The Effect of Glycol Side Chains on the Assembly and Microstructure of Conjugated Polymers

Conjugated polymers with glycol-based chains, are emerging as a material class with promising applications as organic mixed ionic-electronic conductors, particularly in bioelectronics and thermoelectrics. However, little is still known about their microstructure and the role of the side chains in determining intermolecular interactions and polymer packing. Here, we use the combination of electrospray deposition and scanning tunneling microscopy to determine the microstructure of prototypical glycolated conjugated polymers (pgBTTT and p(g2T-TT)) with submonomer resolution. Molecular dynamics simulations of the same surface-adsorbed polymers exhibit an excellent agreement with the experimental images, allowing us to extend the characterization of the polymers to the atomic scale. Our results prove that, similarly to their alkylated counterparts, glycolated polymers assemble through interdigitation of their side chains, although significant differences are found in their conformation and interaction patterns. A model is proposed that identifies the driving force for the polymer assembly in the tendency of the side chains to adopt the conformation of their free analogues, i.e., polyethylene and polyethylene glycol, for alkyl or ethylene glycol side chains, respectively. For both classes of polymers, it is also demonstrated that the backbone conformation is determined to a higher degree by the interaction between the side chains rather than by the backbone torsional potential energy. The generalization of these findings from two-dimensional (2D) monolayers to three-dimensional thin films is discussed, together with the opportunity to use this type of 2D study to gain so far inaccessible, subnm-scale information on the microstructure of conjugated polymers.


Scanning tunnelling microscopy
Measurements were performed on two ultrahigh vacuum (UHV) systems: a variable temperature (VT) and a low temperature (LT) scanning tunnelling microscope (STM), each equipped with an electrospray deposition (ESD) setup. The samples were dissolved in chlorobenzene at concentrations ≈ 0.025 g/L, and methanol was added to a 4:1 volume ratio.
The molecules were deposited from solution by ESD with a 4-stage Molecular Spray system.
The deposition current was monitored on the target substrate and typical deposition charges were 8 pAh for pBTTT and 6pAh for pgBTTT and p(g2T-TT). Au(111)/mica and Ag(111)/mica thin films (Georg Albert PVD, 300 nm thickness) were used as substrates and prepared in UHV by cycles of Argon sputtering (1 kV) and annealing to 500 °C. The measurements for pBTTT were performed on Au(111) and Ag(111) substrates, while all measurements for pgBTTT and p(g2T-TT) were carried out on Au(111). The substate did not affect the overall assembly of the material that formed similar structures on both surfaces. The VT-STM (SPECS Aarhus) was 3 cooled with a flow liquid nitrogen to a temperature of −145 °C, while the LT-STM (CreaTec Fischer & Co. GmbH) was kept at −196 °C by a liquid nitrogen bath cryostat. The first instrument was used to acquire the images for pBTTT, the second one to obtain the data for pgBTTT and p(g2T-TT). All images were acquired in constant current feedback mode with an etched tungsten tip that was treated by argon sputtering after insertion in UHV, to remove the oxide. The STM images were processed with WSxM 1 and Gwyddion, 2 the molecular models were made and optimised in Avogadro, 3 and their comparison with/superposition on the STM images was performed with the LMAPper software. 4

