This dataset contains electrostatic simulation outputs used in the study "Physics-Informed Neural Network Surrogate Model For Capacitive Touch Sensors By Solving Maxwell’s Equations". It supports the development, training, and evaluation of surrogate models for capacitive touch sensing, with particular relevant to physics-informed neural networks. Each CSV file stores the results of one simulation scenario corresponding to a specific finger position in a three-dimensional computational domain. The main purpose of the dataset is to enable the learning and assessment of machine learning models that approximate the electrostatic response of a capacitive touch sensor. The data provide field-based outputs associated with different finger-position configurations and therefore support research on surrogate modelling, simulation-based learning, and generalisation across varying spatial conditions.

METHODOLOGICAL INFORMATION

1. Description of methods used for collection-generation of data: The raw dataset was generated with the finite element method from COMSOL Multiphysics. The simulation model included a sensor positioned in the middle, a simulated finger above the sensor, and a PCB plate below the sensor. The sensor was excited with 3.3V, while the finger and PCB were set as ground. The COMSOL electrostatics solver was then used to compute the spatial distribution of electric potential (V) and electric field components (Ex, Ey, Ez), as well as charge density (rho) on the sensor surface. Different simulations were run for multiple finger positions in 3D space, the corresponding outputs from COMSOL were deposited into two CSV files, and the file names encode the corresponding finger position.

2. Methods for processing the data: This dataset was derived from raw COMSOL outputs. For each finger position, two raw CSV files were available: one containing electric potential and electric field vector, and one containing charge density values on the sensor plate. These two files were merged into a single CSV file for each finger position. During processing, the finger position was added explicitly as dist_x, dist_y, and dist_z, and charge density values were extended to the full spatial coordinate set so that all rows share the same format. The processing was performed in batch using a Python script.

3. Instrument- or software- specific information needed to interpret the data: The raw data were generated using COMSOL Multiphysics (release version 6.2) with AC/DC module. The published dataset was processed from the raw data with python 3.9.18, and now it is provided in CSV format for open reuse. The main columns are x, y, z, dist_x, dist_y, dist_z, V, Ex, Ey, Ez, and rho.

4. Environmental or experimental conditions: The dataset was produced entirely through a computational environment rather than through direct physical acquisition. The simulation assumed an electrostatic setup with the sensor at 3.3 V and both the simulated finger and PCB at ground potential. Raw results depend on the modelled geometry, material settings, and solver configurations defined in COMSOL.

This dataset supports a manuscript (https://doi.org/10.7148/2025-0390) and may be updated in future versions if the related study, documentation, or file organisation changes. Users of this dataset are encouraged to cite both the dataset DOI and the associated manuscript when reusing the data.