Seven New Quasar‑Lens Systems Discovered Using AI in 800,000 Spectra
Astronomers used AI to sift through 800,000 quasar spectra, uncovering seven new quasar‑lens systems and roughly doubling the known sample.
A team of astronomers has uncovered seven new gravitational lens systems where a quasar’s host galaxy also bends the light of a more distant galaxy. They achieved this by training a neural network on synthetic lens spectra and scanning the Dark Energy Spectroscopic Instrument’s catalog of roughly 800,000 quasar observations. The algorithm trimmed the dataset to 200 promising candidates, and human vetting left just seven real lenses—about twice the number previously known. These rare alignments sit billions of light‑years away and open a fresh window onto the mass distribution of quasar‑hosting galaxies. Confirming them will require sharper imaging from facilities like the James Webb Space Telescope.
What happened
Quasars are the ultra‑bright cores of distant galaxies powered by supermassive black holes, and their brilliance often overwhelms the surrounding host galaxy. When a foreground galaxy containing a quasar lies directly in front of a more distant galaxy, the foreground mass bends and magnifies the background light, producing multiple distorted images—a phenomenon known as gravitational lensing.
Finding such chance alignments is difficult because they require a very specific line‑of‑sight geometry. Everett McArthur and collaborators tackled the problem by feeding a neural network a library of mock lenses created from real quasar spectra combined with real background galaxy spectra. The network learned the spectral signature of a blended quasar‑galaxy system and then scanned the DESI archive of ~800,000 quasar spectra, reducing the pool to 200 high‑confidence candidates.
Human inspection of the 200 spectra narrowed the list to seven robust lens candidates. All seven lie at least five to six billion light‑years away, and each shows the characteristic multiple‑image pattern expected from a quasar host lensing a background galaxy. Follow‑up with high‑resolution imaging will be needed to confirm the lensing geometry.
Why it matters
Doubling the census of quasar‑lens systems gives astronomers a much richer laboratory for probing the distribution of dark matter in active galaxies and for measuring the expansion of the universe with lens‑time delays. The success of a synthetic‑training‑set approach also demonstrates that machine‑learning pipelines can extract rare phenomena from massive spectroscopic surveys, paving the way for future discoveries with upcoming facilities such as the Vera C. Rubin Observatory.
- Provides roughly double the number of known quasar‑lens systems, enabling better statistical studies.
- Shows that synthetic training data can overcome the scarcity of real examples in astrophysical machine learning.
- Creates new laboratories for probing dark‑matter distribution in active galaxies at high redshift.
- The sample remains small, so conclusions about the broader population are still limited.
- Confirmation requires expensive high‑resolution observations that are not yet available.
- Reliance on simulated spectra may introduce selection biases that miss atypical lenses.
How to think about it
Researchers interested in rare lens configurations should consider building similar neural‑network classifiers that use blended‑spectrum simulations as training material. After automated filtering, a modest human vetting stage can efficiently isolate the most promising candidates. Once a shortlist is assembled, prioritize follow‑up with space‑based imaging or adaptive‑optics instruments to resolve the multiple images and model the lens mass.
FAQ
How can a quasar host act as a gravitational lens?+
Why are quasar‑lens systems so rare?+
What observations are required to confirm the new candidates?+
- astronomy·3 min readAI uncovers seven quasar gravitational lenses, shedding light on supermassive black hole growth
Artificial intelligence finds seven quasar lenses in DESI data, offering a new way to study how supermassive black holes evolve.
- astronomy·3 min readSDSS‑V DR20 releases 2 million spectra, mapping the Milky Way’s halo in detail
SDSS‑V’s 20th data release provides over 2 million stellar spectra, including 250 k halo stars, opening new windows on the Galaxy’s formation.
- astronomy·3 min readNew 3‑D Mapping Technique Reveals Detailed Distribution of Cosmic Dark Matter
A novel method combines gravitational lensing data to produce a high‑resolution 3‑D map of dark matter, sharpening our view of the universe’s hidden scaffolding.