Self-convolutions of m-step Fibonacci numbers

DefinitionThe \(m\)-step Fibonacci recurrence#✓fully proved

A sequence \(F : \mathbb{Z} \to \mathbb{Q}\) satisfies the \(m\)-step Fibonacci recurrence if \(F_n = F_{n-1} + F_{n-2} + \dots + F_{n-m}\) for every integer \(n\), that is, \(F_n = \sum_{i=0}^{m-1} F_{n-i-1}\). For \(m = 0\) this says \(F_n = 0\) for all \(n\).

def IsHigherFibRecurrence (m : ℕ) (F : ℤ → ℚ) : Prop :=
  ∀ n : ℤ, F n = ∑ i ∈ Finset.range m, F (n - (i : ℤ) - 1)
Used by 2

DefinitionThe standard order-\(m\) Fibonacci sequence#✓fully proved

A sequence \(F : \mathbb{Z} \to \mathbb{Q}\) is the standard order-\(m\) Fibonacci sequence if it satisfies the \(m\)-step Fibonacci recurrence, \(F_i = 0\) for every natural number \(i < m - 1\), and \(F_{m-1} = 1\), where \(m - 1\) is computed in the integers.

def IsStandardHigherFib (m : ℕ) (F : ℤ → ℚ) : Prop :=
  IsHigherFibRecurrence m F ∧
    (∀ i : ℕ, i < m - 1 → F (i : ℤ) = 0) ∧
    F ((m : ℤ) - 1) = 1
Uses 1 · Used by 3

DefinitionWeighted sum of a sequence#✓fully proved

For a sequence \(F : \mathbb{Z} \to \mathbb{Q}\), a natural number \(m\), integers \(N\) and \(l\), and a rational number \(q\), \(S(m,N,l,q) = \sum_{k=0}^{N} q^k F_{k+l}\); for \(N < 0\) this is the empty sum \(0\).

The order \(m\) is carried explicitly to match the paper notation, although the sum itself only uses \(F\), \(N\), \(l\), and \(q\).

def weightedHigherFibSum (F : ℤ → ℚ) (_m : ℕ) (N l : ℤ) (q : ℚ) : ℚ :=
  if N < 0 then
    0
  else
    ∑ k ∈ Finset.range (N.toNat + 1), q ^ (k : ℤ) * F ((k : ℤ) + l)
Used by 3

DefinitionFirst-moment weighted sum#✓fully proved

For a sequence \(F : \mathbb{Z} \to \mathbb{Q}\), a natural number \(m\), integers \(N\) and \(l\), and a rational number \(q\), \(S_2(m,N,l,q) = \sum_{k=0}^{N} k q^k F_{k+l}\); for \(N < 0\) this is the empty sum \(0\). The value does not depend on \(m\).

def weightedHigherFibMomentSum (F : ℤ → ℚ) (_m : ℕ) (N l : ℤ)
    (q : ℚ) : ℚ :=
  if N < 0 then
    0
  else
    ∑ k ∈ Finset.range (N.toNat + 1),
      (k : ℚ) * q ^ (k : ℤ) * F ((k : ℤ) + l)
Used by 2

DefinitionShifted self-convolution#✓fully proved

For a sequence \(F : \mathbb{Z} \to \mathbb{Q}\) and natural numbers \(m\) and \(n\), the shifted self-convolution is \(w_n^{(m)} = \sum_{j=0}^{n} F_{j+m-1} F_{n-j+m-1}\), with the indices computed in the integers.

def higherFibWalk (F : ℤ → ℚ) (m n : ℕ) : ℚ :=
  ∑ j ∈ Finset.range (n + 1),
    F ((j : ℤ) + (m : ℤ) - 1) * F ((n : ℤ) - (j : ℤ) + (m : ℤ) - 1)
Used by 3

AbbreviationFibonacci numbers at integer indices#✓fully proved

For an integer \(n\), \(F_n\) is the \(n\)-th Fibonacci number as a rational number: \(F_0 = 0\), \(F_1 = 1\), \(F_{n+2} = F_{n+1} + F_n\) for \(n \ge 0\), and \(F_{-n} = (-1)^{n+1} F_n\) for \(n > 0\).

