Magnetized Liner Inertial Fusion (MagLIF)

• MagLIF is a fusion concept that uses a cylindrical metallic liner to compress preheated, premagnetized deuterium or deuterium-tritium fuel.
• It combines inertial compression, axial magnetic insulation, and targeted fuel preheat to achieve fusion-relevant conditions with lower implosion velocities and stagnation pressures than traditional laser-driven ICF.
• Recent studies underline robust scaling laws, improved instability controls, and advanced diagnostics, paving the way for reactor-oriented designs using MagLIF.

Magnetized Liner Inertial Fusion (MagLIF) is a magneto-inertial fusion concept in which a cylindrical metallic liner implodes inward around preheated, premagnetized fusion fuel under pulsed-power drive. Its defining rationale is the simultaneous use of inertial compression, axial magnetic insulation, and fuel preheat so that fusion-relevant conditions can be approached at lower implosion velocity, lower stagnation pressure, and lower convergence than conventional laser-driven inertial confinement fusion, while retaining a direct electromagnetic driver architecture (McBride et al., 2015, Ruiz et al., 2022, Alexander et al., 14 Apr 2025).

1. Core configuration and operating sequence

The standard MagLIF configuration combines a cylindrical metal liner, a deuterium or deuterium-tritium fuel fill, axial premagnetization, and laser preheat immediately before radial implosion. In the Sandia Z-machine implementation summarized in the field-injection study, the implosion is driven in a Z-pinch configuration using about $27~\mathrm{MA}$ of electrical current with a rise time of about $100~\mathrm{ns}$, and the original concept requires a thick liner, an axial magnetic field, and fuel preheating just before implosion (Gourdain et al., 2017). In a simplified 1D Kraken representation used for data-analysis studies, the sequence is stated as axial premagnetization by an approximately $10~\mathrm{T}$ field, multi-kilojoule laser preheat through a top-side window, and magnetic compression by a current pulse rising to about $20~\mathrm{MA}$ in about $100~\mathrm{ns}$ (Joseph et al., 2023).

The liner is deliberately thick because thick liners suppress magnetic Rayleigh–Taylor instabilities, but that same thickness complicates magnetic-field penetration and raises transport and stability questions. The attraction of the concept is that the magnetic field reduces electron thermal conduction to the liner wall, the preheat raises the fuel adiabat before compression, and the liner then compresses the magnetized plasma quasi-adiabatically (Gourdain et al., 2017, Alexander et al., 14 Apr 2025).

Several scaling studies and reactor-oriented assessments emphasize that MagLIF operates at much lower implosion velocity than laser ICF. One recent design study states liner velocities of about $v\sim 100~\mathrm{km/s}$, compared with $v\gtrsim 350~\mathrm{km/s}$ for laser-driven inertial fusion, with implosion durations on the order of $\gtrsim 100~\mathrm{ns}$. The same study states that MagLIF can reach ignition at stagnation pressures around $\sim 10~\mathrm{Gbar}$, versus roughly $\sim 100~\mathrm{Gbar}$ for laser ICF (Alexander et al., 14 Apr 2025). This operating point makes magnetic insulation and end-loss control central rather than auxiliary.

2. Fuel energetics, preheat topology, and magnetized transport

A useful reduced description of MagLIF fuel energetics writes the DT internal-energy evolution per unit length as

$100~\mathrm{ns}$0

with alpha heating competing against compressional work, radiative loss, electron and ion conduction, and axial end losses (Alexander et al., 14 Apr 2025). In this formulation, self-heating occurs when the alpha term exceeds the losses. The same source identifies $100~\mathrm{ns}$1 as the key magnetic confinement metric for $100~\mathrm{ns}$2 alpha particles and states that magnetization becomes very effective once $100~\mathrm{ns}$3 (Alexander et al., 14 Apr 2025).

Semi-analytic MagLIF modeling showed early that the geometry of preheat matters as much as the total preheat energy. If only a central fraction of the fuel is preheated, the fuel evolves into a hot spot plus a cold dense shelf. The hot spot is lower density and therefore can reach higher temperature for a given deposited energy, while the shelf buffers the hot region from the liner and reduces radiative, conductive, and magnetic-flux losses (McBride et al., 2015, McBride et al., 2015). Those studies state explicitly that radiative loss rates depend strongly on the radial fraction of the fuel that is preheated, and that preheating only a central portion is often better than preheating the entire fuel uniformly (McBride et al., 2015, McBride et al., 2015).

