<p>This thesis is based on five papers addressing variance reduction in different ways. The papers have in common that they all present new numerical methods.</p><p>Paper I investigates quantitative structure-retention relationships from an image processing perspective, using an artificial neural network to preprocess three-dimensional structural descriptions of the studied steroid molecules.</p><p>Paper II presents a new method for computing free energies. Free energy is the quantity that determines chemical equilibria and partition coefficients. The proposed method may be used for estimating, e.g., chromatographic retention without performing experiments.</p><p>Two papers (III and IV) deal with correcting deviations from bilinearity by so-called peak alignment. Bilinearity is a theoretical assumption about the distribution of instrumental data that is often violated by measured data. Deviations from bilinearity lead to increased variance, both in the data and in inferences from the data, unless invariance to the deviations is built into the model, e.g., by the use of the method proposed in paper III and extended in paper IV.</p><p>Paper V addresses a generic problem in classification; namely, how to measure the goodness of different data representations, so that the best classifier may be constructed. </p><p>Variance reduction is one of the pillars on which analytical chemistry rests. This thesis considers two aspects on variance reduction: before and after experiments are performed. Before experimenting, theoretical predictions of experimental outcomes may be used to direct which experiments to perform, and how to perform them (papers I and II). After experiments are performed, the variance of inferences from the measured data are affected by the method of data analysis (papers III-V).</p>

<p>This thesis is based on studies in which the suitability of various gas chromatography (GC) injection techniques was examined for the determination of polybrominated diphenyl ethers (PBDEs) and carbazoles, two groups of compounds that are thermally labile and/or have high boiling-points. For such substances, it is essential to introduce the samples into the GC system in an appropriate way to avoid degradation and other potential problems. In addition, different types of gas chromatographic column system and mass spectrometric detectors were evaluated for the determination of PBDEs.</p><p>Conventional injectors, such as splitless, on-column and programmed temperature vaporizing (PTV) injectors were evaluated and optimized for determination of PBDEs. The results show on-column injection to be the best option, providing low discrimination and high precision. The splitless injector is commonly used for “dirty” samples. However, it is not suitable for determination of the high molecular weight congeners, since it tends to discriminate against them and promote their degradation, leading to poor precision and accuracy. The PTV injector appears to be a more suitable alternative. The use of liners reduces problems associated with potential interferents such as polar compounds and lipids and compared to the hot splitless injector, it provides gentler solvent evaporation, due to its temperature programming feature, leading to low discrimination and variance.</p><p>Increasing the injection volume from the conventional 1-3 µL to >50 µL offers two main benefits. Firstly, the overall detection and quantification limits are decreased, since the entire sample extract can be injected into the GC system. Secondly, large volume injections enable hyphenation of preceding techniques such as liquid chromatography (LC), solid phase extraction and other kinds of extraction. Large-volume injections were utilized and optimized in the studies included in this thesis.</p><p>With a loop-type injector/interface large sample volumes can be injected on-column providing low risk of discrimination against compounds with low volatility. This injector was used for the determination of PBDEs in air and as an interface for the determination of carbazoles by LC-GC. Peak distortion is a frequently encountered problem associated with this type of injector that was addressed and solved during the work underlying this thesis.</p><p>The PTV can be used as a large volume injector, in so-called solvent vent mode. This technique was evaluated for the determination of PBDEs and as an interface for coupling dynamic sonication-assisted solvent extraction online to GC. The results show that careful optimization of the injection parameters is required, but also that the PTV is robust and yields reproducible results.</p><p>PBDEs are commonly detected using mass spectrometry in electron capture negative ionization (ECNI) mode, monitoring bromine ions (m/z 79 and 81). The mass spectrometric properties of the fully brominated diphenyl ether, BDE-209, have been investigated. A high molecular weight fragment at m/z 486/488 enables the use of 13C-labeled BDE-209 as an internal surrogate standard.</p>

