Explanation of the manuscript's Section 4: constructing the leading-order flow (pages 24 to 45)

Section 4, explained: Constructing the leading order flow (pages 24 to 45)

The problem this section is handed

The force is the momentum residual of the constructed flow, so the task is a blowup whose residual stays smooth through t = 1 [p. 3]. Section 2 pictures a slender self-similar vortex, spiraling inward and flowing out along the axis, joined across an annulus to a decaying exterior swirl; the join leaves an unbounded annular residual that shear-amplified ring pulses must cancel with their averaged momentum fluxes [pp. 3 to 6].

In Section 3's terms, pulses cancel the residual through the rθ and rz entries of their averaged covariance, as a positive combination of two families' covariance vectors [pp. 10, 12]. So Section 4 must produce profiles whose leading tangential residual is minus the radial divergence of a stress T that vanishes near the axis and in the exterior, points strictly inside the cone the two families span, and sits beside a residual-free exterior that can later be cut off [pp. 6, 9, 24]. Later sections need more: analyticity in η near the axis for Section 5 [p. 47], an edge weight for Section 6 [p. 69], positive shear bounds and a cone margin reaching the annulus edges for Section 7 [p. 74], and untouched intervals for Sections 5 and 8 [pp. 33, 34].

What a person would try first, and why it fails

An isotropic self-similar vortex. The natural guess is Leray scaling: lengths like τ^{1/2}, velocities like τ^{-1/2}. Then every ratio the construction runs on is of order one: axial over radial diffusion (q^{2h} in the paper), rotation over radial diffusion (q^{-h}), pulse wavelength over core radius (q^{h/2}) [pp. 4, 8, 11]. Nothing orders corrections or makes pulses short relative to the core (our reading). The paper's axial length τ^{1/2-h}, with swirl and axial speeds τ^{-1/2-h}, instead puts axial transport, radial transport and radial diffusion all at rate τ^{-1} [pp. 4, 5].

A symmetric axial profile. The paper names the failure: under reflection symmetry about z = 0, midplane axial transport of angular momentum vanishes, and u_z = 0 there leaves no axial shear to compensate [p. 5]. In Section 4's terms, with axis data U* = 4η the transport coefficient H* and the axial source Z* both vanish at η = 0, since the axis pressure is even (our reading of [pp. 35, 37]). The paper adds a small upward offset, U* = 4η + j0, and its pressure bounds make Z* positive where H* vanishes [p. 37].

A stress wherever the residual happens to be. Near the axis the swirl is locally a rigid rotation: from l = 1 + XF_X/F and a = 2 - 2l [pp. 26, 27], the swirl shear a vanishes at X = 0, so nothing there can amplify pulses (our reading). The stress must vanish on a core solving the stress-free equations (4.13) [pp. 27, 33]. At the far end, zero exterior residual is not enough: the stress is an integral from the axis and can still carry tails like r^{-2} and r^{-1} [p. 30].

The idea, motivated

One clock for two lengths (M4.1). Take radial length q^{1/2}, axial length q^D and tangential velocities q^{-A}, where A = 1/2 + h, D = 1/2 - h, and 0 < h < 1/100 is fixed [pp. 24, 32]; A + D = 1 keeps axial transport on the clock (our reading of [pp. 5, 26]). As z/τ^D would blow up at fixed z ≠ 0 (our reading), q is defined by τ = q(1 - η²), z = q^D η, keeping η in (-1, 1), with η = ±1 describing t = 1 away from the origin [pp. 7, 25]. Lemma 4.1 turns each time or axial derivative of q^b f(X, η) into a profile operator times an exact power of q [p. 25].

Three profiles, one free function (M4.2, M4.3). Write u_θ = q^{-A}E, u_z = q^{-A}U, r u_r = V0, p = q^{-2A}Π; since X = r²/(2q) is smooth in Cartesian coordinates, factoring √(2X) out of E and X out of V0 gives smoothness across the axis [p. 25]. Incompressibility fixes V0, and the leading radial equation is just pressure against centrifugal force, Π_X = E²/(2X); the rest of it, and axial viscosity, are smaller by q^{2h} and deferred to Section 5, whence its powers q^{2nh} [p. 26]. Free: E, U and the axis value of Π.

Any residual is a stress, but a nonlocal one (M4.4). Pulses supply stress divergences, so the tangential residuals must read -(∂_r + 2/r)T_θ and -(∂_r + 1/r)T_z [p. 9]. For an axisymmetric field this is free: multiply by r² or r and integrate from the axis [p. 27], with zero constants, since others blow up like X^{-2} or X^{-1} [p. 26]. The result is T = q^{-A-1/2}T0, T0 = F(p_s - s): p_s holds the integrated inviscid terms and the shear s = (a, -b_s) holds radial viscosity; zero stress means p_s = s, the pair (4.13) [p. 27]. The price is memory: T0 at X depends on the whole profile inside X.

