Total positivity of Riordan arrays
TheoremBell-type Riordan arrays with a Pólya frequency A-sequence are totally positive#●ready to prove
Let \(g\) be a formal power series over \(\mathbb{R}\) with \(g(0) \neq 0\), write \([x^n]F\) for the coefficient of \(x^n\) in a power series \(F\), and let \(D = (d_{n,k})_{n,k \ge 0}\) be the Bell-type Riordan array \((g(x), x g(x))\), with entries \(d_{n,k} = [x^n]\, g(x) (x g(x))^k\). Let \((a_j)_{j \ge 0}\) be a real sequence, the \(A\)-sequence of \(D\), such that for all \(n, k \ge 0\)
If \((a_j)\) is a Pólya frequency sequence, that is, every minor of its lower Toeplitz matrix (whose entry in row \(n\) and column \(k\) is \(a_{n-k}\) for \(k \le n\) and \(0\) otherwise) is nonnegative, then \(D\) is totally positive: every minor of \(D\) is nonnegative.
The sum stops at \(j = n\) because \(d_{n,m} = 0\) for \(m > n\).
public theorem_wanted bellRiordan_hasNonnegativeMinors_of_aSequence (g : PowerSeries ℝ)
(hg : PowerSeries.constantCoeff g ≠ 0) (a : ℕ → ℝ)
(ha : ∀ n k : ℕ, PowerSeries.coeff (n + 1) (g * (PowerSeries.X * g) ^ (k + 1)) =
∑ j ∈ Finset.range (n + 1),
a j * PowerSeries.coeff n (g * (PowerSeries.X * g) ^ (k + j)))
(hpf : Matrix.HasNonnegativeMinors (Matrix.lowerToeplitz a)) :
Matrix.HasNonnegativeMinors
(Matrix.of fun n k : ℕ => PowerSeries.coeff n (g * (PowerSeries.X * g) ^ k))
public def Matrix.HasNonnegativeMinors {ι κ : Type*} [LinearOrder ι] [LinearOrder κ]
(M : Matrix ι κ ℝ) : Prop :=
∀ (k : ℕ) (r : Fin k → ι) (c : Fin k → κ),
StrictMono r → StrictMono c → 0 ≤ (M.submatrix r c).det
MathlibExt/LinearAlgebra/Matrix/NonnegativeMinors.lean:32
def lowerToeplitz {R : Type*} [Zero R] (a : ℕ → R) : Matrix ℕ ℕ R :=
fun i j => if j ≤ i then a (i - j) else 0
MathlibExt/LinearAlgebra/Matrix/LowerToeplitz.lean:31
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Proposition 5.
Uses 2
DefinitionRiordan array with given A- and Z-sequences#✓fully proved
For real sequences \((a_n)_{n \ge 0}\) and \((z_n)_{n \ge 0}\), an infinite real matrix \(R = (r_{n,k})_{n,k \ge 0}\) is the Riordan array with \(A\)-sequence \((a_n)\) and \(Z\)-sequence \((z_n)\) when \(r_{0,0} = 1\), \(r_{0,k} = 0\) for \(k \ge 1\), and for all \(n, k \ge 0\)
The sums stop at \(j = n\) because the recurrence forces \(r_{n,m} = 0\) for \(m > n\).
def IsRiordanArrayAZ (a z : ℕ → ℝ) (R : ℕ → ℕ → ℝ) : Prop :=
R 0 0 = 1 ∧ (∀ k, R 0 (k + 1) = 0) ∧
(∀ n, R (n + 1) 0 = ∑ j ∈ Finset.range (n + 1), z j * R n j) ∧
∀ n k, R (n + 1) (k + 1) = ∑ j ∈ Finset.range (n + 1), a j * R n (k + j)
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Section 1, equation (1).
