The Cohn–Elkies exponent and sign uncertainty

5.2. The Mellin ansatz🔗

Parameters, shells, perturbation, envelope and polynomials of the construction (Section 4.2).

Definition5.2.1
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.2
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 0
Used by 6
Reverse dependency previews
Preview
Definition 5.2.2
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

Put \lambda = d/2; throughout the construction \epsilon is fixed before d \to \infty. Constants in O_\epsilon(\cdot), \ll_\epsilon may depend on \epsilon but not on d or on the saddle parameter. Introduce cutoffs 0 < a_0 < A < B, a positive-shell amplitude Q > 0, and u_0 = 1 + \dfrac{\epsilon}{4}, U = 1 + \dfrac{\epsilon}{2}, C_0 = A + a_0^{-1}. The requirements as \epsilon \downarrow 0 are a_0 = o(\epsilon), e^{-2A}/A = o(\epsilon), \epsilon A = o(1), A = o(B), BQe^{(u_0-1)B} = o(1) and C_0Q^{-1}e^{-(U-1)(B-A)} = o(1). The realization used (equation (34)) is a_0 = \epsilon^2, A = \log(1/\epsilon), B = \epsilon^{-3}, q_\epsilon = \dfrac{(u_0-1)+(U-1)}{2} = \dfrac{3\epsilon}{8}, Q = e^{-q_\epsilon B}, b(a) = 1 - 2\epsilon(1+a), \beta = u_0 - 1 = \dfrac{\epsilon}{4}. For sufficiently small \epsilon, b > 0 on [a_0, A] and B > A + 1.

Lean code for Definition5.2.1●6 definitions
  • complete
    def CohnElkies.a₀ε (ε : ℝ) : ℝ
    def CohnElkies.a₀ε (ε : ℝ) : ℝ
    The parameter `a₀ = ε²` of the upper bound construction. 
  • complete
    def CohnElkies.Aε (ε : ℝ) : ℝ
    def CohnElkies.Aε (ε : ℝ) : ℝ
    The parameter `A = log (1/ε)` of the upper bound construction. 
  • complete
    def CohnElkies.Bε (ε : ℝ) : ℝ
    def CohnElkies.Bε (ε : ℝ) : ℝ
    The parameter `B = ε⁻³` of the upper bound construction. 
  • complete
    def CohnElkies.Qε (ε : ℝ) : ℝ
    def CohnElkies.Qε (ε : ℝ) : ℝ
    The shell weight `Q = exp (-3 ε B / 8)` of the upper bound construction. 
  • complete
    def CohnElkies.bε (ε a : ℝ) : ℝ
    def CohnElkies.bε (ε a : ℝ) : ℝ
    The parameter `b(a) = 1 - 2 ε (1 + a)` of the upper bound construction. 
  • complete
    def CohnElkies.β (ε : ℝ) : ℝ
    def CohnElkies.β (ε : ℝ) : ℝ
    The parameter `β = ε / 4` of the polynomials `P₊`, `P₋`. 
Definition5.2.2
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 5
Reverse dependency previews
Preview
Definition 5.2.3
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

With the parameters of Definition 5.2.1, the negative shell is (equation (35)) w_s(a) = -\dfrac{b(a)e^{-2a}}{2a^2\cosh a}\,\mathbf 1_{[a_0,A]}(a).

Lean code for Definition5.2.2●1 definition
  • complete
    def CohnElkies.w_s (ε a : ℝ) : ℝ
    def CohnElkies.w_s (ε a : ℝ) : ℝ
    The negative shell density `w_s(a) = -b_ε(a) e^{-2a} / (2a² cosh a)` of report (35), carried
    on `[a₀, A]`. 
Definition5.2.3
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
Statement uses 2
Statement dependency previews
Preview
Definition 5.2.1
Loading preview
Statement dependency preview content is loaded from the rendered-fragment cache.
Used by 6
Reverse dependency previews
Preview
Definition 5.2.4
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

With the parameters of Definition 5.2.1, the positive shell is (equation (35)) w_B(a) = \dfrac{Q}{\cosh a}\,\mathbf 1_{[B,B+1]}(a), and the signed density of the construction is w = w_s + w_B (Definition 5.2.2).

Lean code for Definition5.2.3●1 definition
  • complete
    def CohnElkies.w_B (ε a : ℝ) : ℝ
    def CohnElkies.w_B (ε a : ℝ) : ℝ
    The positive shell density `w_B(a) = Q_ε / cosh a` of report (35), carried on `[B, B+1]`. 
Definition5.2.4
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 3
Reverse dependency previews
Preview
Lemma 5.2.5
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

The signed density w of Definition 5.2.3 determines the even entire function h_\epsilon(\zeta) = \int_0^\infty w(a)\bigl(\cos(a\zeta) - 1\bigr)\,da (equation (36)).

Lean code for Definition5.2.4●1 definition
  • complete
    def CohnElkies.h_ε (ε : ℝ) (z : ℂ) : ℂ
    def CohnElkies.h_ε (ε : ℝ) (z : ℂ) : ℂ
    The entire even Mellin perturbation `h_ε(ζ) = ∫ w(a)(cos(aζ) - 1) da` of report (36). 
Lemma5.2.5
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1used by 1✓L∃∀N

The perturbation h_\epsilon of Definition 5.2.4 is even and real on the real and imaginary axes, with h_\epsilon(iu) = \int_0^\infty w(a)(\cosh(au) - 1)\,da and ih_\epsilon'(iu) = \int_0^\infty w(a)\,a\sinh(ua)\,da.

