Mars 3

Mars 3 flew the same hardware as Mars 2 and had the same job, and it got the part its twin had failed. On 2 December 1971 the descent module separated on the approach hyperbola at 09:14 UT, fired a solid motor fifteen minutes later to add about 120 metres per second and drop itself out of the flyby, spun up for stability and coasted for four and a half hours while the orbiter manoeuvred clear. Entry came at about 5.7 kilometres per second at an angle under ten degrees. The conical shield did the bulk of the braking, a drogue chute pulled out a reefed main chute that opened fully once the module dropped below the speed of sound, the shield was jettisoned and a radar altimeter woke up. At a few tens of metres the parachute was cut away and shoved aside by a small rocket so it would not settle over the station, the soft-landing engine fired, and the sphere separated and dropped the last stretch into thick foam padding. It hit at a reported 20.7 metres per second at about 13:50:35 UT. Four petals opened and righted it. Whatever else is disputed about Mars 3, this part is not: it survived, on a planet that had never let anything survive before.

Then it spoke, and the trouble started. Ninety seconds after touchdown, on schedule, the station switched on both radio channels and both Ya-198 panoramic cameras together and began scanning. What came back was a grey field with nothing in it. Perminov, who was in the room, writes that the signal disappeared in 14.5 seconds, that the same thing happened on the second telephotometer, and that they never worked out why two cameras on independent bands would fail together within a hundredth of a second. NASA's catalogue says twenty seconds, from 13:52:05 to 13:52:25. A 1973 Soviet report gives 19.7 seconds. The 2024 reanalysis by Gektin and Kostachuk argues that all of these measure the same thing, the length of the fragment with a properly formed video structure and a scanning-mirror retrace pulse in it, and that the surviving print of the signal runs at least 62 seconds. Their reading of the record is stranger than either number suggests: for the first twenty seconds or so both cameras were in an abnormal mode, putting out one or two brightness levels instead of the designed 64 and no retrace pulse at all; then some unidentified electrical event occurred and the cameras produced about nineteen seconds of normal video, 77 lines from one and 78 from the other; then the quality collapsed again and the signal effectively vanished.

One correction is worth making before anything else, because it changes the shape of the story. None of this was live. The lander talked to the orbiter, which was simultaneously performing its own orbit insertion burn, recorded the signal on magnetic media, and replayed it to Earth at one quarter speed during sessions between 2 and 5 December. Every dramatic retelling of engineers watching the picture from Mars die on the screen is describing a playback of a recording made hours earlier. That matters because it widens the list of suspects. The Soviet explanation, and the one NASA still repeats, is corona discharge: Perminov found that British radio operators in the Lebanese desert had suffered transmitter failures in dust storms during the Second World War, and reasoned that the same thing could happen on Mars in the middle of a planet-encircling storm. Other candidates on the record are a station tipped over or draped by its own parachute, battery damage, and a touchdown with dangerous horizontal velocity into winds Mariner 9 measured at more than 140 metres per second. And there is a fifth, which cannot be dismissed: the relay itself. NASA's own page concedes that it is not known whether the fault originated with the lander or with the orbiter's communications relay, and the orbiter that night was busy, mispointed after its burn, and had already lost three-axis stabilisation once to contamination on a star sensor.

What was actually in those 79 lines is now a live scientific argument rather than a settled disappointment. The original magnetic recordings are, as far as anyone knows, lost. What survives are scanned phototelegraph prints kept by Arnold Selivanov, who built the cameras, plus a longer and worse fragment that appeared in a BBC documentary. Gektin and Kostachuk rotated the signal ninety degrees, stripped the retrace region, rejoined the split halves of each line, mirrored the print and stretched it, on the assumption that the cameras were running in the faster 1,500-line survey mode, which sweeps a full circle in a little over six minutes, rather than the 6,000-line detailed mode, which would have taken about twenty-five minutes and outlasted the eleven to sixteen minutes the orbiter was above the horizon. Out of that came two small fragments that look like uneven rocky ground, shot through an orange filter and a yellow-green one, with non-overlapping fields of view. Their reconstructed geometry has the station sitting tilted by fifteen to twenty degrees toward the cameras, which would put the horizon above the field of view and leave both cameras staring at dirt between roughly 0.9 and 3.8 metres away. One dark patch might be a dent left by a petal opening. It is a plausible story that explains why Soviet analysts, who knew perfectly well they had camera signal, could not recognise a textured scrap with no horizon, no rock outlines and no piece of their own spacecraft in it. It is also a chain of assumptions applied to a much-copied paper print, and the authors of the 2026 review that summarises it say so.

The lander site itself was found, probably. In late 2012 Vitaly Egorov, who ran the largest Russian-language Curiosity community online, divided a 1.8 billion pixel HiRISE image of the landing ellipse taken in November 2007 into twenty pieces and handed them to volunteers. Phil Stooke told him the flight had been to the east, which cut the search area in half. The group found four objects with the right sizes, shapes and relative arrangement for a parachute, heat shield, retrorocket and four-petalled lander. Alexander Basilevsky put the case to HiRISE principal investigator Alfred McEwen, a follow-up image was taken on 10 March 2013 at different lighting, and Lavochkin engineer Vladimir Molodtsov dug the retrorocket chain length out of the archives at 4.52 metres against 4.8 metres measured on Mars. NASA announced it on 11 April 2013 with McEwen calling it a remarkable match while stating that alternative explanations could not be ruled out. That caveat has never been lifted. The lander is a few pixels across and each petal is two or three, so the imagery cannot settle the tilt, the camera pointing or anything else the signal reconstruction depends on. In August 2013 the IAU named three craters there, Reutov, Tyuratam and Belyov, which is as close to an official acknowledgement as an unconfirmed find gets.