Home·The theorem·The proof in one reading·Architecture·Units·Manuscript·Glossary·FAQ·Work items·Data·Method·Provenance·Changes
← III.8 Connected activities and backward dependence conesPart III·Detailed notesIII.10 One full nonlinear renormalization step: bounds on irrelevant terms →

III.9. The marginal flow: the running coupling and its reset

Contents

1. What this unit proves 1

2. What it uses 4

3. Strategy 4

4. Steps 4

5. Where the difficulty lies 5

6. Notation 5

7. What remains unexplained 5

Code dependencies 5

Lean source 5

Additional source locations 6

Key statements 12

Part III — Ultraviolet construction (renormalization group).

Tracks the running coupling and resets the effective density after each renormalization step.

Status: edited. Condensed from the unit's detailed notes, which were checked against the linked Lean in two rounds and edited; the local limitations still apply. Status meanings.

Overview · All units · Detailed notes · ← Previous: III.8 · Next: III.10 →

Results · Inputs · Strategy · Proof steps · Notation · What remains unexplained · Code dependencies · Lean source · Key statements

Source inventory and conventions

Part: III, Ultraviolet construction (renormalization group). Lean: unit III.9: coupling reset, verified reset, Wilson start-up. Size: 7,215 lines.

1. What this unit proves

Let \(G \subset U (n)\) be closed, with Lie algebra \(\mathfrak{g}\).

Proposition III.9.1 (slope). For simple nonabelian \(\mathfrak{g}\) with orthonormal basis \((b_i)\) in \(\langle X, Y \rangle = - \text{Retr} (XY) / n\), define \(C_G = - (\dim \mathfrak{g})^{- 1} \sum_i \text{tr}_{\mathfrak{g}} ((\text{ad} b_i)^2)\) and \(\kappa = 11 C_G / (24 \pi^2)\). Then \(\kappa > 0\). Status: definition and positivity theorem. Lean: slopeConstant_pos.

Proposition III.9.2 (reference scales). Fix \(\kappa > 0\), \(\bar{\beta} > 1\), an infinite sequence of odd \(L_m \geqslant 3\). Set

\(\displaystyle s_m = \sum_{i < m} \log L_i = \log (1 / a_m), \quad \hat{\beta}_m = \bar{\beta} + \kappa s_m, \quad M_m = e^{(\log \log \hat{\beta}_m)^2},\)
\(\displaystyle t_m = \hat{\beta}_m^{- 1 / 2} M_m, \quad \varepsilon_m = \hat{\beta}_m^{- 11 / 10}, \quad \tilde{\varepsilon}_m = e^{- M_m^{1 / 5}}, \quad B_m = \lceil (\log \hat{\beta}_m)^2 \rceil, \quad \mathfrak{B}_m = [\hat{\beta}_m / 2, 2 \hat{\beta}_m] .\)

Then \(\hat{\beta}_{m + 1} = \hat{\beta}_m + \kappa \log L_m\), \(\hat{\beta}\) is nondecreasing and diverges. Additional scale hypotheses are: \(t_m \leqslant 1\); sides at least \(100 (B_m + 1)\) and \(100 (R_m + 1)\) at levels \(m, m + 1\), for every volume; \(J_m \geqslant 4\) nondecreasing, divergent, increasing by at most one; nondecreasing derivative orders, spatial powers, support exponents and stage factors \(\ell_j\), with \(L_m = \ell_{J_{m + 1}}\); antitone \(0 < \theta_{J_m} \leqslant 1 / J_m\); \(\log L_m / s_m \to 0\); \(\omega_{J_{m + 1}} / \log L_m \leqslant J_{m + 1}^{- 2}\) and \(\log (1 + \text{paths} + \text{width}^{- 1} + \text{columns} + \text{relaxation} + \text{support exponent})_{J_{m + 1}} / \log L_m \leqslant J_{m + 1}^{- 2}\). Status: definitions and elementary theorems; scale hypotheses supplied by III.2. Lean: Calendar.Fits.