Lean code for Lemma5.2.5●5 declarations
  • complete
    theorem CohnElkies.mellinShellPhase_neg (ε : ℝ) (z : ℂ) :
      CohnElkies.h_ε ε (-z) = CohnElkies.h_ε ε z
    theorem CohnElkies.mellinShellPhase_neg (ε : ℝ)
      (z : ℂ) :
      CohnElkies.h_ε ε (-z) =
        CohnElkies.h_ε ε z
  • complete
    theorem CohnElkies.mellinShellPhase_ofReal (ε t : ℝ) :
      CohnElkies.h_ε ε ↑t = ↑(CohnElkies.realOscillatoryShellPhase ε t)
    theorem CohnElkies.mellinShellPhase_ofReal
      (ε t : ℝ) :
      CohnElkies.h_ε ε ↑t =
        ↑(CohnElkies.realOscillatoryShellPhase
            ε t)
  • complete
    theorem CohnElkies.mellinShellPhase_imaginary (ε u : ℝ) :
      CohnElkies.h_ε ε (Complex.I * ↑u) = ↑(CohnElkies.h_εI ε u)
    theorem CohnElkies.mellinShellPhase_imaginary
      (ε u : ℝ) :
      CohnElkies.h_ε ε (Complex.I * ↑u) =
        ↑(CohnElkies.h_εI ε u)
  • complete
    def CohnElkies.h_εI (ε u : ℝ) : ℝ
    def CohnElkies.h_εI (ε u : ℝ) : ℝ
    `h_ε` restricted to the imaginary axis: `∫ w(a)(cosh(au) - 1) da`. 
  • def CohnElkies.saddleSourceShellDerivative (ε u : ℝ) : ℝ
    def CohnElkies.saddleSourceShellDerivative
      (ε u : ℝ) : ℝ
    The derivative `h_ε'(u)` of the real shell phase, report (79). 
Proof for Lemma 5.2.5
uses 0

\cos is even, \cos(iau) = \cosh(au) and \frac{d}{du}\cosh(au) = a\sinh(au); the compactly supported w allows differentiation under the integral sign.

Definition5.2.6
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 0
Used by 4
Reverse dependency previews
Preview
Lemma 5.2.7
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

For \eta > 0 and \lambda > 0, the positive density describing the unperturbed gamma damping is \mu_{\lambda,\eta}(a) = \dfrac{e^{-\eta a}}{a(1 - e^{-2a/\lambda})} for a > 0 (equation (37)).

Lean code for Definition5.2.6●1 definition
  • def CohnElkies.μ_ℓ (ℓ η a : ℝ) : ℝ
    def CohnElkies.μ_ℓ (ℓ η a : ℝ) : ℝ
    The gamma density `μ_{λ,η}(a) = e^{-ηa} / (a (1 - e^{-2a/λ}))` of report (37). 
Lemma5.2.7
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
Statement uses 2
Statement dependency previews
Preview
Definition 5.2.2
Loading preview
Statement dependency preview content is loaded from the rendered-fragment cache.
Used by 2
Reverse dependency previews
Preview
Lemma 5.3.15
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

For every 0 < \epsilon \le 1/4, \lambda > 0, -1 \le u \le U and a \in [a_0, A], the negative shell of Definition 5.2.2 satisfies, with \mu_{\lambda,\eta} from Definition 5.2.6, \lambda|w_s(a)|\cosh(ua) \le (1 - 2\epsilon)\,\mu_{\lambda,1+u}(a) (indeed b(a)e^{(u-1)a}\cosh(ua)/\cosh a \le 1 - 2\epsilon); equation (54).

Lean code for Lemma5.2.7●2 theorems
  • complete
    theorem CohnElkies.upperFirstBranch_shortMeasure_pointwise {ε ℓ u a : ℝ}
      (hε : 0 < ε) (hεsmall : ε ≤ 1 / 4) (hℓ : 0 < ℓ) (ha : 0 < a)
      (hulower : -1 ≤ u) (huupper : u ≤ 1 + ε / 2)
      (hmargin : 0 ≤ CohnElkies.bε ε a) :
      ℓ * -CohnElkies.w_s ε a * Real.cosh (u * a) ≤
        (1 - 2 * ε) * CohnElkies.μ_ℓ ℓ (1 + u) a
    theorem CohnElkies.upperFirstBranch_shortMeasure_pointwise
      {ε ℓ u a : ℝ} (hε : 0 < ε)
      (hεsmall : ε ≤ 1 / 4) (hℓ : 0 < ℓ)
      (ha : 0 < a) (hulower : -1 ≤ u)
      (huupper : u ≤ 1 + ε / 2)
      (hmargin : 0 ≤ CohnElkies.bε ε a) :
      ℓ * -CohnElkies.w_s ε a *
          Real.cosh (u * a) ≤
        (1 - 2 * ε) *
          CohnElkies.μ_ℓ ℓ (1 + u) a
    Report (49): pointwise, the short shell is dominated by `(1 - 2ε)` times the gamma density. 
  • complete
    theorem CohnElkies.upperFirstBranch_shortRatio_le {ε u a : ℝ} (hε : 0 < ε)
      (ha : 0 ≤ a) (hulower : -1 ≤ u) (huupper : u ≤ 1 + ε / 2)
      (hmargin : 0 ≤ CohnElkies.bε ε a) :
      CohnElkies.bε ε a * Real.exp ((u - 1) * a) *
          (Real.cosh (u * a) / Real.cosh a) ≤
        1 - 2 * ε
    theorem CohnElkies.upperFirstBranch_shortRatio_le
      {ε u a : ℝ} (hε : 0 < ε) (ha : 0 ≤ a)
      (hulower : -1 ≤ u)
      (huupper : u ≤ 1 + ε / 2)
      (hmargin : 0 ≤ CohnElkies.bε ε a) :
      CohnElkies.bε ε a *
            Real.exp ((u - 1) * a) *
          (Real.cosh (u * a) / Real.cosh a) ≤
        1 - 2 * ε
    Report (49): on the first branch `u ≤ 1 + ε/2` the short-shell ratio is at most `1 - 2ε`. 
Proof for Lemma 5.2.7
uses 0

