“…The Laplacian flow solution ϕ(t) and the bracket flow solution μ(t) have the same maximal interval of time existence, say (After some preliminaries on G 2 geometry in Section 2, we adapt in Section 3 the machinery developed in [24] to the Laplacian flow case; the results so obtained include (see Sections 3.3 and 3.4 for more precise statements).(i) The norm |Δ ϕ(t) ϕ(t)| ϕ(t) of the velocity of the flow must blow up at any finite-time singularity (compare with [28, Theorem 1.6]).(ii) If μ(t) converges to a Lie bracket λ, as t → T ± , and there is a positive lower bound for the (Lie) injectivity radii of the G-invariant metrics g ϕ(t) on G/K, then (G λ /K λ , ϕ) is a Laplacian soliton (possibly non-homeomorphic to G/K) and (G/K, ϕ(t)) converges in the pointed (or Cheeger-Gromov) sense to (G λ /K λ , ϕ), as t → T ± .(iii) The following conditions on a simply connected (G/K, ϕ) are equivalent.(a) The operator Q ϕ such that θ(Q ϕ )ϕ = Δ ϕ ϕ satisfiesthat is, (G/K, ϕ) is an algebraic soliton.where X D denotes the vector field on G/K defined by the one-parameter subgroup of automorphisms of G attached to the derivation D (in particular, (G/K, ϕ) is a Laplacian soliton).The concept of algebraic soliton has a long and fruitful history in the Ricci flow case, due perhaps to its neat definition as a combination of geometric and algebraic aspects of (G/K, ϕ). It has also been a useful tool to address the existence problem of soliton structures for general curvature flows in almost-Hermitian geometry (see [23]), the symplectic curvature flow (see [9,26]) and the Chern-Ricci flow (see [25]). As in any of these cases, a natural question is how special are algebraic solitons among homogeneous Laplacian solitons.…”