coal Pocahontas

5 O

O NO

4 O

O 2

3 3

N O

O 2

2 O

O N

Relative abundances of various ion classes of Il classes ion various of abundances Relative 5 0 8 6 4 2

14 12 10 45 40 35 30 20 25 15 10

% Relative abundance Relative % abundance abundance Relative Relative % % abundance Relative %

including species containing O, N, and S hete O, N, including species containing dominant for Illinois #6 coal, whereas the neutral nitrogen class (N) comprises almost 40% of the total fo 40% of the comprises almost class (N) nitrogen coal, whereas the neutral #6 dominant for Illinois #3 coal. Pochahontas 2.8. Figure

Illinois #6

in bonds double 19 rings plus ng and 2 oxygens #3 Pocahontas

32 34

Class with Class Alkyl Carbons s #6 and Pocahontas #3 coals. #3 Pocahontas s #6 and

2 30 O Bonds + Double 19 Rings

containi for ions distributions carbon Alkyl 8 6 4 2 0 20 18 16 14 12 10 % Relative Abundance

of Illinoi the ESI FT-ICR mass spectra

2.9. Figure

#3 Pocahontas class:

2 s for ions containing two oxygen atoms, in two oxygen atoms, s for ions containing 18 20 22 24 26 O

much higher aromatic content for Pochahontas #3 content for Pochahontas much higher aromatic

14 16

Rings + Double Bonds 10 12

class: Illinois #6 class: Illinois

plus double bond of rings Distributions 2 O

coal, in accord with its higher rank. Note the #3 coals. and Pocahontas Illinois #6 Figure 2.10. Figure 8 6 4 2

0 12 10 % Relative Abundance

CONCLUSION

In this first analysis of coal by ESI FT-ICR MS, we report the complete chemical compositional characterization of two crude coal samples of different rank, without prior chromatographic separation. Coupling of FT-ICR MS to electrospray ionization affords the selective ionization of polar heteroatom compounds. Ultrahigh mass resolution and mass accuracy, amplified by Kendrick mass analysis, enables unique identification of thousands of distinct elemental compositions (i.e., ~75% of the resolved mass spectral peaks), sorted according to compound "class" (numbers of various heteroatoms), "type" (rings + double bonds), and number of alkyl carbons. Oxygen- and sulfur- containing classes are the most abundant ions in Illinois #6 coal; whereas nitrogen-containing ions dominate the Pocahontas #3 coal mass

spectrum. Some heteroatom classes (e.g., O2 and NO2) are found in both coals. Pocahontas #3 coal exhibits much higher aromaticity but less alkyl substitution, consistent with its previously known lower carbon content and higher rank. The present results open up a new era in coal characterization, by providing a compositional basis from which to compare coal origins,

maturation, processing, and production of ultimate NOx and SOx combustion products. We shall next proceed to focus on the compositional concomitants of coal liquefaction study and coal fractionation, specifically the fate of heteroatom species in the process of coal liquefaction. Coal fractionation will facilitate sorting of coal components by chemical functionality, as well as concentrate low- abundance components for mass analysis. Finally, it is important to recognize that the relative abundances of ions in any mass spectrometric experiment do not necessarily reflect the relative abundances of neutrals from which they derive in the original sample. For example, electrospray ionization does not yield ions from 34 hydrocarbons unless they contain heteroatoms (N,O,S). Therefore, complete compositional analysis will require ionizing coal in several different ways: e.g., electron ionization (for volatiles, including thiophenes), field desorption (for less volatile compounds, including hydrocarbons), and electrospray ionization (for heteroatom-containing compounds). Nevertheless, the observed mass spectral ion classes, types, and alkylation patterns provide so much detail that they are very likely to prove useful in comparing different coals, and ultimately predicting their properties on a compositional basis.

35 Table 2.1. Bulk elemental composition of Pocahontas #3 and Illinois #6 coals. The values are based on the organic components in the coal, and are reported relative to 100 carbon atoms per "formula weight."

Carbon Hydrogen Oxygen Nitrogen Sulfur Pocahontas 100 58.5 2.0 1.2 0.2 Illinois#6 100 77.3 13.1 1.5 1.2

Table 2.2. Some of the identified peaks of Illinois #6 at 469 Da. The measured mass, theoretical mass and assigned elemental compositions are shown within a mass accuracy of ~0.5 ppm. Assigned Composition Theoretical Mass Measured Mass

- C32H21O4 469.1445 469.1443 - C29H25O4S 469.1478 469.1476 - C31H21N2O3 469.1557 469.1556 - C29H25O6 469.1656 469.1654 - C33H25O3 469.1809 469.1807 - C30H29O3S 469.1842 469.1840 - C30H29O5 469.2020 469.2019 - C34H29O2 469.2173 469.2172 - C31H33O2S 469.2206 469.2204 - C31H33O4 469.2384 469.2383 - C32H37O3 469.2748 469.2746 - C30H45O4 469.3323 469.3322 - C32H53O2 469.4050 469.4048

Table 2.3. Some of the identified peaks of Pocahontas #3 at 469 Da. The measured mass, theoretical mass and assigned elemental compositions are shown within a mass accuracy of ~0.5 ppm. Assigned Composition Theoretical Mass Measured Mass - C34H17N2O 469.1346 469.1348 13 - C34H18NO C 469.1428 469.1430 - C36H21O 469.1598 469.1600 - C35H21N2 469.1710 469.1712 13 - C35H22N C 469.1791 469.1793 - C32H25N2O2 469.1922 469.1923 13 - C32H26NO2 C 469.2003 469.2004 - C34H29O2 469.2173 469.2175 - C35H33O 469.2537 469.2536

CHAPTER 3. DETAILED COMPOSITIONAL ANALYSIS AT DIFFERENT STAGES OF COAL LIQUEFACTION BY NEGATIVE-ION ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

Coal, the most abundant, widely distributed, and economical fossil fuel, is a solid with high carbon content but low hydrogen content (usually less than 6%). Coal cannot be used in internal combustion engines (ICN) or turbines because it is solid and has low hydrogen content (average hydrogen content for liquid fuels is ~12.5%). Furthermore, it is difficult to store and transport low quality coal because it is very porous and can self-ignite when exposed to air. Therefore, in order to broaden the applications of coal, it has to be converted into a liquid with higher hydrogen content. Currently, the cost of converting coal into liquid fuels is higher than the cost of refining crude oil. However, this technique coal liquefaction will become more and more attractive as the international demand for energy increases and the oil reservoirs are depleted, especially for some coal-rich countries. The two major methods to convert coal into liquid fuels are carbonization (removal of carbon) and liquefaction (addition of hydrogen). Direct liquefaction, the most efficient route currently available, with catalyst and a hydrogen donor solvent at a very high temperature and pressure can convert coal into artificial petroleum. In indirect liquefaction coal is completely gasified with steam. The gasification products are mixed with H2 and CO with the removal of sulfur-containing species. Over a catalyst at relatively low pressure and temperature, the mixture reacts to produce the final synthesis liquid fuel. By adjusting

37 the composition of catalyst, hydrogen/carbon ratio, temperature, pressure, etc. one can obtain a variety of different products, such as paraffins, olefinic hydrocarbons or alcohols. [61] The liquefaction rate and extent depend heavily on the temperature, pressure, and catalyst. [11, 35, 62-67] Diverse coal liquefaction procedures are tailored to the properties of the specific coal resource or desired final products: e.g., The SRC process was developed to produce cleaner boiler fuels; whereas the CANMET hydrocracking process is intended to co-process coal with petroleum bitumen for conversion into distillate products. [68-72] In addition to converting coal into liquid fuel, coal liquefaction can also efficiently remove heteroatomic compounds. N-, S-, O- and metal- containing compounds contribute to fuel instability during storage [73,

74] and to air pollution upon combustion by release of NOx and SOx gases. [28, 75] Direct coal liquefaction is considered to take place in two consecutive steps: conversion to a soluble form and reduction in molecular weight and removal of heteroatoms by hydrodesulfurization and denitrogenation, [29-31, 76, 77], known as the "upgrading" process. Thus, it is important to monitor the fate of sulfur- and nitrogen- containing compounds during the coal liquefaction process. However, research has been hampered due to the lack of precise structural information on heteroatomic groups in coal. [9, 78-80] Moreover, coal liquefaction processes involve complex multi-stage refining. The development and optimization of coal liquefaction requires a complete evaluation of not only the products but also the distribution of polar heteroatom-containing compounds throughout the entire process. Such detailed compositional information has been unattainable (resid samples) or required months to complete (distillate samples) by GC-MS. Moreover, detailed compositional information relates to the end use of the coal- derived liquid. For example, if used as petroleum feedstock,

38 characterization and hence separation of specific compounds such as phenols is necessary. [81] We have previously demonstrated electrospray ionization (ESI) Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) for analysis of Illinois #6 and Pocahontas #3 pyridine coal extracts. [82] ESI coupled to high-field (9.4 T) FT-ICR MS has resolved and identified elemental compositions (CcHhNnOoSs) of up to literally thousands of NSO-containing compounds (observed as positive or negative molecular ions from basic or acidic species, respectively) in crude oils and their distillates. [43-49]. Here, we extend the ESI FT-ICR MS methods to detailed chemical compositional analysis of polar compounds in fractions isolated from different stages of the coal liquefaction process. We are able to trace all of the heteroatomic species for two very different types of samples (distillation resid and distillation liquid) in the multi-stage complex refining process without any pre- chromatographic separation. This effort lays the groundwork for better understanding of the fate and/or modification of heteroatom containing polar organics in coal liquefaction and therefore improvement of current liquefaction techniques.

EXPERIMENTAL METHODS

The coal liquefaction samples were kindly provided by CONSOL Energy Inc. (Pittsburgh, PA). The materials were all produced in 1996 in a single continuous bench-scale direct coal liquefaction unit run at Hydrocarbon Technologies, Inc. The feedstock to the run was subbituminous coal from the Black Thunder Mine (Wyoming). The early stage sample, PFL [pressure filter liquid) resid, period 19, corresponds to the distillation residue of the PFL stream taken during 19th day of the process. The resid was produced from the recycled PFL stream by batch distillation to an endpoint of ~850º F. The final liquid sample, SOH,

39 period 23, is the separator overhead (SOH) stream and the major net product of the process. It is a coal-derived substitute for petroleum distillate but not a fully refined consumer fuel. The liquefaction process has been described elsewhere. [4, 83]

Sample preparation. The resid sample is a dark solid and the liquid sample is a yellowish liquid. Prior to ESI-FT-ICR MS analysis, a solution of 10 mg of each sample was completely dissolved in 5 mL of pyridine and then diluted with 15 mL of methanol. One milliliter of the solution mixture was removed and spiked with 10 µL of pure (99.9%) ammonium hydroxide. All solvents were HPLC grade (Fisher Scientific, Pittsburgh, PA).

RESULTS AND DISCUSSION

Negative ion ESI FT-ICR MS resolves ~ 7,000 compositionally distinct compounds in the coal resid sample and ~4,000 chemical species in the coal liquid sample (Figure 3.1.) (Positive-ion ESI FT-ICR MS (not shown) yielded far fewer peaks and was not nearly as useful for compositional identification.) Because all detected ions are singly charged, we henceforth denote each ion by its mass in Da rather than its mass-to-charge ratio, m/z. Not surprisingly, the average mass for the coal liquid polar constituents is much lower than that for the coal resid sample. Moreover, the reduced number of polar heteroatomic compounds in the coal liquid sample suggests that the heavier heteroatom- containing species remain in the coal resid of the previous batch distillation process. Therefore, the heteroatom-enriched resid must be recycled to remove those polar compounds.

Kendrick plot. The Kendrick plot [45] graphically sorts compounds horizontally according to their Kendrick nominal mass and

40 vertically according to their Kendrick mass defect. Each horizontal row represents a homologous series of compounds having the same class (number of N, S, and O atoms) and same "type" or degree of saturation

(number of rings plus double bonds), but different numbers of alkyl (CH2) groups. For each class, each additional ring or double bond shifts the corresponding carbon distribution series upward by the Kendrick mass defect of two hydrogen atoms. Finally, each different class has a different Kendrick mass defect, and is thus displaced vertically from members of any other class. Thus, the Kendrick plot simultaneously separates all elemental compositions separately according to their class, type, and carbon distribution. [45] The Kendrick plot has proved especially advantageous for assignment of thousands of elemental compositions in ESI FT-ICR mass spectra of petroleum [48] and humic substances. [84, 85] Kendrick plots for coal resid and coal liquid are shown in Figures 3.2 and 3.3. Although mass accuracy alone suffices for unique elemental composition assignments only up to ~300 Da, the patterns evident from the Kendrick plots allow for extension for homologous series up to 600+ Da, and those series then serve as internal calibration for the entire mass spectrum. The Kendrick plot thus enables complete compositional assignments discussed in the following section. For now, we simply note that the coal resid sample exhibits many more species of high Kendrick mass defect, indicating that coal resid has higher aromatic content than coal liquid. A second benefit of the Kendrick plot is that it quickly exposes anomalous peaks, because they do not fit in appropriate class, type, and/or alkylation series. For the coal liquid sample, for example, the isolated vertical rows of data in Fig 3.3 represent electronic artifacts (e.g., at ~550 Kendrick nominal mass) or non-coal impurities (upper left) because they have atypical mass defects and do not belong to any homologous series. 41

800

700 the Black Thunder Mine (Wyoming). Thunder the Black 600 spectra of early stage (resid, top) and final top) and (resid, early stage of spectra 500 m/z ed peaks and the liquid ~4,000 peaks. peaks. ~4,000 liquid the and peaks ed 400 300 Broadband electrospray ionization FT-ICR mass FT-ICR ionization electrospray Broadband stage (liquid, bottom) of liquefaction of subbituminous coal from coal from of subbituminous liquefaction of stage (liquid, bottom) resolv ~7,000 contains sample The resid Resid Liquid Figure 3.1. Figure

Resid

in the ESI FT-IC for all peaks mass Kendrick

Nominal KendrickMass

defect vs. nominal mass Plot of Kendrick

200 250 300 350 400 450 500 550 600 650 700 0 Kendrick Mass Kendrick Defect 600 400 100 500 300 200 top). 3.1, (Fig. resid coal of mass spectrum

3.2. Figure

Liquid

Nominal Kendrick Mass Kendrick Nominal

FT-ICR in the ESI peaks for all mass Kendrick nominal defect vs. mass of Kendrick Plot 200 250 300 350 400 450 500 550 600 650 700 0 Kendrick Mass Kendrick Defect 200 600 500 400 300 100 mass spectrum of coal liquid (Fig. 3.1, bottom). Figure 3.3. 3.3. Figure

Unidentified 2 O 2 N

O 2 Liquid N Resid 3

3 O classes) (11 resid in classes of different heteroatom-containing abundance Relative

2 O

3.1. in Figure from the data classes) (6 and liquid

3.4. Figure O

0 5 Relative Abundance% 20 15 10 45 40 35 30 25

45 Heteroatomic classes. The distribution of heteroatomic classes for the processed coal was determined by dividing the sum of the relative abundances of all species of a given class by the sum of the relative abundances of all identified species in the mass spectrum (Figure 3.4). Fig. 3.4 reveals that the coal resid sample contains 11 classes, vs. only 7 for the coal liquid. No significant amount of sulfur-containing compounds was found in either sample, presumably because most sulfur-containing compounds were removed earlier in the liquefaction process. Moreover, molecules containing sulfur but not nitrogen (e.g., thiophenes) are typically less polar and therefore not as easily deprotonated in negative-ion electrospray relative to more acidic nitrogen- and oxygen-containing compounds. By the same token,

carboxylic acids, (O2 class) are more acidic and are thus more efficiently deprotonated than so-called "neutral nitrogen" species to yield negative

ions. Therefore, although the O2-class ions are the most abundant in both coal resid and liquid mass spectra, O2-class neutrals are not necessarily most abundant in the original coal samples. The nitrogen-containing compounds are much higher in abundance in the coal resid than in the liquid. Evidently fewer of them are recovered by distillation. For example, NO, N2O and N2O2 classes are not found at all in the coal liquid, and N, NO2 and NO3 classes are only partly represented, whereas all of those classes are seen in coal resid.

Alkylation patterns. Figure 3.5 shows the distribution in number of alkyl carbons for coal resid and liquid, for species containing two oxygen atoms and 8 rings plus double bonds. Coal resid clearly exhibits a wider alkyl carbon distribution as well as a higher average number of alkyl carbons (~30 vs. ~21 for coal liquid). Although hydrogenation would be expected to increase the extent of alkylation, the liquefaction process also reduces the average molecular weight per

46 compound and in the process lowers the number of -CH2 groups per molecule.

Type analysis (rings plus double bonds). The O class, including phenolic compounds, is usually chosen to monitor coal liquefaction process because it is widely present in coal-derived liquids and plays a critical role as an intermediate in many common coal conversion reactions. Figure. 3.6 shows the rings plus double bonds distributions for the O class in the two samples, ranging from ~3-30 for coal resid vs. ~3-15 for coal liquid. Highly aromatic compounds have clearly been removed in producing the liquid product; the highly aromatic compounds either remain in the solid during distillation or are distilled and further hydrotreated (more saturated) in later stages of liquefaction. In any case, the coal resid has not only a wider carbon distribution but also higher aromatic content than the coal liquid, properties consistent with recycling of the coal resid for cracking by hydrotreatment.

Elemental analysis. A mass scale-expansion of ESI FT-ICR mass spectra for both coal resid and liquid near 381 Da (Figure 3.7) shows dramatic changes in relative abundances of various polar species. For even-electron species (as for electrospray ionization), ions of odd (even) mass must contain an even (odd) number of nitrogen atoms, [86] in this case zero. Although several On species appear at comparable abundance in coal resid, most are found at vastly lower abundance (note 10x vertical scale enhancement in the bottom spectrum of Fig. 3.7) relative to species with two oxygens and one ring or double bond in coal liquid.

Interestingly, two new, less saturated (O and O2) species appear after liquefaction--they must have been generated during the hydrotreatment process.

40 42

class of coal resid and liquid, each and of coal resid class 34 36 38

with 8 Rings plus Double 2 O (Resid) Stage Early Bonds:

with 8 Rings plus double 26 28 30 32 2

onds: Final Stage (Liquid) Carbon Number Carbon O b

Distributions Carbon

Carbon number distribution for members of the O the of members for distribution number Carbon 18 20 with 8 rings plus double bonds. Note the much wider carbon distribution for resid compared to compared resid for distribution carbon wider much the Note bonds. double plus 8 rings with liquid.

5 0

20 Figure 3.5. Figure 30 25 15 10

Relative Abundance, % % Abundance, Abundance, Relative Relative

27 Resid Liquid

coal of for the O class bonds comparison

more saturated than the coal resid. resid. than the coal saturated more

double plus rings of Distribution 6

Rings plus Double Bonds for Class O much is liquid The coal liquid. and resid 3 Figure 3.6.

5 25 20 10 15

Relative Abundance (%)

Figure 3.8 shows another mass scale-expansion for ESI FT-ICR mass spectra of coal resid and liquid, this time near an even nominal mass of 430 Da. In each case, virtually all of the peak masses could be matched to within 1.0 ppm of the assigned elemental composition. In fact, analysis for all even-mass species reveals that compounds of class N and NO are either very low in abundance or nondetectable in coal liquid (due either to physical removal through distillation or chemical

modification). In contrast, classes NOx, x > 2, are present in both coal resid and liquid, and are evidently less affected by both physical and

chemical processes. Other classes such as O, O2, etc. can be found in low but detectable abundance in coal liquid.

