A Possible Fifth Planet for HR 8799, Closer In Than the Rest

The HR 8799 system has four directly imaged planets. A new reduction of JWST data has turned up a candidate fifth, at roughly 7 astronomical units from the star, closer in than the other four.

The candidate comes from JWST's NIRISS instrument, run in aperture masking interferometry mode, and it rests on a technique rather than a bigger mirror. NIRISS can be fitted with an aperture mask: a plate with a small number of holes that sits in the beam. Light passing through separate holes interferes, the way it does in a two-slit experiment, and the resulting fringe pattern is sensitive to where the light came from. A companion separated from its star by a small angle shifts the fringes measurably. A star alone produces a pattern with no such shift. The holes are spread across the full 6.5-metre aperture, so the fringes carry detail at the resolution of the whole mirror rather than of a single pinhole.

That resolution is the point of doing it this way, and it can be worked out from the filter and the mirror. The diffraction limit is 1.22λ/D. At 3.8 micrometres, the wavelength of the F380M filter, 1.22 × 3.8 × 10⁻⁶ metres divided by 6.5 metres gives 7.1 × 10⁻⁷ radians, or about 150 milliarcseconds: roughly one twenty-four-thousandth of a degree. That is the finest detail the mirror can deliver in that filter, and it is also where the candidate sits. In a conventional image, the star's own diffracted light at that separation is what sets the floor, and the floor has a number. The candidate's contrast of about 2 × 10⁻⁴, one part in five thousand, is measured just above the 3σ contrast curve at that radius. The obstacle at 150 milliarcseconds is the star, and the contrast curve says how much of one.

What changed is the pipeline, not the telescope. Earlier NIRISS AMI analyses of faint companions ran into systematics: optical and electronic errors that are not random, that sit above the noise floor, and that do not average away with longer exposures. The reduction described in the new arXiv paper models those and removes them, which lets the measurement reach photon noise, the point where the only thing limiting you is how many photons you collected. That is what brings a source at 2 × 10⁻⁴ contrast into view in the F380M filter.

The 3σ line deserves a plain reading. The paper reports the source as lying just above a 3σ contrast curve, and three sigma means the peak is three times the local noise. In any search that covers many resolution elements, a few peaks that high will turn up by chance, so a detection at that level in a single filter is a candidate by definition. Candidate is the word the paper uses.

One argument that pushes toward the bound-companion case is motion. Gaia and Hipparcos have tracked how HR 8799 itself moves across the sky over years. A background star would drift relative to that motion; a planet bound to the star would not. The candidate's separation is compatible with the bound case. Compatible is a consistency check, not a demonstration.

The other argument is dynamics, and it starts with the resonance. A fifth planet in a 3:1 mean motion resonance with planet e would complete three orbits for every one of e's. Kepler's third law ties period to orbital size: radius goes as period to the power two-thirds, so a factor of three in period is a factor of 3^(2/3), about 2.1, in radius. That would put such a planet a little under half of planet e's orbital distance — arithmetic from the resonance condition, not a number the paper reports. What the paper reports is that the candidate's angular position falls close to a stable orbital solution for a fifth planet in that resonance. Stability is doing real work in that sentence. Many positions in a system of four widely separated giants are chaotic over millions of years, and a planet parked in one of them would not have survived long enough to be imaged.

Seven au is a bit wider than Jupiter's orbit at 5.2 au, so this is not a close-in planet of the kind transit surveys find in bulk. It is an inner member of an architecture of widely separated giants, and confirmation would make HR 8799 the first directly imaged five-planet system. The methodological result is what this reduction reaches: photon-noise-limited measurement on a bright star at 150 milliarcseconds. Whether the same pipeline does as well on other targets is not something this paper shows.

What is still open is mostly the detection itself. One filter, one contrast curve, one 3σ peak. The separation is measured; the orbital radius of about 7 au follows from it, and takes the source to be a companion of HR 8799 rather than a background object. The mass is not measured at all. Contrast alone does not give you a mass, and converting one into the other relies on models of how a giant planet cools with age, which is where "a few to several Jupiter masses" comes from. The resonance fit assumes the source is bound and that this particular stable solution is the right one among the several a five-planet system might occupy. Either of two observations would settle it: a repeat epoch showing the source stays fixed relative to HR 8799 as the star moves, or a detection in a second filter at the same place. Failure to reappear in a deeper exposure or at a different wavelength would make it go away. Until then, HR 8799's planet count stands at four.