QUAM-AFM

<p>QUAM–AFM is the largest dataset of simulated Atomic Force Microscopy (AFM) images generated from a selection of 685,513 molecules that span the most relevant bonding structures and chemical species in organic chemistry. QUAM-AFM contains, for each molecule, 24 3D image stacks, each consisting of constant-height images simulated for 10 tip-sample distances (in the relevant imaging range and spanning a variation of 1 Å (0.1 nanometers)) with one of the 24 different combination of AFM operational parameters, resulting in a total of 165 million images with a resolution of 256x256 pixels. The 3D stacks are especially appropriate to tackle the goal of chemical identification within AFM experiments by using deep learning techniques.</p>

<p>The operational parameters include six different values for the cantilever oscillation amplitude (0.40, 0.60, 0.80, 1.00, 1.20, 1.40 Å), 4 values of the elastic constant describing the tilting of the CO tip (0.40, 0.60, 0.80 and 1.00 N/m). The first parameter is freely chosen in the experiments in order to enhance different features of the image, while the last one reflects differences in the attachment of the CO molecule to the metal tip that are routinely observed and has been characterized in the experiments.</p>

<p>The data provided for each molecule includes, besides a set of AFM images, the ball–and–stick depiction, the IUPAC name, the chemical formula, the atomic coordinates, and the map of atom heights. In order to simplify the use of the collection as a source of information, we have developed a Graphical User Interface (GUI) that allows the search for structures by CID number, IUPAC name or chemical formula.</p>

<p>This dataset arises as a product of the research carried out in collaboration between Quasar Science Resources S.L. (https://quasarsr.com) and the Scanning Probe Microscopy Theory & Nanomechanics Research Group (SPMTH) (http://www.uam.es/spmth) at the Universidad Autónoma de Madrid (UAM), funded by the Comunidad de Madrid under the Industrial Doctorate Programme 2017 (project reference IND2017/IND-7793).</p>

<p>The main goal of this dataset is to support the development of deep learning methods for molecular identification through AFM imaging. Once this project has concluded, this dataset is made freely accessible in order to facilitate and to promote research in a range of fields including Atomic Force Microscopy, on-surface synthesis and deep learning applications.</p>

QUAM-AFM是当前规模最大的模拟原子力显微镜（Atomic Force Microscopy, AFM）图像数据集，其生成基础为685513个精选分子——这些分子覆盖了有机化学领域最具代表性的键合结构与化学物种。QUAM-AFM为每个分子提供24组三维图像栈，每组图像栈包含针对10种针尖-样品间距（处于常规成像范围内，间距变化跨度为1埃（0.1纳米））模拟得到的恒高图像，且搭配24种不同的AFM操作参数组合，最终总计生成1.65亿张分辨率为256×256像素的图像。该三维图像栈尤其适用于借助深度学习技术实现AFM实验中的化学分子识别任务。

其操作参数包含6种不同的悬臂振荡振幅（0.40、0.60、0.80、1.00、1.20、1.40 埃），以及4种描述一氧化碳（CO）针尖倾斜特性的弹性常数（0.40、0.60、0.80和1.00 N/m）。其中，悬臂振荡振幅可在实验中自由调整，以强化图像的不同特征；而弹性常数则反映了金属针尖上一氧化碳分子附着方式的差异——这类差异在实验中被常规观测到，且已有相关实验表征结果。

针对每个分子，数据集除提供一系列AFM图像外，还包含球棍模型图、国际纯粹与应用化学联合会（International Union of Pure and Applied Chemistry, IUPAC）命名、化学式、原子坐标以及原子高度映射图。为简化该数据集的使用流程，开发团队搭建了图形用户界面（Graphical User Interface, GUI），支持通过CID编号、IUPAC命名或化学式检索目标分子结构。

本数据集由Quasar Science Resources S.L.（https://quasarsr.com）与马德里自治大学（Universidad Autónoma de Madrid, UAM）下属的扫描探针显微镜理论与纳米力学研究组（Scanning Probe Microscopy Theory & Nanomechanics Research Group, SPMTH）（http://www.uam.es/spmth）合作研发，获得马德里自治区2017年工业博士项目资助（项目编号IND2017/IND-7793）。

本数据集的核心目标是支持基于AFM成像的分子识别深度学习方法开发。本项目结题后，该数据集将免费开放获取，以推动原子力显微镜、表面合成以及深度学习应用等多个领域的研究工作。

搜集汇总

背景与挑战

背景概述

QUAM-AFM是目前最大的模拟原子力显微镜图像数据集，包含来自685,513个分子的1.65亿张高分辨率图像，涵盖有机化学中最相关的键合结构和化学物种。该数据集专门为开发基于深度学习的AFM分子识别方法而设计，提供了丰富的分子图像数据和相关化学信息。