Adelic Carleman–Weierstrass Regularized Determinants (\(\det_2, \det_3\)), Schwinger 2-Cocycle Central Extensions, Segal–Wilson Determinant Bundles, \(L^1(\mathbb{R})\) Weil Positivity, and Formal Bridges to Signals.RiemannSpectral in Lean 4, Rust/PyO3, and mpmath¶
Authors: Wes Turner
Subjects:
Number Theory (
math.NT); Functional Analysis (math.FA); Mathematical Physics (math-ph); Logic & Formal Verification (cs.LO)
Epistemic Status & Scope Declaration:
Verified Infinite-Dimensional Mathematical Core (Machine-Checked in Lean 4, Rust/PyO3, and 40-Digit
mpmath): We formalize 10 zero-sorryLean 4 modules underSignals.SatoGrassmannian(RHGaps,RemainingGaps,CarlemanWeierstrass,SchwingerCocycle,DeterminantBundle,BigCellCocycle,SegalWilsonGr24,WeilPositivity,WeilGaussianL1,RiemannSpectralBridge), proving: (1) the four classical RH analytical gaps (de Bruijn–Newman backward heat flow, Erdős \(\sum 1/p = \infty\) and \(\sum p^{-1/2} = \infty\), Li/Cayley unit-circle characterization, degree-2 Jensen–Pólya Turán equivalence, and Kato–Rellich operator closure), (2) order-\(k\) Weierstrass–Carleman regularized determinants \(\det_k\) (det2for \(\operatorname{Re}(s) > 1/2\),det3convergent and non-vanishing on the critical line \(\operatorname{Re}(s) = 1/2\), and the multiplicative anomaly \(\det_2((1+A)(1+B)) = \det_2(1+A)\det_2(1+B)e^{-\operatorname{Tr}(AB)}\)), (3) the Schwinger 2-cocycle \(\omega(M_{z^{-1}}, M_z) = -1\), central extension \(1 \to \mathbb{C}^\times \to \widehat{\mathrm{GL}}_{(2)}^0(H) \to \mathrm{GL}_{(2)}^0(H) \to 1\), Segal–Wilson determinant bundle \(\mathrm{Det}^* \to \mathrm{Gr}_{(2)}^0(H)\), and exact reduction \(\det_{\mathrm{Fr}}(1 + g(t)_{+-} T_U g(t)_{++}^{-1}) = \tau_U(t_1, t_2, t_3) / p_{23}\), (4) bidirectional \(L^1(\mathbb{R})\) Weil positivity equivalence using shifted Gaussians with an explicit \(-4\pi m(z_0) < 0\) off-line disproof witness, and (5) six formal Lean 4 bridge theorems connecting the infinite-dimensional stack directly toSignals.RiemannSpectral.Open Millennium-Problem & Axiom Boundary (
Signals.Pending.SatoRiemannSpectralPendingBoundary): This paper does not prove the Riemann Hypothesis [Bombieri, 2000]. InSignals.Pending.sato_infinite_dim_and_pde_stack_do_not_close_rh_or_physical_claims, we explicitly record that certifying \(\omega(S^{-1}, S) = -1\), critical-line \(\det_3 \ne 0\), and bidirectional Weil equivalence does not discharge: (i) the four globalRB-6RH bridges (\(\Lambda \le 0\), Hilbert–Pólya spectral realization, global Weil \(L^1\) positivity for \(\zeta\), and adelic trace limits), or (ii) the classical Lidskii trace-classFredholmDetTheoryaxiom bundle inDeterminantBundle.leanawaiting upstream Mathlib trace-ideal theory.
Primary Source Chat Provenance:
data/chats/Gemini_Formalizing-Sato-Grassmannian-in-Lean_20261120-2337.md(13,272lines,sha256: c8eb7b93d618846add0564db6fd4df248aff7a12a6e63d1808712a6673f1a6ab), specifically Blocks 008–011 (lines 4336–7401) and Blocks 013–017 (lines 7756–11306).