Divide the negative density by \mu_{\lambda,1+u}: \dfrac{\lambda|w_s(a)|\cosh(ua)}{\mu_{\lambda,1+u}(a)} = b(a)\,\Theta_\lambda(a)\,e^{(u-1)a}\,\dfrac{\cosh(ua)}{\cosh a}, \Theta_\lambda(a) = \dfrac{1 - e^{-2a/\lambda}}{2a/\lambda} \in (0,1] (by 1 - e^{-x} \le x). For -1 < u \le 1, both e^{(u-1)a} and \cosh(ua)/\cosh a are at most 1, so the ratio is at most b(a) \le 1 - 2\epsilon. For 1 \le u \le U, the inequality \cosh(ua) \le e^{(u-1)a}\cosh a bounds it by b(a)e^{2(u-1)a} \le b(a)e^{\epsilon a}, and b(a)e^{\epsilon a} \le e^{-2\epsilon(1+a)}e^{\epsilon a} \le e^{-2\epsilon} \le 1 - c\epsilon. Thus the taper retains a damping margin of order \epsilon on every contour -1 < u \le U.

Lemma5.2.8
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
Statement uses 2
Statement dependency previews
Preview
Lemma 5.1.2
Loading preview
Statement dependency preview content is loaded from the rendered-fragment cache.
used by 1✓L∃∀N

At the target saddle, the negative shell of Definition 5.2.2 satisfies \int_{a_0}^Aw_s(a)\,a\sinh(u_0a)\,da \longrightarrow \int_0^\infty w_*(a)a\sinh a\,da = -\tfrac12\log\dfrac{\pi}{2} as \epsilon \downarrow 0 (Lemma 5.1.2).

Lean code for Lemma5.2.8●2 declarations
  • complete
    theorem CohnElkies.tendsto_shortShellRadiusContribution :
      Filter.Tendsto CohnElkies.shortShellRadiusContribution
        (nhdsWithin 0 (Set.Ioi 0))
        (nhds (∫ (a : ℝ) in Set.Ioi 0, CohnElkies.wallisRadiusIntegrand a))
    theorem CohnElkies.tendsto_shortShellRadiusContribution :
      Filter.Tendsto
        CohnElkies.shortShellRadiusContribution
        (nhdsWithin 0 (Set.Ioi 0))
        (nhds
          (∫ (a : ℝ) in Set.Ioi 0,
            CohnElkies.wallisRadiusIntegrand
              a))
  • def CohnElkies.shortShellRadiusContribution (ε : ℝ) : ℝ
    def CohnElkies.shortShellRadiusContribution
      (ε : ℝ) : ℝ
    The short-shell contribution `∫_{a₀}^{A} w_s(a) a sinh ((1 + ε/4) a) da` to `log α_ε`. 
Proof for Lemma 5.2.8

The negative shell agrees with w_* of Lemma 5.1.2 up to its taper on [a_0,A]. Since \tanh a \le \min(a,1), the omitted contributions are \int_0^{a_0}|w_*|a\sinh a\,da = O(a_0) and \int_A^\infty|w_*|a\sinh a\,da = O(e^{-2A}/A), both o(\epsilon). The taper changes the integral by \int_{a_0}^A|1 - b(a)||w_*(a)|a\sinh a\,da = O\bigl(\epsilon\int_0^\infty(1+a)e^{-2a}\tanh(a)\,da/a\bigr), which is O(\epsilon). Moving from u = 1 to u = u_0 costs another O(\epsilon): the mean-value theorem gives |\sinh(u_0a) - \sinh a| \le (u_0-1)a\cosh(u_0a), so \int_{a_0}^A|w_s(a)|a|\sinh(u_0a) - \sinh a|\,da \ll \epsilon\int_0^\infty e^{-2a}\cosh(u_0a)/\cosh a\,da \ll \epsilon. Together with (32) this gives the negative-shell displacement.

Lemma5.2.9
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1used by 1✓L∃∀N

The positive shell of Definition 5.2.3 satisfies 0 \le \int_B^{B+1}w_B(a)\,a\sinh(u_0a)\,da \le (B+1)Qe^{(u_0-1)(B+1)} \longrightarrow 0 as \epsilon \downarrow 0.

Lean code for Lemma5.2.9●3 declarations
  • complete
    theorem CohnElkies.positiveShellRadiusContribution_bounds {ε : ℝ} (hε : 0 < ε) :
      0 ≤ CohnElkies.positiveShellRadiusContribution ε ∧
        CohnElkies.positiveShellRadiusContribution ε ≤
          (CohnElkies.Bε ε + 1) * CohnElkies.Qε ε *
            Real.exp (ε / 4 * (CohnElkies.Bε ε + 1))
    theorem CohnElkies.positiveShellRadiusContribution_bounds
      {ε : ℝ} (hε : 0 < ε) :
      0 ≤
          CohnElkies.positiveShellRadiusContribution
            ε ∧
        CohnElkies.positiveShellRadiusContribution
            ε ≤
          (CohnElkies.Bε ε + 1) *
              CohnElkies.Qε ε *
            Real.exp
              (ε / 4 * (CohnElkies.Bε ε + 1))
    Lemma 4.2 of the report: the positive shell's saddle contribution is nonnegative and
    exponentially small. 
  • complete
    theorem CohnElkies.tendsto_positiveShellRadiusContribution :
      Filter.Tendsto CohnElkies.positiveShellRadiusContribution
        (nhdsWithin 0 (Set.Ioi 0)) (nhds 0)
    theorem CohnElkies.tendsto_positiveShellRadiusContribution :
      Filter.Tendsto
        CohnElkies.positiveShellRadiusContribution
        (nhdsWithin 0 (Set.Ioi 0)) (nhds 0)
  • complete
    def CohnElkies.positiveShellRadiusContribution (ε : ℝ) : ℝ
    def CohnElkies.positiveShellRadiusContribution
      (ε : ℝ) : ℝ
    `∫_B^{B+1} w_B(a) a sinh((1 + ε/4)a) da`: the positive shell's saddle-radius contribution. 
Proof for Lemma 5.2.9
uses 0