<p>På SSAB i Borlänge har man upptäckt att avgaser från färgen förorenar ugnarna vid härdning. Avgaserna består av lösningsmedel och andra flyktiga föreningar som sedan förkolnas och faller ner på lackerade plåtband. För att komma runt det här problemet har SSAB tillsammans med Akzo Nobel Industrial Finishes AB, i Gamleby, tillverkat en färg som inte ska ge ifrån sig något lösningsmedel och på så sätt minska bildandet av föroreningar i ugnarna. Färgen heter NOVA GreenCoat och innehåller rapsmetylester, RME, som är ett reaktivt lösningsmedel och ska binda till bindemedlet i färgen och på så sätt hindras från att avgå vid härdning.</p><p>Syftet har varit att undersöka om en viss metod fungerar bra för att samla upp avgaser vid härdning. Avgaserna analyserades med GC-MS och härdningen studerades med FTIR. Arbetet har utförts både i Borlänge och i Gamleby.</p><p>Metoden för uppsamling av avgaserna visade sig fungera, dock inte särskilt effektivt. Den skulle därför behöva optimeras.</p><p>Härdningen kunde till viss del följas med hjälp av FTIR. Förändring i IR-spektrumet syntes för hydroxyltoppen vid cirka 3500 cm^-1. En skillnad kunde ses för en topp vid 2858 cm^-1 som uppkom i det ohärdade materialet innehållande RME, men den försvann under härdning. Toppen ökade med ökad mängd RME och syntes inte alls för de prover som inte innehöll RME.</p><p>Samtliga gjorda analyser indikerade att RME avgick vid härdning. Vid analys med GC-MS kunde de föreningar som avgick från RME identifieras som: metyl-14-metyl pentadekanoat och/eller metyl hexadekanoat, olika former av metyl oktadekanoat samt olika former av metyl eikosanoat. Andra föreningar från klarlacken som avgick var lösningsmedel x och y samt i två av proverna hexametoximetylmelamin, HMMM.</p> / <p>Problems with exhaust gases from paint that contaminates the ovens have been discovered at SSAB in Borlänge.  The exhaust gases contain solvents and other volatiles that vaporize from the paint and later on carbonize and fall down on the next coming sheets. SSAB and Akzo Nobel Industrial Finishes AB have developed a paint that is not supposed to vaporize any solvents and therefore reduce the contaminations of the ovens. The paint is called NOVA GreenCoat and contains rapeseed methyl ester, RME. RME is a reactive solvent that is supposed to react with the binder in the paint and therefore be prevented from leaving during thermal cure.</p><p>The purpose was to investigate whether a specific method worked well, for collecting the exhaust gases during thermal cure, or not. The collected exhaust gases were analyzed by using GC-MS and the cure was studied with FTIR. The work was executed both in Borlänge and Gamleby.</p><p>The method for collecting the exhaust gases worked, unfortunately it wasn’t particularly effective. It therefore needs to be optimized.</p><p>The cure could partially be studied by using FTIR. Changes in the IR-spectrum could be followed by looking at the hydroxyl peak at approximately 3500 cm^-1. A certain difference was observed for a peak at 2858 cm^-1 that appeared in the uncured material containing RME. The peak disappeared during cure. It increased with increased amount of RME and could not be observed at all for the samples that did not contain RME.</p><p>All the analyzed samples indicated that RME vaporize during cure. When analyzed with GC-MS the volatiles that vaporize from RME were identified as: hexadecanoic acid methyl ester and/or penta decanoic acid methyl-14-methyl ester, different forms of octadecanoic acid methyl ester and different forms of eicosanoic acid methyl ester. Other volatiles that vaporize from NOVA GreenCoat were solvent x and solvent y and for two of the samples hexa metoxy methyl melamine, HMMM, was also found.</p>

