Mallows representation of irrational numbers

InductiveThe Mallows alphabet#✓fully proved

The letters of Mallows representations are the three distinct letters \(L\), \(M\) and \(R\), standing for "left", "middle" and "right".

inductive MallowsLetter
  | L
  | M
  | R
  deriving DecidableEq
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 4.
Used by 37
Used byHida's map \(T\) is a bijection from \(I\) onto the Mallows representations · Block-matching lemma · Characterization of Mallows representations · Lagrange's theorem for Mallows representations · Mallows and Stern–Brocot representations are eventually periodic together · Poisson stability of the Mallows and Stern–Brocot representations · Poisson stability of the Mallows representation and the continued fraction · Children of a Mallows-tree vertex · Letter chosen at a Mallows-tree vertex · Labels of the Mallows tree · Labels of the first Mallows-tree vertices · Gaps of the Mallows tree · Mallows representation · Factor complexity of almost every Mallows representation · Words of frequency zero in almost every Mallows representation · Matching fractions of almost every Mallows representation · Words without bounded gaps in almost every Mallows representation · Typical properties of Mallows representations · Extremal measures of Mallows cylinder sets · Measure bound for the letter \(M\) at a fixed position · Mallows representations without the letter \(L\) · Mallows representations without the letter \(M\) · Mallows representations without the letter \(R\) · The first Mallows letter below one half · The map \(T\) · The letter \(M\) in \(T(Q(\mathrm{SB}(\alpha)))\) marks the ends of runs · The Mallows representation from the Stern–Brocot representation · Blocks of \(T(Q(\mathrm{SB}(\alpha)))\) and of \(Q(\mathrm{SB}(\alpha))\) · The finite version \(\dot T\) of \(T\) · Two values of \(\dot T\) · The letter rule of \(T\) · Each rational in \((0, 1)\) occurs once in the Mallows tree, in increasing order on each level · Eventually alternating parities of Stern–Brocot numerators · Eventual period three of the numerator parities · Parity and side of the Stern–Brocot fractions · Transitive Mallows representations and continued fractions · Transitivity of the Mallows and Stern–Brocot representations

InductiveThe parity alphabet#✓fully proved

The parity alphabet consists of the four letters \(O_L\), \(E_R\), \(O_R\) and \(E_L\). The intermediate set \(I\) linking the Stern–Brocot and Mallows representations is a set of infinite sequences over this alphabet.

inductive MallowsParityLetter
  | OL
  | ER
  | OR
  | EL
  deriving DecidableEq
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Used by 18

DefinitionReal value of a fraction encoded as a pair#✓fully proved

The pair \((n, m)\) of natural numbers encodes the fraction with numerator \(n\) and denominator \(m\), whose real value is \(n / m\) (taken to be \(0\) when \(m = 0\)).

noncomputable def natPairToReal (p : ℕ × ℕ) : ℝ :=
  (p.1 : ℝ) / p.2
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2 (fractions n / m labelling the trees).
Used by 8

TheoremThe pair \((1, 2)\) encodes \(1/2\)#✓fully proved

The real value of the fraction encoded by the pair \((1, 2)\) is \(1/2\).

theorem natPairToReal_one_two : natPairToReal (1, 2) = 1 / 2
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.
Uses 1 · Used by 2

DefinitionMediant of a gap#✓fully proved

For a gap bounded by two fractions \(n_1/m_1\) and \(n_2/m_2\), each encoded as the pair (numerator, denominator) of natural numbers, the mediant is the fraction \((n_1 + n_2)/(m_1 + m_2)\), encoded as the pair \((n_1 + n_2, m_1 + m_2)\).

def gapMediant (I : (ℕ × ℕ) × (ℕ × ℕ)) : ℕ × ℕ :=
  (I.1.1 + I.2.1, I.1.2 + I.2.2)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 1 (mediant construction).
Used by 8

TheoremThe mediant of \(0/1\) and \(1/1\)#✓fully proved

With fractions encoded as pairs (numerator, denominator) of natural numbers, the mediant of the gap bounded by \(0/1\) and \(1/1\) is \(1/2\), that is, the pair \((1, 2)\).

theorem gapMediant_root : gapMediant ((0, 1), (1, 1)) = (1, 2)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 1.
Uses 1 · Used by 1

DefinitionMallows' insertion rule#✓fully proved

Let \(n_1/m_1\) and \(n_2/m_2\) be fractions, given as pairs of natural numbers (numerator, denominator). Mallows' insertion rule inserts between them the ordered pair of fractions \(\langle (n_1 + n_2)/(m_1 + m_2),\ (n_1 + 2n_2)/(m_1 + 2m_2) \rangle\) if \(n_1\) is even, and \(\langle (2n_1 + n_2)/(2m_1 + m_2),\ (n_1 + n_2)/(m_1 + m_2) \rangle\) if \(n_1\) is odd. The rule is applied to adjacent fractions \(n_1/m_1 < n_2/m_2\).

def mallowsInsert (a b : ℕ × ℕ) : (ℕ × ℕ) × (ℕ × ℕ) :=
  if a.1 % 2 = 0 then ((a.1 + b.1, a.2 + b.2), (a.1 + 2 * b.1, a.2 + 2 * b.2))
  else ((2 * a.1 + b.1, 2 * a.2 + b.2), (a.1 + b.1, a.2 + b.2))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3 (definition of MT_{i+1}).
Used by 9

TheoremMallows insertion between \(0/1\) and \(1/1\)#✓fully proved

Mallows' insertion rule inserts the pair \(\langle 1/2, 2/3 \rangle\) between the fractions \(0/1\) and \(1/1\), so the first level of the Mallows tree is \(MT_1 = \langle 1/2, 2/3 \rangle\).

theorem mallowsInsert_zero_one : mallowsInsert (0, 1) (1, 1) = ((1, 2), (2, 3))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3.
Uses 1 · Used by 1

TheoremMallows insertion between \(1/2\) and \(2/3\)#✓fully proved

Mallows' insertion rule inserts the pair \(\langle 4/7, 3/5 \rangle\) between the fractions \(1/2\) and \(2/3\) (here the left fraction \(1/2\) has odd numerator).

theorem mallowsInsert_half_twoThirds : mallowsInsert (1, 2) (2, 3) = ((4, 7), (3, 5))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 1.
Uses 1

DefinitionOne refinement step of the Mallows tree#✓fully proved

Let \(a_1, a_2, \ldots, a_k\) be a finite list of fractions, each fraction \(n/m\) recorded as the pair \((n, m)\) of natural numbers. Its refinement is the list obtained by keeping every entry and inserting, between each two consecutive entries \(a_t\) and \(a_{t+1}\), the two fractions that Mallows' insertion rule places between them, in the order the rule gives them. A list with fewer than two entries is its own refinement.

def mallowsRefine : List (ℕ × ℕ) → List (ℕ × ℕ)
  | a :: b :: rest =>
      a :: (mallowsInsert a b).1 :: (mallowsInsert a b).2 :: mallowsRefine (b :: rest)
  | l => l
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3.
Uses 1 · Used by 1

DefinitionFractions inserted by one Mallows step#✓fully proved

Fractions \(n/m\) are written as pairs \((n, m)\) of natural numbers. For a list of fractions \(a_0, a_1, \dots, a_k\), this is the list obtained by taking, for \(j = 0, 1, \dots, k - 1\) in turn, the two fractions \(p_j, q_j\) (in this order) that Mallows' insertion rule inserts between \(a_j\) and \(a_{j+1}\), that is, the list \(p_0, q_0, p_1, q_1, \dots, p_{k-1}, q_{k-1}\). It is empty when the list has fewer than two entries.

def mallowsInserted : List (ℕ × ℕ) → List (ℕ × ℕ)
  | a :: b :: rest => (mallowsInsert a b).1 :: (mallowsInsert a b).2 :: mallowsInserted (b :: rest)
  | _ => []
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3.
Uses 1 · Used by 1

DefinitionMallows levels#✓fully proved

Fractions \(n/m\) are written as pairs \((n, m)\) of natural numbers. Start from the list \(\langle 0/1, 1/1\rangle\) and refine it repeatedly, each refinement step inserting between every two adjacent entries the pair of fractions given by Mallows' insertion rule. For \(i \ge 1\), \(MT_i\) is the list of the fractions inserted at the \(i\)-th step: the fractions that Mallows' rule inserts between adjacent entries of the list obtained after \(i - 1\) steps, taken from left to right, two for each adjacent pair.

After \(k\) steps the list has \(3^k + 1\) entries, so \(MT_i\) has \(2 \cdot 3^{i-1}\) entries.

def mallowsLevel (i : ℕ) : List (ℕ × ℕ) :=
  mallowsInserted (mallowsRefine^[i] [(0, 1), (1, 1)])
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3.
Uses 2 · Used by 2

TheoremThe second Mallows level#✓fully proved

The ordered list of fractions inserted at the second step of Mallows' construction is \(MT_2 = \langle 1/3, 2/5, 4/7, 3/5, 3/4, 4/5\rangle\).

theorem mallowsLevel_one :
    mallowsLevel 1 = [(1, 3), (2, 5), (4, 7), (3, 5), (3, 4), (4, 5)]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 1 and Figure 2.
Uses 1
Statement usesMallows levels

DefinitionChildren of a Mallows-tree vertex#✓fully proved

Let a vertex of the Mallows tree have gap \((a, b)\), where \(a\) and \(b\) are fractions, and let \(\langle p, q \rangle\) be the pair of fractions that Mallows' insertion rule inserts between \(a\) and \(b\). The left, middle and right children of the vertex have gaps \((a, p)\), \((p, q)\) and \((q, b)\) respectively.

def mallowsChild (I : (ℕ × ℕ) × (ℕ × ℕ)) : MallowsLetter → (ℕ × ℕ) × (ℕ × ℕ)
  | .L => (I.1, (mallowsInsert I.1 I.2).1)
  | .M => mallowsInsert I.1 I.2
  | .R => ((mallowsInsert I.1 I.2).2, I.2)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 3.
Uses 2 · Used by 2