At u_0, \sinh(u_0a)/\cosh a \le e^{(u_0-1)a}, hence 0 \le \int_B^{B+1}w_B(a)a\sinh(u_0a)\,da \le (B+1)Qe^{(u_0-1)(B+1)}, and (B+1)Qe^{(u_0-1)(B+1)} = (B+1)e^{-\epsilon B/8 + \epsilon/4} \to 0.

Lemma5.2.10
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
Statement uses 2
Statement dependency previews
Preview
Definition 5.2.2
Loading preview
Statement dependency preview content is loaded from the rendered-fragment cache.
Used by 2
Reverse dependency previews
Preview
Lemma 5.3.19
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

For all sufficiently small \epsilon the shells of Definition 5.2.2 and Definition 5.2.3 are separated: C_0e^{(U-1)A} \le Qe^{(U-1)B}/5000.

Lean code for Lemma5.2.10●1 theorem
  • complete
    theorem CohnElkies.eventually_upper_shell_parameter_margin :
      ∀ᶠ (ε : ℝ) in nhdsWithin 0 (Set.Ioi 0),
        CohnElkies.upperShellShortCoefficient ε *
            Real.exp (ε / 2 * CohnElkies.Aε ε) ≤
          1 / 5000 * CohnElkies.Qε ε * Real.exp (ε / 2 * CohnElkies.Bε ε)
    theorem CohnElkies.eventually_upper_shell_parameter_margin :
      ∀ᶠ (ε : ℝ) in nhdsWithin 0 (Set.Ioi 0),
        CohnElkies.upperShellShortCoefficient
              ε *
            Real.exp
              (ε / 2 * CohnElkies.Aε ε) ≤
          1 / 5000 * CohnElkies.Qε ε *
            Real.exp (ε / 2 * CohnElkies.Bε ε)
    Lemma 4.6: for small `ε` the short shell carries at most `1/5000` of the positive shell. 
Proof for Lemma 5.2.10
uses 0

At u_0, \sinh(u_0a)/\cosh a \le e^{(u_0-1)a}, hence 0 \le \int_B^{B+1}w_B(a)a\sinh(u_0a)\,da \le (B+1)Qe^{(u_0-1)(B+1)}. The amplitude in (34) has exponential slope q_\epsilon strictly between u_0 - 1 and U - 1, so Qe^{(u_0-1)B} = e^{-\epsilon B/8} and Qe^{(U-1)B} = e^{\epsilon B/8}. Since \epsilon B = \epsilon^{-2} while B, C_0 and e^{(U-1)A} grow only polynomially in 1/\epsilon, all three separation quantities are O(e^{-c'/\epsilon^2}).

The shells determine the common envelope; it remains to impose the Fourier symmetries and select the signs. The first gamma pole occurs at t = -i\lambda, i.e. \zeta = -i.

Definition5.2.11
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 2
Reverse dependency previews
Preview
Definition 5.2.13
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

With h_\epsilon from Definition 5.2.4, the envelope is (equation (38)) E_\lambda(t) = \pi^{it/2}\,\Gamma\Bigl(\dfrac{\lambda - it}{2}\Bigr)\,e^{\lambda h_\epsilon(t/\lambda)}.

Lean code for Definition5.2.11●1 definition
  • complete
    def CohnElkies.E (ε ℓ t : ℝ) : ℂ
    def CohnElkies.E (ε ℓ t : ℝ) : ℂ
    The perturbed Gamma envelope `E_λ(t) = π^{it/2} Γ((λ - it)/2) e^{λ h_ε(t/λ)}`, report (38). 
Definition5.2.12
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 3
Reverse dependency previews
Preview
Definition 5.2.13
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

With \beta from Definition 5.2.1, the polynomials are (equation (38)) P_\pm(\zeta) = 1 + \zeta^2 + \beta \pm i\zeta(1+\zeta^2), P_0(\zeta) = -(1+\zeta^2).

Lean code for Definition5.2.12●3 definitions
  • complete
    def CohnElkies.PPlus (ε : ℝ) (z : ℂ) : ℂ
    def CohnElkies.PPlus (ε : ℝ) (z : ℂ) : ℂ
    The polynomial `P₊(ζ) = 1 + ζ² + β + iζ(1 + ζ²)` of the report. 
  • complete
    def CohnElkies.PMinus (ε : ℝ) (z : ℂ) : ℂ
    def CohnElkies.PMinus (ε : ℝ) (z : ℂ) : ℂ
    The polynomial `P₋(ζ) = 1 + ζ² + β - iζ(1 + ζ²)` of the report. 
  • complete
    def CohnElkies.PZero (z : ℂ) : ℂ
    def CohnElkies.PZero (z : ℂ) : ℂ
    The polynomial `P₀(ζ) = -(1 + ζ²)` of the self-Fourier function `f₀` of the report. 
Definition5.2.13
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
Statement uses 2
Statement dependency previews
Preview
Definition 5.2.11
Loading preview
Statement dependency preview content is loaded from the rendered-fragment cache.
used by 1✓L∃∀N

For j \in \{-, 0, +\} the Mellin data are (equation (38)) X_{f_j}(t) = E_\lambda(t)P_j(t/\lambda), with E_\lambda from Definition 5.2.11 and P_j from Definition 5.2.12.

