I love this story.

Let me take you back to the days of physically demanding astronomy. The best light detector was photographic film adhered to glass plates up to 12×12 inches square. During the long, cold night, the astronomer stayed with their telescope. Occasionally pressing fine-adjustment controls, they kept a guide star centered in the cross-hairs in the eyepiece until the exposure was done. When the time came to end the exposure, the astronomer would close the shutter, then carefully slide the glass plate off the back of the telescope. They would store it in carefully labeled trays for later developing, then go and load up the next plate.

To relax, French observers would often smoke a cigarette in the chilly dome.

In 1962, astronomers Daniel Barbier and Nina Morguleff were observing at the Haute-Provence Observatory 90 miles east of Avignon. When the exposure of ordinary G6 star HD 117043 was developed, the spectrogram looked almost normal, but not quite. There were two bright emission lines where, ordinarily, dark absorption lines should be. The lines were due to the element potassium.

One of the miracles of astronomical spectroscopy is that each element has a distinctive fingerprint, showing an unvarying pattern of narrow lines placed with mathematical precision along the wavelength scale. Often, qualities about the lines (narrow or wide, bright emission or dark absorption, and perhaps little wavelength shifts) give away physical conditions in the astrophysical object being studied. In this case, bright emission lines signal something unexpectedly hot.

Barbier and Morguleff observed the star again the next night. The result changed. The potassium emission lines were gone.

There appeared to be a new class of star, the potassium flare star, whose outbursts were brief. Clearly more observations were needed.

And more observations commenced at the Haute-Provence Observatory. One telescope-instrument combination showed no potassium emission, but using the original configuration, two more flares were discovered. No star was observed to flare more than once despite many attempts. The rate was three flares in about 1000 spectrograms. A group of American astronomers joined the chase.

They were worried about one of the stars, which was so hot that neutral potassium should not exist on the star’s surface. The difficulty of detecting neutral potassium where none should exist make them think of an alternative.

The plot twists: they thought of matches. Both “strike anywhere” and “safety” matches contain potassium chlorate (KCLO₃). For making flame, the important part of the chemical is the oxygen, but for possible contamination of potassium spectrograms, the important part is the potassium (K).

The American group (Robert F. Wing, Manuel Peimbert, and Hyron Spinrad) tested match light at the telescope, lighting matches and taking spectrograms. They tested “book matches, kitchen matches, and safety matches.” The results looked strongly like the supposed flare star spectrograms.

And here is the part I love best. A correspondence was struck between Lick Observatory and the Haute-Provence Observatory in order to test the match light hypothesis. Mme Yvette Andrillat did extensive testing, finding that there were some places in the room where stray light could enter the light beam being fed to the spectrograph.

There was one final paper on the topic (1967, PASP 79, 351). The authors conclude with a gracious paragraph.

We are greatly indebted to Mme Y. Andrillat for conducting match tests at the Haute-Provence Observatory, for communicating to us her results in detail, and for reading the manuscript of this paper. We would like to thank Mr. D. Miller and Dr. P. Conti for helping us to obtain the daytime spectrum illustrated in Plate II, and Dr. G. Preston for authorizing our unorthodox use of the 120-inch coudé spectrograph.

This is an example, and a very innocent, collegial example, of self-correction in science. Match light was repeatable and testable, and therefore became the preferred explanation. Astronomers became more careful with their smoking habits, and potassium flare phenomena have not been observed since.