A team at the ELI-NP facility in Romania passed a laser-generated muon beam through a 2-meter-thick reinforced concrete wall and captured the shadow of a lead object placed beyond it. Research on generating muons with lasers was already under way, but this experiment goes further: it distinguished the muon signal in an environment mixed with other radiation and imaged a real object.
In a preprint published on September 23, 2026, M. Dobre, P. Ghenuche and colleagues state that, to the best of their knowledge, this is the first demonstration of imaging in which artificially generated, laser-driven muons dominate the signal.
Although the work has not yet been peer reviewed, it is a concrete step away from conventional muography, which waits for cosmic-ray muons to rain down from nature, and toward a method that fires artificial muons in a chosen direction.
Making muons with a laser and sending them through a thick wall
ELI-NP used a 10 PW-class high-power laser system. In this experiment, about 230 J of energy was concentrated into an extremely short pulse of 23 fs. The system can produce up to one laser pulse per minute.
The laser light itself, however, does not pass through the 2 m concrete wall.
First, an intense laser pulse is fired into a gas to create a plasma, and the strong electric field generated in its wake accelerates electrons. This technique is called laser wakefield acceleration (LWFA).
Next, the electrons, accelerated to high energy, are directed into lead. Bremsstrahlung produces high-energy gamma rays, and the interaction of those gamma rays with atomic nuclei creates muon–antimuon pairs. This production mechanism is known as the Bethe–Heitler process.
Muons are elementary particles similar to electrons, but far heavier, and they lose less energy to radiation as they pass through matter. Muons with sufficient energy can therefore travel through thick rock, metal and other materials.
For naturally occurring cosmic-ray muons, "muography" is already used to probe the interior of objects based on how much passes through and how it scatters.
However, cosmic-ray muons arrive in limited numbers and mainly from overhead. If an artificial muon source can be made practical, it could irradiate from a needed direction at a needed time, potentially changing the imaging time required and the freedom in how equipment is arranged.
On the other hand, muons are not the only thing produced when high-energy electrons strike lead. Large quantities of gamma rays, neutrons and other radiation are also generated and contaminate the detector signal.
In this experiment, a lead muon-production target 40 cm thick along the beam direction was placed first, followed by shielding made of 100 cm of polyethylene, 60 cm of paraffin and other materials. Beyond that stood the 2 m reinforced concrete wall.
This arrangement is designed to let as many muons through as possible while reducing gamma rays, neutrons and other radiation.
In other words, the wall is both an obstacle demonstrating the muons' penetrating power and part of the shielding that cuts unwanted radiation. What was imaged was not the interior of the wall itself, but a lead object placed on the far side of it.
Measuring the "shadow" of lead behind a 2 m wall
The imaging target was a lead object measuring 20 × 25 × 100 cm. It was placed 23 m from the muon source, about 30 cm in front of a detector called μ36 installed in a vehicle.
The team fired the laser 30 times in succession and compared detector signals with and without the lead.
The μ36 has an array of scintillator bars 2.5 cm wide. When a particle passes through, the scintillator emits light, which is read out by electronic circuitry.
The quantity used to build the image was the time during which each scintillator's signal exceeded a threshold. At the position of the lead, the detected signal decreased over a range of 9 to 10 scintillator bars, a width of 22.5 to 25 cm.
This corresponds to the 25 cm width of the lead target.
What was obtained is a transmission image, a "shadow" produced by the lead blocking muons. It is not a three-dimensional reconstruction of the object's interior, nor does it identify the type of material.
What was actually measured here is the transmission image of the lead. How much of the detected signal consists of muons, and what energies of muons arrived, were evaluated by comparison with simulations.
Dividing Sections IV–VI and VIII of Dobre et al.'s paper into what was measured directly by the detector and what was derived from GEANT4 simulation gives the following.
| Item checked | Conditions and result | Basis |
|---|---|---|
| Transmission image of lead | 23 m from the source, 30 shots. Signal decreased over 22.5–25 cm | Measured by the detector |
| Muon contribution to detected signal | 20 m from the source, after shielding. 86.91% of the energy deposited in the detector | Simulation using the measured electron spectrum |
| Energy of muons reaching the detector | About 1.5 GeV or more at production is needed to pass through the apparatus and reach the detector | Comparison of full-apparatus simulation with observations |
The "86.91%" here is not the proportion of muons among the particles that reached the detector by number.
In the simulation, more gamma rays and neutrons than muons reached the detector by count. Even so, because each muon deposits much more energy in the detector, muons accounted for 86.91% of the energy in the detected signal.
This figure was also obtained for a 20 m configuration, which differs from the 23 m configuration used to image the lead. The team did not directly count particles during imaging and measure a "muon purity of 86.91%."
A key basis for judging that the image was produced by muons is a controlled comparison in simulation.
Using the electron spectrum measured during the experiment, the team reproduced the detector response with muons included and with muons removed, leaving only gamma rays and neutrons. Without muons, the simulation could not reproduce the lead shadow observed in the experiment.
