Atomic Force Microscopy Simulations for CO-functionalized tips with Deep Learning

TL;DR

This paper introduces a deep learning approach using CGANs to efficiently simulate AFM images of molecules with CO-functionalized tips, simplifying the process by using only 2D molecular representations.

Contribution

It presents a novel CGAN-based method that bypasses complex force calculations, enabling faster AFM image simulations from simple 2D molecular depictions.

Findings

Accurately reproduces intramolecular contrast for quasi-planar molecules.

Limited performance for molecules with internal torsion due to 2D input constraints.

Provides an efficient alternative to traditional ab initio simulation methods.

Abstract

Atomic Force Microscopy (AFM) operating in the frequency modulation mode with a metal tip functionalized with a CO molecule images the internal structure of molecules with an unprecedented resolution. The interpretation of these images is often difficult, making the support of theoretical simulations important. Current simulation methods, particularly the most accurate ones, require expertise and resources to perform ab initio calculations for the necessary inputs (i.e charge density and electrostatic potential of the molecule). Here, we propose an efficient and simple alternative to simulate these AFM images based on a Conditional Generative Adversarial Network (CGAN), that avoids all force calculations, and uses as the only input a 2D ball--and--stick depiction of the molecule. We discuss the performance of the model when optimized using different training subsets. Our CGAN reproduces accurately the intramolecular contrast observed in the simulated images for quasi--planar molecules, but has significant limitations for molecules with a significant internal torsion, due to the strictly 2D character of the input.