Nanoscale 3D characterisation of soft organic material using conductive scanning probe tomography

The 3D nanostructure of soft and organic materials is a critical issue across multiple disciplines: from understanding biological processes and designing targeted drug delivery systems, physical chemistry and the design of novel electronic devices.^1–5 Across all these fields, scanning probe microscopy (SPM) is an important and widely used characterisation technique, offering high resolution, relatively simple sample preparation, and access to chemical, physical, and electrical properties through the wide range of modalities now available.⁶ However, a key limitation is the fact that it is essentially a surface-sensitive technique.

The aim of achieving 3D resolution using SPM is shared by multiple research groups and many approaches have been explored. Some have sought to exploit sub-surface sensitivity based on advanced modes of electrical, mechanical and optical SPM measurement, which is attractive as a non-destructive method.^7–11 However, the nature of the sensitivity achieved by these techniques is not clear and so quantitative interpretation is not yet possible, and it is uncertain what kind of depth penetration is achievable. Other methods for 3D SPM imaging have relied on ultramicrotomy or ion beam milling for layer-by-layer reconstruction or to extract cross-sections of a sample.^12–15 The results of these techniques are more readily understood, but their reliance on separate techniques for measurement and milling makes them generally inaccessible. Here we consider an approach for scanning probe tomography that uses widely available commercial equipment and a single probe tip to both characterise a surface and remove layers of material. By sequentially scanning the surface and removing layers, it is possible to construct a 3D map of the sample with nanoscale resolution.^16,17

For inorganic samples, a doped diamond tip can be used to abrade the sample whilst also enabling electrically conductive measurements, an approach often referred to as Scalpel SPM.^16,18–20 However, in applying this technique to soft organic materials, a diamond tip is no longer required, nor desirable, since it does not offer the highest spatial resolution. The resolution is fundamentally limited by the tip radius of the probe, where a typical diamond tip has radius ∼100 nm. On the other hand, commercially available Pt-Ir coated tips have radius ∼20 nm. In addition, the conductivity of diamond tips is relatively poor compared to metal coated probes. Since metal coated probes tend to wear more quickly than diamond, in this work significant care was taken to develop a procedure for tomographic characterization without degradation of the probe. Here we have developed a ‘slice and view’ technique using a Pt-Ir tip for use on soft organic materials, which we demonstrate by characterising a nanostructured organic semiconductor blend film in 3D.

The demonstration sample used is a blend of two organic semiconductors, which is well-known in the field of organic electronics: a 1:1 (by weight) blend of nanowire poly(3-hexylthiophene) (nwP3HT) with [6,6]-phenyl-C₆₁-butyric acid methyl ester (PCBM), prepared in dichloromethane at a total concentration of 4 mg/mL, using previously published conditions.²¹ 

The optimum morphology for an organic photovoltaic blend film for efficient charge generation is generally agreed to comprise both intimately mixed regions and purer phase-separated domains.^22,23 Effective extraction of charges additionally requires interconnectivity of phase-separated domains so that photogenerated charges have continuous transport routes out of the device. Due to the tendency of highly planar conjugated molecules to form semi-crystalline π-π stacked domains, controlled self-assembly of nanowire (also called nanofiber or fibril) networks is a promising motif for an optimised morphology.^21,24–26 In this Letter, the use of a nanowire morphology for the P3HT component ensures that there is a clearly defined nanostructure within the sample, in contrast to a homogeneous film. This sample has specific relevance to organic photovoltaic applications, where the development of low-cost, flexible solar cells requires detailed understanding of the nanoscale morphology of blended electron donating and electron accepting organic semiconductors, which is critical to the device performance.^27,28 However, the technique developed in this work is intended for broader application to organic samples, where the 3D nanostructure is an important property and the softness of the material is a key consideration.

The morphology of the sample, with its polymer nanowire network can be seen in Figure 1(a) showing a high-angle annular dark-field transmission electron microscope image (HAADF TEM). The nanowires show a broad distribution of dimensions, but the widths are roughly tens of nanometres with lengths of hundreds of nanometres. The chemical composition of the nanowires in the TEM image can be established using energy dispersive spectroscopy (EDS) to map the distribution of sulfur atoms, since these are present in P3HT but not in PCBM. Figure 1(b) shows the distribution of sulfur in green with the carbon plotted in red, clearly indicating that the network of nanowires are composed of P3HT. Importantly, we note that there are also non-nanowire regions with a high sulfur content (appearing yellow in the figure) likely due to indicating the presence of disordered P3HT in mixed domains, which is important for efficient charge generation.^19,20 However, we can’t fully exclude that a small part of the high sulfur signal may come from a projection effect due to the TEM sample thickness, in case there are nanowires perpendicular to the sample surface.

