摘要： Transmission Electron Microscopy (TEM) can be utilized to understand the morphology of organic bulk heterojunction (BHJ) solar cells and thus aid in improving device performance. We have previously shown that phase separation and formation of crystallinity is to be expected during co-evaporation of small molecule BHJ layers [1]. Using Electron Spectroscopic Imaging (ESI) [2] and electron diffraction, we found a significant influence of substrate and substrate temperature on the morphology of the photoactive layer during the fabrication of F4ZnPc:C60 BHJs. Whether or not the device is fabricated as inverted [3] or non-inverted cell influences crystal growth and, thus, phase separation during film formation. We have found that heating the substrate during BHJ film formation leads to an increase in efficiency for the inverted cell, whereas the non-inverted device shows no improvement. While the ESI measurements showed that substrate heating facilitates phase separation of the two materials, the difference in efficiency of the different device architectures could not be explained by this. Electron diffraction data indicated that crystallinity plays a role here. Since conventional ESI and electron diffraction only provide information about material distribution and crystallinity in a two-dimensional projection of the BHJ layers, high-resolution electron tomography was performed to gain insight into the three-dimensional structure. F4ZnPc:C60 was co-evaporated onto layers of neat F4ZnPc and C60, respectively. The measurements were performed under low-dose and LN2-cryo conditions in an FEI Titan Krios. This was necessary to preserve the sample, and foremost its crystallinity, since carbon-based materials, like C60, are prone to severe damage by electron irradiation. Figure 1 shows a bright-field TEM image of the BHJ on C60 (gold fiducials, seen in black, were used for tilt-series alignment). All images of the acquired tilt-series show crystalline areas such as the ones marked (A,B,C). The crystalline spacing seen here can be identified in the power spectra as characteristic for C60 (red: 0.85 nm, green: 0.5 nm and blue: 0.44 nm). As illustrated, such crystallinity can also be visualized in high-resolution electron tomograms, albeit only for smaller volumes at quite high magnification. To obtain a statistically significant distribution of crystallinity for different cell architecture and cell fabrication, larger volumes need to be analysed. For a given detector size, one needs to apply lower magnifications which results in lower resolution. However, the signature of pure crystals at these imaging conditions are a low variance in 3D, i.e. crystal distributions can easily be obtained from segmented 3D variance maps. A slice through the tomographic reconstruction of such samples can be seen in figure 2. Here, a BHJ film on C60 substrate is compared with a similar section through a tomogram of the BHJ on F4ZnPc. The gold fiducial indicates the top of the BHJ film. The homogeneous, aka crystalline areas are highlighted (red overlay). From the distribution of crystallinity we deduce, that large C60 crystals are found in both device architectures causing a very rough film surface. In the inverted device, these crystals can extend throughout the whole film, using the polycrystalline C60 substrate as seed for crystal growth, whereas the non-inverted BHJ showed C60 crystals starting somewhere in the middle of the film. Correlating this data with device performance, we find that C60 crystals which have grown throughout the BHJ layer are crucial for efficient devices.