‘ I’IIII‘T IE“? m‘ ”my“ $1013? I It; I III I ' ' III III II'UIIIII HI I I “I III: IIN’II‘ I J,“ TAIJIEIIIIEEE W ’ - ..1 i ' .71! -‘~ ’ ~v . w‘ . -" .J ~v'-r- .2 ‘ ' ~— ‘- .. g. .h - .‘v I ' 13 ..::,—V- ,u 2.1. - ~- In. .- ,I - . 3- v" . - IV. ‘. .4 ‘I I. v7. ,- I—~t , .\ . . -_ -.., .' - - ’ .J .4 I“ ' _ ~Vfi . "J . .. A. H ,..‘fi§? .. .,.V_ pr“ , v‘w‘, 4”“ “.1“:- — n_—-- .n. I .. .-..I _ ..-.. _.'.._-..--‘-n. ..h‘- . ..-l a _ --.' - o< 15" -52: e 70:”: ‘ -131, ' "fl:_'v£é{f o... h D ~_ V‘— .. v 3 fl. : .__. -_.A$.‘_....“‘““-: .A. ‘ 4 ..¢ "‘2. > ‘ »,_ ' c- c' I‘ ' ' I w J' . .. ”r.- : .- . A.;--—_-Q-_ ‘2. "- ‘ ‘Mk‘ III III W "I I- I“; ”W . . I ’ If H: :[131‘ Egg-52411.21!“ “3’.” i; 'I II" {7552 WW! I "3&ng I a; J! IfIrqufiIwIII II ii ’I If” JI IIEII “47h, III I I II £ {Iqu’ It? lIr‘wl‘ If F‘I I: “I I'm!“ IIIIL ’IIh} III, ”III” II ' fl! .III'II “I?“ ; I I M ' 3N ': I! ». I III M I I. 'I: .1'0‘4 » . ‘1‘) 1' IIP’III'IUI}. I" I). I WW“ II II :1 II' I I}. 2}?“ J I "II I IE . 207500 72 I IIIIII’I'III'III’“ I III III I: IIIIIIIII . 3LIZBROOARY * [ Michigan State MnivenitmeJ IIIIIIIIIIIIIIIIIIIIII This is to certify that the dissertation entitled The Development of Techniques for Complex Mixture Analysis by Means of Gas Chromatography/Triple Quadrupole Mass Spectrometry presented by Adam Jeremiah Schubert has been accepted towards fulfillment of the requirements for Ph . D. degree in Chemistry flair % fl... Major professor Date/EWZILLI /2’} /?&g/ ( MS U is an Affirmative Action/Equal Opportunily Institution MSU RETURN I NG_M_AU-EL&L§I Place in book drop to LIBRARJES remove this checkout from W your Y‘t’C‘tn‘d. FINES wi ll be charged if book is returned after the date stamped below. r. 7 - - v v.v r - p. -————~-._ - .--.- v-7- .-__-..[ -._-.c—-e- __—-—.._ THE DEVELOPMENT OF TECHNIQUES FOR COMPLEX MIXTURE ANALYSIS BY MEANS OF GAS CHROMATOGRAPHY/TRIPLE QUADRUPOLE MASS SPECTROMETRY By Adam Jeremiah Schubert A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Chemistry 1987 p > 1" 12%" If /J\J [a K? ABSTRACT THE DEVELOPMENT OF TECHNIQUES FOR COMPLEX MIXTURE ANALYSIS BY MEANS OF GAS CHROMATOGRAPHY/TRIPLE QUADRUPOLE MASS SPECTROMETRY By Adam Jeremiah Schubert Current technology in several phases of organic mixture analysis have been integrated to demonstrate the analytical power of the combined technique of gas chromatography/triple quadrupole mass spectrometry (GO/TOMS). This research has included capillary gas chromatography, chemical ionization, triple quadrupole mass spectrometry, and the high speed instrument control and data systems which are required to make such a powerful instrument feasible. The feasibility of GC/TQMS was enhanced by an increase in overall system speed and the development of programs to allow trace-level targeted component analyses on time variant samples introduced via the gas chromatograph. The performance of the instrument control system was achieved by dividing instrument control tasks among multiple processors rather than by increasing the power of the processor being used. This choice provides greater flexibility, modularity, and sufficient power to allow future incorporation of such features as intelligent dual-mode (pulse counting and analog) data acquisition and programmed ion optics. The data acquisition software for the triple quadrupole mass spectrometer was converted from a single microprocessor to a four microprocessor configuration. To facilitate achievement of the goal of targeted component analysis, additional programs were developed to perform multiple reaction monitoring (MRM, analogous to the GC/MS technique of selected ion monitoring) GC/TQMS. Methanol chemical ionization was investigated as a tool for the mass spectrometric determination of trace level polar components in petroleum products. Results from this study indicate that methanol enhances both the selectivity and sensitivity of the ionization when compared to the more conventional technique of methane chemical ionization. Studies on the effect of varying the pressure of the methanol reagent demonstrated a simple approach through which the analyst can adjust both the sensitivity and selec- tivity of the ionization. Detection limits were determined for the determination of several thiophenes in a commercial jet aviation fuel by means of GC/TQMS. The combined use of capillary gas chroma- tography, methanol chemical ionization, and TOMS with specialized data acquisition programs, enabled the detection of these targeted components down to the low parts per million. ACKNOWLEDGEMENTS The preparation of this dissertation and the research behind it was possible only as a result of the help and support of a number of individuals. A brief mention here represents only a small part of the thanks which I owe them. The love and support of my parents has been key in guiding me along the path to graduate school and in seeing me through to the goal that this dissertation represents. The patience, knowledge, and example set for me by my advisor, Professor Christie G. Enke, has been invaluable to the success of my graduate work and a vital influence in my professional career. The guidance of my second-reader, Professor J3 T. Watson, has been of exceptional value in the preparation of this dissertation. The mutual support and aid of all the members of the Enke Research Group was absolutely essential in the perfor- mance of this research. The. assistance of Mark Bauer and Carl Myerholtz was especially valuable. The love and encouragement of some very special friends was vital. A final word of thanks must go to NASA, NSF, and the MSU Department of Chemistry for the financial support which made this possible. ii TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES . . . . . . . . . . . . . . . . . . . Chapter 1. Introduction Chapter 2. The Analysis of Complex Organic Mixtures Criteria for Complex Mixture Analysis Approaches to Complex Mixture Analysis . . . . . Techniques for the Analysis of Complex Organic Mixtures . . . . . . . . . . . . . . . . . . . . Chapter 3. Adaptation of Triple Quadrupole Mass Spectrometry Programs for a Multiple Microprocessor Control System Sweeps . . . . . . . . . . . . . . . . . . Scans Design Criteria for Future Systems Recommendations . . . . . . Conclusions . . . . . . . . . . . . . Chapter 4. Development of Multiple Reaction Monitoring Programs for Triple Quadrupole Mass Spectrometry Summary . . .I. . . Chapter 5. The Relative Efficiency of Methanol Chemical Ionization for Selected Hydrocarbons and Heterospecies Present in Fuels . . . . . . . Introduction Experimental Conditions Comparison of Spectra under Methane and Methanol CI iii vi vii 10 13 20 24 28 31 35 37 38 48 50 50 56 58 Spectra Obtained Using Methane Chemical Ionization 61 Spectra Obtained Using Methanol Chemical Ionization . . . . . . . . . . . . . . . . . 65 Discussion of Full Scan Chemical Ionization Spectra . . . . . . . . . . . . . . . . . . . 76 Relative Sensitivities of Methane and Methanol CI 79 Discussion of Selected Ion Monitoring Data . . . . 86 Summary and Conclusions . . . . . . . . . . . . . 93 Chapter 6. The Use of Multiple Reaction Monitoring Gas Chromatography/Triple Quadrupole Mass Spectrometry for the Detection of Trace Components in Jet Fuels 96 Background . . . . . . . . . . . . . . . . . . . . 98 Experimental Conditions . . . . . . . . . . . . . 99 Preliminary MS and MS/MS Studies . . . . . . . . . 101 Determination of Detection Limits . . . . . . . . 110 Conclusions . . . . . . . . . . . . . . . . . . . 121 Chapter 7. Selective Ionization, Gas Chromatography, and Triple Quadrupole Mass Spectrometry as Tools for the Characterization of Fuels . . . . . . . . 123 Gas Chromatography as a Selectivity Element . . . 124 Mass Analysis as a Selectivity Element . . . . . . 125 Chemical Ionization as a Selectivity Element . . . 127 Tandem Mass Spectrometry as a Selectivity Element 129 Conclusion . . . . . . . . . . . . . . . . . . . . 129 Appendix A. Inter-Processor Synchronization for Multi- Processor Data Acquisition . . . . . . . . . . . . 132 Sweeps and Scans . . . . . . . . . . . . . . . . . 132 Selected Ion Monitoring/Multiple Reaction Monitoring . . . . . . . . . . . . . . . . . 140 iv Appendix B. Program Listings for Single and Multiple Processor Scans and Sweeps . . . . . . . . Sweeps . Neutral Loss Scans LIST OF REFERENCES 155 155 161 173 LIST OF TABLES Table 3.1. Advantages of Distributed Processing Systems . . . . . . . . . . . . . . . . . . . . . 22 Table 5.1. Composition of Test Blend . . . . . . . . . 59 Table 5.2. Retention Times of Blend Components . . . . 81 Table 5. 3. Sensitivities for Several Heterospecies Relative to n- Butylbenzene with Methane CI at 2. 0 X 10 4 torr . . . . . . . . . . . . . . . . . . 82 Table 5.4. Sensitivities for Several Reterospecies Relative to n-Butylbenzene with Methanol CI at 6.5 X 10"5 torr . . . . . . . . . . . . . . . . . . . 83 Table 5. 5. Sensitivities for Several Heterospecies Relative to n- Butylbenzene with Methanol CI at 2.0 x 10- 4 tort O O O O O O O O O O O O 0 O O O O O 84 Table 5. 6. Sensitivities for Several Heterospecies Relative to n- Butylbenzene with Methanol CI at 4.5 X 10 4 torr . . . . . . . . . . . . . . . . . . 85 Table 6.1. Spectra of Pure Thiophenes Ionized by Electron Impact . . . . . . . . . . . . . . . . . 104 Table 6.2. Spectra of Pure Thiophenes using Methane CI . . . . . . . . . . . . . . . . . . . . 108 Table 6.3. Spectra of Pure Thiophenes using Methanol CI . . . . . . . . . . . . . . . . . . . 111 Table 6.4. Reactions for Monitoring Thiophenes in Fuels . . . . . . . . . . . . . . . . . . . . . . 112 Table 6.5. Detection Limits for Thiophenes in Jet Fuels . . . . . . . . . . . . . . . . . . . . . . 115 LIST OF FIGURES Figure 3.1. Distribution of Tasks in the Multiprocessor System . . . . . . Figure 3.2. Timing of Single and Multiprocessor Sweeps Figure 3.3. Timing of Single and Multiprocessor Scans Figure 4.1. The Multiple Reaction Monitoring Experiment . . . . . . . . . . . . . . . . . Figure 5.1. Compounds for Methanol CI Study Figure 5.2a. Methane CI at 2.0 X 10" torr Figure 5.2b. Methane CI at 2.0 X 10“ torr Figure 5.3. Methanol Reagent Ions . . . . . . . . . . Figure 5.4a. Methanol CI at 6.5 X 10'5 torr Figure 5.4b. Methanol CI at 6.5 X 10"5 torr Figure 5.5a. Methanol CI 2.0 X 10‘4 torr Figure 5.5b. Methanol CI 2.0 X 10" torr . Figure 5.6a. Methanol CI 4.5 X 10" torr Figure 5.6b. Methanol CI 4.5 X 10“ torr Figure 5.7. Trends in Selectivity Ratios for Ionization Conditions Studied Figure 5.8. Trends in Sensitivity for Ionization Conditions Studied vii 23 26 29 40 so 62 63 66 69 7o 72 73 74 75 87 91 Figure 6.1. Structures of Pure Thiophenes Studied . . 102 Figure 6.2. Proposed Fragmentation Paths for Substituted Thiophenes . . . . . . . . . . . . . . 106 Figure 6.3a. Detection of Thiophenes in Jet A . . . . . 116 Figure 6.3b. Detection of Thiophenes in Jet A . . . . . 117 Figure 6.3c. Detection of Thiophenes in Jet A . . . . . 118 Figure 6.3d. Detection of Thiophenes in Jet A . . . . . 119 Figure A.l. Interprocessor Synchronization for SIM and MRM . . . . . . . . . . . . . . . . . . . . . . . 144 viii Chapter 1. Introduction The research presented in this dissertation covers several aspects of the development of the multidimensional analytical technique of gas chromatography/triple quadrupole mass spectrometry (GC/TQMS). Emphasis has been placed on practical concerns relevant to developing a working tool for the analysis of trace components in complex organic mixtures. Chapter Two opens with a discussion of what goes into the development of a practical analytical technique. The chapter continues with a discussion of how gas chromato- graphic and mass spectrometric techniques meet the needs for the analysis of complex organic mixtures and closes with the potential of GC/TQMS for addressing this type of problem. Chapter Three begins the discussion of the actual development of the control and data systems for GC/TQMS. It covers the adaptation of TQMS instrument control and data acquisition software from a single microprocessor control system to a multiple microprocessor system. This adaptation provides analysts with greater speed, flexibility, and power to apply to solving a wide range of analytical problems in a reproducible manner. Chapter Four relates the development of software to perform multiple reaction monitoring (MRM), the TQMS analog to the GC/MS technique of selected ion monitoring (SIM). By making highly efficient use of analytical ”run time,” MRM provides a sensitivity for trace analysis which is greatly enhanced relative to that which is available by full-scan TQMS. Chapters Three and Four are supplemented by Appendices A and B. Appendix A presents a very detailed description of the software discussed in Chapter Three and Appendix B consists of annotated listings of several key portions of the software. Chapter Five of this dissertation covers research which investigated the sensitivity and selectivity of methanol as a chemical ionization reagent for the analysis of trace polar compounds which may be present in fuels. Chapter Six applies this methanol chemical ionization work, in conjunction with the GC/TQMS instrument, and the control and data systems, to the analysis of trace level components in a complex organic mixture. Specifically, the chapter presents the use of capillary GC/TQMS with methanol chemical ionization for the determination of low parts per million levels of thiophenes in commercial jet aviation fuel. Chapter Seven brings the dissertation into focus with a discussion of how the different elements of the GC/TQMS instrument individually, and in concert, contribute to the analysis of complex organic mixtures. The chapter closes with some conclusions and a brief discussion of possible 'future work. Chapter 2. The Analysis of Complex Organic Mixtures Criteria for Complex Mixture Analysis The identification and quantitation of trace species in complex mixtures is a classic problem confronting the analytical chemist. This dissertation covers the devel- opment of techniques for complex mixture analysis, thus, it is useful to begin with a discussion of criteria for the evaluation of this type of work. The necessary consider- ations for this type of analysis may be categorized as follows: Qualitative and Quantitative Information — In order to select the compound(s) of interest from a complex sample matrix, sufficient qualitative information must be available to enable a positive identification. Quantitative accuracy requires an analytical signal that is sufficiently repro- ducible as to allow generation of stable working curves. Depending on the nature of the analysis, the working curve may be derived from internal standards, external standards, or standard addition. Sensitivity and Detectability - The quantitative analysis of trace components necessitates both high sensitivity and low detection and quantitation limits. The limit of detection corresponds to the smallest quantity of analyte for which the analytical signal can be reliably distinguished from background (often defined as the amount of analyte giving a signal-to-noise ratio of 3.) For quantitative work, a limit of quantitation, the smallest concentration of analyte for which the working curve is valid, is often defined at a signal-to-noise ratio somewhat higher than the detection limit. Sensitivity is defined as the derivative of signal with respect to analyte concentration (i.e. the slope of the working curve, if it happens to be linear.) The sensitivity of a technique must be sufficient that signals corresponding to the smallest concentration difference of interest can be distinguished. For example, it may be defined that the smallest difference in signals that can be reliably measured is three times the random noise. If it is then necessary to distinguish between analyte concentrations of 1.0 ppm and 1.1 ppm under conditions where the random noise is 10 units of signal, the sensitivity must then be at least 300 units/ppm ( [3 x lO]/[l.l - 1.0] = 300 ). However, measure- ment systems have a finite dynamic range, hence, excessive sensitivity can limit the range of analyte concentrations which may be determined. Analysis Time — Three of the major factors influencing the choice of an analytical method are cost, throughput, and turnaround time. Complex mixture analyses are often required as responses to crisis situations. In such cases, the rapid availability of even preliminary results can be crucial. Two of the biggest cost factors in this type of analysis are the purchase and maintenance of the necessary instrumentation and the time of a skilled analyst. The high costs associated with instrumentation for complex mixture analyses are amortized across the number of samples which can be analyzed in the available instrument time. Usable instrument time in itself can often be the limiting resource. In such cases, decreased demands on instrument time effectively reduce the costs associated with the analysis of an individual sample while maximizing the availability of scarce instrumentation resources. Another scarce resource is the time of skilled analysts who are frequently responsible for both instrument operation and data interpretation. Because of these factors, reducing the analysis time not only serves to reduce the overall time demands on the analyst, but also serves to maximize the portion of time available for interpreting the data, a task generally not amenable to automation. For routine analyses, the predominant time constraint is the required sample throughput, defined as the number of samples which may be processed in a given time interval. Throughput for a given analysis is increased by reducing the analysis time of an individual sample or by increasing the number of operations that may be either overlapped or performed in parallel for a number of different samples. The overall analysis time per sample is generally fixed by the choice of method. Where there are analysis steps that require little operator attention, or where more than one analyst/technician is available, several steps of the analysis may be overlapped -- for instance, an analyst may be interpreting the results of one sample, while an auto— mated instrument is running the next one, and a technician is preparing additional samples. The throughput gains possible with overlapped tasks depends on how many analysis steps may be efficiently overlapped and the amount of time/attention required by the slowest step. Parallel operation refers to the case where a number of samples simultaneously, can be affected for a particular step in the analysis, such as a; preliminary separation. This approach is a very effective way to increase sample throughput in cases where the parallel operation is the major time- consuming task in the analysis and the costs of the dupli- cate apparatus required for the parallel operations are sufficiently small. Sample Preparation - The amount of sample preparation required for an analysis can have significant effects in terms of sample contamination, turnaround time, throughput, and analysis cost. Any manipulation of a sample provides an opportunity for causing undesired alterations in its composition. The magnitude of this problem is multiplied when analyzing for trace components. Sample treatments, such as. extractions and chromatography can also be very time-consuming to the extent of often being the time limiting factor in the analysis. The degree to which sample preparation activities for multiple samples can be either overlapped or performed in parallel can often be the limiting throughput factor when there is a high sample load. Finally, sample treatments consume analyst time, equipment, and supplies, all of which can add greatly to the cost of the analysis. Cost - Before any analytical technique can be selected for solving a real problem, its costs, measured relative to alternate methods for obtaining the desired information for the required sample load, must be commensurate with the value of the information that it provides in the available time. Costs for a given technique can be attributed to the purchase cost of the instrumentation and related equipment, maintenance costs, the physical plant necessary to support the instrumentation, consumable supplies, analyst time for sample preparation, analysis, and interpretation, and training costs necessary to provide the skilled analysts and technicians necessary to support the technique. For routine analyses, where there is a steady sample load, analysis costs are best measured relative to throughput. Within the bounds of available personnel, capital equipment and physical plant, increased throughput can allow the fixed costs associated with an analysis to be spread over a larger number of samples. The realization of these throughput-related benefits, however, requires a sufficient sample load to justify the investment in addi- tionally dedicated resources. That is, it does little good to set up a method for a fifty sample per hour throughput in a laboratory which only analyzes two hundred of these cases per month. Repeatability and Reproducibility - Complex mixture analyses are often complicated by having to work with small amounts of sample, where replicate analyses are not possible. Additionally, the analyst must be able to make interpre— tations based on samples run over a period of time, and at different laboratories. These factors stress the need for repeatability and reproducibility in the analysis. These problems are generally approached by definition of standard method validation protocols and the analysis of standard reference materials both over time and in inter-laboratory comparisons. In the analysis of trace components in complex mixtures, however, time pressures and the need to do "one- of—a-kind" analyses makes it difficult to define standard protocols, and the manufacture and distribution of standards for trace components in complex mixtures is exceedingly difficult. As a result, the need for overall instrument reliability and good analytical practices is especially pronounced. 10 Approaches to Complex Mixture Analysis In evaluating different methods of complex mixture analysis, consideration should be given to three distinct cases of the type of information desired. The first case occurs when the analytical goal is to quickly determine whether or not any items in a small list of targeted items are present and above some pre-determined threshold level. The second case to be considered occurs when detailed information on the identity and amount of a large number of mixture components is sought. A third case is where general information on the overall composition of the mixture is needed. Each of these cases place very different require- ments on the analytical technique to be employed and, thus, all of the complex mixture analysis considerations discussed above must be weighed quite differently. Screening Analysis - In the first case, which is a rapid screening analysis, the principal goals are speed and a yes/no type answer for each targeted item. For this type of analysis, conditions are adjusted to be selective for the targets, thus minimizing the amount of additional infor- mation required for positive identifications. Quantitative information is desired only to the extent of comparing it to a threshold value and the sensitivity and reproducibility need only be sufficient to make a reliable comparison to the action level. Cost considerations are exceedingly important due to the high sample load generally accompanying this type 11 of analysis. Further, high throughput is essential both as a means of spreading out the fixed analysis costs over a larger number of samples and to reduce the effective turnaround time. Total Composition Analysis — The second case is that of a very detailed analysis. For this type of work, the goal is to obtain as much qualitative and quantitative information as is possible. Further, to enable the study of trace components, there are high demands on sensitivity and dynamic range. In order to reliably examine trace compo— nents, it is still quite important to minimize sample preparation. To obtain all of this information from the sample, a price is generally paid in terms of analysis time and cost. Compound Type Analysis - The third case is a compound type analysis. In this approach the goal is to obtain general information on the overall composition of the mixture and the distribution of related components. Type analyses assume that a particular set of signals can be used to characterize the presence and abundance of specific compound types. This approach precludes the analysis of individual compounds which may be present as unknowns and impurities. In fact pre-separations are frequently employed to remove trace components which may interfere with the indicator signals. The great utility of this approach comes from the fact that specific, predetermined information of a statis- tical nature is being sought; This enables the use of highly 12 automated data treatments which reduce the data to a tabular or graphic format which can be easily interpreted. This approach is frequently used in the petroleum industry for the comparison of different samples of a particular type of fuel such as middle distillates, gas-oils, and low olefinic gasolines (1, 2, 3). Data Organization - A further difference in these three analysis cases is the amount of information being gener- ated. For the rapid screening only a small amount of information is generated -- a yes/no for each item in the target list. This simple information list allows for great flexibility in the manner in which the data are reported. In a detailed analysis, however, there can be an over- whelming amount of information generated. This data absolutely must be organized in some convenient a logical manner before any analyst can be expected to make a useful interpretation. Further, not all of the large quantities of data generated are relevant; much data must be weeded out and interpreted by the analyst before the desired infor- mation is obtained. Thus, as a very practical matter, the design of a technique for detailed analysis must consider approaches to reduce the data. The data obtained in a type analysis is readily summarized, but can only provide general information about the mixture as a whole. 