The chemical sensitivity and high resolution of TEM are powerful for morphological analysis, and tomographic methods have been used for impressive 3D characterisation of organic films, however, it is not possible to directly relate this to the electronic performance.^26,29 In contrast, electrical modes of atomic force microscopy have been demonstrated for simultaneous topographical and electrical characterisation, but with limited sub-surface sensitivity.^7,15,30–33

Figure 2 demonstrates the simple case of conductive atomic force microscopy (C-AFM) on the pristine surface of the nwP3HT:PCBM sample. The device architecture is the same as described by Kim et al., with a hole transporting polymer beneath the active layer on an indium tin oxide substrate, allowing efficient hole injection/extraction at the bottom interface.¹⁸ All the measurements were performed with a Bruker Dimension ICON equipped with linear current amplifier (noise level of ∼0.1 pA), in ambient conditions. Commercial Pt/Ir coated probes (Bruker SCM-PIC) with spring constant 0.2 N/m were used and the bias was applied to the conductive substrate with the tip grounded. The nanowires are not easily identified from the surface topography in Figure 2(a), but by measuring the tip current under bias, the nanowires become clearly visible. With a tip-sample bias of ∼50 mV (Sample biased negative compared to the tip) the nanowire domains conduct a hole current, whereas the surrounding regions do not. Figure 2(b) uses a -2 pA threshold to highlight the nanowires with high contrast. As shown in Figure 2c, current-voltage characteristics measured for points on the nanowire and on the surrounding blend region show a clear difference in the magnitude of the current, particularly under negative sample bias (i.e. hole current from tip into sample). Under positive sample bias, no barrier for charge conduction are present for nwP3HT and the blend regions.¹⁵ The precise composition of the blend region is not clear, but the low current measured is consistent with a high charge injection barrier for the PCBM rich region.^15,33 The low bias needed to inject current into the nwP3HT indicates that the contact between the SPM tip and the P3HT is approximately ohmic and the nanowires exhibit near metallic conduction, which suggests that the polymer is oxygen doped due to atmospheric exposure.³³ 

It is important to recognise that the distribution of current measured in C-AFM relates to the interconnectivity of the nanowire network through the whole thickness of the film, making the interpretation of this data non-trivial.^7,34 Scanning probe tomography provides a means to directly map the electrical connectivity of this network through the 3D structure of the film. Measurements were carried out in contact mode using an alternating sequence of ‘slice’ and ‘view’ scans: in a ‘slice’ scan, a high loading force of 5 nN – 10 nN was used to remove material, whereas in a ‘view’ scan, a lower force of around 500 pN was used for surface characterisation.

The depth resolution of scanning probe tomography is determined by the amount of material removed during the ‘slice’ scan. This can be evaluated by considering the surface height of the ‘sliced’ area relative to the surrounding region after the removal of multiple layers. Figure S1(b) shows a ‘view’ scan of the nwP3HT:PCBM surface with a 1.25 μm × 1 μm central region where 20 successive ‘slice’ scans have been performed. There is clearly some variation in the removal rate for different areas, indicating local variation in material properties as might be expected for a blend film, but an average removed depth of 6.9 nm ± 2.9 nm can be considered, corresponding to a removal rate of 0.35 nm ± 0.15 nm per ‘slice’ scan.

The geometry of the tip-sample interaction is difficult to elucidate with clarity and will have a strong effect on the material removal process. The depth to which the tip penetrates the surface of the sample is a key parameter, which we can evaluate approximately by considering how the loading force affects the magnitude of current measured. Figure 3 illustrates the proposed method. In Figure 3(a) a ‘view’ scan is performed with low loading force revealing faint indications of a sub-surface nanowire network with a peak current of ∼ 0.1 pA. Note that measurements are performed in contact mode, so the term “peak current” is used in this manuscript to denote the maximum current detected in contact mode C-AFM. In Figure 3(b) the corresponding ‘slice’ scan is carried out with a relatively high load force, and higher currents are recorded, with a peak of ∼ 0.4 pA indicating that the tip has penetrated through a capping layer and made contact with the nanowires beneath. After this a sequence of alternating ‘slice’ and ‘view’ scans is carried out where the ‘view’ image shows the nanowires with increasing clarity as the capping layer is gradually removed. In this case, we find that after 4 ‘slice’ scans the peak current in the ‘view’ scan matches that of the original ‘slice’ scan, as shown in Figure 3(c), hence we deduce that the penetration depth of the first ‘slice’ scan is equal to the depth of material that has now been removed. This method was repeated 3 more times in different regions to estimate that the penetration depth using 5 nN loading force is in the range 1.4 nm to 1.6 nm. Significantly, this method avoids the issue of lateral tip displacement, which can be problematic for techniques based on the force curve measurements which suffer from the sliding of the tip.^35–37 By appropriate choice of measurement parameters and threshold image analysis (see ESI for more information) each image in our technique is a 2D representation of the surface, making extrapolation of data in the 3^rd dimension relatively easier.

