Filtern
Dokumenttyp
- Vortrag (3)
- Dissertation (1)
Referierte Publikation
- nein (4)
Schlagworte
- Multivariate Datenanalyse (4) (entfernen)
Organisationseinheit der BAM
Eingeladener Vortrag
- nein (3)
The increasing enrichment of water bodies and soils with plastic waste leads to the accumulation of microscopic plastic particles, so-called microplastics (MP). There is an urgent need for analytical methods that help to identify and quantify MP. At present, mainly thermo-analytical and microscopic methods such as micro-infrared spectroscopy or micro-Raman are used for this purpose. The latter are usually tied to time-consuming sample enrichment and preparation, only small sample quantities (micrograms) can be examined and the evaluation of the obtained spectra can be demanding. In the context of this presentation, two approaches are presented which, via the multi-variate analysis of spectroscopic data, allow i) a new methodological approach to screening MP in contaminated soils and ii) an alternative evaluation of large (micro)-spectroscopic data sets.
First, a NIR spectroscopic method is presented which allows MP consisting of polyethylene, polyethylene terephthalate, polypropylene and polystyrene to be detected in the range of up to 0.5 mass percent. Due to short measurement times and robust technology, this approach has the potential, in contrast to thermo-analytical and micro-spectroscopic methods, to examine larger sample quantities with minimal pre-treatment.
The second approach deals with the evaluation of large data sets, as typically obtained as a result of micro-FTIR using modern FPA detectors. The micro-FTIR technique is based on the spectral recording, imaging and subsequent identification of vibration bands typical of synthetic polymers. The image data sets are large and contain spectra of numerous particles of natural and synthetic origin. Exploratory multivariate data analysis has been tested to complement existing approaches based on e.g. spectrum library searches. The core concept used was dimensionality reduction. The results not only represent an orthogonal method for checking the results obtained by an automated library search, but also revealed a group of spectra that were not recorded in the existing spectrum libraries.
Der Beitrag beinhaltet die Vorstellung eines gemeinsamen Projekts zwischen der BAM und der TU Berlin (Process Dynamics and Operations Group). Mittels Raman Spektroskopie werden Hydroformylierungsreaktionen in einer Miniplant der TU prozessbegleitend untersucht. Kalibrationsstrategien, Laboruntersuchungen sowie Vorhersagen von Prozessmessungen basierend auf chemometrischen Modellen werden vorgestellt. Zum gegenwärtigen Zeitpunkt wird die Strategie als erfolgsversprechend bewertet, jedoch sind weitere Untersuchungen insbesondere zum Einfluss der Microemulsion auf die Ramanspektren notwendig.
Entwicklung neuer Sonden für bioanalytische Anwendungen der oberflächenverstärkten Raman-Streuung
(2011)
Surface-enhanced Raman scattering (SERS) has been established as a versatile tool for probing and labeling in analytical applications, based on the vibrational spectra of samples as well as label molecules in the proximity of noble metal nanostructures. The aim of this work was the construction of novel SERS hybrid probes. The hybrid probes consisted of Au and Ag nanoparticles and reporter molecules, as well as a targeting unit. The concept for the SERS hybrid probe design was followed by experiments comprising characterization techniques such as UV/Vis- spectroscopy (UV/Vis), Transmission electron microscopy (TEM) and Dynamic Light Scattering (DLS), respectively. SERS experiments were per- formed for studying and optimizing the plasmonic properties of nanoparticles with respect to their enhancement capabilities. The SERS-probes had to meet following requirements: biocompatibility, stability in physiological media, and enhancement of Raman-signals from Raman reporter molecules enabling the identification of different probes even in a complex biological environment. Au and Ag nanoaggregates were found to be the most appropriate SERS substrates for the hybrid probe design. The utilization of Raman reporters enabled the identification of different SERS probes in multiplexing experiments. In particular, the multiplexing capability of ten various reporter molecules para-aminobenzenethiol, 2-naphthalenethiol, crystal violet, rhodamine (B) isothiocyanate, fiuorescein isothiocyanate, 5,5'dithiobis(2-nitrobenzoic acid), para- mercaptobenzoic acid, acridine orange, safranine O und nile blue was studied using NIR-SERS excitation. As demonstrated by the results the reporters could be identified through their specific Raman signature even in the case of high structural similarity. Chemical separation analysis of the reporter signatures was performed in a trivariate approach, enabling the discrimination through an automated calculation of specific band ratios. The trivariate identification could be a promising method for SERS-multiplexing in analytical