Johan Gaume, then at the Swiss Federal Institute of Technology in Lausanne (EPFL), and Alexander Puzrin, of its sister institute in Zurich (ETH), published their model in January 2021. It answers four objections to the prosecutors' 2020 avalanche verdict: no avalanche signs when the tent was found, a slope too gentle, a release hours after the platform was cut rather than during the digging, and injuries unusual for avalanche victims. Their answer is a small wind slab, a plate of wind-packed snow roughly 5 m by 5 to 9 m and 0.3 to 0.5 m thick, that released from a buried weak layer, a fragile layer of snow crystals on which a slab can slide, inclined at 28 degrees on a local step above the tent, after wind-blown snow accumulating in the lee of the cut had loaded it for 7 to 13.5 hours; it slid 1 to 2 m onto the uphill half of the tent and crushed the sleepers against the ski-reinforced floor.

The equations in the paper's Methods were reimplemented for this page (the authors' code on Zenodo could not be downloaded), and they reproduce every published number to the quoted precision: wind-snow height 0.24 and 0.44 m, delay 7.2 and 13.5 hours, tension crack 4.95 m, slab width 8.8 m. The model is internally sound. The table below lists its inputs and what the record says about each.

What the replication shows

With the paper's own snow parameters, the delayed-release behaviour exists only in a band two to three degrees wide around the assumed angle. The release window is 24.7 to 29.7 degrees; the 9.5-to-13.5-hour window is hit at 26 to 28.5 degrees. At 22 or 23 degrees the slab never releases, however much snow the wind adds. At 25 degrees it releases only if the wind snow never bonds, and then after more than 150 hours. At 30 degrees it fails while the platform is being dug. The regime is sensitive to small changes in the parameters: cohesion of 300 Pa gives failure at the cut, 440 Pa the published delay, 450 Pa no release; 5 cm of slab thickness, one degree of slope or friction, or 30 kg/m³ of density flips the regime. Once in the delayed regime, the time to release is proportional to the deposition rate, which nothing measured constrains. The journal's reviewer made the same point: Dieter Issler, an avalanche scientist who reviewed the paper for the journal, wrote that the agreement between the computed and forensic delays was "coincidental", and the authors' revision "no longer claims that they have reproduced the delay" but back-calculates the deposition rate from it.

Impact. A simple spring-and-damper model of the chest, calibrated as in the paper, gives 20 to 70 per cent chest compression for 50 to 200 kg blocks at 2 m/s on a person lying against a firm floor; the paper's 28 to 34 per cent falls inside that. Contact pressures are 25 to 50 kPa (kilopascals; the atmosphere presses at about 100), two to three orders of magnitude below what breaks skin. So "multiple rib fractures with no external wound" is what a broad, heavy, slow load does, and it is equally what a collapsing snow roof, or three to four metres of static snow, does to someone lying on a stream bed. The injury physics cannot tell the tent from the ravine.

The skull. The one injury the chest physics does not cover is Thibeaux-Brignolle's: a depressed fracture over 9 by 7 cm of the right temporo-parietal bone, with a piece 3 by 3.5 cm driven onto the dura (л.д. 353). Fractures of that region in cadaver tests need 2.5 to 10 kN (kilonewtons; one kilonewton is roughly the weight of 100 kg), mean 5.2, delivered through a flat 5 by 10 cm plate or a 2.5 cm disc (the 1991 tests of D. L. Allsop and colleagues; a 2004 review by N. Yoganandan and F. A. Pintar gives 5.6 to 9.9). A block of snow can transmit no more force than its crushing pressure times the area it bears on. Over the whole 63 cm² of the depression the 2.5 kN floor needs 400 kPa and the 5.2 kN mean 800; over the 50 cm² test plate, 500 kPa; over the 10.5 cm² bone piece alone, 2.4 MPa, a pressure reached by ice or rock and by no snow. Where the crushing pressure of the snow on that slope lies is the open question. Gaume and Puzrin's own impact simulation gives the wind slab a yield pressure of 100 kPa and the older snow 30 kPa, hardening as it compacts (their Methods, Eq. 48); at 100 kPa a slab face bearing on the entire depression delivers 0.6 kN, a factor of four short. Malcolm Mellor's 1975 review of snow mechanics, compiling tests of dry, coherent snow under rapid laboratory loading reads, from his Fig. 17, as about 300 to 700 kPa at 0.40 g/cm³ and 500 to 950 at 0.46, the densities measured at the site; the same figure puts tensile strength at 0.40 near 190 kPa, twenty to thirty times the 6 to 10 kPa the paper assigns its slab, so a young natural slab sits at or below the bottom of that band. Pairing the hardest of those figures, 300 kPa, with a 100 cm² contact larger than the fracture gives 3 kN, which overlaps the threshold: the numbers do not exclude snow. What the arithmetic supports is narrower: a soft or medium slab cannot have made this fracture by itself; snow could have made it only as a hard, well-sintered slab or crust bearing on most of the depression with the head backed by something firm, or through a hard object between the snow and the temple. The examiner's own description, a deep multi-fragment depression with one piece driven in, reads as a concentrated load, which is why Evgeny Buyanov, the engineer whose 2011 book puts the injuries at the tent, puts a camera under his head, why the paper's supplement adopts that reading ("high concentrated force"), and why this finding names a stone or an ice edge in the stream bed. That is an interpretation rather than a measurement: the piece of temporal bone taken at autopsy for study (л.д. 354) was never reported on, and nobody has examined the margins since. The constraint holds equally in the tent and in the ravine, and it does not change any of the percentages. Arithmetic: code/physics/skull_constraint.py.