Proposition III.9.3 (selected histories). A selected history assigns representations \(\rho_{N, p, m}\) for all \(N, p\) and \(m \leqslant N\), with volume-independent paths \(P_{N, m}\), level norms and clip, stage path bounds and loop cap \(2 M_m^{\theta_{J_m}}\), and \(\beta_{N, p, m} \in \mathfrak{B}_m\). Regular activities are local, gauge invariant, vanish at the flat field, with zero affine Hessian on nonwrapping supports and bound \(\varepsilon_m\); forced activities obey II.2 with bound \(\tilde{\varepsilon}_m\); the family respects translations and signed permutations. At the top, \(v_N = 0\), \(\rho_N = e^{- \beta_N S_W}\); at the bottom, \(\beta_0 = \bar{\beta}\). For every \(m < N\),

\(\displaystyle \rho_m = T^{(m)}_{\beta_{m + 1}} \rho_{m + 1}, \qquad | \beta_{m + 1} - \beta_m - \kappa \log L_m | \leqslant \omega_{J_{m + 1}} .\)

For every \(\epsilon > 0\), eventually \(| \beta_{N, p, m} - \hat{\beta}_m | \leqslant \epsilon s_m\) uniformly in \(N \geqslant m, p\). Consequently \(\beta_{N, p, m} / s_m \to \kappa\) uniformly, including \(m = N\), and \(\int \rho_{N, p, m} \hspace{0.17em} d \text{Haar} = Z_W (\Lambda_{N, p} ; \beta_{N, p, N})\) for every \(m \leqslant N\). Status: specification; realized in III.13 using III.9.5 and III.10. Lean: History.Selected.

Proposition III.9.4 (reset along a curve). Fix scales meeting entry thresholds and kernel conditions, compatible horizon, Haar chart, positive post-block constants, planned kinetic data and recipe. For \(N, p, m < N\), let \(\rho'_x\), \(x \in I = [l, u]\), be continuous admissible fine representations, with \(l \leqslant u\); write \(\beta'_x = \beta (\rho'_x)\). Fix stage-pair reset constants \(c > 0\), \(0 < r \leqslant 1\), prescribed fine-stage \(g > 0\), and coarse-stage sensor \(\eta > 0\). Assume:

\(\displaystyle | (\beta'_x + \alpha)\mathcal{K}(V) +\mathcal{L}(V) - \log (H_x (1) / H_x (V)) | \leqslant | \Lambda_m | c \varepsilon_m,\)
\(\displaystyle |\mathcal{E}_{F (x, b)} (V) -\mathcal{E}_{F (x, a)} (V) | \leqslant | \Lambda_m | c \varepsilon_m |b - a| .\)

There is continuous \(d : I \to [- r, r]\) giving continuous admissible \(\rho_x = H_x\), coupling \(\beta'_x + \alpha + d (x)\) and \(v_x = | \Lambda_m |^{- 1} \log H_x (1)\), retaining the stronger forced bound. Without the last bullet, the unique residual root still matches \(1, V\). Status: theorem. Lean: advance.

Proposition III.9.5 (explicit reset; program). Under III.9.4's datum, curve and constants, assume simplicity, \(\beta'_x \in \mathfrak{B}_{m + 1}\), and the early presentation

\(\displaystyle H_x = e^{\tau_x | \Lambda_m | - \beta'_x \mathcal{K}-\mathcal{L}- \sum_i \chi_{X_i} E_{x, i}} \mathcal{G}_x e^{- \alpha \mathcal{S}} .\)

Assume local gauge-invariant smooth regular functions with nonnegative sup-jet bounds \(h_{x, i}\) satisfying, at every anchor, \(\sum_{i : o_i = o} e^{(a + 3) \text{cover}_{B_m} (X_i)} h_{x, i} \leqslant H_x^E\), forced-family conditions and site-pinned bound \(w_x\) at exponent \(a + 1\), translation and point-group symmetry, fixed addresses and III.9.4's continuity/tails. Require priced reset, compatible clip, width and cubic conditions, \(H_x^E \leqslant c \varepsilon_m / 4\), regular/forced reorganization estimates fitting numerical bounds \(e_x, k_x\) that imply both allowances and \(e^{- gM_{m + 1}^2}\), and recipe acceptance and continuation. Then

\(\displaystyle d (x) = \delta_x, \quad t_m^2 | \delta_x | \leqslant H_x^E, \quad v_x = \tau_x - | \Lambda_m |^{- 1} \sum_i E_{x, i} (1) .\)

All III.9.4 conclusions hold. A Gaussian-sector formulation additionally requires integration licenses and equality of that integral with the push. If these inputs hold at every level along every continuous admissible curve with couplings in \(\mathfrak{B}_{m + 1}\), and start-up curves pass initial and retention tests, they give, for every \(N, p\), an admissible continuous Wilson curve on \(\mathfrak{B}_N\) with coupling parameter \(\beta\) and \(v = 0\), and reset outputs along every continuous admissible input curve with couplings in \(\mathfrak{B}_{m + 1}\) on every closed interval, for all \(m < N\). Output couplings need not remain in the coarse band before shooting. Status: conditional theorems; physical inputs supplied by III.10. Lean: Return.reset, program.

