Physics-informed tritium fuel cycle modelling workflow for fusion reactors

This paper presents a multi-fidelity, physics-informed framework based on the open-source PathSim/PathView platform that integrates zero-dimensional, one-dimensional, and high-fidelity multi-dimensional models to enable consistent and flexible tritium fuel cycle analysis for fusion reactors.

The Gist

Imagine you are trying to run a massive, high-tech bakery that never stops baking. But instead of flour and sugar, your main ingredient is Tritium, a rare and radioactive form of hydrogen. This is the challenge facing future fusion power plants: they need to make their own fuel (Tritium) while they are running, because it's too scarce to buy in bulk.

The problem is, managing this fuel is incredibly complicated. It flows through pipes, gets trapped in metal walls, bubbles through liquid metal, and has to be separated from other gases. If you lose too much, the plant shuts down. If you hold onto too much, it becomes dangerous.

This paper introduces a new digital "control tower" called PathSim/PathView that helps engineers design and test how to manage this fuel cycle. Think of it as a video game engine for fusion power plants, but instead of just looking cool, it uses real physics to predict what will happen.

Here is how they built this control tower using three different levels of detail, like zooming in with a camera:

1. The "Bird's Eye View" (The Zero-Dimensional Model)

The Analogy: Imagine looking at your bakery from a helicopter. You can see the total amount of flour in the silo, the amount in the mixers, and the amount in the ovens. You don't see individual grains of flour or how they move inside the mixer; you just know the average amount and how long it takes to get from one place to another.

• What the paper did: They used a simple math model to simulate the entire fuel cycle of a fusion plant (called an "ARC" plant). It treats every part of the system as a simple bucket. It's fast and great for getting a general idea of whether the plant can run, but it's a bit too simple to catch tricky problems.

2. The "Assembly Line" View (The 1D Physics Model)

The Analogy: Now, zoom in on one specific machine: the Bubble Column Reactor. This is a tall tube where liquid metal (holding the fuel) flows down, and gas bubbles flow up to "scoop" the fuel out.
Imagine trying to figure out how well a vacuum cleaner picks up dust. You can't just say "it works." You have to look at the speed of the air, the size of the bubbles, and how sticky the dust is.

• What the paper did: They built a more detailed model of this specific machine using complex equations (like traffic flow equations). They tested it against real-world data to make sure it worked. Then, they plugged this detailed machine into the simple "helicopter view" model.
• The Result: They could see what happens if the machine breaks down for a day. They found that having two machines working in parallel is safer than one big one, because if one stops, the other can speed up to compensate.

3. The "Microscope" View (The High-Fidelity Model)

The Analogy: Finally, zoom in all the way to the atomic level. Imagine looking at a single brick wall in your bakery. You want to see exactly how the heat moves through the bricks, or how a tiny drop of water seeps through a crack. This requires a super-powerful microscope (or in this case, a super-computer simulation called FESTIM).

• What the paper did: They connected this high-powered microscope directly to their control tower. Now, the system can simulate complex things like how Tritium gets "stuck" (trapped) inside the metal walls of the reactor or how it moves through different layers of material.
• The Magic: Usually, these super-detailed simulations take days to run and are too heavy to use in a whole-plant simulation. But this new framework allows them to run these heavy simulations inside the main system model, giving them the best of both worlds: speed and extreme accuracy.

Why This Matters

Before this, engineers had to choose between a fast, simple map (which might miss hidden dangers) or a slow, detailed map (which is too hard to use for the whole plant).

This paper shows how to build a hybrid map. You can use the simple map for the big picture, swap in a detailed map for the tricky parts (like the fuel extractor), and even zoom in with a microscope for the most critical safety checks—all in one single, open-source software package.

In short: They built a universal translator that lets engineers talk to each other whether they are thinking in "big buckets," "flowing pipes," or "atomic particles." This helps ensure that future fusion power plants won't just work in theory, but will actually be able to fuel themselves safely and efficiently in the real world.

Technical Summary

1. Problem Statement

Achieving tritium self-sufficiency is a critical challenge for the viability of future fusion power plants. While simplified, zero-dimensional (0D) residence time models have historically been used for early-stage fuel cycle design, they possess significant limitations:

• Lack of Physical Fidelity: They rely on empirical or assumed residence times and loss fractions, failing to capture the underlying physics of transport, retention, and exchange within subsystems.
• Inability to Model Complexity: They cannot accurately represent spatially varying conditions, non-linear couplings between processes, or time-dependent transients.
• Parameter Limitations: Key components (e.g., extraction systems, permeators) depend on complex parameters like mass transfer coefficients and diffusion rates that cannot be reduced to a single effective residence time.

The paper addresses the need for a framework that can integrate multi-fidelity models—ranging from simple system-level approximations to high-fidelity physics-based simulations—into a unified, dynamic environment.