CONCLUSION

Coupling of electrospray ionization to FT-ICR MS affords the selective ionization of polar heteroatomic compounds. Ultrahigh mass resolution and mass accuracy, amplified by Kendrick mass analysis, enables unique identification of thousands of distinct elemental compositions at different stages in the coal liquefaction process--in this case, coal resid and liquid. Comparison of coal resid and liquid according to compound classes, types, and carbon distributions reveals that coal resid contains more chemically distinct species of higher average molecular weight, as well as a wider and higher distribution in number of CH2 groups, and higher aromatic content than coal liquid. Because the properties of any liquefied fossil fuel ultimately depend on its complete chemical composition, such detailed characterization of compositional differences begin to provide the primary data upon which

50 future improvement of the coal liquefaction process may be based. For example, different catalysts may now be evaluated at the component-by- component level to judge their efficiency of conversion/removal of heteroatom containing species. ESI FT-ICR MS provides a simple, fast analysis with little sample preparation and no prior chromatographic separation. However, only the most basic (acidic) species will be ionized efficiently to produce positive (negative) ions. Thus, hydrocarbons, thiophenes, and other species are not readily accessible. Therefore, a complete compositional analysis of coal resid and liquid will require other ionization methods such as field desorption (recently adapted to FT-ICR mass analysis [87]) or photoionization (interface to FT-ICR MS currently under development in our laboratory). Even those methods preferentially access the aromatic fraction of hydrocarbons. Analysis of low-mass saturated hydrocarbons is currently best accessed by GC/MS.

381.5

(1)

2 381.4 O

(5x)

O(6)

(7)

381.3 2 mass Odd 381. m/z near liquid and resid O (in this case, zero) of nitrogen atoms. of nitrogen atoms. (in this case, zero)

m/z

O(12)

(15) 2 381.2 O (16)

O(21)

(17) 4 381.1 O

coal segments for scale-expanded Mass

number contain an even all species that requires Liquid Resid Figure 3.7.

C. C. O 12

C in place of C in place of

Liquid: of 15 out 16 Peaks Identified O

Resid: 23 out of 24 Peaks Identified

NO 2

is as fo Notation m/z 430. near d and liquid O 2 ain an odd number (in this case, ain an odd number one) of nitrogen

3 N 3 O m/z

4 NO 4 O species have no nitrogens but contain one nitrogens have no species n NO 2 2 O 5 NO

NO N 3 3 O

Mass scale-expanded segments for coal resi for coal segments Mass scale-expanded NO 4

atoms; thus, the remaining O Fig. 3.7. Even mass requires that all species cont that all species Fig. 3.7. Even mass requires 430.10 430.20 430.30 430.40 3.8. Figure

CHAPTER 4. COMPOSITIONAL DETERMINATION OF ACIDIC SPECIES IN ILLINOIS #6 COAL EXTRACTS BY ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

The presence of acidic constituents considerably increases corrosion in high-temperature distillation units in petroleum refining. The carboxylic acids of various heteroatomic classes (i.e., different numbers of N, O. and S atoms) in heavy immature petroleum crude oil have been identified, including naphthenic (alicyclic) acids, naphthenoaromatic acids, and aliphatic acids. [88-91] Those acids are believed to be a primary cause of corrosion in oil refining equipment. [92- 94] Severe corrosion has been observed in coal liquefaction equipment when acid fraction, basic fraction, and water-soluble chlorides are present simultaneously. [95] Moreover, carboxylic acids are major factors in retrograde reactions in the coal liquefaction process. [96-100] Acids are surface active (i.e., they lower the surface tension of the medium, and accumulate at interfaces) and water-soluble, and thus can readily propagate into the environment. [101] Therefore, a detailed chemical composition of acids in coal should enable a better understanding of the behavior of coal during liquefaction and processing, leading to rationally based improvements in those treatments. Crude oils are considered acidic if their total acid number (TAN) exceeds 0.5 mg KOH/g by non-aqueous titration. However, TAN does not reliably predict the corrosivity of crude oil, and does not identify the causative agents. Carboxylic acid can be isolated by simple base extraction, [102] or exposure to KOH impregnated silica gel. [92, 103, 104] Carboxylic acids and phenols have also been extracted

54 quantitatively with anionic surfactants [105, 106] Naphthenic acids in the polar fraction have been obtained by passage through a cyano- modified silica column after elution of nonpolar hydrocarbons. [107] Nonaqueous ion exchange-based methods have long been used to isolate acids. [88, 108-110] Isolated carboxylic acids from crude oil and coal have been analyzed by infrared spectroscopy, [92, 109], gas chromatography, [102, 104] and mass spectrometry. [89-91,109-112]. However, the compositional complexity of the carboxylic acids fraction has prevented detailed analysis. For example, some aliphatic acids can be resolved by GC/MS, but a large "unresolved complex mixture" in the chromatogram is not completely resolved, even when examined by high-resolution mass spectrometry. [110] Preparative thin-layer chromatography coupled with mass spectrometry has the same problem. [113] We previously reported that negative-ion ESI FT-ICR MS resolves more than 10,000 compositionally distinct compounds (enabling assignment of chemical formulas (CcHhNnOoSs) to more than 7,500 species) in a pyridine extract of Illinois #6 coal. [82] In that work, we used pyridine because it has the highest extraction efficiency for organics in coal. Here, by chromatographic isolation of two acid fractions, we are able to characterize the acids more completely than by pyridine extraction alone. The acids are concentrated in different fractions according to their degree of saturation, so that species of undetectable in the pyridine extract are rendered observable.

EXPERIMENTAL METHODS

Isolation of acid fractions. The methods for isolation of acid and acidic asphaltene fractions have been described previously by Qian. [46] Here we applied the method to coal samples with minor adjustments. Five grams of Illinois #6 coal (Argonne National Laboratory) was dissolved

55 in 40 mL toluene:methanol (70:30 v/v) and loaded onto amino-propyl silica (Sigma-Aldrich, Bellefonte, PA). After filtration and solvent washing, the silica with coal sample was dissolved in 40 mL toluene spiked with 30% acetic acid and then extracted in a sonicator water bath for 45 min. After filtration, the extract was water washed to remove residual acetic acids and then rotovapped to remove solvent. The dark yellowish residue was first dissolved in 30 mL hexane and passed through a silica column (Sigma-Aldrich, Bellefonte, PA). More hexane was added to wash the residue until the solvent was colorless. The hexane- soluble fraction is designated as the "acids fraction," and the hexane- insoluble fraction as "acidic asphaltenes" subsequently dissolved in toluene.

Sample preparation. Prior to analysis by ESI-FT-ICR MS, a solution of 10 mg of coal acid or acidic asphaltene sample was dissolved in 3 mL of pyridine and then diluted with 17 mL of methanol. One milliliter of the solution mixture was removed and spiked with 10 µL of pure (99.9%) ammonium hydroxide. Pyridine extraction of the same coal sample is described elsewhere. [82] All solvents were HPLC grade (Fisher Scientific, Pittsburgh, PA).

Mass analysis. Negative-ion ESI FT-ICR mass spectra were acquired with a homebuilt mass spectrometer equipped with a 22 cm diameter bore horizontal 9.4 T magnet (Oxford Corp., Oxney Mead, England), [2, 3, 53, 54] as described previously. [82] Time-domain data sets were processed and Hanning-apodized, followed by a single zero-fill before fast Fourier transformation and magnitude calculation. Frequency was converted to mass-to-charge ratio (m/z) by the quadrupolar electric trapping potential approximation [56, 57] to generate the m/z spectra.

56 RESULTS AND DISCUSSION

Figure 4.1 shows broadband mass spectra of the acid fraction, acidic asphaltenes fraction and pyridine extract of Illinois #6 coal. Because all detected ions are singly charged (based on the observed unit m/z spacing between chemically identical species containing 12Cc vs.

13C12Cc-1), [114] we shall henceforth denote each ion by its mass in Da rather than its mass-to-charge ratio, m/z. Negative-ion ESI FT-ICR MS resolves ~10,000 compositionally distinct compounds in acidic asphaltene fraction, ~5,000 in the acid fraction and ~10,000 in the pyridine extract. Virtually no species above 600 Da are found in the pyridine extract whereas numerous compounds of 600-750 Da are seen in both acidic fractions.

Homologous series and compound class. The coal acids mass spectrum (Figure 4.2) exhibits homologous series of saturated carboxylic acids differing by multiples of 14.01565 Da (i.e., CH2). It is worth noting that those species are not seen in mass spectrum of the pyridine extract under identical experimental conditions. Mass scale-expansion over a 1 Da interval at 619 Da (Figure 4.3) resolves 29 compositionally distinct compounds, of which 28 could be identified by matching to within 1.0 ppm of a putative elemental composition. The highest-mass peak corresponds to the fully saturated carboxylic acid, C31H23O2–. Other classes including O3 and O4 are also represented. Interestingly, the species with highest mass defect (i.e., highest saturation) are most abundant: in general, the coal acids fraction is the least aromatic of the three.

tene fraction;Bottom: Coal pyridine extract. Broadband negative-ion ESI FT-ICR mass spectra of Illinois #6 coal samples. coal #6 Illinois of spectra mass FT-ICR ESI negative-ion Broadband Top: Coal acids fraction; Middle: Acidic asphal Middle: Acidic fraction; Top: Coal acids Figure 4.1. Figure

700 ). ). 2

690

2 CH

680

670 CH

660 2

CH Homologous series of compounds differing in compounds of series Homologous 650 teristic mass spacings of 14.0157 Da mass (CH Da of 14.0157 spacings mass teristic m/z m/z 2

640 CH

Saturated Carboxyl Acids Carboxyl Saturated in Coal Acid Fraction 2

630

620 2

High-mass segment of the negative-ion ESI FT-ICR mass spectrum of saturated spectrum mass FT-ICR ESI of the negative-ion segment High-mass CH

610 degree of alkylation are evident from the charac evident from of alkylation are degree fraction. acids from the coal naphthenic Figure 4.2. 600

- 619.70 2

O 83

H 42

C - 2 619.60

O 71

H 43

C - 2

619.50 O ion ESI FT-ICR mass spectrum for species o spectrum for species mass FT-ICR ESI ion thenic acid is the most abundant species. thenic acid is the most abundant species.

59 H

44 m/z m/z C

619.40

- 3 O Saturated More 43 Higher Mass Defect Mass Defect Higher H

- 4 44 O C

619.30 39 H

of the negative- expansion scale Mass 43

C nominal mass, 619 Da. The fully saturated naph saturated The fully Da. 619 nominal mass,

Figure 4.3. Figure 619.20

Comparison of acid fractions and pyridine extract. At low mass, the pyridine extract yields more detectable compounds, as shown in Figure 4.4, mass segment at 437 Da: 24 resolved components in the pyridine extract, compared to 12 and nine for acidic asphaltenes and the coal acid fraction. However (see the species at 407 Da in Figure 4.5), some acids of low-abundance and/or low pyridine extraction efficiency are detected only in the coal acid fraction (C27H51O2– and C29H43O–) or only in the coal acids and acidic asphaltenes fraction (C28H39O2– and

C27H35O3–). C26H31O4– is not present in the coal acid fraction but is found in the acidic asphaltenes fraction and in the pyridine extract. Generally, the fractionation process isolates and concentrates acidic species from coal. Therefore, by concentrating acids in different coal fractions, we are able to recover and identify a more complete acid composition.

Compound type and carbon distribution. Figure 4.6 shows compound type distributions for the pyridine extract and coal acid fraction. The pyridine extract distribution was scaled relative to the summed abundances of all O3 species as 100% and the coal acids distribution was scaled relative to the summed abundances of all O2 species as 100%. (The acidic asphaltenes distribution (not shown) has a much higher relative abundance of O3 species (see below).) O3 and O2S classes are considered to be acids with oxygen or sulfur heterocycles.

The O3 and O2S distributions in both pyridine extract and coal acid fraction are roughly Gaussian (i.e., not bimodal), suggesting a single dominant aromatic core structure (rather than two or more non-fused cores) for those two classes. On the other hand, the O2 distribution is bi-

61 or tri-modal with up to 6 aromatic rings, suggesting the presence of at least two different core structures. [46] Note that naphthenic acids (various low-molecular-weight fatty acids believed to have cyclopentane ring mainly) and aromatic acids can have the same number of rings and double bonds and overlap in exact mass, and are thus unresolvable by mass measurement alone. Interestingly, the coal acid fraction also contains a high percentage of saturated aliphatic acids. The type (rings plus double bonds) distribution differs markedly from that for acids in a Chinese heavy crude oil. [49] Naphthenic acids are likely the most prevalent acids in Chinese crude; in coal, saturated straight-chain acids are dominant, and mono-aromatic acids are much more abundant than naphthenic acids. Thus, the composition of organic acids in coal differs qualitatively from that of crude oil. Sulfur-containing acids have low (but different) relative abundances in both coal acids and coal asphaltenes fractions. The major sulfur- containing acids classes (see Figure 4.7) found in the coal acids fraction are OS, O2S and O3S, whereas acidic asphaltene has mainly O2S, O3S and O4S. The compounds concentrated in coal acids fraction evidently have fewer heteroatoms (and simpler aromatic cores) than those concentrated in acidic asphaltenes fraction. Moreover, the DBE distribution shows that acidic asphaltenes extend to larger aromatic cores than do coal acids.

The carbon number distribution for O2 species with 19 rings and double bonds (Figure 4.8) in the acidic asphaltenes fraction is shifted to higher carbon number relative to that for the pyridine extract. Similarly, the carbon number distribution for the less aromatic components of the coal acid fraction (not shown) is wider than for compounds of the same double bond equivalents in the pyridine extract.

- 4 - O 2 37 O 9 peaks H 12 peaks 23 24 peaks 28 H C 31 C

- 3

O 29 H

30 itional complexity for the pyridine extract for the pyridine itional complexity C - on ESI FT-ICR mass spectrum for species o species for spectrum mass FT-ICR on ESI 4 O - 2 25 O H m/z 21 29 H C - 5 32 O C - 21 S H 3 28 O C

21 H 28

Mass scale expansion of the negative-i expansion scale Mass

Coal Acids nominal mass, 437 Da. Note the higher compos nominal mass, relative to the twofractions. acid Acidic Asphaltene Acidic

Pyridine Extract 437.06 437.10 437.14 437.18 437.22 437.26 Figure 4.4. Figure

- 2

H 407.50

43 - H 2 O 29 C

39 - H 3 28 of nominal spectra for species mass FT-ICR I O C

extract of Illinois #6 coal. Some acid pyridine 35

- H 4 O 27 tions, and the combined acid fractions contain m/z

31 C

26 C

407.10 407.20 407.30 407.40

of the negative-ion ES expansion Mass scale

Pyridine extract Coal Acids

Acidic Asphaltene more acid compounds than the pyridine extract alone. extract compounds than the pyridine acid more mass, 407 Da coal acids, acidic asphaltenes and asphaltenes acidic Da coal acids, mass, 407 frac only in the two acid detectable are species 407.00

Figure 4.5.

S 2 O

2 O

Extract Pyridine

25 20

O 15

DBE S 5 class is dominant in the pyridine extract whereas extract in the pyridine is dominant class Coal Acid Coal 2

3 0 2 4 6 O 8 10 Relative Abundance% Relative

20 15 10 5 DBE

0 2 4 6 8 Type distributions (rings plus double bonds, or double bond equivalents) for the coal acids the for equivalents) bond or double bonds, double plus (rings Type distributions 10 12 14 more are extract the pyridine in Compounds fraction. acid in the coal dominates class 16

Abundance% Relative unsaturated than in the acid extract. the O of Illinois #6 coal. The O extract pyridine fraction and

4.6. Figure

S 4 O S

3 O

S 2 O

Acidic Asphaltene Acidic OS

30 S 25

4 asses in coal acids and acidic asphaltenes acidic and in coal acids asses O

15 S tenes fraction is more aromatic and more aromatic is more tenes fraction 5 3 O DBE 0.1 0.2 0.3 0.4 0.5

S 2 O

Acid Coal % Abundance Relative

Type distributions for sulfur-containing cl sulfur-containing for distributions Type

0.1 from Illinois #6 coal. The acidic asphal The acidic #6 coal. Illinois from compositionally complex. complex. compositionally 0.2 0.3 0.4 DBE 0.5 Figure 4.7. % Abundance Relative

Acidic

Asphaltenes

d doublebonds, for acidic

38 42

fo distribution number #6 coal. The carbon

species with 19 rings an 2 Species with Species relative to that for the pyridine extract.

2 O 19 Rings plus Double Bonds

Number Carbon

26 30 34

Carbon number distributions for O for distributions number Carbon

Pyridine Extract

18 22 asphaltenes and the pyridine extract of Illinois extract of Illinois and the pyridine asphaltenes to higher mass is shifted asphaltenes acidic

4.8. Figure

0 2 5 4 3 1 9 6 8 7

10 Relative Abundance % Abundance Relative % Abundance Relative

67 CONCLUSION

Overall, all three extracts share similar major heteroatomic classes, but (for a given heteroatomic class and aromaticity) the combined acid fractions display wider distributions in carbon number and double bond equivalents than the pyridine extract. The less aromatic acids are enriched in the coal acid fraction, whereas the more aromatic acids are concentrated in the acidic asphaltenes fraction. For a given heteroatomic class, asphaltenes > pyridine extract > coal acids in aromaticity. Although the pyridine extract contains the widest variety of organics, additional acidic species are found in the other two extracts. In this paper, we extend our characterization of Illinois #6 coal by comparing a standard pyridine extract with two alternative fractions designed to concentrate acidic components: coal acids and acidic asphaltenes. The resulting detailed compositional analysis of coal acids provides detailed distributions of heteroatomic classes, aromaticity, and alkylation of coal. That kind of information establishes a fundamental basis for assessing the role of those acids in coal processing.

CHAPTER 5. TWO AND THREE DIMENSIONAL VAN KREVELEN DIAGRAMS: A GRAPHICAL ANALYSIS COMPLEMENTARY TO THE KENDRICK MASS PLOT FOR SORTING ELEMENTAL COMPOSITIONS OF COMPLEX ORGANIC MIXTURES BASED ON ULTRAHIGH- RESOLUTION BROADBAND FT-ICR MASS MEASUREMENTS

INTRODUCTION

Ultrahigh-resolution Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) makes it possible to resolve up to 10,000 different elemental compositions in a single complex organic mixture, such as petroleum crude oil [48] or coal. [82] For each spectral peak, ICR frequency is converted to mass-to-charge ratio, m/z, by a simple two- parameter equation requiring at least two known m/z values in the spectrum, [56, 57] preferably as internal standards (see below). The internal standards may be known compounds added to the sample, or may consist of an alkylation series of high mass defect common to samples of a given origin. The charge state may be determined from the

m/z spacing between ions differing in composition by 12Cn vs 12Cn-113C.

[114] At the present mass resolving power, m/∆m50% ≈ 300,000, in which

∆m50% is mass spectral peak full width at half-maximum peak height, it is possible to resolve close mass doublets (notably, elemental compositions differing by C3 vs. SH4, 0.0034 Da) up to ~900 Da. The corresponding mass accuracy (<0.001 Da) allows for assignment of a unique elemental composition, CcHhNnOoSs, to singly-charged ions up to ~300-400 Da. At higher mass, assignment based on mass measurement accuracy alone is no longer unique within experimental mass measurement error. To extend unique elemental composition assignment to higher mass, we introduced the "Kendrick plot," namely, a plot of mass defect (difference between exact mass and nominal mass) vs. nominal mass, in

69 which nominal mass is the mass rounded to the nearest integer value.