13 Techniques for the Analysis of Complex Organic Mixtures Many techniques are employed for the analysis of complex organic mixtures. These techniques have wide ranges in capability and convenience for different cases of analysis. These capabilities also vary widely among different instruments of the same type and are constantly evolving. In general, however, they all have some funda- mental strengths and limitations which must be understood before different analysis methods can be properly considered. Mass Spectrometry Applications - One of the first instru— mental techniques to be employed for organic mixture analysis was mass spectrometry. Conventional mass spectro- metry has been demonstrated to be extremely useful for the determination of individual components in simple mixtures. This approach may be used when the analyst has a fairly good idea of the components which are likely to be present, thus allowing the individual spectra to be deconvoluted by comparison with reference spectra. Conventional mass spectrometry has also been widely applied to the performance of type analyses on complex mixtures (4). A type analysis, however, does not yield information on individual compo- nents, but rather approximate totals for groups of struc- tural isomers by means of monitoring a mass which is characteristic of that group. Additionally, in cases of l4 highly complex mixtures in which different compound types having isobaric characteristic peaks are present, the sample often requires a preliminary separation in order to simplify the identification process while still using a low resolu- tion instrument (5). High resolution mass spectrometry can reduce this problem (6), but greater complexity in instru- ment operation tends to restrict usage, especially in routine cases. In general, type analyses are useful only when general information is required about the distribution of compound classes which make up the bulk of a sample of a type which is routinely analyzed. Generally, mass spectrometry with a batch or probe inlet is best suited for the analysis of simple mixtures where interferences between the components are not disab- ling, or for type analyses of more complex mixtures when only general information on the distribution of major sample components is sought. Type analyses, however often require a pre-fractionation of the sample to minimize interferences and, by their algorithms, preclude the identification of any individual trace component. Gas Chromatography Applications — Gas chromatography (60) presents a complementary method for analyzing complex mixtures. Unlike mass spectrometry, where all of the sample components are analyzed simultaneously, GC disperses the individual components of the sample in time, allowing the detector to measure each individual component as it elutes from the column. The GC detector, however, does not 15 generally provide sufficient qualitative information to identify the compound. The only information available is the retention time and peak area for each resolved compo- nent. In general, the retention time for a given compound is a function of the column type, stationary phase, the column oven temperature program, the structure of the compound in question, and its boiling point. The peak area for a given component is dependent upon the functionality of the molecule, its elemental composition, and its concen- tration in the sample. Assuming that the resolution of the chromatogram is sufficient to prevent any significant overlapping of peaks, the retention time, measured relative to well-known reference components in the sample, can be extremely useful for routine analyses of similar mixtures where essentially the same peaks will be present from sample to sample. Retention times alone, while characteristic of specific compounds, are not sufficiently specific to enable the positive identification of an unknown peak in a mixture. Overall, gas chromatography has the advantages of small sample requirements, ease of automation, quantitative results over a wide linear dynamic range, and, frequently, little or no sample pre-separation. However, 60 analyses of complex mixtures suffer form a lack of qualitative infor- mation and long analysis times to attain high resolution. 16 Gas Chromatography/Mass Spectrometry Applications - The use of mass spectrometry alone for mixture analysis presents the problem of deconvoluting the spectra of each component and gas chromatography alone generally provides too little qualitative information for compound identification. The complementary data dimensions of these two techniques, however, allow them to be readily combined to form the extremely useful hybrid technique of gas chromatography/mass spectrometry (GC/MS). As a result, the combined analytical technique of GC/MS has been very highly developed for use in a wide range of applications (7). The combined GC/MS instrument can be viewed as either providing the GC with a highly selective detector in the case of selected ion monitoring or, as providing the mass spectrometer with an inlet which will isolate and provide, in sequence, each of the individual components in a mixture, as in the case of scanning GC/MS. In selected ion monitoring, the mass spectrometer acts essentially as a GC detector of very high sensitivity and, by selection of which mass to monitor, variable selectivity. In the case of scanning GC/MS, the GC inlet allows the mass spectrometer to obtain a spectrum of each of the components in the mixture with the "purity" of each spectrum being determined by the resolution of the GC and the scan frequency of the mass, spectrometer. Use of both the chromatographic retention time and the qualitative 17 information in the mass spectrum of a given peak has been used in metabolic profiling: a technique demonstrated for its great power when applied to complex mixture analysis (8). In any type of GC/MS analysis, however, the of GC analysis time needs to be traded off against the desired chromatographic resolution. Due to its thoroughness, scanning GC/MS analysis is an extremely powerful technique for the analysis of organic mixtures for which little prior information is available. Additionally, the systematic nature of this type of analysis renders it quite amenable to automation. Instruments for GC/MS are also widely avail- able, thus easing the process of interlaboratory cooper- ation. As a consequence of the thoroughness of this type of analysis, however, there is the potential to generate tremendous quantities of data, of which only a fraction may actually be needed by the analyst. Mass Spectrometry/Mass Spectrometry Applications — One method which has been developed to avoid the time/resolution tradeoff inherent in GC/MS, while still essentially allowing for the mass spectrometric identification of the individual components in a mixture is tandem mass spectrometry (MS/MS). One of the more popular and convenient implemen- tations of MS/MS is the triple quadrupole mass spectrometer (TQMS) (9, 10, 11). In TQMS with a batch or probe inlet, the entire sample is available for analysis at all times. Thus, the components of interest may be examined without 18 having to wait for all of the other components to elute from the gas chromatograph. If the goal of the analysis is only the detection or quantitation of a few components, the great bulk of unneeded data frequently encountered in GC/MS is eliminated; indeed, it is never acquired. Additionally, TQMS allows the operator to change any of the system parameters at any time during the analysis. In order to take advantage of this feature, however, considerable operator attention is often required, whereas, in GC/MS, many instruments can run unattended. Problems with TQMS mixture analysis include the difficulty in getting a representative sample from a mixture of components with a wide range of volatilities (12), the necessity for main- taining constant sample composition for the duration of the analysis, and the high probability of having to deconvolute the spectra of the isomeric and isobaric peaks that are likely to be found in any mixture of reasonable complexity. Furthermore, in cases of non—routine samples for which little prior knowledge is available, the time advantage over GC/MS can easily be lost and the amount of data generated can be just as large and, potentially, even more difficult to interpret. Thus, TQMS has its greatest value in complex mixture analysis when only a few, targeted components need to be determined and differentiation of isomers is not required. 19 Gas Chromatography/Mass Spectrometry/Mass Spectrometry Applications - The addition of a GC inlet to the TQMS can frequently simplify the problems of performing TQMS on complex mixtures by dispersing the components in time and by allowing a more representative sampling of mixtures con- taining components with a wide range of volatilities. The time dispersion provided by 60 can greatly facilitate interpretation of the data by allowing isomeric and isobaric components to be individually analyzed and by providing an additional dimension for interpreting experimental results. Additionally, in multiple reaction monitoring GC/TQMS experiments, several different components or component classes can be simultaneously and independently analyzed as they elute from the 60 column. The power of having two stages of mass filtering allows for a wider range of control over the selectivity of detection than is possible with conventional GC/MS. The separation power of the first mass filter further serves to reduce demands on chromatographic resolution by allowing each component to be monitored in a different dimension. This has a net result of decreasing the analysis time. By providing the capability of increased selectivity at lower chromatographic resolution, GC/TQMS minimizes the two major problems of GC/MS, namely the long analysis times and large quantities of data which are generated. Chapter 3. Adaptation of Triple Quadrupole Mass Spectrometry Programs for a Multiple Microprocessor Control System The evolution of organic mass spectrometry from a specialized art employed in the slow scanning batch analysis of gases to a widely employed, high speed technique for mixture analysis and identification of trace quantity samples has been greatly assisted by the concurrent devel- opment of powerful, high speed data systems. Further, the recent evolution of high-powered, low cost microprocessors has, in recent years, forced an entire rethinking of data system design and functionality (13). This is especially true for multidimensional techniques such as MS/MS and GC/MS/MS where the data rates, quantities of data, and the number of instrumental parameters to control reach well beyond that which is feasible to perform manually (l4). Proper design of a data system requires that these demands be met in a user-friendly and economically feasible manner. Achieving these goals necessitates taking full advantage of the diversity of features now available in computer hardware and software. This entails assigning different data system tasks amongst various elements of hardware and software as needed to optimize cost and 20 21 performance. In other words, the use of several specialized, low-cost microprocessors can provide far greater instrument performance and flexibility than a similar cost system involving a single processor which must be chosen as a compromise of conflicting demands. Some of the advantages of such a distributed processing system are summarized in Table 3.1 (15). The hardware (16) and systems software (17) for the triple quadrupole mass spectrometer control system used in this report have been described previously. Briefly, the system consists of four Intel 8088 processors. The first processor, the master, supports the user terminal and disk drives and directs the operation of the three slave proces- sors. The first slave controls all of the ion path ele- ments, the second slave performs peak-finding and controls the graphics display, and the third slave monitors the detector. The distribution of tasks among the four proces- sors in the system is illustrated schematically in Figure 3.1. This chapter covers the conversion of previously developed (18) single processor data acquisition routines for enhanced operation in a multiple processor environment. The result of this conversion has been to create a flexible and powerful TQMS control system. As a first phase of the multi-micro conversion project, all of the user functions of the single microprocessor system were reproduced. Consistency with the old system was 22 Table 3.1 Advantages of Distributed Processing Systems aster Execution * Parallel execution * Less time spent in "overhead" # Simpler addition of hardware controllers and processors Independent Task Execution * Non-interference of tasks * Elimination of task interleaving problems * Elimination of priority assignment problems * Simpler task program modification Modularity of Hardware and Software * Consolidation of related tasks * Simpler extension of instrument capability * Simpler debugging and troubleshooting Empmhm nommoooumapasz one CH mxmma we soapsnwmpmaa H.m musmHm _ 55;... H #955 _ 442.25; «838.... $552 _ mam 20.20.232.28 mMmfluEamEz. _ w A. ___T .4 m E: w. «mum—55:. 2. 025300 m 0. L 4 W M M . . r .~ ~ A H , m .55.... m u><..m . u><._m 20.5”..me 5... 0252... yawn. ESQ 1:... 20. . ........... p ....... r ..... a ...... a ...... _. o 1a: .19 r. i an”....H........H.H..........H.......... . DU? E ................. - A Hv/l. . ................... . 20. 5mm: 20. m 930 m 930 .930 850m 20. 24 obtained for. the most part. However, since the instrument was intended for a small group of sophisticated users, changes were made when needed to provide additional power or capability for future expansion. Once the user interface was implemented, the next major step was to improve the speed of data acquisition by taking advantage of the distributed processing capabilities of the system. Additional objectives were the implementation of features such as real-time graphics, linked device sweeps, and dual-mode data acquisition; features which were not feasible in the single processor arrangement. Two types of data acquisition routines are discussed: sweeps and scans. A third type of data acquisition, repre- sented by selected ion monitoring and multiple reaction monitoring, will be discussed in the following chapter. Details of the data acquisition software operation are presented in Appendix A. Partial listings of the code discussed in the following sections are given in Appendix B. Sweeps Sweeps are used to acquire data while continuously sweeping the value of any ion path device. They save all of the raw intensity values without performing any peak- finding. This operating mode allows the operator to obtain intensity data as a function of mass or of the voltage on any of the ion path elements. The availability of this mode 25 provides a simple method for the operator to sytematically investigate the ion path tuning, use the analog data to verify the quality of the peak finding, and to examine the energetics of ion-molecule reactions under study. A comparison of the software structuring of sweeps under the single and multiple processor configurations is illustrated in Figure 3.2. In this multiple processor scheme there is a fixed division of tasks among the three slave processors. The first slave controls the ion path configuration, the second slave monitors the ion current through control of the Data Acquisistion Board (Bag) (19), and the third slave receives the x-axis information from the first slave and the y-axis information from the second slave and stores them in a data buffer. The first slave system has an 8 megahertz processor and is used to control the voltages on all of the ion path elements, the quadrupole mass selections, and the quadrupole BF/DC states. This slave calculates the value of each step in the sweep and passes this information to the reduction slave. A second function of this processor is to provide time synchronization through use of the AMU Timer (20) as well as synchronization with the detection slave. In addition, the ion path slave controls the sweep experiment by counting the number of steps remaining in the sweep and informing the other processors when the last step has been executed. Q Tuldoo twonlll'Tll-laoo tmmol'Tlaoo tmonll .0: no; 5: “Fa 91.25223 of "In 255323—95 3: 23532: TL 3: . .. V. : V. x. w. UQHQOCOD a: moot—amt. a; :- >OOC;MANV n; :. >ooc.amAuv m; x0 «a .5 «a a «a . m ’ m ) ’ ’ mmacocoa :3 :A: :5 :3. 4:... :2. mic...“ 9 Jo 0; a 30038: 5 — Sam 5 x 5 jocqm u.» 23:5 zloZostomqu mimmpJ 27 The detection slave has a 5 megahertz processor and is used to interface with the Data Acquisition Board (Daq). In this capacity it provides the start signal to the Daq, monitiors its status byte for the completion flag, and readsback a 24-bit intensity value. This processor informs the ion path slave when acquistion of each point is com- pleted and transmits a 32-bit ion intensity value (the 24-bit value from the Daq with an added high byte of zero) to the reduction slave when that slave is ready to accept it. The third slave also has a 5 megahertz processor and is used to receive the current x-value from the ion path slave and the corresponding ion intensity from the detection slave. These values are then stored in a data buffer to be transmitted to the master processor at the end of the sweep. This processor obtains synchronization by waiting for the needed x and y values to be available from the other two slaves. A more detailed discussion of the sweeps programs is provided in Appendix A. Overall, conversion of the sweeps code to a multi—processor system allows for greater modu- larity in the coding by allowing logically separate tasks to be performed on separate processors. This allows these tasks to be programmed as distinct modules as opposed to interleaving them, as would be necessary on a single processor system. Further, the partioning of data acqui- sition tasks allows the ion path processor nearly twice as 28 much time for calculations as would be available on a single processor system. This additional time is taken advantage of in more complex device contol, such as linked—device sweeps. Scans The second form of data acquisition is the scan. Five types of scans are defined -- single stage MS with the first quadrupole or the third quadrupole, the parent ion scan, the daughter ion scan, and the neutral loss scan. The principal difference between a sweep and a scan is that instead of storing raw intensities, a scan performs peak-finding and saves only the mass/intensity pair for each peak. A schematic comparison of scanning with single and multiple processors is given in Figure 3.3. The task distribution for scans is generally the same as for sweeps with the exception that the reduction slave now performs on- the-fly peak finding instead of directly storing each data point into the buffer and the ion path slave sweeps the RF- power level on any RF—only quadrupole at one-half the magnitude of the scanning quadrupole. The timing diagrams in Figure 3.3 were developed for a neutral-loss scan. This is a "worst case" because the 32-bit calculation required to keep the quadrupole 3 mass tightly linked to the quadrupole 1 mass places the greatest load on the ion-path processor. The times indicated for ?PEAK also represent a worst case P23; .3 3223.32. 3.5: Tn 31.: Ilwomi mnTl ‘ . . , . : _ g 5 _ ax ac —— in“... — — {a f :e Tloomi 00— Ill - i T ommi no. «:5 :3. 33.3.... :1: :3. 33.32. :5» 3:; Ilooms. mnTl Iloomi mnTl Iloomi onTl .. -¢_“x , ._ . J} .3353... :5 in: €35.33. —.:: —J} 2.2.534 :: . 9:: $9: 333233 a!“ :5 —# 3:23:33 Tax—5°: 3:53:33 a!“ 5: Tloomi oo—IIIIIV—IIII-ommi oopllllllloomi 009i- .Ommmooi 205m conusmmm cozoosoo Sou 8. 30 condition (the intensity of the current data point is above threshold and the current peak exceeds the minimum peak width). This case generally occurs only for a small fraction of the actual steps. There are nine possible cases for the current peak-finding algorithm. For the multiple processor implementation, these cases range in time from 16 to 67 microseconds. The schematic illustrates that the single processor system cannot realistically perform a neutral loss scan at 100 microseconds per step (1000 amu/sec at 10 steps per amu). Even evaluating a distribution of times required by the different cases of the peak-finding algorithm, the effective scan rate would vary considerably, depending on the number and size of peaks in the scan. In that mass step calculations are being performed during the data acquisition time of the single processor system, little, if any, gain in scan speed would be achieved by reducing the number of A-to-D conversions averaged for each step (the illustrated implementation uses 8 conversions per step). In contrast to the single processor configuration, the multiple processor implementation of the neutral loss scan comes very close to the 100 microsecond per step goal, even assuming a constant worst case for the peak-finding algorithm. As illustrated in the schematic, this is achieved by performing separable tasks in parallel -- while the detection processor is monitoring the nth point, the ion path processor is calculating the masses for step n+1, and 31 the reduction slave is performing the peak-finding for step n-l. Even with the worst case of the peak-finding algorithm, the reduction slave has a cycle time of 110 microseconds, and it is reasonable to assume that the excess 10 microseconds for the worst case can be caught up with even a small abundance of the more favorable cases. The timing illustrated for the ion path processor allows over 60 microseconds for calculations, but even the calculation- intensive neutral loss scan requires only about five microseconds in the multiple-processor implementation. Thus, the multi-processor system can not only perform a high-speed neutral loss scan, but it also has the capability available to execute more sophisticated scanning algorithms. Design Criteria for Future Systems Study of the timing diagrams presented in this chapter reveal some interesting points on the relative merits of single processor and synchronous multiple processor systems. In the case of SWEEPS, the multiple processor system shows very little advantage over the single processor configuration, primarily due to the overhead imposed by the need for inter-processor synchronization. This comes despite the use of specialized inter-processor hardware designed to minimize software overhead. As a result, the only gains resulting from the multiple processor implemen- tation of SWEEP are the increased amount of time available 32 for ion path calculations and improved modularity of the code. However, both implementations allow sufficient calcu— lations time for most any reasonable scan algorithm. The limited memory available with the 8088-based multiple processor system used restricts SWEEPs to narrow mass ranges (currently 100 amu at 10 steps/amu). Scans conserve memory by performing on-the—fly peak finding synchronized with each data point, allowing for wider scan ranges by throwing away potentially useful peak shape information. The use of synchronized, on-the-fly peak- finding, however, requires multiple processors for repeat- able scan rates. Additionally, even with the multiple processor system, the sophistication of the peak-finding algorithm is compromised by the need to operate within scan rate limitiations (1000 amu/sec at 10 steps/amu). The increasing use of 2000 amu quadrupoles and scan rates of over 2000 amu/sec in commercial instruments further limits the utility of a lOOO-point data buffer and underscores the need for fundamental changes to accommodate higher scan rates. With the rapidly decreasing cost of computer memory, a first step in reducing these problems and simplifying the computer system would be to increase the size of the data buffer so that SWEEPs may be performed on a 2000 amu mass window, and peak-finding can either be postponed until after data acquisition is complete or performed asynchronously from the data acquisition. For SWEEPS employing a constant 33 step size, this temporary data buffer need only contain four bytes for each data point, plus a four byte header region containing the starting x value and the step size. For a 2000 amu scan at 10 steps/amu, this corresponds to 80,000 bytes. Many 16—bit processors, such as the 8088 used here, employ memory segmentation schemes that make it very awkward to address a buffer of larger than 65,536 (64K) bytes. The two most viable approaches to this problem are to use a processor that can directly address an 80,000 byte data buffer (such as the Motorola 68000 family), or to devise a scheme to perform peak finding on-the-fly, but asynchronously from the SWEEP. The first solution gives a surface impression'of simplicity in that it can be readily implemented with a single processor. Such an approach, however, would come at the expense of