Gravitational lens candidates have been found in the process of examination of ground-based deep sky images covering the Euclid Deep Field North area.
More than 10,000 online volunteers scrutinized galaxy cutouts available from the Hawaii Two-0 optical survey project.
Their visual classification yielded 51 highly probable strong gravitational lens candidates, which consist of deflections of the light from background galaxies to form bright rings and arcs by foreground galaxies’ masses.
Crowdsourced visual examination increases the number of known gravitational lens candidates fourfold.
Mapping dark sky corridors with volunteer eyes
The published Galaxy Zoo analysis covers six square degrees of the Euclid Deep Field North using H20 Hyper Suprime-Cam imaging.
It took a lot of sorting to find the galaxy cutouts from ground-based Hawaii Two-0 observations that covered more than 41,000 individual galaxies.
Computer programs overlook the presence of gravitational arcs because the background galaxies have irregular shapes and faint optical emission.
More than ten thousand Galaxy Zoo participants visually screened multi-band images of galaxy cutouts to identify signs of distortion of light from the stars that were ignored by computer programs.
Human eyes traced the distortions of the universe that machines failed to see.
Many participant votes identified promising candidate systems, resulting in a trustworthy catalog of probable gravitational lenses.
Bending starlight around massive foreground galaxies
The giant foreground galaxies become a funhouse mirror in space due to their huge gravitational forces.
Such huge gravitational force bends light paths coming from the background galaxies that lie right behind the foreground galaxies.
The deflected light paths bend such that the light is bent directly around the central galaxy, thereby forming either arcs or multiple images or an Einstein ring.
The measurements of the structure of the bent light enable astronomers to calculate the total mass of the giant galaxies, including the dark matter halo surrounding them.
The gravitational lenses help in the detection of dark matter surrounding individual host galaxies.
Detection of 51 additional candidates increases the sample size four-fold for the northern primary field.
Combining human classifications with artificial intelligence
Volunteers fed visual classifications back into the Galaxy Zoo Zoobot, an advanced deep-learning neural network trained on astronomical survey data.
Zoobot operated within an active-learning loop to classify overall galaxy morphology, efficiently ruling out featureless or ordinary galaxies to prioritize complex subjects for human review.
However, the multi-select “Lens or arc” question was explicitly excluded from Zoobot’s training dataset, meaning the AI system did not find or classify the lens candidates.
Instead, the 51 gravitational lenses were uncovered directly through volunteer efforts across the main Galaxy Zoo workflow, community Talk tags, and a dedicated Galaxy Zoo Mobile workflow.
Astronomical experts then vetted these crowdsourced candidates using Hyper Suprime-Cam (HSC) and Euclid imaging.
Combining machine-driven subject prioritization with human visual inspection establishes a highly scalable pipeline for processing future deep sky surveys.

Mapping dark matter across deep space
The light from these 51 lensing candidates had to travel for billions of years through space before hitting the optical sensors on Earth.
Detecting such distorted light from behind enables scientists to study faint galaxies whose light would otherwise be too faint for detection.
Comparing the total mass of galaxies with the observed bent light gives an accurate estimate of the invisible dark matter halos.
Estimating the dark matter halos enables cosmologists to test different theories of dark energy in relation to structure formation in the universe.
The light from the arches exposes the invisible skeleton structure of the universe.
Defining limits of visual candidate classifications
Based on a visual analysis of ground-based optical cutouts, there are 51 interesting strong gravitational lens candidates within the primary survey field.
The ground-based optical cutouts are unable to resolve ultra-faint structures that need very high angular resolution from space.
High-resolution space observations and subsequent spectroscopic surveys are needed.
Faint halos of foreground galaxies and companion stars can sometimes produce similar visual patterns as gravitational arcs by ground-based seeing effects.
Multi-wavelength observations are required in order to verify visual candidate lists before constructing accurate mass maps.
Crowdsourcing efforts generate great lists of candidates, but further observational follow-up is necessary for dark matter physics.
The full study can be read here: Pearson, J., Dickinson, H., Serjeant, S., Walmsley, M., Fortson, L., Kruk, S., … & Oi, N. (2026). Galaxy Zoo: Cosmic Dawn–morphological classifications for over 41 000 galaxies in the Euclid Deep Field North from the Hawaii Two-0 Cosmic Dawn survey. Monthly Notices of the Royal Astronomical Society, 546(3), staf2250.
Read the whole thing?
Get the week's signal, not the noise
Our sharpest reporting on energy, climate and nature — free, once a week.