DefinitionGaps of the Mallows tree#✓fully proved

Fractions \(n/m\) are written as pairs \((n, m)\) of natural numbers. The root of the Mallows tree has gap \((0/1, 1/1)\). If a vertex has gap \((a, b)\) and Mallows' insertion rule inserts \(\langle p, q\rangle\) in \((a, b)\), then its children along the letters \(L\), \(M\) and \(R\) have gaps \((a, p)\), \((p, q)\) and \((q, b)\). The node of a word \(w\) over \(L\), \(M\), \(R\) is the gap of the vertex reached from the root by following the letters of \(w\) from left to right.

def mallowsNode (w : List MallowsLetter) : (ℕ × ℕ) × (ℕ × ℕ) :=
  w.foldl mallowsChild ((0, 1), (1, 1))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 3.
Uses 2 · Used by 1

DefinitionLabels of the Mallows tree#✓fully proved

For a word \(w\) over the letters \(L\), \(M\), \(R\), the label of the Mallows-tree vertex reached from the root by the path \(w\) is the pair \(\langle p, q\rangle\) of fractions that Mallows' insertion rule inserts in the gap \((a, b)\) of that vertex.

def mallowsLabel (w : List MallowsLetter) : (ℕ × ℕ) × (ℕ × ℕ) :=
  mallowsInsert (mallowsNode w).1 (mallowsNode w).2
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 3 and Figure 3.
Uses 3 · Used by 3

TheoremLabels of the first Mallows-tree vertices#✓fully proved

In the Mallows tree, the root is labelled \(\langle 1/2, 2/3\rangle\), its middle child is labelled \(\langle 4/7, 3/5\rangle\) and its right child is labelled \(\langle 3/4, 4/5\rangle\).

theorem mallowsLabel_small :
    mallowsLabel [] = ((1, 2), (2, 3)) ∧ mallowsLabel [.M] = ((4, 7), (3, 5)) ∧
      mallowsLabel [.R] = ((3, 4), (4, 5))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Figure 3.
Uses 2

DefinitionLetter chosen at a Mallows-tree vertex#✓fully proved

Let a vertex of the Mallows tree have gap \((a, b)\) and label \(\langle p, q \rangle\), the pair of fractions that Mallows' insertion rule inserts between \(a\) and \(b\). For a real number \(\alpha\), the letter chosen at this vertex is \(L\) if \(\alpha < p\); otherwise it is \(M\) if \(\alpha < q\), and \(R\) if \(\alpha \ge q\).

noncomputable def mallowsChoose (α : ℝ) (I : (ℕ × ℕ) × (ℕ × ℕ)) : MallowsLetter :=
  if α < natPairToReal (mallowsInsert I.1 I.2).1 then .L
  else if α < natPairToReal (mallowsInsert I.1 I.2).2 then .M else .R
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 4.
Uses 3 · Used by 2

DefinitionGaps along the Mallows path of a real number#✓fully proved

Fractions \(n/m\) are written as pairs \((n, m)\) of natural numbers. For a gap \((a, b)\) let \(\langle p, q\rangle\) be the pair that Mallows' insertion rule inserts in \((a, b)\). For a real number \(\alpha\), the gaps \(I_0, I_1, \dots\) are defined by \(I_0 = (0/1, 1/1)\) and, if \(I_i = (a, b)\), by \(I_{i+1} = (a, p)\) when \(\alpha < p\), \(I_{i+1} = (p, q)\) when \(p \le \alpha < q\), and \(I_{i+1} = (q, b)\) when \(\alpha \ge q\). In the paper's notation, \(I_i\) is the gap of the vertex \(\mathrm{pf}_{i+1}(\alpha)\) of the Mallows tree visited at step \(i + 1\).

noncomputable def mallowsInterval (α : ℝ) : ℕ → (ℕ × ℕ) × (ℕ × ℕ)
  | 0 => ((0, 1), (1, 1))
  | i + 1 => mallowsChild (mallowsInterval α i) (mallowsChoose α (mallowsInterval α i))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 4.
Uses 2 · Used by 3

DefinitionMallows representation#✓fully proved

For a real number \(\alpha\), the Mallows representation \(M(\alpha) = M_1(\alpha) M_2(\alpha) \cdots\) is the infinite word over \(\{L, M, R\}\) that records the path of \(\alpha\) down the Mallows tree. Fractions are written as numerator–denominator pairs. Start with the gap between \(0/1\) and \(1/1\). At step \(i \ge 1\), if the current gap lies between \(a = n_1/m_1\) and \(b = n_2/m_2\), let \((p, q)\) be the pair that Mallows' rule inserts into it: \(p = (n_1 + n_2)/(m_1 + m_2)\) and \(q = (n_1 + 2n_2)/(m_1 + 2m_2)\) if \(n_1\) is even, and \(p = (2n_1 + n_2)/(2m_1 + m_2)\) and \(q = (n_1 + n_2)/(m_1 + m_2)\) if \(n_1\) is odd. Then \(M_i(\alpha) = L\) and the next gap is \((a, p)\) if \(\alpha < p\); \(M_i(\alpha) = M\) and the next gap is \((p, q)\) if \(p \le \alpha < q\); and \(M_i(\alpha) = R\) and the next gap is \((q, b)\) otherwise.

noncomputable def mallowsRep (α : ℝ) (i : ℕ) : MallowsLetter :=
  mallowsChoose α (mallowsInterval α i)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 4.
Uses 3 · Used by 22

TheoremThe first Mallows letter below one half#✓fully proved

For every real number \(\alpha < 1/2\), the first letter \(M_1(\alpha)\) of the Mallows representation of \(\alpha\) is \(L\).

theorem mallowsRep_zero_of_lt {α : ℝ} (h : α < 1 / 2) : mallowsRep α 0 = .L
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 4 and Figure 4.
Uses 7

DefinitionChildren in the Stern–Brocot tree#✓fully proved

Fractions are pairs \((n, m)\) of natural numbers (numerator, denominator), and the mediant of \(a = n_1/m_1\) and \(b = n_2/m_2\) is \(a \oplus b = (n_1 + n_2)/(m_1 + m_2)\). The child of a gap \((a, b)\) in direction \(L\) is the gap \((a, a \oplus b)\), and its child in direction \(R\) is the gap \((a \oplus b, b)\). For the gap of a Stern–Brocot vertex, whose label is \(a \oplus b\), these are the gaps of its left and right children.

def sternBrocotChild (I : (ℕ × ℕ) × (ℕ × ℕ)) : SternBrocotLetter → (ℕ × ℕ) × (ℕ × ℕ)
  | .L => (I.1, gapMediant I)
  | .R => (gapMediant I, I.2)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 1.
Uses 2 · Used by 2

DefinitionLabels of the Stern–Brocot tree#✓fully proved

Fractions are pairs (numerator, denominator) of natural numbers, and the mediant of \(n_1/m_1\) and \(n_2/m_2\) is \((n_1 + n_2)/(m_1 + m_2)\). For a word \(w\) over \(\{L, R\}\), start from the gap \((0/1, 1/1)\) and read \(w\) from left to right, replacing the current gap \((a, b)\), with mediant \(c\), by \((a, c)\) for each \(L\) and by \((c, b)\) for each \(R\). The label of \(w\) is the mediant of the final gap: the label of the vertex reached from the root by the path \(w\) in the left subtree of the Stern–Brocot tree, whose root is \(1/2\).

def sternBrocotLabel (w : List SternBrocotLetter) : ℕ × ℕ :=
  gapMediant (w.foldl sternBrocotChild ((0, 1), (1, 1)))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 1 and Figure 1.
Uses 3 · Used by 2

TheoremSmall labels of the Stern–Brocot tree#✓fully proved

In the left subtree of the Stern–Brocot tree, the root (reached by the empty path) is labelled \(1/2\), and the vertex reached from the root by the path \(LR\) is labelled \(2/5\), as numerator–denominator pairs.

theorem sternBrocotLabel_small :
    sternBrocotLabel [] = (1, 2) ∧ sternBrocotLabel [.L, .R] = (2, 5)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Figure 1.
Uses 2

DefinitionThe Stern–Brocot choice at a vertex#✓fully proved

Let \(\alpha\) be a real number and let \((a, b)\) be a gap with fractions \(a = n_1/m_1\) and \(b = n_2/m_2\) (pairs of natural numbers). The letter chosen at \((a, b)\) is \(L\) if \(\alpha\) is less than the real number \((n_1 + n_2)/(m_1 + m_2)\), the value of the mediant of \(a\) and \(b\) (the label of the Stern–Brocot vertex with this gap), and \(R\) otherwise.

noncomputable def sternBrocotChoose (α : ℝ) (I : (ℕ × ℕ) × (ℕ × ℕ)) : SternBrocotLetter :=
  if α < natPairToReal (gapMediant I) then .L else .R
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 3 · Used by 2

DefinitionGaps along the Stern–Brocot path#✓fully proved

Fractions are pairs (numerator, denominator) of natural numbers, and the mediant \(a \oplus b\) of \(a = n_1/m_1\) and \(b = n_2/m_2\) is \((n_1 + n_2)/(m_1 + m_2)\). For a real number \(\alpha\), the gaps \(I_0(\alpha), I_1(\alpha), \dots\) are defined by \(I_0(\alpha) = (0/1, 1/1)\) and, if \(I_j(\alpha) = (a, b)\), by \(I_{j+1}(\alpha) = (a, a \oplus b)\) when \(\alpha < a \oplus b\) (as real numbers) and \(I_{j+1}(\alpha) = (a \oplus b, b)\) otherwise. The gap \(I_j(\alpha)\) is that of the Stern–Brocot vertex visited at step \(j + 1\), whose label is the fraction \(f_{j+1}(\alpha)\).

noncomputable def sternBrocotInterval (α : ℝ) : ℕ → (ℕ × ℕ) × (ℕ × ℕ)
  | 0 => ((0, 1), (1, 1))
  | j + 1 => sternBrocotChild (sternBrocotInterval α j)
      (sternBrocotChoose α (sternBrocotInterval α j))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 2 · Used by 4

DefinitionFractions on the Stern–Brocot path#✓fully proved

Fractions are pairs (numerator, denominator) of natural numbers, and the mediant \(a \oplus b\) of \(a = n_1/m_1\) and \(b = n_2/m_2\) is \((n_1 + n_2)/(m_1 + m_2)\). For a real number \(\alpha\), start with the gap \((0/1, 1/1)\); at each step pass from the gap \((a, b)\) to \((a, a \oplus b)\) if \(\alpha < a \oplus b\) (as real numbers) and to \((a \oplus b, b)\) otherwise. For \(j \ge 0\), the fraction \(f_{j+1}(\alpha)\) is the mediant of the gap reached after \(j\) steps; in particular \(f_1(\alpha) = 1/2\).

