Computer-Enhanced Analytical Spectroscopy

The Second Hidden Peak Symposium on Computer-Enhanced Analytical Spectroscopy, held in June, 1988, at the Snowbird Resort (Salt Lake City, Utah), centered around twelve keynote lectures delivered by some of the foremost experts and pioneers in this rapidly expanding field. The editor is highly indebted to each of these colleagues for contributing a chapter to the second volume of Computer-Enhanced Analytical Spec­ troscopy. The primary objective of this volume is to present a repre­ sentative cross-section of current activities in the field while balancing out the lighter coverage of some topics and areas in Volume 1. An exciting new topic, remote IR sensing, is covered in Chapters 4 and 5. Deconvolution and signal-processing methods have now been extended to UV/VIS (Chapter 1) and GC/MS (Chapter 3) applications. Furthermore, the development and testing of novel factor analysis techniques in the areas of UV /VIS and IR spectroscopy are discussed in Chapters 2 and 12, respectively. Fundamental aspects of library search techniques are presented in Chapters 7 (MS) and 9 (NMR). Chapters 6, 10, and 11 cover selected uses of expert systems in NMR, IR, and MS, respectively. Finally, an integrated expert system approach to the interpretation of GC/IR/MS data is outlined in Chapter 8. In an attempt to facilitate access to the various topics for the newcomer to the field, the twelve chapters have been organized into two main parts: Unsupervised Methods: Spectral Enhancement, Deconvolu­ tion, and Data Reduction, and Supervised Methods: Expert Systems, Modeling, and Quantitation.

I: Unsupervised Methods: Spectral Enhancement, Deconvolution, and Data Reduction.- 1 Advances in Regression: Use of Models in Spectroscopic Data Analysis.- 1.1. Introduction.- 1.2. Analysis of Zero-Dimensional Spectroscopic Data (Numbers).- 1.2.1. Method of Maximum Likelihood.- 1.2.2. Maximum Likelihood Quantitative Estimates for Peaks.- 1.2.3. Application to Photoacoustic Spectroscopy.- 1.3. Analysis of One-Dimensional Spectroscopic Data (Spectra, Waveforms).- 1.3.1. Spectrophotometry of Mixtures.- 1.3.2. Analysis of Errors in Linear Regression.- 1.3.3. Selecting “Analytical” Wavelengths.- 1.3.4. Weighting Observations in One-Dimensional Linear Regression.- 1.3.5. Application to Time-Resolved Fluorescence Spectroscopy.- 1.4. Two-Dimensional Spectroscopic Measurements.- 1.4.1. Combinations of Correlated and Uncorrelated Dimensions.- 1.4.2. Modeling the Correlated Dimension: pH/UV Data Analysis.- 1.4.3. Acid-Base Mixture Resolution and Error Predictions.- 1.5. Conclusions.- References.- 2 Factor Analysis of Spectro-Chromatographic Data.- 2.1. Introduction.- 2.2. Background.- 2.3. Detection of Overlapped Peaks.- 2.3.1. Estimating the Number of Factors Using RE.- 2.3.2. Estimating the Number of Factors Using IND.- 2.3.3. Estimating the Number of Factors Using REV.- 2.3.4. Detection of Too Many Components.- 2.3.5. Net Signal and Detection of Minor Components.- 2.3.6. Effect of Measurement Error on Net Signal.- 2.3.7. Effect of Chromatographic Resolution on Net Signal.- 2.3.8. Effect of Spectral Dissimilarity on Net Signal.- 2.3.9. Analysis of Simulated Data.- 2.3.10. Analysis of Experimental Chromatrographic Data.- 2.3.11. Conclusions.- 2.4. ITTFA Self-Modeling Curve Resolution.- 2.4.1. Recent Refinements.- 2.4.2. Sample ITTFA Curve Resolution Results.- 2.5. Conclusions.- References.- 3 Isolation of Pure Spectra in GC/MS by Mathematical Chromatography: Entropy Considerations.