Where the curve was measured
The strength of loose asteroid material was tested here on a small object: a bridge of cohesive grains holding between two boulders a metre across, pulled apart in a particle-dynamics simulation until it broke. Across the runs the peak stress fell from about 0.005 pascal to 0.001 pascal as the mean grain diameter grew from 2.5 to 4.5 centimetres, and rose from 0.0008 pascal to 0.0024 pascal as the grains were flattened. For spherical grains, three independent runs at each size returned closely matching values. That is an internal check worth saying out loud.[1]
What carries that curve to an asteroid is a formula. The team writes the failure stress as the connectivity number times the solid fraction times the contact cohesion force, divided by the square of an equivalent grain diameter — a modified Rumpf relation, and the data collapse onto it well. The extrapolation happens in that inverse-square term: the measured range stops at 2.5 centimetres, while the answer quoted for Bennu belongs to grains of a millimetre and less. The relation permits the step, and the simulations do not cover it.[1]
The inputs that were not measured
The model is then fed numbers from the returned samples: an average particle size of 1.2 millimetres, a solid fraction of 69 per cent to 78 per cent, an internal friction angle of about 32 degrees calculated from photographs of regolith pouring out of TAGSAM, and particle-to-particle cohesive forces measured between 0.5 and 4 nanonewtons. Two of the inputs are softer than the rest. The connectivity number was never measured; the paper gives limits instead and writes that the technique needed to measure it was not available for these samples. The 1.2 millimetre average also leaves out fines below the five-pixel imaging threshold, which by the authors' account were micrometres across.[1]
The lower bound on grain size comes from the same direction. Earlier work put the site's cohesive strength below 1 pascal; experimental and theoretical results put cohesive strength at 2 to 4 times tensile strength; dividing brings the surface tensile strength to at most 0.5 pascal, and reading that back through the inverse-square law gives an equivalent diameter above 31.0 micrometres. So the agreement with a strength under 1 pascal partly returns a value that was put in. The check that is not circular sits elsewhere: reconstruction of the TAG crater gives about 0.001 pascal, the model turns that into an equivalent diameter near 0.5 millimetres, and that lands close to the 1.2 millimetre average measured on the samples. That was the test the model could have failed.[1]
The measurements that would move the number
The authors name the limitation that matters most themselves. Their bridges use a narrow size dispersion, the range in which a mean diameter and a connectivity number stay well defined. Bennu's surface is broadly polydisperse, with a particle size distribution slope of about minus 2.12. The one detailed study they cite on cohesive, broadly distributed material finds effective tensile strength rising with dispersity even when the mean particle diameter is unchanged. A mean is doing work here that a mean may not be able to carry.[1]
The way out is written in the paper: direct experimental measurement of the connectivity number, the solid fraction and above all the contact cohesion force, on as many asteroid samples as possible. Once those measurements on the returned Bennu material become public, the strength band quoted for the surface moves off the bounds it currently borrows, and the signal to watch is whether a new estimate can fix the equivalent diameter without leaning on the earlier values below 1 pascal. Meanwhile the raw simulation data sit on Zenodo, the scripts on GitHub and the solver is open source. That is enough for someone else to build the same bridges and arrive at the same numbers, which is more than most modelling papers offer.[1]