noncomputable def sternBrocotFrac (α : ℝ) (j : ℕ) : ℕ × ℕ :=
  gapMediant (sternBrocotInterval α j)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 2 · Used by 6

TheoremThe Stern–Brocot path starts at \(1/2\)#✓fully proved

For every real number \(\alpha\), the first fraction \(f_1(\alpha)\) on the Stern–Brocot path of \(\alpha\) is \(1/2\), as the pair with numerator \(1\) and denominator \(2\).

theorem sternBrocotFrac_zero (α : ℝ) : sternBrocotFrac α 0 = (1, 2)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 1

DefinitionStern–Brocot representation#✓fully proved

For a real number \(\alpha\), the Stern–Brocot representation \(SB(\alpha) = SB_1(\alpha) SB_2(\alpha) \cdots\) is the infinite sequence over \(\{L, R\}\) obtained by walking down the Stern–Brocot tree from its vertex \(1/2\), the vertex with gap \((0/1, 1/1)\). The vertex with gap \((a, b)\) has label the mediant \(a \oplus b\) and children with gaps \((a, a \oplus b)\) (left) and \((a \oplus b, b)\) (right). If \(f_j(\alpha)\) is the label of the \(j\)-th vertex of the walk, then \(SB_j(\alpha) = L\) and the walk moves to the left child when \(\alpha < f_j(\alpha)\); otherwise \(SB_j(\alpha) = R\) and the walk moves to the right child.

noncomputable def sternBrocotRep (α : ℝ) (j : ℕ) : SternBrocotLetter :=
  sternBrocotChoose α (sternBrocotInterval α j)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 3 · Used by 13

TheoremFirst Stern–Brocot letter below one half#✓fully proved

For every real number \(\alpha < 1/2\), the first letter \(SB_1(\alpha)\) of the Stern–Brocot representation of \(\alpha\) is \(L\).

theorem sternBrocotRep_zero_of_lt {α : ℝ} (h : α < 1 / 2) : sternBrocotRep α 0 = .L
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Definition 2.
Uses 7

DefinitionThe map \(\varphi_{SBT}\)#✓fully proved

Let \(a = n_1/m_1\) and \(b = n_2/m_2\) be fractions, given by numerator and denominator, and consider the Stern–Brocot vertex with gap \((a, b)\): its label is the mediant \(f = a \oplus b = (n_1 + n_2)/(m_1 + m_2)\), and its left and right children have labels \(l = a \oplus f\) and \(r = f \oplus b\). Then \(\varphi_{SBT}\) of this vertex is the pair \(\langle l, f \rangle\) if the numerator of \(l\) is even, and \(\langle f, r \rangle\) otherwise.

def sternBrocotPhi (I : (ℕ × ℕ) × (ℕ × ℕ)) : List (ℕ × ℕ) :=
  if (gapMediant (I.1, gapMediant I)).1 % 2 = 0 then
    [gapMediant (I.1, gapMediant I), gapMediant I]
  else [gapMediant I, gapMediant (gapMediant I, I.2)]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3 (definition of φ_SBT).
Uses 1 · Used by 3

DefinitionThe map \(\psi_{SBT}\)#✓fully proved

Encode a Stern–Brocot vertex by its gap \((a, b)\) of fractions: its label is the mediant \(a \oplus b\) (numerators added, denominators added), and its left and right children are the vertices with gaps \((a, a \oplus b)\) and \((a \oplus b, b)\). Let \(v_L\) and \(v_R\) be the left and right children of a vertex \(v\). If the label of \(v_L\) has odd numerator, then \(\psi_{SBT}(v)\) is the list consisting of \(v_L\) followed by the left and right children of \(v_R\); otherwise it is the list consisting of the left and right children of \(v_L\) followed by \(v_R\).

def sternBrocotPsi (I : (ℕ × ℕ) × (ℕ × ℕ)) : List ((ℕ × ℕ) × (ℕ × ℕ)) :=
  if (gapMediant (I.1, gapMediant I)).1 % 2 = 1 then
    [(I.1, gapMediant I), (gapMediant I, gapMediant (gapMediant I, I.2)),
      (gapMediant (gapMediant I, I.2), I.2)]
  else
    [(I.1, gapMediant (I.1, gapMediant I)), (gapMediant (I.1, gapMediant I), gapMediant I),
      (gapMediant I, I.2)]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3 (definition of ψ_SBT).
Uses 1 · Used by 2

TheoremThe maps \(\varphi_{SBT}\) and \(\psi_{SBT}\) at the root#✓fully proved

At the root \(1/2\) of the Stern–Brocot tree (the vertex with gap \((0/1, 1/1)\)), \(\varphi_{SBT}(1/2) = \langle 1/2, 2/3 \rangle\) and \(\psi_{SBT}(1/2) = \langle 1/3, 3/5, 3/4 \rangle\), where the vertices \(1/3\), \(3/5\) and \(3/4\) are the ones with gaps \((0/1, 1/2)\), \((1/2, 2/3)\) and \((2/3, 1/1)\).

theorem sternBrocotPhi_psi_root :
    sternBrocotPhi ((0, 1), (1, 1)) = [(1, 2), (2, 3)] ∧
      sternBrocotPsi ((0, 1), (1, 1)) =
        [((0, 1), (1, 2)), ((1, 2), (2, 3)), ((2, 3), (1, 1))]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 2.3.
Uses 2

DefinitionTransition rule of the parity sequence#✓fully proved

The map \(\mathrm{step}\) sends a letter of the parity alphabet \(\{O_L, E_R, O_R, E_L\}\) and a letter of \(\{L, R\}\) to a letter of the parity alphabet, by \(\mathrm{step}(O_L, L) = O_L\), \(\mathrm{step}(O_L, R) = E_R\), \(\mathrm{step}(E_R, L) = O_R\), \(\mathrm{step}(E_R, R) = O_L\), \(\mathrm{step}(O_R, L) = E_L\), \(\mathrm{step}(O_R, R) = O_R\), \(\mathrm{step}(E_L, L) = O_R\) and \(\mathrm{step}(E_L, R) = O_L\).

def mallowsQStep : MallowsParityLetter → SternBrocotLetter → MallowsParityLetter
  | .OL, .L => .OL
  | .OL, .R => .ER
  | .ER, .L => .OR
  | .ER, .R => .OL
  | .OR, .L => .EL
  | .OR, .R => .OR
  | .EL, .L => .OR
  | .EL, .R => .OL
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3 (definition of Q, Figure 5).
Uses 2 · Used by 2

DefinitionThe parity sequence of a Stern–Brocot word#✓fully proved

Let \(x = x_0 x_1 x_2 \cdots\) be an infinite word over \(\{L, R\}\). Its parity sequence \(Q(x) = Q_0 Q_1 Q_2 \cdots\) over the parity alphabet \(\{O_L, E_R, O_R, E_L\}\) is defined by \(Q_0 = O_L\) and \(Q_{j+1} = \mathrm{step}(Q_j, x_j)\) for every \(j \ge 0\), where \(\mathrm{step}\) is the transition rule of the parity sequence.

Indices start at \(0\) here; the source numbers from \(1\), so \(Q_0\) is the source's \(Q_1(x)\).

def mallowsQ (x : ℕ → SternBrocotLetter) : ℕ → MallowsParityLetter
  | 0 => .OL
  | j + 1 => mallowsQStep (mallowsQ x j) (x j)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 3 · Used by 8

TheoremFirst values of the parity sequence#✓fully proved

Let \(x = x_0 x_1 x_2 \cdots\) be an infinite word over \(\{L, R\}\) with \(x_0 = R\) and \(x_1 = L\). Then the parity sequence \(Q(x) = Q_0 Q_1 Q_2 \cdots\) of \(x\) satisfies \(Q_0 = O_L\), \(Q_1 = E_R\) and \(Q_2 = O_R\).

Indices start at \(0\) here; in the source these values are \(Q_1\), \(Q_2\) and \(Q_3\).

theorem mallowsQ_small (x : ℕ → SternBrocotLetter) (h0 : x 0 = .R) (h1 : x 1 = .L) :
    mallowsQ x 0 = .OL ∧ mallowsQ x 1 = .ER ∧ mallowsQ x 2 = .OR
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 3

DefinitionThe admissible parity sequences \(I\)#✓fully proved

The set \(I\) consists of the infinite sequences \(x = (x_0, x_1, \dots)\) over the alphabet \(\{O_L, E_R, O_R, E_L\}\) such that \(x_0 = O_L\); no two consecutive terms \((x_j, x_{j+1})\) form one of the pairs \((O_L, O_R)\), \((O_L, E_L)\), \((E_R, E_L)\), \((E_R, E_R)\), \((O_R, O_L)\), \((O_R, E_R)\), \((E_L, E_L)\), \((E_L, E_R)\); \(x\) is not eventually constant equal to \(O_L\), nor eventually constant equal to \(O_R\); and no tail of \(x\) is the periodic sequence \((O_L, E_R, O_L, E_R, \dots)\) or \((O_R, E_L, O_R, E_L, \dots)\).

def mallowsAdmissible : Set (ℕ → MallowsParityLetter) :=
  {x | x 0 = .OL ∧
    (∀ j, (x j, x (j + 1)) ∉ [(.OL, .OR), (.OL, .EL), (.ER, .EL), (.ER, .ER), (.OR, .OL),
      (.OR, .ER), (.EL, .EL), (.EL, .ER)]) ∧
    (¬ ∃ j₀, ∀ j, j₀ ≤ j → x j = .OL) ∧ (¬ ∃ j₀, ∀ j, j₀ ≤ j → x j = .OR) ∧
    (¬ ∃ j₀, ∀ t, x (j₀ + 2 * t) = .OL ∧ x (j₀ + 2 * t + 1) = .ER) ∧
    (¬ ∃ j₀, ∀ t, x (j₀ + 2 * t) = .OR ∧ x (j₀ + 2 * t + 1) = .EL)}
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 1 · Used by 4

TheoremThe constant sequence \(O_L O_L O_L \cdots\) is not admissible#✓fully proved

The constant sequence whose every term is \(O_L\) does not belong to the set \(I\) of admissible sequences over \(\{O_L, E_R, O_R, E_L\}\).

theorem const_OL_not_mem_mallowsAdmissible :
    (fun _ => MallowsParityLetter.OL) ∉ mallowsAdmissible
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 2

DefinitionOdd positions of a parity sequence#✓fully proved

Let \(x = x_0 x_1 x_2 \cdots\) be an infinite sequence over the parity alphabet \(\{O_L, E_R, O_R, E_L\}\), and call a position \(j \ge 0\) odd when \(x_j \in \{O_L, O_R\}\). For \(i \ge 0\), the \(i\)-th odd position of \(x\) is the \((i+1)\)-st smallest odd position of \(x\), and it is \(0\) by convention when \(x\) has at most \(i\) odd positions.

