Tabby's star (KIC 8462852)

Kepler observed the star almost continuously for four years. Most of the time the light curve is flat; the events fall mainly in two groups, an isolated deep dip in Quarter 8, midway through the mission, and a cluster of dips in Quarters 16 and 17, the last months, reaching about 20 per cent, with complex and asymmetric shapes and durations of days⁠[1]⁠[11]. The events have no period. Wright and Sigurdsson found the timings of the deepest dips consistent with being random⁠[6].

Planet Hunters volunteers drew attention to the star, and Boyajian and colleagues reported it in 2015 as "where's the flux?". They showed the dips were in the star's light and not an artefact, that the star is an F-type main-sequence star, and that it has no infrared excess of the kind a massive dust disk would produce⁠[1]. Spitzer and ground-based infrared observations later in 2015 confirmed the absence of warm dust⁠[12]⁠[13].

After Kepler, the star was watched from the ground. In May 2017 it began a series of new dips, the first since 2013, and dipped repeatedly until it moved out of view at the end of the year⁠[7]. TESS has since observed it in seven sectors between 2019 and 2024. Nothing like the Kepler or 2017 dips has recurred, but Sector 15 holds a single symmetric dip of 1.1 per cent lasting about 0.7 days on 3 September 2019, which a 2026 radial-velocity study reads as the transit of a giant planet on an orbit of 1,000 days or more⁠[14]⁠[15].

The record is instrumental, and the accounts are the papers that interpret it. They fall into four threads.

The dips. The discovery paper is the reference description of the Kepler events⁠[1]. Bodman and Quillen fitted them with swarms of comets⁠[11]; Neslušan and Budaj with four dust-shrouded bodies on one orbit⁠[16]; Kiefer and colleagues found two near-identical dips 928 days apart⁠[17]; Ballesteros and colleagues proposed Trojan swarms on a 12-year orbit and predicted a return in 2021⁠[18]; Sacco and colleagues a 1,574-day period with a return in October 2019⁠[19]; Bourne, Gary and Plakhov a ringed brown dwarf with a repeat on 27 December 2021⁠[20]. The 2017 campaign, run mainly on the Las Cumbres network by a team of about two hundred authors, named the dips Elsie, Celeste, Skara Brae and Angkor and showed them deeper in blue than red⁠[7]; Deeg and colleagues confirmed the colour with the Gran Telescopio Canarias⁠[8]; Hitchcock and colleagues saw a month-long colour-dependent brightening after one dip⁠[21]. In 2026 three papers returned to the TESS record: Madurga-Favieres and colleagues modelled the 2019 symmetric dip with radial velocities as a planet of about nine Jupiter masses⁠[14], Zhu and colleagues fitted it with a ringed planet⁠[15], and Socas-Navarro found a Bayes factor of about 45 for recurrence of the dip families at 782 ± 80 days⁠[22].

The fade in the Kepler era and after. Montet and Simon measured it in the Kepler full-frame images⁠[4]. Meng and colleagues followed it from the ultraviolet to the mid-infrared in 2015-16 and found its colour shallower than interstellar extinction⁠[9]; Davenport and colleagues found a matching 3 per cent drop in GALEX ultraviolet between 2011 and 2012⁠[23]. Simon and colleagues showed from ASAS and ASAS-SN that the star's brightness since 2006 has wandered rather than declined steadily: brighter in 2006-07, fainter in 2009-13, brighter around 2014, fainter again at 6 mmag per year in 2015-17⁠[24]; Hippke and Angerhausen showed the year-long variations cannot be cyclic with a period under ten years⁠[25]. Schaefer and colleagues followed the decline in four colours to early 2018⁠[26]. No published measurement extends the fade past 2018.