What the memory is (M4.5, M4.6). This threatens any gluing, but Lemma 4.3 shows the memory is finite: beyond local values of U and E, p_s sees only the axis pressure and five cumulative integrals, axial momentum M, angular momentum I, axial transport of angular momentum J, axial momentum flux S and pressure increment Cp [pp. 28, 30]. By Lemma 4.4(i), profiles agreeing beyond X_h and sharing those five functions of η at X_h have the same pressure, radial velocity and stress beyond X_h [p. 29]. A join is five equations in η, after the axis pressure is fixed [pp. 28, 35].

The exterior first (M4.10, M4.11). The exterior must be residual-free with smooth limits at each fixed r > 0 as t → 1; the paper takes a pure swirl solving the radial heat equation exactly, E = c∞X^{-A}ℋ(2d/X) [pp. 6, 33], on our reading heat flow smoothing the power law similarity demands at large X. Normalizing the pressure to vanish at infinity fixes the axis datum Π0(η), so the exterior comes first [p. 35]. Against the tails of the third failed attempt, total moments vanish (4.28), so the stress integrates backward from infinity and is zero beyond X_b [pp. 30, 33, 36]. Four intervals, three pure power laws, are reserved for repairs [p. 35].

Which stresses pulses can make (M4.7). Two families supply T = c1v1 + c2v2 with c1, c2 > 0 and v1, v2 set by the local background [p. 10]. With P_c, J_c the products of p_s with (1, t_s) and (-t_s, 1), along and across the shear, t_s = -b_s/a, and v_s = a + b_s²/a, the admissible cone is v_s > 2, P_c > v_s and (v_s - 2)J_c² < 2(P_c - v_s)² [pp. 30, 31]. The middle inequality is T0·(1, t_s) > 0 [p. 32]: on our reading of the energy identity, pulses draw energy from the shear [p. 12]. The viscous waves need v_s > 2 [p. 31], positivity of the squared growth rate 2aF0²(1 - 2/v_s) [p. 74]. Our reading: with a = -rΩ'/Ω and b_s = W'/Ω (angular velocity Ω, angular momentum Γ = r²Ω, axial velocity W), v_s > 2 is Ω'Γ' + W'² > 0, the Leibovich and Stewartson combination, a cited precedent [p. 2], and without axial shear it is Rayleigh's condition that angular momentum decrease outward. By homogeneity in T0, the edges can carry the condition on T0/|T0| [p. 32].

The core (M4.12). Proposition 4.10 builds a stress-free, η-analytic flow near the axis from the offset data U* = 4η + j0, with an axis swirl shaped so that a regularized factor χ built from H* exceeds .99 wherever |Z*| is small [pp. 36, 37]; on our reading, this keeps one of the two amplification mechanisms active at every height [p. 5]. The core exits with v_s > 2 + c_ex, switches the stress on through a flat factor, and is continued until it equals the outer reference pair with all five integrals matched [pp. 37, 38]; Lemma 4.4(i) then joins it exactly [p. 40].

The gap and the tempting fix. After the join, the admissible cone holds near both edges but only the relaxed cone (P_c > 2 and v_s below a threshold U(P_c, J_c)) in between [pp. 31, 40]. Where the core meets the reference pair the swirl still rises like x^{1/10} and v_s < 1 [pp. 37, 38]; on our reading, angular momentum increases outward there. The tempting fix, steepening the shear, fails: the shear is a radial derivative, so a sustained order-one change across an interval moves E, U, the five integrals and p_s by order one, undoing the join (our reading).

Oscillate instead (M4.13, M4.14). The shear is made of radial derivatives, while Lemma 4.4(ii) bounds changes in p_s with no radial derivative of any difference [pp. 27, 29, 30]. So let the shear oscillate. Lemma 4.11 replaces it at each point by a periodic loop of shears, each admissible with p_s held fixed, averaging to the original shear; the relaxed cone, with admissibility near the ends, is its hypothesis [pp. 38, 39]. Our reading of the geometry: v > 2 is the outside of a disk in the shear plane, a nonconvex set; the loop runs on a circle of constant v above 2, and the original shear, inside it, is the loop's weighted barycenter. The loop is realized as E_N = E0 exp(𝒜(X, η, N log X)/N), U_N = U0 + ℬ(X, η, N log X)/N, with 𝒜, ℬ zero-mean primitives of the loop's deviation [p. 41]. The logarithm fits the operator X∂_X in which the shear is written: X∂_X(N log X) = N, so the modulation moves a and b_s by order one but values and integrals by O(1/N) [pp. 10, 27, 41]. Then p_s moves by O(1/N), inside the loop's strict margin, so the modulated profile is admissible at every point [p. 42], its stress oscillating in log X at frequency N.

