Magnetic equilibrium reconstruction (EFIT)

Overview

Magnetic equilibrium reconstruction is a computational method fundamental to the operation and analysis of magnetic confinement fusion devices, particularly tokamaks and stellarators. It serves as a computational diagnostic, inferring the internal state of the plasma—specifically the magnetic field structure, plasma pressure, and current distribution—from a limited set of external measurements. This process is necessary because direct, comprehensive measurement of these internal parameters is not feasible in the high-temperature plasma environment.

The technique solves the governing equation for magnetohydrodynamic (MHD) equilibrium, the Grad-Shafranov equation, to find a plasma state consistent with experimental data. These data typically include signals from magnetic pickup coils and flux loops located outside the plasma, currents in external magnetic field coils, and the total plasma current. The most widely adopted and influential code for this purpose is the Equilibrium Fitting (EFIT) code, developed at General Atomics. The outputs of EFIT, such as the shape of the last closed flux surface (separatrix), the safety factor (q) profile, and pressure profiles, are critical for real-time plasma control, post-shot physics analysis, stability calculations, and transport studies.

Physics / Mechanism

The physical basis for magnetic equilibrium reconstruction is the principle of static MHD equilibrium, where the plasma pressure gradient force is balanced by the Lorentz force ($\vec{J} \times \vec{B} = \nabla p$). For an axisymmetric toroidal system like a tokamak, this force balance equation can be reduced to a single, two-dimensional, nonlinear, second-order partial differential equation known as the Grad-Shafranov equation:

$$ R^2 \nabla \cdot \left( \frac{1}{R^2} \nabla \psi \right) = -\mu_0 R^2 p'(\psi) - F(\psi)F'(\psi) $$

Here, $\psi$ is the poloidal magnetic flux, which is constant on a magnetic flux surface. $R$ is the major radius, $\mu_0$ is the vacuum permeability, $p(\psi)$ is the plasma pressure, and $F(\psi) = R B_t$ where $B_t$ is the toroidal magnetic field. The primes denote differentiation with respect to $\psi$. The left side represents the magnetic field curvature force, while the right side represents forces from the plasma pressure gradient and the poloidal current.

Reconstruction codes like EFIT solve this equation numerically. The core of the problem is that the two free functions, $p'(\psi)$ and $FF'(\psi)$, which define the pressure and poloidal current profiles, are unknown. The reconstruction process is an inverse problem: it seeks to determine these functions by finding the best fit between the calculated magnetic fields and currents at the measurement locations and the actual experimental data. This is typically formulated as a least-squares minimization problem, where the code iteratively adjusts the profile functions to minimize the chi-squared ($\chi^2$) value:

$$ \chi^2 = \sum_{i} \frac{(C_i - M_i)^2}{\sigma_i^2} $$

where $C_i$ are the calculated values (e.g., magnetic field at a probe location) from the Grad-Shafranov solution, $M_i$ are the measured values, and $\sigma_i$ are the measurement uncertainties. The inputs ($M_i$) include data from magnetic diagnostics (probes, flux loops, Rogowski coils), external coil currents, and often internal constraints like Motional Stark Effect (MSE) measurements of the magnetic field pitch angle or Thomson scattering data for the pressure profile. The code outputs a complete 2D map of the poloidal flux $\psi(R, Z)$, from which key plasma parameters like the q-profile, plasma boundary, beta, and internal inductance are derived.

Historical development

The concept of reconstructing plasma equilibrium from external magnetic measurements dates back to the early days of tokamak research. Early methods in the 1970s used simplified models, such as representing the plasma as a single current filament or using multipole moment expansions. These approaches provided basic information about plasma position and shape but lacked internal detail.

The modern era of equilibrium reconstruction began with the development of numerical solvers for the full Grad-Shafranov equation. A significant milestone was the creation of the EFIT code by Lao L. Lao and colleagues at General Atomics in the mid-1980s. The original EFIT paper, published in Nuclear Fusion in 1985, detailed a robust method using a Picard iteration scheme to solve the Grad-Shafranov equation, fitting to external magnetic data. This implementation proved highly successful and was first applied to the Doublet III (D-III) tokamak.

Throughout the 1990s, EFIT's capabilities were expanded. A crucial enhancement was the incorporation of internal measurements, most notably data from the Motional Stark Effect (MSE) diagnostic. MSE provides localized measurements of the magnetic field pitch angle inside the plasma, placing strong constraints on the internal current profile and the safety factor (q) profile. This transformed EFIT from a tool that primarily determined the plasma boundary to one that could accurately resolve the internal magnetic structure, a necessity for advanced tokamak research.

The code was also optimized for speed, enabling its use in real-time plasma control systems. The first real-time EFIT (rtEFIT) systems were deployed on the DIII-D tokamak, allowing for dynamic control of plasma shape, position, and performance based on detailed equilibrium calculations performed between control cycles (typically every 1-10 ms).

Current status

As of 2026, magnetic equilibrium reconstruction is a mature and indispensable tool for all major magnetic confinement experiments. The EFIT code, or a variant thereof, is the de facto standard used at facilities worldwide, including DIII-D, JET, KSTAR, EAST, and MAST-U. It is also the designated equilibrium reconstruction code for the ITER project.