the many modularity advantages of multiple processor systems, as described earlier in this chapter. More importantly, this would result in considerable dead time between scans to allow for post scan peak—finding, thus negating many of the advantages of a high scan speed. The second approach, performing peak-finding asynchro- nously from data acquisition, can remove the requirement that the 80,000 bytes of data be resident in memory simul- taneously while also eliminating the inter-scan dead time problem. This solution requires the introduction of a second processor to perform most of the peak—finding operations during scan time. To avoid the timing 34 restrictions observed with synchronized on-the—fly peak- finding, mass and intensity data must be made available to the peak—finding processor in a manner other than the point by point mechanism employed in this work. Two approaches to this problem are a dual port first—in-first-out (FIFO) buffer or providing a segment of dual port memory for use as a data buffer. The use of a FIFO avoids the need for an 80,000 byte buffer, thereby opening the design to processors with 64 Kbyte memory segmentation. A FIFO data transfer path will effectively decouple peak—finding from data acquisition so long as the FIFO is sufficiently deep that it never fills. FIFO memory chips are available in a variety of widths and depths and designed for combination to form most any desired configuration. However, the FIFO configuration and support circuitry would need to be custom designed for the given hardware configuration. Additional complications are that the software to reliably read and write the FIFO’s can be just as involved as the interprocessor synchronization used in the current system. Further, the use of FIFO’s make it awkward to employ peak~finding algorithms that require non- sequential access to the raw data (such as correlation functions and centroiding). The problems associated with FIFO memories suggest that shared memory be considered pas an alternative. Commercial microprocessor buses, such as the Multibus and VMEbus, readily provide for shared memory, which can be implemented 35 using most commercial memory boards (the custom inter- processor bus used in this work allows the master processor to access all of address space of the slave processors, but the slaves are unable to address anything other than 64 Kbytes of local address space). Data transfer through use of shared memory can be implemented with little software overhead. This type of data transfer can also be imple- mented using the shared memory space as a large wrap-around buffer, so that processors with 64 Kbyte memory segmentation may still be employed. If this wrap-around buffer is sufficiently large, data in the buffer can be resident long enough to allow the reduction processor to employ non- sequential peak-finding algorithms. Recommendations This work has provided considerable experience in the design of multiple processor control systems. Combined with the evolution in quadrupole mass spectrometry and micro- processor hardware which has occurred since this multi— microprocessor system was designed, a number of recommen- dations can be made towards the design of faster configur- ations with greater power and flexibility. The rich array of commercial, board-level hardware now available should be incorporated wherever practical. This will result in reduced development time and generally improved reliability and documentation. Large RAM buffers in regions of shared 36 memory are now economically viable and can be implemented with standard commercial memory boards. Their use will greatly simplify interprocessor communications and allow peak-finding to be decoupled from instrument control and data acquisition. Such an approach will provide more reproducible scan rates and allow the use of sophisticated peak-finding algorithms. A large wrap-around buffer in shared memory should be used for raw data, with additional buffers available for peak data. These buffers are configured for ready access by additional processors, allowing peak data from one scan to be transparently written to disk or displayed while the next scan is being acquired. A single high-speed processor can perform both the detection and ion path control functions with somewhat reduced modularity if analog-only detection is to be used. An additional processor creates a more modular system by isolating the instrument control and data acquisition functions and provides sufficient power to perform dual-mode detection. The instrument control and data acquisition processors may be synchronized through status bytes located in shared memory. Both the detection and ion path slaves can write directly to the raw data buffer. Finally, a master processor is needed to control the slave processors and provide interactive foreground functionality during data acquisition. 37 Conclusions Data acquisition code for triple quadrupole mass spectrometry has been 'converted to take advantage of the enhanced power of a multiple processor control system. This conversion has allowed for higher speed operation while maintaining modular software. Combined with the modularity of the hardware, this provides a highly flexible system which can be readily tailored to perform demanding and specialized experiments, including such possibilities as intelligent dual-mode data acquisition and sophisticated linked-scanning algorithms. Based on the experience gained in developing this prototype system, recommendations have been made for the design of a faster, more powerful system. Chapter 4. Development of Multiple Reaction Monitoring Programs for Triple Quadrupole Mass Spectrometry A traditional approach for improving the detection limits and reproducibility of GC/MS is selected ion moni- toring (SIM). As opposed to full scan GC/MS, where the mass filter is continuously and repetitively swept through a range of masses bracketing all ions of interest in the sample, SIM involves repetitively cycling the mass analyzer through a small number of pre-determined m/z values. This improves the signal to noise ratio of the analysis by allowing the instrument to spend the bulk of its time monitoring ions of specific interest while not wasting time gathering data in irrelevant regions of the spectrum. As a result, detection limits can be improved by as much as two to three orders of magnitude over those attainable in full scan mode. This approach is especially valuable in con- junction with capillary gas chromatography where sample concentrations in the ion source may vary significantly over the time required to obtain a full mass spectrum. Clearly, this enhanced sensitivity comes at the expense of infor— mation content -- the analyst no longer has full mass spectra available to aid in evaluation of unexpected mixture 38 39 components. However, for targeted analyses, involving mixtures of a type which has been sufficiently charac- terized, this is frequently not a limiting factor. The analog to SIM in tandem mass spectrometry is to cycle through a short list of parent ion/daughter ion pairs, or reactions. This technique is referred to as multiple reaction monitoring (MRM). Through judicious choice of reactions, MRM provides much greater selectivity than is attainable with SIM. Cooks describes MRM as employing a fixed parent ion and cycling only through the daughter ions (21). The use of tandem quadrupoles, however, allows for rapid cycling of both the parent ion and daughter ion masses; in fact, their hardware and control software are identical. To take full advantage of this feature of quadrupole MS/MS, the software designed here allows cycling of both the parent ions and daughter ions in pre-determined reactions. This design is illustrated schematically in Figure 4.1. The design of the selected ion monitoring/multiple reaction monitoring software attempted to incorporate several criteria. These may be summarized as follows -— 1. Provide a user interface which is generally similar to that designed for sweeps and scans. 2. Define four types of selected ion monitoring experiments -- 40 35.....um anhoioz 202-40220 L - " anvo .L.. n Tfi a L... FF} . L L . Lll.llv " L a u .3 82:2 marcnquads D .Timly a r...“ E 00:30” - L L - Llllalllv . L . at. u -m. A" n. L r . L L _ Illlv an”. a... on u 86» 8.6a 848452 mwmsno ¢.H are sawdwmpo momodwo: scapeouwsm mxmoowaosa 41 a. setting quadrupole 1 while quadrupoles 2 and 3 are in HF-only mode (lSIM), b. setting quadrupole 3 while quadrupoles l and 2 are in BF~on1y mode (3SIM), c. keeping quadrupole 3 at a fixed mass while setting quadrupole 1 (PSIM), d. and keeping quadrupole 1 at a fixed mass while setting quadrupole 3 (DSIM). These SIM modes would be analogous to the scanning modes lSCAN, 3SCAN, PSCAN, and DSCAN. Use a single set of utilities for the four SIM modes and multiple reaction monitoring (MRM). Allow for the simple creation and storage of several sets of ions and reactions (elements), with the capability of specifying a different parameter set for each element in a set. Provide a mechanism to allow the user to select from several possible cycle frequencies. Allow the user to set a duration for data acqui- sition while retaining a capability to force a clean termination of the run at any earlier time. 42 7. Use real-time graphics to provide the user with immediate feedback from the current run. 8. Provide a mechanism to scale the real—time display for effective representation of data with a poten- tially wide dynamic range. 9. Define a set of post-run utilities to generate summary reports, provide a range of graphics capabilities from quick and simple to publication quality, allow for archival data storage and generate dumps of the raw data. Consistent with the overall system design, that of separate control and data systems (22), the experimental set up and execution are performed by the control system, while most of the post-run utilities exist as tasks on the data system. Simple graphics and data listing capabilities are provided on the control system, however, to allow for more rapid feedback to the experimenter and to provide some redundancy. Experimental set up consists of two phases -- defi- nition of the elements to be monitored and definition of the various run-time parameters. These steps need to be executed once before the first run is executed. For subsequent runs, however, they only need to be repeated when 43 the experimenter wishes to change the elements or para- meters. Elements are defined through the use of one of two screen editors -- RED, the Reaction EDitor, and SIMED, the Selected Ion Monitoring BDitor. These two editors are very similar. SIMED allows for the definition of up to eight ions to be monitored and a parameter set to use with each of the ions. Physically, the editor presents the user with a table of ions and parameter sets which the user may edit by using the cursor control keys to position the cursor over the entry to be edited and then typing over the current entry. During run-time the system cycles through the list of ions from the beginning through the last non-zero ion entry. The choice of which quadrupole to use for selecting the various ions is made by choice of the appropriate run command (lSIM, 3SIM, PSIM, or DSIM). Any RF-only quadru- poles are set to one-half of the value of the current ion selection. The fixed mass quadrupole in PSIM and DSIM is kept at the "current" mass for the corresponding quadrupole using the parameter set which is active when the run is initiated. All of the run commands use the ion/parameter set table which is active when the run is started. In addition, up to 32 ion/parameter set tables may be stored on disk for future use. Storage and retrieval of these tables is performed by the commands n SIMSAVE and n SIMGET, respectively, where n is the table number, ranging from 0 to 31. 44 Reactions for a multiple reaction monitoring run are defined using the reaction editor, RED. This editor works in essentially the same manner as SIMED. The principal difference is that RED provides for definition of both the parent and daughter ions for each reaction. In addition, it allows the user to specify an RF—value to use for the second quadrupole. This value will default to one-half the lesser of the parent and daughter ion masses. Similar to SIMED, the user may specify a parameter set to be used with each reaction. A maximum of sixteen reaction tables may be saved on the disk. There is a pair of commands, n RSAVE and n RGET, to store and retrieve the nth table, respectively. The second step in setting up a SIM/MRM run is the SSET command. This command prompts the user for inputs. The parameters to be specified are cycle period, scaling for the real-time display, and the maximum duration of the run. Six possible cycle periods are allowed -- 0.1, 0.2, 0.4, 1.0, 2.0, and 4.0 sec/cycle. The display scaling allows the user to specify the vertical range to be used in the real-time display. The default value is 1,048,575 which is the full scale response of the detection electronics. This value is used for all of the elements to be displayed. It is important to note that this value has no effect on the actual data acquisition; it only scales the real-time display. The third item requested is the run duration. In addition to automatically terminating the run after the given time elapses, this value is also used, along with the 45 cycle period, at the start of a run to verify that suffi- cient disk space is available to store the entire run. For each of these parameters, the user is given the choice of either entering a new value or just entering a carriage return to keep the present value. All values entered in SIMED, RED, and SSET are maintained by the computer until they are explicitly changed by the user. An actual run is initiated when the user enters one of the five command words -- lSIM, 381M, PSIM, DSIM, or MRM. The computer then initializes everything and comes back with a prompt for the user to strike any key to begin data acquisition. This provides the experimenter with an opportunity to let the GC oven equilibrate and to make the injection. When the injection is made, a key may be struck and the instrument will go immediately into data acquisition mode. During data acquisition the monitor will have a real- time display of element intensities versus time for each element being analyzed, up to a maximum of five. If more than five elements are being monitored, the system uses the first five entries in the table. This display is erased and restarted after every 1000 data points, thus the time window depends upon the cycle period. The user may allow the experiment to continue until the specified time has elapsed or may stop the analysis at any time by striking the "Q/q" key. Each run is stored as a scan at the end of the most recent experiment in the current data file. This allows the 46 user to access any SIM/MRM run in the same manner as used for scans and sweeps. As with any other data, Data LISTings may be obtained through use of the DLIST command. For SIM/MRM data this produces a raw data dump of the intensity for each element at each time sampled. Because of the sheer volume of data potentially present in a run, DLIST does not perform any normalization or peak-finding. The control system can also generate simple graphics through use of the DISP command. DISP works somewhat differently, however for SIM/MRM data than for scans and sweeps. Instead of using system defaults or the optional DSET (Display SET) command there is the CSET (Chromatogram SET) command. Generally, this command must be executed before every DISP of SIM/MRM data. CSET prompts for user input or a carriage return to retain the current values. It allows the user to specify which elements to display, to a maximum of five, and what time interval to use (the user may enter zero for the end time to default to the end of the run). Additionally, there is a command NEXT which allows the user to display a set of elements as a series of displays containing successive time window of constant width, without having to return to the CSET command. As is the case with sweeps and scans, archival storage of data and generation of high quality plots is performed on the PDP-ll/23 data system. Data may be uploaded to the data system by means of a 16-bit parallel link. Unlike scans and sweeps, however SIM/MRM data cannot be conveniently 47 incorporated into the MSU-LLNL Multi-Dimensional Database structure (23). This is because the data points cannot be represented as an ordered, x—y pair, as required by the multi-dimensional database. Instead, they must be repre- sented as ordered triples, composed of quadrupole 1 mass, quadrupole 3 mass, and intensity. Thus, a new set of utilities had to be written to process these data on the PDP-ll/23 data system. Data from the control system are uploaded to the data system as "experiments". Each experiment consists of some header information, user-supplied comments, and headers and data for one or more "scans", which are presumably related. Each scan corresponds to an MRM/SIM run and consists of a header record, the full set of parameter values for each element monitored, and one data record for each cycle (regardless of how many elements were actually monitored). The control system uploads one experiment at a time up the 16-bit parallel link to the data system. The receiving task on the data system, UPLOAD (24), creates a specially formatted file. The data system command GCTRAN is used to take the file created by UPLOAD, and separate it into one file for each of the scans in the experiment. Each one of the scan files contains the scan header, all of the comments for the experiment, the scan parameter sets, and all of the scan data. In addition, the user is prompted for any additional comments to go into each individual scan file as it is being generated. 48 A second data system command, MRM, may then be used for each of the scans to be processed. The current functions of the MRM routine are to generate a run listing and to generate a special file compatible with MULPLT, a locally developed graphics package (25). In addition to the MULPLT data file, MRM generates a command file which. may be executed by the user to automatically generate default reaction/ion chromatograms for each element which was monitored for the entire run. Each of these chromatograms will be separately normalized and will indicate the maximum for each element. The run listing generated by MRM consists of the date and time for the run, the parameter values used for each element, and the maximum intensity and corres— ponding time for each of the elements. Any data files transferred to the PDP-ll/23 data system may be archived by use of the general file utilities included in the RSX—ll operating system (26). Summary A flexible and easily learned set of programs has been created for use with a multiple microprocessor triple quadrupole mass spectrometry control and data system. The structure and user interface designed into these programs is consistent with the existing programs for sweep and scanning data acquisition which have been described in chapter 3. Additionally, the use of multiple processor allows 49 incorporation of parameter switching for each ion or reaction being monitored; this gives the user a powerful tool for optimized data acquisition for each element while maintaining a sampling rate which is high enough for accurate monitoring of capillary GC peaks. Chapter 5. The Relative Efficiency of Methanol Chemical Ionization for Selected Hydrocarbons and Heterospecies Present in Fuels Introduction A key element in any mass spectrometric analysis is the ionization process employed. Over the years, mass spectro- metrists have developed a large number of ionization techniques, each having advantages for one or another of the wide array of applications in which mass spectrometry is employed. One of the ways various ionization techniques may be classified is by the degree to which they fragment the sample molecule; another is by their selectivity towards different compounds and classes. A third classification is the technique’s efficiency; what fraction of the sample molecules can be converted into characteristic ions? Energetic ionization techniques, of which 70 eV electron impact _is the most common example, frequently result in extensive fragmentation of the sample molecule. High energy methods such as EI are often used to obtain universal ionization since they are capable of putting sufficient energy into most organic molecules to remove an electron. Putting so much energy into the sample molecule, 50 51 however, often comes at the price of creating virtually nothing but fragment ions and thus losing one of the most valuable single pieces of information on the sample molecule —- the molecular weight. The fragmentation patterns found in the El spectrum are highly characteristic of individual compounds and compound types. As a result, they can be used by a trained mass spectrometrist to gain valuable insight into the identity of an unknown molecule. One limitation of such a sample spectrum is that it often consists of a large number of relatively minor peaks. It is very rare that any of these peaks, taken individually, are selective for the individual compound. As a result, they are of limited utility for quantitation from complex samples by selected ion monitoring (SIM.) Thus, ionization techniques which cause large amounts of fragmentation are often inadequate when attemp- ting to quantitate trace components in complex mixtures. In cases where the electron impact spectrum does not yield molecular weight information or creates excessive fragmentation of trace species, chemical ionization (CI) techniques are often employed (27). Essentially, CI employs an indirect approach to ionization. A relatively closed ion volume is flooded with a reagent gas or gas mixture to a pressure typically on the order of l torr (measured inside of the source). This Vcreates a large excess of reagent compared to sample partial pressures which are generally on the order of a few millitorr. A high 52 energy (ca. 300 eV) electron beam is passed through the ion volume. Due to the large excess present, the electron beam essentially ionizes only the reagent gas. The ions thus formed can then react with other reagent molecules present. in the source to form what amounts to a steady-state reagent plasma. Sample molecules present in the source may be ionized through ion-molecule interactions in the reagent plasma. Depending on the choice of reagent gasses, the temperature, the pressure, and the nature of the sample, these inter- actions may include such processes as proton transfer, charge exchange, and hydride abstraction (28). The energy available for ionization and fragmentation by any of these processes depends primarily on the thermodynamics of the particular reaction for the involved species. The choice of reagents, pressure, and temperature allows the analyst to use ionization conditions which will have increased selectivity for the species of interest and, if desired, yield little or no fragmentation. As a result, molecular weight information, which may not have been available in the El spectrum, can be obtained much more readily. Further, by having fewer, more characteristic peaks in the spectrum, CI conditions can often be tailored to be highly selective and sensitive for trace compounds, even in complex mixtures. The primary ionization mechanism involved with many of the more common chemical ionization reagents (such as 53 methane, iso-butane, and ammonia) is proton transfer. In these cases, the energy transfer upon ionization of the sample is due to the difference in the proton affinities of the reagent ion, a protonated Bronsted base (primarily CHs*, Cqui*, and N114+ for methane, iso-butane, and ammonia, respectively) and the sample molecule. When this difference in proton affinities is such that proton transfer from the reagent ion to the sample molecule is exothermic, chemical ionization occurs and the MH+ ion is formed. The appearance of this ion at (M+l)+ in the mass spectrum (M being the molecular weight of the sample) can then be used to obtain the weight of the sample molecule. The difference in proton affinity between the reagent ion and the sample molecule determines the amount of excess energy deposited into the sample molecule upon ionization. This energy is. primarily released by collisional deacti- vation and/or fragmentation of the collisional complex of the sample molecule and reagent ion. The extent to which either of these processes is observed in the spectrum is dictated by their relative rates. Thus, if the analyst seeks to minimize fragmentation and concentrate the ion current into fewer, more abundant peaks, ionization conditions should be selected so as to minimize the amount of excess energy which must be released by the sample ion. This may be done by choosing ionization conditions which yield reagent ions with the highest proton affinity which will still allow efficient transfer of a 54 proton to the sample components of interest. This approach may be particularly valuable in cases where it is desirable to discriminate against low proton affinity components in the sample. One area where the above approach may be particularly useful is the quantitation of trace hetero-atom containing species present in fuels. The presence of such compounds in fuels, even at the parts-per-million or below levels, can have a profound impact on the storage stability and other properties of the fuel. Detection of such components in fuels using conventional ionization techniques is exceed- ingly difficult due to the vast differences in concen- trations between the hydrocarbon and heteroatom species. Even the very low abundance fragmentation products of the hydrocarbons, which appear at virtually every mass in the spectrum, dominate over the base peaks in the spectra of individual trace components. Further, the fuels matrix is sufficiently complex that even capillary GC/MS may be unable to resolve trace components from the background. The heterospecies, with lone pairs of electrons on the sulfur, oxygen, and nitrogen atoms, are characterized by high proton affinities. As a result, they can be suc- cessfully ionized using ”soft" proton transfer chemical ionization conditions. This makes it possible to enhance sensitivity for trace species by concentrating the ion current from a particular compound into a small number of peaks, generally dominated by the MH+ ion. In contrast, the 55 hydrocarbons, especially the saturates, are characterized by low proton affinities. Because of this, the "soft” proton transfer reagents, which work very well at ionizing the heterospecies tend to strongly discriminate against the hydrocarbons, which make up the bulk of the fuel matrix. One ”soft" proton transfer reagent which has been successfully employed in this laboratory is methanol (29). The first reported use of methanol chemical ionization, by Stan (30), was for the study of fragmentation patterns in the mass spectra of organo—phosphorus pesticides. Its use in that project was essentially abandoned due to the fact that it provided too little fragmentation for their studies. It was this lack of fragmentation which was suc- cessfully employed by Bauer (31) in developing screening methods for organo-phosphorus pesticides by MS/MS. Methanol, under CI conditions in the ion source of a mass spectrometer, primarily tends to form a series of proton bound cluster ions of the form (CHaOH)nH*. Under conditions employed in this work, the predominant values of n ranged from one to three. In addition, there is a second series of ions formed corresponding to the general formula [(CHsOH)nH* - 820]. This current study was undertaken to investigate the utility of methanol chemical ionization for the analysis of trace heterospecies in fuels. 