Companion Papers:
Formal Verification & Computational Artifacts:
SignalsLean 4 Library (3,830Lake jobs viamake signals_build, zerosorry):src/signals/Signals/SatoGrassmannian/RHGaps.leanthroughsrc/signals/Signals/SatoGrassmannian/RiemannSpectralBridge.leansrc/signals/Signals/Pending.lean(SatoRiemannSpectralPendingBoundary)
signals-numerical-adapterRust + PyO3 (maturin) Extension (7/7sato_grassmannianRust unit tests passing):src/signals/numerical_adapter/src/sato_grassmannian.rs(evaluate_carleman_euler_products,evaluate_weil_gaussian_quad_form,evaluate_segal_wilson_gr24_tau,evaluate_riemann_spectral_bridge)
Python + SymPy + 40-Digit
mpmath+ PyO3 Verification Suite (8/8pytesttests passing):
Abstract¶
Our prior Riemann spectral papers established finite-dimensional Grassmannian, Li, Jensen–Pólya, de Bruijn–Newman, Weil, and celestial-holography contracts in Signals.RiemannSpectral, isolating four open infinite-dimensional analytical bridges (RB-6). Here we lift that program into genuine infinite-dimensional Hilbert-space analysis in Lean 4 (Signals.SatoGrassmannian.*), formalizing both the exact operator-theoretic machinery of Sato–Segal–Wilson determinant bundles and the exact mathematical gaps preventing naive Euler-product or Hilbert–Pólya constructions from proving the Riemann Hypothesis.
We establish five main results, verified across Lean 4 (zero sorry), Rust/PyO3, SymPy, and 40-digit mpmath:
Formalized RH Gaps & Archimedean Deficiency Analysis (
RHGaps.lean,RemainingGaps.lean): We prove the local de Bruijn–Newman backward-heat collision dichotomy [Rodgers and Tao, 2020], divergence of \(\sum_p p^{-1}\) and \(\sum_p p^{-1/2}\) (proving the prime Euler operator is neither trace-class on \(\operatorname{Re}(s) = 1\) nor Hilbert–Schmidt on \(\operatorname{Re}(s) = 1/2\)), Li’s unit-circle and off-line exponential blowup theorems [Li, 1997], degree-2 Jensen–Pólya Turán hyperbolicity [Griffin et al., 2019], Kato–Rellich symmetric-operator closure [Kato, 1995], and Archimedean log-coordinate dilation deficiency indices (deficiency_plus_unique,deficiency_minus_unique).Order-\(k\) Weierstrass–Carleman Determinants & Critical-Line \(\det_3\) Convergence (
CarlemanWeierstrass.lean): Using Weierstrass primary factors \(E_k(z) = (1 + z)\exp\bigl(\sum_{j=1}^{k-1}(-1)^j z^j / j\bigr)\) [Simon, 2005], we prove that \(\det_2\) converges and never vanishes for \(\operatorname{Re}(s) > 1/2\), obeys the exact multiplicative anomaly \(\det_2((1+A)(1+B)) = \det_2(1+A)\det_2(1+B)e^{-\operatorname{Tr}(AB)}\) (det2Diag_mul_anomaly), and that order-3 \(\det_3\) converges and never vanishes directly on the critical line \(\operatorname{Re}(s) = 1/2\) (det3_eulerPrimes_critical_line), since \(\sum_p p^{-3/2} < \infty\). At the first Riemann zero \(\gamma_1 \approx 14.13472514\), Rust and 40-digitmpmathagree on \(\lvert\det_3(\tfrac{1}{2} + i\gamma_1)\rvert = 1.05685678145613\) to \(2.35 \times 10^{-15}\).Schwinger 2-Cocycle, \(\widehat{\mathrm{GL}}_{(2)}^0(H)\) Central Extension, and Exact Segal–Wilson \(\mathrm{Gr}(2,4)\) \(\tau\)-Function (