abbrev fibQ (n : ℤ) : ℚ :=
  (Int.fib n : ℚ)
Used by 5

DefinitionFibonacci self-convolution#✓fully proved

For \(n \ge 0\), \(v_n\) is the rational number \(v_n = \sum_{j=0}^{n} F_{j+1} F_{n-j+1}\), where \(F_i\) is the \(i\)-th Fibonacci number (\(F_0 = 0\), \(F_1 = 1\)).

This sum encodes algebraically the number of Fibonacci tiling walks.

def fibWalkCount (n : ℕ) : ℚ :=
  ∑ j ∈ Finset.range (n + 1),
    fibQ ((j : ℤ) + 1) * fibQ ((n : ℤ) - (j : ℤ) + 1)
Uses 1 · Used by 2

DefinitionEven part of the Fibonacci self-convolution#✓fully proved

For \(n \ge 0\), \(E_n\) is the rational number \(\sum F_{j+1} F_{n-j+1}\), the sum taken over the even integers \(j\) with \(0 \le j \le n\), where \(F_i\) is the \(i\)-th Fibonacci number (\(F_0 = 0\), \(F_1 = 1\)).

def evenFibWalkCount (n : ℕ) : ℚ :=
  ∑ j ∈ (Finset.range (n + 1)).filter Even,
    fibQ ((j : ℤ) + 1) * fibQ ((n : ℤ) - (j : ℤ) + 1)
Uses 1 · Used by 4

DefinitionOdd part of the Fibonacci walk count#✓fully proved

For a natural number \(n\), this is the sum of \(F_{j+1} F_{n-j+1}\) over the odd \(j\) with \(0 \le j \le n\), where \(F_k\) are the Fibonacci numbers (\(F_0 = 0\), \(F_1 = 1\), \(F_{k+2} = F_{k+1} + F_k\)). It is the part of \(v_n = \sum_{j=0}^{n} F_{j+1} F_{n-j+1}\) that comes from odd \(j\).

def oddFibWalkCount (n : ℕ) : ℚ :=
  ∑ j ∈ (Finset.range (n + 1)).filter Odd,
    fibQ ((j : ℤ) + 1) * fibQ ((n : ℤ) - (j : ℤ) + 1)
Uses 1 · Used by 4

TheoremRecurrence for the shifted self-convolution of higher-order Fibonacci numbers#●ready to prove

Let \(m \ge 1\) and let \((F_n)_{n \in \mathbb{Z}}\) be a sequence of rational numbers with \(F_n = F_{n-1} + \dots + F_{n-m}\) for every integer \(n\), \(F_i = 0\) for \(0 \le i < m - 1\) and \(F_{m-1} = 1\), and for \(n \ge 0\) let \(w_n = \sum_{j=0}^{n} F_{j+m-1} F_{n-j+m-1}\). Then for every integer \(n \ge 0\), \((m+1) w_{n+1} - 2m w_n = (n+m+2) F_{n+m} - (n+m) F_{n+m-1}\).

theorem_wanted higherFibWalk_recurrence
    (m : ℕ) (hm : 1 ≤ m) (F : ℤ → ℚ)
    (hF : IsStandardHigherFib m F) (n : ℕ) :
    ((m : ℚ) + 1) * higherFibWalk F m (n + 1) -
        2 * (m : ℚ) * higherFibWalk F m n =
      ((n : ℚ) + (m : ℚ) + 2) * F ((n : ℤ) + (m : ℤ)) -
        ((n : ℚ) + (m : ℚ)) * F ((n : ℤ) + (m : ℤ) - 1)
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 1.
Uses 2

TheoremClosed form for weighted sums of an \(m\)-step Fibonacci sequence#●ready to prove