The century fade. Schaefer measured 1,338 Harvard plate magnitudes from 1890 to 1989 and reported a fade of 0.164 mag per century with flat check stars⁠[2]. Hippke and colleagues measured the long-term stability of the same archive, found a structural break at 1962 rather than a trend, and put the systematic floor at 0.15-0.2 mag per century⁠[3]; Schaefer replied that their comparison stars were crowded and mixed emulsions⁠[27]. Lund and colleagues extended the test to 644 F stars and found the population's slopes well behaved only once a jump across the 1953-1969 "Menzel gap" in plate-taking is modelled⁠[28]. Three independent archives were then consulted: Sonneberg and Sternberg plates gave the star constant to 0.03-0.04 mag per century from 1934⁠[29]; Maria Mitchell Observatory plates gave +0.12 ± 0.02⁠[30]. The Harvard archive was wholly reprocessed as DASCH Data Release 7 in 2024⁠[31]⁠[32], and this department's re-analysis of DR7 against matched comparison stars is recorded under the question below.

The explanations. Wright and Sigurdsson's survey of the families of solutions, and Wright's 2018 reassessment after the colour results, are the clearest statements of what the readings are and how each is constrained⁠[6]⁠[33]. Wyatt and colleagues showed one exocomet population can produce both dips and the Kepler-era fade, and that a century fade must have ended before 2010⁠[34]; Metzger, Shen and Stone proposed recovery from swallowing a planet⁠[35], and Martinez, Stone and Metzger an evaporating exomoon left over from such a collision⁠[36]; Foukal proposed an intrinsic change in the star's convective flux⁠[37]; Shestakova and colleagues bodies falling in and breaking up⁠[38]. The bound companion was disputed⁠[39] and then confirmed with five years of Keck astrometry⁠[40]; Young and Wyatt tested whether it could drive comets inward and found the odds about one in a thousand⁠[41].

The core data are public. The Kepler long-cadence light curves (30-minute sampling) for all 18 quarters and the full-frame images, from which the four-year fade was measured, are in the Mikulski Archive for Space Telescopes⁠[42]⁠[4]. TESS recorded the star at 2-minute cadence in Sectors 14 and 15 (2019), 41 (2021), 54 and 55 (2022), and in full-frame images in Sectors 75, 81 and 82 (2024)⁠[42]. Yang and Satarla stitched and detrended the 2019 to 2022 2-minute data with a half-day window and found no dip deeper than 0.5 per cent⁠[10]; with a three-day window the same data show the 1.1 per cent symmetric dip of 3 September 2019, and the 2024 sectors show nothing to 0.5 per cent at durations of a quarter to one day.

The colour information comes from the 2017 dips: multi-band photometry from Las Cumbres and other telescopes⁠[7], and spectrophotometry from the Gran Telescopio Canarias showing the dips deeper in blue than in red with an absorption coefficient that fits grains of about 0.0015 to 0.15 micron⁠[8]. One deep dip was followed by a colour-dependent brightening lasting about a month⁠[21]. The slow fade was measured from the ultraviolet to the mid-infrared from 2015 to 2016, at about 22 mmag per year with Swift and about 5 mmag per year at 3.6 to 4.5 microns⁠[9], and continued, at a varying rate, to 2018⁠[26]. Since 2018 the only ground record is the AAVSO database and ASAS-SN; ZTF saturates on a star this bright and Gaia DR3 published no epoch photometry for it.

The long history is in the Harvard plate collection, now served only as DASCH Data Release 7, which gives 2,717 detections of the star between 1890 and 1990⁠[31]. Sonneberg (1934-1995), Sternberg (1895-1995) and Maria Mitchell (1922-1991) plates have each been measured once⁠[29]⁠[30]. The star has a faint companion about 2 arcseconds away, confirmed in 2021 as a bound M dwarf at about 880 au⁠[40]. Radio and optical searches for technosignatures found nothing: no narrowband radio signal above 180 to 300 Jy in a 1 Hz channel between 1 and 10 GHz⁠[43] and no optical laser pulses⁠[44].