This hot-spot/shelf structure is tied to MagLIF’s Lawson-like metrics. In the semi-analytic exploration of DT MagLIF, self-heating begins when

$100~\mathrm{ns}$4

with the threshold lowered relative to conventional ICF because the embedded axial field improves alpha confinement and deposition (McBride et al., 2015). End losses remain an intrinsic design constraint because MagLIF uses an open-ended cylindrical geometry rather than a closed spherical capsule (McBride et al., 2015, Ruiz et al., 2022).

3. Axial magnetic-field generation and flux transport

The axial field is indispensable but technically awkward. The field-injection analysis states that MagLIF requires about $100~\mathrm{ns}$5 initially to reduce heat losses to the liner wall, yet the field must penetrate both the pulsed-power hardware and the liner itself before implosion, implying rise times of tens of microseconds for conventional external coils (Gourdain et al., 2017). Such coils are bulky, occupy space near the load, increase inductance, and reduce the current that the fixed-voltage driver can deliver.

The proposed alternative is to use tilted return-current posts with minimal helicity. Tilting the return posts gives the return current an azimuthal component, which generates an axial magnetic field; the post angle then controls the field strength. The paper’s practical conclusion is that the external $100~\mathrm{ns}$6 coil can be eliminated and replaced by return current posts with minimal helicity if current penetration through the thick liner is faster than naive skin-depth estimates suggest (Gourdain et al., 2017).

The key physics is that at sufficiently large current density, ohmic heating generates temperature and resistivity gradients, and current penetration becomes advective rather than purely diffusive. The paper identifies an effective velocity

$100~\mathrm{ns}$7

so that current is driven into lower-resistivity regions rather than only diffusing according to the usual skin effect (Gourdain et al., 2017). For a $100~\mathrm{ns}$8 aluminum liner, the 1D model shows full penetration in about $100~\mathrm{ns}$9, and 3D PERSEUS simulations show fields larger than $10~\mathrm{T}$0 diffusing across the liner wall in less than $10~\mathrm{T}$1. In those simulations, the axial magnetic field reaches the axis about $10~\mathrm{T}$2 into the current ramp and can reach as high as $10~\mathrm{T}$3 by about $10~\mathrm{T}$4 because of compression by ablated plasma (Gourdain et al., 2017). A common misconception is therefore that thick MagLIF liners categorically prevent rapid premagnetization on implosion timescales; the resistivity-gradient result shows that this conclusion is not generally valid.

Flux evolution during preheat adds a second transport layer. A later theoretical analysis of preheat propagation shows that when the acoustic timescale is much shorter than the conductive timescale, preheat drives a hot inner core, a dense outer shelf, a leading shock, and a magnetic boundary layer localized near the core/shelf interface. In integrated $10~\mathrm{T}$5 and scaled $10~\mathrm{T}$6 FLASH simulations discussed there, roughly $10~\mathrm{T}$7 of the axial flux is lost when the expanding fuel meets the liner, and turning Nernst off during preheat underestimates total flux loss by about $10~\mathrm{T}$8 (Garcia-Rubio et al., 15 Apr 2025).

4. Similarity scaling, current scaling, and rise-time scaling

Because the MagLIF design space is highly multidimensional, one major line of work reduces the problem to dimensionless control parameters. The theoretical framework identifies, among others, the implosion parameter $10~\mathrm{T}$9, the preheat parameter $20~\mathrm{MA}$0, the stability parameter $20~\mathrm{MA}$1, and loss measures $20~\mathrm{MA}$2, $20~\mathrm{MA}$3, and $20~\mathrm{MA}$4, then argues that MagLIF loads can be “incompletely” similarity scaled: many, but not all, dimensionless quantities can be preserved simultaneously (Ruiz et al., 2022). This is not exact similarity; it is a controlled approximation meant to conserve the dominant no-$20~\mathrm{MA}$5 physics while keeping neglected parameters non-essential.

For current scaling at fixed drive timescale, the 20-MA baseline fits reported in the HYDRA study are

$20~\mathrm{MA}$6

$20~\mathrm{MA}$7

$20~\mathrm{MA}$8

For the $20~\mathrm{MA}$9 example, the same paper reports approximately $100~\mathrm{ns}$0, $100~\mathrm{ns}$1, $100~\mathrm{ns}$2, and $100~\mathrm{ns}$3 (Ruiz et al., 2022). In the no-$100~\mathrm{ns}$4 regime, the yield scaling is

$100~\mathrm{ns}$5

and the same study states that at $100~\mathrm{ns}$6 the no-$100~\mathrm{ns}$7 yield reaches the multi-megajoule scale, while alpha-heated simulations suggest roughly $100~\mathrm{ns}$8 (Ruiz et al., 2022).