"Kendrick" mass is defined such that the mass of CH2 is 14.000000. The primary virtue of the Kendrick plot is that it allows for recognition and graphical resolution of various alkylation series, so that unique elemental compositions may be assigned to ions up to ~900 Da (see below). [44, 48, 82] Elemental compositions have now been determined by accurate mass measurement of volatile hydrocarbons (by low-energy electron ionization [37, 38, 40, 41]) and heteroatom-containing aromatics (by electrospray ionization). Electrospray ionization (ESI) coupled to high-field (9.4 T) FT- ICR MS has resolved and identified literally thousands of NSO-containing compounds as both positive and negative ions. [40, 41, 43, 44, 46-48] In fact, the new field of "petroleomics" [115] consists of connecting (and ultimately predicting) the properties and reactivity of petroleum and other fossil or natural organic mixtures based on their detailed chemical composition. Of course, the elemental composition itself reveals the compound "class" (i.e., numbers of N, O, and S atoms), "type" (number of rings plus double bonds, based on the "hydrogen deficiency", z, in the elemental composition expressed as CcH2c+zNnOoSs, because each additional ring or double bond reduces the number of hydrogens by two), and degree of alkylation (namely, the number of -CH2 groups for compounds of a given class and type). The remaining issue is how to sort those compositions in a way that best exposes differences between samples of different nature (e.g., humic substances, coal, petroleum, etc.), origin (e.g., geography, maturation, migration), and treatment (e.g., desulfurization, liquefaction) To that end, Kim et al. [85] recently extended an idea originally devised by van Krevelen, [116] namely, a plot of the molar ratio of hydrogen to carbon as the ordinate and oxygen to carbon ratio as the abscissa. The plot was originally based on bulk elemental analysis of

70 coal, to give one data point per sample; however, even at that stage it was clear that samples of different origin could be distinguished by the location of the data on such a plot. For example, an increase in oxidation shifts the data to the right; an increase in hydrogenation shifts the data upward; dehydration shifts to lower left; etc. [116] Since its introduction in 1950, [45] the van Krevelen diagram has been widely applied in geochemistry to characterize different fossil fuels. [117-121] Kim et al. recognized that ultrahigh-resolution mass spectrometry enables extension of van Krevelen's idea from mole ratio in a bulk sample to atomic ratio for each of the hundreds or thousands of elemental compositions present in a single sample, so that each distinct elemental composition contributes one data point to the van Krevelen plot. They applied the van Krevelen approach to analyze dissolved organic carbon from a black-water stream in the Pinelands of New Jersey. [85] Dissolved organic matter and humic/fulvic substances contain relatively little nitrogen. In contrast, the ESI FT-ICR positive/negative mass spectra of petroleum and coal and their byproducts are often dominated by nitrogen-containing species, because those substances contain numerous basic/"neutral" nitrogen compounds that protonate/deprotonate readily by positive-ion ESI. Here, we therefore extend the van Krevelen plot in two ways: (a) two-dimensional plot of H/C vs. N/C (rather than H/C vs. O/C) atomic ratios, and (b) a three- dimensional plot of H/C vs. N/C vs. O/C atomic ratios. The three- dimensional plot is especially attractive for two reasons. First, it cleanly separates graphically all of the classes for a given sample, and for each class provides a pattern of unsaturation and alkylation. Second, it resolves compositional differences between compounds of the same classes in samples of different nature, origin, and processing. We offer experimental examples chosen from various compound classes in coal; the same classes in coals of different rank or stage of liquefaction; and

71 the same classes for coal vs. petroleum crude oil. In each case, the differences between the samples are graphically resolved.

EXPERIMENTAL METHODS

Sample preparation. Pocahontas #3 and Illinois #6 coal samples were obtained from Argonne National Laboratory and have previously been analyzed by nuclear magnetic resonance (NMR), laser desorption mass spectrometry, fast atom bombardment tandem mass spectrometry, and ESI FT-ICR MS. Subbituminous coal from the Black Thunder Mine (Wyoming) and liquefied coal (produced from the Wyoming coal as a feedstock) samples were kindly provided by CONSOL Energy Inc. (Pittsburgh, PA). The South American heavy crude oil analyzed in this work is the same as that previously characterized by ESI FT-ICR MS. [46, 47] Sample preparation and ESI FT-ICR mass analysis are as described previously. [82]

Mass calibration and data reduction. Please see chapter 2.

RESULTS AND DISCUSSION

Elemental compositions from a plot of Kendrick mass defect vs. Kendrick nominal mass. Once FT-ICR mass spectral peaks are baseline-resolved and internally calibrated (from ICR frequency to mass- to-charge ratio), [56, 57] the masses are converted to Kendrick masses and sorted into groups according to common Kendrick mass defect. [45] In the present examples, all ions are singly-charged, as evident from the absence of 13C-containing species at fractional nominal m/z values. [114]) At our mass resolving power and mass accuracy, unique elemental composition may be assigned to singly-charged ions up to ~300-400 Da. At higher mass, assignment based on mass measurement accuracy alone is no longer unique--however, starting from one unique mass assignment

72 at low mass, the Kendrick plot quickly identifies the other members of a homologous alkylation series, so that mass assignments can extend with confidence to ~900 Da. [45] The Kendrick plot sorts compounds by their Kendrick mass defect and Kendrick mass, thereby producing rows of

data separated horizontally by number of alkyl (CH2) groups and vertically according to degree of saturation (number of rings and double bonds) and "class" (numbers of N, O and S atoms). More than 10,000 elemental compositions in a single oil or coal mass spectrum [48, 82] can be identified rapidly from those groupings, as shown for Pochahontas #3 coal in Figure 5.1.

Sorting of compound classes by use of a van Krevelen diagram. Figure 5.2 shows a van Krevelen diagram consisting of a plot of H/C vs.

O/C atomic ratio, for class O3 compounds in Pochahontas #3 coal. Data points corresponding to compounds in different alkylation series appear along lines that intersect at H/C = 2 on the y-axis (see Fig. 5.2). It's easy to see why. If the elemental composition, CcHhOo, loses (CH2)n, then its

composition becomes Cc-nHh-2nOo. The new H/C (y-value) and O/C (x- value) then become

y = (h-2n)/c-2n); x = o/(c-n) Elimination of n gives y = 2 + (h-2c)/o) x so that the y-intercept for each alkylation series is 2. Similar relations lead to simple rules for families of compounds differing by oxidation, dehydration, decarboxylation, saturation, etc. [85, 116]

O (14) O (17) O (16) O (15) O (18) O (20) O (19) O (21) 22) = O (DBE

#3 coal. Pocahontas nominal mass for

Kendrick NominalMass

#3 Coal Pocahontas

Kendrick defect vs. mass Kendrick .

ure 5.1 g Fi 250 300 350 400 450 500 550 600 650 Kendrick Mass Defect 450 250 500 400 350 300 200

Class O Class in Pocahontas #3

)

see text see

(

Number Increasing Of Ring & Double Bonds

C axis

appea alkylation of varying degrees to ponding

Coal

AtomicO/C Ratio

Homologous Alkylation Series

O of class for compounds diagram Krevelen van Two-dimensional

on the H of 2 at an atomic ratio intersect lines that 0 0.05 0.1 0.15 0.2 g 2 1 0 alon corres series in homologous coal. The compounds H/C AtomicH/C Ratio 1.6 1.2 0.8 0.4 0.2 1.8 1.4 0.6

5.2. Figure

75 For example, among compounds in the same class, those with same number of carbons have identical O/C values. For those species, as the number of rings and double bonds increases, the H/C value decreases to generate a series of data points falling on a single vertical line (see Fig. 5.2). We shall now apply van Krevelen analysis to various complex organic mixtures. Elemental compositions for each of three different compound

classes (O, NO2, and O3) in Wyoming subbituminous coal are represented in the van Krevelen diagram of Figure 5.3. Class O has only one oxygen,

and therefore has the lowest O/C ratios (< 0.05); members of the NO2

class have O/C ratios ranging from 0.05 - 0.1; and class O3 compounds have O/C ratios of 0.07 - 0.15. Unlike the Kendrick mass plot, the van Krevelen diagram graphically separates different classes. For compounds of the same class, the slope of the line for a given alkylation series increases with increasing aromaticity (e.g., compare the two dashed lines in Fig. 5.2). However, two alkylation series with same number of rings plus double bonds but from different classes have different slopes in this diagram (compare the DBE = 10 alkylation series for the three classes in Figure 5.3). Therefore, comparing the degree of saturation based on slope in a van Krevelen diagram is simple only among alkylation series from the same class. In the two-dimensional van Krevelen diagram of Figure 5.3, the various classes are graphically separated because they differ in number of oxygens. However, if two classes each have the same number of oxygens, their constituent O/C ratios will be the same, and their distributions will not be separated on the O/C axis (see Figure 5.4).

Neverthelss, because the NO2 class contains an odd number of nitrogens, it has an odd number of hydrogens, whereas the O2 class members have an even number of hydrogens. Thus, the two classes may be distinguished by their different H/C values (see Figure 5.4). However, members of classes O2S and O2 would completely overlap in the same 76 diagram. Therefore, the traditional two-dimensional van Krevelen diagram can clearly differentiate only classes with different number of oxygens. Alternatively, we may construct a two-dimensional van Krevelen plot in which the abscissa is based on N/C ratio rather than O/C ratio.

Figure 5.5 shows that such a plot successfully separates the N and N2 classes horizontally, based on their different N/C ratios. However, by the same token, that two-dimensional plot would fail to separate classes sharing the same number of nitrogens, e.g., N, NO, and NO2.

Three dimensional van Krevelen diagram. From Figures 5.4 and 5.5, it is obvious that all classes can be separated by extending the van Krevelen diagram from two to three dimensions, whose axes or H/C, N/C, and O/C atomic ratios. The three-dimensional van Krevelen diagram of Figure 5.6 exhibits dramatic graphical resolution between

members of the N, NO and NO2 classes. Although those three classes share the same number of nitrogens, they contain different numbers of oxygens, so that adding a third atomic ratio axis overcomes the limitations of two-dimensional van Krevelen displays. Conversely, if it is important to resolve different sulfur-containing classes (e.g. NS, N2S,

NS2), the third O/C axis may be replaced by S/C ratio. Finally, the advantage of the three-dimensional van Krevelen diagram over the two-dimensional Kendrick plot is the separation elemental compositions into different planes, thereby ensuring complete graphical resolution of different classes from each other. In the Kendrick plot, the higher the Kendrick mass defect, the more aromatic is the compound. Thus the vertical displacement (referring to Kendrick mass defect) reveals the compound type for a given alkylation series. Similarly, in the three-dimensional van Krevelen diagram, data points for the more aromatic compounds lie closer to the H/C axis.

) 3 , and 2

) 2

(O DBE=10

O Class

(NO DBE=10

2 showing members of three classes: O, NO O, classes: of three members showing on based thei separated graphically classes are

Liquefaction Stage Early

Wyoming Subbituminous Coal

NO Class AtomicO/C Ratio

Class O Class

diagram van Krevelen Two-dimensional

DBE =10 (O) 0 0.05 0.1 0.15 0.2 from Wyoming subbituminous coal. The three coal. subbituminous from Wyoming 3 different O/C ratios. O 2 1 0 Figure 5.3. 1.6 1.4 1.2 0.8 0.6 0.4 0.2 1.8

AtomicH/C Ratio

f 2

and NO

C 2

Class NO

Early Stage Liquefaction Stage Early different by their distinguished be still may t ram showing members of classes O members of classes showing ram

Wyoming Subbituminous Coal

AtomicO/C Ratio

O Class

diag Krevelen van Two-dimensional

0.02 0.04 0.06 0.08 0.1 0.12 0.14 2 1 0 2.5 1.5 0.5 from Wyoming subbituminous coal. Because these two classes have the same number o the same number have these two classes Because subbituminous coal. from Wyoming oxygen atoms, they have identical O/C ratios, bu ratios, O/C identical atoms, they have oxygen H/C ratios. H/C ratios.

AtomicH/C Ratio

5.4. Figure

classes are thereby graphicall are thereby classes

Liquefaction Stage Early

N/C AtomicN/C Ratio Wyoming Subbituminous Coal

C ratios. /

0 0.02 0.04 0.06 0.08 0.1 Two-dimensional van Krevelen diagram for Wyoming subbituminous coal, but this time plotted but this coal, for Wyoming subbituminous diagram van Krevelen Two-dimensional

different N due to their arated 1 p 0.4 H/C AtomicH/C Ratio 0.6 1.2 0.8 se with N/C ratio instead of O/C ratio as the abscissa. The N and N and The N the abscissa. as ratio of O/C ratio instead with N/C

5.5. Figure

O/C , from 2 0.10 0.08 0.06 0.04 0.02

0.00

0.020

0.030 N/C

Coal

0.040

0.050

of the classes, N, NO, and NO for members one heteroatomdefinition), (by and is thus shifted to display. letely separated in the three-dimensional

0.060

1.0 N

2 NO

0.8 NO

H/C 0.6 Three-dimensional van Krevelen diagram diagram Krevelen van Three-dimensional

0.4

a different plane. Different classes are thus comp a different plane. Different classes Pocahontas #3 coal. Each class differs by at least by at least differs class Each #3 coal. Pocahontas

5.6. Figure

81 Kim et al. previously proposed a different three-dimensional van Krevelen plot in which the three axes are H/C ratio, O/C ratio, and mass spectral ion relative abundance. [85, 116] However, the relative abundances of ions in a mass spectrum (in particular ions produced by electrospray ionization) do not accurately represent the relative abundances of their parent neutrals in the original sample. For positive- ion ESI, for example, the most basic compounds will yield disproportionately large ion abundances, because ionization arises from protonation of the neutrals. Here, we therefore consider a three- dimensional van Krevelen plot in which all three axes are atomic ratios.

Visual comparisons between various fossil fuels. To this point, we have applied the van Krevelen display to distinguish different compound classes in the same sample. With that capability, it becomes possible to distinguish graphically different samples based on their composition. Figure 5.7 shows a three-dimensional (H/C, N/C, O/C ratios as axes) van Krevelen diagram for two different fossil fuel samples: coal and crude oil. Because the organics in coal are more aromatic than those in crude oil, the coal data correspond to lower H/C ratios and are shifted away from the crude oil data. The organic constituents of any fossil fuel will exhibit a characteristic pattern of unsaturation, and those differences will be manifested in a van Krevelen plot as differences in H/C ratio. Thus, the three-dimensional van Krevelen diagram not only separates classes in the single fossil sample but also differentiates two different fossil sources from each other in one diagram as well. The same approach may be applied to compare fossil fuels of different origin/maturity, such as two coal samples from geographically different areas and different rank. Figure 5.8 is a three-dimensional van Krevelen diagram for members of the three most abundant shared compound classes from two coals: Pocahontas #3 and Illinois #6. Prior Kendrick analysis of these two coals [82] showed that the higher-rank Pocahontas 82 #3 coal is more aromatic and has fewer heteroatoms than the lower-rank Illinois #6 coal. The two coals are easily distinguished graphically in Figure 5.8, based mainly on the lower H/C ratios for the higher-rank Pochahontas #3 coal. Finally, we apply the three-dimensional van Krevelen analysis to another coal, before and after liquefaction. Coal liquefaction involves stages of hydrotreatment, in which heteroatoms are removed and molecular bonds are broken. Figure 5.9 shows dramatic graphical differences resulting from hydrotreatment: many heteroatom-containing species are removed completely, and those remaining are more saturated (as evident from their displacement to higher H/C ratios). Detailed compositional analysis (not shown) also reveals the formation of new species, due to chemical reactions during processing. It is also clear that among these three classes, members of the O class are harder to remove than members of the two nitrogen-containing classes (NO, NO3).

CONCLUSION

In this paper, we have tried to show that two- and especially three- dimensional van Krevelen plots offer two attractive advantages for compositional analysis of complex organic mixtures. First, the van Krevelen plot complements the Kendrick plot, by graphically exaggerating differences between different classes (numbers of N, O, and S atoms). Second, and more important, the van Krevelen plot affords a simple graphical basis for exposing compositional differences between samples of different nature, origin, and processing. One can imagine, for example, distinguishing between two different putative pollutants in an environmental sample, or determining whether oil pollution comes from a spill vs. natural seepage at the site. Alternatively, it may be possible to

83 correlate the reactivity of a given sample (e.g., corrosivity, refinability, etc.) with its van Krevelen graphical pattern. In summary, van Krevelen's original idea was a good one when it could be applied only to bulk composition. The idea is rendered much more generally useful, by extension to thousands of compositions simultaneously in a single sample.

) fo 2 O/C

0.12 0.10 0.08 0.06

0.04 0.02 0.00

0.02

N/C 0.03

(N, NO, and NO classes 0.04 same

Coal Crude Oil

0.05 are more the coal components Because ed). oil, the two fuels are readily distinguished readily are oil, the two fuels

0.06

1.50

fossil fuels: coal (blue) and crude oil (r and crude coal (blue) fossil fuels: N Three-dimensional van Krevelen diagram for the diagram Krevelen van Three-dimensional NO 1.00

H/C different two aromatic than are the constituents of crude the diagram. in graphically

5.7. Figure 0.50

, and

, NO 2

2.0

H/C 1.6 3

2 1.2

NO 0.8 2 Illinois #6 Coal #6 Illinois

NO 0.4

and Pocahontas #3. The higher rank Pocahontas and Pocahontas

(NO three classes of ram for compounds

lower rank and more saturated Illinois #6. Illinois #6. more saturated and lower rank 0.16

0.12

0.08 O/C

0.04 0.02 Three-dimensional van Krevelen diag Krevelen van Three-dimensional

0.04

0.06 0.08

N/C ) from two coals of different rank: Illinois #6 of different coals ) from two 2 O ratios and are H/C smaller have classes of all its shared #3 is more aromatic; thus, members the diagram from those of displaced in

5.8. Figure

#3 Coal Pocahontas

, r 3 undant classes (NO, NO classes undant faction. The number of nitrogen- faction. ers of the three most ab ers from liquefaction of coal. liquefaction. Moreover, members of all classes become of all classes members Moreover, liquefaction. en diagram provides simple graphical evidence fo evidence graphical simple provides diagram en coal at early and final stages of lique

odenitrogenation resulting odenitrogenation

Three-dimensional van Krevelen diagram for memb diagram Krevelen van Three-dimensional

and O) from Wyoming (Thunder mine) Wyoming (Thunder and O) from heteroatom removal and hydr heteroatom removal after dramatically decreases clearly containing species Krevel van three-dimensional more saturated. The 5.9. Figure

CHAPTER 6. COMPARATIVE COMPOSITIONAL ANALYSIS OF UNTREATED AND HYDROTREATED OIL BY ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

Heteroatomic (S, N and O containing) compounds are constituents of unprocessed fuels originating from petroleum crude oil and its

distillates. Those species degrade the environment by release of NOx and

SOx gases during combustion, [75] and lower the fuel stability during storage [73, 74, 122]. Furthermore, nitrogen-containing compounds can inhibit and poison the catalysts in the hydrodesulfurization process. [123-128] Catalytic hydrotreatment is a complex process that can involve: hydrodesulfurization, hydrodenitrogenation, hydrodeoxygenation, hydrocracking reactions, hydrogenation of aromatics, and etc.[129, 130] Two conventional catalysts used for catalytic hydrotreatment are cobalt-molybdenum (Co-Mo) and nickel- molybdenum (Ni-Mo) catalysts, supported on aluminum oxide. Although both types of catalysts can reduce both nitrogen and sulfur compounds, Ni-Mo is preferred for hydrodenitrogenation and Co-Mo is preferred for hydrodesulfurization. Here, we shall examine Ni-Mo catalysis. Although reduction of total nitrogen and sulfur content is observed in catalytically hydrotreated fuels, [130-134] much less is known about the detailed compositional changes of nitrogen- and sulfur-containing polycyclic aromatic hydrocarbons during hydrotreatment. Analysis of shale oils before and after hydrotreatment has shown that oil sulfur is more easily removed during catalytic hydrogenation than oil nitrogen. [129, 134] Hydrodenitrogenation has been suggested to occur by a nucleophilic substitution mechanism or a Hofmann elimination: hydrogenation of the aromatic ring followed by carbon-nitrogen bond scission, which occurs only in saturated rings. In contrast, 88 hydrodesulfurization involves direct carbon-sulfur bond scission that does not require saturation of the aromatic ring. [30, 76, 129] Hence, the development of better hydrodenitrogenation catalysts and technology requires identification of nitrogen-containing compounds that survive hydrotreatment. However, analysis/identification of such compounds is difficult due to the complexity of the hydrocarbon matrix. Time- consuming separation has been needed to isolate and concentrate those compounds prior to analysis. [109, 135-137] We have previously demonstrated that electrospray ionization (ESI) coupled to high-field (9.4 T) Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS) can resolve and identify elemental compositions (CcHhNnOoSs) of up to literally thousands of NSO-containing compounds (observed as positive or negative molecular ions from basic or acidic species, respectively) in crude oils and coal products. [43-49, 82, 115] Here we extend the ESI FT-ICR MS method to detailed elemental composition analysis of three catalytic hydrotreatments on the same untreated fuel sample. We are able to monitor heteroatomic compounds after the hydrotreatment process and compare the removal efficiency of each procedure according to each of the heteroatomic classes, number of rings plus double bonds, and carbon distribution. This effort lays the groundwork for better understanding of the fate of heteroatom- containing polar organics through hydrotreatment and therefore improvement/tuning of hydrotreatment conditions at the molecular level.