56 Experimental Conditions This study employed an ELTQ-400—3 Triple Quadrupole Mass Spectrometer and Control System manufactured by Extranuclear Laboratories Inc., Pittsburgh, PA. Chemical ionization (CI) was employed using an Extranuclear SimulscanT“ ionizer maintained at 200 C. All of the work reported in this chapter used the instrument as a single stage mass spectrometer with the first quadrupole performing the mass filtering and the second and third quadrupoles operating in RF~only mode, tuned to transmit all masses. No collision gas was present in the second quadrupole. Chemical ionization was performed with methane (ultra- high purity) and methanol (anhydrous reagent grade). Methanol for CI was introduced from a 500 m1 expansion volume maintained at ambient temperature. Pressure of the CI reagent gas was monitored by means of Bayard-Alpert type ionization gauge in the source vacuum chamber. Comparison of methane chemical ionization spectra obtained on this instrument with those obtained on instruments with direct readings of the actual ion source pressure indicates that the pressure reading obtained on this gauge is approximately four orders of magnitude lower than the pressure inside of the ion source. Thus, the pressure values, on the order of 10‘4 torr, being reported here, roughly correspond to source pressures of about 1 torr. Electron energy for the chemical ionization was maintained at 300 eV. Sample introduction was achieved with a Hitatchi 633-30 Gas Chromatograph (Hitatchi Ltd., Tokyo, Japan) equipped with a split/splitless injector maintained at 200 C. Separations were achieved using a 30 meter narrow-bore (0.25 mm ID) DB-l chemically bonded-phase fused silica capillary column with a 1 micron coating (J & W Scientific, Rancho Cordova, CA.) The column was interfaced to the ion source by passing it through a transfer line at 250 C to a direct inlet. Helium carrier gas was used at a flow of 1 ml/min (a flow velocity of 40 cm/sec at 200 C.) The column was held isothermal at 200 C. Sample injections of 2 microliters were used, split 25:1. The test blend used in this study was prepared in reagent grade n-Heptane purchased from Mallinckrodt. Blend components were obtained either from Chemservice Inc. (West Chester, PA) or Aldrich Chemical (Milwaukee, WI.) All chemicals were used without further purification. Data interpretation was performed on a PDP-ll/23 mini- computer running the RSX-llM operating system after performing a l6-bit parallel data transfer from the instrument control microcomputer system. Full scan data were interpreted with the aid of a multi-dimensional database package (32). Selected ion monitoring (SIM) data were obtained at a rate of 5 samples per second for each ion; each intensity an average of 128 ADC readings. These 58 SIM data were interpreted with the aid of another locally developed software package (see Chapter 4). A 10 m1 blend of seven components in n-heptane was prepared for this study and its composition is summarized in Table 5.1. The seven compounds used, two hydrocarbons and five heterospecies, were chosen to represent a wide range of functionality and have boiling points in the range of a middle distillate fuel. The structures, molecular weights, and boiling points of these compounds are given in Figure 5.1. Comparison of Spectra under Methane and Methanol CI The first step in this study was to compare the full spectra of each of the components obtained using methane and methanol chemical ionization. The CI with methanol was performed at three different pressures. The comparison of methanol CI spectra at taken at different pressures can be used to sort out the effects of the change in reagent ion composition as a function of pressure. The full scan spectra were taken by repetitive scanning GC/MS of a 2 microliter injection from the test blend, split 25:1. Spectra were taken over the mass range of 42—180 amu at a nominal scan rate of 250 amu/sec. The individual spectra presented here were those scans, selected from the data set, which exhibited the maximum net ion counts for the (M+1)* Table 5.1 Compound n-Heptane Decalin l-Octanethiol n-Butylbenzene l-Phenylpyrrole Benzothiophene p-Methoxyphenol 2-Methylindole 59 Total MassZg 6. 0. 7 22 0769 .156 .342 -—-¢—.— 051 g in 10.00 ml 2M glit 15:1 49.7 0.615 0.574 0.694 0.234 Composition of Test Blend I_I_o_1_e_£ 91.0 0.816 3.74 0.299 0.948 0.840 0.802 /\/\/\ n-Heptone MW-100. BP-98.4C /\/\/\/\H l-Octonethiol saw-146. BP-198C Benzothiophene haw-1:54. BP-221 C Z-Methylindole uw-131. BP-273 60 Decofin haw-1:58. BP-190C n-Butylbenzene uw-134. BP-1BJC ”O OCH3 p-Methoxyphenol saw-124. 8P-243C 1—Phenylpyrrole BP-234C Figure 5.1 Compounds for Methanol Cl Study 61 peak of the corresponding mixture component. Background subtraction was used to remove the CI reagent peaks. Spectra Obtained Using Methane Chemical Ionization The first set of spectra, shown in Figures 5.2a and 5.2b. were taken using methane chemical ionization at a total pressure of 2.0 X 10'4 torr measured in the source vacuum chamber. This pressure reading includes a base pressure of approximately 2 X 10‘5 torr, primarily due to a flow of approximately 1 ml/min (at atmospheric pressure) of helium from the gas chromatograph. Under these conditions, the spectra of all eight compounds show significant amounts of molecular ion formation in addition to the expected MH+ ion. Addition- ally, the hydrocarbons (n-heptane, n-butylbenzene, and decalin) and l-octanethiol show extensive fragmentation, with little difference from their reference EI spectra. These observations, taken with the high concentration of helium in the ion source suggests that charge transfer from He+ (ionization potential 24.587 eV (33)) is Competing with proton transfer from CH5‘. Charge transfer ionization would transfer an amount of energy approximately equal to the ionization potential of the reactant ion to the reactant molecule. In the case of He’, this would amount to 24.587eV and would be expected to generate spectra similar to that observed in El with an Intensity Intensity Intensity Intensity 100 60 20 100 8 20 1 00 BO 60 40 20 1 00 BO 60 40 20 0 62 n-Heptone (mw = 100) 1 wt '1 f'jf' U‘UTr'j—T 31"? ‘I T l T l U ' I l V ' I I Y I U T U I n-Butylbenzene (mw = 134) Decalin (mw MH+ .. I. IWII'TII'YIIII li‘j'll'lUTU'liijrjtj Octanethiol (mw = 146) 1 ML H+ IIdl-FHA#rH-TnF-rWUH-T-i-FH'J:'H v u: l v 141 Ill! r—' = 138) I I fl I 2.0 I I I ' I 1 I l T U r I 40 60 80 100 M/Z Figure 5.20 Methane Cl L IU'UYI'T'jj 120 140 160 180 200 2.0 x 10"4 torr Intensity Intensity Intensity Intensity 1 00 80 60 20 100 so so 40 20 1 00 BO 60 4O 20 1 00 80 60 4O 20 O 63 p—Methoxyphenol (mw - 124) i L [11-1 'III'IITITITI'III—I'III'IIrj IIIIII‘III'III' Benzothiophene (mw - 134) | uH+ I I I I '1 r I I I Ii r1 1-Phenylpyrrole (mw = 143) uH+ j l l I—rII'Irf'III'fiIIIIWIf11TWI—T‘IjIT'TIT' 2-Methylindolw (mw = 131) 1 MM :T I—T 2.0 I I I ' LI I I I “I I TA I—I I I I I f1 1 I I l I I IT I I I I I I j 1 40 so so 100 120 140 160 180 200 M/Z Figure 5.2b Methane CI 2.0 X 10-4 torr 64 equivalent electron energy. Such spectra would be expected to have enhanced molecular ion abundances relative to 70eV EI spectra with somewhat reduced fragmentation (34). This is very similar to what was observed in these methane CI spectra. An alternate explanation for the appearance of the M* peaks would be that the peak-findeer in the data system was splitting up the MH* peaks. This explanation, however, fails to account for the extensive fragmentation observed in several of the spectra. The n-heptane exhibits the typical alkane peaks at 43*, 57* (base peak), and 71*, with smaller amounts of (M-l)* and M*. The small amount of (M+1)* can be mostly attributed to the 13C isotope. N-Butylbenzene does show a significant (M+l)* peak, but this is still smaller than the M* peak. The base peak in the spectrum is due to the butyl ion (57*) with other major peaks at 91* (tropyllium) and 92*. Decalin shows a large (M+l)* peak, but again it is smaller than the M* peak. There is, however, considerably reduced fragmen- tation, presumably due to stabilization from the fused ring structure. The spectrum of l-octanethiol is dominated by the hydrocarbon chain, showing the characteristic alkene pattern of peaks at 43*, 55*, 57*, 69*, and 71* (base peak). This suggests the elimination of H28 as a first step in the fragmentation process. That possibility is also supported by the presence of a significant peak (ca. 25%) at 113*, corresponding- to [(M+l) - HzS]*. This domination by the hydrocarbon peaks is similar to that observed in the El 65 spectra of the straight-chain primary thiols (35). There are also significant peaks at (M—l)*, M*, (M+l)*, and a (M+29)* adduct peak at 175*. The other four compounds studied, p-Methoxyphenol, Benzothiophene, l-Phenylpyrrole, and 2-Methy1indole, as shown in Figure 5.2b, all exhibit higher relative abundances of (M+l)’. The presence, in all four cases, of a large M* ion again suggests that helium charge transfer is competing with proton transfer. Each of the four also shows an adduct peak at (M+29)*. Further, the 2-Methylindole shows a peak of 453 relative abundance at (M-l)*, possibly due to loss of the nitrogen-bound proton. The fragmentation of these heterospecies is much less than that observed for the hydro- carbons. Much of this effect can be attributed to the charge stabilization afforded by the unsaturated ring structures. However, the ion current is still being divided between at least two major peaks in each spectrum. Spectra Obtained Using Methanol Chemical Ionization These methane CI spectra provide a basis for comparison with spectra obtained from the same compounds, under similar instrument conditions, and using methanol at three different pressures as the reagent gas. The three spectra in Figure 5.3 illustrate the ion composition of the reagent gas plasma at three different source chamber pressures, namely 6.5 X 10‘5, 2.0 X 10“, and 4.5 X 10'4 torr. The first pressure Intensity Intensity Intensity 100 80 60 20 100 60 60 20 100 80 60 20 66 6.5 x 10"5 torr + - CHBOHZ d - 4 - . + q l (CH30H)2H T'IIII'VIII'III'IIIrIIIrII—I'III'IIT' CHJOH2+ 2.0 X 10—4 10" q p q cl ‘ ‘ (ca3ou)2H* ‘ IIIIflII'ITI' II‘rI rI'IIrTI—rr'rII'IIvI'I—I I1 614301-12“ 4.5 X 10'"4 torr - l - (CH30H)2H+ d d l . ‘ - . (CH30H)3H* III'III'I’III' IIrIrI'rII'III'II—FIIII'IrI' 0 20 40 60 80 100 120 1 40 1 60 1 80 200 M/Z Figure 5.3 Methanol Reagent Ions 67 was chosen to maximize the absolute intensity of the 33* (CHsOHz*) peak while minimizing the contribution of 65* [(CHaOH)2H]*. There is a small peak (ca. 3% relative abundance) at mass 47*, representing loss of water from 65*. The third pressure, 4.5 X 10" torr, was chosen to maximize the contribution due to the 65* ion. Under these conditions, 65* has a relative abundance of about 60% and 47* has a peak of roughly 30%. A small peak (ca. 3%) at 79* represents loss of water from the methanol cluster ion at 97*, [(CHaOH)3H]*. The second pressure, 2.0 X 10" torr, was chosen as an intermediate between the first and the third. At this pressure, the peaks at 47* and 65* are found to be of roughly equal relative abundance, 25%. These spectra, taken at a source temperature of 200 C and with a helium flow of 1 ml/min from the GC column show significantly less clustering of the methanol than in the work of Bauer (36). In that study, which employed a source temperature of 100 C and no helium flow, the methanol pressure was adjusted so as to maximize the absolute intensities of the 65* and 97* ions. This resulted in a spectrum with the base peak at 65* and the 33* and 97* ions having approximately equal relative abundances at about 80%. The source conditions in the present study were dictated by the need to optimize within conditions appro- priate for GC/MS. The reduced abundances of the higher methanol clusters at elevated temperatures is reasonable 68 considering the weakness of the hydrogen bonds which bind the clusters. Figures 5.4a and 5.4b present the spectra of the blend components obtained with methanol CI at 6.5 X 10‘5 torr (the lowest pressure). Generally reduced fragmentation, relative to that observed with methane CI, is observed, even with this relatively low methanol pressure. As was the case in the methane CI spectra, the extent of fragmentation and the presence of M* ions in several of the spectra suggests that helium charge transfer ionization is again competing with proton transfer, this time from the methanol reagent ions. This is consistent with the high helium concentration resulting from the use of a low methanol pressure. The appearance of the n-heptane spectrum is virtually unchanged from that observed with methane. The additional peak at 65* can be attributed to a residual reagent gas peak which was incompletely removed in the background subtraction. The n-butyl benzene shows greatly reduced fragmentation. The butyl peak at 57*, which was the base peak with methane, has been virtually eliminated, and (M+l)*, at 135* has taken its place. The relative abun- dances of the 91* and 92* ions have also been reduced by about 30-40%. The decalin spectrum shows a greatly reduced M*, in favor of (M+l)*, but the large number of fragment ions are essentially unchanged. l—Octanethiol shows a greatly enhanced (M+1)* ion, which is now the base peak, but the fragment ions, other than their reduced relative Intensity Intensity Intensity Intensity 100 80 60 40 20 100 80 80 20 100 80 60 40 20 100 80 80 4O 20 O n-Heptone (mw =- 100) 1 . l 1 .lL“H+ 'rfiI rfiI—f'TrAI'IIr'fi'Ir'IIII’rjr'III'IIIrIrI] n-Butylbenzene (mw :- 134) 1 ill-1* -l I .. A l I I I II I I II I I 'I rT'T fII I I 'I I rfrrrrTI 'TFI‘. Decalin (mw a 138) 1 A- 'III'III'II IIIrIIr'III‘I Octanethiol (mw = 146) “Hi. 'rII'III'II I'IIIrIII—' O 20 4O 60 80 100 120 140 160 180 200 M/Z Figure 5.40 Methanol on 6.5 x 10-5 torr Intensity Intensity Intensity Intensity 1 00 80 80 20 1 00 80 40 20 1 00 80 60 20 1 00 80 80 40 20 0 70 p-Methoxyphenol (mw - 1424) l - L L 'IrI'III—rrIIrIII'IIfrIII' rI'IIf'IrI'ITr' Benzothiophene (mw =3 134) El uH+ 1L II A i T [II—rrrII'Irr' rTr'III'IIIrIII'III'IrIfl'fiI’ 1-Phenylpyrrole (mw - 143) 1 MH TIfT'IIfi'III'IfIthiI—I}IHII ITIIIrfIr' 2-Methylindole (mw 131) uH* rfrrI—rITITFI'AIIIJrIUTILITIII I'III'III—IIIII o 20 40 so so 100 120 140 160 150 200 M/Z Figure 5.4b Methanol CI 6.5 X 10"5 torr 71 abundance, remain the same. The spectra of the four heterospecies in Figure 5.4b are very similar to those obtained with methane except that the (M+l)* ions are enhanced relative to M*. For p-Methoxyphenol, the M* ion at 124* is completely eliminated. These differences from the methane CI spectra observed for the polar compounds and n-butylbenzene occur despite the high concentration of helium in the ion source. This suggests that the efficiency of proton transfer from the methanol reagent ions to these compounds is higher, relative to helium charge exchange, than is the case for the non- polars. Increasing the methanol pressure to 2.0 X 10‘4 torr, as illustrated in Figures 5.5a and 5.5b. generally results in further reduced fragmentation. The exceptions are the n-heptane and the decalin. The spectrum of n~heptane is essentially unchanged once more, and the decalin actually shows increased fragmentation. All of the heterospecies, including l-octanethiol, as well as the n-butylbenzene, show increased relative abundances for (M+l)*. The M* peaks are now eliminated from the spectra of benzothiophene, l-phenyl- pyrrole, and 2-methy1indole. The third set of methanol CI spectra, taken at a pressure of 4.5 X 10“ torr, are presented in Figures 5.6a and 5.6b. In this set, the fragmentation of decalin has undergone further increase. Additionally, n~butylbenzene is 1 HI‘ I... n—Heptane (mw =- 100) 100 80 60 Intensity 20 o g ,4_ ,A lguH* ‘rfr'rrrrv IjtrIIrIIIIIrIrIfiI'I’Ir'rII'IIr' n-Butylbenzene (mw =- 134) 100 us* 80 80 20 Intensity o 3 LrF‘FFF‘n-rr'rfi Decalin (mw a 138) 100 80 60 20 Intensity C: 8 z I + Octanethiol (mw = 146) “H... 100 , 80 80 20 W 80 100 120 140 160 180 200 M/Z Intensity C) 8 O N O 3 8 Figure 5.5a Methanol CI 2.0 x 10"4 torr Intensity Intensity Intensity Intensity ‘1 CA: p-Methoxyphenol (mw - 124) 100 MH* 80 60 20 l O1r!r'trtTrrfi'rrr'tivIrrr'uTrrrrvrrrvtitr' Benzothiophene (mw =- 134) 100 Ml-l* so so 40 20 1 IL. rIT'IIT'IrT' o 'IT—I'II—rrIII'frI'IIIrrI—rrIr 1-Phenylpyrrole (mw = 143) 100 NH" so so 40 20 l o A. [TIItrTI'III'II'I'IIIII'IIII'III'IIItIII'III' 2-Methylindole (mw = 131) 100 MH* so so ‘ 4o 20 L l r o IrI'III'III'ITI'IIIrrII'III'IIII—III'TII' 20 40 60 80 100 120 140 180 180 200 M/Z Flgure 5.5b Methanol CI 2.0 x 10" torr Intensity Intensity Intensity Intensity 100 80 80 20 100 80 80 20 1 no so so 20 100 80 60 20 ~31 .Cz. n-Heptone (mw =- 100) l 1 .luni' 'IIr'IrTTrLI}rI I'ITII'III'IIItIIT'rfirtrfT' I n—Butylbenzene (mw =- 134) MH* All 1 rIrItI’IIfIII'rIr'II [rII'II 'TIIrIII'IrW' Decalin (mw =- 138) M+ IFT'IrI'TI‘fi' Octanethiol (mw = 146) “H... l W..Tl..r' 40 60 80 100 120 140 160 180 200 M/Z Figure 5.6o Methanol CI 4.5 x 10‘4 torr Intensity Intensity Intensity Intensity 100 80 80 40 20 100 80 60 40 20 100 80 80 40 20 100 80 80 20 1 75 p—Methoxyphenol (mw - 124) Mll* 'III'IrI'IIfTI—I’IirIIIIA'IIII rr‘I—rIlIrITIII' l l Benzothiophene (mw =- 134) lufl'l' l A L i L I—IIIrIrfir'IIItIIIrII—r'IrrTT—Ifi 'III—rfiI’I'IIfi‘ 1—Phenylpyrrole (mw = 143) MH* L 'fiIItTII'Ij’I'Irr'IIrrIIr'IIT'lrIrIII—rrII' 2-Methyllndole (mw - 131) 1 MH" fT I r I I I ' I I I r I I I I I I I I I I I I I I r t I I I r I I I ' I I I l O 20 40 BO 80 100 120 140 150 180 200 ”/2 Figure 5.6b Methanol CI 4.5 x 10‘4 torr 76 now showing increased fragmentation. The normalized spectra of all the other compounds remain essentially unchanged. In summary, increasing the methanol pressure results in greatly reduced fragmentation and elimination of the M* peaks. This may be attributed to an effective dilution of helium in the ionization region and the fact that higher methanol pressures increase the probability that any helium ions formed would collide with, and ionize, methanol molecules rather than sample molecules. The fact that this is not observed for l-octanethiol, heptane, and decalin suggests that these compounds are not very sensitive to methanol CI, possibly due to low proton affinities, and, instead, the bulk of the ionization of these molecules is the result of helium charge transfer. Discussion of Full Scan Chemical Ionization Spectra Comparison of the spectra taken under methane CI with those obtained using methanol CI show significant differ- ences in behavior for the different compound classes studied. The observed differences all appear quite reason- able considering the diversity of the eight compounds under study. The spectra of n-heptane and decalin are the least affected by the choice of ionization conditions. Changes in the spectral patterns observed for these two compounds may be explained by poor proton transfer efficiency from the 77 methanol reagents ions to these saturated hydrocarbons. As a result, charge exchange ionization from helium (present in the ionization region from the gas Chromatograph) appears to be the dominant ionization mechanism. This effect is particularly noticeably at the highest methanol pressure (i.e. the lowest helium concentration) where the spectra of the polar compounds show virtually no fragmentation and the non-polar compounds undergo extensive fragmentation. The third hydrocarbon studied, n-butylbenzene, shows a significant change under all of the methanol CI conditions studied. At all three methanol pressures, the n-butyl- benzene shows enhanced abundance of the MH* ions and reduced abundances for M*, (M-H)*, and fragmentation products. This difference in behavior from that observed for n-heptane and decalin can probably be attributed to the stabilizing affect of the aromatic ring. There 'appear to be no significant differences in the spectral patterns obtained using the three different methanol pressures. Perhaps the most interesting compound studied is l- octanethiol. It’s spectra contain both a fragmentation pattern typical of an alkene and an enhanced abundance of the MH* ion, as would be expected for a hetero—atom containing species. Switching from methane to methanol CI retains the alkene fragmentation pattern, but at reduced relative abundance, while the increase in the abundance of the MH* ion is such that it becomes the base peak. The contributions of M* and (M-H)* are both greatly reduced. 78 Again, there are no major differences in the spectral patterns obtained under the three different methanol pressures. The remaining four compounds in this study —- p-methoxyphenol, benzothiophene, l-phenylpyrrole, and 2~methylindole -— all show very similar trends. Generally, they all exhibit minimal fragmentation under any of the four ionization conditions studied. The one major trend is a reduction in the relative abundance of the M* ion from being approximately equal to the MH* intensity under methane CI, to a reduced abundance in the low- and medium-methanol CI spectra, and finally being eliminated in the high-pressure methanol spectra. This may be explained as a reduction in the occurrence of charge exchange ionization as the concen- tration of helium in the ion source is effectively reduced. In all four cases, the high—pressure methanol spectra are virtually fragment free and show only the MH* ion and the related isotope peaks. In general, methanol CI is observed to result in reduced fragmentation relative to the spectra obtained using methane CI. This effect holds true for n—butylbenzene and all of the heterospecies studied. Due to the apparent occurrence of helium charge exchange ionization in this work, the spectral patterns for the saturated hydrocarbons appear basically unaffected by the choice of ionization conditions, though their sensitivities appear to be reduced under methanol CI conditions. 