Nunki: Interferometry in a Close Binary

• Nunki is a B-type stellar system identified as Sigma Sagittarii and recently revealed to be a nearly equal-mass binary at about 69 pc.
• High-angular resolution interferometry using H.E.S.S. and VLTI has provided sub-milliarcsecond angular-diameter measurements and precise orbital parameters.
• Advanced intensity interferometric techniques, including second- and third-order analyses, underpin its role as a benchmark for studying mass transfer and binary evolution.

Nunki, the star $\sigma$ Sagittarii, is a nearby B-type stellar system whose observational status has changed substantially with recent high-angular-resolution work. In the H.E.S.S. stellar intensity interferometry campaigns it served as a bright southern target for sub-milliarcsecond angular-diameter measurements in the blue optical, while later VLTI-based interferometry established that it is a close, nearly equal-mass binary at $d \approx 69\ \mathrm{pc}$. The system is now relevant both as an interferometric benchmark and as a candidate nearby pathway to eventual core collapse (Vogel et al., 2024, Waisberg et al., 17 Mar 2026).

1. Identification and stellar parameters

Nunki is identified as Sigma Sagittarii, $\sigma$ Sgr, and is described as a B-type star. The 2026 orbital study states that the paper’s spectral analysis supports a B2.5 V classification with effective temperature $T_{\rm eff} \approx 18.5$–$19\ \mathrm{kK}$ and $\log g \approx 4.0$, while the 2023 H.E.S.S. two-colour analysis adopts $T_{\rm eff} = 18987\ \mathrm{K}$ and $\log(g) = 4.0$ and uses a linear limb-darkening coefficient $u_\lambda = 0.385$ interpolated from Claret & Bloemen (2011). In the H.E.S.S. target table Nunki is listed with B-band magnitude $B = 1.923$, and the same study cites a literature angular diameter of $d \approx 69\ \mathrm{pc}$0 from Underhill et al. (1979), stated to have been measured at infrared wavelengths (Vogel et al., 2024).

Distance is central to the system’s later astrophysical interpretation. Hipparcos gives a parallax $d \approx 69\ \mathrm{pc}$1, corresponding to $d \approx 69\ \mathrm{pc}$2, and the orbital analysis adopts $d \approx 69\ \mathrm{pc}$3 at $d \approx 69\ \mathrm{pc}$4. The same paper argues that this proximity makes Nunki closer than Spica and Bellatrix, both at approximately $d \approx 69\ \mathrm{pc}$5, and therefore the closest core-collapse progenitor candidate to the Sun (Waisberg et al., 17 Mar 2026).

The early H.E.S.S. intensity-interferometry treatment did not convert the measured angular diameter into a physical radius. That paper explicitly states that it does not provide a parallax for Nunki and therefore does not compute a linear radius, while the later two-colour H.E.S.S. paper likewise states that it does not compute or quote a physical radius for Nunki from a distance or parallax (Zmija et al., 2023).

2. Angular-diameter measurements with H.E.S.S.

Nunki was one of the first stellar targets measured with the H.E.S.S. optical intensity interferometer. In April 2022, two Phase I telescopes, CT3 and CT4, were instrumented with external optical systems using a $d \approx 69\ \mathrm{pc}$6 interference filter with $d \approx 69\ \mathrm{pc}$7, two Hamamatsu R11265U-300 PMTs per telescope, and digitization at $d \approx 69\ \mathrm{pc}$8. Nunki was observed on 2022-04-18, 2022-04-22, and 2022-04-23 for a total of $d \approx 69\ \mathrm{pc}$9 of on-source data. The spatial-correlation curve was fit with a uniform-disk model, giving $\sigma$0 at $\sigma$1, with fitted zero-baseline coherence $\sigma$2 and reduced $\sigma$3 (Zmija et al., 2023).

The second H.E.S.S. campaign, in 2023, extended the setup to three telescopes, CT1, CT3, and CT4, and introduced truly simultaneous two-colour operation. Each telescope used a dichroic beamsplitter and two interference filters, providing simultaneous measurements at $\sigma$4 with $\sigma$5 and at $\sigma$6 with $\sigma$7. Nunki was observed on 2023-05-06 with $\sigma$8 and on 2023-05-10 with $\sigma$9, for a total on-source time of $T_{\rm eff} \approx 18.5$0. All three baseline combinations, 1–3, 1–4, and 3–4, were used on the nights when all telescopes were active, and the 1–3 combination had the largest projected baselines during the campaign, resulting in smaller correlation amplitudes, consistent with expectations from $T_{\rm eff} \approx 18.5$1 (Vogel et al., 2024).