89 EXPERIMENTAL METHODS

The original fuel sample is a 1:1 mixture of LCO (light cycle oil) and RCO (refined chemical oil). The three products after hydrotreatment with a Ni-Mo catalyst are denoted EI-41, EI-42, and EI-43. For EI-41, the

conditions are: 540-600 psi partial H2 pressure, first pass at 690 °F and second pass at 725 °F (same hydrogen pressures), after which fuel was then cut by distillation between 180-270 °F. EI-42 was subjected to the same treatment, except that the sample was collected at the end of the run, as the reactor had begun to plug up with coke. EI-43 was treated as for EI-41 and EI-42, but the distillation cut was taken after only one pass at 690 °F. The untreated LCO/RCO mixture is black and heavy whereas the three hydrotreated distillates are light yellow.

Sample preparation. A solution of 10 mg of each sample was completely dissolved in 10 mL of toluene and then diluted with either 10 mL of methanol (for positive-ion ESI) or 10 mL of acetonitrile (for negative-ion ESI). One milliliter of the solution mixture was removed and spiked with either 10 µL of pure (99.9%) acetic acid (for positive-ion ESI)

or 10 µL of 99.9% NH4OH (for negative-ion ESI). All solvents were HPLC grade (Fisher Scientific, Pittsburgh, PA).

Mass analysis & Kendrick mass and data reduction. Please see chapter 2.

RESULTS AND DISCUSSION

Untreated fuel. Positive-ion and negative-ion ESI FT-ICR MS provide access to basic and acidic compound compositions in the untreated fuel 1:1 mixture of light cycle oil: refined chemical oil (Figure 6.1.) Because all detected ions are singly charged, we shall henceforth

90 denote each ion by its mass in Da rather than its mass -to-charge ratio, m/z. Three hydrotreated samples EI-41, 42 and 43 were also subjected to positive- and negative-ion ESI FT-ICR MS, but yield of negative ions was very low. Evidently any of the three hydrotreatments removes most of the acidic species. Therefore, we shall focus only on the fate of the basic constituents. The distribution of basic heteroatomic classes for the untreated fuel was determined by dividing the sum of the relative abundances of all species of a given class by the sum of all the relative abundances of species from all classes in the mass spectrum (Figure 6.2). Of the 9 observed major classes, most contain nitrogen. Only two sulfur-

containing classes: O4S, O5S are detected, both as sodiated (rather than protonated) ions. Other sulfur-containing species are either low in abundance in the original mixture, or have low ionization efficiency relative to pyridine-type heterocycles with highly basic nitrogen lone-pair electrons. [77]

Species removed or generated by hydrotreatment. Of the three hydrotreatments, the EI-41 process leaves the fewest heteroatomic compounds whereas the EI-43 process leaves the most, presumably due to more extensive reduction for EI-41 due to two passes through the catalyst vs. one pass for EI-43. For example, mass scale-expansion for species of 373 Da nominal (nearest-integer) mass (Figure 6.3) shows complete removal of all heteroatomic species by EI-41, whereas E-42 and E-43 leave 3 and 2 heteroatomic species, respectively. Another mass scale-expansion for species of 413 Da nominal mass is shown in Figure 6.4. EI-41 &42 treatments successfully remove most

+ of the heteroatomic compounds except for C24H38O4Na . Although such non-nitrogen compounds are in low abundance in the untreated sample, their relative abundances increase markedly after E-41 or E-42 treatment because virtually all other species have been removed. On the 91 other hand, EI-43 treatment not only leaves behind many more heteroatomic species that EI-41 or EI-42, but also generates several new

+ 13 + species (e.g., C28H33N2O and C27H34N3 C ) that were not present originally. Similar behavior is seen for species of 466 Da nominal mass (Figure 6.5); namely, EI-43 removes most of the heteroatomic compounds in the original sample but also produces several new species

+ + such as C31H36N3O and C32H40N3 . As would be expected from catalytic reduction, the newly formed compounds are less typically aromatic than the original compounds. Thus, the incomplete EI-43 hydrotreatment partially reduces some of the original heteroatomic aromatic compounds rather than removing them entirely. Because the E-41 two-stage hydrotreatment removes most of the heteroatomic species, we shall next focus on the other two treatments that resulted in less complete catalytic conversion.

Effect of single-stage hydrotreatment on aromaticity and carbon distribution. The effect of EI-43 (single-stage) hydrotreatment on the distribution of rings plus double bonds for each of the three most abundant heteroatomic classes (N3, N2O and N3O) is shown in Figure 6.6. Each distribution is determined by dividing the sum of the abundances of all species of a given class by the sum of the abundances of all identified species in the mass spectrum. For all three classes, EI- 43 treatment preferentially eliminates species of higher aromaticity (i.e., shifts the DBE values downward). After EI-43 treatment, the most abundant class is N3, with a narrow DBE distribution centered at 14 rings plus double bonds.

92 refined chemical oil. Non-basic refined chemical ion (bottom)ultrahigh-resolution mass basic compounds as positive ions (bottom). (bottom). positive ions as compounds basic ) mixture of light cycle oil and cycle oil of light ) mixture Broadband negative-ion (top) and positive- (top) and negative-ion Broadband spectraof an untreated1:1 (v/v ions (top) and as negative detected are compounds Figure 6.1.

O 3

3 Untreated Fuel: Basic Classes O 2 N

2 O

O 2 N

positive (i.e., detected as classes nine major basic heteroatomic Relative abundances of

Figure 6.1. of ions) in the untreated fuel sample

6.2. Figure

5 0 Relative Abundance (%) 25 20 15 10

373.

+ C 13 3 N

373.3 30

H + 24

O 2 C m/z talytic processes (see text). EI-41 removes text). (see processes talytic

N 29

Note the different selective (bottom). ments + H 2 -43 has high abundant species left. left. species abundant -43 has high 25 N C 373.2 21 H

+ O

N 17 373.1 H 26

and afte (top) untreated Da, for 373 mass, at nominal expansion scale Mass EI-41 EI-42 EI-43 Untreated

each of three different catalytic hydrotreat the three ca by each species of various removal EI all heteroatomic compounds whereas

6.3. Figure

Na 5 413.4

O + C 46 13 H

3 22 N C + 34 H Na

4 27 O C 413.3

38 H 24 m/z

+ 2 + 2

O + 2 ments (bottom). Note that the EI-43 process EI-43 Note that the (bottom). ments N O N 25 2 413.2

29 H N H

30 33 27 lly detected (see text). lly C H C +

28 O C 2 N 21

H 413.1

EI-41 EI-42 EI-43 Mass scale expansion at nominal mass, 413 Da, for untreated (top) and and afte (top) untreated for Da, 413 nominal mass, at expansion scale Mass Untreated

each of three different catalytic hydrotreat different catalytic of three each origina not species generates new 6.4. Figure

466.

+ 3 N EI-41 EI-42 EI-43

40 Untreated H

32 C + +

2 O 3 O + 3 3 N 466.3 N N 36 32 28 H H H 31 30 + 33

C C C C 13 m/z O 2 N 25

lytic hydrotreatments (bottom). H 32

466.2 C + + C O 3

13 2 N O 24

2 H N

32 21 C Mass scale expansion at nominal mass, 466 Da, for untreated (top) and (top) untreated 466 Da, for mass, at nominal expansion scale Mass H 31 C

466.1 after each of three different cata Figure 6.5. Figure

O 3

EI-43 Untreated 3

O 2

for each of its three most abundan 6.1,

20 Rings plus double bonds distributions before and after EI-43 catalytic EI-43 after and before distributions bonds double Rings plus 15 DBE

2 4 10 6 8 10 12 14 heteroatomic classes. Highly aromatic compounds are partially reduced to yield to reduced partially are compounds Highly aromatic classes. heteroatomic text). (see aromatic species of less abundance relative increased Figure of the sample of hydrotreatment Figure 6.6. 6.6. Figure % Abundance Relative

98 Figure 6.7 shows carbon distributions for N3 compounds having

DBE values of 14, 17 and 20. In the untreated sample, N3 species with 17 DBE are most abundant whereas species with 14 DBE are least

abundant (with 22-26 carbons). After single-stage hydrotreatment, N3

species with 17 and 20 DBE decrease significantly, whereas N3 species with 14 DBE (now with 21-33 carbons) dominate. The effective increase

in carbon number for N3 species with 14 DBE is likely due to the partial reduction of molecules of higher aromaticity (e.g., 17 DBE, with more rings and also higher alkylation) rather than alkylation of species of a given aromaticity. It is also possible that non-basic nitrogen compounds are converted into basic compounds during hydrotreatment; [126, 134] .i.e, some heteroatomic compounds are removed during catalysis but other, less aromatic compounds are also created, in accord with the two- step theory of hydrodenitrogenation: hydrogenation of the nitrogen- containing aromatic ring, followed by carbon-nitrogen bond scission after ring saturation. [30, 76, 131] The single pass at 690 °F may only partly saturate the aromatic rings or convert non-basic compounds into basic ones, but the second step of carbon-nitrogen bond scission is incomplete.

Effect of two-stage hydrotreatment on aromaticity and carbon distribution. EI-41 and 42 both have two passes through the same catalyst, at 690 °F and then at 725 °F. However, during EI-42 process, the reactor began to plug up with coke, thereby reducing catalytic efficiency, resulting in incomplete removal of many heteroatomic species relative to the unhampered E-41 process. Interestingly, the compounds remaining after EI-42 treatment actually have higher aromatic content than the untreated sample, as seen in Figure 6.8 for species from the most abundant heteroatomic class, N2O. Evidently the less aromatic species are more efficiently removed by the two-pass hydrotreatment.

99 Figure 6.9 shows carbon distribution of three selected classes after EI-42 treatment. The relative abundances of N2O carbon distributions with 13 and 25 DBE decrease whereas N2O carbon distribution with 30 DBE increases compared with the untreated sample. Again, it appears that the less aromatic compounds are the easiest to remove.

CONCLUSION

Detailed elemental composition comparisons of the type shown here provide a new and rational basis for optimizing parameters for hydrotreatment of commercial oils. Identification of species that increase or decrease in abundance, and (especially) which new species are produced, should greatly improve understanding of differential catalytic efficiency for removal and/or conversion of potentially all of the chemical constituents of an oil feedstock.

100

17 DBE ( ) 17 DBE 14 DBE ( ) 14 DBE ( ) 20 DBE

class class with 14, 17 and 20 class Components 3 3

EI-43

Untreated

N Basic

Carbon Number

Carbon distributions for members of the N the of members for distributions Carbon

20 22 24 26 28 30 32 34 36 38 20 22 24 26 28 30 32 34 36 38

0 8 6 4 2 0 3.0 2.0 1.0

hydrotreatment. EI-43 one-stage after and before bonds, double rings plus Relative Abundance (%) Abundance Relative

6.7. Figure

101

EI-42

Untreated

O class, before and before O class, 2

26 28 30 32 34 36

DBE 24

20 O ClassO After Compounds

2 Rings + Double Bonds for N EI-42 Hydrotreatment 18

of the N members for bonds distributions Rings plus double

14 16

hydrotreatment. EI-42 after two-stage 0 8 4 2 6 Figure 6.8.

14 18 16 12 10

Relative Abundance (%)

102

Untreated EI-42

30 DBE

O class with 13, 25 and 30 rings and 30 25 and 13, O class with 2

25 DBE

Number Carbon O CompoundsO After EI-42

13 DBE N the of members for distributions Carbon

20 25 30 35 40 45 50 E-42 hydrotreatment. after two-stage before and double bonds 3 1 6 5 4 2 0

Figure 6.9.

Relative Abundance (%)

103 CHAPTER 7. COMPOSITION OF EXPLOSIVES BY ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

Analytical techniques for analysis of explosives have typically focused on detection of active ingredient(s). For example, dogs have been trained for chemical detection of explosives as well as supplements for instrumental detection. [138] Thin layer chromatography (TLC) was one of the first techniques applied to explosive analysis[139, 140] and is still used today. Gas chromatography (GC) is usually coupled with chemiluminescence, [141, 142] mass spectrometric, electron capture, [143, 144] or flame ionization detection. [145] However, care is obviously needed in volatilizing (by heating) thermally unstable explosive materials for GC analysis. In contrast, HPLC[146-149] may be conducted at room temperature and is less likely to lead to chemical decomposition of the sample. Size exclusion chromatography (SEC), [150] ion chromatography (IC), [151]capillary electrophoresis (CE), [152] electron monochromator mass spectrometry, [153] and supercritical fluid chromatography (SFC) [154] have all been used for explosive analysis. Ion mobility spectrometry plays an important role in the detection of trace levels of nitro-organic explosives. [155, 156] However, mass spectrometry is probably the most informative method for identifying organic compounds in trace amounts. Consequently, GC/MS, [157] liquid chromatography-mass spectrometry (LC/MS), [158] and tandem mass spectrometry (MS/MS) [159] are now widely applied to organic explosives analysis. [160-165] Unfortunately, the low mass resolving power, m/∆m50% ≈ 500-2000 (in which m is ion mass and ∆m50% is the mass spectral peak full width at half-maximum peak height) of typical quadrupole mass analyzers (MS or MS/MS) typically limits compound identification to a few "target" active

104 ingredients of explosives. However, in addition to one or two primary explosive ingredients, commercial explosives may contain sulfur, woodmeal, dyes, engine oil, and trace amounts of ingredients characteristic of each manufacturer. [166] Determination of the source and environmental fate of explosives thus requires identification not only of the active agent(s) but also of the dozens of other compounds present in actual explosive products. Especially for forensic applications, it is important to identify an explosive from its post-blast residue. For example, GC/MS identification of the brand of European dynamites based on the relative abundances of different dinitrotoluene isomers has been attempted, [167] and similar attempts with smokeless powder based on HPLC coupled to electrospray ionization (ESI) quadrupole ion trap mass spectrometry have been reported. [168] However, the success of such approaches depends on constant relative abundances before and after explosion; moreover, the explosion tends to vaporize many of the components of the original explosive.

Here, we show that the ultrahigh resolving power (m/∆m50% > 200,000) and mass accuracy (<1 ppm) of electrospray ionization Fourier transform ion cyclotron resonance (ESI FT-ICR) mass spectrometry allow for a definitive identification of various species in military explosives before and after explosion. We examine chemically pure RDX, HDX, and TNT, as well as commercial military TNT, smokeless powder (active ingredient, nitroglycerin), Powermite (active ingredient, ammonium nitrate), and military C4 (active ingredient, RDX). Except for military C4, several characteristic active and non-active ingredients are recovered from the post-blast residue. For C4, the original ingredients are chemically modified, but still detectable.

105 EXPERIMENTAL METHODS

Sample Preparation. Three explosive compounds, 2,4,6- trinitrotoluene (TNT), cyclo-1,3,5-trimethylene-2,4,6-trinitramine (RDX), and cyclo-tetramethylene-tetranitramine (HMX) were purchased from Supelco-Sigma-Aldrich (Bellefonte, PA). Nitroglycerin is the active explosive in smokeless powder. Military samples for comparison before and after explosion included TNT, C4, powermite, and smokeless powder obtained from the Florida Department of Law Enforcement bomb squad. Each compound or military sample was dissolved in acetonitrile at 1000 µg/mL. Each soil sample after explosion was Soxhlet-extracted in acetonitrile for 24 h; the extract was then reduced by evaporation from 150 mL to ~1 mL. Prior to ESI-FT-ICR MS analysis, a solution of each explosive pure compound was prepared by dissolving 20 µL of the original solution in 80 µL of acetonitrile spiked with 0.5% ammonium acetate. For better electrospray of additives, the military explosives before and after explosions were prepared as above but with 0.5% ammonium hydroxide rather than ammonium acetate. Spectra were internally calibrated with respect to a mixture of fatty acids (Sigma- Aldrich, Bellefonte, PA) and HP mix (Agilent Palo Alto, CA), and elemental compositions were assigned from accurate mass measurements. All solvents were pesticide grade or better and obtained from Fisher Scientific (Pittsburgh, PA). Prior to placement of an explosive, about 100 g of soil was collected and placed in 125 mL widemouth amber jar to serve as a blank. Explosives were buried and a ~100-foot detonation cord attached. Soil samples (~100 g) were then collected at the post-detonation point where the soil appeared black and charred. Post-blast soil samples were treated as follows: Plant matter, rocks, sticks, etc. were removed. Soil was then ground in a mortar and pestle to give a uniform texture.

Fifteen grams of the ground soil was then mixed with 10 g NaSO4. The

106 solid mixture was placed into a cellulose extraction thimble for Soxhlet extraction. The extraction was carried out with 150 mL of acetonitrile for 24 h. After extraction, the sample was concentrated to 10 mL by evaporation. Samples were then further concentrated by evaporation under dry nitrogen to a volume of 1 mL. Unexploded explosives were directly dissolved in acetonitrile to give a concentration of ~1 mg/mL prior to analysis. Mass Analysis. Mass analysis was performed with a homebuilt FT-ICR mass spectrometer equipped with a 22 cm diameter bore horizontal 9.4 T magnet (Oxford Corp., Oxney Mead, England). [2] Data were collected and processed by a modular ICR data acquisition system (MIDAS). [53, 54] Negative ions were generated from a microelectrospray source equipped with a 50 µm i.d. fused silica micro ESI needle. Samples were infused at a flow rate of 600 nL/min. Typical ESI conditions were: needle voltage, -1.7 kV; tube lens, -390 V; and heated capillary current, 3 A. Ions were accumulated externally [3] in a linear octopole ion trap for 10 s and transferred through rf-only multipoles to a 10 cm diameter, 30 cm long open cylindrical Penning ion trap. Multipoles were operated at 2 MHz at a peak-to-peak rf amplitude of 170 V. Broadband frequency-sweep ("chirp") dipolar excitation[55] (70 kHz to 1.27 MHz at a sweep rate of 150 Hz/µs and peak-to-peak amplitude, 190 V) was followed by direct mode image current detection that yielded 4 Mword time-domain data. 2 Mword time-domain data sets were processed and Hanning-apodized, followed by a single zero-fill before fast Fourier transformation and magnitude calculation. Frequency was converted to mass-to-charge ratio (m/z) by the quadrupolar electric trapping potential approximation[56, 57] to generate the m/z spectra.