In the source, positions are numbered from \(1\) and the odd positions are enumerated as \(o(1) < o(2) < \cdots\); the \(i\)-th odd position here is \(o(i+1) - 1\).

noncomputable def mallowsOddIndex (x : ℕ → MallowsParityLetter) : ℕ → ℕ :=
  Nat.nth fun j => x j = .OL ∨ x j = .OR
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 1 · Used by 7

TheoremThe first odd position of a sequence starting with \(O_L\)#✓fully proved

Let \(x = x_0 x_1 x_2 \cdots\) be an infinite sequence over the parity alphabet \(\{O_L, E_R, O_R, E_L\}\) with \(x_0 = O_L\), and call a position \(j \ge 0\) odd when \(x_j \in \{O_L, O_R\}\). Then the \(0\)-th odd position of \(x\), the smallest odd position, is \(0\).

theorem mallowsOddIndex_zero {x : ℕ → MallowsParityLetter} (h : x 0 = .OL) :
    mallowsOddIndex x 0 = 0
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3 (o(1) = 1).
Uses 2

DefinitionThe letter rule of \(T\)#✓fully proved

For a truth value \(b\) and letters \(a, a'\) of \(\{O_L, E_R, O_R, E_L\}\), the letter \(\tau(b, a, a') \in \{L, M, R\}\) is defined as follows. If \(b\) is true, it is \(L\) when \(a = O_L\) and \(R\) otherwise. If \(b\) is false, it is \(M\) when \(a \ne a'\), \(L\) when \(a = a' = O_R\), and \(R\) when \(a = a' \ne O_R\).

If \(o(1) < o(2) < \cdots\) are the positions of the letters \(O_L\) and \(O_R\) in a sequence \(x\), then \(T_i(x) = \tau(b, x_{o(i)}, x_{o(i+1)})\), where \(b\) is true exactly when \(o(i + 1) = o(i) + 1\).

def mallowsTLetter (adj : Bool) (a b : MallowsParityLetter) : MallowsLetter :=
  if adj then (if a = .OL then .L else .R)
  else if a = b then (if a = .OR then .L else .R) else .M
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3 (definition of T_i).
Uses 2 · Used by 2

DefinitionThe map \(T\)#✓fully proved

Let \(x = \langle x_1, x_2, \dots \rangle\) be a sequence over \(\{O_L, E_R, O_R, E_L\}\) and let \(o(1) < o(2) < \cdots\) be the positions \(j\) with \(x_j \in \{O_L, O_R\}\). The sequence \(T(x) = \langle T_1(x), T_2(x), \dots \rangle\) over \(\{L, M, R\}\) is defined as follows. If \(o(i + 1) = o(i) + 1\), then \(T_i(x)\) is \(L\) when \(x_{o(i)} = O_L\) and \(R\) otherwise. If \(o(i + 1) \ne o(i) + 1\), then \(T_i(x)\) is \(M\) when \(x_{o(i)} \ne x_{o(i + 1)}\), \(L\) when \(x_{o(i)} = x_{o(i + 1)} = O_R\), and \(R\) when \(x_{o(i)} = x_{o(i + 1)} \ne O_R\).

If there are only finitely many such positions, \(o(i)\) is taken to be \(1\) for every \(i\) larger than their number.

noncomputable def mallowsT (x : ℕ → MallowsParityLetter) (i : ℕ) : MallowsLetter :=
  mallowsTLetter (decide (mallowsOddIndex x (i + 1) = mallowsOddIndex x i + 1))
    (x (mallowsOddIndex x i)) (x (mallowsOddIndex x (i + 1)))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 3.
Uses 4 · Used by 5

DefinitionThe finite version \(\dot T\) of \(T\)#✓fully proved

Let \(y = \langle y_1, \dots, y_n \rangle\) be a finite word over \(\{O_L, E_R, O_R, E_L\}\) and let \(p_1 < p_2 < \cdots < p_k\) be the positions \(j\) with \(y_j \in \{O_L, O_R\}\). Then \(\dot T(y) = \langle t_1, \dots, t_{k-1} \rangle\) is the word over \(\{L, M, R\}\) (empty if \(k \le 1\)) given as follows. If \(p_{i+1} = p_i + 1\), then \(t_i\) is \(L\) when \(y_{p_i} = O_L\) and \(R\) otherwise. If \(p_{i+1} \ne p_i + 1\), then \(t_i\) is \(M\) when \(y_{p_i} \ne y_{p_{i+1}}\), \(L\) when \(y_{p_i} = y_{p_{i+1}} = O_R\), and \(R\) when \(y_{p_i} = y_{p_{i+1}} = O_L\).

def mallowsTFin (y : List MallowsParityLetter) : List MallowsLetter :=
  let os := (List.range y.length).filter fun j => y.getD j .ER = .OL ∨ y.getD j .ER = .OR
  (os.zip os.tail).map fun pq =>
    mallowsTLetter (decide (pq.2 = pq.1 + 1)) (y.getD pq.1 .ER) (y.getD pq.2 .ER)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 4.3 (definition of Ṫ).
Uses 3 · Used by 2

TheoremTwo values of \(\dot T\)#✓fully proved

The finite version \(\dot T\) of \(T\) satisfies \(\dot T(\langle O_L, E_R, O_R \rangle) = \langle M \rangle\) and \(\dot T(\langle O_L, O_L \rangle) = \langle L \rangle\).

theorem mallowsTFin_small :
    mallowsTFin [.OL, .ER, .OR] = [.M] ∧ mallowsTFin [.OL, .OL] = [.L]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 4.3.
Uses 3

DefinitionPositively Poisson stable sequence#✓fully proved

A sequence \(x = (x_0, x_1, x_2, \dots)\) with values in a set \(A\) is positively Poisson stable if there is an index \(n_0\) such that for every length \(\ell\) there is a shift \(p \ge 1\) with \(x_{n_0 + p + t} = x_{n_0 + t}\) for all \(t < \ell\); that is, every initial block of the tail \((x_{n_0}, x_{n_0 + 1}, \dots)\) occurs again at a later position.

This is the shift-space form of the condition that the positive orbit of \(x\) under the shift \(\sigma\) meets the \(\omega\)-limit set of \(x\). The point that recurs is \(\sigma^{n_0} x\), not necessarily \(x\) itself: for example, \((0, 1, 1, 1, \dots)\) is positively Poisson stable.

def IsPositivelyPoissonStable {A : Type*} (x : ℕ → A) : Prop :=
  ∃ n₀ : ℕ, ∀ ℓ : ℕ, ∃ p : ℕ, 0 < p ∧ ∀ t < ℓ, x (n₀ + p + t) = x (n₀ + t)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 4.2.
Used by 4

TheoremConstant sequences are positively Poisson stable#✓fully proved

For every set \(A\) and every \(c \in A\), the constant sequence \((c, c, c, \dots)\) is positively Poisson stable.

theorem isPositivelyPoissonStable_const {A : Type*} (c : A) :
    IsPositivelyPoissonStable (fun _ : ℕ => c)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Section 4.2.
Uses 1

TheoremStern–Brocot tree: labels, order and paths#●ready to prove

Consider the subtree of the Stern–Brocot tree with root \(1/2\), and identify each vertex with its path \(w\) from the root, a finite word over \(\{L, R\}\). Then: (i) every rational \(q\) with \(0 < q < 1\) is the label of exactly one vertex; (ii) for all words \(u, v_1, v_2\) over \(\{L, R\}\) with \(v_1\) and \(v_2\) of the same length, the label of \(u L v_1\) is less than the label of \(u R v_2\); (iii) if the label of the vertex \(w\) has continued fraction expansion \([0; a_1, \ldots, a_\ell]\) with last partial quotient \(a_\ell \ge 2\), then \(w = L^{a_1 - 1} R^{a_2} L^{a_3} \cdots\), made of \(\ell\) blocks alternating between \(L\) and \(R\) starting with \(L\), where the \(k\)-th block has length \(a_k\), except that the first block has length \(a_1 - 1\) and the last has length \(a_\ell - 1\); if \(\ell = 1\), then \(w = L^{a_1 - 2}\).

public theorem_wanted sternBrocotTree_label_unique_order_path :
    (∀ q : ℚ, 0 < q → q < 1 →
      ∃! w : List SternBrocotLetter, natPairToReal (sternBrocotLabel w) = q) ∧
    (∀ u v₁ v₂ : List SternBrocotLetter, v₁.length = v₂.length →
      natPairToReal (sternBrocotLabel (u ++ .L :: v₁)) <
        natPairToReal (sternBrocotLabel (u ++ .R :: v₂))) ∧
    ∀ (w : List SternBrocotLetter) (a : List ℕ),
      (∀ k (hk : k < a.length),
        (GenContFract.of (natPairToReal (sternBrocotLabel w))).partDens.get? k = some (a[k] : ℝ)) →
      (GenContFract.of (natPairToReal (sternBrocotLabel w))).partDens.get? a.length = none →
      w = (List.range a.length).flatMap fun k =>
        List.replicate (a.getD k 0 - (if k = 0 then 1 else 0) - (if k + 1 = a.length then 1 else 0))
          (if Even k then SternBrocotLetter.L else SternBrocotLetter.R)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 1.
Uses 3

TheoremStern–Brocot representation and continued fractions#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational with continued fraction expansion \(\alpha = [0; a_1, a_2, \ldots]\), let \(f_1(\alpha), f_2(\alpha), \ldots\) be the fractions labelling the vertices visited by its Stern–Brocot path, and let \(SB(\alpha) = SB_1(\alpha) SB_2(\alpha) \cdots\) be its Stern–Brocot representation. Then (i) \(f_j(\alpha) \to \alpha\) as \(j \to \infty\); and (ii) \(SB(\alpha) = L^{a_1 - 1} R^{a_2} L^{a_3} R^{a_4} \cdots\), that is, for all \(j \ge 1\) and \(k \ge 0\) with \(a_1 + \cdots + a_k \le j < a_1 + \cdots + a_{k+1}\), the letter \(SB_j(\alpha)\) is \(L\) if \(k\) is even and \(R\) if \(k\) is odd.

