Kinetic theory with a relaxation time. Using a Debye spectrum with one effective sound speed v, κ = (kB/2π²v)(kBT/ℏ)³ ∫0θ/T x⁴eˣ/(eˣ−1)² τ(x) dx with x = ℏω/kBT. Scattering rates add (Matthiessen): 1/τ = B ω² T e−θ/3T + A ω⁴ + (v/D)(1−p): Umklapp, isotope/point defects, and the wire surface. Here v = 6000 m/s, n = 5×10²&sup8; m-3 (so θ = 658 K), A = 2.4×10-45 s³ and B = 3.8×10-19 s/K, chosen so the bulk value is close to silicon's ~150 W/m/K at 300 K. These are illustrative fit values, not a measured material table.
Boundary limit. If every phonon scatters only at a diffusely reflecting wall, the mean free path is the mean chord 4V/S = D for a cylinder (Cauchy). Then κCas = (1/3) C v D = 2π²kB⁴T³D/(15ℏ³v²) at low T (Casimir 1938). The Monte Carlo panel measures this mean free path directly: random flight in the wire cross-section, diffuse (cosine-law) re-emission with probability 1−p, mirror reflection with probability p.
Simplifications. One averaged phonon branch (no dispersion, no separate transverse/longitudinal modes), no Normal-process correction of the Callaway model, wire modelled as an infinite circular cylinder of diameter D (the 3D model is hexagonal for looks), no phonon confinement or surface oxide, and the 1−p factor is the simple diffuse-fraction form, not Ziman's (1+p)/(1−p). Real silicon wires of ~50 nm measure far below bulk (tens of W/m/K), which this model reproduces only in order of magnitude.
Models and textures were built in Blender and baked with Cycles; Three.js r160 renders them.