The separatrix operational space of next-step fusion experiments: From ASDEX Upgrade data to SPARC scenarios is the 2025 Nuclear Materials and Energy paper in which 📝Thomas Eich and fourteen co-authors turn 6,688 📝ASDEX Upgrade measurements into a model, then use that model to pick an ELM-free operating point for 📝SPARC.
The requirement it serves is blunt: a fusion power plant has to run without type-I 📝edge localized modes and with a detached 📝divertor, because periodic ELM heat pulses erode and damage divertor surfaces faster than a plant can tolerate. The separatrix operational space — SepOS, recast here in simpler parameters — is the tool for finding where a machine can sit in the type-I-ELM-free quasi-continuous exhaust regime.
The measured half is ASDEX Upgrade. Against an extended database of 6,688 individual measurements from that tokamak, SepOS reproduces how the 📝H-mode boundary shifts with 📝plasma current and magnetic field strength, and correctly separates 📝L-mode, H-mode and density-limit 📝disruption. Every number in that half came off an operating machine. The generalizing step is normalization: scaled to the L–H minimum density and temperature, the H-mode boundaries of multiple devices are nearly identical, which is what licenses carrying a result from one machine to another. The L–H minimum density SepOS predicts also matches an experimentally determined multi-machine scaling, an independent check on the framework.
The predicted half is SPARC. Applying the normalized framework, the authors identify a viable quasi-continuous-exhaust operating point at 📝separatrix density 4×10²⁰ m⁻³, separatrix 📝electron temperature 156 eV and turbulence parameter α_t = 0.7 — a point sitting solidly inside the QCE 📝operational space measured on ASDEX Upgrade. That operating point is a projection for a tokamak that has not yet produced a plasma; its credibility rests entirely on the 6,688 measurements that are not projections. The same normalization is offered for 📝ARC, 📝ITER and 📝DEMO. Alongside Eich the byline includes 📝Thomas Body, 📝Tom Looby, 📝Adam Kuang, 📝Matt Reinke, 📝Alex Creely, 📝Jon Hillesheim and 📝Jerry Hughes.
We did not measure this on SPARC — nobody has. We measured it on ASDEX Upgrade, 6,688 times, and this paper is the arithmetic that carries those measurements onto a machine we are still building.