integrated modeling is the practice of coupling separate plasma physics codes — transport, equilibrium, heating and 📝current drive, pedestal and boundary — into a single workflow, iterated until each model's output is consistent with every other model's input.
The pieces cannot honestly be solved apart, because each one's answer is another one's assumption. Core temperature in a 📝plasma depends on turbulent transport; turbulent transport depends on the gradients and on the magnetic equilibrium; the equilibrium depends on the pressure and current profiles, and the current profile contains a 📝bootstrap current generated by the pressure gradient itself; radio-frequency power deposits where the density and temperature place the resonance; and the 📝pedestal sets the boundary condition on the core while the 📝scrape-off layer sets the boundary condition on the pedestal. Run any one code with assumed inputs and its answer is conditional on assumptions the others would contradict. Integrated modeling closes that loop and iterates it to a fixed point.
In practice a driver couples a transport solver — TRANSP, ASTRA, JINTRAC or TORAX — to an equilibrium solver, a reduced turbulence model such as TGLF, a pedestal model such as EPED, a full-wave code such as TORIC for 📝ion cyclotron resonance heating (ICRH), and a fast-ion module such as NUBEAM, with a boundary code beneath. The reduced turbulence model is the compromise that makes the loop tractable: nonlinear 📝gyrokinetics at every timestep is computationally out of reach, so a cheap model runs the iteration and a first-principles code spot-checks the result. Frameworks built to standardise the coupling, such as the STEP tool in OMFIT (arXiv:2305.09683), interface every code through a common 📝ITER-IMAS data structure so codes can be swapped or reordered without rewriting the workflow.
Integrated modeling is how a machine that does not yet exist acquires a number. 📝Pablo Rodriguez-Fernandez, who leads the MFE Integrated Modeling group at the 📝MIT Plasma Science and Fusion Center, produced the 2020 projection of 📝Q (fusion energy gain factor) ≈ 9 for 📝SPARC from TRANSP coupled to TGLF and EPED with TORIC and NUBEAM, checked afterwards against nonlinear CGYRO — one chapter of 📝The SPARC Physics Basis (2020). 📝Nathan Howard's 📝ARC performance work scoped an operating point at 1.13 gigawatts zero-dimensionally, then ran it through three independent integrated codes — TRANSP, ASTRA and TORAX — which returned 900 to 1300 megawatts. The spread is the point: a range across independent code chains states a projection more honestly than a single figure, and every number here is a prediction awaiting the machine that would test it. The instruments that would test it are designed the same way, through a 📝synthetic diagnostic applied to these very outputs.