Lean code for Definition5.2.13●4 definitions
  • complete
    def CohnElkies.spectrum (ε ℓ : ℝ) (P : ℂ → ℂ) (t : ℝ) : ℂ
    def CohnElkies.spectrum (ε ℓ : ℝ) (P : ℂ → ℂ)
      (t : ℝ) : ℂ
    The spectrum `X_P(t) = E_λ(t) P(t/λ)` of a polynomial factor `P` (report (38)); `XPlus` and
    `XMinus` are the cases `P = P₊, P₋`. 
  • complete
    def CohnElkies.XPlus (ε ℓ t : ℝ) : ℂ
    def CohnElkies.XPlus (ε ℓ t : ℝ) : ℂ
    `X₊(t) = E_λ(t) P₊(t/λ)`. 
  • complete
    def CohnElkies.XMinus (ε ℓ t : ℝ) : ℂ
    def CohnElkies.XMinus (ε ℓ t : ℝ) : ℂ
    `X₋(t) = E_λ(t) P₋(t/λ)`. 
  • def CohnElkies.XZero (ε ℓ t : ℝ) : ℂ
    def CohnElkies.XZero (ε ℓ t : ℝ) : ℂ
    `X₀(t) = E_λ(t) P₀(t/λ)`, report (38). 
Definition5.2.14
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 2
Reverse dependency previews
Preview
Lemma 5.2.17
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

The radial profiles are the inverse Mellin transforms of Definition 5.2.13 (equation (38)), f_j(r) = \dfrac{r^{-\lambda}}{2\pi}\int_{\mathbb{R}}X_{f_j}(t)r^{it}\,dt (r > 0), extended to r = 0 by the value (42) of Lemma 5.2.19.

Lean code for Definition5.2.14●4 definitions
  • def CohnElkies.mellinProfile (ε ℓ : ℝ) (P : ℂ → ℂ) (c r : ℝ) : ℂ
    def CohnElkies.mellinProfile (ε ℓ : ℝ)
      (P : ℂ → ℂ) (c r : ℝ) : ℂ
    The radial profile `f_P` of a Mellin datum `M_P`: its inverse Mellin transform for `r ≠ 0`,
    with the prescribed real value `c` at the origin. The profiles `f₊`, `f₋`, `f₀` of the report
    are the cases `P = P₊, P₋, P₀` (with `c = f_P(0)` given by `poleResidue_zero`). 
  • def CohnElkies.fPlus (ε ℓ r : ℝ) : ℂ
    def CohnElkies.fPlus (ε ℓ r : ℝ) : ℂ
    The radial profile `f₊` of the report, the inverse Mellin transform of `M₊`; definitionally
    `mellinProfile ε ℓ (PPlus ε) (originValue ε ℓ)`. 
  • def CohnElkies.fMinus (ε ℓ r : ℝ) : ℂ
    def CohnElkies.fMinus (ε ℓ r : ℝ) : ℂ
    The radial profile `f₋` of the report, the inverse Mellin transform of `M₋`; definitionally
    `mellinProfile ε ℓ (PMinus ε) (originValue ε ℓ)`. 
  • def CohnElkies.fZero (ε ℓ r : ℝ) : ℂ
    def CohnElkies.fZero (ε ℓ r : ℝ) : ℂ
    The radial profile `f₀` of the report: the inverse Mellin transform of `M₀` for `r ≠ 0`, with
    `f₀(0) = 0` (report (42)); definitionally `mellinProfile ε ℓ PZero 0`. 
Lemma5.2.15
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 2
Reverse dependency previews
Preview
Lemma 5.2.17
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

The polynomials of Definition 5.2.12 satisfy P_-(-\zeta) = P_+(\zeta), P_0(-\zeta) = P_0(\zeta), \overline{P_j(\zeta)} = P_j(-\bar\zeta), P_\pm(-i) = \beta > 0 and P_0(-i) = 0.

