Z machine: the world's largest pulsed X-ray machine

2020-09-22T00:00:00+00:00

My sick, unbridled hyperfocus on the Z machine continues. Writing it down so I don't forget.

The equation I use as a signature is $\partial_\mu F^{\mu\nu} = j^\nu$, the compact form of Maxwell's equations with a source: the current $j$ tells the field $F$ how to arrange itself. The Z machine, at Sandia National Laboratories, is that equation pushed to the limit. A huge axial current runs through a cylindrical target, creates a magnetic field around it, and the Lorentz force per unit volume, $\mathbf{J} \times \mathbf{B}$, points inward. The target is crushed by the very current flowing through it. That is called a Z-pinch, and it is where the machine gets its name.

I put this research together in March 2026, as part of a survey of physics and tools. In October I checked every number against its source and every repository against the GitHub API. Several were wrong, and the corrections are in the text.

One shot, in numbers

According to Sandia's page about Z, the machine fires about 200 shots a year:

Quantity	Value
Current	about 26 million amperes
Peak X-ray emission	350 terawatts
X-ray energy	2.7 megajoules
Peak electrical power	about 80 terawatts
Pulse length	about 100 nanoseconds

The last two come from Sandia's energy page, which compares the 80 terawatts with "all the power plants in the world combined, about 15 terawatts". That comparison does not hold for electricity. According to Ember's report, low-carbon generation was 40.9% of the 2024 total and came to 12,609 TWh, which gives about 30,800 TWh for the year. Divided by the 8,760 hours in a year, the average is about 3.5 terawatts. Whatever the baseline, for those 100 nanoseconds Z exceeds the entire electricity generation of the planet: about 5 times against Sandia's 15 terawatts, more than 20 times against the real average.

The five fronts

Stockpile stewardship

This is the main mission. The United States has not run a real nuclear test since 1992, and Z produces pressures and temperatures similar to those of a detonation, in a small sample, to validate the physics models used in simulation. In Sandia's 2026 accomplishments report, the plutonium team describes reduced-scale targets to compress plutonium to unprecedented pressures, new diagnostics to measure temperature during compression, and more plutonium mass allowed per shot.

Fusion with MagLIF

MagLIF stands for magnetized liner inertial fusion. A beryllium tube about 5 millimetres across, with 7.5 millimetres of useful height, is filled with deuterium, given an axial magnetic field of 10 tesla, and hit by a laser pulse that preheats the gas. Then Z drives about 19 million amperes through it in roughly 100 nanoseconds, and the tube implodes at about 70 km/s.

According to the review by Yager-Elorriaga et al., in Nuclear Fusion in 2022, the first integrated experiments, in 2013, reached 2×10¹² DD neutrons, and the best result since then was 1.1×10¹³. The same review says 2D simulations point to yields of several megajoules on a 60 MA machine, and gain high enough for fusion energy above 65 MA. These are simulation projections for a machine that does not exist yet.

Materials under extreme conditions

This is the most versatile front: measuring how a material responds to very high pressure and temperature, which is called an equation of state. In 2007, Sandia announced that Z had frozen water in nanoseconds, compressing thin samples to tens of thousands of atmospheres until they became ice VII, an ice hotter than the boiling point of water. In 2008, Knudson et al. used shocks on Z to map the high-pressure phases of carbon, including the melting of diamond.

Fundamental science and astrophysics

Most of the matter in the universe is in conditions that on Earth only appear in an experiment like this. One example is the white dwarf photosphere that Falcon et al. recreated on Z, heating hydrogen with the machine's X-rays to test the models used to estimate the mass of these stars. The Z Fundamental Science Program opens the machine to outside researchers. Each yearly call hands out shots over two years: it used to be 14 per call, and in the 2025 and 2026 calls it went up to about 21.

Fusion companies

In March 2024, First Light Fusion became the first private company to fire a shot on Z. In July 2024, Fuse signed a research agreement with Sandia and said it plans to build larger commercial versions of Z. In October 2025, Pacific Fusion fired four shots with simplified targets, aluminium bonded to plastic, and measured how the magnetic field diffuses through the target. The result came out in February 2026.

What comes next

In a January 2020 presentation, Daniel Sinars, then director of Sandia's pulsed power center, said the lab had proposed to the NNSA a machine of more than 60 MA, which could be ready around 2032 or later. That is the range where, according to the MagLIF review, yields would reach megajoules. I found no public source on whether that machine was approved.

Open repositories

A caveat before the tables: stars here are much lower than in web and AI. A high energy density physics code with 300 stars can have more scientific impact than a JavaScript library with 100,000. What counts is publication, validation against experiment, and use in national laboratories.