79 The reduced sensitivity of methanol chemical ionization for non-polar compounds can be used to discriminate against the hydrocarbon matrix in a fuel sample, thereby facil- itating the study of heterospecies. As a result, the costly and time-consuming preliminary separations required by less selective techniques can, potentially, be avoided. Further, the reduced fragmentation observed for the heterospecies can be used to simplify the spectra of fuels, yielding a single characteristic ion for each heterocompound of interest. Relative Sensitivities of Methane and Methanol CI After examination of the normalized spectra of the blend components under different ionization conditions, the next area for examination was the relative sensitivities of the different ionization techniques. As discussed earlier, an experimental goal was to find ionization conditions which maximize sensitivity for heterospecies which may be present at trace levels in fuels, while minimizing contributions to the ion current due to the hydrocarbon species which make up .the bulk of the fuel matrix; in other words, to develop a technique that is selective for the heterospecies. To obtain this information, selected ion monitoring (SIM) was employed. Peak areas were determined for the (M+l)* ions from each of the components in the previously described test blend. The choice of ions was made on the basis that both methane and methanol CI were observed to 80 produce ions at (M+l)* for all of the compounds in the test blend. Conditions for the GC were identical to those previously described for the full scan experiments. Six ions were simultaneously monitored, representing the (M+l)* ions for each of the mixture components (n-butylbenzene and benzothiophene are isobaric with molecular weights of 134, but are readily distinguished on the basis of their reten- tion times.) Triplicate runs were made at each of the four ionization conditions described above. Observed retention times are reported in Table 5.2. To compare the relative sensitivities of the trace heterospecies under these ionizing conditions, n-butyl~ benzene was chosen as a reference compound. This choice was made because n-butylbenzene is expected to be present in middle distillate fuels at the high parts per million level, it was observed to yield large relative abundances of (M+l)* with all of the ionization conditions employed in this study, and it produced more reproducible spectra than could be obtained for the alkanes (which show very poor sensi- tivity with methanol CI). Thus, measurement of the ionization efficiency for each of the heterospecies, relative to that observed for n-butylbenzene, will be an indicator of the selectivity of a given ionization technique for heterospecies in a hydrocarbon matrix. Results obtained using this approach are reported in Tables 5.3 - 5.6. They are presented in terms of the ratio of the peak area of each component to that obtained for n-butylbenzene for the 81 Table 5.2 Retention Times of Blend Components Compound Retentiog Time (gec) n-Butylbenzene 98.2 i 0.3 l-Octanethiol 105.4 1 0.3 Decalin 110.6 1 0.3 p—Methoxyphenol 113.5 : 0.2 Benzothiophene 123.2 f 0.3 l-Phenylpyrrole 135.6 1 0.2 Z-Methylindole 161.5 f 0.3 82 Table 5.3 Sensitivities for Several Heterospecies Relative to n-Butylbenzene with Methane CI at 2.0 X 10“ torr App; Corrgcted Senpitivity Compound u in' 33312 32112_ (areang) n-Butylbenzene 2.74 1.00 l 0.3 1 0.2 l-Octanethiol 0.16 0.1210.01 2.1 1 0.1 0.6 1 0.4 p~Methoxyphenol 0.071 0.6 1 0.1 23 1 4 7 1 l Benzothiophene 0.087 2.1 1 0.6 70 1 20 20 1 5 l-Phenylpyrrole 0.029 1.4 1 0.4 130 1 40 40 1 10 2~Methy1indole 0.072 3 1 1 100 1 40 30 1 10 83 Table 5.4 Sensitivities for Several Heterospecies Relative to n-Butylbenzene with Methanol CI at 6.5 X 10“5 torr 1553 Corrected Sensitivity Compound u in' £2112 £3112 (arealpg) n-Butylbenzene 2.7 1.00 1.00 0.9 1 0.3 l-Octanethiol 0.16 0.1910.01 3.3 1 0.2 2.9 1 0.2 p-Methoxyphenol 0.071 0.1810.01 6.8 1 0.4 5.9 1 0.4 Benzothiophene 0.087 0.4510.03 14.3 1 0.1 12.5 1 0.1 l—Phenylpyrrole 0.029 0.3810.03 36 1 3 31 1 3 2-Methylindole 0.072 0.5210.04 20 1 2 l7 1 2 84 Table 5.5 Sensitivities for Several Heterospecies Relative to n-Butylbenzene with Methanol CI at 2.0 X 10“ torr 1552 Corrgcteg Sensitivity Compound u in' 33112 32112 (arealpg) n-Butylbenzene 2.74 1.00 1.00 1.3 1 0.3 l—Octanethiol 0.16 0.2910.01 5.1 1 0.2 6.5 1 0.3 p—Methoxyphenol 0.071 0.7710.07 30 1 3 39 1 4 Benzothiophene 0.087 0.8510.05 27 1 2 35 1 3 l-Phenylpyrrole 0.029 1.8 1 0.1 70 1 10 90 1 10 2—Methylindole 0.072 2.8 1 0.1 105 1 4 135 1 5 85 Table 5.6 Sensitivities for Several Heterospecies Relative to n-Butylbenzene with Methanol CI at 4.5 X 10“ torr 1553 Corrected Sensitivity Compound u in' £211p_ £5112 (arealpg) n—Butylbenzene 2.74 1.00 1.00 0.26310.006 l-Octanethiol 0.16 0.20810.007 3.7 1 0.1 0.97 1 0.03 p-Methoxyphenol 0.071 l.7310.07 67 1 3 17.6 1 0.8 Benzothiophene 0.087 1.2910.03 40.7 1 0.9 10.7 1 0.2 1~Pheny1pyrrole 0.029 5.3 1 0.2 500 1 20 130 1 5 2-Methy1indole 0.072 9.6 i 0.5 370 : 2b 97 t 5 86 amounts indicated and also the sensitivity in peak area per picogram of component. This approach minimizes uncer- tainties in the data resulting from fluctuations in instrument sensitivity due to factors other than the ionization conditions and serves to emphasize how the use of selective ionization conditions can enhance sensitivity for trace components in complex mixtures. The mass corrected ratios are a measure of the selectivity of each set of ionization conditions because they directly compare the signal obtained for each analyte to that obtained from an equal weight of a typical interference, with all other analysis conditions being held constant. Discussion of Selected Ion Monitoring Data The area ratio approach to sensitivity comparison becomes somewhat distorted due to the major variations in the fragmentation pattern of n-butylbenzene across the four different ionization conditions employed. It is, none- theless, a valuable approach considering that quantitative studies are generally done on the basis. of peak areas obtained by selected ion monitoring of one, or perhaps two, ions for each compound of interest. As a result, however, these data are best evaluated by comparing the trends across ionization conditions for each of the heterospecies in the blend. This is done graphically in Figure 5.7. Area Rotlo to n-Butylbenzene 87 1000.0 - E- 111111 III a l- . . 1-Phenylpyrrole 100.0-l :- 2-Methyllndole 0 Benzothlephene r I'TITIT— J _' L D p-Methoxyphenol 10.01 :- 2 I .1 /\ l. .l p A 1-Octonethlol ‘°° I 1 F1 r I I 1 l .0 100 200 300 400 500 Methanol Pressure (microtorr) Methane CI 200 mlerotorr Figure 5.7 Selectlvlty Factor vs. Ionization Mode 88 Observing the trends for the five heterospecies studied shows that l-Octanethiol is the only one to increase in selectivity factor going from methane to low pressure methanol CI. The selectivity factors of the other four heterospecies are seen to actually fall by factors of roughly three to five under that change of conditions. It may be recalled that under these two ionization conditions, the spectrum of l-octanethiol was observed to be dominated by the alkyl portion of the molecule, showing an alkane fragmentation pattern, while the other compounds were showing very little fragmentation. The fact that the selectivity factors of the heterospecies actually declined under this change in ionization conditions can be attributed primarily in the dramatic increase in the relative abundance of the (M+l)+ ion for n—butylbenzene from about 10% of total ionization with methane to almost 503 under low pressure methanol CI. Further examination reveals that the selectivity factors of the five heterospecies studied show a general increase with increasing methanol pressure. These increases are quite dramatic for the p-methoxyphenol, l-phenylpyrrole, and 2-methylindole for which the selectivity factors increase approximately 10, 14 and 18-fold, respectively, in going from the low to the high methanol pressure. In contrast, benzothiophene shows a much less pronounced increase in relative sensitivity with increased methanol pressure, going from 14.3 at the low pressure to 89 40.7 at the high pressure - an overall improvement of less than threefold. It can be further noted that even at high methanol pressure, relative sensitivity is lower than that which was obtained using methane, the only one of the five heterospecies studied to behave in this manner. l-Octanethiol also shows a trend different from the other heterospecies studied. Its selectivity factor is very low and exhibits relatively little change, with corrected area ratios of 3.3 at low pressure and 3.7 at high methanol pressure. The selectivity factor of l—octanethiol is actually greatest at the middle methanol pressure, with a value of 5.1. As Figure 5.7 shows, l-octanethiol exhibits the lowest relative response of all the heterospecies studied at under all four ionization conditions employed. The anomalous behavior of l-octanethiol, however does not seem unreasonable in light of the full mass spectra, examined earlier, which were dominated by the alkyl portion of the molecule. The selectivity of an ionization technique, however, is of little use if that technique is not sufficiently sen- sitive to produce a reliable signal at the concentration range of interest. Tables 5.3 - 5.6 also present the sensitivity of each of the four ionization conditions studied for each of the five heterospecies in the test blend. These numbers were obtained by calculating the product of the sensitivity of n-butylbenzene and the corrected area ratio for each of the heterospecies. To 90 better illustrate the trends across the different ionization conditions, these values are plotted in Figure 5.8. Four of the five heterospecies studied show reduced sensitivity in going from methane to low pressure methanol as the chemical ionization reagent. This can be attributed to fragmentation resulting from helium charge exchange ionization. l-Octanethiol, which shows a significantly increased relative abundance of the MH* ion under low pressure methanol CI conditions, is the one exception. It shows nearly a five-fold gain in sensitivity in going from methane to low pressure methanol CI. Increasing the methanol pressure from 6.5 X 10’5 to 2.0 X 10‘“ torr, as measured in the source vacuum chamber, results in major increases in sensitivity for all five heterospecies. This can be attributed to a general increase in the overall ionization efficiency, illustrated by a nearly 50% increase in the sensitivity for n—butylbenzene. Recall too, that all five of the compounds exhibit increased selectivity in going to the medium pressure methanol. The medium pressure methanol is the most sensitive ionization condition for .all but one of the compounds studied, the one exception being l-phenylpyrrole. With the exception of l-phenylpyrrole, all of the heterospecies showed a major loss in sensitivity as the methanol pressure was increased to the highest value used. This can be primarily attributed to reduced ion extraction efficiency resulting from the high source pressures Sensltlvity (area / pg) 53 1 1aaa.a- ,- 1 . ‘ f' < P < r- 4 n < +- 1oo.o-; P L" 1 Z d I- .. r. 1-Phenylpyrrale ‘ ' l’ 1 Z-Uethyiindole < ’ 0 Benzothiophene 10.01 3' i C D p-Hethoxyphenol 4 K r 1.0 1 {5' 1 L j t A i-Octanethiol I P . b O n-Butyibenzene 1 3 °-‘ r r r fir r r r r O 100 200 300 400 500 Methanol Pressure (microtarr) Figure 5.! Sensitivity vs. Methane Cl 200 microtorr Ionization Mode 93 Summary and Conclusions This study examined the methanol chemical ionization mass spectra of compounds representing a wide range of functionality and representative of species which may be present in a middle distillate fuel. These spectra exhi— bited reduced fragmentation when compared to spectra of the same compounds obtained with the more conventional technique of methane chemical ionization. The degree of fragmentation using methanol CI was observed to generally decrease with increased methanol pressures, where proton-bound clusters of methanol are present. The optimum methanol pressure varied for the different compounds studied, with higher optimum pressures observed for the more polar compounds. In general, the overall optimum pressure was found to occur at levels at, or somewhat higher, than those generally employed in this laboratory for methane chemical ionization. Additional studies examined the selectivity and sensitivity of methane and methanol chemical ionization for the MH* ions of the same compounds studied above. Under these conditions, the selectivity factors of all five heterospecies, measured relative to that of the represen- tative hydrocarbon, n-butylbenzene, increase with increasing pressure of the methanol CI reagent. For all but l~octane— thiol, 'the highest relative selectivity under methanol CI conditions, is measured at the highest methanol pressure 94 studied. The selectivity factor for all of the hetero— species studied, except for benzothiophene, is higher at the maximum methanol pressure than that observed using methane CI. All five of the heterospecies studied exhibit higher sensitivity at the middle methanol pressure than with methane as the CI reagent. The ionization process employed is a crucial factor in determining the overall selectivity and sensitivity of techniques such as GC/MS, MS/MS, and GC/MS/MS. The power of multi-dimensional methods for trace analysis depends upon each independent dimension contributing some selectivity for the species of interest. Small selectivity gains can often be made without requiring severe sensitivity tradeoffs. The combination of even small enhancements, however, allows the analyst to tailor a highly selective technique while retaining good sensitivity. This study has quantitated the selectivity versus sensitivity tradeoff resulting from the use of methanol chemical ionization for the analysis of five diverse heterospecies which may be present in a middle distillate fuel. .The optimum methanol pressure to employ in such an analysis in a fuel is largely dependent upon which addi- tional analytical dimensions are also available. In low resolution GC/MS, for instance, background from unresolved mixture components is often a limiting factor. In such a case, the use of high-pressure methanol, with its very high selectivity, is indicated, especially for compounds such as 95 2—methylindole and l—phenylpyrrole. The analyst must realize, however, that this increase in selectivity is coming at a stiff price in terms of sensitivity. The use of additional selectivity ‘elements such as highuresolution chromatography, high-resolution mass spectrometry, or tandem mass spectrometry serves to lower the selectivity demand placed on the ionization process. As a result, the analyst could then use the medium pressure methanol CI conditions which, though less selective, are much more sensitive. For a targeted analysis, the methanol pressure can be fine- tuned along with other instrument elements employed to allow the maximum overall selectivity and sensitivity. Such an optimization of the selectivity can greatly reduce inter— ferences from the hydrocarbons which compose the fuels matrix, thereby allowing the heteroatom species to be analyzed without the need for time-consuming and costly preliminary separations. Thus, methanol chemical ionization provides a sensitive and selective technique for the analysis of selected trace heterospecies present in a low polarity matrix, such as a fuel. Chapter 6 The Use of Multiple Reaction Monitoring Gas Chromatography/Triple Quadrupole Mass Spectrometry for the Detection of Trace Components in Jet Fuels A number of studies have shown that the storage stabil- ity and thermal stability of jet aviation fuels can be greatly affected by low levels of heteroatom-containing species (36, 37, 38, 39, 40). The storage stability of a fuel is determined by the extent of polymerization and sediment formation upon long term exposure to temperatures in the range of 50-125 C. Thermal stability is the degree of deposit formation resulting from short term exposure to temperatures in the ZOO—400 C range such as in the hot section of a jet engine. These problems have considerable economic and safety implications in the aviation industry and are expected to become increasingly serious in the future as lower quality crude oils and non—petroleum sources (e.g., shale oil and coal liquefaction products) enter the hydrocarbon pool used in the refining of these fuels. Jet fuels present a complex hydrocarbon matrix containing large numbers of isomeric and isobaric species spanning a wide range of concentrations. As a result, at unit mass resolution, it is virtually impossible to obtain 96 97 unique peaks in the electron impact spectrum; one generally finds a peak at virtually every mass. The use of full— scanning GC/MS greatly simplifies the mass spectra and thus their interpretation, however, even with high resolution capillary chromatography, the time resolution of individual components in the bulk of a fuel sample is incomplete. This is especially true when relatively large sample injections are required to enable detection of low-level components. Further, the peaks which are not resolved are likely to represent isomeric and isobaric compounds whose spectra are especially difficult to deconvolute, or trace components which are lost in competition with the matrix. The use of chemical ionization can serve to reduce much of the general background and increase sensitivity by concentrating the ion current into fewer peaks. This sensitivity, however, is obtained at the cost of the structural information necessary to distinguish closely related compounds present in the sample and likely to be co-eluting. Triple quadrupole mass spectrometry (TQMS), coupled with capillary gas chromatography, provides a powerful tool for addressing these characteristic problems of complex mixture analysis. Selective reactions (parent ion/daughter ion pairs) may be multiplexed in time over the course of a chromatographic run to provide for the simultaneous and independent analysis of co—eluting peaks in different data dimensions. Selective reactions also have the advantage of reducing the requirements for chromatographic resolution, 98 effectively allowing for larger sample injections and more rapid temperature programming. Additionally, the structural information available from the CAD spectra in TQMS allows chemical ionization to be used while retaining a mechanism for obtaining structural information. The limited selectivity of mass spectrometry for trace quantities of heteroatom-containing species in complex hydrocarbon mixtures has, in the past, required a presep- aration of the aromatic and polar species from the aliphatic constituents which make up the bulk of fuels. In addition to being a time-consuming process, preseparation provides many opportunities for altering the distribution of trace components in a sample. In this study, the power of capillary GC/TQMS has enabled the detailed analysis of selected trace heteroatom- containing species in jet fuels without any sample pretreat- ment. The thiophenes were selected as the targeted compound class for this research because they have been previously reported in jet fuels (41) and there is evidence of their participation in the thermal degradation of fuels (42, 43). Background Batch MS/MS has been previously reported for the anal- ysis of thiophene in gasoline (44) and middle distillate fuels (45) and for the determination of organosulfur compounds in crude oil fractions (46). In the first study, 99 McLafferty and Bockhoff claim a detection limit of < 25 ppm using 70 eV EI on a MIKES instrument. The authors also suggest that GC/MS/MS could be a valuable technique for the analysis of high complexity mixtures. Myerholtz, in the second study, reports the use of a single neutral loss scan for the rapid screening of several thiophenes in middle distillate fuels using TQMS and 20 eV El. This study did not report any quantitative information or attempt to identify individual thiophene components. The third study, by Hunt and Shabanowitz, uses methane and N0 CI to obtain qualitative profiles of the 370 to 535 0F cut from different crude samples. This work was primarily concerned with the dibenzothiophenes and the di- and tri-thienyls prevalent in that particular crude fraction and reported very different behavior from the alkylthiophenes which are more likely to be present in refined middle-distillate fuels. Experimental Conditions This study employed an ELTQ-400-3 Triple Quadrupole Mass Spectrometer and Control System manufactured by Extranuclear Laboratories Inc., Pittsburgh, PA. Both electron impact (EI) and chemical ionization (CI) were employed using an Extranuclear SimulscanTM ionizer maintained at 200 C. Electron impact spectra were obtained at both high (70 eV) and low (20 eV) energies. Chemical ionization was 100 performed with methane (ultra-high purity) and methanol (anhydrous reagent grade). Methanol for CI was introduced to the ion source by leaking it through a needle valve from a 500 ml expansion volume maintained at ambient temper— ature. Collisionally activated dissociation (CAD) spectra were obtained using argon (99.9 X) at laboratory pressures ranging from 1-4 x 10‘5 torr in the analyzer chamber. For pure components, liquid samples were introduced by leaking them through a needle valve and solids were intro- duced by means of a solids probe. Fuel samples were introduced by means of a Hitatchi 633-30 Gas Chromatograph (Hitatchi Ltd., Tokyo, Japan) equipped with a split/- splitless injector maintained at 200 C and a 30 meter narrow-bore DB-l chemically bonded phase fused silica capil- lary column with a 1 micron coating (J & W Scientific, Rancho Cordova, CA.) The column was interfaced to the ion source by passing it through a transfer line at 250 C to a direct inlet. Helium carrier gas was used at a flow of 1 ml/min (a flow velocity of 40 cm/sec at 200 C.) The column was temperature programmed from 50 to 250 C at 6 C/min. Pure components used in this study were obtained from either Chemservice Inc. (West Chester, PA) or Aldrich Chemi— cal (Milwaukee, WI) and were used without further purifi; cation. Fuel samples were obtained from Dr. Gary Seng at NASA-Lewis Research Center and were analyzed without any pretreatment. 101 Data interpretation was performed on a PDP-ll/23 mini- computer running the RSX—llM operating system after performing a lB-bit parallel upload from the instrument control microcomputer system. Full scan data were inter- preted with the aid of a multi-dimensional database package (47). Selected ion monitoring (SIM) and multiple reaction monitoring (MRM) data were interpreted with the aid of another locally developed software package (see Chapter 4.) Preliminary MS and MS/MS Studies The first stage in this work was a MS and MS/MS characterization of several thiophenes under different ionization conditions. Structures for the thiophenes used in this study are given in Figure 6.1. These particular thiophenes were chosen because they represent the type of functionality expected for thiophenes in a refined fuel and they cover the molecular weight and boiling point ranges typical of the middle distillate fuels. The EI spectra of the thiophenes were obtained at electron energies of 70 and 20 electron volts. CAD spectra of the principal ions in the RI spectra were also obtained at both electron energies. These data are summarized in Table 6.1. Considerable fragmentation was observed at both 102 Thiophene MW=84. BP=84C Z—Methylthiophene S-Methylthiophene MW=98. BP=113 MW-QB. BP-114C 2.5 Dimethylthioph‘ene Z-Ethylthiophene MWai12. BP=134C MW=112. BP=133C A Q.I Benzothiophene MW=134. BP-221C Figure 5.1 Structures of Pure Thiophenes Studied 103 electron energies for each of the thiophenes investigated with the exception of benzothiophene which, at 70 eV, produced only the molecular ion and which was not ionized at 20 eV. The mass spectral behavior of the thiophenes studied fell into two groups. The first group consists of thiophene itself and the alkyl substituted thiophenes. Members of this group tended to undergo neutral losses of 45 (corres- ponding to CH3) and also formed 45+ both in the source and under CAD with argon at Plab = 2 x 10‘5 torr and collision energies on the order of 20 eV. The formation of 45+ is generally considered characteristic of aliphatic organo- sulfur compounds (48). The neutral loss of 45 from the thiophenes is quite reasonable in that it results in the formation of Cst-R+, a substituted aromatic ring where R is the ring substituent of the parent thiophene. The 3—methyl- thiophene was found to give essentially the same pattern as other members of this group despite the difference in position of the ring substituent, even at the low electron energy. The one disubstituted thiophene examined, 2,5-dimethylthiophene, gave the same peaks as were charac- teristic of other thiophenes studied, plus additional peaks at 77+, 78+, and 65*. These peaks probably correspond to ring rearrangement incorporating a methyl group to form benzene and phenyl ions. These ions exhibit substantially reduced abundance under in the 20 eV spectra“ The principal reaction of the substituted thiophenes is the cleavage of 104 Table 6.1 Spectra of Pure Thiophenes Ionized by Electron Impact Thiophene 70 eV EI 84 (100), 58 (70), 45 (40), 39 (35) 20 eV EI 84 (95), 58 (100), 45 (60), 39 (60) 70 eV CAD(84+) 84 (100), 58 (50), 45 (60) 20 eV CAD(84*) 84 (100), 58 (60), 45 (60), 39 (15) Z—Methylthiophene 70 eV 51 98 (60), 97 (100), 45 (18), 39 (10) 20 eV 51 98 (65), 97 (100) 70 eV CAD(97*) 97 (100), 53 (25), 45 (10) 20 eV CAD(97*) 97 (100), 53 (60), 45 (25) 3-Methylthiophene 70 eV EI 98 (59), 97 (100), 53 (10), 45 (21) 20 eV 81 98 (75), 97 (100), 45 (5) 70 eV CAD(97+) 97 (100), 53 (50), 45 (35) 20 eV CAD(97*) 97 (100), 53 (35), 45 (15) 2,5—Dimethylthiophene 70 eV 31 112 (92), 111 (100), 97 (75), 77 (22), 59 (27), 45 (18) 20 eV E1 112 (90), 111 (100), 97 (50), 77 (10), 59 (10) 70 eV CAD(1ll+) 111 (100), 97 (15), 78 (30), 77 (50), 67 (60), 65 (15), 45 (30), 41 (10) 20 eV CAD(lll*) 111 (100), 97 (20), 78 (15), 77 (25), 67 (25), 65 (10), 45 (15), 41 (10) 2—Ethylthiophene 70 eV El 112 (36), 97 (100), 45 (19) 20 eV El 112 (60), 97 (100) 70 eV CAD(112*) 112 (25), 97 (100) 20 eV CAD(112*) 112 (15), 97 (100), 53 (5) 70 eV CAD(97*) 97 (100), 53 (90), 45 (25) 20 eV CAD(97*) 97 (85), 69 (10), 53 (100), 45 (40) Benzothiophene 70 eV El 134 (100), 90 (10), 89 (15) 70 eV CAD(134*) 134 (100) 105 the substituent group gamma to the sulfur. Proposed mechanisms for this group are given in Figure 6.2. This observed behavior is consistent with previous studies of the El spectra of the alkylthiophenes (49). The second type of thiophene studied was benzothio- phene. As would be expected, its molecular ion is highly stabilized by the fused aromatic ring and gives much less fragmentation in the primary spectrum, even at 70 eV. This compound also undergoes a neutral loss of 45* in its CAD spectrum, although less readily than the alkyl thiophenes. In that the experimental goal was to maximize sensi- tivity by multiple reaction monitoring, an ionization method was sought which would reduce fragmentation from that observed with electron impact, even at low energy. As illustrated in Chapter Five, ionization techniques which minimize fragmentation while maintaining the overall ionization efficiency can improve sensitivity by concen- trating the ion current into essentially one ion for each compound under study. Preferably, this ion would also carry molecular weight information. Since the actual analysis would use multiple reaction monitoring (MRM) of the targeted compounds, the loss of structural information in the primary spectrum was not a problem. The combined information available with the retention time from a capillary gas chromatograph and the choice of reactions to monitor was expected to provide sufficient selectivity to identify the 106 masocmodca cavspavmpsm mom enema compensasmamm emmonomm m.o muswfim 59.73 5.2 ..m +. Cm rl .mém w C .8 36 mo Imam: 67.5370qu 20E .m .._o 80.. 