public theorem_wanted sternBrocotRep_tendsto_and_contFract (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) (a : ℕ → ℕ)
    (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    Filter.Tendsto (fun j => natPairToReal (sternBrocotFrac α j)) Filter.atTop (nhds α) ∧
      ∀ j k, ∑ i ∈ Finset.range k, a i ≤ j + 1 → j + 1 < ∑ i ∈ Finset.range (k + 1), a i →
        sternBrocotRep α j = if Even k then .L else .R
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 2.
Uses 4

TheoremCharacterization of Stern–Brocot representations#●ready to prove

A sequence over \(\{L, R\}\) is the Stern–Brocot representation \(\mathrm{SB}(\alpha)\) of some irrational \(\alpha \in (0, 1)\) if and only if it is neither eventually constantly \(L\) nor eventually constantly \(R\).

public theorem_wanted image_sternBrocotRep :
    (fun α => sternBrocotRep α) '' {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α} =
      {x | (¬ ∃ j₀, ∀ j, j₀ ≤ j → x j = .L) ∧ ¬ ∃ j₀, ∀ j, j₀ ≤ j → x j = .R}
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 1.
Uses 2

TheoremEach rational in \((0, 1)\) occurs once in the Mallows tree, in increasing order on each level#●ready to prove

Each rational in \((0, 1)\) occurs once in the Mallows tree, in increasing order on each level. For a finite word \(w\) over \(\{L, M, R\}\) let \(\langle p_w, q_w \rangle\) be the label, a pair of fractions, of the vertex of the Mallows tree reached from the root by the path \(w\); fractions are compared by their real values. Then: (i) for every rational number \(r\) with \(0 < r < 1\) there is exactly one pair \((w, s)\) of a path \(w\) and a side \(s \in \{\text{left}, \text{right}\}\) such that the fraction on side \(s\) of the label of \(w\) (\(p_w\) for left, \(q_w\) for right) has value \(r\); (ii) \(p_w < q_w\) for every path \(w\); (iii) for all words \(u\), \(v_1\), \(v_2\) over \(\{L, M, R\}\) with \(v_1\) and \(v_2\) of the same length and every pair \((b, c)\) among \((L, M)\), \((L, R)\), \((M, R)\), we have \(q_{u b v_1} < p_{u c v_2}\).

Together, (ii) and (iii) say that on each level \(\mathrm{MT}_i\) of the tree the left-to-right order of the fractions agrees with their increasing order.

public theorem_wanted mallowsTree_label_unique_order :
    (∀ q : ℚ, 0 < q → q < 1 → ∃! p : List MallowsLetter × Bool,
      natPairToReal (if p.2 then (mallowsLabel p.1).2 else (mallowsLabel p.1).1) = q) ∧
    (∀ w : List MallowsLetter,
      natPairToReal (mallowsLabel w).1 < natPairToReal (mallowsLabel w).2) ∧
    ∀ (u v₁ v₂ : List MallowsLetter) (b c : MallowsLetter), v₁.length = v₂.length →
      (b, c) ∈ [(.L, .M), (.L, .R), (.M, .R)] →
      natPairToReal (mallowsLabel (u ++ b :: v₁)).2 < natPairToReal (mallowsLabel (u ++ c :: v₂)).1
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 3.
Uses 3

TheoremMallows levels from the Stern–Brocot maps#●ready to prove

Encode each vertex of the Stern–Brocot tree (left subtree, root \(1/2\)) by its gap, so that the root is the gap \((0/1, 1/1)\), and extend \(\psi_{\mathrm{SBT}}\) and \(\varphi_{\mathrm{SBT}}\) to lists of vertices by applying them to each entry and concatenating the results in order. Then for every \(i \ge 1\) the ordered list \(MT_i\) of fractions inserted at step \(i\) of Mallows' construction is \(MT_i = \varphi_{\mathrm{SBT}}(\psi_{\mathrm{SBT}}^{\,i-1}(\langle 1/2\rangle))\), where \(\psi_{\mathrm{SBT}}^{\,i-1}\) is the \((i-1)\)-fold iterate and \(\langle 1/2\rangle\) is the list whose only entry is the root.

The paper quotes this identity from Hida, J. Comb. 14 (2023).

public theorem_wanted mallowsLevel_eq_phi_psi_iterate (i : ℕ) :
    mallowsLevel i =
      ((fun l : List ((ℕ × ℕ) × (ℕ × ℕ)) => l.flatMap sternBrocotPsi)^[i]
        [((0, 1), (1, 1))]).flatMap sternBrocotPhi
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 1.
Uses 3

TheoremCharacterization of Mallows representations#●ready to prove

A sequence over \(\{L, M, R\}\) is the Mallows representation \(M(\alpha)\) of some irrational \(\alpha \in (0, 1)\) if and only if it is neither eventually constantly \(L\) nor eventually constantly \(R\).

public theorem_wanted image_mallowsRep :
    (fun α => mallowsRep α) '' {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α} =
      {x | (¬ ∃ i₀, ∀ i, i₀ ≤ i → x i = .L) ∧ ¬ ∃ i₀, ∀ i, i₀ ≤ i → x i = .R}
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 4.
Uses 2

TheoremHida's map \(Q\) is a bijection onto \(I\)#●ready to prove

Let \(\mathrm{SB}(\alpha)\) be the Stern–Brocot representation of a real number \(\alpha\), let \(Q\) be Hida's map from sequences over \(\{L, R\}\) to sequences over \(\{O_L, E_R, O_R, E_L\}\), and let \(I\) be the set of admissible sequences over \(\{O_L, E_R, O_R, E_L\}\). Then \(Q\) maps the set of Stern–Brocot representations \(\{\mathrm{SB}(\alpha) : \alpha \in (0, 1) \text{ irrational}\}\) bijectively onto \(I\): for every irrational \(\alpha \in (0, 1)\) the sequence \(Q(\mathrm{SB}(\alpha))\) lies in \(I\), \(Q\) is injective on this set of representations, and every element of \(I\) equals \(Q(\mathrm{SB}(\alpha))\) for some irrational \(\alpha \in (0, 1)\).

public theorem_wanted bijOn_mallowsQ :
    Set.BijOn mallowsQ
      ((fun α => sternBrocotRep α) '' {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α}) mallowsAdmissible
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 5.
Uses 5

TheoremParity and side of the Stern–Brocot fractions#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, let \(f_1(\alpha), f_2(\alpha), \dots\) be the fractions on its Stern–Brocot path, and let \(SB(\alpha) = \langle SB_1, SB_2, \dots \rangle\) be its Stern–Brocot representation. Let \(Q(SB(\alpha)) = \langle Q_1, Q_2, \dots \rangle\) be the sequence over \(\{O_L, E_R, O_R, E_L\}\) with \(Q_1 = O_L\) in which \(Q_{j+1}\) is determined by \(Q_j\) and \(SB_j\) as follows: from \(O_L\), \(L \mapsto O_L\) and \(R \mapsto E_R\); from \(E_R\), \(L \mapsto O_R\) and \(R \mapsto O_L\); from \(O_R\), \(L \mapsto E_L\) and \(R \mapsto O_R\); from \(E_L\), \(L \mapsto O_R\) and \(R \mapsto O_L\). Then for every \(j \ge 1\): (i) the numerator of \(f_j(\alpha)\) is odd if and only if \(Q_j \in \{O_L, O_R\}\); (ii) \(f_j(\alpha)\) is the left or the right element of the label of some vertex of the Mallows tree; (iii) for every vertex of the Mallows tree, if \(f_j(\alpha)\) is the left element of its label then \(Q_j \in \{O_L, E_L\}\), and if it is the right element then \(Q_j \in \{O_R, E_R\}\).

Since every fraction is an element of at most one Mallows label (Proposition 3), (ii) and (iii) give the paper's equivalence: \(f_j(\alpha)\) is the left (resp. right) element of its Mallows label if and only if \(Q_j \in \{O_L, E_L\}\) (resp. \(\{O_R, E_R\}\)).

public theorem_wanted sternBrocotFrac_parity_side (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α)
    (j : ℕ) :
    (Odd (sternBrocotFrac α j).1 ↔
      mallowsQ (sternBrocotRep α) j = .OL ∨ mallowsQ (sternBrocotRep α) j = .OR) ∧
    (∃ w : List MallowsLetter,
      sternBrocotFrac α j = (mallowsLabel w).1 ∨ sternBrocotFrac α j = (mallowsLabel w).2) ∧
    ∀ w : List MallowsLetter,
      (sternBrocotFrac α j = (mallowsLabel w).1 →
        mallowsQ (sternBrocotRep α) j = .OL ∨ mallowsQ (sternBrocotRep α) j = .EL) ∧
      (sternBrocotFrac α j = (mallowsLabel w).2 →
        mallowsQ (sternBrocotRep α) j = .OR ∨ mallowsQ (sternBrocotRep α) j = .ER)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 6.
Uses 6

TheoremGaps between odd letters of admissible sequences#●ready to prove

Let \(x = x_0 x_1 x_2 \cdots\) be a sequence in the admissible set \(I\) over the letters \(O_L\), \(E_R\), \(O_R\), \(E_L\), and for \(i \ge 0\) let \(p_i\) be the \(i\)-th position (counting from \(0\)) at which \(x\) has one of the letters \(O_L\), \(O_R\). Then for every \(i \ge 0\), \(p_{i+1} > p_i + 1\) holds if and only if \(p_{i+1} = p_i + 2\).