107

RESULTS AND DISCUSSION

TNT, RDX, and HMX. The electron-withdrawing nitro group(s) of explosives and their byproducts and metabolites stabilize their gas-phase anions. Thus, all spectra shown were run in negative ion mode. For example, the deprotonated molecular ion, [M-H]- is the "base" (i.e., most abundant) species in the negative-ion ESI FT-ICR mass spectrum of TNT (not shown). The dominant negative ions for RDX and HMX (see Figures 7.1 and 7.2) are [M+45]- and [M+59]-, for which the elemental composition assignments have been controversial. No molecular ions were observed for RDX or HMX. In 1994, they were assigned (correctly,

- - 12 as it turns out) to [M+HCO2] and [M+CH3CO2] (i.e., differing by CH2) by Casetta and Garofolo [169], then inexplicably reassigned in 1997 by

- - 14 Yinon and Yost as [M+NO2-H] and [M+NNO2-H] (differing by N), [170] and discussed but not assigned in 2000 by Asbury et al. [156] Here, the ultrahigh mass accuracy of FT-ICR MS unequivocally establishes the mass difference between the two most abundant RDX and HMX anions

14 as CH2 (14.0157 Da) rather than N (14.0031 Da), confirming the original Casetta and Garofolo assignment. Moreover, based on internal calibration with a mixture of fatty acids and an HP mixture, we are able to assign the two main RDX and HMX anionic species definitively as

- - [M+HCO2] and [M+CH3CO2] . The [M+45]- ion may be assigned as an

- HCO2 adduct from the self-decomposition of RDX, and [M+59]- is assigned as RDX adducted to a deprotonated hydroxy-oxirane (i.e., a three membered oxygen heterocycle with another oxygen on one carbon). In support of that assignment, it is worth noting that a proposed mechanism for decomposition of HMX and RDX includes the formation of

HONO and HCN. [171] HCN will react with H2O under basic conditions to

- produce HCOO and NH3. [172] Interestingly, we find that the same anions are produced from RDX and HMX when ammonium acetate is

108 replaced by ammonium hydroxide in the electrospray solution. The likely explanation is that the explosives were originally dissolved in

- acetonitrile (CH3CN), which could react with H2O to form CH3CO2 to serve as an adduct. Other species in the RDX and HMX negative-ion mass spectra were

- - - assigned as [RDX + Cl] , [HMX+Cl] , and [HMX +CH3CO2 + H2CO] . Each experimental mass matched that for the assigned elemental composition to within 0.0001 Da. Although no chlorine-containing compounds were added to the samples, the chloride adduct to RDX (and, at lower abundance, to HMX) could derive from chloride leached from the glass vials, or from a sample impurity. Chloride adducts are commonly encountered in negative ion electrospray even when no readily apparent source of chloride ions can be identified. A likely source is the glass vials in which the samples are stored. Military TNT. An explosion obviously produces numerous combustion products at the expense of species present in the original explosive mixture. Nevertheless, Figure 7.3 shows that many components of military TNT may be recovered post-blast and detected by ESI FT-ICR MS. All observed ions are singly-charged, as evident by the

12 ~1 Da separation between the signals from Cn (monoisotopic) and

12 13 C C1 nuclides of each chemically distinct ion. In the low-mass range, 200-300 Da (Fig. 7.3, left), the identified species include deprontonated TNT molecular ion, [TNT-H]-, and numerous other initially present constituents listed in Table 7.1. Note that the agreement to within ±0.1 mDa (<1 ppm) between the measured mass and the mass computed from the putative elemental composition constitutes unique identification of that composition in each case. Various other original components of the military TNT explosives in pre- and post-blast samples are seen in the higher mass range, 300-500 Da (Fig. 7.3, right), and some of their elemental compositions are listed in Table 7.1. Those species are not

109 observed in the soil blanks. According to the elemental composition assignments of the species observed in the high mass range, one of the most abundant compound classes has the same number of rings or double bonds, with peaks separated by 44.0262 Da (e.g., ethylene oxide

units, -CH2CH2O-). The neutral species are tentatively assigned as a polymer series that contains sulfur. That series and other high mass components (taken together) serve as a marker for TNT, because they appear to be unique to the military TNT spectra. The successful recovery and identification of species other than TNT itself augurs well for the prospect of identifying the origin of such an explosive from its characteristic constituents. Even in the post-blast residue, we are still able to detect trace amounts of additives as a standard to identify the origin of the explosive products. The elemental compositions of the explosive metabolites and additives were assigned at a mass accuracy of ±1 ppm. Smokeless Powder. Figure 7.4 shows mass spectra of smokeless powder before and after combustion. Because smokeless powder is a relatively low-power explosive, numerous high-mass species (representing non-active ingredients) remained after extensive combustion (see Fig. 7.4 and Table 7.2). The active ingredient,

nitroglycerin, yields an NO3 adduct peak at 288.991 Da. Interestingly, the nitroglycerin signal is more prominent in the post-blast residue than in the original smokeless powder. It is interesting to note that all of the sulfur-containing series (separated by -CH2CH2O-) found in military TNT, and a few more, are found in smokeless powder--the two explosives evidently share some (but not all) non-active ingredients. Thus, it appears feasible to distinguish the two explosives based on non-active (as well as active) ingredients.

110

ive. The mass difference between the two between the difference The mass ive. N (14.0031). 14 (14.0157 Da), not not Da), (14.0157 2 ESI FT-ICR mass spectrum of RDX explos ESI FT-ICR mass spectrum major peaks is CH Figure 7.1. Figure

111

- ]

- 450 COO]

CO 3 2

COO] 3

400 [HMX+H +CH [HMX+CH

- 350

[HMX+HCO are peaks major The two osive.

300

m/z m/z

[HMX+HCOO]

HMX 250

2 9.4 T ESI FT-ICR MS FT-ICR T ESI 9.4 2

O O

N N . - ]

2 200 N N CO 3

HMX expl of spectrum mass ESI FT-ICR

N N

150

and [HMX+CH N 2 N Figure 7.2. 7.2. Figure

2 O

112

m/z

- - S

NO 8 S

2 7 O O O 37 4 NO

33 H H 2 H 22 C 20

C O C N

4 2

H O 2 and after (bottom, plotted upside

TNT C

- O S

4 6 -

H S O 2 5 C 29 O

H 25

18 H

C 300 500 16 C

After Explosion

Before Explosion MS FT-ICR 9.4 T ESI

[TNT-H]

- ]

Low-mass (200-300 Da, left) and high-mass (300-500 Da, segments of right) the ESI FT-ICR mass

spectra of species found in military TNT before (top, plotted normally) normally) TNT before (top, plotted in military found of species spectra down) detonation. detonation. down) 200 300 [TNT-NH

7.3. Figure

113 Powermite. Mass spectra of Powermite before and after detonation are shown in Figure 7.5, and some common constituents are listed in Table 7.3. Again, there is a good match between pre- and post-

- blast spectra. The major peaks correspond to a series of [(NaNO3)nNO3] clusters, presumably from the active ingredient (nitrate). Similarly, Zhao

- et al. have previously observed HNO3 clusters [(HNO3)nNO3] .[173]

- Additional series of nitrogen-containing ingredients are probably NO2

adducts of a series separated by 30.0106 Da (CH2O) mass spacings-- again, apparently specific to Powermite. Note that the non-active ingredient species differ markedly from those for TNT and smokeless powder.

Military C4 Explosive. Figure 7.6 compares mass spectra of military C4 before and after explosion. The only matched species in the two spectra derive from the active ingredient, RDX: [RDX+NO3]-,

- [RDX+H2CO+HCO2] and [RDX+H2CO+CH3CO2]-. Those species were also observed in the mass spectrum of pure RDX, but are more abundant in the C4 spectrum, presumably due to the different solvent (ammonium acetate vs. ammonium hydroxide, which was verified by running pure RDX in ammonium hydroxide) and the presence of other species in the C4 sample. Because C4 is a high-power explosive, the original ingredients evidently combust and/or vaporize upon detonation. It is possible that characteristic components could be collected farther from the blast site.

114

NO O m/z 2

- S O N 9

2 O

O After Explosion 41 Before Explosion H 24

- C

S 8 O in military smokeless powder, before and before powder, in military smokeless

O 4 500 600 H

37 2 -

H S C Nitroglycerin 7 22 O O C 4 33 H

2 H

C 20 O 4 C H

400 2 -

OC S 4 6 H O 2 - 29 C S

5 H

O 18

25 C H 300 of species found mass spectra ESI FT-ICR

16 - 3 Smokeless Powder C

after combustion, plotted as in Figure 7.3. 7.3. as in Figure plotted combustion, after 9.4 T ESI FT-ICR MS FT-ICR T ESI 9.4

Figure 7.4.

NG+NO

115

m/z

1000

) After Explosion After

3 Before Explosion 3 NO und in military Powermite, before and und in military Powermite, 4

3 NaNO

NaNO

500

9.4 T ESI FT-ICR MS FT-ICR T ESI 9.4 Powermite (NH

NaNO

mass spectra of species fo ESI FT-ICR

7.3. in Figure as plotted explosion, after 7.5. Figure

100

116

m/z

600

ilitary C4, before and afte

Explosion After Before Explosion

C4 Military

MS FT-ICR T ESI 9.4 m of species found in a

- ] 2

- ]

2 CO

- ]

CO+HCO CO+CH spectr ESI FT-ICR mass 2 2

[RDX+NO

7.3. Figure as in plotted explosion, 200 400

7.6. Figure [RDX+H [RDX+H

117

CONCLUSION

We have demonstrated the potential of electrospray ionization combined with FT-ICR mass analysis for discrimination and possible identification of military explosives. The unique value of high-resolution mass analysis is that it can identify elemental compositions of both active and non-active ingredients of the explosives. For example, we were able to correct prior misassignments of the most abundant negative ion species in RDX and HMX. More important, characteristic ingredients can be detected from post-blast explosive residues of each explosive. Future work will focus on analysis of the same explosive product from different manufacturers to see if the origin of an explosive can be determined by mass analysis.

118 Table 7.1. Matched ions observed by ESI FT-ICR MS, before and after explosion of military TNT. The series of the sulfur-containing species appear to constitute a unique marker for this explosive product.

Formula Measured Theoretical Difference Identity Mass (Da) Mass (Da) (mDa) - - C7H3N2O6 210.9997 210.9996 +0.1 [TNT-NH2] - C7H2N3O6 223.9950 223.9949 +0.1 - - C7H4N3O6 226.0106 226.0105 +0.1 [TNT-H] - C6H2N3O7 227.9899 227.9898 +0.1 - C7H4N3O7 242.0056 242.0054 +0.2 - - C16H31O2 255.2330 255.2329 +0.1 [Hexadecanoic acid-H?] - C16H25O5S 329.1429 329.1428 +0.1 - C18H29O6S 373.1690 373.1690 0.0 - C20H33O7S 417.1953 417.1952 +0.1 - C22H37O8S 461.2215 461.2214 +0.1

Table 7.2. Matched ions observed by ESI FT-ICR MS, before and after explosion of military smokeless powder. Formula Measured Theoretical Difference Identity Mass (Da) Mass (Da) (mDa) - - C3H5N3O9 288.9911 288.9909 +0.2 [Nitroglycerin+NO3] - C16H25O5S 329.1429 329.1428 +0.1 - C18H29O6S 373.1690 373.1690 0.0 - C20H33O7S 417.1953 417.1952 +0.1 - C24H41O5S 441.2682 441.2680 +0.2 - C22H37O8S 461.2215 461.2214 +0.1 - C26H45O6S 485.2940 485.2942 -0.2 - C24H41O9S 505.2476 505.2476 0.0

119 Table 7.3. Matched ions observed by ESI FT-ICR MS, before and after explosion of military Powermite.

Peak # Assigned Measured Theoretical Difference Composition Mass (Da) Mass (Da) (mDa) - 1 (NaNO3)2NO3 231.9435 231.9035 0.0 - 2 C14H14NO6 292.0826 292.0826 0.0 - 3 (NaNO3)3NO3 316.9211 316.9212 -0.1 - 4 C15H14NO7 320.0774 320.0775 -0.1 - 5 C15H16NO7 322.0932 322.0932 0.0 - 6 C16H18NO8 352.1037 352.1037 0.0 - 7 C20H18NO6 368.1139 368.1139 0.0 - 8 C21H20NO7 398.1244 398.1245 -0.1 - 9 (NaNO3)4NO3 401.8987 401.8988 -0.1 - 10 C22H22NO8 428.1350 428.1351 -0.1 - 11 C23H24NO9 458.1455 458.1456 -0.1 - 12 (NaNO3)5NO3 486.8763 486.8764 -0.1 - 13 C24H26NO10 488.1561 488.1562 -0.1 - 14 (NaNO3)6NO3 571.8538 571.8540 -0.2 - 15 (NaNO3)7NO3 656.8314 656.8316 -0.2 - 16 (NaNO3)10NO3 826.7872 826.7869 +0.3

120

CHAPTER 8. CHARACTERIZATION OF VEGETABLE OILS: DETAILED COMPOSITIONAL FINGERPRINTS DERIVED FROM ELECTROSPRAY IONIZATION FOURIER TRANSFORM ION CYCLOTRON RESONANCE MASS SPECTROMETRY

INTRODUCTION

Once the adverse effects of saturated fats upon blood lipids were discovered, vegetable oils have largely supplanted highly saturated animal fats in many food preparations. The composition and hence health benefits of vegetable oil vary according to the vegetable from which the oil is extracted. Thus, the authenticity of vegetable oils is important from both commercial and health perspectives. For instance, there is financial incentive to adulterate relatively expensive olive oil with other, cheaper oil(s). Such adulteration can cause toxic oil syndrome and affect thousands of people. [174]For example, the so-called Spanish toxic oil syndrome, in which olive oil is adulterated by rapeseed oil that has been denatured with aniline for industrial use, has caused over 500 human deaths. [175] Thus, a rapid and accurate method is needed to detect such adulteration. Vegetable oil is a very complex mixture (thousands of chemically distinct components), with fatty acids and di- and triglyerides as major constituents, and sterols, alcohols, wax esters, etc. as minor constituents. Those components differ in presence and relative abundance in each vegetable oil, and can thus in principle be used to characterize and distinguish vegetable oils of different source and/or processing. In practice, the problem is that the mixtures are so chemically complex that it is not been possible to separate and identify (let alone quantitate) individual species. Most current methods for detecting oil adulteration are based on chromatographic analysis. Specifically, fatty acids composition is

121 traditionally used in the food industry as an indicator of purity. Fatty acids are typically first converted to fatty acid methyl esters to increase their volatility to enable gas-liquid chromatography (GLC) analysis with capillary columns. [176-179] Stable carbon isotope analysis is another important method in authenticity assessment, [175, 180] based on the premise that each vegetable oil has its own unique pattern of naturally occurring stable isotopes of carbon. Triglyceride composition has also been used to measure the quality and purity of vegetable oils. High- performance liquid chromatograph (HPLC) is widely accepted for analysis of triglycerides; [181-183]; high-temperature GLC is also commonly used. [184-187] Alternatively, non-aqueous reversed-phase liquid chromatography (RPLC) based on a silver ion HPLC system with flame ionization detection has been used to characterize triglycerides in natural oil. [188, 189] Chromatographic methods are currently widely used for the qualitative and quantitative analysis of sterols, which comprise a major portion of the unsaponifiable matter. [190, 191] GC-electron ionization mass spectrometry [192] and on-line LC-GC-flame ionization has detected different sterols in vegetable oils. [193] Tocopherols, natural antioxidants that stabilize oils, serve as biomarkers for different oils. HPLC coupled with UV detection can analyze tocopherols after saponification. [194] GC-isotope ratio MS can reportly identify volatile compounds in vegetable oils. [195, 196] The above-listed techniques typically require complicated and time-consuming isolation procedures. For example, GLC analysis requires that fatty acids first be converted to methyl esters; stable carbon isotope analysis requires combustion of the whole oil sample before analysis; techniques based on quantitative analysis of particular chemical fractions requires prior chromatographic separation to isolate triglycerides, sterols, tocopherols etc. Moreover, because none of these methods resolves individual chemical constituents, some kinds of adulteration can go undetected. For example, mixing oils of similar fatty 122 acids composition can defeat detection based on the fatty acids fraction; alternatively, adulteration with desterolized oils can foil analysis based on sterolic fraction determination. Fortunately, it is the very complexity of such natural mixtures that renders them identifiable, provided that one can resolve and identify each of the chemical components of the mixture, as previously demonstrated for identification of various arson accelerants based on their detailed chemical composition fingerprints. [197] Here we show that ultrahigh

mass resolving power (m/∆m50% >350,000, in which ∆m50% is mass spectral peak full width at half-maximum peak height) and mass accuracy (<1 ppm) of electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry (ESI FT-ICR MS) [1] allows fast, definite assignments of components having thousands of different elemental compositions in vegetable oil without any prior sample extraction, separation, or chemical derivatization. We resolve and identify components of canola, olive, and soybean oils. The oils may be distinguished according to the compound distributions within any of each of several chemical families (fatty acids, di- and triacylglycerols, tocopherols). The methods presented here are modeled after prior successful ESI FT-ICR MS analysis of tens of thousands of chemical components of petroleum crude oil and its distillates. [48, 115]

EXPERIMENTAL METHODS

Sample preparation. Commercial canola (Crisco®), olive (Wesson®), and soybean (Publix®) oils were purchased from Publix in Tallahassee, FL. Twenty mg of each was dissolved in 20 mL chloroform- methanol (1:1) solution. One mL of the solution was spiked with 10 µL of pure (99.9%) ammonium acetate for subsequent negative-ion ESI FT-ICR MS analysis. All solvents were HPLC grade (Fisher Scientific, Pittsburgh, PA).

123

Mass analysis & Mass calibration and data reduction. Please see chapter 2.

Vertical scaling of mass spectra. In the absence of a stable- isotopic internal standard for each of the hundreds of components in the mass spectra, one must decide how to scale the peak heights in mass spectra from different samples. Unlike most other kinds of spectra, the absolute signal magnitude for ions of a particular mass-to-charge ratio, m/z, in a mass spectrum depends not only upon the concentration of the corresponding neutral analyte in the original sample, but also on the presence of other chemical constituents in that sample. In positive-ion ESI, for example, ions are typically formed by protonation of neutrals in the original liquid sample; thus, the most basic compounds will be ionized most efficiently, and could effectively suppress the signal that would be obtained from less basic compounds. Therefore, if (as in the present examples) one seeks to determine the presence of an adulterant added to an analyte, one needs to identify signals from the adulterant that are not found in the analyte. One cannot rely on relative abundances of components that are common to the adulterant and analyte. On the other hand, if one is trying to determine the relative abundances of different fatty acids, then scaling is not a problem. In that case, one is interested only in the ratio of the signal magnitudes for different fatty acids.

124

RESULTS AND DISCUSSION

Negative-ion ESI FT-ICR MS Elemental compositions. Negative ion ESI FT-ICR MS resolves 3000-4000 compositionally distinct compounds (enabling unique

assignment of chemical formulas (CcHhNnOoSsPp) to most species) for each of the three vegetable oils (Fig. 8.1). Because all detected ions are singly charged (as evident from the unit m/z spacing between chemically

12 13 12 identical species containing Cc vs. C Cc-1, [114] we shall henceforth denote each ion by its mass in Da rather than its mass-to-charge ratio, m/z. In negative-ion electrospray ionization analytes typically deprotonate to become negatively charged. Thus, acidic species dominate the mass spectrum: e.g., carboxylic acids, alcohols, etc. Acidic heteroatomic classes. Once the elemental compositions of all components have been assigned, then it becomes possible to sort the species according to compound "class", i.e., the number(s) of heteroatoms: NnOoSsPp. The distribution of each heteroatomic class for vegetable oils is determined by dividing the sum of the relative abundances of all species of a given class by the sum of the relative abundances of all detectable species in the negative-ion mass spectrum (Figure 8.2). All three vegetable oils contain the same major classes, e.g. class N, including NO, NO2, N2O, etc. Canola and olive oils have almost identical compositional distributions and evidently contain highly similar acidic components. In contrast, soybean oil differs significantly from the

other two oils by its high relative abundance of O2 species (>50% vs. 11- 13%). For all three oils the oxygen-containing classes are dominant (>95% of the total abundance). Class P includes species with multiple oxygens, such as PO6, PO8, suggesting the presence of phosphate compounds in the vegetable oils.