Lean code for Lemma5.2.15●8 theorems
  • complete
    theorem CohnElkies.minusPolynomial_neg (ε : ℝ) (z : ℂ) :
      CohnElkies.PMinus ε (-z) = CohnElkies.PPlus ε z
    theorem CohnElkies.minusPolynomial_neg (ε : ℝ)
      (z : ℂ) :
      CohnElkies.PMinus ε (-z) =
        CohnElkies.PPlus ε z
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.PZero_neg (z : ℂ) :
      CohnElkies.PZero (-z) = CohnElkies.PZero z
    theorem CohnElkies.PZero_neg (z : ℂ) :
      CohnElkies.PZero (-z) =
        CohnElkies.PZero z
    `P₀` is even: the self-Fourier symmetry of report (40). 
  • complete
    theorem CohnElkies.plusPolynomial_conj (ε : ℝ) (z : ℂ) :
      (starRingEnd ℂ) (CohnElkies.PPlus ε z) =
        CohnElkies.PMinus ε ((starRingEnd ℂ) z)
    theorem CohnElkies.plusPolynomial_conj (ε : ℝ)
      (z : ℂ) :
      (starRingEnd ℂ) (CohnElkies.PPlus ε z) =
        CohnElkies.PMinus ε
          ((starRingEnd ℂ) z)
  • complete
    theorem CohnElkies.minusPolynomial_conj (ε : ℝ) (z : ℂ) :
      (starRingEnd ℂ) (CohnElkies.PMinus ε z) =
        CohnElkies.PPlus ε ((starRingEnd ℂ) z)
    theorem CohnElkies.minusPolynomial_conj (ε : ℝ)
      (z : ℂ) :
      (starRingEnd ℂ)
          (CohnElkies.PMinus ε z) =
        CohnElkies.PPlus ε ((starRingEnd ℂ) z)
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.PZero_conj (z : ℂ) :
      (starRingEnd ℂ) (CohnElkies.PZero z) =
        CohnElkies.PZero (-(starRingEnd ℂ) z)
    theorem CohnElkies.PZero_conj (z : ℂ) :
      (starRingEnd ℂ) (CohnElkies.PZero z) =
        CohnElkies.PZero (-(starRingEnd ℂ) z)
  • complete
    theorem CohnElkies.plusPolynomial_neg_I (ε : ℝ) :
      CohnElkies.PPlus ε (-Complex.I) = ↑(CohnElkies.β ε)
    theorem CohnElkies.plusPolynomial_neg_I (ε : ℝ) :
      CohnElkies.PPlus ε (-Complex.I) =
        ↑(CohnElkies.β ε)
    `P₊(-i) = β`, report (42). 
  • complete
    theorem CohnElkies.minusPolynomial_neg_I (ε : ℝ) :
      CohnElkies.PMinus ε (-Complex.I) = ↑(CohnElkies.β ε)
    theorem CohnElkies.minusPolynomial_neg_I (ε : ℝ) :
      CohnElkies.PMinus ε (-Complex.I) =
        ↑(CohnElkies.β ε)
    `P₋(-i) = β`, report (42). 
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.PZero_neg_I : CohnElkies.PZero (-Complex.I) = 0
    theorem CohnElkies.PZero_neg_I :
      CohnElkies.PZero (-Complex.I) = 0
    `P₀(-i) = 0`, report (42): the residue of `M₀` at `z = 0` vanishes. 
Proof for Lemma 5.2.15
uses 0

Direct substitution of -\zeta, \bar\zeta and \zeta = -i.

Lemma5.2.16
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 2
Reverse dependency previews
Preview
Lemma 5.4.1
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

On the imaginary axis the polynomials of Definition 5.2.12 are real (equation (39)): P_+(iu) = \beta + (1-u)^2(1+u), P_-(iu) = \beta + (1-u)(1+u)^2, P_0(iu) = u^2 - 1. Consequently P_+(iu) > 0 for every u > -1, whereas \beta = u_0 - 1 gives P_-(iu_0) = -\beta(3 + 4\beta + \beta^2) < 0 and P_0(iu_0) = \beta(2+\beta) > 0, and these signs persist for all u \ge u_0.

Lean code for Lemma5.2.16●6 theorems
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.plusPolynomial_imaginary (ε u : ℝ) :
      CohnElkies.PPlus ε (Complex.I * ↑u) =
        ↑(CohnElkies.β ε) + (1 - ↑u) ^ 2 * (1 + ↑u)
    theorem CohnElkies.plusPolynomial_imaginary
      (ε u : ℝ) :
      CohnElkies.PPlus ε (Complex.I * ↑u) =
        ↑(CohnElkies.β ε) +
          (1 - ↑u) ^ 2 * (1 + ↑u)
    `P₊(iu) = β + (1 - u)² (1 + u)` is real. 
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.minusPolynomial_imaginary (ε u : ℝ) :
      CohnElkies.PMinus ε (Complex.I * ↑u) =
        ↑(CohnElkies.β ε) + (1 - ↑u) * (1 + ↑u) ^ 2
    theorem CohnElkies.minusPolynomial_imaginary
      (ε u : ℝ) :
      CohnElkies.PMinus ε (Complex.I * ↑u) =
        ↑(CohnElkies.β ε) +
          (1 - ↑u) * (1 + ↑u) ^ 2
    `P₋(iu) = β + (1 - u) (1 + u)²` is real. 
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.PZero_imaginary (u : ℝ) :
      CohnElkies.PZero (Complex.I * ↑u) = ↑u ^ 2 - 1
    theorem CohnElkies.PZero_imaginary (u : ℝ) :
      CohnElkies.PZero (Complex.I * ↑u) =
        ↑u ^ 2 - 1
    `P₀(iu) = u² - 1` is real. 
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.plusPolynomial_imaginary_re_pos {ε u : ℝ} (hε : 0 < ε)
      (hu : -1 < u) : 0 < (CohnElkies.PPlus ε (Complex.I * ↑u)).re
    theorem CohnElkies.plusPolynomial_imaginary_re_pos
      {ε u : ℝ} (hε : 0 < ε) (hu : -1 < u) :
      0 <
        (CohnElkies.PPlus ε
            (Complex.I * ↑u)).re
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.minusPolynomial_imaginary_re_neg {ε u : ℝ} (hε : 0 < ε)
      (hu : 1 + ε / 4 ≤ u) : (CohnElkies.PMinus ε (Complex.I * ↑u)).re < 0
    theorem CohnElkies.minusPolynomial_imaginary_re_neg
      {ε u : ℝ} (hε : 0 < ε)
      (hu : 1 + ε / 4 ≤ u) :
      (CohnElkies.PMinus ε
            (Complex.I * ↑u)).re <
        0
  • theoremdefined in CohnElkies/Parameters.lean
    complete
    theorem CohnElkies.PZero_imaginary_re_pos {ε u : ℝ} (hε : 0 < ε)
      (hu : 1 + ε / 4 ≤ u) : 0 < (CohnElkies.PZero (Complex.I * ↑u)).re
    theorem CohnElkies.PZero_imaginary_re_pos
      {ε u : ℝ} (hε : 0 < ε)
      (hu : 1 + ε / 4 ≤ u) :
      0 <
        (CohnElkies.PZero (Complex.I * ↑u)).re
Proof for Lemma 5.2.16
uses 0

Direct substitution of \zeta = iu.