Since the positions increase, this says that consecutive positions of \(O_L\) or \(O_R\) are at distance one or two.

public theorem_wanted mallowsOddIndex_gap (x : ℕ → MallowsParityLetter)
    (hx : x ∈ mallowsAdmissible) (i : ℕ) :
    mallowsOddIndex x i + 1 < mallowsOddIndex x (i + 1) ↔
      mallowsOddIndex x (i + 1) = mallowsOddIndex x i + 2
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 7.
Uses 3

TheoremHida's map \(T\) is a bijection from \(I\) onto the Mallows representations#●ready to prove

Let \(I\) be the set of admissible sequences over \(\{O_L, E_R, O_R, E_L\}\), let \(T\) be Hida's map from sequences over \(\{O_L, E_R, O_R, E_L\}\) to sequences over \(\{L, M, R\}\), and let \(M(\alpha)\) be the Mallows representation of a real number \(\alpha\). Then \(T\) maps \(I\) bijectively onto the set \(\{M(\alpha) : \alpha \in (0, 1) \text{ irrational}\}\): for every \(x \in I\) the sequence \(T(x)\) lies in this set, \(T\) is injective on \(I\), and for every irrational \(\alpha \in (0, 1)\) there is \(x \in I\) with \(T(x) = M(\alpha)\).

public theorem_wanted bijOn_mallowsT :
    Set.BijOn mallowsT mallowsAdmissible
      ((fun α => mallowsRep α) '' {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α})
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 8.
Uses 5

TheoremStern–Brocot labels of the Mallows representation vertices#●ready to prove

For irrational α ∈ (0, 1) and i ≥ 1, φ_SBT(f_{o(i)}(α)) = pf_i(α). The vertex f_{o(i)}(α) is encoded by its gap, pf_i(α) is the label ⟨p, q⟩ inserted in the gap mallowsInterval α (i - 1), written as the list [p, q], and o is computed on Q(SB(α)).

public theorem_wanted sternBrocotPhi_oddIndex (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α)
    (i : ℕ) :
    sternBrocotPhi (sternBrocotInterval α (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i)) =
      [(mallowsInsert (mallowsInterval α i).1 (mallowsInterval α i).2).1,
        (mallowsInsert (mallowsInterval α i).1 (mallowsInterval α i).2).2]
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Lemma 1.
Uses 7

TheoremThe Mallows representation from the Stern–Brocot representation#●ready to prove

For every irrational number \(\alpha\) with \(0 < \alpha < 1\), the Mallows representation \(M(\alpha)\) is obtained from the Stern–Brocot representation \(\mathrm{SB}(\alpha)\) by applying the map \(Q\), which gives a sequence over \(\{O_L, E_R, O_R, E_L\}\), and then the map \(T\), which gives a sequence over \(\{L, M, R\}\): \(T(Q(\mathrm{SB}(\alpha))) = M(\alpha)\).

public theorem_wanted mallowsT_mallowsQ_sternBrocotRep (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    mallowsT (mallowsQ (sternBrocotRep α)) = mallowsRep α
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 2.
Uses 5

TheoremBlock-matching lemma#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, let \(x = \mathrm{SB}(\alpha)\) be its Stern–Brocot representation, let \(q = Q(x)\) be the image of \(x\) under Hida's map \(Q\) to sequences over \(\{O_L, E_R, O_R, E_L\}\), and let \(T(q)\) be the image of \(q\) under Hida's map \(T\) to sequences over \(\{L, M, R\}\); all sequences are indexed from \(0\). Let \(o(i)\) be the \(i\)-th position, counting from \(0\), at which \(q\) takes the value \(O_L\) or \(O_R\). Then for all natural numbers \(\ell\), \(j_1\), \(j_2\), \(i_1\), \(i_2\), with \(-\) denoting truncated subtraction of natural numbers: (i) \(x_{j_1 + t} = x_{j_2 + t}\) for all \(t < \ell\) together with \(q_{j_1} = q_{j_2}\) holds if and only if \(q_{j_1 + t} = q_{j_2 + t}\) for all \(t \le \ell\); (ii) \(T(q)_{i_1 + t} = T(q)_{i_2 + t}\) for all \(t < \ell\) together with \(q_{o(i_1)} = q_{o(i_2)}\) holds if and only if \(q_{o(i_1) + t} = q_{o(i_2) + t}\) for all \(t \le o(i_1 + \ell) - o(i_1)\); (iii) if \(T(q)_{i_1 + t} = T(q)_{i_2 + t}\) for all \(t < \ell\) and \(q_{o(i_1)} = q_{o(i_2)}\), then \(o(i_2 + k) - o(i_2) = o(i_1 + k) - o(i_1)\) for all \(k \le \ell\).

The paper assumes \(\ell \ge 1\), \(j_1 < j_2\) and \(i_1 < i_2\); the statement is trivial for \(\ell = 0\), symmetric in the two positions, and trivial for equal positions.

MetaMathlibExt.block_match_sternBrocot_mallowsQ · 22 lines
public theorem_wanted block_match_sternBrocot_mallowsQ (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α)
    (ℓ j₁ j₂ i₁ i₂ : ℕ) :
    (((∀ t < ℓ, sternBrocotRep α (j₁ + t) = sternBrocotRep α (j₂ + t)) ∧
        mallowsQ (sternBrocotRep α) j₁ = mallowsQ (sternBrocotRep α) j₂) ↔
      ∀ t ≤ ℓ, mallowsQ (sternBrocotRep α) (j₁ + t) = mallowsQ (sternBrocotRep α) (j₂ + t)) ∧
    (((∀ t < ℓ, mallowsT (mallowsQ (sternBrocotRep α)) (i₁ + t) =
          mallowsT (mallowsQ (sternBrocotRep α)) (i₂ + t)) ∧
        mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₁) =
          mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₂)) ↔
      ∀ t ≤ mallowsOddIndex (mallowsQ (sternBrocotRep α)) (i₁ + ℓ) -
          mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₁,
        mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₁ + t) =
          mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₂ + t)) ∧
    (((∀ t < ℓ, mallowsT (mallowsQ (sternBrocotRep α)) (i₁ + t) =
          mallowsT (mallowsQ (sternBrocotRep α)) (i₂ + t)) ∧
        mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₁) =
          mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₂)) →
      ∀ k ≤ ℓ, mallowsOddIndex (mallowsQ (sternBrocotRep α)) (i₂ + k) -
          mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₂ =
        mallowsOddIndex (mallowsQ (sternBrocotRep α)) (i₁ + k) -
          mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₁)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Lemma 2.
Uses 7

TheoremBlocks of \(T(Q(\mathrm{SB}(\alpha)))\) and of \(Q(\mathrm{SB}(\alpha))\)#●ready to prove

Let \(X\) be one of the letters \(O_L, E_R, O_R, E_L\), let \(u = u_1 \cdots u_\ell\) be a finite word over \(\{L, R\}\), and let \(\dot Q^X(u) = \langle X_0, X_1, \dots, X_\ell \rangle\) be the word with \(X_0 = X\) and \(X_k = \delta(X_{k-1}, u_k)\) for \(1 \le k \le \ell\), where \(\delta\) is the transition rule that defines the map \(Q\) from sequences over \(\{L, R\}\) to sequences over \(\{O_L, E_R, O_R, E_L\}\). Assume that \(X_\ell\) is \(O_L\) or \(O_R\), and let \(n\) be the length of the word \(\dot T(\dot Q^X(u))\), where \(\dot T\) is the finite version of the map \(T\) from sequences over \(\{O_L, E_R, O_R, E_L\}\) to sequences \(T(x) = \langle T_1(x), T_2(x), \dots \rangle\) over \(\{L, M, R\}\). Let \(\alpha\) be an irrational number with \(0 < \alpha < 1\), let \(\mathrm{SB}(\alpha)\) be its Stern–Brocot representation, write \(x = Q(\mathrm{SB}(\alpha)) = \langle x_1, x_2, \dots \rangle\), let \(o(1) < o(2) < \cdots\) be the positions \(j\) with \(x_j \in \{O_L, O_R\}\), and let \(i_0 \ge 1\). Then the \(n\) letters \(T_{i_0}(x), \dots, T_{i_0 + n - 1}(x)\) form the word \(\dot T(\dot Q^X(u))\) and \(x_{o(i_0)} = X\) if and only if the \(\ell + 1\) letters \(x_{o(i_0)}, x_{o(i_0) + 1}, \dots, x_{o(i_0) + \ell}\) form the word \(\dot Q^X(u)\).

The source assumes \(X \in \{O_L, O_R\}\) and \(\ell \ge 1\). For \(X \in \{E_R, E_L\}\) both sides are false, since \(x_{o(i_0)}\) is \(O_L\) or \(O_R\); for \(\ell = 0\) both sides say \(x_{o(i_0)} = X\).

public theorem_wanted mallowsT_startsWith_iff (X : MallowsParityLetter) (u : List SternBrocotLetter)
    (hlast : (u.scanl mallowsQStep X).getLast? = some .OL ∨
      (u.scanl mallowsQStep X).getLast? = some .OR)
    (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α) (i₀ : ℕ) :
    (List.ofFn (fun t : Fin (mallowsTFin (u.scanl mallowsQStep X)).length =>
          mallowsT (mallowsQ (sternBrocotRep α)) (i₀ + t)) =
        mallowsTFin (u.scanl mallowsQStep X) ∧
      mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₀) = X) ↔
    List.ofFn (fun t : Fin (u.length + 1) =>
        mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i₀ + t)) =
      u.scanl mallowsQStep X
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Lemma 3.
Uses 9

TheoremThe letter \(M\) in \(T(Q(\mathrm{SB}(\alpha)))\) marks the ends of runs#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, let \(x = Q(\mathrm{SB}(\alpha))\) be the sequence over \(\{O_L, E_R, O_R, E_L\}\) obtained from the Stern–Brocot representation \(\mathrm{SB}(\alpha)\), and let \(o(1) < o(2) < \cdots\) be the positions \(j\) with \(x_j \in \{O_L, O_R\}\). Let \(e_1, e_2, \ldots\) be positive integers such that \((x_{o(1)}, x_{o(2)}, \ldots) = O_L^{e_1} O_R^{e_2} O_L^{e_3} \cdots\), that is, for every \(k \ge 0\) and every \(i\) with \(e_1 + \cdots + e_k < i \le e_1 + \cdots + e_{k+1}\) (the empty sum being \(0\)), \(x_{o(i)}\) is \(O_L\) if \(k\) is even and \(O_R\) if \(k\) is odd. Then for every \(i \ge 1\), the \(i\)-th letter \(T_i(x)\) of \(T(x)\) is \(M\) if and only if \(i = e_1 + \cdots + e_k\) for some \(k \ge 1\).