TheoremEntries in row 1 of a Riordan array#✓fully proved
If \(R = (r_{n,k})_{n,k \ge 0}\) is the Riordan array with \(A\)-sequence \((a_n)\) and \(Z\)-sequence \((z_n)\), for real sequences \((a_n)\) and \((z_n)\), then \(r_{1,0} = z_0\) and \(r_{1,1} = a_0\).
theorem IsRiordanArrayAZ.one_zero_and_one_one {a z : ℕ → ℝ} {R : ℕ → ℕ → ℝ}
(h : IsRiordanArrayAZ a z R) : R 1 0 = z 0 ∧ R 1 1 = a 0
def IsRiordanArrayAZ (a z : ℕ → ℝ) (R : ℕ → ℕ → ℝ) : Prop :=
R 0 0 = 1 ∧ (∀ k, R 0 (k + 1) = 0) ∧
(∀ n, R (n + 1) 0 = ∑ j ∈ Finset.range (n + 1), z j * R n j) ∧
∀ n k, R (n + 1) (k + 1) = ∑ j ∈ Finset.range (n + 1), a j * R n (k + j)
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:52
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Section 1, equation (1).
Uses 1
DefinitionCoefficient matrix of a Riordan array#✓fully proved
For real sequences \((z_n)_{n \ge 0}\) and \((a_n)_{n \ge 0}\) with generating functions \(Z(x)\) and \(A(x)\), the coefficient matrix \(J(Z(x), A(x))\) is the infinite real matrix, with rows and columns indexed by \(n, k \ge 0\), whose entry in row \(n\) and column \(k\) is \(z_n\) if \(k = 0\), is \(a_{n+1-k}\) if \(1 \le k \le n + 1\), and is \(0\) if \(k > n + 1\). Thus row \(n\) reads \(z_n, a_n, a_{n-1}, \ldots, a_0, 0, 0, \ldots\)
def riordanCoeffMatrix (z a : ℕ → ℝ) : Matrix ℕ ℕ ℝ :=
Matrix.of fun n k => if k = 0 then z n else if k ≤ n + 1 then a (n + 1 - k) else 0
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Section 1.
TheoremFirst two columns of the coefficient matrix of a Riordan array#✓fully proved
For real sequences \((z_n)\) and \((a_n)\) with generating functions \(Z(x)\) and \(A(x)\), and for every \(n \ge 0\), the coefficient matrix \(J(Z(x), A(x))\) has entry \(z_n\) in row \(n\), column \(0\), and entry \(a_n\) in row \(n\), column \(1\).
theorem riordanCoeffMatrix_zero_one (z a : ℕ → ℝ) (n : ℕ) :
riordanCoeffMatrix z a n 0 = z n ∧ riordanCoeffMatrix z a n 1 = a n
def riordanCoeffMatrix (z a : ℕ → ℝ) : Matrix ℕ ℕ ℝ :=
Matrix.of fun n k => if k = 0 then z n else if k ≤ n + 1 then a (n + 1 - k) else 0
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:74
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Section 1.
Uses 1
TheoremTotal positivity of a Riordan array from its coefficient matrix#●ready to prove
Let \((a_n)\) and \((z_n)\) be real sequences with generating functions \(A(x)\) and \(Z(x)\), and let \(R\) be the Riordan array with \(A\)-sequence \((a_n)\) and \(Z\)-sequence \((z_n)\). If the coefficient matrix \(J(Z(x), A(x))\) is totally positive, that is, every minor of it is nonnegative, then every minor of \(R\) is nonnegative.
public theorem_wanted riordanArray_hasNonnegativeMinors_of_coeffMatrix {a z : ℕ → ℝ}
{R : ℕ → ℕ → ℝ} (hR : IsRiordanArrayAZ a z R)
(hJ : Matrix.HasNonnegativeMinors (riordanCoeffMatrix z a)) :
Matrix.HasNonnegativeMinors (Matrix.of R)
def IsRiordanArrayAZ (a z : ℕ → ℝ) (R : ℕ → ℕ → ℝ) : Prop :=
R 0 0 = 1 ∧ (∀ k, R 0 (k + 1) = 0) ∧
(∀ n, R (n + 1) 0 = ∑ j ∈ Finset.range (n + 1), z j * R n j) ∧
∀ n k, R (n + 1) (k + 1) = ∑ j ∈ Finset.range (n + 1), a j * R n (k + j)
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:52
public def Matrix.HasNonnegativeMinors {ι κ : Type*} [LinearOrder ι] [LinearOrder κ]
(M : Matrix ι κ ℝ) : Prop :=
∀ (k : ℕ) (r : Fin k → ι) (c : Fin k → κ),
StrictMono r → StrictMono c → 0 ≤ (M.submatrix r c).det
MathlibExt/LinearAlgebra/Matrix/NonnegativeMinors.lean:32
def riordanCoeffMatrix (z a : ℕ → ℝ) : Matrix ℕ ℕ ℝ :=
Matrix.of fun n k => if k = 0 then z n else if k ≤ n + 1 then a (n + 1 - k) else 0
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:74
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Proposition 1 (
jis-2025-25.7.5-prop-1).