SchwingerCocycle.lean,DeterminantBundle.lean,BigCellCocycle.lean,SegalWilsonGr24.lean): We construct the Lie-algebra Schwinger 2-cocycle \(\omega(X, Y) = \operatorname{Tr}(X_{+-}Y_{-+} - Y_{+-}X_{-+})\), prove its Jacobi identity and non-triviality \(\omega(M_{z^{-1}}, M_z) = -1\) (schwingerCocycle_shiftInv_shift), construct the group central extension \(1 \to \mathbb{C}^\times \to \widehat{\mathrm{GL}}_{(2)}^0(H) \to \mathrm{GL}_{(2)}^0(H) \to 1\) and Segal–Wilson determinant bundle \(\mathrm{Det}^* \to \mathrm{Gr}_{(2)}^0(H)\) moduloFredholmDetTheory[Segal and Wilson, 1985, Pressley and Segal, 1986], and prove insegalWilson_tau_gr24_eq_tauUthat the infinite-dimensional Fredholm determinant \(\det_{\mathrm{Fr}}(1 + g(t)_{+-} T_U g(t)_{++}^{-1})\) reduces identically to the finite Schur polynomial \(\tau_U(t_1, t_2, t_3) / p_{23}\) (0.0Rust/SymPy mismatch).Bidirectional \(L^1(\mathbb{R})\) Shifted-Gaussian Weil Positivity Equivalence (
WeilPositivity.lean,WeilGaussianL1.lean): Upgrading abstract Mellin positivity [Weil, 1952, Connes, 1999] to genuine \(f \in L^1(\mathbb{R})\) test functions built from shifted Gaussians \(g_\delta(x) = e^{-(x-\delta)^2}\),weilPositivity_L1_iff_allRealZerosproves that a conjugate-symmetric finite zero multiset lies entirely on \(\mathbb{R}\) if and only if the Weil Hermitian form is nonnegative for all \(f \in L^1(\mathbb{R})\), exhibiting an explicit \(-4\pi m(z_0) < 0\) witness for any off-line zero (evaluated numerically at \(-8\pi = -25.132741 < -4\pi = -12.566371\) against the first 10 true Riemann zeros).Formal Lean 4 Bridge to
Signals.RiemannSpectral(RiemannSpectralBridge.lean): Six machine-checked bridge theorems unify the finiteSignals.RiemannSpectralmodule withSignals.SatoGrassmannian.*across Li/Cayley, dBN backward-heat collisions, Jensen–Pólya Madelung–Bohm vs. Turán, Wronskian Hirota KP lifts, finite-to-infinite Kato–Rellich promotion, and celestial chiral Casimir reality.
1. Architecture of the Adelic Carleman, Determinant-Bundle, Weil, & Bridge Stack¶
flowchart TB
subgraph Gaps["RH Gaps & Carleman Regularization"]
O1["RHGaps.lean (Block 008)\ndBN collision, Erdős ∑1/p = ∞,\nCayley/Li, Jensen-2, Kato–Rellich"]
O2["RemainingGaps.lean (Block 009)\nCarleman det_2 (σ > 1/2),\nArchimedean deficiency indices"]
O3["CarlemanWeierstrass.lean (Block 010)\nOrder-k E_k(z), det_2 anomaly,\ncritical-line det_3 ≠ 0 (∑p^-3/2 < ∞)"]
end
subgraph Bundle["Central Extension & Segal–Wilson Det* Bundle"]
D1["SchwingerCocycle.lean (Block 011)\nIsHS, GL_2^res, Jacobi identity,\nω(S^-1, S) = -1"]
D2["DeterminantBundle.lean (Block 014)\n1 → C* → GL̂_(2)^0 → GL_(2)^0 → 1,\nDet* → Gr_(2)^0(H), segalWilson_tau_bigCell"]
D3["BigCellCocycle.lean & SegalWilsonGr24.lean\n(Blocks 015–016)\nBig-cell splitting, det_Fr = τ_U / p_23"]
end
subgraph Weil["L¹(ℝ) Weil Positivity & RiemannSpectral Bridge"]
W1["WeilPositivity.lean & WeilGaussianL1.lean\n(Blocks 013, 017)\nL¹(ℝ) shifted Gaussians ↔ AllRealZeros\n(-4π m(z_0) < 0 off-line witness)"]
W2["RiemannSpectralBridge.lean\nFormal Lean 4 bridge linking\nSignals.RiemannSpectral ↔ SatoGrassmannian"]
W3["Pending.lean\nSatoRiemannSpectralPendingBoundary\n(4 open RH bridges + FredholmDetTheory)"]
end
O1 --> O2 --> O3 --> W2
D1 --> D2 --> D3 --> W2
W1 --> W2 --> W3
2. Carleman–Weierstrass Regularization Across \(\operatorname{Re}(s) = 1\), \(1/2\), and \(1/3\)¶
A central gap proven in RHGaps.lean (lines 4336–5037) is that the unregularized Euler product \(\det_1(1 - P_s) = \prod_p (1 - p^{-s}) = \zeta(s)^{-1}\) fails to define a trace-class Fredholm determinant on the boundary \(\operatorname{Re}(s) = 1\) because \(\sum_p p^{-1} = \infty\), and fails to define a Hilbert–Schmidt Carleman determinant \(\det_2\) on the critical line \(\operatorname{Re}(s) = 1/2\) because \(\sum_p (p^{-1/2})^2 = \sum_p p^{-1} = \infty\).