Rise-time scaling modifies the trade space in a different way. When the current-rise time or implosion timescale is increased while preserving similarity, longer-rise-time MagLIF requires a larger-radius liner, lower initial gas density, weaker axial field, longer liner height, and more total preheat energy (Ruiz et al., 2022). The rise-time study reports the numerical fits

$100~\mathrm{ns}$9

together with the weak load-voltage scaling

$v\sim 100~\mathrm{km/s}$0

rather than the ideal $v\sim 100~\mathrm{km/s}$1, because preserving end losses forces the imploding height to increase (Ruiz et al., 2022). The consequence is that longer-rise-time pulsed-power generators are not automatically easier for MagLIF, even if they are simpler electrically.

5. Instabilities, mix, and stagnation morphology

MagLIF is unusually sensitive to coupled instability and transport effects because liner integrity, ablative interfaces, magnetized conduction, and open-end losses all interact near stagnation. A transport analysis of a MagLIF-like plasma with a hot magnetized fuel region adjacent to a cold dense liner shows that increasing the liner atomic number $v\sim 100~\mathrm{km/s}$2 reduces both energy and magnetic-flux losses in the fuel for small and moderate magnetization values. In the same model, mass diffusion is confined within a thin layer at the ablated border, concentration-gradient diffusion and baro-diffusion are the predominant mechanisms, and mass ablation, energy loss, magnetic-flux loss, and liner–fuel mass diffusion scale as $v\sim 100~\mathrm{km/s}$3 for large $v\sim 100~\mathrm{km/s}$4 (García-Rubio et al., 2018). The paper also notes that the width of the diffusion layer may be comparable to the turbulent mixing layer resulting from Rayleigh–Taylor instability at the ablated border (García-Rubio et al., 2018).

Observed stagnation morphology remains an active interpretive problem rather than a settled diagnostic. Self-emission x-ray images show double-helical or bifurcated helical structures in the stagnated plasma, while radiographs show corresponding helical structure in the imploding liner. The underlying physics linking the liner’s double helix to the plasma’s bifurcated double helices is stated to be not yet known; the hypotheses discussed include helical magnetic Rayleigh–Taylor feed-through and Taylor relaxation or self-organization driven by conserved magnetic and cross helicities (Glinsky et al., 2019). The point is not merely descriptive: morphology encodes instability, mix, and possibly nonlinear self-organization, but quantitative comparison requires a metric rather than visual inspection.

A more recent extension adds anisotropic magnetized viscosity to MagLIF simulations. That work presents the first implementation of the full Braginskii magnetized viscosity tensor for arbitrary magnetic-field orientation in the Pacific Fusion branch of FLASH, then shows in MagLIF-relevant calculations that magnetized viscosity damps vortical structures, converts kinetic energy in those vortical structures into thermal energy, mitigates Rayleigh–Taylor instabilities, and preserves yield in seeded-perturbation simulations. In the $v\sim 100~\mathrm{km/s}$5 perturbation case, the largest improvement reported is about $v\sim 100~\mathrm{km/s}$6 yield preservation (Sam et al., 22 Apr 2026). This suggests that some instability saturation pathways in magnetized liners are transport-limited rather than purely hydrodynamic.

6. Diagnostics, data analysis, and code validation

MagLIF diagnostics are heterogeneous enough that comparison methodology has become a research subject in its own right. One experimental-data-driven framework for stagnation imaging uses a database of $v\sim 100~\mathrm{km/s}$7 image plate scans from $v\sim 100~\mathrm{km/s}$8 different experiments and studies how signal-to-noise ratio, registration uncertainty, and instrument resolution alter image metrics in a chosen feature space. Using the Mallat Scattering Transform (MST), that work shows that real-space denoising is more robust than off-strand subtraction, that small registration errors affect the metric approximately linearly, and that resolution differences matter less than true morphology differences in this dataset (Lewis et al., 2023).

The broader MST program addresses the absence of a systematic quantitative morphology metric for MagLIF stagnation images. In synthetic-class tests, the MST representation achieved misclassification probability below $v\sim 100~\mathrm{km/s}$9, enabling experiment-to-experiment and simulation-to-experiment comparisons together with uncertainty-aware parameter inference for helical morphology (Glinsky et al., 2019). This does not resolve the origin of the helical structures, but it provides a common coordinate space in which those structures can be compared.

