Layerwise Deposition-Field Generative Modeling: A Process-Conditioned Framework for Three-Dimensional Generation Through Thin-Film Deposition and Surface Growth
Abstract
Three-dimensional generative modeling has predominantly represented objects through explicit geometric primitives, point clouds, implicit fields, radiance fields, or learned volumetric representations. However, many physical structures are formed through sequential material deposition, surface growth, interface formation, and repeated processing operations rather than through a single geometric construction step. This paper proposes Layerwise Deposition-Field Generative Modeling (LDF-GM), a process-conditioned framework for three-dimensional generation inspired by physical vapor deposition (PVD), chemical vapor deposition (CVD), and related thin-film fabrication processes. LDF-GM represents an evolving structure using a material field, a surface-growth field, an interface state, and a sequence of deposition operations. Each operation updates the current structure according to a learned transition model conditioned on the existing surface, material properties, deposition parameters, and process history. A hierarchical latent-variable formulation jointly models the number of deposition stages, stage-specific process conditions, and the resulting three-dimensional structure. The framework supports variable layer counts, heterogeneous materials, nonuniform thickness, conformal coatings, interface-dependent growth, and accumulated surface morphology. A physics-informed objective can incorporate material conservation, geometric consistency, process-conditioned growth priors, and interface regularization. This paper provides the mathematical formulation, a practical training strategy, a generation algorithm, and an experimental protocol for evaluating the proposed method against geometry-first generative models and process-based growth simulators. Since no implementation or benchmark experiments are reported here, the method's effectiveness and novelty remain hypotheses requiring empirical validation. LDF-GM is intended to investigate whether representing three-dimensional generation as a sequence of material-deposition state transitions can improve process controllability, structural consistency, and the generation of physically plausible layered geometries.
Related
- Deposition-Based 3D Generative Modeling: A Framework for Three-Dimensional Object Generation Through Gas-Phase Particle Deposition, Local Adsorption, and Branching Growth
- Fluvial Erosion–Deposition Generative Model (FED-GM): A Process-Driven Framework for Three-Dimensional Terrain Generation Through Sediment Transport and Geomorphic Feedback
- Granular Flow and Deposition Generative Modeling: A Physics-Driven Framework for Three-Dimensional Shape Generation Through Discrete Particle Dynamic
- Reaction-Driven Three-Dimensional Generative Computing: A Framework for Constructing 3D Objects Through Local Chemical Transformation, Deposition, and Structural Growth
- Collective Biological Construction: A Multi-Unit Generative Framework for Three-Dimensional Shape Formation Through Local Deposition and Collective Structural Evolution
BibTeX
@misc{zhang2026layerwise,
title = {{Layerwise Deposition-Field Generative Modeling: A Process-Conditioned Framework for Three-Dimensional Generation Through Thin-Film Deposition and Surface Growth}},
author = {Jincheng Zhang},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23636346},
url = {https://zenodo.org/records/23636346},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Layerwise Deposition-Field Generative Modeling: A Process-Conditioned Framework for Three-Dimensional Generation Through Thin-Film Deposition and Surface Growth
AU - Jincheng Zhang
PY - 2026
DA - 2026/10/25
PB - Zenodo
DO - 10.5281/zenodo.23636346
UR - https://zenodo.org/records/23636346
AB - Three-dimensional generative modeling has predominantly represented objects through explicit geometric primitives, point clouds, implicit fields, radiance fields, or learned volumetric representations. However, many physical structures are formed through sequential material deposition, surface growth, interface formation, and repeated processing operations rather than through a single geometric construction step. This paper proposes Layerwise Deposition-Field Generative Modeling (LDF-GM), a process-conditioned framework for three-dimensional generation inspired by physical vapor deposition (PVD), chemical vapor deposition (CVD), and related thin-film fabrication processes. LDF-GM represents an evolving structure using a material field, a surface-growth field, an interface state, and a sequence of deposition operations. Each operation updates the current structure according to a learned transition model conditioned on the existing surface, material properties, deposition parameters, and process history. A hierarchical latent-variable formulation jointly models the number of deposition stages, stage-specific process conditions, and the resulting three-dimensional structure. The framework supports variable layer counts, heterogeneous materials, nonuniform thickness, conformal coatings, interface-dependent growth, and accumulated surface morphology. A physics-informed objective can incorporate material conservation, geometric consistency, process-conditioned growth priors, and interface regularization. This paper provides the mathematical formulation, a practical training strategy, a generation algorithm, and an experimental protocol for evaluating the proposed method against geometry-first generative models and process-based growth simulators. Since no implementation or benchmark experiments are reported here, the method's effectiveness and novelty remain hypotheses requiring empirical validation. LDF-GM is intended to investigate whether representing three-dimensional generation as a sequence of material-deposition state transitions can improve process controllability, structural consistency, and the generation of physically plausible layered geometries.
ER -