125

650

Olive Oil Canola Oil Soybean Oil

550

m/z m/z

450

Negative-Ion ESI FT-ICR MS FT-ICR ESI Negative-Ion 350

o mass spectra negative-ion FT-ICR ionization electrospray Broadband

acidic components of canola oil (top), olive oil (middle), and soybean oil (bottom). and soybean oil (bottom). olive oil (middle), components of canola oil (top), acidic 250 8.1. Figure

126

-S -N

-P

10 O

9 O oils. for the three vegetable classes omic 8

7 O

5 O

Olive Oil 3 Soybean Oil

Canola Oil O Relative abundances of various ion heteroat ion various of Relative abundances

Acidic Species Detected by Negative-Ion ESI

0 4 8

4 0 8 16 12 0 16 12 60 40 20 8.2. Figure

127

Fatty acids. Fatty acids are a major component of vegetable oils. Figure 8.3 shows a mass spectral segment containing four fatty acids, having the same number of carbons (18) but different degree of

saturation. Let Cc:n denote a fatty acid with c carbons and n double bonds. The fatty acid relative abundances are diagnostic for each of the

three oils. Canola: C18:1>C18:2>C18:0>C18:3; Olive: C18:1>C18:2>C18:0 (no

C18:3); Soybean: C18:2>C18:1>C18:0>C18:3. Because these fatty acids have the same functional group and similar structure, their ionization efficiencies should be comparable. Therefore, the mass spectral relative abundances should reflect the relative concentrations in the original sample. Our mass spectral results match the fatty acid compositions of these three vegetable oils obtained by traditional GC and HPLC analyses, [178]but are obtained much faster and without the need for derivatization (into methyl esters) or prior wet chemical separation. It is worth noting that the most abundant species in canola and olive oils is C18:1 but in soybean oil is C18:2. Thus, the highest-magnitude peak in Figure 8.1 (from which the other peak heights are scaled) is from a different component in canola and olive oil (m/z 281) than in soybean oil (m/z 279), demonstrating one of the difficulties in trying to compare ion abundances for different samples. Finally, note that the signals at

13 12 even masses in Figure 8.3 represent the C Cc-1 forms of the

12 corresponding Cc species observed at odd masses.

128

285

284

18:0 18:0 18:0 283 C Fatty Acids Fatty C C

282 8.1, showing relative abundances o 8.1, showing

281 18:1 18:1 18:1 C C

C m/z

280

18:2

18:2 18:2 279 C C C

278

18:3 fatty acids in three vegetable oils. 18:3 277 18 C C Figure segments from scale-expanded Mass

Olive Oil Soybean Oil Soybean Canola Oil

C various

Figure 8.3.

129 Tocopherols. Tocopherols are natural antioxidants with a phenol functional group that can be deprotonated in negative-ion ESI. Figure 8.4 shows the mass spectral segment containing tocopherols in soybean oil. β,γ-tocopherol is the most abundant, whereas there is only a small amount of α-tocopherol. In the mass scale-expanded segment, it is clear that high mass resolution is needed to resolve multiple peaks at one nominal (nearest-integer) mass. Figure 8.5 shows that tocopherol composition also provides a biomarker to distinguish different vegetable oils, e.g., β,γ-tocopherol has a very high abundance in both canola and soybean oils relative to olive oil. Thus, just as soybean oil is easily distinguished from canola and olive oils based on fatty acid relative abundances, olive oil is easily distinguished from canola and soybean oils based on the relative abundance of β,γ-tocopherol.

POSITIVE-ION ESI FT-ICR MS Elemental compositions. Negative-ion ESI FT-ICR MS resolves and enables identification of ~3000 compositionally distinct acidic compounds in each of the three vegetable oils. Additional non-acidic compounds can be detected by positive-ion ESI to produce [M+H]+ or

+ [M+NH4] ions. Figure 8.6 shows mass scale-expanded segment of the positive-ion ESI FT-ICR mass spectrum of soybean oil at 657 Da nominal mass. The ultrahigh resolving power and mass accuracy of FT-ICR reveal multiple "isobaric" elemental compositions at the same nominal mass, thereby providing an extraordinarily detailed fingerprint for each oil. Triacylglycerols & diacylglycerols. In triacylglycerols, a fatty acid is esterified to each of the three hydroxyls of the glycerol backbone. In positive-ion ESI, a triacylglycerol can be protonated to produce a quasimolecular ion, [M+H]+. Figure 8.7 shows that for canola and soybean oils, C54:6 (54 total carbons in the three fatty acids with 6 double bonds) is most abundant, whereas C54:5 is the highest for olive oil.

130 In diacylglycerols, two fatty acids are esterified to two of the three glycerol hydroxyls. Their compositions are also characteristic for the three vegetable oils. In the mass scale-expanded segment in Figure 8.8,

C36:2 is most abundant for both canola and olive oils whereas C36:4 with 4 double bonds is highest for soybean oil. Both the di- and triacylglycerols thus exhibit distinctive patterns of relative abundances of species containing the same number of carbons.

Detection of intentional adulteration. The ESI FT-ICR mass spectral relative abundances of particular homologous components are characteristic of a particular vegetable oil. Thus, it should be possible to detect adulteration of one (e.g., expensive) vegetable oil by addition of another (e.g., inexpensive) oil. To test that idea, we mixed olive oil with different proportions of soybean oil (olive:soybean weight ratio: 1:1, 2:1, 3:1, 4:1 and 5:1). Negative-ion ESI FT-ICR mass spectra of those mixtures shows a marked increase in the relative abundance of β,γ- tocopherol relative to pure olive oil, readily detected even in the 5:1 mixture (not shown). Figure 8.9 shows that C18:3 fatty acid (m/z 277), absent in pure olive oil, is immediately apparent even at 5:1 olive:soybean ratio, with increasing relative abundance on increasing proportion of soybean adulterant whereas the peak appears when it is adulterated. Moreover, positive-ion ESI FT-ICR MS exposes dramatic changes in both triacylglycerol and diacylglycerol component abundances on addition of soybean oil to pure olive oil. Figure 8.10 shows that the most abundant triacylglycerol, C54:5, is the most abundant in pure olive oil whereas C54:6 increases rapidly on addition of soybean oil. Thus, the ratio C54:6/C54:5, is a sensitive indicator of the presence (and relative proportion) of soybean adulterant in olive oil. Similar effects are seen for the diacylglycerol relative abundances.

131

- 2 O

435 49 429.4 H

29 C -tocopherol) 429.3 430 α m/z (

429.2 425

429.1 negative-ion ESI FT-ICR mass spectrum o spectrum mass FT-ICR ESI negative-ion

- 420 2 O 47

m/z H

415 28 -tocopherol) C

, Negative-Ion ESI Negative-Ion

of three tocopherols. undances (

410

O -

2 405

O Tocopherols in Soybean Oil Soybean in Tocopherols

R1 R3 45

H Mass scale-expanded segment of the segment scale-expanded Mass 27 -tocopherol) O 400 C R2 δ H

(

soybean oil, showing relative ab oil, showing relative soybean

Figure 8.4. Figure

132

415.6

ily be distinguished from ily be distinguished

415.5 Olive Oil

Canola Oil Soybean Oil

-tocopherol)

abundance showing relative 8.1, Figure ( 415.4 - 2

- O m/z 9 47 O le oils: olive oil can read le oils: olive 415.3

H 31 28 H

C 20 C - 415.2

27 H

from segments scale-expanded Mass 415.1 19

-tocopherol in three vegetab , C

of of oils. and soybean canola

Figure 8.5.

133

77 657.62

H +

47 4

C mponents of soybean mponents O 77 H 43 + 2 C O

657.58 73 H + 46 ment for non-acidic co 5 C O Oil Soybean

73 m/z

H + tal compositions at the same nominal mass. at the same nominal mass. tal compositions 3

42 O

69 657.54

H + Positive-IonMS ESI FT-ICR

6 45

O C

69 + H P

4 41

O C 70

H Positive-ion ESI FT-ICR mass spectral seg mass spectral ESI FT-ICR Positive-ion 657.50 41

elemen multiple of resolution the Note oil. 8.6. Figure

134

Olive Oil

Canola Oil

890 Soybean Oil

positive-ion ESI FT-ICR mass spectrum o mass ESI FT-ICR positive-ion

885

Triacylglycerols m/z

54:5

C 880

54:6

C 54:7 C of the segments Mass scale-expanded

54:8 C 875 triacylglycerols. of various abundances showing relative oils, three vegetable Figure 8.7.

135

610

Olive Oil

Canola Oil Oil Soybean

36:1 7, but for diacylglycerols, providing yet diacylglycerols, 7, but for C 605

36:2

m/z

36:3 C

600

36:4 C

as in Figure 8. segments Mass spectral

36:5 C

oils. three vegetable fingerprint to distinguish another Diacylglycerols

8.8. Figure

136

100 80 60 40 20

282

278

274

Relative Abundance %

oil. oil and soybean ures of olive

Pure Soybean

Fatty acid distributions for mixt

18:2 C 18:3 Figure 8.9. Figure

Acids Fatty

Pure Olive

137

100 80 60 40

20 884

880

876

Relative Abundance %

Pure Soybean Oil

oil. soybean oil and of olive for mixtures distributions Triacylglyerol

54:5 C 54:6 54:7 C C

Figure 8.10. Pure Olive Oil Triacylglycerols

138

CONCLUSION

In this first analysis of vegetable oils by ESI FT-ICR MS, we report the complete chemical compositional characterization of canola, olive and soybean oils, for both negative and positive ions, without prior chromatographic separation. Coupling of FT-ICR MS to electrospray ionization affords the selective ionization of acidic (negative mode) and basic (positive mode) heteroatom-containing compounds. Ultrahigh mass resolution and mass accuracy enable unique identification of thousands of distinct elemental compositions in the three samples.

Fatty acids dominate the negative-ion mass spectra. C18 fatty acids (with various numbers of double bonds) show different patterns in these three oils, to yield a fingerprint to differentiate them from each other. Similarly, tocopherol composition serves to differentiate the oils based on negative-ion mass spectra (β,γ-tocopherol has a very high relative abundance in both canola and soybean oil but low relative abundance in olive oil). Triacylglycerols & diacylglycerols provide an independent fingerprint based on positive-ion mass spectra. We also demonstrate this technique for detection of intentional adulteration of olive oil with soybean oil. In this technique, it is essential to be able to resolve multiple elemental compositions at a single nominal mass for two reasons: (a) to assign the correct elemental composition to each component; and (b) to ensure that a given signal arises from only a single component, thereby eliminating the overlap and interferences that characterize all other chemical and spectroscopic analyses. Moreover, a "fingerprint" based on literally thousands of chemical constituents provides unprecendented

139 detail for correlating vegetable oil origin and/or adulteration with inherent chemical composition. Relative abundances within each of several chemical families (e.g., fatty acids, di- and triacylglycerols, tocopherols, etc.) offer multiple independent bases for comparisons. Limiting relative abundances to each chemical family avoids problems due to differences in ionization efficiency for different families). The present results introduce a new, rapid way to analyze and characterize vegetable oils, either pure or as mixtures. No prior chemical pretreatment, extraction, or chromatographic separation is required, so that sample preparation is minimal. Some obvious next extensions will be to compare compositional differences between different commercial brands of the same vegetable oil, and to compare raw vs. processed oils.

140

CHAPTER 9. LIMITATIONS AND CONCLUSION

This work has demonstrated the implementation and great potential of ESI-FT-ICR MS in the analysis of complex organic mixtures. However, there are always limitations in any technique. First, electrospray ionization selectively ionizes only polar heteroatomic compounds and fails to generate ions from non-polar species. Therefore, different ionization techniques are being developing in our group to gain access to non-polar components such as field desorption ionization, atomospheric photoionization and atomospheric chemical ionization. Second, we are not capable of quantitative analysis due to the following factors: (a) ESI ionization efficiency; (b) lack of standard compounds; (c) isomers; (d) extraction solvent limits. Quantitative data is possible only when two compounds have similar electrospray ionization efficiency in the sample. However, ionization efficiency varies widely among polar compounds. For instance, a test of model compounds shows that ionization efficiency of carboxylic acids is 1,000 times higher than that of neutral nitrogen compounds in negative ESI. [49] In addition, the presence of efficiently ionized neutrals in a complex mixture are likely to influence the ionization efficiency with each of other neutrals. If standard compounds with similar structure and functional groups to the target compounds in the mixture are obtainable, quantitative analysis is becoming possible. Unfortunately, such compounds are hard to find: e.g., we know little about most of the isomeric forms of the detectable compounds in petrochemical samples. Isomers are chemical species with same number and type of atoms but with different chemical bonding patterns. Mass spectrometry can provide only the elemental composition but cannot differentiate isomers. Two isomers appear in the mass spectrum at the same mass. Finally, solvent extraction efficiency is 141 another limit to samples like coal extract. Pyridine is chosen due to its high efficiency for extraction of organics from coal. But no solvent is able to extract organics equally from coal. So even if all the extractable organics have the same ionization efficiency, their relative abundance as ions in mass spectrum does not reflect their relative concentrations of the corresponding neutrals in original sample. Due to all these factors, quantitative analysis of complex mixtures by ESI-FT-ICR MS is hard to achieve. Nevertheless, this technique has the best advantages in qualitative characterization, and quantitation should be possible within members of a homologous chemical family. Selective ionization targets compounds of interest in a bulk mixture without implementation of chromatography. Ultrahigh resolving power enables resolution of thousands of peaks in one spectrum and accurate mass allows unambiguous assignment of elemental compositions. Such detailed compositional information enables characterization of complex organic mixtures.

142

REFERENCES

1. Marshall, A.G., C.L. Hendrickson, and G.S. Jackson, Fourier Transform Ion Cyclotron Resonance Mass Spectrometry: A Primer. Mass Spectrom. Rev., 1998. 17: p. 1-35.

2. Senko, M.W., et al., Electrospray Ionization FT-ICR Mass Spectrometry at 9.4 Tesla. Rapid Commun. Mass Spectrom., 1996. 10: p. 1824-1828.

3. Senko, M.W., et al., External Accumulation of Ions for Enhanced Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. J. Am. Soc. Mass Spectrom., 1997. 8: p. 970-976.

4. Berkowitz, N., An introduction to coal technology. Second ed. Vol. 1. 1994, San Diego: Academic press, Inc. 398.

5. Swaine, D.J., Modern methods in bituminous coal analysis: trace elements. Crit. Rev. Anal. Chem., 1985. 15: p. 315-346.

6. Lafleur, A.L. and Y. Nakagawa, Multimode Size Exclusion Chromatography with Poly(Divinylbenzene) N-Methylpyrrolidinone for the Characterization of Coal-Derived Mixtures. Fuel, 1989. 68(6): p. 741-752.

7. Chambrion, P., R. Bertau, and P. Ehrburger, Effect of Polar Components on the Physicochemical Properties of Coal-Tar. Fuel, 1995. 74(9): p. 1284-1290.

8. Herod, A.A., et al., Effect of LiBr addition to 1-methyl-2- pyrrolidinone in the size-exclusion chromatography of coal-derived materials. Energy & Fuels, 1998. 12(1): p. 174-182.

9. Lazaro, M.J., et al., Calibration of size exclusion chromatography in 1-methyl-2-pyrrolidinone for coal-derived materials using standards and mass spectrometry. Energy & Fuels, 1999. 13(6): p. 1212-1222.

10. Charlesworth, J.M., A Review of Sample Preparation Methods, and Analysis of Coal Derived Materials by Gas-Chromatography. Fuel

143 Processing Technology, 1987. 16(2): p. 99-162.

11. Lai, W.C. and C.S. Song, Temperature-Programmed Retention Indexes for Gc and Gc-Ms Analysis of Coal-Derived and Petroleum- Derived Liquid Fuels. Fuel, 1995. 74(10): p. 1436-1451.

12. Padlo, D.M., R.B. Subramanian, and E.L. Kugler, Hydrocarbon class analysis of coal-derived liquids using high performance liquid chromatography. Fuel Processing Technology, 1996. 49(1-3): p. 247-258.

13. Domin, M., et al., Large molecular mass materials in coal-derived liquids by Cf-252-plasma and matrix-assisted laser desorption mass spectrometry. Energy & Fuels, 1998. 12(3): p. 485-492.

14. Winans, R.E., Pyrolysis Fast-Atom-Bombardment Tandem Mass- Spectrometry Characterization of Coals. Journal of Analytical and Applied Pyrolysis, 1991. 20: p. 1-13.

15. Winans, R.E., Organic Structural Components within the Argonne Premium Coal Samples. Abstracts of Papers of the American Chemical Society, 1994. 207: p. 6-6.

16. Winans, R.E. and N.A. Tomczyk, Linkages between aromatic structures in the Argonne premium coal samples. Abstracts of Papers of the American Chemical Society, 1997. 213: p. 103-103.

17. Hunt, J.E., et al., Molecular characterization of hydrocarbons from soots in low temperature flames of biphenyl and Illinois No. 6 coal. Abstracts of Papers of the American Chemical Society, 2001. 221: p. U510-U510.

18. Botto, R., E. and D. Axelson, E., Solid State 13C Relaxation Studies of Coals. Preprints, Fuel Chem. Div., Am. Chem. Soc., 1988. 33(3): p. 50-57.

19. Muntean, J.V., L.M. Stock, and R.E. Botto, Improving the Reliability of Quantitative Solid State [13]C NMR Analysis of Coal. Energy & Fuels, 1988. 2(1): p. 108-110.

20. Wind, R.A., J. Antoni, and G.M. Gary, H NMR Zeeman Relaxation in Premium Coals and Other Coals of Various Rank. Fuel, 1989. 68(9): p. 1189-1197.

21. Pruski, M., L. Rosa, and B.C. Gerstein, Effect of Sample Spinning

144 on Detection of [13]C NMR in Coals. Energy & Fuels, 1990. 4(2): p. 160-165.

22. Guillen, M.D., C. Diaz, and C.G. Blanco, Characterization of coal tar pitches with different softening points by H-1 NMR - Role of the different kinds of protons in the thermal process. Fuel Processing Technology, 1998. 58(1): p. 1-15.

23. Cebolla, V.L., et al., Journal of Aoac International, 2000. 83(6): p. 1474-1479.

24. Fabianska, M.J., TLC study of the polar compounds present in extractable organic matter of different origin. Jpc-Journal of Planar Chromatography-Modern Tlc, 1998. 11(2): p. 108-113.

25. Wasserman, S.R., R.E. Winans, and R. McBeth, Iron species in argonne premium coal samples: An investigation using x-ray absorption spectroscopy. Energy & Fuels, 1996. 10(2): p. 392-400.

26. Evans, J.R., et al., Analysis of Eight Argonne Premium Coal Samples by X-ray Fluorescence Spectrometry. Energy & Fuels, 1990. 4(5): p. 440-442.

27. Wertz, D.L., X-ray Analysis of the Argonne Premium Coals. 1. Use of Absorption/Diffraction Methods. Energy & Fuels, 1990. 4(5): p. 442-447.