Proposition III.9.6 (Wilson start-up). For compatible clip/norm pairs (\(0 < t \leqslant 1\), \(c_{\text{dir}} \leqslant 1\), \(B \geqslant 2\), \(te^{(B - 2) / c_{\text{rel}}} \geqslant 2\)), Wilson top paths, point-group-averaged regular activities \(\beta (1 - c_{\text{dir}}) R_p\) on small-field plaquettes and forced activities \(e^{- \beta \hspace{0.17em} \text{charge} (D, U)}\) on padded large-field components give \(\rho^{\text{start}}_{\beta} (U) = e^{- \beta S_W (U)}\) for every real \(\beta\) and configuration \(U\), with \(v = 0\). For every \(G, k, q\), some \(C > 0\), \(0 < \delta \leqslant 1\) give residual sup-jet bound \(Ct^5\) for every norm choice of order \(k\) and power \(q\), every torus, plaquette and site of that plaquette when \(t \leqslant \delta\). At each fixed stage with \(c_{\text{dir}} \leqslant 1\), some \(\beta^{\ast}\) ensures compatibility, small threshold and both norm allowances for every realized choice of levels and every \(N\) at that stage with \(\hat{\beta}_N \geqslant \beta^{\ast}\). The band curve is continuous; admissibility additionally requires Wilson top paths, their stage bounds and recipe tests. Given old planned data and a revised horizon with Wilson top quadratic datum, unchanged lower data and initial guesses, bounded Wilson top paths can replace the old top while preserving \(\alpha\) and planned conditions. Status: theorem. Lean: Start.evaluates.

Proposition III.9.7 (thresholds). For a fixed pre-block profile and simple \(G\), sensors \(\eta_j > 0\) precede block sizes. Fix block factors, \(0 < \theta_j \leqslant 1\), reset constants, decay forecasts and post-block costs. Assume, for each \(i \leqslant j \leqslant i + 1\), \(8 f_{ji} \leqslant q_i \ell_j^{\min}\) and \(u_{ji} (x) \to 0\), where \(f_{ji}\) is the inherited coefficient, \(\ell_j^{\min}\) the preassigned block-size floor, and \(q_i = 1 / [64 (1 + A_i)]\) for the regular reorganization coefficient \(A_i \geqslant 0\). Every threshold record \(V\) admits \(U \geqslant V\) such that every \(W \geqslant U\) ensures pricing, cubic, width, reorganization and inherited inequalities beyond thresholds; clip compatibility requires \(c_{\text{dir}} \leqslant 1\). Realized levels must share these data and satisfy \(W^{\text{entry}}_{J_{m + 1}} \leqslant \hat{\beta}_m\). Physical bounds also require \(4 q_i \leqslant c_{ji}\) and \(4 g_j \leqslant \text{rate}_{ji}\). Start-up thresholds are likewise monotone when every used clip has \(c_{\text{dir}} \leqslant 1\). Status: conditional numerical theorem. Lean: Time.beforeWait.

Proposition III.9.8 (marker sums). Fix \(M, \mu \in \mathbb{N}\), \(a \in \mathbb{R}\), \(c \geqslant 0\). On a four-torus let arbitrarily indexed polymers \(X_i\) have \(f_i \geqslant 0\) and subsets of finitely many marker names, at most \(\mu\) markers per site. Every positive-weight polymer contains one of its markers' sites. If every marker-rooted sum \(\sum_{i : v \in \text{markers} (i)} e^{(a + 1) \text{cover}_M (X_i)} f_i \leqslant c\), then every site-pinned sum \(\sum_{i : x \in X_i} e^{a \text{cover}_M (X_i)} f_i \leqslant \mu e (2 (16 M + 9))^4 c\), uniformly in volume. All nonnegative sums must be summable. Status: theorem. Lean: root_hit.

