Posted by The search for vanishing stars is less like astronomy and more like archive work. You take a photograph of the sky from 1952, you take a digital image of the same patch from last year, you line them up to sub-arcsecond precision, and you ask which points of light are in the first and not the second. Do that for a few hundred million catalogued objects and you will end up with a pile of candidates, almost all of which are junk.
Sorting the junk from the residue is the actual scientific work, and it is where technosignature research overlaps with something much more mundane: data forensics on seventy-year-old glass.
The Archive Is the Instrument
Between 1949 and 1958 the 48-inch Schmidt telescope at Palomar photographed the entire northern sky onto glass plates for the Palomar Observatory Sky Survey. Those plates were later scanned and turned into the Digitized Sky Survey, and they are the reason this line of research is possible at all. No modern instrument can give you a snapshot of the sky from before the space age. The plates can.
Modern comparison data comes from Pan-STARRS, from Gaia’s astrometric catalogue, from the Zwicky Transient Facility running on the very same Palomar telescope. The baseline you get by chaining these together is seventy years, which is long enough to catch changes that no single observing campaign would ever notice.
Long baselines are the whole point. A star that faded over four decades looks completely static in a three-year survey.
What VASCO Actually Did
The project that made this a recognised field is VASCO — Vanishing and Appearing Sources during a Century of Observations — led by Beatriz Villarroel. The method is unglamorous and rigorous. Start from an old catalogue built on the Palomar plates. Cross-match it against a deep modern survey. Flag every source with no counterpart. Then throw away everything with a boring explanation.
The attrition is severe. Hundreds of millions of starting objects came down to a candidate list in the low hundreds, and even those were not being presented as anything more than unexplained. Villarroel’s group has been explicit that the expected yield of the whole exercise is astrophysics — rare variables, obscured stars, possibly the direct collapse of a massive star into a black hole with no visible supernova. The technosignature framing is a secondary motivation, and it is honest about being a long shot.
Most Candidates Die in the First Hour
Photographic plates are physical objects that sat in darkrooms and got handled. They carry scratches, emulsion flaws, dust specks, chemical stains and reflections, any of which reads as a point source to a scanner. The standard defence is to require the object to show up in more than one independent exposure at the same coordinates, which kills most defects immediately.
Then there are asteroids. A main-belt object crossing the field during a long exposure leaves a point or a short streak that will never be at those coordinates again. Anything within a few degrees of the ecliptic gets extra scrutiny for that reason.
Proper motion is the subtler killer. Nearby stars move measurably against the background over decades — Barnard’s Star shifts more than ten arcseconds a year, which over the span between the Palomar plates and a modern image puts it a long way from where it started. A star that appears to have vanished from its 1952 position may simply be sitting somewhere else in the same frame. Gaia’s astrometry has made this class of false positive largely solvable, which was not true fifteen years ago.
What survives all of that is a short list. That is the useful output.
The 1952 Plate
One VASCO result has stayed genuinely awkward. On a plate exposed on 19 July 1952, several star-like points appear in a tight group; on another exposure of the same field taken under an hour later, they are gone. Nine of them, in a region where nothing is catalogued now.
Explanations have been offered and none has stuck cleanly. Contamination on the emulsion is the mainstream candidate, though the alignment and point-like quality of the sources make it an uncomfortable fit. Villarroel has floated glints from objects in Earth orbit, which would be striking given the date, and that suggestion has been contested at length. The case is unresolved rather than solved, and it is a good illustration of what these searches actually produce: not answers, but items that refuse to go away.
There is a broader lesson buried in that plate. Historical astronomical data was never collected with this kind of question in mind, which is precisely what makes it valuable — nobody was looking for anything in particular, so nothing was selected out. The trade-off is that the metadata is patchy, the photometry is rough by modern standards, and the people who could explain an odd feature on a given night died decades ago.
Why This Counts as SETI at All
Classical SETI waits for a transmission. Somebody out there has to point something at us, at a frequency we happen to be listening on, during a window when we happen to be looking. That is a lot of coincidences.
Technosignature astronomy drops the requirement for intent. If a civilisation builds anything large enough to affect the light we receive from a star, we might detect the consequence without anyone deciding to communicate. Archival searches take this a step further by using data that already exists — no telescope time to apply for, no target list to negotiate, just computation against catalogues that were assembled for other reasons entirely.
That is the honest appeal of vanishing stars as a search strategy. It is cheap, it uses the deepest time baseline available to astronomy, and the failure mode is discovering something about stellar evolution instead.
Machines Doing the First Pass
The scale forces automation. Cross-matching hundreds of millions of sources across surveys with different depths, resolutions and photometric systems is not something a human does by eye, and the Vera C. Rubin Observatory will make the problem larger by orders of magnitude once it starts imaging the southern sky every few nights.
So classifiers rank the anomalies and humans work down the list. It matters that the algorithm never decides anything is artificial — it decides that an object behaves unlike its neighbours, which is a statistical statement. Everything after that is done by people pulling original plate scans and requesting follow-up observations. VASCO has also put candidates in front of citizen-science volunteers, on the reasonable grounds that human eyes are still better at spotting a scanning artefact than most classifiers are.
What Would Have to Happen
A convincing case would need several independent things to line up: the source verified in multiple historical exposures, genuine absence in deep modern imaging, no proper-motion or transient explanation, and ideally an infrared measurement that is inconsistent with the obvious dust model. Independent teams would then need to reproduce all of it.
Nothing has come close, and it would be surprising if anything did soon. The value of the work is that it is testable, cheap and produces real astrophysics regardless of the outcome, which is more than can be said for most attempts to answer this question. SETIworld tracks the vanishing stars candidates and the follow-ups, including the many that end as a smudge on a piece of glass from the 1950s.