- 3.1. Introduction.- 3.2. Alternating Regression.- 3.2.1. TTie Overall Algorithm.- 3.2.2. Selecting the Starting Basis.- 3.2.3. Constraining the Solution.- 3.3. Optimal Preprocessing of Data.- 3.3.1. The Role of the Apodization Function.- 3.3.2. Selecting the Optimal Band-Pass.- 3.4. Balancing the “Fit” and Entropy.- 3.4.1. The Object Function in Spectrum Reconstruction..- 3.4.2. Measuring the Fit.- 3.4.3. Measuring the Entropy.- 3.5. Applications of AR.- 3.5.1. Steroid Mixtures.- 3.5.2. Aroma Compounds.- 3.6. Conclusions.- References.- 4 Signal Processing Techniques for Remote Infrared Chemical Sensing.- 4.1. Introduction.- 4.2. The Passive Remote Sensing Problem.- 4.3. Frequency-Domain Algorithms.- 4.3.1. The Linear Discriminant.- 4.3.2. Improved Frequency-Domain Filtering Techniques.- 4.3.3. Apodization Methods for Digital Filtering.- 4.4. Time-Domain Digital Filtering.- 4.4.1. Linear Finite-Impulse Response Fibers.- 4.4.2. Matrix Filters.- 4.5. Maximum Entropy Method Transformations.- 4.6. Digital Signal Processing Hardware.- 4.6.1. General-Purpose Processors versus the DSP.- 4.6.2. DSP Architecture.- 4.6.3. Classes of DSPs.- 4.6.4. DSP Hardware for Remote Chemical Sensors.- 4.7. Conclusions.- References.- 5 Imaging Spectrometry of the Earth: A Breakthrough and a Nightmare.- 5.1. Introduction.- 5.2. Background.- 5.3. Spectral Properties of Earth Surface Materials.- 5.3.1. Minerals.- 5.3.1.1. Electronic Processes.- 5.3.1.2. Vibrational Processes.- 5.3.2. Vegetation.- 5.4. Imaging Spectrometry.- 5.4.1. Sensors.- 5.4.1.1. Airborne Imaging Spectrometer.- 5.4.1.2. Airborne Visible and Infrared Imaging Spectrometer.- 5.4.1.3. The High-Resolution Imaging Spectrometer.- 5.5. Imaging Spectrometer Data Analysis.- 5.5.1. The Nightmare.- 5.5.2. Analysis Status.- 5.5.2.1. Present Approach.- 5.6. Conclusions.- References.- 6 Computer-Enhanced Nuclear Magnetic Resonance Spectroscopy.- 6.1. Introduction.- 6.2. History.- 6.2.1. Modem Computer Technology and NMR Data Acquisition.- 6.3. Current Enhancement of Data Reduction in NMR Spectroscopy.- 6.3.1. Curve Fitting.- 6.3.2. Maximum Entropy Processing.- 6.3.3. Linear Prediction Processing.- 6.3.4. Baseline Conditioning.- 6.3.5. Other Methods for One-Dimensional NMR Spectra.- 6.3.6. 2-D Spectroscopy.- 6.4. New Trends in NMR Software.- 6.5. Coming Trends in Computer Hardware.- 6.6. Laboratory Computer Networking.- 6.7. Conclusions.- References.- II: Supervised Methods: Expert Systems, Modeling, and Quantitation.- 7 Computer Identification of Mass Spectra.- 7.1. Introduction.- 7.1.1. Computer Identification Algorithms.- 7.1.2. Matching versus Interpretative Systems.- 7.2. Experimental.- 7.3. Discussion.- 7.3.1. The Reference File.- 7.3.2. Probability-Based Matching (PBM).- 7.3.2.1. The Statistical Basis of PBM.- 7.3.2.1a. Data Weighting.- 7.3.2.1b. Performance Evaluation.- 7.3.2.1c. Reliability Ranking.- 7.3.2.2. The Impurity/Artifact Problem.- 7.3.2.2a. Peak Flagging.- 7.3.2.2b. Reverse Searching.- 7.3.2.2c. Subtraction of Retrieval References.- 7.3.2.3. The Problem of Abundance Value Variation.- 7.3.2.4. Examples.- 7.3.2.5. Exact-Mass PBM (EPBM).- 7.3.2.6. PBM Speed.- 7.3.3. Self-Training Interpretive and Retrieval System (STIRS).- 7.3.3.1. The STIRS Approach.- 7.3.3.2. Automated Substructure Prediction.- 7.3.3.3. Prediction of Molecular Weight.- 7.4. Conclusions.- 7.4.1. Performance.- 7.4.2. Future.- 7.4.3. Significance.- References.- 8 A Distributed Expert System for Interpretation of GC/IR/MS Data.- 8.1. Introduction.- 8.1.1. Background.