MagLIF has also become a testbed for statistical methods aimed at hidden-variable discovery. A Bayesian/Gaussian-process framework was applied to a 1D Kraken-generated MagLIF dataset in which neutron yield on a log scale was measured against deposited laser energy while a hidden degradation and tune-up structure was imposed on the true laser energy. In that controlled setting, the method correctly identifies sudden tune-up-related shifts, reconstructs the underlying energy–yield relationship closely, but “does not provide much insights into the degradation effect on the laser energy” (Joseph et al., 2023). This reflects a real experimental concern: MagLIF campaigns are expensive, low-shot-rate, and subject to substantial run-to-run variability.

Predictive design increasingly relies on validated radiation-MHD tools rather than only reduced models. A 2025 validation study reports six FLASH benchmarks ranging from single-mode and multi-mode liner instabilities to a fully integrated MagLIF shot and a 60-MA scaling study. For shot z2977, FLASH predicts a yield of $v\gtrsim 350~\mathrm{km/s}$0 neutrons against an experimental value of about $v\gtrsim 350~\mathrm{km/s}$1, with a burn-weighted ion temperature of $v\gtrsim 350~\mathrm{km/s}$2 compared with the experimental $v\gtrsim 350~\mathrm{km/s}$3, and the paper concludes that the latest FLASH capability obtains good agreement with experimental data, theoretical results, and leading design-code results across all six benchmarks (Ellison et al., 14 Apr 2025).

7. Prospective extensions and reactor-oriented studies

Several recent studies extend MagLIF beyond its baseline formulation by exploiting the fact that stagnated MagLIF plasmas are magnetized, collisional, and geometrically cylindrical. One example is favorable collisional demixing of ash and fuel. For alpha ash in an equal D–T fuel mix, the stationary relation becomes

$v\gtrsim 350~\mathrm{km/s}$4

which implies ash concentration in the dense, cool edge and fuel enrichment in the hot core. In the representative MagLIF-like case with $v\gtrsim 350~\mathrm{km/s}$5, $v\gtrsim 350~\mathrm{km/s}$6, $v\gtrsim 350~\mathrm{km/s}$7, $v\gtrsim 350~\mathrm{km/s}$8, and $v\gtrsim 350~\mathrm{km/s}$9, the burn fraction changes from $\gtrsim 100~\mathrm{ns}$0 to $\gtrsim 100~\mathrm{ns}$1, corresponding to $\gtrsim 100~\mathrm{ns}$2 increase in neutron yield. The same paper states that enhancements can saturate near $\gtrsim 100~\mathrm{ns}$3 and that around $\gtrsim 100~\mathrm{ns}$4 may be accessible in more aggressive high-field cases (Ochs et al., 2018).

Another extension adapts fast ignition to MagLIF. Because the hotspot is cylindrical rather than spherical and the axial field suppresses transverse conduction, the required ignitor can be longer-pulse and lower-energy than in conventional laser ICF. The long-pulse fast-ignition model reports ignition with a $\gtrsim 100~\mathrm{ns}$5 pulse and, in one case, $\gtrsim 100~\mathrm{ns}$6. It also states that seed fields of $\gtrsim 100~\mathrm{ns}$7 can be compressed to kiloteslas in experiments and even tens of kT in simulations, so that the same field that insulates the fuel may also collimate ignitor electrons (Wang et al., 13 Feb 2026).

Reactor-oriented scaling studies push these ideas to higher-current drivers. One pulser-driven inertial-fusion assessment states that scaling from $\gtrsim 100~\mathrm{ns}$8 to $\gtrsim 100~\mathrm{ns}$9–$\sim 10~\mathrm{Gbar}$0 enables net facility gain, introduces a Demonstration System designed to deliver more than $\sim 10~\mathrm{Gbar}$1 and store approximately $\sim 10~\mathrm{Gbar}$2, and reports modeled load delivery of roughly $\sim 10~\mathrm{Gbar}$3, $\sim 10~\mathrm{Gbar}$4, and about $\sim 10~\mathrm{Gbar}$5 to the load region. In example $\sim 10~\mathrm{Gbar}$6 targets with $\sim 10~\mathrm{Gbar}$7 height, $\sim 10~\mathrm{Gbar}$8 DT ice liner, $\sim 10~\mathrm{Gbar}$9, and $\sim 100~\mathrm{Gbar}$0 preheat, the simulated outcomes are $\sim 100~\mathrm{Gbar}$1 and $\sim 100~\mathrm{Gbar}$2 for a beryllium liner and $\sim 100~\mathrm{Gbar}$3 and $\sim 100~\mathrm{Gbar}$4 for an aluminum liner (Alexander et al., 14 Apr 2025). These are projected regimes rather than present experimental performance, but they show how MagLIF’s core architecture—thick liner, axial field, preheat, and direct pulsed-power drive—has become the basis for ignition-scale and facility-gain design studies.