The follow-up evidence. Puzrin and Gaume returned in March and September 2021 and January 2022, and reported in a 2022 Comment: a 9 cm drone model showing "an average slope of about 20 degrees" with metre-scale steps over 30; two slab avalanches on an eastern slope less than 3 km from the tent on 29 January 2022, the smaller one "practically invisible after less than an hour" under wind-transported snow; independent expert work for the prosecutors (Popovnin's 2019 snow measurements, Pigoltsina's computed 9 to 12 m/s wind and more than 150 cm of snow above the tent) consistent with their assumptions. A further slab was photographed by Dmitriy Borisov, a hiker who visits the pass, about 700 m from the tent on Kholat Syakhl itself on 7 January 2023. These observations settle three things: that steps steeper than 28 degrees exist above every candidate tent position, so the model's terrain requirement is met at the scale of the steps; that slab avalanches occur in the area; and that small ones vanish within an hour. They do not show the snow requirement, a weak layer lying on such a step in February 1959, and the avalanches observed were on a different slope under different conditions, as the authors say: the slab there "was softer, the slope was steeper, it was not undercut from below and the trigger was probably different".

The counter-case, and what survives of it

The strongest opponents are people who have examined the slope and its snow in winter. Vladimir Borzenkov, an aviation engineer with six winter expeditions between 2012 and 2019, measured the snow surface at 16 to 17 degrees along its whole length with about 20 degrees in the last 5 m under the shoulder, snow density 458 kg/m³, depth 0.8 m at the tent rising to 2.2 to 2.5 m twenty metres upslope, hard crust throughout, and in four winters of pits found no weak layer and no depth hoar; a shear test in March 2019 gave a strength more than ten times the driving stress. Vladislav Karelin, a 1959 searcher, showed that the prosecutors' trace-evidence expert in 2020 reversed cause and effect on the tent poles, and that the entrance pole stood with its guy lines intact while the poles for the rear guys were still in place. Sergey Sogrin, also a searcher, testified in 1959 that the slope was 15 to 18 degrees and later that he looked specifically for depth hoar under the crust and found none.

Scored against the replication: Borzenkov's density, depth profile and wind climatology survive as measurements, and the last two argue against the model rather than for a bigger avalanche, since denser snow and stronger wind shorten the modelled delay to minutes. His 16 to 17 degrees is the winter snow surface, which is the smoothed average and does not speak to the ground steps beneath it. His pits show that in four recent winters the snowpack was nowhere near failure; they cannot show what it was in 1959. His claim that a slab would have left debris does not hold: a 12 to 20 m³ slab stopping on the flat platform leaves 0.3 to 0.5 m that 25 days of wind reworks, and the 2022 observation of a trace erased within an hour is direct. Karelin's point about the poles is correct as mechanics but does not discriminate: a half-collapsed tent with the far guys torn is what either a small slab or a 20 to 25 m/s gust leaves. The objection that the footprints start 15 to 40 m below the tent does not discriminate either; hard crust records no prints at the tent, with or without a slab.

The decisive question is therefore not the slope, which the drone survey has answered in the model's favour, but the snow: whether a buried weak layer lay on one of those steps for about 5 to 7 m above the cut in early February 1959. The winter observations at the site point in different directions. Popovnin, measuring for the prosecutors in March 2019, found snow properties consistent with the model's assumptions and a stone ridge under the tent site, and still told Komsomolskaya Pravda that the version needed «чёткие доказательства. Не аргументы, а доказательства», clear proofs, not arguments, and had none. Borzenkov, from pits cut to the ground in 2010, 2014, 2015 and 2019, the first by an expedition he briefed but did not join, found strong bonding between every layer, no depth hoar, stones protruding 10 to 40 cm from the ground under the pack, and in a March 2019 shear test on the visually weakest deep layer a strength more than ten times the driving stress; Sogrin recalls cutting sections in 1959 and finding none. None of these was a pit line up the step above the fixed 1959 position, so none settles it. On the steps themselves the two sides' figures are close: Borzenkov concedes steps of 28 to 30 degrees but puts them at no more than 5 to 6 m long, while the paper's 4-to-6 m height implies 8 to 12 m of slope length; the modelled slab is 5 m long and its stressed zone 7 m, so on his figures the terrain is marginal and on theirs ample. The paper's mechanism is possible on this slope. Whether it occurred at the tent on 1 February 1959 depends on the snowpack, which was not measured; on the delay, which is fitted rather than predicted; and on the injuries, which cannot be placed.

Two points follow from this. The positive case for the slab, terrain met and slabs seen nearby, does not place the crushing injuries at the tent: the model's own impact calculation is satisfied as well by a collapse in the ravine. And the objections to the slab, the unmeasured weak layer and the fitted delay, do not place them in the ravine: if the slab is wrong the injuries still need a cause, and the ravine collapse is argued from the examiner's statements on mobility and survival time (л.д. 381–383) and from Nikitin's 2019 reading of the bones; the slab's failure plays no part in it. The 75 per cent for the ravine depends on those statements.