Let \(m \ge 1\) and let \(F : \mathbb{Z} \to \mathbb{Q}\) satisfy the \(m\)-step Fibonacci recurrence \(F_k = F_{k-1} + \dots + F_{k-m}\) for every integer \(k\). For every natural number \(n\), every integer \(l\) and every rational number \(q\) with \(q^{m+1} - 2q + 1 \ne 0\), the weighted sum \(S(m,n,l,q) = \sum_{k=0}^{n} q^k F_{k+l}\) equals

\[ \frac{(q-1)\Bigl(q^{n+1}F_{n+l+1} - F_l + \sum_{i=1}^{m-1}\sum_{j=1}^{i} q^{i+1-j}\bigl(q^{n+1}F_{n+l-j+1} - F_{l-j}\bigr)\Bigr)} {q^{m+1} - 2q + 1}. \]
theorem_wanted weightedHigherFibSum_eq
    (m : ℕ) (hm : 1 ≤ m) (F : ℤ → ℚ)
    (hF : IsHigherFibRecurrence m F)
    (n : ℕ) (l : ℤ) (q : ℚ)
    (hq : q ^ (m + 1) - 2 * q + 1 ≠ 0) :
    weightedHigherFibSum F m (n : ℤ) l q =
      ((q - 1) *
        (q ^ (n + 1) * F ((n : ℤ) + l + 1) - F l +
          ∑ i ∈ Finset.Icc 1 (m - 1),
            ∑ j ∈ Finset.Icc 1 i,
              q ^ (i + 1 - j) *
                (q ^ (n + 1) * F ((n : ℤ) + l - (j : ℤ) + 1) -
                  F (l - (j : ℤ))))) /
        (q ^ (m + 1) - 2 * q + 1)
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 2.
Uses 2

TheoremExplicit formula for the shifted self-convolution of higher-order Fibonacci numbers#●ready to prove

Let \(m \ge 1\) and let \((F_n)_{n \in \mathbb{Z}}\) be a sequence of rational numbers with \(F_n = F_{n-1} + \dots + F_{n-m}\) for every integer \(n\), \(F_i = 0\) for \(0 \le i < m - 1\) and \(F_{m-1} = 1\). For \(n \ge 0\) let \(w_n = \sum_{j=0}^{n} F_{j+m-1} F_{n-j+m-1}\); for integers \(N\), \(l\) and rational \(q\) let \(S(N, l, q) = \sum_{k=0}^{N} q^k F_{k+l}\) and \(S_2(N, l, q) = \sum_{k=0}^{N} k q^k F_{k+l}\), both equal to \(0\) when \(N < 0\). Put \(q = (m+1)/(2m)\). Then for every integer \(n \ge 0\),

\[ w_n = \left(\frac{2m}{m+1}\right)^{n} + \frac{(2m)^{n-1}}{(m+1)^{n}} \bigl((m+2) S(n-1, m, q) + S_2(n-1, m, q) - m S(n-1, m-1, q) - S_2(n-1, m-1, q)\bigr). \]

At \(n = 0\) the weighted sums have upper limit \(-1\) and are empty.

MathlibExt.NumberTheory.Recurrences.HigherFibonacciWalksWanted.higherFibWalk_eq_weighted_sums · 16 lines
theorem_wanted higherFibWalk_eq_weighted_sums
    (m : ℕ) (hm : 1 ≤ m) (F : ℤ → ℚ)
    (hF : IsStandardHigherFib m F) (n : ℕ) :
    higherFibWalk F m n =
      (2 * (m : ℚ) / ((m : ℚ) + 1)) ^ n +
        ((2 * (m : ℚ)) ^ ((n : ℤ) - 1) / ((m : ℚ) + 1) ^ n) *
          (((m : ℚ) + 2) *
              weightedHigherFibSum F m ((n : ℤ) - 1) (m : ℤ)
                (((m : ℚ) + 1) / (2 * (m : ℚ))) +
            weightedHigherFibMomentSum F m ((n : ℤ) - 1) (m : ℤ)
              (((m : ℚ) + 1) / (2 * (m : ℚ))) -
            (m : ℚ) *
              weightedHigherFibSum F m ((n : ℤ) - 1) ((m : ℤ) - 1)
                (((m : ℚ) + 1) / (2 * (m : ℚ))) -
            weightedHigherFibMomentSum F m ((n : ℤ) - 1)
              ((m : ℤ) - 1) (((m : ℚ) + 1) / (2 * (m : ℚ))))
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 3.
Uses 4