28. Meyers, R.A., Coal Structure. 1982, New York: Academic Press.

29. Winans, R.E., N.A. Tomczyk, and J.E. Hunt, Changes in nitrogen containing molecules during the coalification process. Abstracts of Papers of the American Chemical Society, 1996. 212: p. 7-7.

30. Prins, R., M. Jian, and M. Flechsenhar, Mechanism and kinetics of hydrodenitrogenation. Polyhedron, 1997. 16(18): p. 3235-3246.

31. Callejas, M.A. and M.T. Martinez, Hydroprocessing of a Maya residue. Intrinsic kinetics of sulfur-, nitrogen-, nickel-, and vanadium-removal reactions. Energy & Fuels, 1999. 13(3): p. 629- 636.

32. Veloski, R.A., R.J. Lynn, and R.F. Sprecher, Characterization of Nitrogen-Containing Species in Coal- and Petroleum-Derived Products by Ammonia Chemical Ionization High-Resoluton Mass Spectrometry. Energy Fuels, 1997. 11: p. 137-143.

145

33. Damste, J.S.S., et al., Organosulfur compounds in sulfur-rich Rasa coal. Energy & Fuels, 1999. 13(3): p. 728-738.

34. Neill, P.H., Y.J. Xia, and R.E. Winans, Identification of the Heteroatom Containing Compounds in the Benzene Methanol Extracts of the Argonne Premium Coal Samples. Abstracts of Papers of the American Chemical Society, 1989. 198: p. 23-23.

35. Winans, R.E., R.L. Mcbeth, and P.E. Melnikov, Abstracts of Papers of the American Chemical Society, 1993. 206: p. 57-Geoc.

36. Hsu, C.S., Z. Liang, and J.E. Campana, Hydrocarbon Characterization by Ultrahigh Resolution FTICRMS. Anal. Chem., 1994. 66: p. 850-855.

37. Guan, S., A.G. Marshall, and S.E. Scheppele, Resolution and Chemical Formula Identification of Aromatic Hydrocarbons Containing Sulfur, Nitrogen, and/or Oxygen in Crude Oil Distillates. Anal. Chem., 1996. 68: p. 46-71.

38. Rodgers, R.P., et al., Resolution, Elemental Composition, and Simultaneous Monitoring by Fourier Transform Ion Cyclotron Resonance Mass Spectrometry of Organosulfur Species Before and After Diesel Fuel Processing. Anal. Chem., 1998. 70: p. 4743-4750.

39. Rodgers, R.P., et al., 5.6 Tesla Fourier Transform Ion Cyclotron Resonance Mass Spectrometer for Analysis of Volatile Complex Mixtures. Rev. Sci. Instrum., 1998. 69: p. 2278-2284.

40. Rodgers, R.P., et al., Jet Fuel Chemical Composition, Weathering, and Identification as a Contaminant at a Remediation Site, Determined by Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Anal. Chem., 1999. 71: p. 5171-5176.

41. Rodgers, R.P., et al., Complete Compositional Monitoring of the Weathering of Transportation Fuels based on Elemental Compositions from Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Environ. Sci. Technol., 2000. 34: p. 1671- 1678.

42. Zhang, D. and J.B. Fenn, Electrospray mass spectrometry of fossil fuels. Int. J. Mass Spectrom., 2000. 194: p. 197-208.

43. Rodgers, R.P., et al., Molecular Characterization of Petroporphyrins

146 In Crude Oil by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Can. J. Chem., 2001. 79: p. 546-551.

44. Hughey, C.A., et al., Elemental Composition Analysis of Processed and Unprocessed Diesel Fuel by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Energy & Fuels, 2001. 15: p. 1186-1193.

45. Hughey, C.A., et al., Kendrick Mass Defect Spectroscopy: A Compact Visual Analysis for Ultrahigh-Resolution Broadband Mass Spectra. Anal. Chem., 2001. 73: p. 4676-4681.

46. Qian, K., et al., Resolution and Identification of Elemental Compositions for more than 3000 Crude Acids in Heavy Petroleum by Negative-Ion Microelectrospray High Field Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Energy & Fuels, 2001. 15: p. 1505-1511.

47. Qian, K., et al., Reading Chemical Fine Print: Resolution and Identification of 3000 Nitrogen Containing Aromatic Compounds from a Single Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrum of Heavy Petrolueum Crude Oil. Energy & Fuels, 2001. 15: p. 492-498.

48. Hughey, C.A., R.P. Rodgers, and A.G. Marshall, Resolution of 11,000 Compositionally Distinct Components in a Single Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrum of Crude Oil. Anal. Chem., 2002. 74: p. 4145-4149.

49. Hughey, C.A., et al., Identification of Acidic NSO Compounds in Crude Oils of Different Geochemical Origins by Negative Ion Electrospray Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Org. Geochem., 2002. 33: p. 743-759.

50. Stenson, A.C., et al., Ionization and Fragmentation of Humic Substances in Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Anal. Chem., 2002: p. 4397-4409.

51. Freitas, M.A., et al., High resolution Fourier transform ion cyclotron resonance mass spectrometry for the analysis of complex humic and fulvic acids. Abstracts of Papers of the American Chemical Society, 2001. 221: p. U465-U465.

147 52. Kujawinski, E.B., P.G. Hatcher, and M.A. Freitas, High-resolution Fourier transform ion cyclotron resonance mass spectrometry of humic and fulvic acids: Improvements and comparisons. Analytical Chemistry, 2002. 74(2): p. 413-419.

53. Senko, M.W., et al., A High-Performance Modular Data System for FT-ICR Mass Spectrometry. Rapid Commun. Mass Spectrom., 1996. 10: p. 1839-1844.

54. Blakney, G.T., et al. Further Improvements To The MIDAS Data Station For FT-ICR Mass Spectrometry. in Proc. 49th Amer. Soc. Mass Spectrom. Conf. on Mass Spectrom.&Allied Topics. 2001. Chicago,IL: Amer. Soc. Mass Spectrom.

55. Marshall, A.G. and D.C. Roe, Theroy of Fourier Transform Ion Cyclotron Resonance Mass Spectroscopy: Response To Frequencey-Sweep Excitation. J. Chem. Phys., 1980. 73: p. 1581- 1590.

56. Ledford, E.B., Jr., D.L. Rempel, and M.L. Gross, Space Charge Effects In Fourier Transform Mass Spectrometry. Mass Calibration. Anal. Chem., 1984. 56: p. 2744-2748.

57. Shi, S.D.-H., et al., Comparison and interconversion of the two most common frequency-to-mass calibration functions for Fourier transform ion cyclotron resonance mass spectrometry. Int. J. Mass Spectrom., 2000. 195/196: p. 591-598.

58. Hannis, J.C. and D.C. Muddiman, A Dual Electrospray Ionization Source Combined with Hexapole Accumulation to Achieve High Mass Accuracy of Biopolymers in Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. J. Am. Soc. Mass Spectrom., 2000. 11(10): p. 876-883.

59. Kendrick, E., A Mass Scale Based on CH2 = 14.0000 for High Resolution Mass Spectrometry of Organic Compounds. Anal. Chem., 1963. 35(13): p. 2146-2154.

60. Hsu, C.S., K. Qian, and Y.C. Chen, An Innovative Approach to Data Analysis In Hydrocarbon Characterization by On-line Liquid Chromatography-Mass Spectrometry. Anal. Chim. Acta, 1992. 264: p. 79-89.

61. Song, C.S. and H.H. Schobert, Non-fuel uses of coals and synthesis of chemicals and materials. Fuel, 1996. 75(6): p. 724-736.

148

62. Mukherjee, D.K. and P.B. Chowdhury, Catalytic Effect of Mineral Matter Constituents in a North Assam Coal on Hydrogenation. Fuel, 1976. 55(1): p. 4-8.

63. Marzec, A., et al., Coal Reactive Sites Involved in Hydrogen Transfer from an H-Donor Studied by Field-Ionization Mass- Spectrometry. Fuel Processing Technology, 1990. 26(1): p. 53-66.

64. Joseph, J.T., et al., Coal Maceral Chemistry .1. Liquefaction Behavior. Energy & Fuels, 1991. 5(5): p. 724-729.

65. Ferrance, J.P., R.P. Warzinski, and B. Bockrath, The effect of catalyst, hydrogen pressure, and solvent in coal liquefaction under simulated preheater conditions. Abstracts of Papers of the American Chemical Society, 1998. 216: p. U874-U874.

66. Kouzu, M., M. Onozaki, and S. Oi, Conceptual study of hydrogen donor solvent in the NEDOL coal liquefaction process. Kagaku Kogaku Ronbunshu, 2002. 28(2): p. 125-136.

67. Zhang, L., et al., Properties and liquefaction activities of ferrous sulfate based catalyst impregnated on two Chinese bituminous coals. Fuel, 2002. 81(7): p. 951-958.

68. Zielke, C.W., W.A. Rosenhoover, and E. Gorin, Direct Zinc-Chloride Hydrocracking of Sub-Bituminous Coal and Regeneration of Spent Melt. Advances in Chemistry Series, 1976(151): p. 153-165.

69. Boehm, F.G., R.D. Caron, and D.K. Banerjee, Coprocessing Technology Development in Canada. Energy & Fuels, 1989. 3(2): p. 116-119.

70. Benito, A.M., et al., Hydrotreating of Distillates from Spanish Coal- Liquefaction. Erdol & Kohle Erdgas Petrochemie, 1993. 46(2): p. 78-80.

71. Fernandez, I., et al., Application of Petroleum Processing Technology to the Upgrading of Coal Syncrude. Fuel, 1995. 74(1): p. 32-36.

72. Hu, H.Q., et al., Catalytic liquefaction of coal with highly dispersed Fe2S3 impregnated in-situ. Energy & Fuels, 2001. 15(4): p. 830- 834.

149 73. Worstell, J.H., S.R. Daniel, and G. Frauenhoff, Deposit formation in liquid fuels. 3. The effect of selected nitrogen compounds on diesel fuel. Fuel, 1981. 60(485-487).

74. Chmielowiec, J., P. Fischer, and C.M. Pyburn, Characterization of precursors which cause light instability in hydroprocessed gas oils. Fuel, 1987. 66: p. 1358-1363.

75. Wen, B., M. He, and C. Costello, Simultaneous Catalytic Removal of NOx, SOx, and CO from FCC Regenerator. Energy & Fuels, 2002. 16(5): p. 1048-1053.

76. Weller, K., et al., Homogeneous Models for Hydrodenitrogenation Catalysis. Polyhedron, 1997. 16(18): p. 3139-3163.

77. Kabe, T., A. Ishihara, and W. Qian, Hydrodesulfurization and hydrodenitrogenation. 1999, Tokyo: WILEY.

78. Snape, C.E., et al., Influence of Organic Coal Structure on Liquefaction Behavior - an Update with Emphasis on Low Severity Conditions. Fuel Processing Technology, 1990. 24: p. 119-125.

79. Yuan, Q.C., et al., Investigation of extracts of coal by supercritical extraction. Fuel, 1998. 77(11): p. 1237-1241.

80. Yu, L.E., L.M. Hildemann, and S. Niksa, Characteristics of nitrogen-containing aromatic compounds in coal tars during secondary pyrolysis. Fuel, 1999. 78(3): p. 377-385.

81. Erdmann, K., T. Mohan, and J.G. Verkade, HOLC and 31P NMR analysis of phenols in coal liquefaction oils. Energy & Fuels, 1996. 10: p. 378-385.

82. Wu, Z., et al., Resolution of 10,000 Compositionally Distinct Components in Polar Coal Extracts by Negative-Ion Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry. Energy & Fuels, 2003. 17: p. 946-953.

83. Robbins, G.A., S.D. Brandes, and R.A. Winschel, A Characterization and Evaluation of Coal Liquefaction Process Streams. 1997, CONSOL Inc.: Library.

84. Stenson, A.C., A.G. Marshall, and W.T. Cooper, Exact Masses and Chemical Formulas of Individual Suwannee River Fulvic Acids from Ultrahigh Resolution Electrospray Ionization Fourier Transform Ion

150 Cyclotron Resonance Mass Spectra. Anal. Chem., 2003. 75: p. 1275-1284.

85. Kim, S., R. Kramer, and P. Hatcher, Graphical Method for Analysis of Ultrahigh-Resolution Broadband Mass Spectra of Natural Organic Matter, the Van Krevelen Diagram. Anal. Chem., 2003. 75(20): p. 5336-5344.

86. McLafferty, F.W. and F. Turecek, Interpretation of Mass Spectra. 1993, Sausalito, CA: University Science Books. 371.

87. Schaub, T.M., et al., High-Resolution Field Desorption Ionization Fourier Transform Ion Cyclotron Resonance High-Resolution Mass Analysis of Non-Polar Molecules. Anal. Chem., 2003. 75: p. 2172- 2176.

88. Jones, D.M., et al., Determination of Naphthenic Acids in Crude Oils Using Non-Aqueous Ion Exchanges Solid-Phase Extraction. Anal. Chem., 2001. 73: p. 703-707.

89. Seifert, W. and R. Teeter, Identification of Polycyclic Naphthenic, Monoaromatic, and Diaromatic Crude Oil Carboxylic Acids1. Anal. Chem., 1970. 42(2): p. 180.

90. Seifert, W. and R. Teeter, Identification of Polycylic Aromatic and Heterocyclic Crude Oil Carboxylic Acids. Anal. Chem., 1970. 42(7): p. 750.

91. Dzidic, I., et al., Determination of Naphthenic Acids in California Crudes and Refinery Wastewaters by Fluoride-Ion Chemical Ionization Mass- Spectrometry. Anal. Chem., 1988. 60(13): p. 1318-1323.

92. Ramljak, Z., et al., Separation of Acids from Asphalts. Anal. Chem., 1977. 49(8): p. 1222-1225.

93. Miller, S., M. Gardiner, and C. Ward, In-Line Inspection Detects Early Cracking on Canadian Crude-Oil Line. Oil & Gas Journal, 1998. 96(39): p. 90.

94. Filby, R. and S. Olsen, A Comparison of Instrumental Neutron- Activation Analysis and Inductively-Coupled Plasma-Mass Spectrometry for Trace-Element Determination in Petroleum Geochemistry. Journal of Radioanalytical and Nuclear Chemistry- Articles, 1994. 180(2): p. 285-294.

151 95. Bagga, P.S., C.F. Harris, and K.L. Baumert, Corrosion of Fractionation Towers in Coal Liquefaction Plants. Energy Res. Abstr., 1984. 9(17): p. 38.

96. Manion, J., D. McMillen, and R. Malhotra, Decarboxylation and Coupling Reactions of Aromatic Acids under Coal-Liquefaction Conditions. Energy & Fuels, 1996. 10(3): p. 776-788.

97. Suuberg, E., D. Lee, and J. Larsen, Temperature-Dependence of Crosslinking Processes in Pyrolyzing Coals. Fuel, 1985. 64(12): p. 1668-1671.

98. Serio, M., et al., Kinetics of Volatile Product Evolution in Coal Pyrolysis - Experiment and Theory. Energy & Fuels, 1987. 1(2): p. 138-152.

99. Solomon, P., et al., Cross-Linking Reactions During Coal Conversion. Energy & Fuels, 1990. 4(1): p. 42-54.

100. Gozmen, B., et al., Direct Liquefaction of High-Sulfur Coals: Effects of the Catalyst, the Solvent, and the Mineral Matter. Energy & Fuels, 2002. 16(5): p. 1040-1047.

101. Lai, J.W.S., et al., Factors That Affect the Degradation of Naphthenic Acids in Oil Sands Wastewater by Indigenous Microbial Communities. Environmental Toxicology and Chemistry, 1996. 15(9): p. 1482-1491.

102. Barth, T., L. Moen, and C. Dyrkorn, Comparison of Acid Numbers and Carboxylic Acid Molecular Compositions in Biodegraded and Normal Crude Oils. Abstracts of Papers of the American Chemical Society, 1998. 3: p. 134-136.

103. Douglas, A. and P. TG, Rapid Separation of Fatty Acids From Fossil Lipids by Impregnated Adsorbent Thin-Layer Chromatography. J. Chromatog, 1969. 43(2): p. 241.

104. Jaffe, R., P. Albrecht, and J. Oudin, Carboxylic-Acids as Indicators of Oil Migration .2. Case of the Mahakam Delta, Indonesia. Geochimica Et Cosmochimica Acta, 1988. 52(11): p. 2599-2607.

105. Seifert, W., et al., Analysis of Crude Oil Carboxylic Acids after Conversion to Their Corresponding Hydrocarbons. Anal. Chem., 1969. 41: p. 1639-1646.

152 106. Seifert, W. and R. Teeter, Carboxylic Acids in a California Petroleum - Identification of Structure Type. Chemistry & Industry, 1969. 41: p. 1464.

107. Yepez, O., et al., Reaction Kinetics of Venezuelan Naphthenic Acids with High Surface Iron. Abstracts of Papers of the American Chemical Society, 1998. 3: p. 114-122.

108. Jewell, D., et al., Ion-Exchange, Coordination, and Adsorption Chromatographic Separation of Heavy-End Petroleum Distillates. Anal. Chem., 1972. 44(8): p. 1391-1395.

109. Green, J.B., et al., Separation of Liquid Fossil-Fuels into Acid, Base and Neutral Concentrates .1. An Improved Nonaqueous Ion- Exchange Method. Fuel, 1984. 63(9): p. 1290-1301.

110. Tomczyk, N., et al., On the Nature and Origin of Acidic Species in Petroleum. 1. Detailed Acid Type Distribution in a California Crude Oil. Energy & Fuels, 2001. 15(6): p. 1498-1504.

111. Schulten, H., A. Marzec, and S. Czajkowska, Mass-Spectrometric and Chemometric Hemometric Studies of Thermoplastic Properties of Coals .2. Field-Ionization Mass-Spectrometry of Coals. Energy & Fuels, 1992. 6(1): p. 103-108.

112. Rudzinski, W., L. Oehlers, and Y. Zhang, Tandem Mass Spectrometric Characterization of Commercial Naphthenic Acids and A Maya Crude Oil. Energy & Fuels, 2002. 16(5): p. 1178-1185.

113. Seifert, W. and R. Teeter, Preparative Thin-Layer Chromatography and High Resolution Mass Spectrometry of Crude Oil Carboxylic Acids. Anal. Chem., 1969. 41(6): p. 786.

114. Senko, M.W., S.C. Beu, and F.W. McLafferty, Automated Assignment of Charge States from Resolved Isotopic Peaks for Multiply Charged Ions. J. Am. Soc. Mass Spectrom., 1995. 6: p. 52- 56.

115. Marshall, A.G. and R.P. Rodgers, Petroleomics: The Next Grand Challenge for Chemical Analysis. Acc. Chem. Res., 2004. 37: p. 0000-0000.

116. van Krevelen, D.W., Graphical-Statistical Method for the Study of Structure and Reaction Processes of Coal. Fuel, 1950. 29: p. 269- 284.

153

117. Hatcher, P.G., et al., Solid-State C-13 Nuclear Magnetic-Resonance Studies of Coalified Gymnosperm Xylem Tissue from Australian Brown Coals. Org. Geochem., 1989. 14(2): p. 145-155.

118. Majid, A., B. Sparks, D, and J.A. Ripmeester, Characterization of Solvent-Insoluble Organic-Matter Isolated from Alberta Oil Sands. Fuel, 1991. 70(1): p. 78-83.

119. Curiale, J.A. and M.R. Gibling, Productivity Control on Oil-Shale Formation - Mae-Sot-Basin, Thailand. Org. Geochem., 1994. 21(1): p. 67-89.

120. Bostick, N.H. and T.A. Daws, Relationships Between Data from Rock-Eval Pyrolysis and Proximate, Ultimate, Petrographic, and Physical Analyses of 142 Diverse United States Coal Samples. Org. Geochem., 1994. 21(1): p. 35-49.

