Value chain · CLM-0863
In plain terms
A published figure suggesting that modelling a solar power satellite beam needs ten million million times more rays than the available software could handle turns out to describe the cost of one way of doing the sum, not the difficulty of the problem.
The claim as recorded
SRC-320's thirteen-order-of-magnitude scale gap - "problem sizes around 1 * 10^7 rays" against "1 * 10^20 rays" required - measures the cost of one EVALUATION STRATEGY and not the difficulty of the physics. The paper's own electromagnetic kernel is a surface integral: it states the propagation algorithm is "a derivation of the Rayleigh-Sommerfeld Diffraction Integral", cited to Goodman's Introduction to Fourier Optics p.44. So the comparison is not ray optics against wave optics; both that method and a transform evaluation compute the same class of aperture integral, and "ray" in the paper denotes a geometry query rather than a geometrical-optics ray. The acceleration the paper reports lies entirely in the triangle factor - "from O(Rays) * O(Triangles) to O(Rays) * O(sqrt(Triangles))" - and nothing in the paper reduces O(Rays); its stated route to the remaining gap is more hardware.
Evidence behind this: T0-analysis. Close reading of SRC-320's stated algorithm and complexity claims, plus a full-text search of SRC-320 and SRC-323 for the two figures and for any discretisation criterion. The convergence figures quoted below are calculated here rather than taken from the source; the arithmetic is CALC-087, verified independently from a cold start by a verifier sharing no code with the producer (VER-1090).
Scope and limits
This does not establish that a transform evaluation is sufficient for every SBSP geometry. Ray tracing represents terrain interaction and an aperture integral does not. The converged point counts are properties of the quadrature scheme, not of the problem - a cruder rule needs about a thousand times more points for the same answer - and they are resolved only to a factor of four, so the reported ratios must not be read as measured exponents. The deep case, sag 10,000 wavelengths, did NOT converge, and that was a memory ceiling on one machine rather than a property of the problem. A helix was never modelled, so CASSIOPeiA's own geometry remains untested. Polarisation, mutual coupling, element discreteness and near-field behaviour are all outside this work.
What it assumes
No assumptions are recorded against this claim.
Where it comes from
This is the project's own working, shown in full:
s2.2 verbatim: "The propagation algorithm selected as part of the SCOPES development is a derivation of the Rayleigh-Sommerfeld Diffraction Integral [12], with changes to support polarimetric (covering all field polarisations) modelling of radio propagation." Its [12] is Goodman, Introduction to Fourier Optics, p.44. s2.1 verbatim: "The available open source model (LyceanEM) was capable of handling problem sizes around 1 * 10^7 rays ... and modelling at the desired scale would require 1 * 10^20 rays [6]." s2.1 verbatim on the acceleration: "This approach allows the computational complexity to go from O(Rays) * O(Triangles) to O(Rays) * O(sqrt(Triangles))." s4 verbatim: "future development will be needed to support convenient parallelisation onto multi GPU high performance [computing resources]".
Related records
- Answers question
- QST-016 not published in this release
Cite this
energyaperture.com/c/clm-0863
CLM-0863, Energy Aperture, as of 2026-09-05.