OctoInformBeta

OpenAI preprint, September 2026 · §7. From the action inequality to spectral moments · §7

Lemma 7.1: Shifted-resolvent realization for spectral exponent γ>1/2\gamma > 1/2 and almost everywhere nonnegative W∈Lγ+1/2(ℝ)W \in L^{\gamma+1/2}(\mathbb{R}), with discrete negative spectrum when γ<3/2\gamma < 3/2

Self-adjoint realization and negative spectrum of Schrödinger operators.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Consider the formal Schrödinger operator −d2dx2−W-\frac{d^2}{dx^2} - W on ℝ\mathbb{R}, where γ>1/2\gamma > 1/2 is a spectral exponent and W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} is an almost everywhere nonnegative potential in Lγ+1/2(ℝ)L^{\gamma+1/2}(\mathbb{R}). At issue is how to realize −d2dx2−W-\frac{d^2}{dx^2} - W as a self-adjoint operator on L2(ℝ)L^2(\mathbb{R}) via its quadratic form, and whether for γ∈(1/2,3/2)\gamma \in (1/2, 3/2) its negative spectrum consists of isolated eigenvalues of finite multiplicity with only finitely many below −a-a for every a>0a > 0.

Sobolev Graph and Quadratic Form

The complex inner product on L2(ℝ)L^2(\mathbb{R}) is the inner product ⟨f,g⟩L2=∫ℝf(x)¯g(x)dx\langle f, g \rangle_{L^2} = \int_{\mathbb{R}} \overline{f(x)}\,g(x)\,dx, conjugate-linear in its first argument. The product space L2(ℝ)×L2(ℝ)L^2(\mathbb{R}) \times L^2(\mathbb{R}) is the Cartesian product of L2(ℝ)L^2(\mathbb{R}) with itself under the ℓ2\ell^2-product norm ∥(f,g)∥L2×L2=(∥f∥L22+∥g∥L22)1/2\|(f, g)\|_{L^2 \times L^2} = (\|f\|_{L^2}^2 + \|g\|_{L^2}^2)^{1/2}.✦ A smooth test function φ∈Cc∞(ℝ;ℂ)\varphi \in C_c^\infty(\mathbb{R};\mathbb{C}) is an infinitely differentiable complex-valued function on ℝ\mathbb{R} with compact support.✦ For each φ∈Cc∞(ℝ;ℂ)\varphi \in C_c^\infty(\mathbb{R};\mathbb{C}), the weak-derivative constraint ℓφ:L2(ℝ)×L2(ℝ)→ℂ\ell_\varphi : L^2(\mathbb{R}) \times L^2(\mathbb{R}) \to \mathbb{C} is the continuous linear functional defined by (1): ℓφ(f,g)=⟨φ′¯,f⟩L2+⟨φ¯,g⟩L2.\ell_\varphi(f, g) = \langle \overline{\varphi'}, f \rangle_{L^2} + \langle \overline{\varphi}, g \rangle_{L^2}.(1)✦ The Sobolev graph ΓH1\Gamma_{H^1} is the complex subspace of L2(ℝ)×L2(ℝ)L^2(\mathbb{R}) \times L^2(\mathbb{R}) equal to ⋂φ∈Cc∞(ℝ;ℂ)ker⁡ℓφ\bigcap_{\varphi \in C_c^\infty(\mathbb{R};\mathbb{C})} \ker \ell_\varphi.✦