45.5% a .2 9.9.; 5 % + :11 mg: m .. e. .9. 9 no 80.. 2332a .612. 92 , MN + A... fig 511%: T m 9 em\\\o.: a a . ; 9 as. do 292.2“.qu 107 targeted components. Consideration of these criteria suggested the use of chemical ionization (CI). The most widely used reagent gas for chemical ioni- zation is methane. Methane CI is expected to ionize vir- tually all organic compounds, yielding a base peak at (M+H)‘, where M is the molecular weight of the compound under study, but with much less fragmentation than in El (50). The primary and CAD mass spectra of the thiophenes were obtained with methane CI at 200 C and helium present from the GC column (column flow ca. 1 ml/min.) These pri- mary spectra, summarized in Table 6.2, showed extensive fragmentation of the alkyl thiophenes, forming significant (20 ~ 100% of the base peak) abundances of 97+ and as much as 90% relative abundance of 111+ in the case of 2,5-dimethylthiophene. Thiophene and benzothiophene showed very little fragmentation. The CAD spectra of the (M+H)* ions of thiophene and the alkyl thiophenes all showed neutral losses of 44 mass units (corresponding to elimi- nation of CS.) In addition, the alkyl thiophenes all showed a neutral loss of 34 mass units (corresponding to 828.) The principal CAD product of the (M+H)+ ion of 2~ethylthiophene was loss of 28 mass units (elimination of ethylene, C284.) The major CAD product of 97+ from the alkyl thiophenes was loss of 44 mass units to yield the aromatic species 53*, which was also observed in the El spectra. CAD of the (M+H)* ion from benzothiophene resulted in the neutral loss i 108 Table 6.2. Spectra of Pure Thiophenes using Methane CI Thiophene CI 113 (5), 87 (5), 86 (7), 85 (100), 84 (14) CAD(85*) 85 (100), 45 (28), 42 (32) 2-Methylthiophene CI 127 (4), 101 (3), 100 (8), 99 (100), 98 (14), 97 (12) CAD(99*) 99 (100), 65 (20), 55 (37) 3—Methy1thiophene (3), 100 (8), 99 (100), 98 (9), CI 127 (5), 101 97 (13) CAD(99+) 99 (100), 65 (10), 2-Ethylthiophene CI 141 (10), 114 (8), 111 (5), 97 (20), CAD(113+) 113 (100), 98 (5), 1,2-Dimethylthiophene CI 141 (1), 114 (9), 111 (19) CAD(113+) 113 (100), 98 (3). 69 (19), 45 (8) Benzothiophene 55 113 (100), 85 (31) 85 (65), 79 (32), CI 135 (100), 134 (25) CAD(135*) 135 (100), 91 (37) (45) 113 (100), 79 (7) 77 (12), 112 (22), 112 (26), 109 of 44 mass units, presumably to form the highly stable tropyllium ion at 91*. Overall, the methane CI generally produced a smaller number of peaks in the primary mass spectra of the thiophenes being studied. However, using source conditions simulating those present in a GC/MS/MS analysis, a few fragment ions still accounted for a large fraction of the ion current. Additionally, since methane is capable of ionizing virtually all organic compounds (51), it provides little if any gain in selectivity over EI when dealing with a mixture as complex as a fuel where there are large numbers of isomeric and isobaric compounds that are not completely resolved, even with high resolution chromatography. This is especially true when analyzing for trace components which suffer from interferences from even very low abundance reactions of the major mixture components. Methanol has been previously reported as providing extremely little fragmentation when employed as a CI reagent gas (52, 53). Under GC/MS/MS conditions in the source (200 C and ca. 1 ml/min of helium flow) optimum ionization of the thiophenes occurs when the methanol pressure is adjusted to provide approximately equal abundances of the ions at masses 33+ (CH30H2’) and 65+ (corresponding to (CHaOH)2H+, the methanol dimer cluster ion.) Other significant reagent ions present under these conditions are 97+ (the methanol trimer cluster ion (CH30H)3H*,) 79+ (loss of water from 97*) and 47+ (loss of water from 65+.) 110 Using the source conditions described above, methanol CI gave virtually no fragmentation in the primary mass spectrum of any of the thiophenes studied. All of the primary spectra consisted only of the (M+H)* ion from the sample and reagent ions. The CAD spectra of (M+H)+ from each of the thiophenes gave the same product ions as were the case for methane CI. Methanol CI spectra of the thiophenes, and the corresponding CAD spectra, are listed in Table 6.3. An additional, very significant, advantage of methanol chemical ionization is its high selectivity. As reported in the previous chapter, methanol CI has been found to discri- minate against the hydrocarbons, and gives essentially no ionization of the saturates which make up the bulk of the matrix in a fuel sample. As a result, matrix interferences are greatly reduced, thus lowering the requirements for chromatographic resolution and enabling shortened analysis times. Determination of Detection Limits 0n the basis of studies of the MS/MS characteristics of the thiophenes under methanol CI conditions described above, reactions may be chosen for the detection of these compounds by means of multiple reaction monitoring. This choice of reactions is summarized in Table 6.4. The loss of 44 u (CS) was chosen for the detection of thiophene, the methyl- thiophenes, and benzothiophene. The 113* ~~> 79+ reaction 111 Table 6.3 Spectra of Pure Thiophenes using Methanol CI Thiophene CI 87 (6), 86 (9), 85 (100), 84 (32), 58 (4), 45 (4), 39 (2) CAD(85*) 85 (100), 59 (2), 45 (49), 41 (82), 39 (19) 2-Methylthiophene CI 101 (6), 100 (8), 99 (100), 85 (2) CAD(99*) 99 (100), 97 (2), 84 (6), 65 (22), 55 (78), 45 (7), 39 (1) 3-Methylthiophene CI 101 (5), 100 (9), 99 (100), 85 (l) CAD(99*) 99 (100), 84 (8), 65 (12), 55 (65) 2-Ethylthiophene CI 114 (8), 113 (100), 111 (2), 97 (10), 85 (8) CAD(113*) 113 (100), 98 (6), 85 (86), 79 (8) ' 1,2-Dimethy1thiophene CI 114 (10), 113 (100) CAD(113*) 113 (100), 98 (8), 79 (35), 77 (ll), 69 (25), 45 (5) Benzothiophene CI 137 (5), 136 (12), 135 (100) CAD(135*) 135 (100), 91 (53) 112 Table 6.4 Reactions for Monitoring Thiophenes in Fuels Compound Parent Ion Daughter Ion Thiophene 85* 41* 2-Methylthiophene 99* 55* 3-Methylthiophene 99* 55* Z-Ethylthiophene 113+ ' 85+ 2,5-Dimethylthiophene 113* 79* Benzothiophene 135* 91* if 113 (loss of 34 u, 828) was used for detection of 2,5~dimethyl- thiophene since it was found to be much more sensitive. Similarly, for 2-ethy1thiophene, the 113* —-> 85* reaction (loss of 28 mass units, cleavage of the ethyl side chain) was also employed for reasons of sensitivity. Potential interferences for thiophene and the alkyl thiophenes are the cycloparaffins and the olefins which are isobaric. Monitoring the reaction corresponding to the neutral loss of 34 mass units (H25) from 2,5-dimethyl- thiophene should discriminate against both of these inter- fering compound types. The reactions chosen for thiophene and the methylthiophenes (loss of 44 u) and for 2~ethy1- thiophene (loss of 28 mass units) are not as selective as the loss of 34 reaction. However, these reactions are much more sensitive and, since the targeted compounds have boiling points at the low end of the middle distillate boiling range, the overall analysis will be sufficiently selective to attain unambiguous identifications. Benzo— thiophene is isobaric with the Caesubstituted benzenes and both compound classes readily undergo neutral losses of 44 mass units (corresponding to CS for benzothiophene and C3H8 for the substituted benzenes) to form the highly stabilized ion of mass 91. Benzothiophene does have a somewhat higher boiling point, however, so it can be resolved chromato— graphically. (It is interesting to note that benzothiophene cannot be resolved from the Cq-substituted benzenes by batch 114 MS/MS at unit mass resolution, in that they are isobaric and both undergo the loss of a 44 u neutral.) Using these reactions, detection limits for thiophenes in jet fuels were determined by preparing a series of blends of Jet A (the standard commercial jet fuel) with known amounts of thiophene, 2-methylthiophene, 3-methylthiophene, 2-ethylthiophene, 2,5-dimethylthiophene, and benzothiophene. Triplicate analyses were performed on each blend by methanol chemical ionization GC/MS/MS using multiple reaction monitoring. The detection limit is defined as the lowest analyzed concentration where the average ratio of the integrated peak area to the background area was greater than 2:1. The detection limits so determined are reported in Table 6.5. Retention times reported are the average of all gc/ms runs. Typical single reaction chromatograms used to determine these detection limits are given in Figures 6.3a through 5.3a. 115 Table 6.5 Detection Limits for Thiophenes in Jet Fuels Retention Detection Signal to Compound Time (sec) Limit (ng) Background Thiophene 87.7 2.5 2.2 2-Methy1thiophene 111.3 2.2 6.9 3-Methylthiophene 113.7 2.1 5.0 Z-Ethylthiophene 147.6 1.7 4.5 2,5-Dimethylthiophene 148.8 1.5 3.5 Benzothiophene 352.1 7.6 4.3 116 ”#41 11562 90 —-> 55. (fix/291 ng Z-Methylthiophene «”273 ng B-Methylthiophene TII'IUII' 113 —-) 05 5219 ng Z-Ethylthiophene 113.9 79. W1 97 n8 2 , S-Dimethylthiophene 0 135 ,-> 01 5986' 159 ng Benzothiophene o 1. l 1 Time (see) Figure 6.3a Detection of Thiophenes in Jet A d 50 100 150 200 250 300 350 uoo 117 asl-> 41 <:"’<’:‘:il~lv0 ng Thiophene 5179 90 -> 55 é—ls—126 ng 2-Methylthiophene 61—118 ng 3-Methylthiophene 113 -#> 85 51/94 ng 2-Ethylthiophene 113 -> 70 2’85 ng 2.5-Dimethylthiophene 135 -) 01 68 ng Benzothiophene O 50 100 150 200 250 300 350 400 Time (sec) Figure 6.3b Detection of Thiophenes in Jet A 118 -> 807 85 41 55‘ ‘fl\‘14 ng Z-Methylthiophene 5197 ‘7—13 ng 3—Methylthiophene 0"r'I—rlI'IrIYrtr‘V'l'l'IleIri'lrfI'I‘UI'Yi'I—I 2681 113 —) 85' ‘$2110 ng 2-Ethylthiophene ofirr'UIrl'VFlI'YlUY'UUIIIIUYU‘IUIIIIrII' 1284 113 '-> 70 ‘kZ——9 ng 2,5-Dimethy1thiophene lllllLll. O T" ' '7 rrxuis':effi 135-9 91‘ 3939 8 ng Benzothiophene OWUUIIIIF'rtlI'V'Irj‘tifi'r‘ril' 0 50 100 150 200 250 300 350 400 Time (see) Figure 6.3a Detection of Thiophenes in Jet A 119 as» 4 3115 1 2.5 ng Thiopzeneg 0 I.'tr'rr'llr'Ir'lTVIIYlUr'llIlrlIll'll'U1 101 9° 3 55‘ 5 42.2 ng 2-Methylthiophene ‘;;—2.1 ng 3-Methylthiophene o 'UIIU'A‘Ii'rrIIIU'lUlI'IIII. 113' -> 85' 831 01 1.2 ng Benzothiophene oWIIITIVU'U'IT'IUIl'IIrIr‘I'I'r'fiT'AI'II—l 0 50 100 150 200 250 300 350 400 Time (see) Figure 6.3a Detection of Thiophenes in Jet A 120 At the highest spike level (Figure 6.3a), the reactions for native thiophene and the 01 and C2 substituted thiophenes are observed to be virtually free of inter- ferences from the Jet A hydrocarbon compounds. Several interfering peaks, however, are observed in the recon- structed single reaction chromatogram. for benzothiophene. The use of capillary gas chromatography sufficiently separates these potential interferences, attributable to the isobaric C4 substituted benzenes, so as not to be a problem. The chromatography also serves to easily resolve the 2- methyl- and 3~methy1thiophene peaks, while the choice of reactions for the 2-ethyl- and 2,5-dimethylthiophene insures that they are readily distinguished. As the thiophenes spike level is decreased (going from Figure 6.3a to Figure 6.3d), the relative abundance of species reactions interfering with thiophene and benzo~ thiophene determinations are observed to increase. (The source of the interfering 85* —> 41* reactions are the isobaric CeH13* ions formed from fragmentation of alkane molecules present in much higher relative abundances than thiophene.) From this work, it is observed that the selectivity of this analysis is sufficient that the limiting factor for detectivity of the thiophenes is the overall instrument sensitivity. Thus, detection limits for this technique should improve directly with improvements in instrument design. 121 Conclusions The detection of trace level sulfur compounds in a fuel is an especially difficult analytical problem not only due to the complexity of the matrix, but also due to the fact that these compounds are isobaric with hydrocarbons, and their fragments, which are normally present in much higher abundances. Existing studies of the effect of thiophenes on jet fuel stability have found them to be deleterious at the parts per thousand level (54, 55). These levels correspond to detection limits for the thiophenes using batch MS with chromatographic pre—separations or capillary GC/MS (56). Batch MS/MS at unit mass resolution has tremendous potential for dramatic improvements in both detectivity and analysis time, but will fail in cases where isobaric species also undergo isobaric neutral losses (e.g. benzothiophene and the C4 substituted benzenes). BEQQ instruments would appear to have interesting potential in this area for target compounds which can be resolved at 10,000 to 20,000 resolution. The use of capillary GC/MS/MS, demonstrated in this work, allows for detection limits on the order of 25ppm (for native thiophene) where previous techniques had been limited to parts per thousand. Application of this work will allow testing to determine whether or not thiophenes effect fuel stability at levels down to 25ppm. This ability to determine the existence of such threshold values would be of 122 tremendous economic impact in the establishment of workable specifications for future fuels refined from low—grade hydrocarbon sources. Chapter 7. Selective Ionization, Gas Chromatography, and Triple Quadrupole Mass Spectrometry as Tools for the Characterization of Fuels The research presented in this dissertation has encompassed several aspects of bringing a powerful and elegant analytical instrument together and applying it to a specific problem. The GC/TQMS instrument together with its control and data systems and other accessories, however, is expensive and, to take full advantage of its capabilities, requires an analyst with more knowledge and skill than would be required for a single-stage GC/MS. The obvious and pragmatic question then becomes "How often is all of this power really necessary?" Method development for complex mixture analysis can be viewed as a process of tying a series of "selectivity elements" together until a sufficient number of inter— ferences have been removed that the signal of interest may be reliably measured. Depending on the particular problem, these "selectivity elements" may include such tools as chromatographs, mass analyzers, and monochromators, physical separations, such as sieveing, or chemical procedures, such as solvent extraction or derivatization reactions. An 123 124 understanding of the sample at hand and the goals of the analysis enables the analyst to choose the most effective set of tools to apply to the problem. For the problem of determining part per million levels of thiophenes in jet fuels, the combined use of capillary gas chromatography, selective chemical ionization, and two stages of mass analysis separated by collisionally activated dissosciation (CAD) were deemed necessary. Each of these individual selectivity elements made a unique and necessary contribution to the analysis. Gas Chromatography as a Selectivity Element The combination of gas chromatography with mass spectrometry has been immensely successful not only for the. temporal dispersion of the sample, but also for the repro- ducibility of sample introduction. This comes at a price, however, of having to analyze a time-varying sample, limiting the available sample size, and increasing the analysis time. The use of MS/MS often reduces the demand on the chromatography to that of providing for reproducible sampling while using the primary mass filter to distinguish co-eluting species. This approach falls down, however, when it is necessary to distinguish co-eluting components which are isomeric or, in quadrupole MS/MS, isobaric. 125 An excellant example of this problem occurs in the work reported in Chapter 6. In that instance, the use of capillary gas chromatography was essential in preventing the butylbenzene isomers from interfering with the detection of benzothiophene. Not only do the butylbenzenes have the same molecular weight (134 amu) as benzothiophene, they also exhibit virtually identical MS/MS activity. In both cases the MH* ion (the only significant peak in the primary CI mass spectra of these compounds) readily undergoes a neutral loss of 44 amu to form identical daughter ions -- the highly stabilized 91* species. There are no other significant CAD products for either compound type. Mass Analysis as a Selectivity Element From a chromatographer’s viewpoint, the addition of a mass spectrometric detector to a Chromatograph adds a second data dimension which can enable positive identification of solutes as they elute from the column. In addition, the mass spectrometer can frequently aid in deconvoluting incompletely resolved chromatographic peaks through inter— pretation of appropriate single-ion chromatograms. This second point effectively relaxes the demand placed on the chromatographic resolution. Consequently, detection of lower concentrations of analytes are possible because lower demand on chromatographic resolution generally allows for the injection of larger samples. Additionally, decreased 126 demand for chromatographic resolution frequently allows for more rapid analyses. As a practical matter, mass spectrometric deconvolution of partially resolved chromatographic peaks requires the existence of at least a distinctive peak in the mass spectrum of each of the incompletely resolved components and that signals for the components be reasonably close in magnitude. Thus, this approach becomes especially difficult when attempting to distinguish isomeric or, at unit mass resolution, isobaric compounds. These limitations on mass spectral deconvolution of chromatographic peaks create an especially difficult situation for the determination of trace thiophenes in fuels. Not only is it exceedingly difficult to obtain complete chromatographic resolution for all the matrix components, the hydrocarbons tend to form ions at virtually every unit mass position in the spectrum. Many of these peaks may be quite small, nevertheless, they can be suffi- ciently large to interfere with the detection of isobaric trace components. (The alkylthiophenes are isobaric with the alkenes and the cycloalkanes; the benzothiophenes are isobaric with the alkylbenzenes.) Where there are significant chemical differences between the analytes and the matrix, chemical ionization and MS/MS can help to alleviate these problems. By simplifying the mass spectra, chemical ionization can reduce the incidence of isobaric peaks in the mass spectra. Selective 127 CI reagents can be of further service by increasing the signal of the analyte relative to that for the matrix. Tandem mass spectrometry approaches the problem from the opposite end, taking advantage of the fact that even isobaric ions are likely to form unique fragment ions as a result of the CAD process. Chapter 6 demonstrated the power of the simultaneous application of selective chemical ionization and MS/MS by detecting part per million levels of thiophenes in spiked jet fuels. Chemical Ionization as a Selectivity Element Chemical ionization, with an appropriate choice of reagents, can be a powerful selectivity tool in cases where there is a difference in chemistry between the analyte and the matrix which can readily be exploited. The analytical goal of determining trace level polar compounds in a non- polar matrix is a perfect case for selective ionization. Polar compounds, with their relatively high proton affinities, can be ionized by a weak proton donor, such as protonated methanol, while minimizing ionization of non- polar compounds, which have relatively low proton affinities. An entirely different approach would have been required had the goal been to detect trace level non—polars in a polar matrix (e.g., hydrocarbons in a fermentation broth.) Further, selective ionization processes would not work at all when the analyte and the matrix are chemically 128 similar (e.g., the determination of a particular alkane isomer in a fuel sample.) The research presented in Chapter 5 quantitates an instance of the sensitivity and selectivity implications inherent in the selection of ionization conditions for a mass spectrometric analysis. The work also illustrates the trade~off between sensitivity and selectivity when using methanol chemical ionization. An understanding of this relationship, taken together with knowledge of the selec- tivity of the rest of the analytical system, allows the analyst to essentially "tune" the methanol pressure to a value which supplies a sufficient contribution to the selectivity of the overall method while retaining the maximum possible sensitivity. Extending this approach to an entire multi-dimensional analysis method allows the devel- opment of highly selective techniques which do not require a severe sensitivity trade—off at any one element of the The methanol work developed trends in the ionization process which can be developed by varying the extent of cluster ion formation. Future work covering additional analyte functionality and the effects of varying source temperature may offer extended insights into the process. Study of cluster ions formed by different polar molecules may reveal different selectivity patterns to be exploited for different analytical problems. 129 Tandem Mass Spectrometry as a Selectivity Element As a tool for mixture analysis, tandem mass spectro- metry with CAD can frequently facilitate the distinction of isomeric and isobaric ions by differences in their fragmen- tation patterns. By programming a triple quadrupole mass spectrometer to monitor specific CAD reactions, the analyst can obtain a highly selective analysis of chromatographic effluents with signal-to-noise advantages analogous to that for selected ion monitoring in single stage GC/MS. Additional modes of selectivity are available through choice of collision energy, collision pressure, and the use of reactive collision gases. In the analysis of thiophenes in jet fuel, the chromatographic resolution and ionization conditions provided sufficient selectivity that the collision conditions could be tuned for maximum sensitivity. Conclusion This dissertation has demonstrated a specific problem which has become more tractable for routine analysis because of the availability of GC/TQMS. The use of this instrument allowed part per million level determination of thiophenes in jet fuels without requiring any lengthy preparative scale liquid chromatographic separations as would traditionally be employed. One can readily imagine equally complex problems 130 in the analysis of such complex matrices as biological systems and hazardous wastes, as well as other petroleum- based products and intermediate streams, where such an approach would be equally valuable. The thiophenes work presented an example of the analysis of polar components in a non-polar matrix. This suggests testing the applicability of GC/TQMS to the analysis of non-polar analytes in a polar matrix, or the even more difficult problem of non-polar targets in a non- polar sample. The latter challenge would be of particular value in the petroleum industry, where the traditional mass spectro- metric type analyses fail to provide sufficient detail to accurately model petroleum refining processes. Specific problems would be distinction between isomeric types such as the alkenes and cycloalkanes and 5-member and 6-member cyclic systems. This work demonstrates GC/TQMS to be a powerful tool that is available to solve some very difficult analytical problems. Truly, one crucial mark of a skilled analyst is an awareness of all the tools at his or her disposal and the ability to choose the simplest one which meets the sensit— ivity, selectivity, precision, accuracy, speed, and cost constraints placed on obtaining the desired information. i APPEND ICES Appendix A Inter-Processor Synchronization for Multi-Processor Data Acquisition The power of a multiple microprocessor control system comes from the ability to divide the control tasks among the different processors. This division of tasks allows for higher speed operation by performing different time-critical activities in parallel instead of intermingling them on a single processor. Applying this approach to simultaneous instrument control and data acquisition, however, requires that these independent processors be closely coordinated. This appendix serves as a follow up to chapters three and four, providing a detailed discussion of how the multi- processor data acquisition software achieves the needed synchronization. Sweeps and Scans The most basic form of data acquisition on the multiple microprocessor control system is the sweep. A sweep records ion current as function of the voltage applied to any single element of the ion path while everything else is held 132 133 constant (quadrupole masses are a special case in that mass is being swept rather than voltage). The control system stores sweep data as ordered pairs of raw voltage (or mass) and intensity; that is, no peak-finding or any other data treatment is performed. The operator is responsible for commanding the quadrupoles for the desired RF-only or DC/RF mode. Additionally, the current software does not allow for any sweeps with linked device. Since raw data is being retained, sweep width is limited to 1024 steps. Scans refer to the traditional mass spectrometric data acquisition mode, where ion intensity is monitored as a function of mass and mass—intensity pairs are saved only for peaks. This allows for a virtually unlimited scan width, the sole limitation being 1024 peaks per scan (this number should be quite adequate even for a 2000 amu quadrupole, assuming unit mass resolution). Five scan types are defined lSCAN - the first quadrupole is scanned and the other two are placed in RF-only mode, BSCAN - the third quadrupole is scanned while the first two quadrupoles are kept in RF-only mode, PSCAN - parent scan, the third quadrupole is set to transmit a fixed mass, the first quadrupole is scanned, and the center quadrupole is kept in RF- only mode, 134 DSCAN - or daughter scan, the first quadrupole is set to transmit a single mass, the third quadru- pole is scanned, and the center quadrupole is placed in RF-only mode, and NSCAN - neutral loss scan, the first and third quadrupole are scanned simultaneously with a fixed mass offset and the second quadrupole is retained in RF-only mode. In each of these modes any RF-only quadrupoles are linked to the scanning quadrupole, scanning at one-half the RF power level of the scanning quadrupole. The settings and scan windows for each of the ion path elements are maintained in the Variable Device Parameter Table (VDPT). The VDPT allows entries for a default value, start and end values, and step size for up to thirty-two ion path elements, with a complete set of these values being memory resident for each of the four defined ionization modes. Entries in this table are normally modified by means of the Parameter Editor, PED, and sixteen of these tables, with titles, may be maintained on the disk. It is through the VDPT that the user specifies the scan limits and step size to be used for any device to be swept or scanned, and the values at which all of the other ion path elements are maintained. 