TheoremA total positivity criterion for Riordan arrays#●ready to prove
Let \(P, Q, A, Z\) be formal power series over \(\mathbb{R}\) with \(P(0) \ne 0\), write \([x^n]F\) for the coefficient of \(x^n\) in a power series \(F\), and let \(a, b, r, s, t \ge 0\) be real numbers. Suppose that
and
and let \(R\) be the Riordan array whose \(A\)-sequence is \(([x^n]A)_{n \ge 0}\) and whose \(Z\)-sequence is \(([x^n]Z)_{n \ge 0}\). Let \(J(Q(x), P(x))\) be the coefficient matrix whose row \(n \ge 0\) reads \([x^n]Q, [x^n]P, [x^{n-1}]P, \dots, [x^0]P, 0, 0, \dots\). Call a matrix totally positive when every minor of it is nonnegative. If \(J(Q(x), P(x))\) is totally positive, \(s^2 \ge 4t\) and \(as + a\sqrt{s^2 - 4t} \ge 2b\), then \(R\) is totally positive.
Since \(P(0) \ne 0\), the series \(P(x) - rx\) and \(P(x)^2 - rxP(x)\) are invertible, so the two hypotheses say \(A = (P^2 + sxP + tx^2)/(P - rx)\) and \(Z = ((P^2 + sxP + tx^2)Q + (aP + bx)(P - xQ))/(P^2 - rxP)\).
public theorem_wanted riordanArray_hasNonnegativeMinors_of_params (P Q A Z : PowerSeries ℝ)
(hP : PowerSeries.constantCoeff P ≠ 0) {a b r s t : ℝ} (ha : 0 ≤ a) (hb : 0 ≤ b)
(hr : 0 ≤ r) (hs : 0 ≤ s) (ht : 0 ≤ t)
(hA : A * (P - PowerSeries.C r * PowerSeries.X) =
P ^ 2 + PowerSeries.C s * PowerSeries.X * P + PowerSeries.C t * PowerSeries.X ^ 2)
(hZ : Z * (P ^ 2 - PowerSeries.C r * PowerSeries.X * P) =
(P ^ 2 + PowerSeries.C s * PowerSeries.X * P + PowerSeries.C t * PowerSeries.X ^ 2) * Q +
(PowerSeries.C a * P + PowerSeries.C b * PowerSeries.X) * (P - PowerSeries.X * Q))
{R : ℕ → ℕ → ℝ}
(hR : IsRiordanArrayAZ (fun n => PowerSeries.coeff n A) (fun n => PowerSeries.coeff n Z) R)
(hJ : Matrix.HasNonnegativeMinors
(riordanCoeffMatrix (fun n => PowerSeries.coeff n Q) (fun n => PowerSeries.coeff n P)))
(hst : 4 * t ≤ s ^ 2) (hab : 2 * b ≤ a * s + a * Real.sqrt (s ^ 2 - 4 * t)) :
Matrix.HasNonnegativeMinors (Matrix.of R)
def IsRiordanArrayAZ (a z : ℕ → ℝ) (R : ℕ → ℕ → ℝ) : Prop :=
R 0 0 = 1 ∧ (∀ k, R 0 (k + 1) = 0) ∧
(∀ n, R (n + 1) 0 = ∑ j ∈ Finset.range (n + 1), z j * R n j) ∧
∀ n k, R (n + 1) (k + 1) = ∑ j ∈ Finset.range (n + 1), a j * R n (k + j)
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:52
public def Matrix.HasNonnegativeMinors {ι κ : Type*} [LinearOrder ι] [LinearOrder κ]
(M : Matrix ι κ ℝ) : Prop :=
∀ (k : ℕ) (r : Fin k → ι) (c : Fin k → κ),
StrictMono r → StrictMono c → 0 ≤ (M.submatrix r c).det
MathlibExt/LinearAlgebra/Matrix/NonnegativeMinors.lean:32
def riordanCoeffMatrix (z a : ℕ → ℝ) : Matrix ℕ ℕ ℝ :=
Matrix.of fun n k => if k = 0 then z n else if k ≤ n + 1 then a (n + 1 - k) else 0
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:74
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Theorem 2 (
jis-2025-25.7.5-thm-2).
