(124), p. 35 — E(α, R) = (1/δ)(1 + dα, cα(1 + dα); −bα, 1 + aα − bcα²) and 0 < det E(α, R) = 1/δ = 1/(1 + |tr R|α + adα²) < 1
ProvedNAGFlow.GSSpectrum.eq_124gauss-seidelmatrix-splittingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1spectral-radius
Let and
with . Split with its lower triangular part and . For every step size , with ,
and
This is the explicit form of the Gauss–Seidel step for the scalar model, from which Lemma A.1 and the spectral bounds for and are derived.
Formalization Note. The paper leaves implicit (it is the step size of (41)); it is a hypothesis here. and are obtained from by the lower-triangular split of the definitions file, which on this gives exactly the printed matrices.
Preamble
import Mathlib import Definitions.Def_NAGFlow_GSSpectrum_Setting
Formal statement
namespace NAGFlow.GSSpectrum
open Matrix
/-- (124) and the display after it, App. A, p. 35. For R = (−a c ; −b −d) with a, b, c, d ≥ 0 and
tr R < 0 < det R, and a step size α > 0, the Gauss–Seidel matrix E(α, R) = (I − αM)⁻¹(I + αN)
(M the lower triangular part of R, N = R − M) equals
(1/δ)(1 + dα, cα(1 + dα) ; −bα, 1 + aα − bcα²) with δ = (1 + aα)(1 + dα), and
0 < det E(α, R) = 1/δ = 1/(1 + |tr R|α + adα²) < 1. -/
theorem eq_124 (a b c d α : ℝ) (ha : 0 ≤ a) (hb : 0 ≤ b) (hc : 0 ≤ c) (hd : 0 ≤ d)
(htr : (Rmat a b c d).trace < 0) (hdet : 0 < (Rmat a b c d).det) (hα : 0 < α) :
ER α (Rmat a b c d) = (1 / ((1 + a * α) * (1 + d * α))) •
!![1 + d * α, c * α * (1 + d * α); -(b * α), 1 + a * α - b * c * α ^ 2] ∧
0 < (ER α (Rmat a b c d)).det ∧
(ER α (Rmat a b c d)).det = 1 / ((1 + a * α) * (1 + d * α)) ∧
1 / ((1 + a * α) * (1 + d * α)) = 1 / (1 + |(Rmat a b c d).trace| * α + a * d * α ^ 2) ∧
1 / (1 + |(Rmat a b c d).trace| * α + a * d * α ^ 2) < 1 := by sorry
end NAGFlow.GSSpectrum
Source
Luo & Chen, arXiv:1909.03145v4, App. A, proof of Theorem 2.1, (124) and the display after it, p. 35
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.