Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Certified Syracuse descent on 13 progressions inside n≡27(mod32)n \equiv 27 \pmod{32}n≡27(mod32)

Proved
syracuse_descent_progressions_twentyseven_mod32

by Zexuan Liu · Sep 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

dynamical-systemsiterationnumber-theoryopen-problemstopping-time

For every nnn in any of the 13 arithmetic progressions listed below, the Syracuse orbit of nnn drops below its starting point within at most 777 steps:

∃ t≤7:Tt(n)<n.\exists\, t \le 7 : \quad T^{t}(n) < n .∃t≤7:Tt(n)<n.

These are exactly the residue classes modulo a power of 222 up to 2122^{12}212, inside n≡27(mod32)n \equiv 27 \pmod{32}n≡27(mod32), on which the Terras affine certificate closes. Together they cover a proportion 0.64840.64840.6484 of that class; the complement is left open in the companion statement.

The progressions:

  • n≡59(mod128)n \equiv 59 \pmod{128}n≡59(mod128)
  • n≡123(mod256)n \equiv 123 \pmod{256}n≡123(mod256)
  • n≡219(mod256)n \equiv 219 \pmod{256}n≡219(mod256)
  • n≡347(mod1024)n \equiv 347 \pmod{1024}n≡347(mod1024)
  • n≡507(mod1024)n \equiv 507 \pmod{1024}n≡507(mod1024)
  • n≡923(mod1024)n \equiv 923 \pmod{1024}n≡923(mod1024)
  • n≡1019(mod4096)n \equiv 1019 \pmod{4096}n≡1019(mod4096)
  • n≡1435(mod4096)n \equiv 1435 \pmod{4096}n≡1435(mod4096)
  • n≡1787(mod4096)n \equiv 1787 \pmod{4096}n≡1787(mod4096)
  • n≡2203(mod4096)n \equiv 2203 \pmod{4096}n≡2203(mod4096)
  • n≡2587(mod4096)n \equiv 2587 \pmod{4096}n≡2587(mod4096)
  • n≡2907(mod4096)n \equiv 2907 \pmod{4096}n≡2907(mod4096)
  • n≡3675(mod4096)n \equiv 3675 \pmod{4096}n≡3675(mod4096)

The certificate

Let T(n)=(3n+1)/2 v2(3n+1)T(n) = (3n+1)/2^{\,v_2(3n+1)}T(n)=(3n+1)/2v2​(3n+1) be the Syracuse (accelerated Collatz) map, which sends an odd nnn to the odd part of 3n+13n+13n+1.

Terras's observation is that a residue class fixes an initial segment of the orbit's 222-adic data. Write n=2kq+rn = 2^k q + rn=2kq+r. Then 3n+1=(3r+1)+3⋅2kq3n+1 = (3r+1) + 3\cdot 2^k q3n+1=(3r+1)+3⋅2kq, so v2(3n+1)=v2(3r+1)=:a1v_2(3n+1) = v_2(3r+1) =: a_1v2​(3n+1)=v2​(3r+1)=:a1​ as soon as a1<ka_1 < ka1​<k, and

T(n)  =  3r+12a1+3⋅2 k−a1q.T(n) \;=\; \frac{3r+1}{2^{a_1}} + 3\cdot 2^{\,k-a_1} q .T(n)=2a1​3r+1​+3⋅2k−a1​q.

Iterating, as long as the running total Vi=a1+⋯+aiV_i = a_1 + \dots + a_iVi​=a1​+⋯+ai​ stays below kkk, the valuations aia_iai​ are the ones read off the representative rrr, and

Tt(n)  =  3tn+ct2Vt,ct  =  2VtTt(r)−3tr  ≥  0,T^{t}(n) \;=\; \frac{3^{t} n + c_t}{2^{V_t}}, \qquad c_t \;=\; 2^{V_t} T^{t}(r) - 3^{t} r \;\ge\; 0 ,Tt(n)=2Vt​3tn+ct​​,ct​=2Vt​Tt(r)−3tr≥0,

an affine function of nnn on the whole class.

Two further points make the certificate sharp:

  • Truncated last step. When the remaining budget j=k−Vt−1j = k - V_{t-1}j=k−Vt−1​ is at most ata_tat​, the valuation is no longer pinned down, but 2j∣3nt−1+12^{j} \mid 3n_{t-1}+12j∣3nt−1​+1 still holds, so T(nt−1)≤(3nt−1+1)/2jT(n_{t-1}) \le (3n_{t-1}+1)/2^{j}T(nt−1​)≤(3nt−1​+1)/2j. A one-sided bound is all a strict inequality needs, so this final step may be spent even though it is not exact.
  • The least representative decides. Since ct≥0c_t \ge 0ct​≥0, the inequality Tt(n)<nT^{t}(n) < nTt(n)<n for every nnn in the class is equivalent to ct<(2Vt−3t) rc_t < (2^{V_t} - 3^{t})\,rct​<(2Vt​−3t)r, i.e. to the same inequality at the smallest member rrr of the class. So checking the representative certifies the entire progression.

Consequently each progression listed below carries a finite, purely computational witness, and the bound t≤Tmax⁡t \le T_{\max}t≤Tmax​ is uniform across the class.

Formalization note. No oddness or positivity hypothesis is needed: each congruence already forces nnn odd and nnn at least the stated residue. The step bound is stated as t≤7t \le 7t≤7 so that the lemma composes with a case split without re-deriving a bound.

Preamble
import Mathlib
import Definitions.Def_syracuseStep
Formal statement
theorem syracuse_descent_progressions_twentyseven_mod32 (n : ℕ)
    (h : n % 128 = 59 ∨
      n % 256 = 123 ∨
      n % 256 = 219 ∨
      n % 1024 = 347 ∨
      n % 1024 = 507 ∨
      n % 1024 = 923 ∨
      n % 4096 = 1019 ∨
      n % 4096 = 1435 ∨
      n % 4096 = 1787 ∨
      n % 4096 = 2203 ∨
      n % 4096 = 2587 ∨
      n % 4096 = 2907 ∨
      n % 4096 = 3675) :
    ∃ t : ℕ, t ≤ 7 ∧ syracuseStep^[t] n < n := by sorry
Source
https://en.wikipedia.org/wiki/Collatz_conjecture; supporting lemma for the prove2.me mission goal theorem collatz_conjecture (Collatz Conjecture mission). Riho Terras, A stopping time problem on the positive integers, Acta Arith. 30 (1976), 241-252, Section 2 (the coefficient/parity vector and the affinity of T^t on residue classes mod 2^k); Jeffrey C. Lagarias, The 3x+1 Problem and Its Generalizations, Amer. Math. Monthly 92 (1985), 3-23, Section 2, https://websites.umich.edu/~lagarias/3x%2B1.html

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me