121. Walker, R., M. Mastalerz, and P. Padgett, Quality of Selected Coal Seams from Indiana: Implications for Carbonization. Int. J. Coal Geology., 2001. 47(3-4): p. 277-286.

122. Benyamna, A., et al., Upgrading of Distillate Fractions of Timahdit Moroccan Shale Oil Over a Sulfided NiO-MoO3/GAMMA-Al2O3 Catalyst. Fuel, 1991. 70(7): p. 845-848.

123. Albro, T., P. Dreifuss, and R. Wormsbecher, Quantitative- Determination of Sulfur-Compounds in FCC Gaslines by AED - A Study of the Effect of Catalyst Type and Catalytic Conditions on Sulfur Distribution. HRC-Journal of High Resolution Chromatography, 1993. 16(1): p. 13-17.

124. Dorbon, M., et al., Identification of Carbazoles and Benzocarbazoles in A Coker Gas Oil and Influence of Catalytic Hydrotreatment on Their Distribution. Fuel, 1984. 63(4): p. 565- 570.

125. Fathoni, A. and B. Batts, A Literature-Review of Fuel Stability Studies with A Particular Emphasis on Shale Oil. Energy & Fuels, 1992. 6(6): p. 682-693.

126. Nagai, M. and T. Kabe, Selectivity of Molybdenum Catalyst in Hydrodesulfuization, Hydrodenitrogenation, and Hydrodeoxygenation - Effect of Additives on Dibenzothiophene Hydrodesulfurization. Journal of Catalysis, 1983. 81(2): p. 440- 449. 154

127. Girgis, M. and B. Gates, Reactivities, Reaction Networks, and Kinetics in High-Pressure Catalytic Hydroprocessing. Industrial &Engineering Chemistry Research, 1991. 30(9): p. 2021-2058.

128. Furimsky, E. and F. Massoth, Deactivation of Hydroprocessing Catalysts. Catalysis Today, 1999. 52(4): p. 381-495.

129. Katzer, J.R. and R. Sivasubramanian, Process and Catalyst Needs for Hydrodenitrogenation. Catalysis Reviews-Science and Engineering, 1979. 20(2): p. 155-208.

130. Murti, S., et al., Detailed Characterization of Heteroatom- Containing Molecules in Light Distillates Derived from Tanito Harum Coal and its Hydrotreated Oil. Fuel, 2002. 81(17): p. 2241- 2248.

131. Chishti, H. and P. Williams, Aromatic and Hetero-Aromatic Compositional Changes During Catalytic Hydrotreatment of Shale Oil. Fuel, 1999. 78(15): p. 1805-1815.

132. Williams, P. and H. Chishti, Reaction of Nitrogen and Sulphur Compounds During Catalytic Hydrotreatment of Shale Oil. Fuel, 2001. 80(7): p. 957-963.

133. Laredo, G., et al., Nitrogen Compounds Characterization in Atmospheric Gas Oil and Light Cycle Oil from a Blend of Mexican Crudes. Fuel, 2002. 81(10): p. 1341-1350.

134. Bej, S., A. Dalai, and J. Adjaye, Comparison of Hydrodenitrogenation of Basic and Nonbasic Nitrogen Compounds Present in Oil Sands Derived Heavy Gas Oil. Energy & Fuels, 2001. 15(2): p. 377-383.

135. Schmitter, J., et al., Selective Isolation of Nitrogen Bases from Petroleum. Anal. Chem., 1983. 55(11): p. 1685-1688.

136. Snyder, L., B. Buell, and H. Howard, Nitrogen and Oxygen Compound Types in Petroleum - Total Analysis of a 700-850 Degrees F Distillate from a California Crude Oil. Anal. Chem., 1968. 40(8): p. 1303-1317.

137. Hsu, C. and K. Qian, High-Boiling Aromatic-Hydrocarbons Characterized by Liquid-Chromatography Thermospray Mass- Spectrometry. Energy & Fuels, 1993. 7(2): p. 268-272.

155

138. Kenneth, G., Furton; Lawrence, J, Myers, The Scientific Foundation and Efficacy of the Use of Canines as Chemical Detectors for Explosives. Talanta, 2001. 54: p. 487-500.

139. Jenkins, R.Y., H,J, The Identification of Explosives in Trace Quantities on Objects Near an Explosion. Explosivestoffe, 1970. 18: p. 139-141.

140. Parker, R., G; McOwen,j ,M; Cherolis, J, A, Analysis of Explosives and Explosives Residues, Part 2: Thin Layer Chromatography. Journal of Forensic Science, 1975. 20: p. 254-256.

141. Douse, J., M, F, Trace Analysis of Explosives at the Low Picogram Level Using Silica Capillary Column Gas Chromatography with Thermal Energy Analyser Detection. Journal of Chromatography, 1983. 256: p. 359-362.

142. Fine, D., H; Yu,W,C; Goff, U,E; Bender,E,C; Reutter,D,J, Picogram Analysis of Explosive Residues Using the Thermal Energy Analyzer(TEA). Journal of Forensic Sciences, 1984. 29: p. 732-746.

143. Douse, J., M,F, Trace Analysis of Explosives at the Low Picogram Level by Silica Capillary Column Gas-liquid Chromatography with Electron-capture Detection. Journal of Chromatography, 1981. 208: p. 83-88.

144. Andrasko, J., Characterization of Smokeless Powerder Flakes from Fired Cartridge Cases and from Discharge Patterns on Clothing. Journal of Forensic Sciences, 1992. 37: p. 1030-1047.

145. Sekavja, C., M; Strobel,R,A; Tontarski, R,E. The Systematic Identification of Smokeless powerders: An Update. in Proceedings of The Third Symposium on Analysis and Detection of Explosives. 1989.

146. Lyter, A., H, A High Performance Liquid Chromatographic(HPLC) Study of Seven Common Explosive Materials. Journal of forensic Sciences, 1983. 28: p. 446-450.

147. Lloyd, J., B,F, High-performance Liquid Chromatography of Organic Explosives components with Electrochemical Detection at A Pendant Mercury Drop Electrode. Journal of Chromatography, 1983. 257: p. 227-236.

148. King, R., M, The Work of the Explosives & Gunshot Residues Unit of 156 the Forensic Science Service(UK). Advances in the Analysis and Detection of Explosives. 1993. 91-100.

149. Wallace, C., L; Midkiff,C,R, Smokeless powder Characterization, an Investigative Tool in Pipe Bombings. Advances in the Analysis and Detection of Explosives. 1993. 29-39.

150. Lloyd, J., B,F, Detection and Differentiation of Nitrocellulose Traces of Forensic Science Interest with Resiductive Mode Electrochemical Detection at a Pendant Mercury Drop Electrode Coupled with Size Exclusion Chromatography. Analytical Chemistry, 1984. 56: p. 1907-1912.

151. Reutter, D., J; Buechele,R,C; Rudolph,T,L, Ion Chromatography in Bombing Investigations. Analytical Chemistry, 1983. 55: p. 1468A- 1472A.

152. Hargadon, K., A; McCord,B,R, Explosive Residue Analysis by Capillary Electrophoresis and Ion Chromatography. Journal of Chromatography, 1992. 602: p. 241-247.

153. Laramee, J.A., et al., Electron monochromator mass spectrometer instrument for negative ion analysis of electronegative compounds. Mass Spectrom. Rev., 1996. 15(1): p. 15-42.

154. Francis, E., S; Eatough, D,J; Lee, M,L, Capillary Supercritical Fluid Chromatography with Nitro- and Nitroso-specific Chemiluminescence Detection. Journal of Microcolumn Separations, 1994. 6: p. 395-401.

155. Ewing, R., G; Atkinson, D,A; Eiceman,G,A; Ewing, G,J, A Critical Review of Ion Moblity Spectrometry for the Detection of Explosives and Explosive Related Compounds. Talanta, 2001. 54: p. 515-529. 156. Asbury, G., R; Klasmeier,J; Hill,H,H,Jr, Analysis of Explosives using Electrospray Ionization/Ion Mobility Spectrometry(ESI/IMS). Talanta, 2000. 50: p. 1291-1298.

157. Karasek, F., W; Clement,R,e, Basic Gas Chromatography-Mass Spectrometry. Principles and Techniques. 1988.

158. Yergev, a., L; Edmonds,C,G; Lewis,I,A,S; Vestal,M,L, Liquid Chromatography/Mass Spectrometry. Techniques and Applications. 1990.

159. Busch, K., L; Glish,G,L; McLuckey,S,A, Mass Spectrometry/Mass

157 Spectrometry: Techniques and Applications of Tandem Mass Spectrometry. 1988, New York: VCH.

160. Burn, D., T; Lewis,R,J, Analysis and Characterisation of Nitroglycerine-Based Explosives by Gas Chromatography-Mass Spectrometry. Analytica Chimica Acta, 1995. 307: p. 89-95.

161. Barshick, S., A; Griest,W,H, Trace Analysis of Explosives in Seawater Using Solid-Phase Microextraction and Gas Chromatography/Ion Trap Mass Spectrometry. Analytical Chemistry, 1998. 70: p. 3015-3020.

162. Garofolo, F.L., A; Migliozzi, V; Tallarico, C, Quantitative analysis of Thermostable Explosive Compounds by Combined Liquid Chromatography/Tandem Mass Spectromtry. Rapid Communications in Mass Spectrometry, 1996. 10: p. 1273-1277.

163. Beberich, D., B; Yost,R,A; Fetterole,D,D, Analysis of Explosives by Liquid Chromatography/Thermospray/Mass Spectrometry. Journal of Forensic Sciences, 1988. 33: p. 946-959.

164. Voyksner, R., D; Yinon,J, Trace Analysis of Explosives by Thermospray High-performance Liquid Chromatography-mass Spectrometry. Journal of Chromatography, 1986. 354: p. 393-405.

165. Hobbs, J., R; Conde,E. Techniques for the Detection of Explosives. in Proceedings 3rd International symposium on Analysis and Detection of Explosives. 1989.

166. Hoffman, C., M: Byall,E,B, Identification of Explosive Residues in Bomb Scene Investigations. Journal of Forensic Science, 1973: p. 54-62.

167. Kolla, P.S., A, Identification of Dynamite Explosives in Post Explosion. Journal of Forensic Science, 1995. 40(3): p. 406-411.

168. Mathis, J., A: McCord,B,R. Smokeless powder Comparsion Using HPLC/ESI-ITMS. in 49th ASMS. 2001. Chicago,IL.

169. Casetta, B. and F. Garofolo, Characterization of Explosives by Liquid Chromatography/Mass spectrometry and Liquid Chromatography/Tandem Mass Spectrometry Using Electrospray Ionization and Parent-Ion Scanning Techniques. Org. Mass Spectr., 1994. 29: p. 517-2-525.

158 170. Yinon, J., A, J.E. McClellan, and A. Yost, Electrospray Ionization Tandem Mass Spectrometry Collision-induced Dissociation Study of Explosives in an Ion Trap Mass Spectrometer. Rapid Communications in Mass Spectrometry, 1997. 11: p. 1961-1970.

171. Chakraborty, D., et al., Mechanism for Unimolecular Decompostion of HMX an ab Initio Study. J. Phys. Chem. A, 2001. 105: p. 1302- 1314.

172. Morrison, R., T and R. Boyd, N, Morrsion and Boyd Organic Chemistry, ed. t. Edition. 72.

173. Zhao, X. and J. Yinon. Forensic Identification of ammonium Nitrate by Electrospray Ionization-Tandem Mass Spectrometry(ESI- MS/MS). in 49th ASMS. 2001. Chicago: Amer. Soc. Mass Spectrom.

174. Abaitua, I. and M. Posada, Toxic Oil Syndrome: Current Knowledge and Future Perspectives. European series No.42. 1992, Copenhagen: WHO Regional publication. 27.

175. Mottram, H., R., et al., High-Resolution Detection of Adulteration of Maize Oil Using Multi-Component Compound-Specific 13C Values of Major Components and Discriminant Analysis. Rapid Commun. Mass Spectrom., 2003. 17: p. 706-712.

176. Browse, J., P.J. Mccourt, and C.R. Somerville, Fatty-Acid Composition of Lea Lipids Determined after Combined Digestion and Fatty-Acids Methyl-Ester Formation from Fresh Tissue. Anal. Biochem., 1986. 152(1): p. 141-145.

177. Garces, R. and M. Mancha, One-Step Lipid Extraction and Fatty- Acid Methyl-Esters Preparation from Fresh Plant-Tissues. Anal. Biochem., 1993. 211(1): p. 139-143.

178. Jeyashoke, N., K. Krisnangkura, and S.T. Chen, Microwave Induced Rapid Transmethylation of Fatty Acids for Analysis of Food Oil. J. Chromatogr. A, 1998. 818(1): p. 133-137.

179. Kittiratanapiboon, K., N. Jeyashoke, and K. Krisnangkura, Forecasting Retention Times of Fatty Acid Methyl Esters in Temperature-Programmed Gas Chromatography. J. Chromatogr. Sci., 1998. 36(11): p. 541-546.

180. Kelly, S., et al., Assessing the Authenticity of Single Seed Vegetable

159 Oils Using Fatty Acid Stable Carbon Isotope Ratios (C-13/C-12). Food Chem., 1997. 59(2): p. 181-186.

181. Mu, H.L., H. Sillen, and C.E. Hoy, Identification of Diacylglycerols and Triacylglycerols in a Structured Lipid Sample by Atmospheric Pressure Chemical Ionization Liquid Chromatography/Mass Spectrometry. J. Amer. Oil Chemists Soc., 2000. 77(10): p. 1049- 1059.

182. Mottram, H., R., S.E. Woodbury, and R.P. Evershed, Rapid Commun. Mass Spectrom., 1997. 11(12): p. 1240-1252.

183. Kimpe, K., P.A. Jacobs, and M. Waelkens, J. Chromatogr. A, 2001. 937(1-2): p. 87-95.

184. Carelli, A.A. and A. Cert, Comparative-Study of the Determination of Triacylglycerol in Vegetable-Oils Using Chromatographic Techniques. J. Chromatogr. A, 1993. 630(1-2): p. 213-222.

185. Antoniosi, N.R., E. Carrilho, and F.M. Lancas, Fast Quantitative- Analysis of Soybean Oil in Olive Oil by High-Temperature Capillary Gas-Chromatography. J. Amer. Oil Chemists Soc., 1993. 70(10): p. 1051-1053.

186. Proot, M., P. Sandra, and E. Geeraert, Resolution of Triglycerides in Capillary SFC as A Function of Column Temperature. J. High- Resolution Chromatogr. & Chromatography Communications, 1986. 9(3): p. 189-192.

187. Russo, M.V., G. Goretti, and A. Soriero, Triglyceride Analysis with Microcapillary Gas Chromatography Columns. Annali di Chimica, 1996. 86(5-6): p. 281-291.

188. Plattner, R.D., High-Performance Liquid-Chromatography of Triglycerides - Controlling Selectivity with Reverse Phase Columns. J. Amer. Oil Chemists Soc., 1981. 58(5): p. 638-642.

189. Neff, W.E., et al., Analyses of Vegetable Oil Triacylglycerols by Silver Ion High-Performance Liquid-Chromatography with Flame Ionization Detection. J. Chromatogr., 1994. 17(18): p. 3951-3968.

190. Kalo, P. and T. Kuuranne, Analysis of Free and Esterified Sterols in Fats and Oils by Flash Chromatography, Gas Chromatography and Electrospray Tandem Mass Spectrometry. J. Chromatogr. A, 2001. 935(1-2): p. 237-248.

160

191. Verleyen, T., et al., Influence of the Vegetable Oil Refining Process on Free and Esterified Sterols. J. Amer. Oil Chemists Soc., 2002. 79(10): p. 947-953.

192. Brumley, W.C., et al., Mass-Spectrometry and Identification of Sterols in Vegetable-Oils as Butyryl Esters and Relative Quantitation by Gas-Chromatography with Flame Ionization Detection. J. Assoc. of Official Anal. Chemists, 1985. 68(4): p. 701- 709.

193. Grob, K., et al., The Detection of Adulteration with Desterolized Oils. Fett Wissenschaft Technologie-Fat Science Technology, 1994. 96(9): p. 341-345.

194. Dionisi, F., J. Prodolliet, and E. Tagliaferri, Assessment of Olive Oil Adulteration by Reversed-Phase High-Performance Liquid Chromatography Amperometric Detection of Tocopherols and Tocotrienols. J. Amer. Oil Chemists Soc., 1995. 72(12): p. 1505- 1511.

195. Frank, C., et al., GC-IRMS in the Authenticity Control of the Essential Oil of Coriandrum-Sativuml. J. Agric. & Food Chem., 1995. 43(6): p. 1634-1637.

196. Aparicio, R., M.T. Morales, and V. Alonso, Authentication of European Virgin Olive Oils by Their Chemical Compounds, Sensory Attributes, and Consumers' Attitudes. J. Agric. & Food Chem., 1997. 45(4): p. 1076-1083.

197. Rodgers, R.P., et al., Compositional Analysis for Identification of Arson Accelerants by Electron Ionization Fourier Transform Ion Cyclotron Resonance High-Resolution Mass Spectrometry. J. Forensic Sci., 2001. 46: p. 268-279.

198. Emmett, M.R. and R.M. Caprioli, Micro-Electrospray Mass Spectrometry: Ultra-High-Sensitivity Analysis of Peptides and Proteins. J. Am. Soc. Mass Spectrom., 1994. 5: p. 605-613.

199. Chalmers, M.J., et al., Liquid Chromatography-Fourier Transform Ion Cyclotron Resonance Mass Spectrometric Characterization of Protein Kinase C Phosphorylation. J. Proteome Res., 2003. 2: p. 373-382.

161

BIOGRAPHICAL SKETCH

EDUCATION:

Florida State University, Tallahassee, FL Ph.D. in Analytical Chemistry, April 2004

Fudan University, Shanghai, P.R.China B.S. in Chemistry, June 1998

Publications

Wu, Z.; Strohm, J.J.; Song, C.; Rodgers, R.P.; Marshall, A.G. “Comparative Compositional Analysis of Untreated and Hydrotreated Oil by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Submitted to Energy & Fuels, March 2004.

Wu, Z.; Rodgers, R.P.; Marshall, A.G. “Detailed Compositional Analysis at Different Stages of Coal Liquefaction by Negative-Ion Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Submitted to Environmental Science & Technology, March 2004.

Wu, Z.; Rodgers, R.P.; Marshall, A.G. “Compositional Determination of Acidic Species in Illinois #6 Coal Extracts by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Submitted to Energy & Fuels, March 2004.

Wu, Z.; Rodgers, R.P.; Marshall, A.G. “Characterization of Vegetable Oils: Detailed Compositional Fingerprints Derived from Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Submitted to Journal of Agricultural and Food Chemistry, March 2004.

Wu, Z.; Rodgers, R.P.; Marshall, A.G. “Two and Three Dimensional van Krevelen Diagrams: A Graphical Analysis Complementary to the Kendrick

162 Mass Plot for Sorting Elemental Compositions of Complex Organic Mixtures Based on Ultrahigh-Resolution Broadband FT-ICR Mass Measurements”, Accepted by Analytical Chemistry, February 2004.

Wu, Z.; Jernström, S.; Hughey, C.A.; Rodgers, R.P.; Marshall, A.G. “Resolution of 10,000 Compositionally Distinct Components in Polar Coal Extracts by Negative-Ion Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Energy & Fuels, (2003), 17, 946-953.

Wu, Z.; Hendrickson, C.L.; Rodgers, R.P.; Marshall, A.G. “Composition of Explosives by Electrospray Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry”, Analytical Chemistry, (2002), 74, 1879- 1883.

163