Lemma5.2.17
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 5
Reverse dependency previews
Preview
Lemma 5.2.18
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

For every sufficiently small \epsilon > 0 and every integer d \ge 1, with \lambda = d/2, the inverse Mellin integrals of Definition 5.2.14, initially defined for r > 0, extend to f_j \in \mathcal{S}_{\mathrm{rad}}(\mathbb{R}^d;\mathbb{R}) for j \in \{-,0,+\}. More generally the pole t = -i(\lambda+2n), n \ge 0, contributes to f_j(r) the term (equation (41)) 2\pi^{\lambda/2+n}\,r^{2n}\,\dfrac{(-1)^n}{n!}\,e^{\lambda h_\epsilon(\zeta_n)}\,P_j(\zeta_n), where \zeta_n = -i(1 + 2n/\lambda) (CohnElkies.poleResidue, CohnElkies.mellinData_nthPole_decomposition).

Lean code for Lemma5.2.17●4 definitions
  • complete
    def CohnElkies.plusSaddleSchwartz {ε : ℝ} (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    def CohnElkies.plusSaddleSchwartz {ε : ℝ}
      (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    Report Lemma 4.3: `f₊` is a Schwartz function on `ℝᵈ`. 
  • complete
    def CohnElkies.minusSaddleSchwartz {ε : ℝ} (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    def CohnElkies.minusSaddleSchwartz {ε : ℝ}
      (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    Report Lemma 4.3: `f₋` is a Schwartz function on `ℝᵈ`. 
  • def CohnElkies.zeroSaddleSchwartz {ε : ℝ} (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    def CohnElkies.zeroSaddleSchwartz {ε : ℝ}
      (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      CohnElkies.TestFunction d
    Report Lemma 4.3: `f₀` is a Schwartz function on `ℝᵈ`. 
  • complete
    def CohnElkies.mellinProfileSchwartz {ε : ℝ} (hε : 0 < ε) {d : ℕ}
      (hd : 0 < d) (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) {P : ℂ → ℂ}
      (hP : CohnElkies.IsSaddlePolynomial ε P) {c : ℝ}
      (hc : ↑c = CohnElkies.poleResidue ε (↑d / 2) P 0) :
      CohnElkies.TestFunction d
    def CohnElkies.mellinProfileSchwartz {ε : ℝ}
      (hε : 0 < ε) {d : ℕ} (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε)
      {P : ℂ → ℂ}
      (hP : CohnElkies.IsSaddlePolynomial ε P)
      {c : ℝ}
      (hc :
        ↑c =
          CohnElkies.poleResidue ε (↑d / 2) P
            0) :
      CohnElkies.TestFunction d
    Report Lemma 4.3: `x ↦ f_P(‖x‖)` is a Schwartz function on `ℝᵈ` for every saddle polynomial
    `P`, when `f_P(0)` is the residue at `z = 0`. 
Proof for Lemma 5.2.17
Proof uses 2
Proof dependency previews
Preview
Lemma 3.1.3
Loading preview
Proof dependency preview content is loaded from the rendered-fragment cache.

Fix d and \epsilon. Compact support of w makes h_\epsilon entire, and on each horizontal line t = s + i\tau it satisfies |h_\epsilon((s+i\tau)/\lambda)| \le 2\int_0^\infty|w(a)|\cosh(a\tau/\lambda)\,da, so the perturbation is bounded on every fixed horizontal strip. Uniformly for \tau in compact pole-free intervals, the polynomial decay of \Gamma along vertical lines (|\operatorname{Im} z|^k|\Gamma(z)| \le \Gamma(\operatorname{Re} z + k), from Lemma 3.1.3 and |\Gamma(z)| \le \Gamma(\operatorname{Re} z)) gives |X_{f_j}(s+i\tau)| \le C_{d,\epsilon,\tau}(1+|s|)^{-2} (the report has the sharper (1+|s|)^{(\lambda+\tau-1)/2+3}e^{-\pi|s|/4}), so the vertical sides of rectangular contour shifts tend to zero. The only poles of the integrand are those of \Gamma((\lambda - it)/2), at t = -i(\lambda + 2n), n = 0,1,2,\ldots. Shifting the contour upward to \operatorname{Im} t = \tau > 0 gives f_j(r) = O_\tau(r^{-\lambda-\tau}) as r \to \infty, for every \tau, also after differentiation in r: rapid decay. Shifting downward past the poles, the residue \operatorname{Res}_{z=-n}\Gamma(z) = (-1)^n/n! (equivalently 2i(-1)^n/n! in the variable t) gives the terms (41), an expansion of f_j in even powers r^{2n} with a remainder of arbitrarily high order; thus f_j extends to a smooth radial function on \mathbb{R}^d, and it is Schwartz. Conjugate symmetry X_{f_j}(-t) = \overline{X_{f_j}(t)} for real t (from Lemma 5.2.15, realness of h_\epsilon on \mathbb{R}, and \Gamma(\bar z) = \overline{\Gamma(z)}) makes the extension real.