2. Methodology

The authors propose a multi-fidelity, physics-informed framework built on the open-source PathSim (Python-based dynamical system simulator) and PathView (web-based graphical interface) platforms. The workflow integrates three distinct levels of modeling fidelity:

A. Low-Fidelity: 0D Residence Time Model

• Approach: Uses a system of ordinary differential equations (ODEs) to model tritium inventory ($I_i$) in each component based on inflow, outflow, residence time ($\tau_i$), and losses.
• Application: Reproduced the fuel cycle of an ARC-class fusion power plant to establish a baseline system-level description.
• Role: Serves as a computationally inexpensive tool for rapid preliminary design and feasibility studies.

B. Intermediate-Fidelity: 1D Physics-Informed Component Model

• Approach: Developed a coupled 1D ODE model for a Gas-Liquid Contactor (GLC) Bubble Column Reactor (BCR) used for tritium extraction from liquid Lithium-Lead (LiPb).
• Physics: Solves mass balance equations for both liquid and gas phases, incorporating dispersion, advection, and mass transfer flux ($J_T$) driven by Sievert's law.
• Integration: Implemented as a custom block within PathSim, allowing the BCR to function as the Tritium Extraction System (TES) within the larger dynamic fuel cycle.

C. High-Fidelity: Multi-Dimensional FEM Coupling

• Approach: Coupled FESTIM (an open-source finite element code for hydrogen isotope transport) directly to the system model.
• Physics: Simulates complex phenomena including diffusion, trapping, surface interactions, and multi-material interfaces in 1D and higher dimensions.
• Integration: PathSim acts as a coordinator, wrapping FESTIM simulations as blocks within the global timestepping loop, enabling the exchange of boundary conditions (e.g., pressure, concentration) between the system model and the detailed physics solver.

3. Key Contributions

1. Unified Multi-Fidelity Framework: Demonstrated the consistent combination of 0D, 1D, and multi-dimensional models within a single open-source environment (PathSim/PathView).
2. Dynamic System Integration: Successfully integrated a detailed 1D BCR model into a full fusion fuel cycle, allowing for the simulation of dynamic behaviors (e.g., startup transients, maintenance downtime) that static models cannot capture.
3. FESTIM-PathSim Coupling: Established a workflow for directly coupling high-fidelity finite element transport codes with system-level simulators, bridging the gap between component physics and system performance.
4. Open-Source Ecosystem: Provided a modular, extensible platform where high-fidelity models can be swapped for surrogate models or simplified physics depending on simulation needs, facilitating collaboration.

4. Key Results

• ARC Baseline Validation: The 0D residence time model successfully reproduced the fuel cycle behavior of an ARC-class plant, showing good agreement with previous literature (Meschini et al.), with minor inventory differences attributed to the neglect of radioactive decay and non-radioactive losses for simplicity.
• BCR Model Validation: The 1D BCR model was validated against recent literature (Mohan et al.). It successfully reproduced dimensionless liquid concentration and gas molar fraction profiles for both "Closed-Closed" and "Open-Closed" boundary conditions.
• Parametric Analysis (BCR):
  • Diameter Effects: Under "Open-Closed" conditions, extraction efficiency initially increases with column diameter (due to improved hydrodynamics) but decreases beyond a certain point due to axial dispersion (back-mixing). "Closed-Closed" conditions showed a plateau in efficiency.
  • Height Effects: Extraction efficiency is highly dependent on column height, increasing from ~2.5% at 1m to >13% at 5m.
• Dynamic Scenarios:
  • Series vs. Parallel: A single 3m BCR showed slightly higher steady-state efficiency (10%) than three 1m columns in series (9%) due to hydrostatic pressure effects.
  • Downtime Simulation: Simulating a shutdown of one BCR in a parallel configuration showed that the remaining column's increased flow rate maintained extraction efficiency, preventing blanket inventory accumulation. Conversely, a single BCR shutdown caused immediate inventory accumulation in the blanket and depletion in storage.
• FESTIM Integration: Successfully simulated hydrogen diffusion through a slab and a depleted source problem, showing excellent agreement between FESTIM results and analytical solutions.

5. Significance

This work represents a significant step forward in fusion fuel cycle analysis by moving beyond purely empirical system models.

• Design Optimization: It enables designers to perform "what-if" analyses that account for complex physical phenomena (e.g., spatial gradients, material interfaces) without sacrificing the ability to simulate full plant dynamics.
• Predictive Capability: The framework supports predictive analysis for tritium doubling times and startup strategies by capturing transient behaviors and non-linear couplings.
• Future Scalability: The modular architecture lays the foundation for integrating other critical physics tools, such as OpenMC (neutronics) and OpenFOAM (fluid dynamics), and utilizing surrogate models to accelerate high-fidelity evaluations.
• Community Impact: By being open-source, the framework fosters collaboration and standardization in the fusion community, allowing researchers to share and validate complex fuel cycle models.