In CarlemanWeierstrass.lean (lines 5783–6609), we generalize to the order-\(k\) Weierstrass–Carleman factor [Simon, 2005]:
For \(\lambda_p(s) = -p^{-s}\):
\(k = 1\) (trace class, \(\mathcal{I}_1\)): Converges absolutely for \(\operatorname{Re}(s) > 1\) (\(\sum_p p^{-\sigma} < \infty\)), recovering \(\zeta(s)^{-1}\).
\(k = 2\) (Hilbert–Schmidt, \(\mathcal{I}_2\)): Converges absolutely and never vanishes for \(\operatorname{Re}(s) > 1/2\) (
RemainingGaps.lean, lines 5115–5667;CarlemanWeierstrass.lean), obeying the multiplicative anomaly:
\(k = 3\) (Schatten \(\mathcal{I}_3\)): Converges absolutely and never vanishes for \(\operatorname{Re}(s) > 1/3\), and in particular everywhere on the critical line \(s = \tfrac{1}{2} + i t\) (
det3_eulerPrimes_critical_line), because \(\sum_p p^{-3/2} < \infty\). Crucially, because \(\det_3(\tfrac{1}{2} + i t) \ne 0\) identically, the zeros of \(\zeta(\tfrac{1}{2} + i t)\) cannot arise from vanishing of individual Euler factors — they live entirely in the divergent counterterm exponential \(\exp\bigl(\sum_p (p^{-s} + \tfrac{1}{2} p^{-2s})\bigr)\)!
3. Schwinger 2-Cocycle, \(\widehat{\mathrm{GL}}_{(2)}^0(H)\) Central Extension, and Exact Segal–Wilson \(\tau\)-Function¶
On the Hilbert–Schmidt restricted Lie algebra \(\mathfrak{gl}_{(2)}(H)\) (SchwingerCocycle.lean, lines 6746–7401), the cross-block commutator trace defines the Schwinger 2-cocycle:
We prove \(\omega\) is skew-symmetric, obeys the Lie-algebra 2-cocycle Jacobi identity (schwingerCocycle_jacobi, centralBracket_jacobi), and evaluates on the commuting shift pair \((S^{-1}, S)\) on \(L^2(S^1)\) to:
yielding the canonical Heisenberg central term \([\widehat{S}^{-1}, \widehat{S}] = -c\) (schwingerCocycle_shiftInv_shift).
At the group level (DeterminantBundle.lean, lines 8157–9706; BigCellCocycle.lean, lines 9811–10002), relative to the classical trace-class Fredholm determinant axiom bundle FredholmDetTheory [Simon, 2005], we construct the exact central extension:
the dual determinant line bundle \(\mathrm{Det}^* \to \mathrm{Gr}_{(2)}^0(H)\) with its canonical holomorphic section \(\sigma\) vanishing precisely off the big cell (canonicalSection_ne_zero_iff_bigCell), the weight-\(+1\) equivariant action of \(\widehat{\mathrm{GL}}_{(2)}^0(H)\), and the big-cell Segal–Wilson \(\tau\)-function (segalWilson_tau_bigCell).
In SegalWilsonGr24.lean (lines 10014–11022), specializing to the 3-time KP flow \(g(t) = \exp(t_1 z + t_2 z^2 + t_3 z^3)\) and a big-cell 2-plane \(U \in \mathrm{Gr}_{\mathbb{C}}(2,4)\) (\(p_{23} \ne 0\)) with finite-rank graph operator \(T_U : H_+ \to H_-\), we prove segalWilson_tau_gr24_eq_tauU:
closing the full bridge from infinite-dimensional Fredholm determinant bundles back to the exact Schur-polynomial KP \(\tau\)-function of Paper 1.
