# Conditional diffusion model predicts variability in focused-ion-beam fabrication

> Researchers at the University of Southampton developed a diffusion model that predicts a distribution of possible outcomes for micro- and nanostructures milled by focused ion beam, rather than a single result. It was trained on 100 paired designs and scanning-electron-microscope images.

- Canonical URL: https://newsverum.com/en/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026
- Publisher: Newsverum
- Section: Science
- Country: United Kingdom
- Published: 2026-10-13T01:40:12+00:00
- Updated: 2026-10-13T01:40:12+00:00
- Source review period: 2026-10-09
- Other editions: [Türkçe](https://newsverum.com/tr/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md), [العربية](https://newsverum.com/ar/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md), [Français](https://newsverum.com/fr/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md), [Español](https://newsverum.com/es/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md), [Deutsch](https://newsverum.com/de/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md), [Português](https://newsverum.com/pt/news/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026.md)

*The source does not state the event date; the period shown is the source review period.*

Focused-ion-beam milling can produce different edge and linewidth outcomes from repeated runs of the same design because of effects such as beam-current fluctuations and redeposition. The team conditioned a diffusion model on a binary design mask and generated a set of plausible scanning-electron-microscope-like outputs instead of one deterministic prediction.

For line trenches in a gold film, the model reproduced measured linewidth distributions and orientation-dependent defect behavior. Yet milling often failed for narrow trenches aligned parallel to the beam’s raster lines. The results use 100 paired training examples and a limited set of test geometries; this is a research tool for studying process tolerances, not a guarantee of fabrication performance.

## Data

### Measurements reported by the study

The table reports only findings within this study’s experimental or model scope.

Methodology: Quantity and scope are taken from the methods and results of the first arXiv version dated 9 October 2026; they should not be generalized beyond the study.

| Reported study quantity | Value | Study-specific scope |
|---|---|---|
| Paired mask and sample images in the training data | 100 | 100 design/SEM-image pairs from focused-ion-beam milling experiments on gold film; model outputs are simulations, not additional fabricated samples. |

Sources: <https://arxiv.org/abs/2610.13420> · Download data: [JSON](https://newsverum.com/api/v1/articles/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026/data/study-measurements?lang=en&format=json) · [CSV](https://newsverum.com/api/v1/articles/conditional-diffusion-model-predicts-fib-nanofabrication-variance-2026/data/study-measurements?lang=en&format=csv)

## Sources

- [arXiv](<https://arxiv.org/abs/2610.13420>) (Source checked: 2026-10-13; Published: 2026-10-09)

---

Cite the canonical URL above. [Editorial standards and corrections](https://newsverum.com/en/editorial)