TheoremTotal positivity of Bell-type Riordan arrays from a Pólya frequency series#●ready to prove
Let \(P\) and \(A\) be formal power series over \(\mathbb{R}\) with \(P(0) \ne 0\), write \([x^n]F\) for the coefficient of \(x^n\) in a power series \(F\), and let \(r, s, t \ge 0\) be real numbers with \(s^2 \ge 4t\). Suppose that \(P\) is a Pólya frequency series, that is, every minor of the lower Toeplitz matrix of its coefficients (whose entry in row \(n\) and column \(k\) is \([x^{n-k}]P\) for \(k \le n\) and \(0\) otherwise) is nonnegative, and that
Let \(R\) be the Riordan array with \(A\)-sequence \((a_n)_{n \ge 0}\) and \(Z\)-sequence \((z_n)_{n \ge 0}\), where \(a_n = [x^n]A\) and \(z_n = a_{n+1}\) (a Bell-type Riordan array). Then \(R\) is totally positive: every minor of \(R\) is nonnegative.
Since \(P(0) \ne 0\), the series \(P(x) - rx\) is invertible, so the hypothesis on \(A\) says \(A = (P^2 + sxP + tx^2)/(P - rx)\).
public theorem_wanted bellRiordanArray_hasNonnegativeMinors_of_pf (P A : PowerSeries ℝ)
(hP : PowerSeries.constantCoeff P ≠ 0) {r s t : ℝ} (hr : 0 ≤ r) (hs : 0 ≤ s) (ht : 0 ≤ t)
(hpf : Matrix.HasNonnegativeMinors (Matrix.lowerToeplitz fun n => PowerSeries.coeff n P))
(hA : A * (P - PowerSeries.C r * PowerSeries.X) =
P ^ 2 + PowerSeries.C s * PowerSeries.X * P + PowerSeries.C t * PowerSeries.X ^ 2)
{R : ℕ → ℕ → ℝ}
(hR : IsRiordanArrayAZ (fun n => PowerSeries.coeff n A)
(fun n => PowerSeries.coeff (n + 1) A) R)
(hst : 4 * t ≤ s ^ 2) :
Matrix.HasNonnegativeMinors (Matrix.of R)
public def Matrix.HasNonnegativeMinors {ι κ : Type*} [LinearOrder ι] [LinearOrder κ]
(M : Matrix ι κ ℝ) : Prop :=
∀ (k : ℕ) (r : Fin k → ι) (c : Fin k → κ),
StrictMono r → StrictMono c → 0 ≤ (M.submatrix r c).det
MathlibExt/LinearAlgebra/Matrix/NonnegativeMinors.lean:32
def lowerToeplitz {R : Type*} [Zero R] (a : ℕ → R) : Matrix ℕ ℕ R :=
fun i j => if j ≤ i then a (i - j) else 0
MathlibExt/LinearAlgebra/Matrix/LowerToeplitz.lean:31
def IsRiordanArrayAZ (a z : ℕ → ℝ) (R : ℕ → ℕ → ℝ) : Prop :=
R 0 0 = 1 ∧ (∀ k, R 0 (k + 1) = 0) ∧
(∀ n, R (n + 1) 0 = ∑ j ∈ Finset.range (n + 1), z j * R n j) ∧
∀ n k, R (n + 1) (k + 1) = ∑ j ∈ Finset.range (n + 1), a j * R n (k + j)
WantedExt/Combinatorics/Enumerative/BellRiordanTotalPositivityWanted.lean:52
Sources
- Lili Mu, Yuanyuan Xing and Sai-Nan Zheng, A New Criterion for the Total Positivity of Riordan Arrays, Journal of Integer Sequences 28 (2025), Article 25.7.5, Corollary 6 (
jis-2025-25.7.5-cor-6).
Uses 3
Loading dependency explorer…
The dependency explorer needs JavaScript and its graph data.