Campaign and band	$T_{\rm eff} \approx 18.5$2	$T_{\rm eff} \approx 18.5$3
2022, $T_{\rm eff} \approx 18.5$4	$T_{\rm eff} \approx 18.5$5	—
2023, $T_{\rm eff} \approx 18.5$6	$T_{\rm eff} \approx 18.5$7	$T_{\rm eff} \approx 18.5$8
2023, $T_{\rm eff} \approx 18.5$9	$19\ \mathrm{kK}$0	$19\ \mathrm{kK}$1

The 2023 study fits both uniform-disk and limb-darkened-disk models and reports that the LD diameters are approximately $19\ \mathrm{kK}$2 larger than the UD values, as expected for linear limb darkening. For Nunki, the wavelength dependence is significant: the $19\ \mathrm{kK}$3 diameters are smaller than the $19\ \mathrm{kK}$4 diameters beyond the quoted uncertainties. The paper discusses two possibilities, namely unaccounted systematics or more complex stellar physics. It also notes that the 2022 and 2023 $19\ \mathrm{kK}$5 results are consistent within $19\ \mathrm{kK}$6, with substantially reduced uncertainty in 2023 due to occasional three-telescope operation and improved treatment of the zero-baseline amplitude (Vogel et al., 2024).

3. Intensity-interferometric formalism and H.E.S.S. implementation

The H.E.S.S. Nunki measurements are framed in the standard second-order formalism of stellar intensity interferometry. The 2023 two-colour study writes the temporal correlation as

$19\ \mathrm{kK}$7

with the Siegert relation

$19\ \mathrm{kK}$8

and separability $19\ \mathrm{kK}$9. For a uniform disk of angular diameter $\log g \approx 4.0$0,

$\log g \approx 4.0$1

so the observable is the squared visibility,

$\log g \approx 4.0$2

Using spatial frequency $\log g \approx 4.0$3, the same paper writes $\log g \approx 4.0$4 and $\log g \approx 4.0$5 with $\log g \approx 4.0$6 (Vogel et al., 2024).

For H.E.S.S., the filtered-light coherence time is expressed as

$\log g \approx 4.0$7

The expected zero-baseline amplitude in the 2023 setup is

$\log g \approx 4.0$8

with $\log g \approx 4.0$9 at $T_{\rm eff} = 18987\ \mathrm{K}$0 and $T_{\rm eff} = 18987\ \mathrm{K}$1 at $T_{\rm eff} = 18987\ \mathrm{K}$2, where the factor $T_{\rm eff} = 18987\ \mathrm{K}$3 accounts for unpolarized light. Numerically, the paper gives $T_{\rm eff} = 18987\ \mathrm{K}$4 at $T_{\rm eff} = 18987\ \mathrm{K}$5 and $T_{\rm eff} = 18987\ \mathrm{K}$6 at $T_{\rm eff} = 18987\ \mathrm{K}$7. For Nunki, however, the UD-fit amplitudes are $T_{\rm eff} = 18987\ \mathrm{K}$8 at $T_{\rm eff} = 18987\ \mathrm{K}$9 and $\log(g) = 4.0$0 at $\log(g) = 4.0$1, and the authors attribute the lower measured values to residual bandwidth mismatch among filters. To stabilize fits, they inserted a zero-baseline black point using colour-wise weighted averages across all targets, namely $\log(g) = 4.0$2 at $\log(g) = 4.0$3 and $\log(g) = 4.0$4 at $\log(g) = 4.0$5 (Vogel et al., 2024).

The instrumental implementation changed between campaigns but retained the same core logic: digitized PMT photocurrents, offline correlation, geometric delay correction, and fitting of the integrated bunching peak rather than its sampled height. In 2023 the PMT photocurrents were digitized at $\log(g) = 4.0$6 over a $\log(g) = 4.0$7 range, triggered and time-synchronized via the H.E.S.S. White Rabbit PPS and $\log(g) = 4.0$8 clock. Each triggered run acquired $\log(g) = 4.0$9 per channel, repeating every $u_\lambda = 0.385$0 for a duty cycle of approximately $u_\lambda = 0.385$1. The temporal $u_\lambda = 0.385$2 peak was modeled as a Gaussian with globally fixed widths $u_\lambda = 0.385$3 at $u_\lambda = 0.385$4 and $u_\lambda = 0.385$5 at $u_\lambda = 0.385$6, and uncertainties were derived by refitting the injected peak across 160 independent noise intervals to obtain a $u_\lambda = 0.385$7 distribution of integrals (Vogel et al., 2024).

