DeepMixing – Quantification of the mixing and interfacial heterocharacteristics of nanoparticle aggregates forming in an aerosol mixing zone

Abbildung einer dreidimensionalen Rekonstruktion zweier Heteroaggregate mit zugehörigem Größenverhältnis von 25 Nanometern.
The figure shows the three-dimensional reconstruction of two heteroaggregates of tungsten oxide and titanium oxide particles. The geometric reconstruction was carried out using STEM tomography and the identification of the individual particles was performed using artificial intelligence trained by machine learning.

Project Leader :

Prof. Dr.-Ing. habil. Lutz Mädler
University of Bremen

Prof. Dr. Andreas Rosenauer
University of Bremen

Dr. Nicolae Bârsan
University of Tübingen

The properties of nanomaterials depend heavily on their structure and chemical composition. One example is aggregates of particles with diameters in the range of a few to a few tens of nanometres. Their properties change when particles of two different materials are mixed together (heteroaggregate). Due to the different properties, it is important to know for each sample whether particles of different materials are completely mixed or whether particles of the same material are bound together in clusters.

In this project, we will investigate films of nanoparticle heteroaggregates for use as gas sensors, which are important tools for analysing air quality or detecting leaks in pipelines. The films are produced by means of so-called double-flame spray pyrolysis. Starting materials of the two materials are sprayed separately into two flames, where they first form nanoparticles, then clusters and, after overlapping of the two flames, heteroaggregates. At the start of the project, we will investigate mixtures of tin oxide and cobalt oxide with regard to their suitability for the detection of gases such as carbon monoxide, hydrogen, nitrogen dioxide and acetone. This combination of materials has already shown promising results, but a systematic investigation is still pending. To close this gap, we will systematically investigate sensors in which we vary the mixture of nanoparticles by varying parameters of the sample synthesis.

The expected differences in mixing must be analysed very precisely. In the previous funding period (DeepMixing I), we developed methods for characterising the mixture using scanning transmission electron microscopy (STEM) based on machine learning. So far, these methods have only been applied to material combinations with high material contrast in the STEM images. Zinc-cobalt oxide heteroaggregates require new hybrid methods, as zinc oxide and cobalt oxide are hardly distinguishable in STEM images. We will develop these methods to measure the mixture in previously unanalysable heteroaggregates. To this end, we will integrate X-ray spectroscopy and tomographic reconstruction into the measurement and analysis.

The combination of material synthesis, sample characterisation and functional application within one project provides a great opportunity to develop a theoretical model of the sensing behaviour of nanoparticle heteroaggregate films. In the semiconductor industry, many devices utilise the fact that two layers of material with different doping form an interface called the p-n junction. The p-n junction is very well understood and described theoretically. However, it has not yet been proven whether p-n junctions between nanoparticles behave in the same way. Our aim is to clarify this open question and thus make a major contribution to this field of research.