Challenges Related to Tritium Transport in Fusion Reactors:

from component to system level

Remi Delaporte-Mathurin

Plasma Science and Fusion Center, MIT, USA

MATERIAL

Half-life: 12 years

\mathrm{T} \rightarrow \mathrm{He} + \mathrm{e}^-

☢

Consumption of a 1 GWth fusion reactor (1 year)
50 kg

+ radiological safety & regulatory issues

The breeding blanket

\mathrm{n} + ^6\mathrm{Li} \rightarrow \mathrm{T} + \mathrm{He} + 4.8 \ \mathrm{MeV}

\mathrm{n} + ^7\mathrm{Li} \rightarrow \mathrm{T} + \mathrm{He} + \mathrm{n} - 2.5 \ \mathrm{MeV}

\mathrm{TBR} = \mathrm{\frac{tritium \ produced}{tritium \ consumed}} > 1

\mathrm{D} + \mathrm{T} \rightarrow \mathrm{He} + \mathrm{n}

The breeding blanket is only one component of the fuel cycle

Samuele Meschini et al 2023 Nucl. Fusion 63 126005

Component

Material

What is the required TBR? Startup tritium inventory?

What are the minimum dimensions of the extractor?

How much T is retained in the plasma facing materials?

What is the T concentration at the blanket outlet?

What is the diffusivity of EUROFER?

Trapping properties of tungsten?

Permeation reduction factor of coating X?

System

Multi-scale tritium design

Design issues propagate

Challenge #1: the residence time method is not predictive

Challenge #3: Experiments for Property Measurements Are Not Standardised

Challenge #4: Data on Transport Properties is Scarce and Variable

Challenge #2: Component-Level Testing Remains Limited

Component

Material

System

Outline

Component

Material

System

Challenge #1: the residence time method is not predictive

Context: system fuel cycle modelling and the residence time method

\underbrace{\frac{d I_i}{dt}}_\text{inventory evolution} = \underbrace{\sum F_\mathrm{in}}_\text{tritium in} - \underbrace{\frac{I_i}{\tau_i}}_\text{tritium out}

Residence time

For each component \(i\)

Goal: to model tritium mass fluxes between components

Samuele Meschini et al 2023 Nucl. Fusion 63 126005

Illustration with a simple fuel cycle model

Burns tritium

Breeds tritium (TBR)

Tritium Extraction System

Breeding Blanket

Plasma

Storage

neutrons

Source: remdelaportemathurin.github.io/interactive-plots/fuel_cycle/

Startup inventory

Tritium Extraction System

Breeding Blanket

Plasma

Storage

Limitations

How to estimate the residence time?
What if there are several phases in the blanket (eg. LiPb and coolants)?
Great simplifications need to be made for permanent trapping

Current situation

System-level tritium models often use fixed residence times for each subsystem.
Approach is empirical and lacks connection to material physics.

Risks & implications

Path forward

Large uncertainties and increased safety and performance risks.
Inconsistent assumptions across organisations → poor comparability.
Weak technical basis for regulatory review.

Move towards physics-informed component-level models

Challenge #1: the residence time method is not predictive

Path forward

Move towards physics-informed component-level models

Challenge #1: the residence time method is not predictive

Other physics-informed models:

Simpler 1D models
Surrogate models
Empirical experimental laws
...

FESTIM model of an ARC's breeding blanket

COMSOL model of a bubble column T extractor

Teng Wang et al 2025 J. Phys.: Conf. Ser. 3011 012054

Dark et al, Tritium 2025, 10.13140/RG.2.2.28729.84321

Component

Material

System

Challenge #2: Component-Level Testing is limited

Current situation

only few integrated tritium breeding tests + extraction tests exist under "relevant" combined conditions
other hydrogen experiments are focused on coupon tests

Risks & implications

Path forward

Large uncertainty in system-level tritium recovery predictions.
Missing data for code validation and licensing.