4. Bidirectional \(L^1(\mathbb{R})\) Weil Positivity via Shifted Gaussians¶
In WeilPositivity.lean (lines 7756–8004) and WeilGaussianL1.lean (lines 11030–11306), we formalize the Weil explicit-formula quadratic form [Weil, 1952] on a conjugate-symmetric finite zero multiset \((Z, m)\):
Using real shifted Gaussians \(g_\delta(x) = e^{-(x-\delta)^2} \in L^1(\mathbb{R})\) with closed-form Mellin/Fourier transform \(\widehat{g_\delta}(z) = \sqrt{\pi}\,e^{i \delta z - z^2/4}\) (mellinTransform_shiftedGaussian) and an explicit Vandermonde separation shift \(\delta\) (exists_delta_exp_injOn), weilPositivity_L1_iff_allRealZeros proves both directions without assuming compactly supported or Schwartz test functions:
Whenever any off-line zero \(z_0 \in Z\) has \(\operatorname{Im}(z_0) \ne 0\), the proof constructs an explicit finite linear combination of shifted Gaussians achieving \(Q_{Z, m}(f) = -4\pi m(z_0) < 0\) (or \(-8\pi\) summed over the conjugate pair \(\{z_0, \bar{z}_0\}\)).
5. Formal Lean 4 Bridge to Signals.RiemannSpectral (RiemannSpectralBridge.lean)¶
To unify our existing finite Signals.RiemannSpectral library with the new infinite-dimensional Signals.SatoGrassmannian.* stack, RiemannSpectralBridge.lean proves six exact Lean 4 bridge theorems:
Bridge Theorem in |
Connects |
To |
Mathematical Meaning |
|---|---|---|---|
|
|
|
\(|1 - 1/\rho|^2 = \mathrm{liTransformModulusSq}(\sigma_{\mathrm{dev}}, \gamma)\); unit circle \(\iff \sigma_{\mathrm{dev}} = 0\) |
|
|
|
\(H_t(\Delta z)\) at \(z_2 = -2\) equals \(-\operatorname{Re}(\mathrm{localDoubleZero}(t + z_0/2, \Delta z))\) |
|
|
|
Degree-2 Jensen real-rootedness \(\iff J_{\mathrm{osmotic}} + J_{\mathrm{Bohm}} \ge 0\) |
|
|
|
Real Wronskian minors lift to \(\mathcal{Q}_4\) and solve Hirota’s KP-II bilinear equation |
|
|
|
Finite real diagonal spectra promote to everywhere-defined |
|
|
|
For \(\gamma \ne 0\), \(\operatorname{Im}(h(h-1)) = 0 \iff |1 - 1/\rho|^2 = 1 \iff \sigma_{\mathrm{dev}} = 0\) |
6. Tri-Language Numerical Verification (mpmath 40-Digit vs. Rust/PyO3)¶
Running pytest -q tests/test_sato_grassmannian_sympy_pyo3.py cross-verifies the Lean 4 theorems against compiled Rust/PyO3 (signals_numerical_adapter) and 40-digit mpmath, including live evaluation on the first 10 nontrivial Riemann zeros computed by py_find_first_n_riemann_zeros(10) from riemann_spectral.rs:
Quantity Audited |
Parameters |
40-Digit |
Rust/PyO3 ( |
Absolute Discrepancy |
|---|---|---|---|---|
Classical strip \(|\det_1(2 + 5i)|\) |
\(168\) primes (\(p \le 1000\)) |
\(1.167258402893353\) (vs. \(1/|\zeta| = 1.16726702\)) |
\(1.167258402893354\) |
\(1.03 \times 10^{-15}\) |
Classical strip \(|\det_2(2 + 5i)|\) |
\(168\) primes (\(p \le 1000\)) |
\(0.978948714184111\) |
\(0.978948714184107\) |
\(3.81 \times 10^{-15}\) |
Critical-line \(|\det_3(\tfrac{1}{2} + i\gamma_1)|\) |
\(\gamma_1 = 14.13472514\), \(239\) primes (\(p \le 1500\)) |
\(1.056856781456128\) |
\(1.056856781456131\) |
\(2.35 \times 10^{-15}\) |
Critical-line \(\sum_{p \le 1500} p^{-3/2}\) |
\(\sigma = 1/2\), \(239\) primes |
Strictly convergent (\(< 2\)) |
\(0.843891898264092\) |
Exact finite tail |
\(\det_2\) multiplicative anomaly residual |
\(\sigma = 1/2, t = \gamma_1\) |
\(0\) ( |
\(1.77 \times 10^{-15}\) |
\(1.77 \times 10^{-15}\) |
Segal–Wilson \(\det_{\mathrm{Fr}}\) vs. \(\tau_U/p_{23}\) |
\(t = (0.35, -0.20, 0.15)\) |
\(0.918618732284581\) |
\(0.918618732284581\) |
\(0.0\) |
First 10 true Riemann zeros: Li/Cayley |
\(\gamma_1 \dots \gamma_{10}\), \(\sigma_{\mathrm{dev}} = 0\) |
\(0\) ( |
max err \(2.22 \times 10^{-16}\) |
\(2.22 \times 10^{-16}\) |