<p>Freudenberg Household Products AB in Norrköping are manufacturer of sponge cloths with the well-known brand names of Wettex^® and Vileda^®. The production is based on the viscose fiber process and involves a high chemical demand. Recent customer complaints involve a diffuse smell from the cloths that is like a “garage odor” and occurs after a few uses. The company’s theory is that the smell derives from a chemical used in the process called Exxal 9.</p><p>The aim was to screen the washing water from two sections and the cloth before and after wash for the presence of Exxal 9 and other prominent components. The washing water samples consisted of a salt solution from one section and a water condensate from another section. A method to qualitatively and quantitatively examine the production samples was developed. To evaluate the variation over a short period of time, twelve samples were taken during four weeks. The focus for the analysis lay on production line Wx4, but comparisons with two other production lines, Wx7 and SL1, were also made. The method of choice was gas chromatography in combination with two different detectors; mass spectrometer for identification and flame ionization detector for quantification.</p><p>Exxal 9 could be identified in both of the washing water sections but in very various concentrations. At the production line Wx4, the mean concentration in the mother lye was 61.96 µl/l whereas the mean concentration in the condensate was 0.24 µl/l. The comparison between the different production lines showed significant variations, where Wx4 had the highest concentration. In the cloths, Exxal 9 could only be found before it had been washed. The concentration in the cloths was not high enough for quantification. In both the washing waters and cloths, additional unknown peaks were found. Attempts to identify all the unknowns were made but only two compounds were included in the commercial library; 2-ethyl-1-hexanol and 2-(2-butoxyethoxy)-ethanol.</p>

<p>The performance of different types of nebulisers: Vulkan direct injection nebuliser (Vulkan DIN), direct injection high efficiency nebuliser (DIHEN), microconcentric nebuliser coupled to cyclonic or double pass spray chamber (MCN-C or MCN-DP, respectively) was investigated and compared when used for sample introduction to ICP-MS or ICP-OES. With ICP-OES, in axial viewing mode, intensity distributions across the radius of the plasma (radial intensity profiles) were determined for different spectral lines with Esum 1.85-15.41 eV to determine fundamental plasma properties for various matrices using Vulkan DIN and MCN-C. The results showed that with the MCN-C the ionisation temperature (Tion) was about the same across the measured region of the plasma (±3.0 mm) whereas with the Vulkan DIN the Tion was significantly lower in the centre of the plasma. A large deviation from local thermodynamic equilibrium, as well as deteriorated stability, was observed for the plasma when using the Vulkan DIN.</p><p>With ICP-MS noise power spectra (NPS) were generated to identify sources of noise. NPS showed that the magnitude of white noise for the tested sample introduction systems decreased in the following order: Vulkan DIN > DIHEN > MCN-C > MCN-DP. This order follows the decrease of mean droplet size and span of the size distribution, indicating that the white noise is caused by spatial and temporal non-uniform desolvation and ionisation. Another source of noise arose from the peristaltic pump and the magnitude of pump interference noise decreased in the following order: DIHEN > MCN-C/DP > Vulkan DIN. Mains power interference noise and 1/f noise were lower for the direct injection nebulisers compared to the spray chamber systems. The contribution or effects of these noise components on relative standard deviations of steady-state ion-count rate and isotope ratio measurements is discussed in this thesis.</p><p>Aerosols generated by the Vulkan DIN and the DIHEN were also directly characterised using Particle Dynamic Analysis. The Vulkan DIN produced particles with a mean diameter of ~30 µm and a size distribution between 2-80 µm. With the DIHEN the corresponding values were ~11 µm and 1-40 µm, respectively, with a few particles at 55-78 µm. The mean velocity of particles from the Vulkan DIN was ~10 m s-1 and from the DIHEN ~18 m s-1. The lower velocity allows longer residence time counteracting the effects of the larger droplet size.</p>

<p>In the growing field of proteomics identification of proteins by tandem mass spectrometry (MS/MS) is performed by matching experimental mass spectra against calculated spectra of all possible peptides in a protein database. One problem with this approach is the false-positive identifications. MS-based proteomics experiments are further affected by a rather poor efficiency typical in the range of 10-15%, implicating that only a low percentage of acquired mass spectrometric data is significantly identified and assigned a peptide sequence.</p><p>In this thesis improvement in spectrum specificity is accomplished by using a combination of high-accuracy mass spectrometry and techniques that will yield complementary sequence information. Performing collision-activated dissociation (CAD) and electron capture dissociation (ECD) upon the same peptide ion will yield such complementary sequence information. Implementing this into a proteomics approach and showing the advantages of using complementary fragmentation techniques for improving peptide identification is shown. Furthermore, a novel database-independent score is introduced (S-score) based upon the maximum length of the peptide sequence tag derived from complementary use of CAD and ECD. The S-score can be used to separate poor quality spectra from good quality spectra. An-other aspect of the S-score is the development of the ‘reliable sequence tag’ which can be used to recover below threshold identifications and for a reliable backbone for de novo sequencing of peptides.</p><p>A novel proteomics-grade de novo sequencing algorithm has also been developed based upon the RST, which can retrieve peptide identification with the highest reliability (>95%). Furthermore, a novel software tool for unbiased identifications of any post-translational modifications present in a peptide sample is introduced (ModifiComb). Combining all the tools described in this thesis increases the identification specificity (>30 times), recovers false-negative identifications and increases the overall efficiency of proteomics experiements to above 40%. Currently one of the highest achieved in large-scale proteomics.</p>

