A divergent-light 22° halo from a Long March 6A reentry

Around 19:51 (UTC+8) on 4 June 2026, the RedNote user "岚烟喵喵喵" (ID:6109395002) photographed a bright divergent-light 22° halo at Qishui Bay, Hainan (19.6416°N, 110.9883°E). The light source was probably the reentering stage of the Long March 6A (Y25), which shone through high clouds about twelve minutes after lifting off from Taiyuan Satellite Launch Center at 19:39. The strong divergence made the 22° halo appear noticeably distorted rather than circular (Fig. 1). No precise orbital data was available for the reentering stage, thus an approximate trajectory was reconstructed from the government maritime navigation warning (Qiong Navigation Warning 82/26)¹ and from image fitting. 

Fig. 1 Frame 300, showing the divergent 22° halo

The trajectory reconstruction here is rough. No aerodynamics are used. It is a purely geometric back-projection with drag and reentry deceleration ignored, so this is an approximate trajectory. There is only one viewpoint, so the direction to the source is constrained but its distance is not. And the fit was done by eye, so errors are large.

The observer was located at Qishui Bay from terrain in the video and the RedNote post, with a small cove ahead of them matched against Google Earth to find the coordinates (19.6416°N, 110.9883°E) as shown in Fig. 2. Because the event was at night and low over the sea, no calibration stars were available, so the camera was calibrated against the terrain instead. SRTM 30 m DEM data gives the three-dimensional coordinates (latitude, longitude, height) of the surrounding landscape. Manually aligning the terrain outline visible in the video, the two small hills on the left, with the DEM-rendered terrain solves for the camera's field of view and pointing. Once calibrated, any pixel in the frame can be converted to an azimuth and elevation.

Five pixels along the glowing track were measured this way, giving five sightlines. The debris is assumed to move along a straight line in space. However, single-station angles cannot fix the distance scale. The official drop zone was therefore used as the trajectory endpoint, and the distance to the debris at each of the five points was then solved by least squares. The fitted points have a collinearity residual of about 1.3 km, which is small compared with the roughly 200 km range, although it should not be interpreted as a formal uncertainty. The result places the glowing reentry low in the east-southeast, at roughly 70 to 34 km altitude and 230 to 195 km away. This still rests on the assumption that the path is a straight line.

Fig. 2 Google Earth view of the observation site, matching the terrain in the video

The light passed through a layer of high cloud between the observer and the reentry. Himawari-9 cloud data for the nearest time (11:40 UTC) in Fig. 3 shows the sightlines crossing cirrus with tops around 13 to 15 km and an optical depth of roughly 0.5 to 1.4. This supports the presence of a suitable high-cloud layer.

Fig. 3 Himawari-9 cloud-top height at 11:40 UTC 

The parallel-light halo data was generated with ZHANG Jiajie's Lumice, and the divergent-light display was then computed from it using Lefaudeux's method². Because of the poor quality of the original video and the presence of only a 22° halo, the simulation used a single regular hexagonal crystal habit with a c/a ratio of 2.5. Several simplifications should be noted. The simulation accounts for neither the brightness of the light source nor that of the halo, and the optical transmittance of the cloud plays no role in it. The ice crystals are assumed uniform throughout the cloud layer, which was set between 13200 and 14700 m, with a maximum simulated distance of 300 km from the observer. The result is therefore meant only to show that the halo geometry matches the video, not as a precise photometric comparison. As can be seen in Fig. 4, the simulation aligns fairly well with the observation. 

Fig. 4 Frame 303 with the simulation result

Elevation 10.37°, distance 195.8 km

In divergent light, the crystals that can send 22° halo rays have positions that form an elongated cigar-shaped locus between the eye and the source, known as Minnaert’s Cigar⁴. A flat cloud layer therefore cuts this locus in different ways depending on the source elevation and distance.

Divergent-light halos from natural high cirrus are rare; a case illuminated by a rocket reentry seems even more unusual. A closely related case was published very recently by Haussmann³, in which a bright fireball produced a 22° halo whose radius was shrunk to about 20° by divergent effect. In that case, however, the source was near the zenith, so the cirrus layer cut Minnaert’s Cigar almost straight above and the halo remained circular, only smaller. Here the source was low in the sky, so the cloud layer cuts the cigar obliquely. This is the slanted-cut case that Haussmann³ pointed to as probably not yet observed, and it distorts the halo rather than simply shrinking it. It should be noted that no measurement as accurate as Haussmann's was possible here, since there was no simultaneous parallel-light halo in the same frame to serve as a reference. 

The Hainan event is only one of many possible divergent-light configurations. As the source elevation, distance, and halo type vary, the Minnaert cigar is cut in different ways, producing shapes that have rarely or never been recorded. To show how sensitive these forms are, I include a few simulations below.

Fig. 5 shows an animated 22° halo with the source at 30 m distance and the cloud top at 100 m, the cloud base rising from 0 to 3 m, and the elevation set to 0°. The halo shrinks into a "light ball".

Fig. 5 22° halo, source distance 30 m, cloud top 100 m, cloud base rising from 0 to 3 m

Fig. 6 shows odd-radius halos, with an elevation of −1°, a distance of 100 m, and a cloud layer from 0 to 5 m.

Fig. 6 Divergent odd-radius halos

Fig. 7 shows the diffuse halo of raypath 3-5-6-7-3, at elevation −25.6°, distance 100 m, and a cloud layer from 1 to 30 m. The halo takes two forms, one surrounds the whole horizon, the other looks like a normal diffuse halo. Here the cigar subtends such a large angle that it looks like a "tomato" (Fig. 8). Raypath filtering was applied.