2. What it uses

3. Strategy

Extract the isotropic quadratic coefficient of the early regular activities into \(\beta\). Subtract flat values and quadratic projections, compensate path terms, and absorb failed-mask discrepancies into the forced gas. A scalar equation identifies this exact reorganization with the continuous reset; shooting chooses the bare coupling with endpoint \(\bar{\beta}\).

4. Steps

Under the corresponding hypotheses in §1:

  1. Taylor expansion through order four gives the Wilson residual; connected large-field components give the exact gas. Start.gas_raw.

  2. A small-field stripe gives \(\mathcal{K} (V) \geqslant | \Lambda | \eta t^2\). Pricing and the activity-motion bound make \(f (x, d) = (\beta'_x + \alpha + d) \mathcal{K} (V) +\mathcal{L} (V) +\mathcal{E}_{F (x, d)} (V) - \log (H_x (1) / H_x (V))\) increasing with slope at least \(\mathcal{K} (V) / 2\), yielding a unique continuous root. paidResponse.

  3. Simplicity and lattice symmetry make the projected Hessian isotropic; exact masked splitting preserves every density value and gives \(d = \delta\). Local.exponent.

  4. Fixed-stage asymptotics make the reset inequalities eventual; prefix maxima combine their thresholds monotonically. collect_upper.

  5. Shooting restricts intervals to keep couplings in their bands and reach \(\beta_0 = \bar{\beta}\); summing drift errors gives uniform running. Shoot.trajectory.

5. Where the difficulty lies

Two-point matching requires full masked reorganization to imply exact transfer. Activity-norm continuity must survive hard configuration masks. The running estimate gives an \(o (s_m)\) remainder, without a bounded remainder or two-loop expansion.

6. Notation

\(N, p, m\): cutoff, volume, level; \(| \Lambda_m |\): site count; \(\alpha = \alpha_{N, m}\): one-loop shift; \(a\): support exponent; \(\text{cover}_M\): polymer cover size; \(J_m\): stage; \(\omega_j\): drift allowance; \(R_m\): patch radius. \(\mathcal{K}, \mathcal{L}, \mathcal{E}, \mathcal{G}\): kinetic, loop, regular and forced-gas terms; \(\mathcal{S}\): masked quadratic receipt; \(\delta\): isotropic Hessian coefficient; \(\chi_X\): small-field indicator. Admissible means prescribed paths/norms/clip, lawfulness and recipe acceptance. Curve continuity means continuous coupling and vacuum constant, fixed supports, inventories and eligibility tests, and continuity in weighted regular sup-jet/forced uniform norms; regular jets are measured on small-field supports at the prescribed order and power.

7. What remains unexplained

Expand reorganization constants and connected-polymer estimates. Nonlinear inputs belong to III.4/III.10; analytic existence and uniqueness within the specified continuation class belong to III.13.

Detailed notes: the full account of this unit, with proof sketches for every step, is on the detailed notes page.

Code dependencies

Code dependencies (from the compiled Lean). These lists follow references between declarations allocated to units; they can include shared definitions, record fields and alternative constructions. They are not a mathematical reading order. Only unit pairs with at least 25 references are shown; see the counting rules.

References to: I.1, III.1, III.2, III.15, III.16, II.2, III.6, II.1, III.10, III.3, III.5, III.4, VI.1.

Referenced by: III.13, III.10, III.12, III.2, III.11, VII.2, III.4, VIII.1, III.15, III.14, VII.4.

Lean source

Source allocation: 7,215 attributed lines; 1,261 line-counted declarations; 1,678 assigned declaration names. Of the assigned names, 1,582 are in the compiled closure of the seventeen headline results. These quantities have different scopes.

The folders and result chains below are source inventories. Their line counts cover each whole group, including any share allocated to another unit; percentages use attributed lines plus one per assigned name. They need not sum to the unit total.

Additional source locations

Further pinned locations for the definitions and comparisons discussed in this unit.

Key statements

In the selected statement order. K- identifiers name this generated list; proposition and lemma numbers in the outline are separate. Each sentence is an automatically drafted summary of a Lean statement; follow its link for the exact hypotheses and conclusion. The outline status does not certify these summaries. Main result and key result describe importance, not correctness. Full names are in the statement table.

All units · Overview of the proof · Repository

← III.8 Connected activities and backward dependence conesPart III·Detailed notesIII.10 One full nonlinear renormalization step: bounds on irrelevant terms →