4. From single-star interpretation to resolved close binary

The 2022 H.E.S.S. study treated Nunki as a single dominant source for interferometric analysis. It explicitly notes that the Washington Double Star catalog lists a much fainter companion of magnitude $u_\lambda = 0.385$8, which is negligible for the photon flux and for the measured spatial coherence, and it reports no detectable binary signature or oblateness at the sampled baselines and signal-to-noise. The excellent reduced $u_\lambda = 0.385$9 of the uniform-disk fit was taken as evidence that a single, symmetric disk at $B = 1.923$0 adequately described the data in the first visibility lobe (Zmija et al., 2023).

Later interferometric work changed that picture. The 2026 VLTI-based analysis states that Nunki is a binary of two near-equal B stars with isochrone masses $B = 1.923$1 and $B = 1.923$2. The K-band flux ratio is $B = 1.923$3 (Ab/Aa), and the H-band flux ratios from PIONIER are $B = 1.923$4–$B = 1.923$5, consistent with equal luminosities for hot stars. The data set combines a 2024 VLTI/GRAVITY detection at projected separation approximately $B = 1.923$6, three 2017 VLTI/PIONIER epochs, and a 1991 MAPPIT measurement, yielding a well-determined Keplerian orbit with $B = 1.923$7, $B = 1.923$8, $B = 1.923$9, $d \approx 69\ \mathrm{pc}$00, $d \approx 69\ \mathrm{pc}$01, and $d \approx 69\ \mathrm{pc}$02, with the standard $d \approx 69\ \mathrm{pc}$03 degeneracy because no radial velocities are used. The periastron distance is $d \approx 69\ \mathrm{pc}$04, and the dynamical total mass is $d \approx 69\ \mathrm{pc}$05 at $d \approx 69\ \mathrm{pc}$06 for $d \approx 69\ \mathrm{pc}$07 (Waisberg et al., 17 Mar 2026).

The later paper also explains why Nunki remained difficult to identify spectroscopically as a binary. The projected rotational velocity is $d \approx 69\ \mathrm{pc}$08, broad metal lines have $d \approx 69\ \mathrm{pc}$09, and the radial-velocity difference between components has semi-amplitude $d \approx 69\ \mathrm{pc}$10. At the archival spectral epochs cited in the paper, the expected separations are approximately $d \approx 69\ \mathrm{pc}$11 and $d \approx 69\ \mathrm{pc}$12, both far smaller than the intrinsic line widths. The result is that an equal-mass, rapidly rotating binary can appear virtually spectroscopically undetectable (Waisberg et al., 17 Mar 2026).

5. Third-order intensity correlations and closure phase

Nunki also became one of the first astrophysical targets used to test third-order optical intensity interferometry with H.E.S.S. In the 2023 H.E.S.S. optical intensity interferometer campaign, three of the $d \approx 69\ \mathrm{pc}$13 Cherenkov telescopes, CT1, CT3, and CT4, formed a single closure-phase triangle. The third-order analysis used the $d \approx 69\ \mathrm{pc}$14, $d \approx 69\ \mathrm{pc}$15 band; a simultaneous second band at $d \approx 69\ \mathrm{pc}$16 was recorded but not used for the $d \approx 69\ \mathrm{pc}$17 analysis because of low signal-to-noise. Nunki was observed for a total of $d \approx 69\ \mathrm{pc}$18 in 2023, but because of a broken amplifier only $d \approx 69\ \mathrm{pc}$19 of three-telescope data could be used for $d \approx 69\ \mathrm{pc}$20 (Zmija et al., 15 Dec 2025).