applications. Multivariate methods such as Principal Components Analysis (PCA) and Hierarchical Cluster Analysis (HCA) were as well applied for discrimination and imaging of the reporter signatures. With the help of multivariate imaging methods based on cluster analysis, it could be for the first time demonstrated that such methods provided a fast identification of various SERS hybrid probes inside the biological matrix, this was demonstrated using living 3T3 cells. Further, in a duplex imaging approach, the probes fulfill the requirements for the sensitive detection of both the specific reporter signatures and intrinsic information coming from eukaryotic cells. The results of cluster methods and principal components approaches for discrimination indicate that fast, multivariate evaluation of whole sets of multiple probes is feasible, beyond the visual inspection of individual spectra that has been practiced so far. This suggests multiplexing applications with SERS hybrid nanoprobes and SERS tags in very high density sensing and biological imaging applications, where fast read-out is required. The pH-sensitivity of SERS-Tags that consisted of different reporter molecules attached to aggregated Au and Ag nanoparticles in the range between pH∼3-10 was studied. It could be demonstrated that the reporter molecules provided pH-dependent SERS signatures and could therefore be suitable for the sensitive pH-detection, e.g., inside cellular compartments. The construction of targeted SERS probes was based on the integration of a goat-anti- mouse antibody as targeting element Antibodies were coupled to Au and Ag nanoprobes surrounded by a Bovine Serum Albumin (BSA) coating which served as a carrier for the covalent linkage of a Raman reporter molecule and the targeting units at the same time In experiments with BSA and a conjugated reporter the spectra of the BSA-coupled re- porters provided an indication of the secondary structure of BSA which is related to the BSA-reporter coupling procedure In in v i t ro -experiments with BSA-coupled nanoprobes inside 3T3 cells reporter signatures and intrinsic information from the cellular matrix could be delivered BSA enabled the coupling of reporter molecules as well as the targeting of antibodies and served as stabilizer of the gold nanoaggregates The functionality of the coupled antibodies after their integration into the SERS probe was retained This was verified by the results of a direct Enzyme-Linked Immunosorbent Assay (ELISA) Conjugates with implemented reporter molecules could be characterized using SERS The application of the complete probes suggested a use of these novel biocompatible stable and targeted SERS probes that can be excited out-of- resonance also for other bioanalytical applications. On the basis of the constructed SERS hybrid probes comprising a large number of BSA- coupled reporters could e g be implemented for automated high-througput immuno- assays where they are arranged on a microstructured device for the simultaneous and multilevel SERS-readout in one step .
Die Wettbewerbsfähigkeit der Prozessindustrie basiert auf der Sicherung der geforderten Produktqualität bei einer optimalen Nutzung von Anlagen, Rohstoffen und Energie. Ein Weg zur wissensbasierten Produktion führt über die Betrachtung der wesentlichen Apparate-, Prozess- und Freigabedaten aus Betrieben und Labors. Das Potenzial dieser Daten wird heute vielfach noch nicht konsequent für ein umfassendes Verständnis der Produktion genutzt.
Neben Fragen zur Datenerfassung, Datenkonnektivität und Datenintegrität müssen solche Daten für eine ganzheitliche Prozessanalyse zunächst mit Kontextinformationen zusammengebracht werden. Datenquellen enthalten Zeitwertpaare, aber auch diskrete Daten aus LIMS (Laboratory Information Management Systems) oder ELN (Electronic Laboratory Notebooks) und werden zunehmend durch 2D- und 3D-Daten aus der Chromatographie-Massenspektrometrie oder bildbasierter Analytik ergänzt.
Für die automatisierte Merkmalsextraktion, etwa zur Extraktion chemischer Informationen aus den oben genannten Datenquellen werden multivariate Werkzeuge und Algorithmen genutzt. Multivariate Statistiken wie PCA (Principle Component Analysis), PLS (Partial Least Squares) und LDA (Latent Discriminant Analysis) bilden die erste Grundlage für die Datenanalyse. Für diese Verfahren sind heute Datenvorbehandlungsschritte nötig. Die Modellbildung geschieht manuell und ist sehr aufwendig.
Können diese Daten im Zeitalter von ML (Machine Learning) und KI (Artificial Intelligence) anderweitig sinnvoll genutzt werden und ohne klassische Modellbildung? Die Bezeichnung „Big Data“ als Voraussetzung für datengetriebene Auswerteverfahren ist für die Prozessindustrie allerdings unpassend, denn auch bei mengenmäßig großen Datensätzen liegen für Kampagnen typischerweise nur Informationen über einige Batches mit einer Serie von Messdaten vor, die genügend Varianz für eine datengetriebene Auswertung aufweisen – nicht vergleichbar mit den Datenmengen im WWW oder von großen Internet-Konzernen.