135 Four additional parameters necessary for data acqui- sition are set through the ASET (acquisition set) command. These are the threshold, minimum peak width, maximum peak width, and scan rate. The scan rate may be set to one of thirteen different nominal values ranging from 1 to 10,000 points per second (corresponding to 0.1 to 1000 amu/sec for a mass sweep or scan at 10 steps per amu). The other three parameters are used by the "on-the-fly” peak finding algorithm. All four of these parameters are downloaded to the appropriate slave processors whenever they are changed. All of the actual data acquisition code resides on the three slave processors -— ion path, reduction, and detec- tion. At the start of a sweep or scan, the master processor downloads the necessary parameters onto the parameter stacks of each slave immediately prior to giving them the commands needed to perform the selected type of scan or sweep. Ion Path Processor The ion path processor is responsible for setting the appropriate RF/DC status of each of the quadrupoles for the selected scan type (or do nothing for a sweep) and scanning the appropriate mass or voltage. During a scan, the ion path is also responsible for scanning the RF power on the RF-only quadrupoles. At the start of a sweep, the master has placed the number of the device to be swept, the scan limits and the step size on the ion path slave’s parameter stack. For a scan, the master places the scan limits and 136 step size on the stack but the device number is not needed. In either case, the ion path clears its software status byte, stops the fast scan task, initializes the AMU Timer (57), and converts the scan limits and step size into a starting DAC value, a step size in DAC units, and a number of steps, and sets any scanned devices to their initial values. The AMU Timer is a dedicated piece of hardware which provides a timed synchronization signal to approximate the nominal scan rate. It is initialized using the scan rate parameter which was defined by ASET. After a settling time of roughly 10 milliseconds, the ion path slave can enter the scanning loop. At this point, the current DAC value of the scanned device is stored in a processor register. During a sweep, a second register contains the address of the device to be scanned. Each of the five scan routines has encoded in it the addresses of the appropriate quadrupole mass DAC’s. The first step in each loop iteration is to check that the detection slave has finished sampling the previous data point. This is done by testing a register on the Link and Sync Output Module (58) going to the detection slave. After this has been verified, the ion path can write the new values out to the DACs (for the first step this will just be a repeat of the initial value). For scans, the values to send to RF-only quadrupoles are calculated by dividing the mass of the scanned quadrupole by two (a single shift right). After writing out the new data values, the ion path 137 then waits for the signal from the AMU timer and then uses the Link and Sync Output module to inform the detection slave that it can now proceed with acquisition of the next data point. The current value of the scanned device can now be sent to the reduction slave, again using the Link and Sync Output Module. The last step in the loop is to increment the register containing the current data point by the step size to calculate the value of the next data point. This loop is then repeated for each step. The ion path slave now informs all the other processors of the completion of the scan by setting its software status byte to a 1. It then waits for the detection slave to indicate that it is done with this last data point and then restores all of the ion path devices to their initial values. When the reinitialization is complete, the ion path slave can return to its idle state of monitoring the command FIFO from the master processor. Detection Processor The detection slave functions identically for both scans and sweeps. This processor is responsible for controlling the data acquisition board (DAq) (59), trans- mitting intensity values on to the reduction slave, and maintaining synchronization with the ion path processor. At the start of a scan or sweep, the detection slave executes a hardware reset of the DAq and sets the number of conversions that it is to average for each data point. This number of 138 ADC conversions is determined by the rate parameter requested by the user through ASET. Scan initialization is completed by indicating to the ion path that it is ready for the next data point; this is done through the use of a Link and Sync Output Module. The scan loop begins with the acquisition processor then monitoring the Link and Sync Output module, waiting for a signal from the ion path slave indicating that the ion path has been properly configured for the next data point. The acquisition processor then gives the start command to the DAq and waits for the it to indicate completion. This slave then informs the ion path processor that it is through with the current data point (using the Link and Sync Output module), reads the value of this data point from the DAq, clears it, and transfers the value on to the display/reduc- tion processor by means of a Link and Sync Output module. The last step in the data acquistion loop has the acqui- sition slave check the software status byte from the ion path slave to see if it has been set to the value 1, which would indicate the end of the sweep or scan. If the end of scan is indicated, the DAq is reset and the acquisition processor returns to monitoring the command FIFO from the master processor. If there is no end of scan indication, the detection slave loops back to the point where it was waiting for the ion path ready signal on the Link and Sync Input module. 5 139 Reduction Processor The processor most affected by whether this is a scan 1 or a sweep is the reduction slave. For a sweep,. it only needs to read the x-axis value from the ion path slave and the y-axis value from the detection slave and place them into the data buffer. For a scan, rather than putting all of the raw data into the data buffer, it must perform "on- the-fly" peak-finding and store only the actual peaks. Scan and sweep initialization are identical. The master places the device number of the scanned device on the parameter stack of the reduction slave. During scan initialization, this number is stored for later use, the data buffer and peak-finding parameters are initialized, and the processor performs a read from the Link and Sync Input Module (60) from the detection slave to inform that slave that it is ready to receive the first data point. For a sweep, the run-time loop begins by reading the x- axis value sent from the ion path slave via a Link and Sync Input Module, and the corresponding intensity value is read from the Link and Sync Input Module of the detection slave. These values are then stored in the next available location in the data buffer. The last step in the loop is to check the software status byte from the ion path slave. If this byte has been set to l, the scan is complete. If not, the processor returns to the start of the loop. At the end of the scan, the reduction slave must convert each of the x-axis values which were sent from the 140 ion path slave in DAC units to physical units. During the process of performing these conversions, the reduction slave also determines the maximum intensity value and the total ion current for the sweep. When this is done, the slave sets its software status byte to a l, informing the master that data is now available for uploading. Following this, the ion path’s software status byte is reset to zero, the sweep data are transferred to the graphics buffer, and the peak-finding variables are re-initialized. When all of this is complete, the reduction slave can return to monitoring the FIFO with commands queued from the master. Selected Ion Monitoring/Multiple Reaction Monitoring Because of the hierarchical design of the TQMS computer systems (61), the tasks comprising the MRM package were divided between the multi-microprocessor "control system" and the PDP-ll/23 "data system". In keeping with this overall system design, the control system was given respon— sibility for instrument set-up, data acquisition, storage, real-time graphics, and generation of simple data listings and display. The data system was given responsibility for generation of experimental summaries, advanced graphics output, and archival storage. On both systems, the MRM tasks were, insofar as possible, designed to be compatible and logical extensions of programs developed for the handling of full scan data. 141 Because the control system code was written in FORTH (62, 63, 64) it was a straight—forward matter to create a family of related routines to perform selected ion moni- toring and multiple reaction monitoring. This family includes commands to perform SIM with just one of the quadrupoles, leaving the other two in RF—only mode, SIM with one of the quadrupoles fixed at selecting a single mass while jumping the other mass-resolving quadrupole, and MRM. The two single-quadrupole scan modes are invoked by the commands lSIM and 381M. These command names were chosen to stress their analogy to the single quadrupole full scan commands lSCAN and 3SCAN, which are used to scan the first quadrupole and the third quadrupole, respectively, while leaving the two non-scanned quadrupoles in RF-only mode. Similarly, the two cases of leaving one quadrupole fixed and jumping the other while performing CAD in the center quadrupole, are given the commands PSIM and DSIM as anal- ogies to the parent ion scan, PSCAN, and the daughter ion scan, DSCAN, commands. There is no NSIM command provided as an analog to the NSCAN mode because, as will be explained later there was no experimental advantage over selecting the same reactions in the MRM mode. The control system portion of the SIM/MRM software package can be divided into three parts -- experiment definition, run time, and post-experiment. The experiment set-up phase consists of defining the ions/reactions to be cycled (which will be referred to as elements), the cycle 142 frequency, a scale-factor to be used for the real-time display, and the length of time to collect data. The run time phase consists of a command for each of the five data modes (lSIM, 381M, PSIM, DSIM, and MRM). Each of these commands performs data acquisition, storage, and real-time graphics. The third phase, post—experiment, consists of subroutines to the general~purpose system commands for generation of Data LISTings (DLIST) and simple graphics DISPlays (DISP). Definition of the elements to monitor is done with a pair of screen editors, SIMED and RED, the Selected Ion Monitoring Editor and the Reaction Editors, respectively. These names were chosen to go with the Parameter Editor, PED, the Method Editor, MED, and the Sequence Editor, SED. In addition to selecting a list of up to eight elements to monitor, the editors allow for specification of a parameter set to be used with each element. Further, up to thirty- two sets of ions and sixteen sets of reactions may be stored on the disk for future use (using the SIMSAVE, SIMGET, RSAVE, and RGET commands). The current version of the software, however, does not provide for different sets of elements to be chained to cover different portions of a GC run. Additional experiment definition takes the form of prompted input in the SSET (for SIM set-up) command. Within this command, the user defines the cycle period to be employed (six values, ranging from 0.1 sec/cycle to 4 143 sec/cycle are allowed), the length of time to acquire data, and a scale factor for the real-time display. Based on the information entered in the experiment set- up phase, all five of the run commands (lSIM, 381M, PSIM, DSIM, and MRM) check that there is sufficient disk space available to handle the cycle period and experiment duration specified. Additionally, the number of ADC samples to average for each data point is determined by the cycle frequency, number of elements to monitor, and whether this is a SIM or MRM mode. For a given sampling rate and number of ions to monitor, the SIM modes can generally perform more averaging than would be possible using the corresponding MRM conditions. This is because the SIM modes are only concerned with the jumping of one quadrupole (remembering that one quadrupole is static in the PSIM and DSIM modes). This explains why an NSIM mode is not implemented; it would require that two quadrupoles he jumped, making it little more than a special case of MRM where all of the neutral losses happen to be the same. Each of the five run commands performs data acqui— sition, storage, and simple real-time graphics. The inter- processor synchronization required to perform these functions is illustrated schematically in Figure A.l. Again, the inherent modularity of the FORTH language allows each of the five commands to be constructed from common subroutines, some of which are the same as those used for scan and sweep data acquisition. Because of FORTH’s - v ——.-. .--.-r-.o.¢- r....- .-..-—... ._.. -4. won-occa- .——-. — —-—- ».- «.- r... 1 l L Ion Path SCYCLE5=Icycles >1 Z=0,0SIPS-l l ?VSAHPLED .1511. 5......“ LEAD FARANSlE) iéll 5 HS KSTABLE .- . . w I‘fip —~—.-.—.—1r—- W‘— .— --«1 AMU TIHER 1¢1¢1 Detection YEAHPLED i?DAiA- -AVAi LABLE' l AEAD DATA SCYCLES=SEYCLE5-l l SCYELESZD SEieEND END-GF-RUN Figure A.1 l...g* YSAHPLED ?ACKN Reduction l ' r iCCYCLE=tCYCLEvl T K..___.,. A. ?RESET l A I=0,tSlH5-l (£3 49—7 ?XSTABLE i l I DAIA REE l SIAAI 9A0 . ?DAlA-ACK } DATAACK ?lDAiAD l l ?YDATA DATA REC’D ?END VDATA M .J ?PLOT STORE END-OF-RUN l LOOP ._. l Q -—~ .— -—-¢-_ DATA-AVAiL ?DATAACK l sna-or-Auu l l ' ADAIA , l g AATA-Acx ' ?END ....J l . .....1__..__, F. “1 lDATAPDISK END-GF-RUNl 1 Logic Flow for Multiprocessor SIM/MRM 145 threaded dictionary structure and linking mechanisms, a single copy of each of these shared subroutines is available for use by any higher level command. This allows appli— cations as complex as the entire TQMS control system to be entirely memory resident, a major factor contributing to the overall speed of the system. Detection Processor The run time code for the data acquisition processor, slave 3, is the simplest of the four processors involved in the multi-micro system. Actually, it can use the exact same code as is used for data acquisition in sweep and scan modes. Acquisition for SIM and MRM begins with commanding a hardware reset of the Data Acquisition Board (DAq) and setting the number of ADC conversions that the DAq hardware is to average for each data point. The acquisition proces- sor then monitors the Link and Sync Output module waiting for a signal from the ion path slave indicating that the ion path has been properly configured for the next data point. The acquisition processor then gives the start command to the DAq and waits for the DAq to indicate completion. Slave 3 then informs the ion path slave that it is through with the current data point (using the Link and Sync Output module), reads the value of this data point from the DAq, clears it, and transfers the value on to the display/reduc- tion processor by means of a, Link and Sync Output module. The last step in the data acquistion loop has the l46 acquisition slave check the software status byte from the ion path slave to see if it is indicating the end of the experiment. If the end of experiment is indicated, the DAq is reset and the acquisition processor returns to monitoring the command FIFO from the master processor. If there is no end of scan indication, slave 3 then 100ps back to the point where it was waiting for the ion path ready signal on the Link and Sync Input module. Ion Path Processor During run time, slave l, the ion path processor, is responsible for cycling the quadrupoles through the elements being monitored, activating the appropriate ion path parameter set for each element, counting the number of cycles which have been executed so as to determine when the end of the experiment has been reached, and control of the time period for each cycle by synchronization with the AMU Timer module. At the beginning of the experiment, the ion path slave is initialized by having the master processor download the list of elements for monitoring, the parameter sets to be switched in, the number of elements to be monitored, the number of cycles to execute, and a code for the cycle period. All of the parameter values as well as the masses to use are converted into DAC units prior to the start of the analysis. Slave 1 stores the number of cycles to execute in a memory location known to the master and does a 147 decrement and test for zero at the end of each cycle. This arrangement allows the master to overwrite the current value if the operator requests an early termination of the experiment. The code for the cycle period is used to set up the AMU Timer. Unlike scans and.sweeps, where the AMU Timer is used to effect a step-by-step time synchronization, SIM and MRM use the AMU Timer to synchronize cycles. Since the AMU Timer can only be set up for a small set of specific and discrete time intervals, this approach makes it possible to optimize the amount of averaging performed by the DAq not only for the selected cycle period, but also on the basis of the number of elements being monitored, and the type of monitoring being performed. For a series of experiments, this feature allows the operator to add or subtract elements from the list while maintaining a constant cycle period. This becomes especially significant at short cycle times, where only a limited amount of averaging can be performed; the operator has the freedom to trade off the amount of data to be obtained (i.e. the number of elements to monitor) in a run against the degree of averaging to be performed. The actual data acquisition code consists of a singly- nested loop structure. The outer loop is executed for each cycle and the inner loop is executed for each element within a cycle. Each pass through the outer loop initializes and executes the inner loop, waits for the synchronization signal from the AMU Timer, decrements the cycle counter, and 148 exits when the cycle counter comes down to zero. The inner loop begins by monitoring the Link and Sync Output module from the detection slave, waiting for the signal indicating that it is done with the previous data point. Upon receipt of this signal, the ion path slave sets the quadrupoles as needed for the next ion or reaction, switches in the appropriate set of parameters, and allows a five millisecond settling time. At the end of the settling period, the detection slave is informed that the ion path is now stable by means of the Link and Sync Output module. At this point, the ion path processor also uses the Link and Sync Output module to pass the index number of the current element to the display/reduction processor. The inner loop is then repeated for each element which is being monitored. After all of the cycles have been executed, the ion path slave sets a bit in its software status byte to inform the other slaves of completion and restores the ion path to its state prior to the experiment.- When these tasks are complete, the slave returns to monitoring the command FIFO from the master processor. Display/Reduction Processor The third slave, the display/reduction processor combines the x-axis information being sent it by the ion path slave and the y‘axis information being sent it by the detection slave. This combined information is then used to provide real-time graphics and is relayed back to the master 149 processor at the end of each cycle. This slave is initial- ized by reading the number of elements being monitored, clearing its software status byte, and setting up the graphics screen. During run time, the display/reduction processor also uses a singly nested loop structure, the outer loop being executed once for each cycle, and the inner loop being executed for each element within a cycle. The outer loop begins by checking whether the master processor has finished reading the data from the previous cycle. After this has occurred, as indicated by the master clearing slave 2’s software status byte, the inner loop is executed. On completion of the inner loop, the software status byte is set, informing the master that data is now available, and the display/reduction processor checks the software status byte of the ion path slave to see whether there are additional cycles to be executed. If there are additional cycles to execute, the cycle number is checked to see if it is an integral multiple of 1000. If it is, the graphics screen is re-initialized. When the end of the experiment is indicated, the display/reduction processor returns to monitoring the command FIFO from the master. On each cycle, the inner loop is executed once for each element being monitored. This loop begins by reading the index of the current element, which is passed to it by means of the Link and Sync Input Module coming from the ion path slave. Next,.the corresponding intensity value is read from the Link and Sync Input module coming from the detection 150 processor. This intensity value is stored in an array indexed by the value which had been provided by the ion path. If this index is less than or equal to five, the intensity value is scaled and plotted for the appropriate element in the real-time display, otherwise the slave loops back for the next element. The run time tasks for the master processor include reading data values from the display/reduction slave as each cycle is executed, formatting these data into a proper record for the data file, writing this data out to disk, and monitoring the keyboard for requests for early termination of the run. This is a unique aspect of MRM and SIM data acquisition; the master has no run-time tasks during sweeps or scans. All five of the SIM/MRM run commands behave in essen- tially the same manner. The first step is to download required information to each of the slave processors. In the case of the parameter tables and the lists of elements to monitor, this is done by direct memory transfers. The remaining information is downloaded through the command FIFOs, pushing the information onto the stack of the appropriate slave. The downloading process is followed by commands to start each of the slaves. Next, the master clears the display/reduction slave software status byte, updates the current experiment header, creates a new scan header for the active data file, and enters its data acquisition loop. This loop is executed once for each cycle 151 and begins with the master monitoring the keyboard to see if the "Q/q" key has been struck. If that occurs, the cycle counter on the ion path slave is set to 1, so that execution will terminate at the end of the current cycle. In any case, the master breaks out of the keyboard monitor loop when slave 2, display/reduction, sets its software status byte to indicate that a full cycle of data has been acquired. The master then reads the intensity values from the slave by means of interprocessor reads and clears the slave 2 software status byte when it’s done. These values are then