- 8.1.2. Design Philosophy.- 8.1.3. System Architecture.- 8.2. Spectral Data Experts.- 8.2.1. Role of the Data Expert.- 8.2.2. Requirements as Servers.- 8.2.3. The IR Expert.- 8.2.3.1. Design.- 8.2.3.2. IR Rules.- 8.2.3.3. Results.- 8.2.4. The MS Expert.- 8.2.5. Structure Confirmation.- 8.2.6. The Librarian.- 8.3. The Controller.- 8.3.1. Chemical Knowledge.- 8.3.2. Interface with Data Experts.- 8.3.3. Search Strategy.- 8.3.3.1. Initial Context.- 8.3.3.2. Functional Group Analysis.- 8.3.3.3. Hypotheses.- 8.3.3.4. Structure Generation.- 8.3.3.5. Search Results.- 8.4. The Reasoner.- 8.4.1. Requirements for the Reasoner.- 8.4.2. Belief Systems.- 8.4.2.1. Dempster-Shafer Theory.- 8.4.2.2. Bayesian Probabilities.- 8.4.2.3. What Is Evidence?.- 8.4.3. Explanations.- 8.5. Conclusions.- References.- 9 Computer-Aided Solutions to 13C NMR Spectral Interpretation Problems.- 9.1. Introduction.- 9.2. Overview of Computer-Based Spectral Interpretation Strategies.- 9.3. 13C NMR Library Searching.- 9.3.1. Overview of Library Searching Methods.- 9.3.2. Overall Search Design.- 9.3.2.1. Clustering and Partitioning.- 9.3.2.2. Line Mapping.- 9.3.2.3. Scoring.- 9.3.3. Evaluation of Search Results.- 9.3.4. Reverse Searching.- 9.3.4.1. Reverse Search Design.- 9.3.4.2. Evaluation of Reverse Search Results.- 9.3.5. Characterization of Library Searching Performance.- 9.4. 13C NMR Spectral Simulation.- 9.4.1. 13C NMR Spectra of Carbohydrates — A Simulation Example.- 9.4.2. Spectrum Simulation Based on Parametric Modeling.- 9.4.3. Assembly of Data.- 9.4.4. Molecular Mechanics Computations.- 9.4.5. Definition of Atom Groups.- 9.4.6. Design of Structural Parameters.- 9.4.6.1. Distance Parameters.- 9.4.6.2. Energy Parameters.- 9.4.6.3. Induction Parameters.- 9.4.6.4. Topological Complexity.- 9.4.7. Development of Models.- 9.4.8. Test of Predictive Ability.- 9.4.9. Characterization of Results.- 9.5. Conclusions.- References.- 10 Expert System for Interpretation of the Infirared Spectra of Environmental Mixtures.- 10.1. Introduction.- 10.2. Comparison of intRpret and PAWMI.- 10.2.1. PUSHSUB.- 10.2.2. STO.- 10.2.3. AUTOGEN.- 10.2.4. PAIRS.- 10.2.5. PAIRSPLUS.- 10.3. Application to Air Monitoring.- 10.4. Conclusions.- References.- 11 Approaches to Pyrolysis/Mass Spectrometry Data Analysis of Biological Materials.- 11.1. Introduction.- 11.2. Multivariate Py/MS Data Analysis.- 11.3. Applications of Multivariate Analysis.- 11.4. A Rule-Building Expert System.- 11.5. Py/MS Expert System.- 11.6. Classification of Bacteria.- 11.7. Conclusions.- References.- 12 Theory and Application of the CIRCOM Software for Quantitative Spectroscopic Analysis.- 12.1. Introduction.- 12.1.1. Historical Aspects.- 12.1.2. Current Trends.- 12.2. General Issues of Quantitative Infrared Analysis.- 12.2.1. Suitability of Infrared Data for Quantitation.- 12.2.2. The Effect of Spectral Bandwidths.- 12.2.3. Calibration Requirements of Indirect Methods.- 12.2.4. Limitations of Multiple Linear Regression.- 12.3. The Computational Steps in CIRCOM.- 12.3.1. General Approach.- 12.3.2. Data Compression into a Correlation Matrix.- 12.3.3. Eigenanalysis and Derivation of Factor Loadings.- 12.3.4. Derivation of Eigenspectra.- 12.3.5. MLR with Factor Loadings.- 12.3.6. Determination of Factor Loadings for an Unknown.- 12.3.7. Calculation of Property Values in Unknown Samples.- 12.4. CIRCOM Calibration of a Synthetic Data Set.- 12.5. CIRCOM Calibration of a Simple Chemical System.- 12.6. CIRCOM Calibration of a Difficult Chemical System.- 12.7. Conclusions.- References.