Molecular dynamics simulations
The used molecular dynamics (MD) force field is based on OPLS-AA, 5 with inter-thiophene dihedrals parameterised to fit with density functional theory (DFT) at the B3LYP/6311G** level of theory, 6 and can be seen in Fig. S1. . It should be noted that in the DFT simulation the side chains were approximated by a simple methyl group for pBTTT and by a hydroxy group for pgBTTT and p(g2T-TT). This explains why the potential energy surfaces for the T-T dihedrals are essentially identical for pBTTT and pgBTTT in Fig. S1. Glycol side chain parameters are from work by Woods et al. 7 and are based on OPLS-aa atom types, with other backbone bonded parameters from work by M. Moreno et al. and R. Bhatta et al. 8,9 The force field used is closely related to that presented in a recent MD simulation study of oxy-bithiophene oligomers by N. Siemons et al., 10 and it is therefore expected that the validation performed on that force field is also valid for the one used in this study. For the full force field parameterisation, please see the public GitHub Repository.
Simulations were performed using GROMACS2018.2. 11,12 Electrostatics and Van-der-Waals forces are computed using a cut-off scheme. Particle-Mesh Ewald Summation, whilst more accurate for bulk systems, scales poorly with vacuum space and hence is avoided here.
Hydrogen bonds are constrained using a LINCS algorithm, allowing for 2 fs time steps. Initial configurations were prepared using GROMACS 2018.2 and Packmol. 13 Trajectories were analysed using PyMOL. 14 Initial energy minimisation was performed on all simulations using the steep integration method and a time step of 1 fs, until the change in energy was less than numerical accuracy. The gold substrate was modelled using a 12-6 Van-der-Waals potential with parameters obtained from. 15 All simulations were performed with 20-mers on a Au substrate, with Au parameters from work by L. Su. 16 In all simulations, initial configurations were energy-minimised in vacuum and placed 3 Å above the surface of the substrate, measured from the substrate to the centre of mass of the oligomer. During the subsequent NVT simulations, the oligomers were deposited onto the substrate, being driven by their Van-der-Waals interaction, after allowing for some in-vacuum equilibration. For simulations of low molecular density areas, oligomers were simulated in NVT for 5 ns using a velocity-rescale algorithm for temperature coupling. 17 For simulations of high molecular density areas, after vacuum energy minimisation, 48 oligomers were placed 3 Å above a 60 nm × 60 nm surface. The deposition occurred during the 5 ns NVT simulations, which were followed by a subsequent 20 ns NPT simulation that was performed using semi-isotropic pressure coupling to apply pressure only parallel to the substrate. For NPT, pressure coupling is ensured using a Parrinello-Rahman barostat. 18,19 To ensure sufficient statistical sample, all simulations of single oligomers were performed over 36 replicas. Simulations of multiple polymers were performed over 4 replicas, each with 48 polymers. 5 Figure S3. STM image of pBTTT on Au(111) showing that, even in areas of lower local polymer coverage, pBTTT forms compact two-dimensional (2D) islands, leaving areas of bare exposed substrate. The wavy appearance of the gold surface is given by the Au(111) herringbone reconstruction. 6

Molecular models used for fitting the STM images
The molecular models used to fit the polymers in the STM images by means of the LMAPper software were created in the Avogadro molecular editor. Separated models were used for the side chains and for the backbone and their configuration was locally minimised with the MMFF94 force field of Avogadro to find optimal bond lengths and dihedral angles. In particular, the models of the alkyl side chains were based on isolated C14 chains in their fully extended linear configuration while different configurations of the triethylene glycol (TEG) sidechains were used with various degrees of chain bending (see Section 7 of the SI). and T-TT single C-C bonds, with respect to the all-trans lowest energy configuration) were also considered. All these individual models where then composed together in LMAPper in order to best fit the features appearing in the STM images.
The structure of a few of these models was also optimised in MD to check for a qualitative agreement between the force field used for the MD, and that used for the geometry optimisations. For the case of planarised units (such as those adsorbed on the gold surface), the geometry optimised structures for both the MD force field and the MMFF94 force field showed close qualitative agreement. It is important to note however, that the planarised structures modelled in this study, do not correspond to minimum-energy 'free structures' (i.e. not adsorbed onto the gold surface), as many dihedrals do not have their minima at 0 o . 7

Kinks in the polymer backbones
Based on the molecular models created in Avogadro, it was found that any significant kink in the polymer backbone must arise because of one or more T-T bonds in a cis configuration (see Fig. S4). This is due to the fact that the T-T bonds are oriented at an angle of about 25 with respect to the overall straight direction of the backbone (see Fig. S4b). On the other hand, the T-TT bonds are almost parallel to the backbone axis, so that their cis configuration does not produce as much of a distinct kink in the backbone (Fig. S4biii).

Statistics of cis and trans bonds in the polymer backbones
A statistic of the backbone conformation was collected from the experimental STM data following two different procedures, depending on the resolution of the image. When possible, the polymers were fully fitted with optimised molecular models for both the backbones and the side chains (see Section 4). This allowed to estimate the number of both T-T and T-TT trans and cis conformations. An example of this procedure is shown by the fitted images in Figs. 1c and 1d, Fig. 4 and Figs. 5e and 5f. On the other hand, when the resolution of the STM images was not high enough to allow a detailed and complete fit but sufficient that backbone kinks could be recognised, these latter were simply counted in the images, as they are equivalent to the number of cis T-T bonds. As can be seen from Fig. S4biii, the angle introduced by a cis T-TT bond is smaller and therefore harder to detect with certainty, so the count for the cis T-TT bonds was not included in this case. The combined results of the two methods are reported in Table S1. The total number of T-T and T-TT bonds was evaluated by direct counting in the first procedure, adopted for high-resolution images. For lower resolution images, the total length of each evaluated polymer was measured and divided by the length of the pBTTT repeat unit; this ratio is a good estimation of the total number of T-T and T-TT bonds. Note that no data are present in Table S1 for isolated pBTTT, as no such polymer has been observed. Table S1. Frequencies of cis and trans bonds in the backbones of surface adsorbed pBTTT-C14, pgBTTT and p(g2T-TT) polymers, based on the analysis of STM images. The total frequency of kinks is the sum of the frequencies of cis T-T and cis T-TT bonds.