The state of the art includes several key features:

• Real-Time Control: rtEFIT is routinely used for feedback control of critical plasma parameters. This includes precise control of the plasma boundary shape to optimize divertor performance, management of the plasma-wall gap, and control of stability parameters like the normalized beta ($\beta_N$).
• Kinetic EFIT: For high-performance plasmas, where the pressure can be anisotropic and significant fast ion populations exist, standard MHD is insufficient. Kinetic EFIT incorporates pressure data from multiple diagnostics (e.g., Thomson scattering for thermal electrons, charge exchange recombination spectroscopy for thermal ions, and neutral particle analyzers for fast ions) to construct a more accurate, non-thermal pressure profile. This is essential for accurately assessing the stability of plasmas approaching the Lawson criterion.
• Uncertainty Quantification (UQ): There is a growing focus on quantifying the uncertainties in reconstructed parameters. Techniques like Bayesian inference are being applied to provide error bars on outputs like the q-profile and the location of the separatrix, which is critical for validating physics models and designing future devices.
• Integration with other models: EFIT is a core component of larger computational workflows. Its outputs are used as inputs for complex stability codes (like DCON and GATO), transport models (like TRANSP), and heating and current drive simulations.

Notable implementations

• General Atomics (DIII-D): The original developer and a leading user of EFIT. The DIII-D National Fusion Facility has pioneered many of its advanced applications, including the first use of rtEFIT for plasma control and the development of kinetic EFIT. Their work continues to push the code's capabilities.
• ITER Organization: The ITER project has adopted EFIT as its primary tool for equilibrium reconstruction and plasma control. Significant effort is underway to adapt and validate the code for the unique scale and operating conditions of ITER, including its metallic wall and long-pulse operation. The ITER Plasma Control System will rely heavily on rtEFIT for basic machine protection and advanced performance optimization.
• JET (UKAEA/EUROfusion): The Joint European Torus has used EFIT for decades. The large database of JET discharges analyzed with EFIT has been instrumental in developing the physics basis for ITER, particularly in studies of H-mode, ELMs, and plasma disruptions.
• NSTX-U (PPPL): On spherical tokamaks like the National Spherical Torus Experiment Upgrade, where the geometry is more extreme (low aspect ratio), equilibrium reconstruction is particularly challenging. Specialized versions of EFIT and other codes like LRDFIT have been developed to handle these configurations.
• Private Fusion Companies: Several private companies, such as Commonwealth Fusion Systems and Tokamak Energy, utilize EFIT or similar in-house codes for the design and operation of their high-field compact tokamaks. Accurate reconstruction is vital for achieving and diagnosing the high-pressure, high-performance regimes these devices target.

Open challenges

Despite its success, magnetic equilibrium reconstruction faces several ongoing challenges:

• 3D Effects: The standard Grad-Shafranov model is strictly 2D (axisymmetric). Real tokamaks have non-axisymmetric features from toroidal field ripple, error fields, and MHD instabilities like Edge Localized Modes (ELMs) or tearing modes. Reconstructing these 3D structures requires more complex 3D equilibrium codes (e.g., V3FIT, STELLOPT), which are computationally far more expensive and less mature than EFIT.
• Fast Transients: During rapid events like disruptions or ELM crashes, the plasma is not in a static equilibrium. While EFIT can provide a snapshot, it does not capture the dynamics. Reconstructing the plasma state during these transients requires time-dependent models and is an active area of research.
• Stellarator Equilibrium: While the principles are similar, reconstruction in non-axisymmetric stellarators is inherently a 3D problem. Codes like V3FIT are used, but the computational cost and the increased number of required measurements make it significantly more difficult than in tokamaks.
• Burning Plasmas: In a burning plasma like ITER's, a large and spatially distributed population of alpha particles will contribute significantly to the plasma pressure. Accurately modeling this kinetic component and its effect on the equilibrium in real-time will be a challenge for kinetic EFIT implementations.
• Diagnostic Integration: Integrating a diverse set of diagnostics, each with its own uncertainties and spatial/temporal resolution, into a single consistent equilibrium remains a complex task. Robustly handling conflicting data or diagnostic failures, especially in a real-time control context, is a critical engineering problem.

Outlook

The 5-15 year trajectory for magnetic equilibrium reconstruction will be driven by the needs of next-generation devices like ITER and fusion power plants. The primary focus will be on increasing the fidelity, robustness, and speed of the calculations.

We can expect significant advances in real-time processing. The use of GPU acceleration and machine learning (ML) surrogates is poised to dramatically reduce computation time. ML models, trained on extensive databases of EFIT runs, could provide near-instantaneous equilibrium solutions, enabling more sophisticated control algorithms. These ML-based approaches are already being tested, with promising results for accelerating specific components of the reconstruction workflow.

For physics analysis, the integration of more complex physics will continue. Fully 3D and time-dependent reconstructions will become more common as computational power increases, allowing for routine analysis of MHD instabilities and other non-axisymmetric phenomena. The development of integrated modeling suites, where equilibrium reconstruction is tightly coupled with transport, stability, and heating models, will provide a more holistic and self-consistent picture of the plasma state.

For ITER and future power plants, reliability will be paramount. This will drive further development in uncertainty quantification and automated diagnostic validation to ensure the control system is always acting on the most accurate possible picture of the plasma. Ultimately, EFIT and its successors will evolve from a diagnostic tool into the core of the 'brain' of a fusion power plant, essential for safe, stable, and optimized operation.

References

1. Theory and operation of the DIII-D tokamak — General Atomics (2005)
2. Reconstruction of current profile parameters and plasma shapes in tokamaks — Nuclear Fusion (1985)
3. Real time analysis of the DIII-D tokamak plasma — Review of Scientific Instruments (1991)
4. Plasma equilibrium and control in the ITER tokamak — Plasma Physics and Controlled Fusion (2015)
5. Magnetic equilibrium reconstruction in the presence of ferritic steel — Nuclear Fusion (2004)
6. Kinetic equilibrium reconstruction of DIII-D discharges — Nuclear Fusion (2000)
7. Bayesian equilibrium reconstruction at JET — Nuclear Fusion (2019)
8. Use of the EFIT code for routine equilibrium analysis of TEXTOR discharges — Fusion Engineering and Design (2001)