Tutorials

Instructors: Maxim Ziatdinov (PNNL), Sarah Allec (PNNL), Semanti Mukhopadhyay (PNNL)
Format: Hands-on with laptops (bring your laptop)
Level: Introductory; no prior AI/ML background required

Motivation
Experimental workflows central to the AVS community – surface spectroscopy, scanning-probe and electron microscopy, hyperspectral mapping, and the many forms of process tuning that go on around a vacuum chamber – can generate more data than any individual researcher can interpret in real time. Agentic AI systems, in which large language models plan, call analysis tools, dynamically write and execute custom code, and reason over results, are rapidly maturing into practical research partners for these tasks. This tutorial introduces AVS participants to the agentic-AI paradigm through hands-on use of the open-source platform developed for physical-science research, while keeping the discussion broad enough to remain relevant as the underlying agentic-AI ecosystem evolves.

Why this matters to AVS participants

Hands-on examples will center on agent-driven automated analysis applied to techniques familiar to the AVS audience, including:

• Surface and thin-film spectroscopy – automated peak deconvolution, chemical-state assignment, and literature-aware fitting.
• Scanning-probe and electron microscopy – automated atomic-resolution image analysis, defect and phase identification, and series-level outlier detection.
• Atom-probe tomography – agent-assisted exploration of 3D reconstructions, composition profiles, and clustering/segregation analyses.
• Hyperspectral imaging – automated spectral-image unmixing, phase mapping, and material system-tailored focused analysis with dynamically generated code.
• Experimental planning and optimization – using accumulated measurements, prior literature, and data-driven methods to suggest informative next experiments.
• Closing the loop with first-principles simulation – coupling experimental observations to DFT calculations so that an agent can propose candidate structures, set up and run simulations, and feed the results back into the next round of experimental decisions.
• Serendipity angle – automatically scoring findings against published work so that experimental effort focuses on what is genuinely new.

The specific demos will be refreshed close to the event to reflect the current state of the agentic-AI ecosystem and the tooling available at that time.

Tutorial Outline
The tutorial is organized into loosely timed blocks rather than a rigid schedule, so the material can be updated as the field evolves:

1. Foundations. What an “agent” is in practical terms – language model plus tools plus autonomy levels – and how this differs from a chatbot or a classical pipeline. Short live demonstration.
2. Analysis on your data. Participants run agent-driven analyses on representative surface-science and microscopy datasets, then read and critique the auto-generated reports.
3. From analysis to decisions. Connecting analysis outputs to planning: how an agent can use prior data and the literature to suggest the next experiment, and where human judgment should override.
4. Bridging experiment and theory. A walkthrough of an experiment-to-DFT loop – moving from an experimental observation (e.g., a defect or interface seen in microscopy or spectroscopy) to a candidate atomic structure, and an automatically prepared first-principles calculation input file.
5. Extending and integrating. Adding domain-specific skills and custom tools, and interoperating with the wider agentic ecosystem (e.g., MCP-based integrations with other AI clients and external data sources).

A short closing discussion will cover limitations, failure modes, and where human oversight remains essential.

Learning Outcomes
Participants will leave able to (1) install and run an agentic system on their own data, (2) read and critically evaluate an agent-generated analysis report, (3) understand how experimental analysis, planning, and first-principles simulation can be linked into a single agent-mediated loop, and (4) extend such a system with a domain-specific skill or tool relevant to their technique.

Logistics
Sample datasets, notebooks, and supporting materials will be provided during the tutorial. A limited number of API keys will be available for participants working through the provided examples.

Instructors: Derek T. Anderson (University of Missouri), Matthias J. Young (University of Missouri), Matthew R. Maschmann (University of Missouri)

Recent advances in large language models make it possible for researchers to build customized scientific machine learning workflows using natural language rather than traditional software engineering to generate code. For surface and materials scientists, this creates an exciting opportunity to engage with machine learning in a more accessible and domain-relevant way, especially when paired with a clear conceptual foundation and practical strategies for applying these tools effectively. This 90-minute hands-on tutorial acts as a launchpad for this transition by pairing an accessible introduction to machine learning concepts with a practical “vibe coding” tutorial for building applied AI workflows for materials and surface science domains with no-to-low prior coding experience. Participants will learn the basics of machine learning and how to use LLM-assisted code development to construct data ingestion, preprocessing, model training, evaluation, and visualization pipelines while maintaining scientific rigor. The workshop emphasizes how to identify problems suited for machine learning/deep learning, frame materials questions in computational terms, and distinguish deep learning workflows from LLM-based or agentic systems. Through practical examples and guided exercises, attendees will also learn prompting, debugging, validation, and refinement strategies for AI-generated code. By the end of the session, participants will leave with reusable workflows, example prompts, and a clearer path for responsibly applying AI-assisted machine learning to their own datasets.