Pay back the integrals, then read off the edges (M4.9, M4.15, M4.8). Bumps on the first power-law patch restore the five integrals exactly, since there the linear part of the moment map is a matrix of distinct-power integrals, invertible because a nonzero combination of m distinct powers has at most m - 1 positive zeros; a contraction handles the rest [pp. 34, 42]. The repair costs another O(1/N) in the shear quantities, and N is fixed last, after every radius, width and bump [pp. 40, 43]. The flat edge factorizations give a smooth unit direction up to X_a and X_b, a uniform margin κ, and a weight ζ built to vanish exactly as fast as the stress at both edges [p. 44]. Theorem 4.6 records all this, leaving two patches untouched for Sections 5 and 8 [pp. 32 to 34, 45].

What breaks without each move

These counterfactuals are our reading unless a page is tagged.

• M4.1 (q, Lemma 4.1): no exact powers, so no term is provably leading, and z ≠ 0 at t = 1 is an unbounded edge in η.
• M4.2 (factors √(2X), X): the velocity is not smooth across the axis, so neither is the force for t < 1.
• M4.3 (incompressibility, radial balance): a nonzero divergence or a leading radial residual survives outside the stress form, and the q^{2h} expansion is lost.
• M4.4 (radial integration): no stress for pulses to target; nonzero axis constants make it singular [p. 26].
• M4.5 (five integrals): matching means matching the whole interior profile, so no finite repair protects the exterior.
• M4.6 (joining lemma): without (i), gluing alters the exterior pressure, radial velocity and stress; without (ii), the shear change leaves p_s uncontrolled.
• M4.7 (admissible cone): Proposition 7.5 could need a negative squared amplitude; at v_s ≤ 2 the reference growth rate is not positive [p. 74].
• M4.8 (Theorem 4.6): later constants would depend on q or the band, and edge amplitudes would go uncontrolled.
• M4.9 (moment solve): the integrals cannot be restored, so repairs leak into the exterior; at λ = 0 two weights coincide [p. 42].
• M4.10 (heat exterior): a residual or a missing limit at fixed r > 0 breaks the spatial cutoff [p. 6], and the axis pressure is unknown.
• M4.11 (total moments): zero exterior residual still leaves r^{-2} and r^{-1} stress tails [p. 30].
• M4.12 (offset core): without j0 both amplification mechanisms fail on one layer [p. 5]; without the core, nothing regular and stress-free matches the exterior.
• M4.13 (prescribed mean): the primitives are not periodic, log E drifts by order one, and profiles, integrals and p_s all move.
• M4.14 (frequency N): the loop stays a family of shears no profile has; at small N the O(1/N) errors exceed the margin.
• M4.15 (repair, edges): the O(1/N) defect alters the exterior stress and heat tail; without Step 4, the edge direction, κ and ζ are missing.

Three questions the section leaves open

1. Is the modulation necessary, and is the relaxed cone its exact reach? The loop is needed only because the joined profile fails the admissible cone on a stretch [pp. 38, 41]. Could a profile be admissible on the whole annulus? Is the relaxed cone exactly the set of barycenters of admissible shears at fixed p_s? It matters because the stress's radial derivatives grow like powers of N, and the answer decides which joined profiles are repairable.

2. How small must h be? The order of choices forces h < e^{-T_d}, T_d = e^{M_d} + 10 [p. 35]: doubly exponentially small in the exterior's axial-decrease parameter [p. 36]. Is that forced by the mechanism or by this exterior construction? It matters because h sets the anisotropy τ^h, the swirl Reynolds number τ^{-h} and the pulse scale separation q^{h/2} [pp. 4, 11]; a moderate explicit h would make the rate quantitative and numerically checkable.

3. Does the core need analyticity in η? The stress-free equations mix parameter and radial derivatives, with a regular singular point at the axis [p. 145], and the core and Section 5's recursion both use an analytic inner collar [pp. 33, 47]. Would a Gevrey or finite-regularity class do, and is the core locally unique given its axis data? It decides whether the leading flow is robust or finely tuned.

Page tags

Pages cited, in order: 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 47, 69, 74, 145.

Extraction: displays on pp. 27, 31 and 40 to 42 were flattened; the fraction in (4.13) on p. 27 was read from context. No page was otherwise garbled.