TheoremSelf-convolution of higher-order Fibonacci numbers#●ready to prove

Let \(m \ge 1\) and let \((F_n)_{n \in \mathbb{Z}}\) be a sequence of rational numbers with \(F_n = F_{n-1} + \dots + F_{n-m}\) for every integer \(n\), \(F_i = 0\) for \(0 \le i < m - 1\) and \(F_{m-1} = 1\). For \(p \ge 0\) let \(w_p = \sum_{j=0}^{p} F_{j+m-1} F_{p-j+m-1}\); for integers \(N\), \(l\) and rational \(q\) let \(S(N, l, q) = \sum_{k=0}^{N} q^k F_{k+l}\) and \(S_2(N, l, q) = \sum_{k=0}^{N} k q^k F_{k+l}\), both equal to \(0\) when \(N < 0\). Put \(q = (m+1)/(2m)\). Then for every integer \(n \ge 2m - 2\), writing \(p = n - 2m + 2\),

\[ \sum_{i=0}^{n} F_i F_{n-i} = w_p = \left(\frac{2m}{m+1}\right)^{p} + \frac{(2m)^{p-1}}{(m+1)^{p}} \bigl((m+2) S(p-1, m, q) + S_2(p-1, m, q) - m S(p-1, m-1, q) - S_2(p-1, m-1, q)\bigr). \]

The source states the formula for all \(n \ge 0\), but for \(m \ge 2\) it fails when \(n < 2m - 2\): at \(m = 2\) and \(n = 0\) the left side is \(0\) and the right side is \(9/16\).

MathlibExt.NumberTheory.Recurrences.HigherFibonacciWalksWanted.higherFib_selfConvolution_eq_weighted_sums · 21 lines
theorem_wanted higherFib_selfConvolution_eq_weighted_sums
    (m : ℕ) (hm : 1 ≤ m) (F : ℤ → ℚ)
    (hF : IsStandardHigherFib m F) (n : ℕ) (hn : 2 * m - 2 ≤ n) :
    (∑ i ∈ Finset.range (n + 1), F (i : ℤ) * F ((n : ℤ) - (i : ℤ))) =
        higherFibWalk F m (n - (2 * m - 2)) ∧
      (∑ i ∈ Finset.range (n + 1), F (i : ℤ) * F ((n : ℤ) - (i : ℤ))) =
        (2 * (m : ℚ) / ((m : ℚ) + 1)) ^ ((n : ℤ) - 2 * (m : ℤ) + 2) +
          ((2 * (m : ℚ)) ^ ((n : ℤ) - 2 * (m : ℤ) + 1) /
              ((m : ℚ) + 1) ^ ((n : ℤ) - 2 * (m : ℤ) + 2)) *
            (((m : ℚ) + 2) *
                weightedHigherFibSum F m ((n : ℤ) - 2 * (m : ℤ) + 1)
                  (m : ℤ) (((m : ℚ) + 1) / (2 * (m : ℚ))) +
              weightedHigherFibMomentSum F m
                ((n : ℤ) - 2 * (m : ℤ) + 1) (m : ℤ)
                (((m : ℚ) + 1) / (2 * (m : ℚ))) -
              (m : ℚ) *
                weightedHigherFibSum F m ((n : ℤ) - 2 * (m : ℤ) + 1)
                  ((m : ℤ) - 1) (((m : ℚ) + 1) / (2 * (m : ℚ))) -
              weightedHigherFibMomentSum F m
                ((n : ℤ) - 2 * (m : ℤ) + 1) ((m : ℤ) - 1)
                (((m : ℚ) + 1) / (2 * (m : ℚ))))
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 4.
Uses 4