The source assumes that \(O_L\) and \(O_R\) both occur infinitely often among the \(x_{o(i)}\); this is exactly what makes such \(e_k\) exist, so the run decomposition is taken as a hypothesis.

public theorem_wanted mallowsT_eq_M_iff_runEnd (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α)
    (e : ℕ → ℕ) (he : ∀ k, 0 < e k)
    (hrun : ∀ k i, ∑ t ∈ Finset.range k, e t ≤ i → i < ∑ t ∈ Finset.range (k + 1), e t →
      mallowsQ (sternBrocotRep α) (mallowsOddIndex (mallowsQ (sternBrocotRep α)) i) =
        if Even k then .OL else .OR)
    (i : ℕ) :
    mallowsT (mallowsQ (sternBrocotRep α)) i = .M ↔
      ∃ k, 1 ≤ k ∧ i + 1 = ∑ t ∈ Finset.range k, e t
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Lemma 4.
Uses 6

TheoremMallows and Stern–Brocot representations are eventually periodic together#●ready to prove

For every irrational \(\alpha \in (0, 1)\), the Mallows representation \(M(\alpha)\) is eventually periodic if and only if the Stern–Brocot representation \(\mathrm{SB}(\alpha)\) is eventually periodic.

public theorem_wanted isEventuallyPeriodic_mallowsRep_iff_sternBrocotRep (α : ℝ)
    (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    IsEventuallyPeriodic (mallowsRep α) ↔ IsEventuallyPeriodic (sternBrocotRep α)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 3.
Uses 5

TheoremLagrange's theorem for Mallows representations#●ready to prove

For every irrational \(\alpha \in (0, 1)\), the Mallows representation \(M(\alpha)\) is eventually periodic if and only if \(\alpha\) is a real quadratic irrational, that is, algebraic of degree \(2\) over \(\mathbb{Q}\).

public theorem_wanted isEventuallyPeriodic_mallowsRep_iff_quadratic (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    IsEventuallyPeriodic (mallowsRep α) ↔ IsRealQuadraticIrrational α
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 2.
Uses 4

TheoremPoisson stability of the Mallows and Stern–Brocot representations#●ready to prove

For every irrational \(\alpha \in (0, 1)\), the Mallows representation of \(\alpha\) is positively Poisson stable if and only if the Stern–Brocot representation of \(\alpha\) is positively Poisson stable.

public theorem_wanted isPositivelyPoissonStable_mallowsRep_iff_sternBrocotRep (α : ℝ)
    (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    IsPositivelyPoissonStable (mallowsRep α) ↔ IsPositivelyPoissonStable (sternBrocotRep α)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 4.
Uses 5

TheoremPoisson stability of the Mallows representation and the continued fraction#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, with continued fraction expansion \(\alpha = [0; a_1, a_2, \dots]\). Then the Mallows representation of \(\alpha\) is positively Poisson stable if and only if the sequence \((a_1, a_2, \dots)\) of its partial quotients is positively Poisson stable.

public theorem_wanted isPositivelyPoissonStable_mallowsRep_iff_contFract (α : ℝ)
    (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) (a : ℕ → ℕ)
    (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    IsPositivelyPoissonStable (mallowsRep α) ↔ IsPositivelyPoissonStable a
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 3.
Uses 3

TheoremTransitivity of the Mallows and Stern–Brocot representations#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational. Every finite word over the alphabet \(\{L, M, R\}\) occurs as a factor of the Mallows representation of \(\alpha\) if and only if every finite word over the alphabet \(\{L, R\}\) occurs as a factor of the Stern–Brocot representation of \(\alpha\); that is, the Mallows representation of \(\alpha\) is transitive if and only if its Stern–Brocot representation is. The empty word occurs trivially.

public theorem_wanted transitive_mallowsRep_iff_sternBrocotRep (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    (∀ w : List MallowsLetter, InfiniteWord.IsFactor w (mallowsRep α)) ↔
      ∀ w : List SternBrocotLetter, InfiniteWord.IsFactor w (sternBrocotRep α)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 5.
Uses 5

TheoremTransitive Mallows representations and continued fractions#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational with continued fraction expansion \(\alpha = [0; a_1, a_2, \ldots]\). The Mallows representation of \(\alpha\) is transitive, that is, every finite word over \(\{L, M, R\}\) occurs in it as a factor, if and only if every finite word over the positive integers occurs as a factor of the sequence \(a_1 a_2 a_3 \cdots\).

public theorem_wanted transitive_mallowsRep_iff_contFract (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α)
    (a : ℕ → ℕ) (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    (∀ w : List MallowsLetter, InfiniteWord.IsFactor w (mallowsRep α)) ↔
      ∀ w : List ℕ, (∀ c ∈ w, 0 < c) → InfiniteWord.IsFactor w a
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 4.
Uses 3

TheoremMallows representations without the letter \(M\)#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, with continued fraction expansion \(\alpha = [0; a_1, a_2, \ldots]\). The letter \(M\) does not occur in the Mallows representation \(M(\alpha)\) if and only if \(a_{2k}\) is even for every \(k \ge 1\).

public theorem_wanted mallowsRep_ne_M_iff (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α)
    (a : ℕ → ℕ)
    (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    (∀ i, mallowsRep α i ≠ .M) ↔ ∀ k, Even (a (2 * k + 1))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 9.
Uses 2

TheoremMallows representations without the letter \(R\)#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, with continued fraction expansion \(\alpha = [0; a_1, a_2, \ldots]\). The letter \(R\) does not occur in the Mallows representation \(M(\alpha)\) if and only if \(a_2 = 1\), \(a_3\) is even, and the sequence \(a_4, a_5, \ldots\) is a concatenation of blocks, each of the form \((2, b)\) or \((1, c, 1, b)\) with \(b\) even. Equivalently, \((a_1, a_2, \ldots) = (e'_1, 1, 2e'_2)\, w_1\, (2e'_3)\, w_2\, (2e'_4) \cdots\) with positive integers \(e'_k\) and each \(w_k\) equal to \((2)\) or \((1, e''_k, 1)\) for a positive integer \(e''_k\).

public theorem_wanted mallowsRep_ne_R_iff (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α)
    (a : ℕ → ℕ)
    (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    (∀ i, mallowsRep α i ≠ .R) ↔
      a 1 = 1 ∧ Even (a 2) ∧ ∃ s : ℕ → ℕ, s 0 = 3 ∧ ∀ k,
        (a (s k) = 2 ∧ Even (a (s k + 1)) ∧ s (k + 1) = s k + 2) ∨
        (a (s k) = 1 ∧ a (s k + 2) = 1 ∧ Even (a (s k + 3)) ∧ s (k + 1) = s k + 4)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 10.
Uses 2

TheoremMallows representations without the letter \(L\)#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, with continued fraction expansion \(\alpha = [0; a_1, a_2, \ldots]\). The letter \(L\) does not occur in the Mallows representation \(M(\alpha)\) if and only if \(a_1 = 1\), \(a_2\) is odd, and the sequence \(a_3, a_4, \ldots\) is a concatenation of blocks, each of the form \((2, b)\) or \((1, c, 1, b)\) with \(b\) even. Equivalently, \((a_1, a_2, \ldots) = (1, 2e'_1 - 1)\, w_1\, (2e'_2)\, w_2\, (2e'_3) \cdots\) with positive integers \(e'_k\) and each \(w_k\) equal to \((2)\) or \((1, e''_k, 1)\) for a positive integer \(e''_k\).

public theorem_wanted mallowsRep_ne_L_iff (α : ℝ) (hα : α ∈ Set.Ioo 0 1) (hirr : Irrational α)
    (a : ℕ → ℕ)
    (ha : ∀ k, (GenContFract.of α).partDens.get? k = some (a k : ℝ)) :
    (∀ i, mallowsRep α i ≠ .L) ↔
      a 0 = 1 ∧ Odd (a 1) ∧ ∃ s : ℕ → ℕ, s 0 = 2 ∧ ∀ k,
        (a (s k) = 2 ∧ Even (a (s k + 1)) ∧ s (k + 1) = s k + 2) ∨
        (a (s k) = 1 ∧ a (s k + 2) = 1 ∧ Even (a (s k + 3)) ∧ s (k + 1) = s k + 4)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 11.
Uses 2

TheoremThe numerator parities are not eventually constant#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational and let \(f_1(\alpha), f_2(\alpha), \dots\) be the fractions on its Stern–Brocot path. Then there are no natural numbers \(c\) and \(j_0\) such that the numerator of \(f_j(\alpha)\) has remainder \(c\) modulo \(2\) for all \(j \ge j_0\): the parities of these numerators are not eventually constant.

public theorem_wanted sternBrocotFrac_parity_not_eventually_const (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    ¬ ∃ c j₀, ∀ j, j₀ ≤ j → (sternBrocotFrac α j).1 % 2 = c
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 12.
Uses 1

TheoremEventually alternating parities of Stern–Brocot numerators#●ready to prove

Let \(\alpha\) be an irrational number with \(0 < \alpha < 1\), let \(f_1(\alpha), f_2(\alpha), \dots\) be the fractions on its Stern–Brocot path, and let \(M(\alpha) = \langle M_1(\alpha), M_2(\alpha), \dots \rangle\) be its Mallows representation. The following are equivalent: (a) there is \(j_0\) such that for every \(j \ge j_0\) the numerators of \(f_j(\alpha)\) and \(f_{j+1}(\alpha)\) have different parities; (b) the letter \(M\) occurs infinitely often in \(M(\alpha)\), and, writing \(m_1 < m_2 < \cdots\) for the positions \(i\) with \(M_i(\alpha) = M\), there is an integer \(k \ge 1\) such that for every \(t \ge 2k\) and every \(i\) with \(m_t < i < m_{t+1}\) we have \(M_i(\alpha) = R\) if \(t\) is even and \(M_i(\alpha) = L\) if \(t\) is odd.