The important point supporting this muon imaging is that the team did not merely show that the detector produced a signal, but tested whether other radiation alone could explain the same result.
The beam spread was also measured. The full width at half maximum was 4.51 m at 20 m from the source and 6.35 m at 29 m, close to the simulated values of 4.43 m and 6.27 m.
Beam measurements were also made at the farthest point, 42 m. This does not mean an object 42 m away was imaged, however. The ability of muons to travel far and the ability to image small objects at high resolution are separate performance characteristics.
A 23 fs laser does not mean the image is completed in 23 fs
In this imaging, signals were accumulated over 30 laser shots.
Because the laser's maximum repetition rate is one shot per minute, a simple calculation puts 30 shots at about 30 minutes. This is only a rough figure obtained by dividing 30 shots by one shot per minute, and does not represent the total experiment time, including setup and comparison measurements without the lead.
Using an extremely short laser pulse of 23 fs is entirely different from the image being completed in 23 fs.
The possibility of shortening imaging time with an artificial muon source has long been examined through numerical simulation.
In research published by Luke Calvin and colleagues in Frontiers in Physics in 2023, assuming ELI-NP's 10 PW laser system, an electron beam charge of 12 nC and a repetition rate of one shot per minute, about 10,000 muons per shot were predicted to reach the detector plane, allowing a transmission image in about 10 minutes.
However, the target assumed in that study was a uranium sample surrounded by a 2 cm-thick lead wall, and the result came from numerical simulation. The configuration also used multiple magnets to select muons, so the conditions differ from the shielding structure and lead imaging in the present experiment.
For that reason, the "about 10 minutes" figure cannot be applied directly to this experiment, and imaging speed cannot be compared from the two sets of numbers alone. The 2023 study itself also noted that methods using multiple large magnets carry cost and space burdens.
From "making" muons to "imaging" with them
Research on laser-driven muons did not begin suddenly with this work.
In 2025, research was reported in which muons were produced from laser-accelerated electrons and the electrons produced by their decay were detected.
Furthermore, in a preprint released in July 2026 by the BELLA team at Lawrence Berkeley National Laboratory in the US, the tracks of individual muons were reconstructed, and for some, energy was estimated from the amount they bent in a magnetic field.
The BELLA research advanced techniques for measuring individually where generated muons travel and how much energy they carry.
The ELI-NP work is different: its significance lies in capturing the shadow of a real object using the signal from many muons.
As both the development of artificial muon sources and the techniques for distinguishing and measuring the generated muons from other radiation advance, we move closer to non-destructive testing that makes quantitative use of transmission and scattering.
Practical use will also require detectors that can handle intense beams
The μ36 used here was originally built to record natural cosmic-ray muons one at a time.
With laser-driven muons, however, many particles arrive in a short time. Under these conditions, nonlinearity, in which the electronic circuit's output stops being proportional as more particles arrive, becomes a problem.
The team states explicitly that this nonlinear response has not yet been sufficiently evaluated. They list as future work the development of detectors that can handle higher particle densities and measure multiple simultaneously arriving particles individually.
Nor does a scintillator bar width of 2.5 cm mean that a practical device would automatically achieve a spatial resolution of 2.5 cm.
Image quality depends on many conditions, including the number of muons that arrive, background radiation such as gamma rays and neutrons, the response of the detection circuitry, and temperature. Having confirmed the shadow of lead 25 cm wide, the next stage will need to establish how small a defect can be identified, and whether stable results can be obtained when measurements are repeated under the same conditions.
Other efforts aimed at industrial use are also under way.
On August 17, 2026, Ideon Technologies announced that it had joined a separate international consortium centered on the University of Texas at Austin and plans a five-week experiment at ELI-NP. It aims to develop detectors and data-processing technology for intense artificial muon beams, with a commercial system as the longer-term goal.
The figure of "tens of thousands of muons per shot" described by the company is a simulation-based prediction, not a value measured in the work by Dobre and colleagues. There is also currently no basis for treating the two as the same experiment.
Cargo inspection, non-destructive testing of large structures and surveys of underground resources are envisioned as future applications of a controllable artificial muon source.
However, what was mounted in the vehicle this time was the detector, not the 10 PW laser itself.
Even if laser wakefield acceleration can produce high-energy electrons over shorter distances than before, the entire apparatus, including the high-power laser, the muon-production section and the radiation shielding, will not necessarily be easy to move.
Whether this can develop into a practical muon imaging device cannot be judged from the laser's peak power alone.
It will require delivering enough muons to the target, irradiating repeatedly at the needed frequency, preparing detectors that can accurately measure such an intense beam, and stably reproducing the shape and density of objects.
If the size of the equipment, its installation and the burden of radiation shielding can also be brought within realistic bounds, an artificial muon source could become a new non-destructive testing technology, one that examines the interior of thick structures from a chosen direction without relying solely on naturally falling cosmic rays.