TheoremEven and odd parts of the Fibonacci self-convolution agree at odd length#●ready to prove

Let \(F_n\) denote the Fibonacci numbers (\(F_0 = 0\), \(F_1 = 1\)) and, for \(n \ge 0\), let \(E_n\) and \(O_n\) be the sums of \(F_{j+1} F_{n-j+1}\) over the even and over the odd integers \(j\) with \(0 \le j \le n\). For every integer \(k \ge 1\), \(E_{2k-1} = O_{2k-1}\).

theorem_wanted evenFibWalkCount_odd_length_eq_oddFibWalkCount
    (k : ℕ) (hk : 1 ≤ k) :
    evenFibWalkCount (2 * k - 1) = oddFibWalkCount (2 * k - 1)
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Lemma 5.
Uses 2

TheoremEven and odd parts of the Fibonacci self-convolution at even length#●ready to prove

Let \(F_n\) denote the Fibonacci numbers (\(F_0 = 0\), \(F_1 = 1\)) and, for \(n \ge 0\), let \(E_n\) and \(O_n\) be the sums of \(F_{j+1} F_{n-j+1}\) over the even and over the odd integers \(j\) with \(0 \le j \le n\). For every integer \(k \ge 1\), \(E_{2k} = O_{2k} + F_{2k+2}\).

theorem_wanted evenFibWalkCount_even_length_eq_odd_add_fib
    (k : ℕ) (hk : 1 ≤ k) :
    evenFibWalkCount (2 * k) =
      oddFibWalkCount (2 * k) + fibQ (2 * (k : ℤ) + 2)
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Lemma 5.
Uses 3

TheoremHalving the Fibonacci self-convolution at odd length#●ready to prove

Let \(F_n\) denote the Fibonacci numbers (\(F_0 = 0\), \(F_1 = 1\)) and, for \(n \ge 0\), let \(v_n = \sum_{j=0}^{n} F_{j+1} F_{n-j+1}\), and let \(E_n\) and \(O_n\) be the sums of the same terms \(F_{j+1} F_{n-j+1}\) over the even and over the odd integers \(j\) with \(0 \le j \le n\). For every integer \(k \ge 1\), \(E_{2k-1} = v_{2k-1}/2\) and \(O_{2k-1} = v_{2k-1}/2\).

theorem_wanted fibWalkCounts_odd_length_eq_half
    (k : ℕ) (hk : 1 ≤ k) :
    evenFibWalkCount (2 * k - 1) = fibWalkCount (2 * k - 1) / 2 ∧
      oddFibWalkCount (2 * k - 1) = fibWalkCount (2 * k - 1) / 2
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 6.
Uses 3

TheoremHalving the Fibonacci self-convolution at even length#●ready to prove

Let \(F_n\) denote the Fibonacci numbers (\(F_0 = 0\), \(F_1 = 1\)) and, for \(n \ge 0\), let \(v_n = \sum_{j=0}^{n} F_{j+1} F_{n-j+1}\), and let \(E_n\) and \(O_n\) be the sums of the same terms \(F_{j+1} F_{n-j+1}\) over the even and over the odd integers \(j\) with \(0 \le j \le n\). For every integer \(k \ge 1\), \(E_{2k} = (v_{2k} + F_{2k+2})/2\) and \(O_{2k} = (v_{2k} - F_{2k+2})/2\).

theorem_wanted fibWalkCounts_even_length_eq
    (k : ℕ) (hk : 1 ≤ k) :
    evenFibWalkCount (2 * k) =
        (fibWalkCount (2 * k) + fibQ (2 * (k : ℤ) + 2)) / 2 ∧
      oddFibWalkCount (2 * k) =
        (fibWalkCount (2 * k) - fibQ (2 * (k : ℤ) + 2)) / 2
Sources
  • Jan Sedlák and Antonín Slavík, "Walks in Tiled Rectangles and Self-Convolutions of Higher-Order Fibonacci Numbers", Theorem 6.
Uses 4
9 fully proved8 ready to prove

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.