Tip-enhanced Raman spectroscopy (TERS) integrates the scanning probe microscopy and surface-enhanced Raman spectroscopy (SERS). Similar to SERS, the enhancement of the analyte molecule’s Raman scattering in TERS greatly benefits from the inhomogeneous near field induced by the plasmon resonance of the metallic nanojunction. However, the TERS spectrum performs differently than SERS does. The TERS spectra strongly correlate with the tip positions as a result of the extremely confined near-field in angstrom volume. The combination of the sensitivity and the spatial resolution makes TERS a promising technique for microscopic characterization. 

A comprehensive demonstration of these unique features is TERS imaging, which correlates the Raman intensity with the tip position for a specific normal mode. TERS images exhibit sub-nanometer resolution and are mode specific, which are considered as visualizations of molecular vibrations. However, interpreting TERS images is challenging because the hotspot pattern does not always align with the vibrating atoms. Therefore, it is necessary to rationalize the relation between TERS images and the vibrational modes. In particular, it is critical to understand the near-field localization necessary to achieve the atomic resolution, and the molecular property that is locally probed by the near field. To address these two aspects, we have devoted our efforts to developing a consistent and intuitive description of TERS. 

Previously we used a hybrid atomistic electrodynamics/quantum mechanics method (termed DIM/QM) to simulate TERS images of small molecules such as benzene and porphyrin (ACS Nano 11, 5094-5102,2017). We found that an atomically sharp tip confines the near field in an extremely small volume with a diameter of less than half a nanometer. With such a field confinement, the hotspot size in TERS images can be as small as an atom. The molecular polarizabilities that are needed to calculate Raman intensities are solved self-consistently in the presence of the confined near field induced by the plasmonic nanojunction. However, the molecular polarizability is a nonlocal property. It was unclear what property of the vibrating atoms is locally probed by the tip. Moreover, repeated self-consistent calculations are required for every tip position, which limits the application of DIM/QM simulations to large molecules that are more interesting. 

In our current work (Nat. Commun. 10, 2567, 2019), we showed that the TERS images reflect local sub-molecular density changes induced by the confined near-field during the Raman process, and can be explained by a locally integrated Raman polarizability density (LIRPD) model. The Raman intensity is proportional to the square of the molecular polarizability change, i.e. Raman polarizability, in a given normal mode. The Raman polarizability is obtained from the local integration of Raman polarizability densities within the near-field spatial distribution. The polarizability density can be readily calculated from the perturbed electron density within the framework of time-dependent density functional theory. We showed that Raman polarizability densities are distributed along the atomic vibration vectors, and the spatial variation of TERS intensities can be explained by these Raman polarizability densities locally enhanced by the highly localized near field. 

The LIRPD model provides an intuitive description of TERS imaging mechanism. Furthermore, this method makes the electronic structure calculations independent of the tip position, which significantly reduces the overall computational cost. Together with the flexibility in tuning the near-field distribution, we were able to interpret TERS images of large molecules like H₂TBPP. Furthermore, it is now possible to efficiently scrutinize TERS images of complex molecule and determine their structure and vibrations as seen in experiments (e.g., Nature 568, 78-82, 2019).  

This blog is contributed by Pengchong Liu, Xing Chen, and Lasse Jensen.