In other words, from the \(2k\)-th occurrence of \(M\) on, \(M(\alpha) = M R^{e'_1} M L^{e'_2} M R^{e'_3} M \cdots\) with all \(e'_i \ge 0\).

public theorem_wanted parity_eventually_alternating_iff (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    (∃ j₀, ∀ j, j₀ ≤ j → (sternBrocotFrac α (j + 1)).1 % 2 ≠ (sternBrocotFrac α j).1 % 2) ↔
      {i | mallowsRep α i = .M}.Infinite ∧ ∃ k, 1 ≤ k ∧ ∀ t, 2 * k - 1 ≤ t → ∀ i,
        Nat.nth (fun n => mallowsRep α n = .M) t < i →
        i < Nat.nth (fun n => mallowsRep α n = .M) (t + 1) →
        mallowsRep α i = if Odd t then .R else .L
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 13.
Uses 3

TheoremEventual period three of the numerator parities#●ready to prove

Let \(\alpha \in (0, 1)\) be irrational, let \(f_1(\alpha), f_2(\alpha), \dots\) be the fractions on its Stern–Brocot path, with numerators \(p_1, p_2, \dots\), and let \(M(\alpha) = \langle M_1, M_2, \dots \rangle\) be its Mallows representation. Then there is \(j_0 \ge 1\) with \(p_{j+3} \equiv p_j \pmod 2\) for all \(j \ge j_0\) if and only if one of the following holds: (a) there is \(i_0 \ge 1\) such that \(M_{i_0 + 2t} = L\) and \(M_{i_0 + 2t + 1} = R\) for all \(t \ge 0\); (b) the letter \(M\) occurs infinitely often in \(M(\alpha)\), and so do the letters other than \(M\); writing \(u_1, u_2, \dots\) for the sequence obtained from \(M(\alpha)\) by deleting every \(M\), there is \(s \ge 1\) with \(u_{s + 2t} = L\) and \(u_{s + 2t + 1} = R\) for all \(t \ge 0\); and, writing \(m(1) < m(2) < \cdots\) for the positions of \(M\) in \(M(\alpha)\), there is \(k \ge 1\) such that for all \(k' \ge k\) we have \(M_{m(2k') - 1} = R\), \(M_{m(2k') + 1} = L\), \(M_{m(2k'+1) - 1} = L\) and \(M_{m(2k'+1) + 1} = R\).

public theorem_wanted parity_eventually_period_three_iff (α : ℝ) (hα : α ∈ Set.Ioo 0 1)
    (hirr : Irrational α) :
    (∃ j₀, ∀ j, j₀ ≤ j → (sternBrocotFrac α (j + 3)).1 % 2 = (sternBrocotFrac α j).1 % 2) ↔
      (∃ i₀, ∀ t, mallowsRep α (i₀ + 2 * t) = .L ∧ mallowsRep α (i₀ + 2 * t + 1) = .R) ∨
      ({i | mallowsRep α i = .M}.Infinite ∧ {i | mallowsRep α i ≠ .M}.Infinite ∧
        (∃ t₀, ∀ t,
          mallowsRep α (Nat.nth (fun n => mallowsRep α n ≠ .M) (t₀ + 2 * t)) = .L ∧
          mallowsRep α (Nat.nth (fun n => mallowsRep α n ≠ .M) (t₀ + 2 * t + 1)) = .R) ∧
        ∃ k, 1 ≤ k ∧ ∀ k', k ≤ k' →
          mallowsRep α (Nat.nth (fun n => mallowsRep α n = .M) (2 * k' - 1) - 1) = .R ∧
          mallowsRep α (Nat.nth (fun n => mallowsRep α n = .M) (2 * k' - 1) + 1) = .L ∧
          mallowsRep α (Nat.nth (fun n => mallowsRep α n = .M) (2 * k') - 1) = .L ∧
          mallowsRep α (Nat.nth (fun n => mallowsRep α n = .M) (2 * k') + 1) = .R)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 14.
Uses 3

TheoremTypical properties of Mallows representations#●ready to prove

Each of the following holds for Lebesgue-almost every \(\alpha \in (0, 1)\), provided \(\alpha\) is irrational: (i) the Mallows representation \(M(\alpha)\) is not eventually periodic; (ii) \(M(\alpha)\) is positively Poisson stable; (iii) \(M(\alpha)\) is transitive, that is, every finite word over \(\{L, M, R\}\) occurs in \(M(\alpha)\).

public theorem_wanted mallowsRep_ae_properties :
    (∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      ¬ IsEventuallyPeriodic (mallowsRep α)) ∧
    (∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      IsPositivelyPoissonStable (mallowsRep α)) ∧
    ∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      ∀ w : List MallowsLetter, InfiniteWord.IsFactor w (mallowsRep α)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Proposition 15.
Uses 5

TheoremFactor complexity of almost every Mallows representation#●ready to prove

For almost every irrational \(\alpha \in (0, 1)\), with respect to Lebesgue measure, the factor complexity of the Mallows representation \(M(\alpha)\) satisfies \(p_{M(\alpha)}(n) = 3^n\) for every \(n \ge 0\).

public theorem_wanted mallowsRep_ae_factorComplexity :
    ∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      ∀ n, wordFactorComplexity (mallowsRep α) n = 3 ^ n
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 5.
Uses 3

TheoremWords without bounded gaps in almost every Mallows representation#●ready to prove

For Lebesgue-almost every \(\alpha \in (0, 1)\), if \(\alpha\) is irrational, then there is a finite word \(w\) over \(\{L, M, R\}\) that does not occur relatively densely in the Mallows representation \(M(\alpha) = M_1(\alpha) M_2(\alpha) \cdots\): there is no \(\ell > 0\) such that, for every \(i \ge 0\), the window \(M_{i+1}(\alpha) \cdots M_{i+\ell}(\alpha)\) contains an occurrence of \(w\).

public theorem_wanted mallowsRep_ae_not_relativelyDense :
    ∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      ∃ w : List MallowsLetter, ¬ ∃ ℓ, 0 < ℓ ∧ ∀ i, ∃ j, i ≤ j ∧ j + w.length ≤ i + ℓ ∧
        InfiniteWord.factorList (mallowsRep α) w.length j = w
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Corollary 6.
Uses 3

TheoremWords of frequency zero in almost every Mallows representation#●ready to prove

Let \(w\) be a finite word over \(\{L, M, R\}\) that contains the letter \(M\) or contains both \(L\) and \(R\). Then for almost every irrational \(\alpha \in (0, 1)\), with respect to Lebesgue measure, \(w\) has frequency \(0\) in the Mallows representation \(M(\alpha) = M_0(\alpha) M_1(\alpha) M_2(\alpha) \cdots\): if \(N_n\) is the number of positions \(i < n\) with \(M_i(\alpha) M_{i+1}(\alpha) \cdots M_{i+|w|-1}(\alpha) = w\), then \(N_n / n \to 0\) as \(n \to \infty\).

public theorem_wanted mallowsRep_ae_frequency_zero (w : List MallowsLetter)
    (hw : MallowsLetter.M ∈ w ∨ (MallowsLetter.L ∈ w ∧ MallowsLetter.R ∈ w)) :
    ∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α →
      Filter.Tendsto (fun n : ℕ => ((((Finset.range n).filter fun i =>
          InfiniteWord.factorList (mallowsRep α) w.length i = w).card : ℕ) : ℝ) / n)
        Filter.atTop (nhds 0)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 6.
Uses 3

TheoremMatching fractions of almost every Mallows representation#●ready to prove

For almost every irrational \(\alpha \in (0, 1)\), with respect to Lebesgue measure, the Mallows representation \(M(\alpha) = M_0(\alpha) M_1(\alpha) M_2(\alpha) \cdots\) satisfies, for every \(\ell \ge 0\),

\[\lim_{n \to \infty} \frac{|\{i < n : M_i(\alpha) = M_{i+\ell}(\alpha)\}|}{n} = 1,\]

that is, its \(\ell\)-th matching fraction is \(1\).

For \(\ell = 0\) the statement is trivial; the source states it for \(\ell \ge 1\).

public theorem_wanted mallowsRep_ae_matchingFraction_one :
    ∀ᵐ α ∂(MeasureTheory.volume.restrict (Set.Ioo (0 : ℝ) 1)), Irrational α → ∀ ℓ : ℕ,
      Filter.Tendsto (fun n : ℕ => ((((Finset.range n).filter fun i =>
          mallowsRep α i = mallowsRep α (i + ℓ)).card : ℕ) : ℝ) / n) Filter.atTop (nhds 1)
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 7.
Uses 2

TheoremExtremal measures of Mallows cylinder sets#●ready to prove

Let \(\ell \ge 0\). For a word \(w = w_1 \cdots w_\ell\) over \(\{L, M, R\}\), let \(C_w\) be the set of irrational \(\alpha \in (0, 1)\) whose Mallows representation begins with \(w\), that is, \(M_t(\alpha) = w_t\) for \(1 \le t \le \ell\). Over the \(3^\ell\) words \(w\) of length \(\ell\), the Lebesgue measure of \(C_w\) attains the maximum value \(1/(\ell + 1)\) and the minimum value

\[\frac{4\sqrt{2}}{(1 + \sqrt{2})^{2\ell + 2} - (1 - \sqrt{2})^{2\ell + 2}}.\]

For \(\ell = 0\) both values equal \(1\).

public theorem_wanted mallowsRep_cylinder_measure_extrema (ℓ : ℕ) :
    IsGreatest (Set.range fun w : Fin ℓ → MallowsLetter =>
        MeasureTheory.volume.real
          {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α ∧ ∀ t : Fin ℓ, mallowsRep α t = w t})
      (1 / (ℓ + 1)) ∧
    IsLeast (Set.range fun w : Fin ℓ → MallowsLetter =>
        MeasureTheory.volume.real
          {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α ∧ ∀ t : Fin ℓ, mallowsRep α t = w t})
      (4 * √2 / ((1 + √2) ^ (2 * ℓ + 2) - (1 - √2) ^ (2 * ℓ + 2)))
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 8.
Uses 2

TheoremMeasure bound for the letter \(M\) at a fixed position#●ready to prove

For every \(i \ge 1\), the set of irrational \(\alpha \in (0, 1)\) whose Mallows representation has \(i\)-th letter \(M_i(\alpha) = M\) has Lebesgue measure less than \(3 - 2\sqrt{2}\).

public theorem_wanted mallowsRep_eq_M_measure_lt (i : ℕ) :
    MeasureTheory.volume.real {α : ℝ | α ∈ Set.Ioo 0 1 ∧ Irrational α ∧ mallowsRep α i = .M} <
      3 - 2 * √2
Sources
  • Takanori Hida, The Mallows Representation of the Irrational Numbers, INTEGERS 25 (2025), Article A78, Theorem 9.
Uses 2
47 fully proved34 ready to prove

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.