The University of California, San Diego (UC San Diego) announced on September 25, 2026 a research result showing that changing how light is applied can reverse the magnetization of magnetic films stacked thicker than before. Even in a film of nine platinum (Pt)/cobalt (Co) pairs, the magnetization could be switched regardless of the rotation direction of the circularly polarized light.
The result suggests that, in raising the density and speed of magnetic recording, how light is applied can be an important design element alongside material composition.
However, the experiments used repeated light pulses to switch the magnetization. The laser's pulse width was 120 femtoseconds, but that does not mean a memory write would take 120 femtoseconds.
Why can tightly focused light switch even thick magnetic films?
The paper, by Muhammad Waleed Khalid, Abdoulaye Ndao, Eric E. Fullerton and colleagues, was published as a peer-reviewed article in Nature Communications on September 15.
The subject is a thin film made of alternating layers of 0.7 nm platinum and 0.6 nm cobalt. The paper treats one Pt/Co bilayer as a single unit and denotes the number of repeats as N.
N = 9 means "nine Pt/Co pairs were stacked." It does not mean the whole metal structure, including underlayers and protective films, has nine layers.
In magnetic recording, the direction of magnetization corresponds to the information values 0 and 1.
All-optical magnetization switching, which changes the magnetization direction with light, does not require an external magnetic field for rewriting. It has therefore been studied as a way to combine photonic circuits with magnetic memory.
But simply shining light does not always align the magnetization of the entire illuminated area in one direction.
Inside a magnetic film there are "magnetic domains," regions in which the magnetization points the same way. Earlier studies with Pt/Co multilayers mainly focused on relatively thin films. They used circularly polarized light and exploited the slight difference in absorption between right- and left-handed polarization to move domain walls, the boundaries between domains.
As the number of stacked layers increases, however, the size of the domains that are stable inside the film also changes. Multiple domains then tend to remain in the illuminated area, making it difficult to reverse the whole area in a single direction.
In this study, the team strongly focused a "circularly polarized vector beam," whose polarization state varies by position, with an objective lens.
They did not shorten the wavelength of the light itself. Instead, they made the illuminated region smaller and controlled how the light's energy is distributed within it.
According to the team, rapidly heating a small region weakens the force that keeps the magnetization in a fixed direction, making it easier for a reversed domain to form.
Repeated light pulses then make that domain grow. The magnetic field created by the surrounding unilluminated region also acts to stabilize the magnetization state after cooling.
The team believes that when light is focused onto a sufficiently small area, this heating and the influence of the surrounding field affect the switching more strongly than the absorption difference between right- and left-handed circular polarization.
In other words, tight focusing does more than merely shrink the write area. It changes the process by which domains are born, grow, and settle into a final magnetization state.
As a result, the same material was shown to support either switching that depends on the circular polarization handedness or switching that does not, depending on the conditions.
Narrowing the light spot to 0.96 µm changes how switching occurs
For a film with three Pt/Co pairs, the team compared spot diameters ranging from 1.92 µm to 0.96 µm.
With large spots, the final magnetization direction differed between right- and left-handed circularly polarized light. When the spot diameter was reduced to about 1.36 µm or less, this difference disappeared, and both polarizations switched the magnetization in the same direction.
The spot diameter here is defined as the position where the light intensity falls to 1/e² of its peak value at the center.
Although 0.96 µm corresponds to 960 nm, this is not the size of one memory bit or the minimum spacing at which adjacent bits can be placed.
Figures 3 and 4 of the paper show that how switching occurs depends on how tightly the light is focused and on the layer structure of the magnetic film.
| Number of Pt/Co repeats N | Magnetization switching confirmed in the paper |
|---|---|
| 1–2 pairs | Strongly dependent on circular polarization handedness |
| 3 pairs | Shifts to switching independent of handedness as the spot is made smaller |
| 4 pairs | Handedness-independent switching confirmed at the spot sizes tested |
| 5–9 pairs | Switching occurs when light is focused to about the wavelength; multiple domains remain with larger spots |
| 10 pairs or more | Striped domains remain, making uniform switching difficult, according to the authors |
Note the differences in measurement conditions when reading this table.