The stars below were read from the GitHub API on 13 October 2026. The first version of this list had approximate numbers, and several were inflated: yt showed 4.6 thousand and has 559, ParaView 4.2 thousand and has 1,712, WarpX 2.8 thousand and has 495. Two repositories changed owner: WarpX moved to BLAST-WarpX and EPOCH to epochpic.

Particle-in-cell (PIC)

Repository	Stars	What it does
ComputationalRadiationPhysics/picongpu	795	Relativistic PIC, portable across architectures, built for supercomputers
BLAST-WarpX/warpx	495	Mesh-refined PIC on top of AMReX; won the 2022 Gordon Bell Prize
SmileiPIC/Smilei	450	Relativistic electromagnetic PIC, used for laser-plasma and astrophysics
epochpic/epoch	259	Warwick's PIC for laser-plasma interaction
fbpic/fbpic	212	Spectral quasi-3D PIC in cylindrical geometry, CPU and GPU
osiris-code/osiris	85	Relativistic PIC from UCLA and IST Lisbon, opened in January 2023 under AGPL
vonkarmaninstitute/pantera-pic-dsmc	37	PIC with DSMC for non-equilibrium gas and plasma

Magnetohydrodynamics and radiation hydrodynamics

Repository	Stars	What it does
DedalusProject/dedalus	712	Spectral framework in Python for partial differential equations
PrincetonUniversity/athena	371	Athena++, MHD with radiation and general relativity, with mesh refinement
AMReX-Astro/Castro	343	Compressible hydrodynamics with radiation and MHD, on AMReX
amrvac/amrvac	104	MPI-AMRVAC, parallel mesh refinement for hydro and MHD
zenitani/OpenMHD	69	A collection of MHD codes in Fortran, good for study
PlasmaControl/FreeMHD	26	Free-surface liquid metal MHD, on OpenFOAM
lanl/HARD	21	Los Alamos radiation hydrodynamics for astrophysics, from 2024

The open code closest to a Z experiment is FLASH, from the Flash Center in Rochester, with a high energy density physics version. Its successor, Flash-X, has a private repository: access is requested by email with your GitHub username, because the funder requires a list of who downloads the code.

Equation of state

Repository	Stars	What it does
feos-org/feos	201	Equations of state in Rust with Python bindings, more for thermodynamics than for high pressure
lanl/singularity-eos	40	Equations of state portable across CPU and GPU, from Los Alamos
lanl/eospac	10	Routines to read SESAME equation-of-state tables

Particle transport

Repository	Stars	What it does
openmc-dev/openmc	1,124	Monte Carlo neutron and photon transport
Geant4/geant4	865	Passage of particles through matter, from CERN

Molecular dynamics and electronic structure

Repository	Stars	What it does
lammps/lammps	3,094	Massively parallel molecular dynamics, from Sandia itself
gromacs/gromacs	968	High-performance molecular dynamics; development happens on GitLab
QEF/q-e	701	Quantum ESPRESSO, electronic structure with DFT; a GitLab mirror

Computing infrastructure

Repository	Stars	What it does
kokkos/kokkos	2,700	C++ programming model portable across CPU and GPU
trilinos/Trilinos	1,438	Linear algebra and solvers, from Sandia
AMReX-Codes/amrex	818	Block-structured mesh refinement, the base of WarpX and Castro
AMReX-Astro/Nyx	105	Cosmological hydro with N-body, on AMReX

Visualization and analysis

Repository	Stars	What it does
Kitware/ParaView	1,712	Scientific visualization on VTK, for large simulation data
yt-project/yt	559	Analysis and visualization of volumetric data in Python
visit-dav/visit	534	Visualization of mesh data, from LLNL

Plasma and magnetic fusion

Repository	Stars	What it does
PlasmaPy/PlasmaPy	717	Python package for plasma physics research and teaching
freegs-plasma/freegs	145	Free-boundary Grad-Shafranov solver, tokamak equilibrium
iterorganization/IMAS-Python	73	Python access to the IMAS data model, from ITER
iterorganization/SOLPS-ITER	66	Plasma edge simulation
tokamak-energy/gsfit	39	Grad-Shafranov equilibrium reconstruction in Rust and Python, from Tokamak Energy
iterorganization/DINA-IMAS	14	The DINA plasma simulator

ITER opened IMAS on GitHub under open licences; SOLPS-ITER and DINA arrived in November 2025.

Lists and organizations

Repository	Stars	What it does
paulromano/awesome-nuclear	508	Curated list of open software for nuclear science and engineering
nikosarcevic/HEP-ASTRO-COSMO	157	Packages for high energy physics, astroparticles and cosmology
pmocz/awesome-astrophysical-simulation-codes	125	List of astrophysical simulation codes

Two organizations are worth a full visit: fusion-energy, with 45 repositories of fusion software, and the UCLA Plasma Simulation Group, with QuickPIC, UPIC and teaching codes.

Brenner Cruvinel - física