Measuring the mechanical properties of molecular conformers

Scanning probe-actuated single molecule manipulation has proven to be an exceptionally powerful tool for the systematic atomic-scale interrogation of molecular adsorbates. To date, however, the extent to which molecular conformation affects the force required to push or pull a single molecule has not been explored. Here we probe the mechanochemical response of two tetra(4-bromophenyl)porphyrin conformers using non-contact atomic force microscopy where we find a large difference between the lateral forces required for manipulation. Remarkably, despite sharing very similar adsorption characteristics, variations in the potential energy surface are capable of prohibiting probe-induced positioning of one conformer, while simultaneously permitting manipulation of the alternative conformational form. Our results are interpreted in the context of dispersion-corrected density functional theory calculations which reveal significant differences in the diffusion barriers for each conformer. These results demonstrate that conformational variation significantly modifies the mechanical response of even simple porpyhrins, potentially affecting many other flexible molecules.

Type I conformer translated in the <211>-direction. In this case no significant difference in the energy barrier was found between these two different approaches to describe dispersion, see Figure S6.

Phase errors during 2D ∆f measurement
Maintaining optimal PLL feedback is essential during collection of the lateral ∆f (x, z) data. For normal measurements of ∆f (z), variations in ∆f are gradual, and generally follow a smooth Lennard-Jones like profile which can easily be tracked by the PLL. The tip trajectory when taking ∆f (x) measurements, however, is significantly different. In this case rapid variations in ∆f can occur due to varying reactivity across a surface or molecule.
Therefore as the tip laterally moves across the surface the PLL must track sudden changes in ∆f which can lead to large errors in phase regulation, even with significantly increased bandwidth. To reduce the effect of phase errors on the recorded ∆f channel we increased the phase bandwidth from 15Hz (image acquisition) to 40Hz until deviations in the phase were reduced to ∼10 • or below. In Figure S7 we attempt to quantify the effect of the phase error(B) on the recorded ∆f (A) measured during a manipulation attempt. In principle the relationship between phase and frequency shift can be readily determined from a frequency sweep measurement as shown in Figure S7 C. If we approximate the Q of the tuning fork to remain unchanged as the tip interacts with the surface, an estimate for the phase-induced frequency error can be made as shown in Figure S7 D. Due to the high Q factor for the tuning fork at 5K temperatures and the high PLL bandwidths the error is shown to be minimal ( 0.05Hz). It is important to note, however, that at higher temperatures this error will be much more pronounced. For instance, frequency sweeps measured at 77K temperatures (which reduce the Q to ∼8000) show this error will increase to several 100mHz, thus becoming significant enough to affect calculated forces.

Apparent dissipation
In all measurements the oscillation amplitude was maintained at a constant value. Therefore in principle non-conservative dissipation can be measured. In the majority of cases where the oscillation excitation was recorded, dissipation above the molecule was not observed. Only in a minority of cases, particularly for the much larger frequency shifts recorded above Type II molecules, did we observe significant dissipation signals. Examination of the transfer function, however, demonstrated that this was largely due to apparent dissipation.
Labuda et al [4] have convincingly demonstrated that the transfer function of the piezoacoustic excitation system is rarely flat, leading to measurements of apparent dissipation at specific frequency shift values unrelated to physical tip-sample processes. In Figure S8 we show an in situ measurement of the transfer function recorded for a qPlus sensor immediately following measurements that recorded dissipation. It is clear that a significant peak in apparent dissipation is recorded between ∆f values of 10-20Hz. This corresponds exactly to the ∆f recorded above the Br 4 TPP molecule, suggesting that the recorded dissipation signal has no physical origin. It is therefore clear that apparent dissipation, in addition to the phase channel, must always be analysed to make meaningful conclusions from measured dissipation.