Circumstellar dust in clumps is the explanation the colour data support. A dip deeper in blue than red means the occulter scatters short wavelengths more, which fine dust does and solid bodies, planets or opaque panels do not⁠[7]⁠[8]. The grains implied are small enough to be pushed out of the system by starlight, so they must be renewed within days by something travelling with them⁠[8]. Candidate sources include a family of comets from a broken-up body, which can reproduce depths and durations of the later Kepler dips with tens to hundreds of large comets but not the Quarter 8 dip⁠[11]; Trojan swarms around a ringed planet, which predicted a return in 2021⁠[18]; bodies falling towards the star and breaking up⁠[38]; and debris from a planet the star has swallowed⁠[35]⁠[36]. Driving comets inward by the bound companion through the eccentric Kozai-Lidov mechanism was tested and found to have a probability of about one in a thousand of producing one such star in the Kepler field⁠[41]. The 2019 symmetric dip is a different kind of event, and if it is a planetary transit⁠[14]⁠[15] the system has a giant planet whose orbit the dust models will have to accommodate.

The slow fade is the harder part. Meng and colleagues found its colour dependence less steep than interstellar reddening and attributed it to circumstellar dust⁠[9]. Metzger and colleagues proposed that the star is recovering from swallowing a planet⁠[35]. Wright and Sigurdsson rated an intervening cloud in the interstellar medium, or an object with a large disk between us and the star, as the most plausible natural explanations for dips and fade together⁠[6]. The Kepler and post-Kepler fades rest on other instruments and are not in doubt⁠[4]⁠[9], though since 2006 the brightness has gone up as well as down⁠[24].

Whether the star also faded through the twentieth century is still open, and the department's DASCH DR7 re-analysis says why. The 0.166 mag per century slope is reproduced, and against 49 comparison stars of the same brightness and colour on the same plates, screened for neighbours, the star has the steepest slope and the largest offset across the 1953-1969 gap; but it is only about 2 sigma above them (one star in 49 is steeper), the comparison stars share a drift of +0.03 mag per century, and the star's own record is flat from 1890 to 1953. The signal is a drop of about 0.07 mag between the 1950s and the 1970s beyond what matched stars show, not a steady century-long fade⁠[2]⁠[3]⁠[28]. Independent archives disagree about even that: Sonneberg sees no step⁠[29], Maria Mitchell sees a steady fade⁠[30].

Intrinsic variability of the star was left open by the 2017 campaign, which noted that its data might also be consistent with variations in the photosphere⁠[7]; Foukal's convective-flux model is the one worked-out version⁠[37], and no known stellar process reproduces the whole light curve⁠[4]. Artificial megastructures were proposed as a hypothesis to be tested, not a conclusion, and predicted grey (colour-independent) dips for large opaque structures⁠[5]. The chromatic dips and the null technosignature searches have removed most of the support for that reading⁠[7]⁠[43].

A deep dip observed with spectroscopy across a wide wavelength range would measure the dust directly: its grain size, composition and the gas, if any, that comes with it. A comet family predicts water and carbon monoxide gas and grains of a cometary mix; debris from a destroyed planet predicts rock-forming silicates; an intervening interstellar cloud predicts sodium and other absorption lines that change during dips.

The 2019 symmetric dip gives the first dated prediction on this star that can be checked: the allowed orbital periods of Madurga-Favieres and colleagues and the 782-day recurrence of Socas-Navarro each name future transit windows, and TESS returns to the field in 2027⁠[14]⁠[22]. A second transit at a predicted time would establish the planet; an observed window with none would retire it.

For the century fade, more Harvard statistics will not do it. What would is agreement between archives: the Sonneberg, Sternberg and Maria Mitchell plates reduced with the same matched-comparison method and tested for a 0.07-0.10 mag step between the 1950s and the 1970s; and, on the Harvard side, a per-plate differential against stars within a few arcminutes using DR7's measured colour terms and a Gaia neighbour screen, to rule out a blend on the coarse post-1962 patrol plates⁠[31]. For the modern fade, the 2018-2026 ground record (AAVSO, ASAS-SN) has never been reduced; eight more years would show whether the 6 mmag per year decline continued or reversed⁠[24]. A population of analogue stars would turn a single case into a statistical problem; a survey for analogues has so far published two lists of candidates⁠[45]⁠[46].