In the experiment comparing N = 1 to 9 in Figure 4, the numerical aperture (NA), which indicates focusing performance, was fixed at 0.80. The dependence on spot diameter itself was examined in separate experiments, such as Figure 3.
The table therefore summarizes results from several experiments; it is not a comparison of all stack counts at exactly the same spot diameter.
Nor can films be switched by the same method indefinitely as they are made thicker.
Magnetic memory must both retain recorded magnetization stably for a long time and reliably reverse it only when needed.
The result shows conditions under which optical switching is possible even in Pt/Co multilayers thicker than before. It is not a study that measured any extension of how long the magnetization is retained.
Even with a 120-femtosecond laser, writing does not finish in 120 femtoseconds
The laser used in the experiments had a wavelength of 800 nm, a pulse width of 120 femtoseconds, and a repetition rate of 1 kHz.
That is, extremely short light pulses were delivered 1,000 times per second.
In the experiment comparing spot diameters on the three-pair Pt/Co film, the sample was moved at 1 µm per second so that enough pulses overlapped on the same surrounding region.
Using 120-femtosecond light pulses is entirely different from finishing the writing of magnetic information in 120 femtoseconds.
The authors have not confirmed a phenomenon in which a single light pulse reliably switches the magnetization of this Pt/Co film.
They explain that multiple pulses create a reversed domain, and the process of its gradual growth determines the final magnetization state.
Calculating a memory's write speed from the laser pulse width alone would therefore ignore the time needed for domains to form and grow.
The sample scanning speed used in the experiments likewise does not indicate an upper limit on the speed of future storage devices.
These were experimental conditions set to observe under what circumstances switching occurs.
To judge whether this could serve as high-speed memory, the total time required for one write must be measured, including how many pulses are needed to reliably switch a targeted region and at what interval each pulse can be delivered.
The university's announcement also cites the researchers' estimate that optical switching could be "more than 1,000 times" faster than methods using an external magnetic field.
This does not mean that the read/write performance of a whole memory was measured with the present setup and a 1,000-fold speedup was achieved.
The potential future speedup and the magnetic-material switching phenomenon demonstrated here need to be considered separately.
The phrase "without an external magnetic field" also calls for care.
According to the paper's methods section, before the experiments began, a permanent magnet was used to align the magnetization of the whole sample either up or down. Light was then used to switch the magnetization from that state.
The switching mechanism is also thought to involve the field produced by the magnetic film surrounding the illuminated region.
In other words, the action of magnetic fields has not become entirely unnecessary; the point is that no external field needs to be applied when rewriting the magnetization.
Practical memory will require smaller light sources and verified write performance
The same research team also studied the interaction between light and Co/Pt thin films in Physical Review B in April 2024.
That experiment reported that magnetization switching was governed mainly by the spin angular momentum of light associated with its polarization, with little effect from the direction or magnitude of the orbital angular momentum related to optical vortices.
The present study does not contradict that conclusion.
It is better understood as showing that changing conditions such as how tightly light is focused and how many Pt/Co layers are stacked lets the effects of heat and the surrounding magnetic field appear more strongly.
In other words, not only the magnetic material itself but also "what kind of light is applied, and how" is an important design element for controlling magnetization switching.
If dependence on the handedness of circular polarization can be reduced, the optical system for rewriting magnetization might be simplified.
However, the experiments also precisely controlled the polarization distribution and focus of the light. This does not mean the same switching occurs with any laser.
One major obstacle to practical use is the light source.
UC San Diego explains that the specialized ultrafast laser used here is difficult to integrate directly into a computer chip.
The team plans to look for materials that show similar switching with lasers that are easier to build into electronic circuits, and to study optical structures that can focus light down to several hundred nanometers.
The figure of several hundred nanometers has not been achieved; it is a goal for future research.
Evaluating this as magnetic memory will also take more than miniaturizing the optics.
Can the targeted spot be reliably switched with the required number of light pulses? Does performance degrade when the same spot is rewritten repeatedly? Can the written magnetization state be retained for long enough? Can writing be done without affecting adjacent recording regions?
Energy consumption also needs to be evaluated not just as the light energy absorbed by the magnetic film, but for the whole system, including the laser and optics.
Only when these conditions are met can the combination of magnetic film and optics shown here be assessed as a practical memory technology whose recording density, write speed, and power consumption can be compared with existing technologies.