The formalism extends the second-order method to closure phase. For three stations,

$d \approx 69\ \mathrm{pc}$21

and at zero delay the paper writes

$d \approx 69\ \mathrm{pc}$22

where $d \approx 69\ \mathrm{pc}$23 is the closure phase. The H.E.S.S. analysis computed

$d \approx 69\ \mathrm{pc}$24

offline from raw waveforms and modeled the pairwise two-photon contributions in the $d \approx 69\ \mathrm{pc}$25 map with three Gaussian tubes. For Nunki, the fitted squared visibilities from the $d \approx 69\ \mathrm{pc}$26 map are $d \approx 69\ \mathrm{pc}$27, $d \approx 69\ \mathrm{pc}$28, and $d \approx 69\ \mathrm{pc}$29, implying a maximum possible three-photon signal $d \approx 69\ \mathrm{pc}$30 for $d \approx 69\ \mathrm{pc}$31 (Zmija et al., 15 Dec 2025).

After subtraction of the fitted two-photon ridges, the residual map showed no significant central peak. The residual RMS for Nunki is $d \approx 69\ \mathrm{pc}$32, approximately 230 times worse than the amplitude of the expected three-photon term even for $d \approx 69\ \mathrm{pc}$33, so no meaningful constraint on $d \approx 69\ \mathrm{pc}$34 could be set. The off-signal RMS scales as $d \approx 69\ \mathrm{pc}$35 after correcting for rate fluctuations, indicating shot-noise-limited performance. The same paper argues that Nunki is astrophysically instructive because a sufficiently sensitive third-order measurement could discriminate a single-star model, for which the time-averaged $d \approx 69\ \mathrm{pc}$36, from a binary model in which the closure phase toggles between $d \approx 69\ \mathrm{pc}$37 and $d \approx 69\ \mathrm{pc}$38 and a representative geometry yields $d \approx 69\ \mathrm{pc}$39 (Zmija et al., 15 Dec 2025).

6. Evolutionary interpretation and nearby core-collapse candidacy

The 2026 orbital study places Nunki in an evolutionary context that extends beyond interferometric morphology. Because the system is a close, eccentric, nearly equal-mass binary with $d \approx 69\ \mathrm{pc}$40 and periastron distance $d \approx 69\ \mathrm{pc}$41, the authors argue that mass transfer will begin at periastron rather than after circularization. Using Eggleton’s Roche-lobe approximation with $d \approx 69\ \mathrm{pc}$42, they obtain $d \approx 69\ \mathrm{pc}$43 and $d \approx 69\ \mathrm{pc}$44, so a stellar radius of approximately $d \approx 69\ \mathrm{pc}$45–$d \approx 69\ \mathrm{pc}$46 at periastron will initiate eccentric Roche lobe overflow. Their MIST-based estimate places the primary at $d \approx 69\ \mathrm{pc}$47–$d \approx 69\ \mathrm{pc}$48 in approximately $d \approx 69\ \mathrm{pc}$49, with $d \approx 69\ \mathrm{pc}$50 and a helium core of approximately $d \approx 69\ \mathrm{pc}$51 (Waisberg et al., 17 Mar 2026).

The same work argues that tidal circularization is too slow for the orbit to become circular before interaction. Using dynamical tide theory with a radiative envelope, it quotes

$d \approx 69\ \mathrm{pc}$52

far exceeding evolutionary timescales. It then compares the orbital energy available before coalescence, $d \approx 69\ \mathrm{pc}$53, with the envelope binding energy at onset, $d \approx 69\ \mathrm{pc}$54. Since the latter is a few times larger, envelope ejection is described as disfavored, and a merger into a single star of mass $d \approx 69\ \mathrm{pc}$55 is presented as a plausible outcome of eccentric Roche lobe overflow and common-envelope evolution (Waisberg et al., 17 Mar 2026).

This evolutionary argument is the basis for the designation of Nunki as the closest core-collapse progenitor candidate to the Sun. The paper explicitly compares it with Spica and Bellatrix, both at approximately $d \approx 69\ \mathrm{pc}$56, and notes that Bellatrix does not show a close equal-mass companion with K-band flux ratio higher than $d \approx 69\ \mathrm{pc}$57 in the reported VLTI/GRAVITY data. The interpretation remains conditional in several respects: the authors state that common-envelope ejection cannot be ruled out a priori, that the spin-orbit misalignment inference rests on $d \approx 69\ \mathrm{pc}$58 together with the low orbital inclination and the absence of a decretion disk, and that the time from merger to core collapse as well as the eventual supernova subtype depend on the post-interaction mass-loss history. This suggests that Nunki is best understood not as a completed evolutionary case, but as a nearby, unusually well constrained progenitor candidate whose future still depends on unresolved interaction physics (Waisberg et al., 17 Mar 2026).