Learning Objectives

Participants will:

1. Understand the distinction between AI, ML, DL, LLMs, and agentic AI systems
2. Learn when ML/DL is appropriate for scientific research problems
3. Develop practical natural language-based (vs coding) prompting strategies for low-to-no code AI-assisted software development
4. Build and validate end-to-end DL workflows using scientific datasets
5. Gain experience generating data analysis and visualization tools with AI assistance
6. Leave with reusable example workflows and educational resources

Target Audience

This workshop is intended for:

• Experimental researchers in surface science and materials science
• Industry scientists and engineers
• Graduate students and postdoctoral researchers

Instructor: Erwin Kessels (Eindhoven)

Atomic-scale processing has become essential in the semiconductor industry, where continued device scaling through advanced patterning, the introduction of new materials, the transition to three-dimensional architectures, and heterogeneous integration increasingly require the ability to deposit, remove, and modify materials with atomic-level precision. At the same time, the relevance of these methods extends well beyond semiconductor manufacturing, with important applications in energy technologies, catalysis, optics, quantum devices, sensors, and advanced nanomaterials.

Atomic layer deposition (ALD) is firmly established as a method for preparing ultrathin films with precise thickness control, excellent conformality, and high uniformity. Atomic layer etching (ALE), in turn, is becoming increasingly important for controlled material removal, surface cleaning, and profile engineering. In parallel, area-selective deposition and related approaches are attracting rapidly growing interest as enabling concepts for selective processing.

This tutorial will introduce the basic concepts of atomic-scale processing, with a focus on ALD, ALE, and related surface-controlled methods. It will address the surface chemistry that underpins these processes, as well as the plasma-surface interactions involved in plasma-enhanced approaches. Key concepts enabling self-limiting and highly controlled processing will be discussed, including saturation, conformality, directionality, and selectivity. The similarities and differences between ALD and ALE will be highlighted, and the tutorial will also show how ALD, ALE, area-selective deposition, and surface preparation steps can be combined in advanced process flows for nanoscale and three-dimensional device fabrication.

Special attention will be given to why and when atomic-scale processing methods are useful, what limitations and misconceptions exist, and how process and reactor choices influence material properties and device-relevant performance. Examples from key material systems and applications will be used to illustrate both the opportunities and challenges of atomic-scale processing in current and future technologies. 

Learning Objectives
Participants will:

1. Understand the basic principles of atomic-scale processing, including ALD, ALE, and related surface-controlled methods
2. Learn how self-limiting and cyclic processes enable precise control over film growth, material removal, surface modification, and selectivity
3. Gain insight into the role of surface chemistry and plasma-surface interactions.
4. Recognize the similarities and differences between ALD and ALE, and understand how related concepts can be extended toward area-selective deposition and other selective processing approaches.
5. Learn how ALD, ALE, surface preparation, and selective processing steps can be combined in advanced process flows for nanoscale and three-dimensional device fabrication
6. Develop an awareness of key applications, limitations, misconceptions, and future opportunities for atomic-scale processing in semiconductor manufacturing and beyond

Target Audience
This tutorial is intended for graduate students, postdoctoral researchers, other researchers, and engineers working in academia or industry, as well as anyone interested in an accessible, engaging, and enjoyable introduction to atomic-scale processing. It is designed to be useful for both newcomers to the field and more experienced participants who would like to broaden or refresh their perspective.

Instructors: Freek Molkenboer (TNO), Keith Middleman (STFC Daresbury Laboratory)  and Giulia Lanza (SLAC)

Vacuum technology plays an essential role in many scientific and industrial applications, from surface analysis to semiconductor manufacturing. This tutorial offers a clear and accessible introduction for students and newcomers who want to understand the basics of vacuum technology.

We will cover the main pressure regimes, how pumping systems work, and how vacuum levels are measured. Practical examples show how vacuum environments support modern technologies, including a brief look at ultra‑clean vacuum and why it matters for areas like EUV lithography.

The goal of this 1–2 hour session is to give attendees a solid starting point, helping them a better understand on core concepts of vacuum and vacuum systems.