Hyperspectral imaging

Both optical images and the related chemical images were combined and a chemometric investigation of the combined images was done using the software package Imagelab (Epina Software Labs). Principal component and cluster analysis could be performed for the Raman and the EDX images based on selected spectral descriptors. Chemical maps have been designed based on the same descriptor sets. EDX descriptors were based on single elements, Raman descriptors on CH-organic, soot, sulfates, carbonates, …

Site, samples & methods

Released particles from three different salt lakes (Fig. 1) were collected on alumina foils inside the Teflon chamber using a Sioutas impactor. While the ultra-fine fraction of the released particles is missing, the surrounding coarse mode could be sampled. Vibrational mapping of an area of 100 x 100 µm at a lateral resolution of 1 µm was performed using a Horiba LabRam 800HR Raman microscopy. The same area was analyzed using a Quanta FEI 200 electron microscope. Besides the high-resolution image, the elemental composition could be investigated by energy-dispersive X-ray spectroscopy.

Introduction

Western Australia was originally covered by natural eucalyptus forests, but land-use has changed considerably after large scale deforestation. Thus the ground-water level rose and brought dissolved salts and minerals to the surface. Nowadays a great plenty of salt lakes with pH levels reaching from 2.5 to 7.1 cover the land, which is in use for wheat farming and livestock. Missing rain periods cause dryness, which reason can be seen in the formation of ultrafine particles from salt lakes, which increase the cloud condensation nuclei and prevent therefore rainfall (Junkermann, ACP, 2009). Several field campaigns have been conducted between 2006 and 2013 where new particle formation has been observed and correlated to the salt lakes. To identify the formed particles directly with the chemistry of and above the salt lakes, a 1.5 m³ Teflon® chamber was brought above the lake in 2012 and 2013 (Fig. 2).

Hyperspectral Raman and SEM-EDX imaging of size-segregated aerosols, released above Australian salt lakes JOHANNES OFNER1, KATHARINA A. KAMILLI2, ELISABETH EITENBERGER1, JOHANN LOHNINGER1, GERNOT FRIEDBACHER1, BERNHARD LENDL1 AND ANDREAS HELD2

1 INSTITUTE OF CHEMICAL TECHNOLOGIES AND ANALYTICS, VIENNA UNIVERSITY OF TECHNOLOGIES, AUSTRIA 2 JUNIOR PROFESSORSHIP IN ATMOSPHERIC CHEMISTRY, UNIVERSITY OF BAYREUTH, GERMANY

Results

All coarse-mode fractions are dominated by NaCl and silicates. Lake Dune exhibits CaSO4 as a main additional fraction, Lake Orr CaCO3. Both Ca salts contribute to the coarse-mode of Lake Shot. Most of the particles are covered by an organic layer, thus indicating that nucleating organic material contributes to the composition of the coarse mode. The Raman signal of soot might derive from burned organic, which happened during the Raman mapping by the laser.

Fig. 2 – Mobile Teflon® chamber above the surface of Lake Orr in Western Australia.

Fig. 1 (above) –Satellite image (Google Earth) of the area of Lake King and the surrounding salt lakes, indicating the three sampling sites .

Data processing of hyperspectral datasets

Combined Raman and SEM-EDX imaging allows to investigate all aerosol components of interest (Fig. 3). The use of chemometric methods like principal components analysis (Fig. 4) and cluster analysis (Fig. 5) unravels lateral distributions and links between single components or elements. Inorganic particles like NaCl and CaSO4 can easily be identified. Further, organic particles can be distinguished from organic coatings of inorganic particles. Chemical maps, designed based on the chemometric approach described above, gain access to single particle identification.

Fig. 3 – combined total intensities of Raman (red) and EDX (yellow) – both methods cover nearly the whole dataset – example of Lake Dune.

Fig. 4 – Principal component analysis of the Raman dataset of Fig. 3 – indicating 3 main species in the PC2/PC3 plot.

organics

soot

inorganics

The authors like to thank the German research foundation within RU 763 HALOPROC for funding of the field campaigns as well as the Epina Software Labs for ImageLab.

Conclusions

The combination of the hyperspectral imaging methods of Raman spectroscopy and SEM-EDX allows a detailed morphological study of the composition of aerosol coarse-mode particles and assist the interpretation of the interaction between nucleation- and coarse-mode. Both methods chemically characterize nearly 100% of the segregated particles. A future multi-sensor hyperspectral chemometric approach will assist a combined analysis treatment of the dataset for a better understanding of aerosol formation and atmospheric heterogeneous reactions on particle surfaces.

Fig. 5 – Cluster analysis of the EDX dataset. Colored sub-clusters indicate NaCl, sulfates and carbonates.

Fig. 6 – Raman chemical images of organics (CH) in yellow, soot (maybe CH, burned by Raman laser)) in red, sulfates in green and carbonates in orange of aerosol particles from the three different salt lakes (Lake Dune (left image), Lake Orr (center image) and Lake Shot (right image)).

Fig. 7 – SEM-EDX chemical images of NaCl in blue, CaSO4 in green, CaCO3 in orange and silicates in purple of aerosol particles from the three different salt lakes (Lake Dune (left image), Lake Orr (center image) and Lake Shot (right image)).