Comparing Raman quantifications in water-ethanol-propanol mixtures – How much is it?

Harald Matthias Fitzek, Lukas Hageneder, Alexander Reiger, Dorian Brandmüller, Florian Martin Trummer, Elias Michael Henögl, Alexander Pranter, Daniel Kollreider, Rupert Bachler, Marcel Simhofer

Research output: Contribution to conferencePoster

Abstract

In 2021, students had the following task at our advanced lab exercise for Raman spectroscopy:“In our current laboratory exercise research project, we will build a robust quantitative model of some chemical mixture. To do this both a calibration and a test data set will be provided. On these data sets several different approaches for building quantitative Raman models are tested and finally the accuracy and detection limits of the best model will be determined. Our goal is to demonstrate that even in the limited time available during a lab exercise it is possible to build a quantitative Raman model with sufficient confidence for real life applications.”Over the course of this lab exercise, which was highly focused on data treatment, students tested a variety of quantitative approaches on the example of water-ethanol-propanol mixtures. These include simple univariate approaches such as band height, band area or simple band fits, as well as multivariate approaches ranging from common least square fit (simplest) to neural networks (most complex). On top of this, one group designed a specialized quantification approach based on the particular physics of molecular interaction in thewater-ethanol-propanol mixture. All approaches were tested on the same data set (both training and testdataset), which is shown in the figure below, where (a) is the measurement setup, (b) the pure componentspectra (c) a heat map of the spectra of the calibration data set and (d) shows all the concentrations bothin the calibration and test datasets. The combined work of these student groups can be used to comparethe accuracy of each method, since all approaches were tested on the same measurement data set. The results showed that most methods have a similar performance, with common-least-square-fit achieving the best results and interestingly, simple band height slightly outperforming neural networks.
Original languageEnglish
Pages47
Publication statusPublished - Apr 2023
EventASEM Workshop 2023: Workshop on Advanced Electron Microscopy - University of Vienna, Wien, Austria
Duration: 13 Apr 202314 Apr 2023

Conference

ConferenceASEM Workshop 2023
Country/TerritoryAustria
CityWien
Period13/04/2314/04/23

ASJC Scopus subject areas

  • General Materials Science

Fields of Expertise

  • Advanced Materials Science

Treatment code (Nähere Zuordnung)

  • Basic - Fundamental (Grundlagenforschung)

Fingerprint

Dive into the research topics of 'Comparing Raman quantifications in water-ethanol-propanol mixtures – How much is it?'. Together they form a unique fingerprint.

Cite this