A team from the Instituto de Astrofísica de Canarias (IAC) has tested a new technique capable of rapidly identifying the distortions that light undergoes as it passes through Earth’s atmosphere. The development, carried out by the Laboratorio para la Innovación en Optomecánica (LIOM) within the framework of the CELESTE project, could in the future contribute to the creation of more compact and simpler adaptive optics systems to improve the quality of astronomical observations.

The technology combines an optical fibre approximately one centimetre long with artificial intelligence techniques. The fibre transforms distortions in the light into a kind of luminous fingerprint, which can then be interpreted by a neural network to determine how the light has been distorted before reaching the measuring instrument. The system carries out this process in just a few milliseconds, a particularly important speed because the atmosphere changes continuously and any correction must be made virtually in real time.

Correcting the effects of the atmosphere

Light from stars and other astronomical objects reaches Earth almost undistorted. However, as it passes through the atmosphere, it encounters small variations in the properties of the air that continuously alter its path.

In optics, the shape taken by light as it propagates is described by its wavefront. Atmospheric turbulence distorts this wavefront and causes astronomical images obtained from the ground to lose sharpness.

To compensate for this effect, large telescopes use adaptive optics systems, a technology capable of measuring these distortions and continuously correcting them using deformable mirrors that make slight adjustments to their shape.

In conventional systems, the sensor that measures the distortions and the science camera do not always receive exactly the same light through the same optical path. This can introduce small additional differences that limit the precision of the correction, a problem that is especially important when trying to observe very faint objects next to much brighter ones, as is the case in the search for and study of exoplanets.

The approach presented by the team addresses this problem by measuring the distortions directly from the light that reaches the point where the image is formed.

A one-centimetre fibre as a sensor

The new technique uses a very short multimode optical fibre, approximately one centimetre long. This type of fibre allows light to travel along several internal paths at the same time. When these paths recombine at the output, they produce a characteristic light pattern that changes depending on how the light was distorted when it entered the fibre.

The length of the fibre is key. Through simulations and laboratory tests, the team has shown that when the fibre is sufficiently short, it better preserves the information needed to distinguish one distortion from another. This is especially relevant in astronomy because, unlike the highly controlled light of a laboratory laser, starlight contains a range of wavelengths.

Auxiliadora Padrón-Brito, a researcher at the IAC and the Universidad de La Laguna (ULL) and first author of the study, explains: “The fibre converts distortions in the light into a pattern that we can read as a fingerprint. What is interesting is that we are able to preserve that information using an extremely simple and compact element, opening up the possibility of developing sensors that are easier to integrate into future instruments.”

Distinguishing distortions that may look the same

Some types of distortions pose an additional difficulty: they can produce virtually identical images even when the distortion has opposite characteristics. Wavefront sensors must be able to distinguish between them.

The study shows that passing the light through the short fibre helps to break this similarity. The different paths followed by the light inside the fibre cause distortions that might otherwise be indistinguishable to produce different fingerprints at the output.

The team demonstrated this behaviour using defocus, one of the best-known optical distortions, as an example. With a short fibre, the system retained information that made it possible to distinguish between situations that could appear ambiguous if only the image itself were observed. When the fibre was much longer, this ability was significantly reduced.

This result is one of the key features of the new sensor concept, as it makes it possible to recover information that other methods working directly on the image may have difficulty distinguishing.

Artificial intelligence to interpret the light fingerprint

Once this luminous fingerprint has been obtained, the next challenge is to translate it and determine what distortions the light has undergone.

To do this, the team trained a convolutional neural network, a machine-learning algorithm capable of learning the relationship between thousands of different light patterns and the distortions that produced them.

In the simulations, the network learned to recognise ten basic types of optical distortion and was able to reconstruct them with high accuracy. In addition, each estimate takes only a few milliseconds.

“The atmosphere is changing continuously, so it is not enough to measure a distortion accurately: you have to do it quickly enough to react before the conditions change again,” explains Natalia Arteaga-Marrero, a researcher at the IAC and the Universidad de La Laguna (ULL) and co-author of the study. “The results show that this combination of optical fibre and machine learning has the potential to operate on those timescales.”

In future developments, this capability could make it possible to integrate the sensor into an adaptive optics system that continuously measures distortions in the light and sends that information to a deformable mirror responsible for compensating for them.

A more compact and simpler system

Compared with other wavefront sensor concepts, the proposed solution stands out for the simplicity of the elements it requires: a short section of optical fibre, several optical components to form the image, a camera and a machine-learning-based processing system.

The fibre is also a passive, compact and low-cost element. Its short length could also help the device remain stable against vibrations or small movements, an important consideration in an observatory environment.

The study also suggests that improved temperature control could help maintain the system’s calibration over longer periods. This is also easier to achieve because of the short length of the fibre.

Another possibility opened up by the concept is to use the same signal to obtain two types of information: on the one hand, to determine how the light has been distorted and, on the other, to recover the astronomical image itself.

This would allow distortion measurement and scientific observation to share the same optical path, reducing possible errors introduced by using different paths.

Applications beyond astronomy

Although the work is mainly presented in the context of astronomical instrumentation, the technique could also have applications in free-space optical communications.

These systems transmit information using beams of light that travel through the atmosphere and are also affected by turbulence. Being able to measure and compensate for these distortions rapidly could help improve the quality and efficiency of the links.

The study also points to possible future applications in quantum optical communications, using a suitable reference channel to carry out the measurements.

From proof of concept to real observations

The results presented still constitute a demonstration of the technique’s feasibility. Part of the fibre’s behaviour has been experimentally verified in the laboratory and, more recently, the researchers have taken a further step by also demonstrating experimentally that the system can identify distortions introduced into the light using machine learning. This test reproduces under laboratory conditions what had previously been validated mainly through simulations.

The next steps will include more complete experimental tests, integrating the sensor with a real correction system and assessing how it performs during continuous operation.

The team also plans to study its performance with more complex distortions, improve the long-term stability of the device and explore other fibre and waveguide geometries. A particularly important stage will ultimately be to test the system under real astronomical observing conditions.

A development within the framework of CELESTE

The work is part of CELESTE, a project of the Instituto de Astrofísica de Canarias aimed at developing new technological capabilities and advanced instrumentation.

In this context, the participation of the Laboratorio para la Innovación en Optomecánica (LIOM) contributes to the development of solutions combining optics, photonics, instrumentation and machine learning to address new scientific and technological challenges.

The work has been published in the scientific journal Optics Express under the title Focal-plane wavefront sensing with narrowband light using a short multi-mode fiber. The author team comprises Auxiliadora Padrón-Brito, Natalia Arteaga-Marrero, Ian Cunnyngham and Jeff Kuhn, all of whom are members of the team involved in this development. Padrón-Brito and Arteaga-Marrero are affiliated with the IAC and the Universidad de La Laguna (ULL); Cunnyngham is affiliated with the Institute for Astronomy at the University of Hawai‘i at Mānoa; and Kuhn is affiliated with the IAC, the ULL and the Hawaiian institution.

LIOM’s R&D&I activities are also supported by the Cabildo Insular de Tenerife through the collaboration agreement “Apoyo a las actuaciones I+D+I en el espacio de cooperación IACTEC”, in addition to funding from the European Union through project reference 101087032.

Scientific article

Auxiliadora Padrón-Brito, Natalia Arteaga-Marrero, Ian Cunnyngham and Jeff Kuhn
Focal-plane wavefront sensing with narrowband light using a short multi-mode fiber
Optics Express, Vol. 34, No. 6, 2026
DOI: 10.1364/OE.580986