The record for the conversion efficiencies of singe junction organic photovoltaic cells has recently leapt from about 14 % [1] to 16 % [2] within one year. This huge increase is not only owed to optimizing the acceptor molecules in terms of electronic structures but also an improved blend morphology.  Its optimization is typically a result of educated guesses combined with trial and error instead of direct investigation of the cell’s morphology. The reason is that donor polymers and acceptor molecules show similar chemical composition and properties. Thus, standard microscopy methods fail to generate contrast between the different materials [3]. Most studies rely on atomic force microscopy, deducing the blend morphology from the surface profile of the sample.   

We introduce a novel technique to visualize the morphology of fullerene (FA) and non-fullerene acceptor (NFA) bulk heterojunctions at the nanoscale. The method is based on spatially resolved secondary (SE) and backscattered electron (BSE) spectra, which directly relate to material properties [4]. Unsupervised machine learning reveals similarities within the datasets and allows to assign them to the different phases. Exemplary results for FA and NFA blends based on SE and BSE spectra are displayed in fig. 1. An interfacial mixed phase can be identified in both cases separating donor and acceptor domains.

The spectra are acquired with the aberration corrected prototype ultra-low voltage scanning electron microscope (ULVSEM) Zeiss DELTA® [5]. This microscope can retain high resolution down to beam electron energies ≥ 20 eV. This is radically lower compared to energies of ≥1 keV and 30 – 300 keV in conventional SEM and transmission electron microscopy, respectively. Several benefits for the investigation of organic electronics result from ultra-low landing energies: (i) New contrast mechanisms arise; (ii) the interaction volume of electron probe and sample is drastically reduced. Thus, signal mixing across interfaces is diminished and the signal originates directly from the surface. (iii) We observe decrease of beam damage, reducing the alteration of the sample during investigation. We expect that the opportunity to gain insight into local electronic properties from nano-resolved electron spectra will lead to a full understanding of interfacial effects and charge separation processes in the near future.