total T-T cis T-T cis T-T (%)
total T-TT cis T-TT cis T-TT (%) The differences in kink frequencies between pBTTT and its glycolated derivatives and the role played by intermolecular interactions in determining this, are discussed in the main paper.
Here we only note that a cis T-T bond in p(g2T-TT) is accompanied by a significant steric clash between the TEG side chains of the corresponding thiophenes, while this is not the case for pgBTTT (see Fig. S5). This is because, being attached to the 3 and 3' positions, the TEG chains are closer together in p(g2T-TT) than in pgBTTT. As a consequence, when restricting the comparison to p(g2T-TT) and pgBTTT, one would expect a lower frequency of cis T-T bonds in the former polymer than in the latter, as is effectively observed in the experiments (see Table   S1). Cis T-TT bonds do not cause any steric clash (see Fig. S5), but DFT calculation predicted a slightly higher torsional potential energy for p(g2T-TT) than for pbTTT (see Fig. S1).

Inter-backbone distance distributions in high-density areas for pgBTTT and p(g2T-TT)
Inter-backbone distances have been evaluated in high-coverage areas for a series of STM images where locally collinear backbones were observed. Regions of the STM images where three or more polymer backbones were seen to be parallel to one another were chosen to collect a statistics of the distances between backbones, measured normally to the direction

Models of interacting glycolated side chains
The helical structure reported by French et al. 20  described procedure were marked by a lower degree of overlap and thus an expected lower interaction energy (one example is shown in Fig. S8g).
In a second step, selected pairs of TEG chains were used to describe the interaction between polymers by attaching them in their "carved out" conformation onto geometry optimised models of the backbones. Out of all the possible configurations that can be found, in Figs. S9 and S10 (for pgBTTT and p(g2T-TT), respectively) we report the cases that allow to obtain the minimum and the maximum inter-backbone distance. The models in the right column of both figures (e, g, i) are characterised by having the pairs of TEG chains almost perpendicular to the direction of the backbones, resulting in a larger separation between the backbones and a lower local side chain density. The opposite is true for the models in the left column of both

Fitting of molecular models on STM images for pgBTTT and p(g2T-TT)
In  14 and red in Fig. S12c). Similarly to the procedure followed in the case of pBTTT, these positions were used to guide the fitting of individual BTTT repeat units (i.e. the backbone's monomer units), taking care that their length and shape was consistent with the experimentally observed "wiggliness" of the backbone (see green lines in Figs. S11c and red ones in Fig. S12c).
Once the position of the monomers BTTT was determined within the backbones, a pair of side chains, chosen from those described above, was attached at each exit-point (green and red dots in Figs. S11c and S12c) and rotated until the best match was found with the features displayed in the STM images. Figs. S11d and S12d show the different types of side chains pairs that best matched the STM images, as shown in Figs. 6e and 6f of the main text (reported again as Figs. S11b and S12b, for clarity). In Figs. S11d and S12d, each pair of TEG chains is surrounded by a box, whose colour corresponds to those defined in Figs. S9 and S10. The good agreement of the fit with the STM data allows to satisfactorily identify all features in the images. This procedure can only be successfully attempted in ordered areas of the samples, where high-resolution imaging was possible.   Every simulated replica contains at least one backbone kink, with most containing more. The backbone kinks are clearly identified as being due to a cis T-T bond. Figure S54. A selection of typical configurations observed when pgBTTT is simulated in the low-density limit. Similar rates of backbone kinking are observed when compared to the case of pg(2T-TT); however, the angles formed by the backbone kinks are less pronounced. This is likely due to the reduced steric hindrance that occurs upon a cis T-T bond due to the side chain pairs being positioned further away from each other than in the case of pg(2T-TT).