The loading force used here is lower than that reported elsewhere due to the softness of the organic sample, and the need for controlled removal of material.^18,38 We find that higher loading forces tend to result in very non-uniform removal of material and contamination of the tip, similarly to what was observed when using diamond probes on hard surfaces.²⁰ Our results indicate that a scratch test at the nanoscale follows a very similar process to that observed in mesoscale and macroscale scratching of a viscoelastic polymer film.³⁹ At low force, no material removal is observed. After a certain threshold force, which will depend on the materials under study and the geometry of the tip-sample interaction, material removal takes place. If too much force is used, or if the scratch process is too fast, tearing of the film occurs, which can lead to uncontrolled material removal and unwanted damage to the sample. The process in hard surfaces is not completely different, except from the viscoelastic behavior for polymer systems. In both cases one would expect the tip to penetrate into the film to a greater depth than the depth of the resulting scratch. This can be largely assigned to the relatively lower lateral force, as compared to the vertical force on the surface, during the slicing process. However, in a viscoelastic polymer film the apparent vertical indentation can be even higher due to compression of the film.^40,41 Indeed, our results show that the tip penetration depth is 3 – 4 times greater than the thickness of material removed in a ‘slice’ scan, which could potentially lead to some smearing of the features and loss of resolution in the exposed layer. In practice we see little evidence of this, with clear continuity of features and dimensions through the sequence of layers, which we attribute to the correct choice of measurement force and speed parameters to ensure recovery of the viscoelastic film.

A 3D interpolation of the conductive scanning probe tomography results is given in Figure 4. The image on the left represents a slab of the sample with dimensions 100 nm × 180 nm × 5 nm, where the conductive nwP3HT is plotted in blue, and the insulating matrix is plotted in translucent yellow, such that buried nanowires appear green. From left to right the images represent sequential removal of surface layers. This sample volume shows two nanowire domains, one at the surface and one buried. The top surface nanowire (aligned diagonally from the bottom left corner) disappears gradually as material is removed such that it has completely disappeared in the final image. Simultaneously, the second feature (towards the top of the image) is not visible at the surface in the first image and is gradually exposed as material is removed, becoming clearly visible by the time 5 nm of material has been removed. From these observations we deduce that the first nanowire was on average 5 nm thick, which agrees with estimates obtained from X-ray studies and topographical SPM.^42–44 We also confirm that the low conductivity of the blended phase is sufficient to block conduction between the tip and a nanowire buried 5 nm beneath the surface and so preserve contrast between surface and sub-surface features, in this case.

Since the first reports of polymer nanowire-based devices, it has been generally assumed that the high charge carrier mobilities arise from efficient charge transport along the nanowires and that the nanowires are electrically connected.²¹ However, this model has not been verified through direct experimental evidence. Specifically, there is evidence of sub-surface sensitivity in some cases suggesting that there is significant conductivity through the non-nanowire matrix.⁷ In our case, we find a strong electrical contrast between the surface and sub-surface nanowires, demonstrating that the conduction in this sample is confined to the nanowire network itself. Additionally, it has been suggested that electrical conduction between nanowires is poor resulting in higher than expected bimolecular recombination.⁴⁵ Here we are able to probe the electrical properties of the inter-nanowire charge conduction directly, and in this case we find that the current measured on the two nanowires shown in this sample are comparable, indicating that these structures are electrically well-connected. These specific conclusions are limited to the sample system used in this Letter, but the methods shown will be more widely applicable to nanoscale studies of the relationship between morphology and electrical performance.

In summary, this Letter has demonstrated that conductive scanning probe tomography can be used to directly map the 3D nanostructured charge transport network in an organic blend film optimised for photovoltaic applications. This method is also more generally applicable to the whole class of nanostructured soft organic samples, since in every case the functional properties are strongly dependent on the 3D nanostructure.

See supplementary material for discussion on the material removal using slice and view technique and discussion on the reconstruction of 3D data from 2D C-AFM current maps.

This work was funded by the UK Department for Business, Energy and Industrial Strategy (BEIS) through the National Measurement System, the European metrology research programme (EMRP) project NEW01-TReND and the EMPIR project 16ENG03-HyMET. The EMPIR initiative is co-funded by the European Union’s Horizon 2020 research and innovation programme and the EMPIR Participating States. We thank Andrew Pollard (NPL) for useful discussions.