More component level experiments such as LIBRA
A dedicated fuel cycle component test facility

Challenge #2: Component-Level Testing is limited

Component

Material

System

Challenge #3: Property Measurements are not Standardised

Context: hydrogen transport properties

\frac{\partial c_m}{\partial t} = \nabla \cdot (D \nabla c_m) - k\ c_m \ (n - c_t) + p \ c_t

c= K_S \ \sqrt{P}

Governing equations for H transport

Material properties

Properties can be obtained experimentally (gas driven permeation, plasma-driven permeation, absorption, thermo-desorption, electro-permeation...)
Simple in principle, complex in practice.

\varphi = k_d \ P - k_r \ c^2

MIT's SHIELD permeation rig

Current situation

No standards for experiments (except electro-permeation ASTM)
No standard way of disseminating results (units, dimensions...): see Challenge #4
Hard to quantify tritium retention/trapping
Few inter-laboratory comparisons (round Robins).

Risks & implications

Path forward

Poor reproducibility and large data scatter.
Difficult to identify reliable reference values.
Increased risk for dimensioning of components.

Develop standardised experiments & protocols

for GDP, TDS, PDP...

Challenge #3: Property Measurements are not standardised

Component

Material

System

Challenge #4:

Data on transport properties is scarce and variable

Context: hydrogen transport properties (next)

Tungsten's diffusivity

LiPb solubility

Different tungsten grades?
Different setups?
Isotopic effect?
Influence of trapping? effective diffusivity?

4 orders of magnitude variation!
Composition?
Difficulties measuring liquids?
Mistakes in unit conversion?

Source: HTM database v0.17

Current situation

diffusivity, solubility, and trapping parameters vary by up to orders of magnitude in literature.
several sources of truth (lit. reviews) and no unified database

Risks & implications

Path forward

unreliable component-level simulations
poor reproducibility of model predictions.

Establish a large hydrogen transport properties database

Challenge #4: Data on Tritium Properties is Scarce and Variable

Path forward

Establish a large hydrogen transport properties database

Challenge #4: Data on Tritium Properties is Scarce and Variable

Citable and Persistent

Each dataset has a DOI and version number.
Enables reproducibility and traceable references in reports, models, and standards.

Peer-Reviewed and Curated

Data vetted by subject-matter experts before inclusion.
Documented acceptance criteria (e.g. residuals, experimental validation quality).

Standardised Format

Machine-readable (e.g. JSON, CSV, or YAML) and interoperable with modelling codes
Consistent units, naming conventions, and metadata schema (e.g. composition, phase, microstructure).

Version-Controlled, Openly Licensed, and Transparent

Public Git repository or data platform (e.g. Zenodo, GitHub).
Each update logged with changelogs and clear licence (CC-BY or similar).
Extensible.
Full provenance of every data point

Needs to be:

The HTM database

✅Citable and Persistent

✅Version-Controlled, Openly Licensed, and Transparent

✅Peer-Reviewed and Curated

✅Standardised Format

✅Compatible with modelling workflows

github.com/remdelaportemathurin/h-transport-materials

But lacks one crucial ingredient: incentives from research programmes

HTM dashboard

https://htm-dashboard-uan5l4xr6a-od.a.run.app/

Component

Material

System

Bonus Challenge:

Open-science and transparency

Data availability and open science

"dAtA WiLl bE aVaiLabLe uPoN rEqUeSt"

Current situation

Many modelling tools and datasets are closed-source
Lack of transparency when disseminating results.

Risks & implications

Path forward

Limited reproducibility
Duplication of effort
Slow progress toward standardisation and regulatory trust.

Support open-source initiatives and provide incentives to open-science
Provide clear guidelines on how to do open-science

Challenge #5: Open-Science and Transparency

Unpopular opinion: open-science is more than publishing a paper at the end of a project

https://ossfe.org/

Thank you!

Any question?

✉️ remidm@mit.edu

github.com/f estim-dev