<p>Electrospray ionization (ESI) mass spectrometry (MS) in combination with liquid chromatography (LC) is an excellent tool for the identification of drug metabolites. Utilizing this hyphenated technique in combination with proper sample pretreatment, the metabolic pathways of the analgesic drug ketobemidone were investigated in human urine and rat microdialysate from blood and brain. Two novel phase I metabolites (ketobemidone N-oxide and meta-hydroxymethoxyketobemidone) and three novel phase II metabolites (glucuronic acid conjugates of ketobemidone, norketobemidone and hydroxymethoxyketobemidone) were identified in human urine. Further, norketobemidone and ketobemidone N-oxide were identified in rat microdialysate from brain after regional distribution of ketobemidone in striatum. This indicates that the brain itself has the possibility to metabolize ketobemidone. </p><p>Synthetic ketobemidone metabolites were used for comparison of retention times and tandem MS spectra with the possible metabolites recovered from the biological samples. The conjugated metabolites were identified by accurate mass measurements and tandem MS spectra of the aglycones. The accuracy of the estimated masses was better than 2.1 ppm for two out of three conjugates in presence of internal standard.</p><p>On-line micro-SPE was successfully used for trapping and desalting of the microdialysates. The small SPE pre-column made it possible to inject approximately 100 times more sample on the analytical column compared to injection without pre-column. Selective trapping was demonstrated for the polar catechol amine metabolite, dihydroxyketobemidone, which forms covalent complexes with phenylboronic acid (PBA). A fluorinated silica type stationary phase was the only column out of several tested that was able to separate ketobemidone and all relevant phase I metabolites. </p><p>Liquid chromatography and mass spectrometry are independently valuable tools in the field of analytical pharmaceutical chemistry. The present study showed that the combination of LC-MS, with its excellent selectivity and sensitivity, offers an outstanding tool in the qualitative analysis of drugs and metabolites in biological fluids. </p>

This thesis is based on five papers addressing variance reduction in different ways. The papers have in common that they all present new numerical methods. Paper I investigates quantitative structure-retention relationships from an image processing perspective, using an artificial neural network to preprocess three-dimensional structural descriptions of the studied steroid molecules. Paper II presents a new method for computing free energies. Free energy is the quantity that determines chemical equilibria and partition coefficients. The proposed method may be used for estimating, e.g., chromatographic retention without performing experiments. Two papers (III and IV) deal with correcting deviations from bilinearity by so-called peak alignment. Bilinearity is a theoretical assumption about the distribution of instrumental data that is often violated by measured data. Deviations from bilinearity lead to increased variance, both in the data and in inferences from the data, unless invariance to the deviations is built into the model, e.g., by the use of the method proposed in paper III and extended in paper IV. Paper V addresses a generic problem in classification; namely, how to measure the goodness of different data representations, so that the best classifier may be constructed. Variance reduction is one of the pillars on which analytical chemistry rests. This thesis considers two aspects on variance reduction: before and after experiments are performed. Before experimenting, theoretical predictions of experimental outcomes may be used to direct which experiments to perform, and how to perform them (papers I and II). After experiments are performed, the variance of inferences from the measured data are affected by the method of data analysis (papers III-V).