formatted into a data record which includes the mass settings of the first and third quadrupoles for each element (storing a —l for an RF-only quadrupole). The data record is written out to disk at the end of the active data file and the master increments its cycle counter. The last step in the loop is to test the software status byte from the ion path slave to see if it’s indicating the end of the experiment. When the end of the experiment is indicated, the master updates the experiment and scan headers in the data file to reflect the number of data records which were actually obtained. Once this is completed, the master is done executing the MRM/SIM and can return to accepting new commands from the user. At the end of a SIM/MRM experiment, the data are stored as a scan within the most recently defined experiment in a file maintained by the FORTH Database Management System (65). This scan is stored in a manner which is compatible 152 with the data structures used by sweep and scan data. This allows all of the data file utilities, such as experiment directories, scan directories, data listings and display commands, to be used in essentially the same manner as they are for scans and sweeps. Directories, experiment deletion, and scan deletion are entirely identical. Data listings are generated by a special subroutine to the normal command, DLIST, and no additional commands are required. Displays are also generated by a special subroutine, this time to the normal display command, DISP. The greater flexibility required by the display subroutine, however, requires a special display set-up command CSET (chromatogram set-up) to be used instead of the DSET command which is used to format sweep and scan displays. The CSET command allows the user to specify which elements to display (to a maximum of five at one time) and over which time interval. Each chroma- togram is separately normalized over the time interval used for the current display. Additionally, there is a command NEXT which allows the user to display a set of elements as a series of displays containing successive time windows of constant width, without having to return to the CSET command. As is the case with sweeps and scans, archiving of data and generation of high quality plots is performed on the PDP-ll/23 data system. Data may be uploaded to the data system by means of a lG-bit parallel link. Unlike scans and sweeps, however, MRM/SIM data cannot be conveniently 153 incorporated into the MSU-LLNL Multi—Dimensional Database structure (66). This is because the data points cannot be represented as an ordered, x-y pair, as required by the multi-dimensional database. Instead, they must be repre- sented as ordered triples, composed of quadrupole 1 mass, quadrupole 3 mass, and intensity. Thus, a new set of utilities had to be written to process these data on the PDP-ll/23 data system. Data from the control system are uploaded to the data system as "experiments". Each experiment, a series of 64~ byte records, consists of some header information, user— supplied comments, and headers and data for one or more "scans", which are presumably related. Each scan corre- sponds to an MRM/SIM run and consists of a header record, the full set of parameter values for each element monitored, and one data record for each cycle (regardless of how many elements were actually monitored). The control system uploads one experiment at a time up the 16-bit parallel link to the data system with transmission of a checksum for each record. The receiving task on the data system, UPLOAD (67), creates a specially formatted file consisting of each 64- byte record as it is received. A pair of tasks, GCTRAN and GCXFEH, were written to take the file created by UPLOAD, and separate it into one direct-access file for each of the scans in the experiment. Each one of the scan files contains the scan header, all of the comments for the experiment, the scan parameter sets, and all of the scan 154 data. In addition, the user is prompted for any additional comments to go into each individual scan file as it is being generated. Appendix 8 Program Listings for Single and Multiple Processor Scans and Sweeps This appendix is included to provide partial listings of the programs discussed in Chapter 3. Sweeps Single Processor Register Usage - 0 - scratch 1 - counts steps 2 - current x value I - address of swept device W — pointer to next slot in data buffer Labels - AMUF/F - address to reset amu timer . STEP - step size, in DAC units, for swept device TESTMUX — connects CPU test input to amu timer daqbaseZ - address of lowest data byte from data acquisition board daqstat — address of data acquisition board status byte startdaq — address to start data acquisition 155 156 Code for SWEEP loop - BEGIN 2 I) MOV l # TESTMUX MOV B WAIT AMUF/F STA B 0 0 SUB 0 DEC startdaq STA B 2 0 MOV STOS 0 0 SUB STOS BEGIN daqstat LDA B 04 #B 0 TEST 0: NOT END 0 0 SUB intensity double word daqbase2 4 + LDA B STOS daqbaseZ LDA B daqbaseZ 2+ 0 HI MOV B STOS daqf/fclr LDA B STEP 2 ADD LOOP Multiple Processors Ion Path Processor Register Usage - - scratch Slur-10 I Write out the new x value Select the amu timer Wait for the amu timer Reset the amu timer Start the data acquisition board Get the current x value And place in data buffer Flag word of data buffer Get data acquisition status And test for completion Exit loop when bit is set Clear highest byte of Get second highest byte Store high word of intensity Lowest byte of intensity And next lowest byte Store low word of intensity Reset data acquisition board Increment x value Repeat until end of sweep - count number of steps store current x value ~ address of device being swept Labels - AMUF/F - STEP - TESTMUX - XDATA - XSTB - YF/F - xstable - 157 address to reset amu timer the step size, in DAC units, for swept device multiplexes CPU test input to amu timer or Link and Sync module address to write x value to reduction micro strobe to transmit x value to reduction micro address to reset ok from detection micro address to send ok to sample to detection slave Code for SWEEP loop - CREATE ?YSAMPLED 0 0 SUB TESTMUX STA B WAIT YF/F STA B RET CREATE XSTABLE l # 0 MOV TESTMUX STA B WAIT AMUF/F STA B xstable STA B RET CREATE >PEAK 2 XDATA MOV XSTB STA B reduction BET Subroutine checks for detection ' micro done sampling previous point Clear accumulator and select The Link and Sync module Wait for an ok from the detection processor Clean the y status flip-flop And return to calling program Subroutine syncs with amu timer and gives an ok to sample to detection micro Select the amu timer Wait for the amu timer And reset it Send ok to detection slave And return to calling program Subroutine to.send current x value to reduction micro Write the current x value Strobe the value over to the micro And return to calling program BEGIN ?YSAMPLED CALL 2 W) MOV XSTABLE CALL >PEAK CALL STEP 2 ADD LOOP 158 Check that detection slave is done Write the new x value out to the swept device Sync with the amu timer and send ok to sample to the detection micro Send current x value to the reduction micro Calculate the next step value Loop through the desired number of steps Detection Processor Register Usage — 0 - scrat 2 ch, high word of intensity value - low word of intensity value Labels - PF/F SSTAT - TESTMUX XF/F - YDATA - YSTROBE daqbaseZ - daqf/fclr daqstat — startdaq - address to clear the reduction sync. flip-flop base address of processor software status bytes multiplexes CPU test input to the synchronization flip flops of the ion path and reduction micros address to clear ion path sync. flip-flop address for sending a data word to the reduction micro address to transmit data to the reduction micro address of low data byte from the data acquisition board address to clear the completion flip-flop from the data acquisition board status byte of the data acquisition board address to start the data acquisiton board Code for SWEEP loop - CREATE ACQUIRE 0 0 SUB 0 DEC startdaq STA BEGIN daqstat LDA B 04 # 0 TEST B 0: NOT END daqbase2 2 MOV B daqbase2 2+ 2 HI MOV B 0 0 SUB daqbase2 4 + LDA B daqf/fclr LDA B RET BEGIN 0 #B TESTMUX MOV WAIT XF/F STA B CALL ACQUIRE YSAMPLED STA B YDATA STA 2 YDATA MOV YSTROBE STA B 1 #B TESTMUX MOV WAIT PF/F STA B SSTAT 1+ LDA B 0 SHR CS END Subroutine to control the data acquisition board Send start command to the data acquistion board Get the status byte Test the completion bit Exit when bit is set Get the lowest byte And the next lowest byte Clear the top byte of the intensity Get the third intensity byte Clear the data acquisition board And return to calling program Start of acquisition loop Select the ion path monitor Wait for ok from the ion path micro And reset Obtain an intensity value Give the ion path micro an ok to step the x value Send intensity high word to the reduction micro Send intensity low word to the reduction micro Strobe the values across Select reduction micro monitor Wait for acknowledge from the reduction micro And reset the the flip-flop Get the ion path micro’s software status byte Look at its low bit Carry set implies the SWEEP has been completed 160 Reduction Processor Register Usage — 0 ~ scratch, x axis value, low intensity word 2 - high word of intensity W - pointer to next location in the data buffer Labels - RSTROBE - address to strobe over intensity data from the detection slave SSTAT - base address of processor software status bytes XDATA - address of data sent from ion path slave XF/F - address to reset the ion path flip-flop XSTROBE - contains status of flip-flop indicating x axis data available YDATA - address of intensity data sent from the detection micro address to reset the detection flip-flop contains status of flip-flop indicating intensity data available YF/F YSTROBE Code for SWEEP loop - CREATE ?XDATA Subroutine to read the x axis value sent from the ion path slave BEGIN XSTROBE LDA B Monitor the strobe flip-flop 0 SEE B from the ion path micro CS Carry bit sent indicates data available END XF/F LDA B Clear the flip-flop XDATA LDA And read the x value RET Return to the calling program CREATE ?YDATA BEGIN YSTROBE LDA B 0 SHR B CS END RSTROBE LDA B YF/F LDA B YDATA 2 MOV YDATA LDA RET BEGIN ?XDATA CALL STOS o o SUB STOS ?YDATA CALL STOS 2 o MOV STOS SSTAT 1+ LDA 0 SHR CS END Single Processor Register Usage - svduncroedo l - scratch - loop counter, — scratch, 161 Subroutine to read the 32-bit intensity value from the detection slave Monitor the strobe flip—flop from the detection micro Carry bit set indicates data available Strobe the data in Clear the flip-flop Read in the high word Read in the low word Return to the calling program Get the x axis value And store it in the data buffer Clear the accumalator And store in the flags word Get the intensity value Store the low word Move in the high word And store it Get the software status byte for the ion path slave Check.its low bit Carry bit set implies that the SWEEP is complete Neutral Loss Scans intensity low word intensity high word intensity high word — stack pointer - intensity low word, pointer — pointer into data buffer Labels - (thld) AMUF/F FRAC FRAC 2+ MlDAC MZDAC M3DAC MASSl MASSS MASS3FRAC STEP TESTMUX daqstat kflag max-peak mwidth pwidth startdaq xlast xmax xvalue ylast yprev 162 the working intensity threshold address to reset the amu timer integer portion of the quad 3 mass step size fractional size address of address of address of portion of the quad 3 mass step the quad 1 mass dac the quad 2 rf-level the quad 3 mass dac the quad 1 mass for the next step the quad 3 mass for the next step the fractional part of the quad 3 mass for the next step the step size for quadrupole 1 mass multiplexes the amu timer to the CPU test input the data acquisition board status byte flag cleared when on a peak tail running intensity maximum for current peak the maximum allowed peak width the minimum allowed peak width address to start the data acquisition board counts the peak width quad 1 mass value at the current peak maximum current quad 1 mass value counts consecutive points above threshold intensity for the previous step dac Code for NSCAN Loop - CREATE NEWTHRESR max-peak 2+ 1 SUB max-peak 2 SBB 0< IF xvalue LDA xmax STA I max—peak 2+ MOV U max-peak MOV U U OR 0: IF I SEE I (thld) MOV threshold ELSE 8000 # (thld) MOV THEN THEN RET CREATE ?RISING yprev 2+ 1 SUB yprev 2 SBB 0< NOT IF I yprev 2+ MOV U yprev MOV ELSE 0 # kflag MOV THEN RET 163 Subroutine to check for a new peak maximum Is the current intensity value greater than the current peak maximum -ve implies a new maximum Get the current mass value And save it Save the intensity low word And the intensity high word Look if the high word is zero True if high word is zero Take half of the current intensity And use it as the new High word is non-zero Use 32k as new threshold Return to the calling program Subroutine to check for a rising peak Compare the current intensity to the previous point 2 0 implies a falling peak Save the intensity low word And the intensity high word The peak is rising Clear the tailing peak flag Return to the calling program 164 CREATE ?PEAK daqbase2 l MOV B daqbase2 2+ 1 HI MOV B 2 2 SUB daqbase2 4 + 2 MOV B 2 U MOV l I MOV (thld) l SUB 0 # 2 SBB 0< NOT IF ylast INC kflag LDA 0 0 OR 0: IF xlast INC xlast LDA 0 mwidth CMP CS IF xmax # I MOV 4 # l MOV MOVS'REP 0 0 SUB xmax STA xlast STA max-peak STA max-peak 2+ STA 1 # kflag MOV ELSE NEWTHRESH CALL THEN ELSE ?RISING CALL THEN Subroutine to perform peak- finding Get lowest intensity byte And the next lowest byte Clear intensity high word And get top intensity byte Make a copy of the high word And the low word Test if current point is above the threshold 3 implies above threshold Count number of points above threshold Check the tailing peak flag Zero implies peak was falling Counts points since start of peak Get the peak width And compare it with the maximum peak width Carry set implies peak exceeds maximum width Address of beginning of data Move 4 words into data buffer Do it Clear the accumulator And re-initialize M Set the tailing peak flag Check if this is a new maximum Possible tailing peak 165 ELSE ylast l MOV 0 # ylast MOV 1 1 OR 0: NOT IF xlast LDA 0 pwidth CMP B CS IF xmax # I MOV 4 # l MOV MOVS REP 0 0 SUB xmax STA xlast STA max—peak STA max-peak 2+ STA l # kflag MOV THEN THEN 0 # kflag MOV THEN RET Get the current value And clear it out Check it for zero # implies previous point was above threshold Get number of points over threshold And compare to minimum peak width Carry set implies minimum width satisfied - save the peak Start of data area Number of word to move Move them into data buffer Put a zero in accumalator And re-initialize " H n Set the tailing peak flag Clear the tailing peak flag And return to the calling program BEGIN MASSl 2 MOV 2 MlDAC MOV MASS3 LDA M3DAC STA 0 SHR MZDAC STA l #B TESTMUX MOV WAIT AMUF/F STA B 0 0 SUB 0 DEC startdaq STA B 2 xvalue MOV STEP 2 ADD 2 MASSl MOV MASSB LDA MASS3FRAC 2 MOV FRAC 2+ 2 ADD FRAC 0 ADC MASSS STA 2 MASS3FRAC MOV BEGIN daqstat LDA B 04 #B 0 TEST 0: NOT END 1 PUSH ?PEAK CALL 1 POP LOOP Multiple Processors Ion Path Register Usage - - scratch — loop counter quad 1 mass SHNv-do I 166 value dac value 1 mass to the 3 mass Get And Get And Use the quad write it the quad write it to the dac half the mass 3 value for the quad 2 level And write it to the dac Select the amu timer And wait for it Reset the amu timer Clear the accumulator And decrement to -1 Start the data acquisition board Save the quad 1 mass value Add in the next step And save the new value Get the current quad 3 mass value the current quad 3 fractional mass in the fractional step size the integer part, with carry out of the fraction Save the integer part And the fractional part Get Add And Get status of the data acquisition board Test the completion bit Bit set implies data is available Loop until data available Save the loop counter Do the peak finding Restore the loop counter Execute the next step - fractional part of quad 3 mass - integer part of quad 3 mass 167 Labels - FRAC - integer part of quad 3 mass step size FRAC 2+ - fractional part of quad 3 mass step size MIDAC - address of quad 1 mass dac MZDAC - address of quad 2 rf level dac MBDAC - address of quad 3 mass dac STEP - step size for quad 1 mass Code for NSCAN Loop - Subroutines ?YSAMPLED, XSTABLE, and >PEAK are the same as used in the SWEEP loop. BEGIN ?YSAMPLED CALL Wait until the detection micro is done with the previous point 2 MlDAC MOV Set the new quad 1 mass W 0 MOV Get the new quad 3 mass M3DAC STA And set quad 3 to this value 0 SHR Divide it by two MZDAC STA And set the quad 2 rf level XSTABLE CALL Inform the detection micro that the next step is ready >PEAK CALL Send the current quad 1 mass to the reduction micro STEP 2 ADD Calculate the next quad 1 mass value FRAC 2+ I ADD Calculate the fractional part of the next quad 3 mass . value FRAC W ADC Calculate the integer part of the next quad 3 mass value LOOP Loop through the desired number of steps Detection Same code as used for SWEEP 168 Reduction Register Usage - 0 - scratch I - intensity low word 2 - intensity high word U — intensity high word S - stack pointer I — intensity high word, pointer W — pointer to data buffer Labels - (thld) - the current intensity threshold kflag - flag set to indicate a tailing peak max-peak - intensity of current peak maximum mwidth - the maximum allowed peak width pwidth - the minimum allowed peak width xlast - counts peak width xmax - quad 1 mass at maximum intensity of current peak xvalue — current quad 1 mass ylast - counts points above threshold yprev — intensity for the previous step 169 Code for NSCAN Loop - CREATE NEWTHRESH Subroutine to check for a new max-peak 2+ 1 SUB max-peak 2 SBB 0a IF xvalue LDA xmax STA I max-peak 2+ MOV U max-peak MOV U U OR 0:. IF I SHR I (thld) MOV threshold ELSE 8000 # (thld) MOV THEN THEN CREATE ?RISING yprev 2+ 1 SUB yprev 2 SBB 0< NOT IF I yprev 2+ MOV U yprev MOV ELSE 0 # kflag MOV THEN peak maximum Is the current intensity value greater than the current peak maximum -ve implies a new maximum Get the current mass value And save it Save the intensity low word And the intensity high word Look if the high word is zero True if high word is zero Take half of the current intensity And use it as the new High word is non-zero Use 32k as new threshold Return to the calling program Subroutine to check for a rising peak Compare the current intensity to the previous point 2 0 implies a falling peak Save the intensity low word And the intensity high word The peak is rising Clear the tailing peak flag Return to the calling program 170 BEGIN ?XDATA CALL Get the quad 1 mass value xvalue POP And save it ?YDATA CALL Get the intensity value 2 POP The high word 1 POP And the low word 2 U MOV And make copies 1 I MOV (thld) l SUB Check if the intensity is 0 # 2 SBB Above threshold 0< NOT 3 implies above threshold IF ylast INC Count points above threshold flag LDA Get the tailing peak flag 0 0 OR Check it for zero 0: 0 implies not a tail IF xlast INC Count the peak width xlast LDA And get the value 0 mwidth CMP Test if we’ve exceeded the maximum width CS Carry set if wide peak IF xmax # I MOV Start of data area 4 # l MOV Number of words to move MOVS REP Move peak data into data buffer 0 0 SUB Get a zero xmax STA And re-initialize xlast STA ' " max—peak STA " max-peak 2+ STA .' " l # kflag MOV Set the tailing peak flag ELSE Peak is not too wide NEWTHRESH CALL Check for a new maximum and adjust (thld) THEN ELSE Tail of a previous peak ?RISING CALL See if data is rising again. If so, clear kflag THEN END 17] ELSE ylast 1 MOV 0 # ylast MOV l 1 OR 0: NOT IF xlast LDA 0 pwidth CMP B CS IF ’ xmax # I MOV 4 t l MOV MOVS REP 0 0 SUB xmax STA xlast STA max-peak STA max-peak 2+ STA l # kflag MOV THEN THEN 0 # kflag MOV THEN SSTAT 1+ LDA B 0 SHR CS Peak is below threshold Get the count above threshold And zero the counter Check if the previous point was above threshold ¢ 0 implies it was Get the peak width And see if it meets the minimum width Carry set implies peak is wide enough Start of data area 4 words to save Transfer to data buffer Get a zero And re-initialize H H '9 Set the peak tail flag Peak ended below threshold. Cancel the flag Get the ion path micro software status byte Get its low bit Carry set implies end of scan LIST OF REFERENCES LIST OF REFERENCES ASTM Standard Test Method 0 2425-83, ”Hydrocarbon Types in Middle Distillates by Mass Spectrometry," Annual Book of ASTM Standards, American Society for Testing and Materials, Philadelphia, PA, 1984, Volume 5.02, p 376. ASTM Standard Test Method D 2786-81, "Hydrocarbon Types Analysis of Gas-Oil Saturates Fractions by High Ionizing Voltage Mass Spectrometry,” ibid., p 737. ASTM Standard Test Method D 2789-81, "Hydrocarbon Types in Low Olefinic Gasoline by Mass Spectrometry," ibid., p 751. Aczel, T., Rev. Anal. Chem;, 1972, 1, 226. Hastings, S.H., Johnson, B.H., and Lumpkin, H.E., Anal. Chem;, 1956, 28, 1243. Aczel, T., Allan, D.E., Harding, J.H., and Knipp, E.A., Anal. Chem., 1970, 42, 341. Burlingame, D.L., Dell, A., and Russell, D.H., Anal. Chem., 1982, 54, 29. Issachar, 0., Holland, J.F., and Sweely, C.C., Anal. Chem., 1982, 54, 29. Yost, R.A. and Enke, 0.6., J. Amer. Chem. Soc., 1978, 100, 2274. 173 10. ll. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. Yost, R.A. and Enke, C.G., Anal. Chem;, 1979, 51, 1251A. Hunt, D.F., Shabanowitz, J., and Giordani, A.B., Anal. Chem., 1980, 52, 386. Shen, J. and Bradley, M.P.T., Anal. Chem., 1976, 48, 2291. Enke, C.G., Proc. 28th IUPAC Conference, Vancouver, B.C., (1981). Kirby, H., Sokolow, S., and Steiner, U., ”Data Systems for Tandem Mass Spectrometry", in Tande! Mass Spectrometry, edited by F.W. McLafferty, John Wiley & Sons, New York, 1983. Enke, C.G., ibid. Newcome, B.H., PhD Dissertation, Michigan State University, 1983. Myerholtz, C.A., PhD Dissertation, Michigan State University, 1983. Myerholtz, C.A., ibid. Newcome, B.H., private communication. Newcome, B.H., PhD Dissertation, Michigan State University, 1984. Busch, K.L. and Cooks, R.G., ”Analytical Appli- cations of Tandem Mass Spectrometry", in Tandem Mass Spectrometry, McLafferty, F.W., editor, John Wiley & Sons, Inc., New York, 1983. 174 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. Michigan State Gregg, Anal. Chem., Myerholtz, C.A., PhD Dissertation, University, 1983. Crawford, R.W., Brand, H.R., Wong, C.M., H.R., Hoffman, P.A., and Enke, C.G., 1984, 56, 1121. Hoffman, P.A., private communication. Atkinso n, T. V., private communication. Digital Equipment Corporation, Harriso Spectro 1983, p McLaffe n, A.G., Chemical Ionization Mass Maynard, MA. metry, CRC Press, Inc., 63. rty, F.W., Boca Raton, FL, Interpretation of Mass Spectra, Third Edition, University Science Books, CA, 1980, p 92. Valley, Bauer, Univers Stan, H. 104. Bauer, Crawfor H.R., H 1984, 5 CRC Handbook of Chemistry and Physics, M.R., Maste ity, 1983. Mill rs Thesis, Michigan State —J., Fresenius Z. Anal. Chem., 1977, 287, M.R., ibid. d, R.W., Brand, H.R., Wong, C.M., Gregg, offman, P.A 6, 1121. ., and Enke, C.G., Anal. Chem., editor, CRC Press, Boca Raton, 175 FL, 1981, Weast, R. C. p E~65. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. Watson, J. T., Introduction to Mass Spectrometry, Raven Press, New York, 1985. Beynon, J.H., Saunders, R.A., and Williams, A.E., The Mass Spectra of Organic Molecules, Elsevier Publishing Co., London, England, 1968. Bauer, ibid. Dahlin, H.E., Daniel, S.H., and Worstell, J.H., Fuel, 1981, 60, 477~480. Tutubalina, V.P., Gabdrakhmanov, F.G., and Korotkova, E.G., Deposited Doc. 1980, SPSTL 460th-D80. Hazlett, R.N. and Hall, J.M., Prepr. - Am. Chem. Soc.,Div. Pet. Cheap, 1981, 26 2., 613-619. Worstell, J.H. and Daniel, S.R., Fuel, 1981, 60, 481-484. Taylor, W.F., Ind. Eng; Chem., Prod. Res. Develop., 1976, 15, 64. Bol’shakov, G.F., Izv. Vyssh. Uchebn. Zaved., Neft Gaz, 1976, 19 5., 51-53. Chertkov, Y.B., Chem. Technol. Fuel Oils, 1976, 154. Heneman, F.G., Gov. Rep. Announce. Index, 1981, 81(26), 5656. McLafferty, F.W. and Bockoff, F.M., Anal. Chem., 1978, 50(1), 69. 176 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. Myerholtz, C.A., PhD Dissertation, Michigan State University, 1984. Hunt, D.F., and Shabanowitz, J., Anal. Chem.; 1982, 54, 574. Crawford, R.W., Brand, H.R., Wong, C.M., Gregg, H.R. Hoffman, P.A., and Enke, C.G., Anal. Chem;, 1984, 56, 1121. McLafferty, F.W., Interpretation of Mass Spectra, Third Edition, University Science Books, Mill Valley, California, 1980, p 286. Beynon, J.H., Saunders, R.A., and Williams, A.E., The Mass Spectra of Organic Molecules, Elsevier Publishing Co., Amsterdam, The Netherlands, 1968, p 354. Harrison, A.G., Chemical Ionization Mass Spectrometry, CRC Press, Inc., Boca Raton, FL, 1983, p 63. ibid., p 32. Stan, H.—J., Fresenius Z. Anal. Chem., 1977, 287, 104. Bauer, M. R., Masters Thesis, Michigan State University, 1984. 'Chertkov, Y.B., Chem; Technol. Fuel Oils, 1976, 154. Heneman, F.C., Gov. Rep. Announce. Index, 1981, 81 26., 5656. Page, R. P., private communication. 177 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. Newcome, B.H., PhD Dissertation, Michigan State University, 1983. ibid. Newcome, B.H., private communication. Newcome, B.H., PhD Dissertation, Michigan State University, 1983. Myerholtz, C.A., PhD Dissertation, Michigan State University, 1983. Moore, C.H., Astron. Astrophys. Suppl., (1974), 15, 497. James, J.S., Byte, (1980), 5, No. 8, 100. Hicks, S.M., Electronics, March 15, 1979, p. 114. FORTH, Inc., Hermosa Beach, CA. Crawford, R.W., Brand, H.R., Wong, C.M., Gregg, H.R., Hoffman, P.A., and Enke, C.G., Anal. Chem., 1984, 56, 1121. Hoffman, P.A., private communication. 178