Fig. 7 Divergent diffuse halos

Fig. 8 A "cigar" with a radius of 140°

All these examples assume the crystals are arranged in infinite flat layers. In reality the crystals can be distributed in very different and interesting ways, such as on windshields, where the halo shapes depend dramatically on the lamp elevation and the windshield's position between observer and source⁵. Marko Riikonen has captured many such windshield cases, including elliptical halos and reflection-view displays⁶.

An interactive HTML file is provided, allowing the simulation to be viewed together with the video, with tools to let the user experiment with the Minnaert's cigars. The original video and the simulation images are also included in the link.

https://drive.google.com/drive/folders/1Z8f5_HufUhPua_f9Rq92FNDS316jD3Cz?usp=sharing

Acknowledgements. I thank the Chinese halo community for help in reaching the original poster, Sedimentary-Cloud (沉积云) for help with the divergent-light simulation, and SONG Xi Pei for help with the rocket data. The original video is used with permission from the observer. AI assistance was used in programming and some other parts of this work.

References

¹ Qiong Navigation Warning 82/26, Maritime Safety Administration of the People's Republic of China (Qinglan MSA), issued 3 June 2026. https://www.msa.gov.cn/html/cnmsa/hxaq/article/2026/808345637af44bf3831ab02000ecc176.html

² N. Lefaudeux, "La simulation des halos divergents / Divergent light halos simulation", opticsaround, 27 July 2013. https://opticsaround.blogspot.com/2013/07/la-simulation-des-halos-divergent.html

³ A. Haussmann, "Divergent light effects in ice crystal halos created by fireballs: a case study of a 22° halo with a 20° radius," Appl. Opt. 65, C42–C47 (2026). https://doi.org/10.1364/AO.582021

⁴ J. O. Mattsson, L. Bärring, and E. Almqvist, "Experimenting with Minnaert's Cigar," Appl. Opt. 39, 3604–3611 (2000). https://doi.org/10.1364/AO.39.003604

⁵ W. Tape, "Windshield Halos," in Streetlight Halos (2010), Chapter 13. https://scholarworks.alaska.edu/uaf_mathstats_facpubs/3/

⁶ M. Riikonen, "26/27 Jan 2026, Rovaniemi… elliptical halo," X. https://x.com/RiikonenMarko/status/2016261603444793506

Divergent light tangent arc on sparsely falling diamond dust

In Cluj-Napoca (Hungarian: Kolozsvár), Romania it is a common winter weather situation to have a dense layer of fog sitting on the city, with temperatures a few degrees below 0⁰C throughout the day with only the surrounding higher hills and mountains rising above this fog. High levels of humidity combined with low temperatures and aerosols floating in the air often results in industrial snowfall or sometimes (especially near the edge of the fog) diamond dust. On the evening of 2nd December we had the above described conditions but the weather was about to change, with gentle southern wind slowly sweeping out the fog.

Around midnight as I prepared to go to bed, I looked through my window and instantly noticed the unmistakable tiny sparkles of diamond dust under the streetlights. It was too sparse to form consistent halos (although the most intense moments produced visible lower pillars), but the sparkles hinted to the presence of the lower tangent arc. After a quick set up, I began taking videos of the area below the closest lamp (at 8-10 m from me, and approx. 8 deg. below eye level). The videos were around 4 min each, but many of them were rendered useless by the sudden drops in the amount of the diamond dust. Between two videos I also snapped a quick picture of a distant lamp that shows the pillar roughly as it was visible to the naked eye (see below).

A clearly visible pillar formed on diamond dust in the light of a distant lamp

On my computer I extracted every frame from the videos and saved them as jpg files (separately for each video), and stacked them together using a StarStax (a free startrail-making software) in ’lighten’ mode, to get the brightest value for each pixel in the resulting image (this is important as averaging the frames would wash out the individual sparkles that make up the halo). With this method I got an integrated image from each video that contained all the sparkles that appeared throughout that video. The videos were taken at various focal lengths ranging from 18 mm to 55 mm. When the first stacking results came out, I was simply delighted to see the lower tangent arc, and did not think much about it. However, on the next day I noticed something fishy: the tangent arc and the lamp comfortably fitted into the same field of view at 55 mm, which was not usual: this arc had to be way closer to the light source than a normal tangent arc. With the help of the solar disk (picture taken with the same equipment and settings, and resized to match the dimensions of the video), I measured and got approx. 7 degrees for the distance of the tangent arc from the lamp (see below).

Measuring the distance between the lamp and tangent arc. The solar disk on the top was used as a reference for the 0.5 deg. unit of the scale

This was completely new to me as I’ve never encountered any observations that describe tangent arcs so close to the light source, but I suspected the divergent light as the cause, since I knew that it can considerably alter optical phenomena compared to their familiar counterparts caused by paraller light rays. With the help of Walter Tape’s detailed article on streetlight halos (1), I confirmed that it is indeed the divergent light that caused the unusual appearance of the tangent arc, its distance from the lamp being normal for the lamp’s elevation relative to the horizon (since my observation of the below-horizon lamp with a lower tangent arc is equivalent to an above-horizon light source with an UTA at same angular distance from the horizon).

Based on Tape’s article, I also concluded that the strip connecting the lamp with the tangent arc is not a simple pillar, but part of the tangent arc. The slight widening at its center corresponds with the figures in the article.

On the following link you can watch a short piece taken from the video used for the picture on which the measurement was made, and the (sped-up) image stacking process using the frames from the original video: (2).

Beáta Ujvárosi