This thesis is based on studies in which the suitability of various gas chromatography (GC) injection techniques was examined for the determination of polybrominated diphenyl ethers (PBDEs) and carbazoles, two groups of compounds that are thermally labile and/or have high boiling-points. For such substances, it is essential to introduce the samples into the GC system in an appropriate way to avoid degradation and other potential problems. In addition, different types of gas chromatographic column system and mass spectrometric detectors were evaluated for the determination of PBDEs. Conventional injectors, such as splitless, on-column and programmed temperature vaporizing (PTV) injectors were evaluated and optimized for determination of PBDEs. The results show on-column injection to be the best option, providing low discrimination and high precision. The splitless injector is commonly used for “dirty” samples. However, it is not suitable for determination of the high molecular weight congeners, since it tends to discriminate against them and promote their degradation, leading to poor precision and accuracy. The PTV injector appears to be a more suitable alternative. The use of liners reduces problems associated with potential interferents such as polar compounds and lipids and compared to the hot splitless injector, it provides gentler solvent evaporation, due to its temperature programming feature, leading to low discrimination and variance. Increasing the injection volume from the conventional 1-3 µL to &gt;50 µL offers two main benefits. Firstly, the overall detection and quantification limits are decreased, since the entire sample extract can be injected into the GC system. Secondly, large volume injections enable hyphenation of preceding techniques such as liquid chromatography (LC), solid phase extraction and other kinds of extraction. Large-volume injections were utilized and optimized in the studies included in this thesis. With a loop-type injector/interface large sample volumes can be injected on-column providing low risk of discrimination against compounds with low volatility. This injector was used for the determination of PBDEs in air and as an interface for the determination of carbazoles by LC-GC. Peak distortion is a frequently encountered problem associated with this type of injector that was addressed and solved during the work underlying this thesis. The PTV can be used as a large volume injector, in so-called solvent vent mode. This technique was evaluated for the determination of PBDEs and as an interface for coupling dynamic sonication-assisted solvent extraction online to GC. The results show that careful optimization of the injection parameters is required, but also that the PTV is robust and yields reproducible results. PBDEs are commonly detected using mass spectrometry in electron capture negative ionization (ECNI) mode, monitoring bromine ions (m/z 79 and 81). The mass spectrometric properties of the fully brominated diphenyl ether, BDE-209, have been investigated. A high molecular weight fragment at m/z 486/488 enables the use of 13C-labeled BDE-209 as an internal surrogate standard.

The performance of different types of nebulisers: Vulkan direct injection nebuliser (Vulkan DIN), direct injection high efficiency nebuliser (DIHEN), microconcentric nebuliser coupled to cyclonic or double pass spray chamber (MCN-C or MCN-DP, respectively) was investigated and compared when used for sample introduction to ICP-MS or ICP-OES. With ICP-OES, in axial viewing mode, intensity distributions across the radius of the plasma (radial intensity profiles) were determined for different spectral lines with Esum 1.85-15.41 eV to determine fundamental plasma properties for various matrices using Vulkan DIN and MCN-C. The results showed that with the MCN-C the ionisation temperature (Tion) was about the same across the measured region of the plasma (±3.0 mm) whereas with the Vulkan DIN the Tion was significantly lower in the centre of the plasma. A large deviation from local thermodynamic equilibrium, as well as deteriorated stability, was observed for the plasma when using the Vulkan DIN. With ICP-MS noise power spectra (NPS) were generated to identify sources of noise. NPS showed that the magnitude of white noise for the tested sample introduction systems decreased in the following order: Vulkan DIN &gt; DIHEN &gt; MCN-C &gt; MCN-DP. This order follows the decrease of mean droplet size and span of the size distribution, indicating that the white noise is caused by spatial and temporal non-uniform desolvation and ionisation. Another source of noise arose from the peristaltic pump and the magnitude of pump interference noise decreased in the following order: DIHEN &gt; MCN-C/DP &gt; Vulkan DIN. Mains power interference noise and 1/f noise were lower for the direct injection nebulisers compared to the spray chamber systems. The contribution or effects of these noise components on relative standard deviations of steady-state ion-count rate and isotope ratio measurements is discussed in this thesis. Aerosols generated by the Vulkan DIN and the DIHEN were also directly characterised using Particle Dynamic Analysis. The Vulkan DIN produced particles with a mean diameter of ~30 µm and a size distribution between 2-80 µm. With the DIHEN the corresponding values were ~11 µm and 1-40 µm, respectively, with a few particles at 55-78 µm. The mean velocity of particles from the Vulkan DIN was ~10 m s-1 and from the DIHEN ~18 m s-1. The lower velocity allows longer residence time counteracting the effects of the larger droplet size.