Weil \(L^1\) form on 10 true critical-line zeros |
3 shifted Gaussians |
\(\ge 0\) ( |
\(\mathrm{Re} = 5.02 \times 10^{-43}\), \(\mathrm{Im} = 0.0\) |
Exact real \(\ge 0\) |
Weil \(L^1\) off-line disproof witness |
\(z_0 = \gamma_1 \pm 0.15 i\) |
\(-8\pi = -25.132741 < -4\pi = -12.566371\) |
\(-25.13274122871833\) |
\(< 10^{-14}\) |
7. Open Boundaries Recorded in Signals.Pending.SatoRiemannSpectralPendingBoundary¶
In accordance with the repository’s evidence-modeling discipline, Signals.Pending.sato_infinite_dim_and_pde_stack_do_not_close_rh_or_physical_claims formally records four explicit open gates:
Four Open RH Bridges (
RiemannHypothesisPendingBoundary): Proving \(\Lambda \le 0\) globally, constructing an adelic Sato trace limit, establishing global Weil/Li/Jensen–Pólya positivity across all zeros of \(\zeta(s)\), and constructing a genuinely self-adjoint Hilbert–Pólya operator whose spectrum equals \(\{\gamma_n\}\) remain open Millennium-problem milestones.Trace-Class Lidskii
FredholmDetTheoryBundle:DeterminantBundle.leancleanly axiomatizes Simon’s trace-class Fredholm determinant properties (detFr, multiplicativity,detFr_rankOne_nilpotent_eq_one,detFr_blockLowerTriangular_two) as a typeclass interface pending full Schatten-class / Lidskii trace theory in upstream Mathlib.Global Infinite-Zero Weil Positivity:
WeilGaussianL1.leanproves bidirectional equivalence for finite conjugate-symmetric zero multisets; extending the off-line witness to the infinite zero set of \(\zeta(s)\) requires the unconditional Weil explicit formula convergence estimates on \(L^1(\mathbb{R})\) test functions.Physical-Model Boundary: Exact mathematical equivalence between
Signals.RiemannSpectralandSignals.SatoGrassmannianis an analytical/algebraic result and makes no unverified physical claims.
References¶
G Segal and G Wilson. Loop groups and equations of kdv type. 1985. Referenced in docs/grassmannian-quantum-causal-jev-bundle-transformer-videogen-paper.md.
Andrew Pressley and Graeme Segal. Loop Groups. Oxford Mathematical Monographs. Oxford University Press, 1986. URL: https://global.oup.com/academic/product/loop-groups-9780198535614.
Barry Simon. Trace Ideals and Their Applications. Volume 120 of Mathematical Surveys and Monographs. American Mathematical Society, 2nd edition, 2005. URL: https://dx.doi.org/10.1090/surv/120.
André Weil. Sur les “formules explicites” de la théorie des nombres premiers. Communications du Séminaire Mathématique de l'Université de Lund, Tome Supplémentaire:252–265, 1952.
Enrico Bombieri. Problems of the millennium: the Riemann hypothesis. Clay Mathematics Institute, 2000. URL: https://www.claymath.org/millennium/riemann-hypothesis/.
Brad Rodgers and Terence Tao. The de Bruijn–Newman constant is non-negative. Forum of Mathematics, Pi, 8:e6, 2020. doi:10.1017/fmp.2020.6.
Xian-Jin Li. The positivity of a sequence of numbers and the Riemann hypothesis. Journal of Number Theory, 65(2):325–333, 1997. doi:10.1006/jnth.1997.2137.
Michael Griffin, Ken Ono, Larry Rolen, and Don Zagier. Jensen polynomials for the Riemann zeta function and other sequences. Proceedings of the National Academy of Sciences, 116(23):11103–11110, 2019. doi:10.1073/pnas.1902572116.
Tosio Kato. Perturbation Theory for Linear Operators. Classics in Mathematics. Springer, 2nd edition, 1995. URL: https://dx.doi.org/10.1007/978-3-642-66282-9.
Alain Connes. Trace formula in noncommutative geometry and the zeros of the Riemann zeta function. Selecta Mathematica, 5(1):29–106, 1999. URL: https://dx.doi.org/10.1007/s000290050042, arXiv:math/9811068, doi:10.1007/s000290050042.