Lemma5.2.18
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1used by 1✓L∃∀N

The extensions of Lemma 5.2.17 satisfy (equation (40)) \widehat{f_-} = f_+ and \widehat{f_0} = f_0.

Lean code for Lemma5.2.18●2 theorems
  • complete
    theorem CohnElkies.saddleSource_fourier_minus_eq_plus {ε : ℝ} (hε : 0 < ε)
      {d : ℕ} (hd : 0 < d) (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε)
      (fminus fplus : CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x = CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x = CohnElkies.fPlusFun ε d x) :
      FourierTransform.fourier fminus = fplus
    theorem CohnElkies.saddleSource_fourier_minus_eq_plus
      {ε : ℝ} (hε : 0 < ε) {d : ℕ}
      (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε)
      (fminus fplus :
        CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x =
            CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x =
            CohnElkies.fPlusFun ε d x) :
      FourierTransform.fourier fminus = fplus
    Report (40): `f̂₋ = f₊`. 
  • complete
    theorem CohnElkies.fourier_zeroSaddleSchwartz {ε : ℝ} (hε : 0 < ε) {d : ℕ}
      (hd : 0 < d) (horder : CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      FourierTransform.fourier
          (CohnElkies.zeroSaddleSchwartz hε hd horder) =
        CohnElkies.zeroSaddleSchwartz hε hd horder
    theorem CohnElkies.fourier_zeroSaddleSchwartz
      {ε : ℝ} (hε : 0 < ε) {d : ℕ}
      (hd : 0 < d)
      (horder :
        CohnElkies.a₀ε ε ≤ CohnElkies.Aε ε) :
      FourierTransform.fourier
          (CohnElkies.zeroSaddleSchwartz hε hd
            horder) =
        CohnElkies.zeroSaddleSchwartz hε hd
          horder
    Report Lemma 4.3 and (40) for `P₀`: `𝓕 f₀ = f₀`, the self-Fourier property of `f₀`. 
Proof for Lemma 5.2.18
Proof uses 4
Proof dependency previews
Preview
Lemma 3.3.8
Loading preview
Proof dependency preview content is loaded from the rendered-fragment cache.

Because h_\epsilon is even, Lemma 5.1.1 gives m_\lambda(t)E_\lambda(-t) = E_\lambda(t); with P_-(-\zeta) = P_+(\zeta) and P_0 even (Lemma 5.2.15), m_\lambda(t)X_{f_-}(-t) = X_{f_+}(t) and m_\lambda(t)X_{f_0}(-t) = X_{f_0}(t). By Lemma 3.3.13, X_{\widehat{f_-}} = X_{f_+} and X_{\widehat{f_0}} = X_{f_0}, and injectivity of the Mellin transform on the critical line (Lemma 3.3.8) gives (40).

Lemma5.2.19
Group: Mellin ansatz (18)
Group member previews
Preview
Definition 5.2.1
Loading preview
Group member preview content is loaded from the rendered-fragment cache.
uses 1
Used by 3
Reverse dependency previews
Preview
Theorem 5.3
Loading preview
Reverse dependency preview content is loaded from the rendered-fragment cache.
✓L∃∀N

The extensions of Lemma 5.2.17 satisfy (equation (42)) f_+(0) = f_-(0) = 2\pi^{\lambda/2}e^{\lambda h_\epsilon(i)}\beta > 0 and f_0(0) = 0.

Lean code for Lemma5.2.19●4 declarations
  • complete
    def CohnElkies.originValue (ε ℓ : ℝ) : ℝ
    def CohnElkies.originValue (ε ℓ : ℝ) : ℝ
    The common value `f_±(0) = 2π^{λ/2} e^{λ h_ε(i)} β` at the origin, report (42). 
  • complete
    theorem CohnElkies.saddleSource_zero_pos {ε : ℝ} (hε : 0 < ε) {d : ℕ}
      (fminus fplus : CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x = CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x = CohnElkies.fPlusFun ε d x) :
      0 < (fminus 0).re ∧ 0 < (fplus 0).re
    theorem CohnElkies.saddleSource_zero_pos {ε : ℝ}
      (hε : 0 < ε) {d : ℕ}
      (fminus fplus :
        CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x =
            CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x =
            CohnElkies.fPlusFun ε d x) :
      0 < (fminus 0).re ∧ 0 < (fplus 0).re
    Report (42): the common value `f₊(0) = f₋(0) > 0` at the origin. 
  • complete
    theorem CohnElkies.saddleSource_zero_eq {ε : ℝ} {d : ℕ}
      (fminus fplus : CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x = CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x = CohnElkies.fPlusFun ε d x) :
      fminus 0 = fplus 0
    theorem CohnElkies.saddleSource_zero_eq {ε : ℝ}
      {d : ℕ}
      (fminus fplus :
        CohnElkies.TestFunction d)
      (hminus :
        ∀ (x : CohnElkies.Euclidean d),
          fminus x =
            CohnElkies.fMinusFun ε d x)
      (hplus :
        ∀ (x : CohnElkies.Euclidean d),
          fplus x =
            CohnElkies.fPlusFun ε d x) :
      fminus 0 = fplus 0
  • complete
    theorem CohnElkies.fZero_zero (ε ℓ : ℝ) : CohnElkies.fZero ε ℓ 0 = 0
    theorem CohnElkies.fZero_zero (ε ℓ : ℝ) :
      CohnElkies.fZero ε ℓ 0 = 0
Proof for Lemma 5.2.19
uses 0

The term n = 0 of (41) is the value at r = 0; evenness gives h_\epsilon(-i) = h_\epsilon(i), and P_\pm(-i) = \beta, P_0(-i) = 0 give (42).