The Sobolev graph Hilbert space ℋ1\mathcal{H}^1 is the subspace ΓH1\Gamma_{H^1} equipped with the inner product inherited from L2(ℝ)×L2(ℝ)L^2(\mathbb{R}) \times L^2(\mathbb{R}).✦ The inclusion map ι:ℋ1→L2(ℝ)\iota : \mathcal{H}^1 \to L^2(\mathbb{R}) is the bounded linear operator restricting the first coordinate projection from L2(ℝ)×L2(ℝ)L^2(\mathbb{R}) \times L^2(\mathbb{R}) onto L2(ℝ)L^2(\mathbb{R}) to ℋ1\mathcal{H}^1✦, and the gradient map D:ℋ1→L2(ℝ)D : \mathcal{H}^1 \to L^2(\mathbb{R}) is the bounded linear operator restricting the second coordinate projection to ℋ1\mathcal{H}^1.✦ The Sobolev space H1(ℝ)H^1(\mathbb{R}) consists of pairs (v,v′)(v, v') of square-integrable complex-valued functions on ℝ\mathbb{R} satisfying (2): ∫ℝv(x)φ′(x)dx=−∫ℝv′(x)φ(x)dxfor all φ∈Cc∞(ℝ;ℂ).\int_{\mathbb{R}} v(x)\,\varphi'(x)\,dx = -\int_{\mathbb{R}} v'(x)\,\varphi(x)\,dx \quad \text{for all } \varphi \in C_c^\infty(\mathbb{R};\mathbb{C}).(2)✦ The Sobolev graph identification Φ:ℋ1→H1(ℝ)\Phi : \mathcal{H}^1 \to H^1(\mathbb{R}) is the linear map sending each u∈ℋ1u \in \mathcal{H}^1 to the pair (ιu,Du)(\iota u, Du).✦

For a spectral exponent γ∈ℝ\gamma \in \mathbb{R}, an admissible potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} at exponent γ\gamma is a real-valued function on ℝ\mathbb{R} that is nonnegative for almost every x∈ℝx \in \mathbb{R} and belongs to Lmax⁡(γ+1/2,0)(ℝ)L^{\max(\gamma+1/2, 0)}(\mathbb{R}).✦ For a real-valued potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R}, the Schrödinger quadratic form h−Wh_{-W} on H1(ℝ)H^1(\mathbb{R}) is the quadratic form defined for all (u,u′),(v,v′)∈H1(ℝ)(u, u'), (v, v') \in H^1(\mathbb{R}) by (3): h−W((u,u′),(v,v′))=∫ℝu′(x)¯v′(x)dx−∫ℝW(x)u(x)¯v(x)dx.h_{-W}\bigl((u, u'), (v, v')\bigr) = \int_{\mathbb{R}} \overline{u'(x)}\,v'(x)\,dx - \int_{\mathbb{R}} W(x)\,\overline{u(x)}\,v(x)\,dx.(3)✦ An unbounded linear operator AA on L2(ℝ)L^2(\mathbb{R}) with domain 𝒟(A)⊆L2(ℝ)\mathcal{D}(A) \subseteq L^2(\mathbb{R}) is associated with the form h−Wh_{-W} if, for all f,g∈L2(ℝ)f, g \in L^2(\mathbb{R}), one has f∈𝒟(A)f \in \mathcal{D}(A) and Af=gAf = g if and only if there exists (u,u′)∈H1(ℝ)(u, u') \in H^1(\mathbb{R}) with u=fu = f in L2(ℝ)L^2(\mathbb{R}) such that h−W((v,v′),(u,u′))=⟨v,g⟩L2h_{-W}\bigl((v, v'), (u, u')\bigr) = \langle v, g \rangle_{L^2} for every (v,v′)∈H1(ℝ)(v, v') \in H^1(\mathbb{R}).✦

Operator Realization and Spectral Notions

Potential decomposition data d=(b,V)d = (b, V) for a measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} consists of a real constant b≥0b \ge 0 and a pointwise nonnegative integrable function V:ℝ→ℝV : \mathbb{R} \to \mathbb{R} with ∫ℝV(x)dx≤1/32\int_{\mathbb{R}} V(x)\,dx \le 1/32 such that |W(x)|≤b+V(x)|W(x)| \le b + V(x) for all x∈ℝx \in \mathbb{R}.✦ Every spectral exponent γ>1/2\gamma > 1/2 and admissible potential WW at exponent γ\gamma determine potential decomposition data d(γ,W)d(\gamma, W) for WW.✦

Given a measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} with potential decomposition data d=(b,V)d = (b, V), the potential-weighted value function of u∈ℋ1u \in \mathcal{H}^1 is wd,u(x)=|W(x)|(ιu)(x)w_{d,u}(x) = \sqrt{|W(x)|}\,(\iota u)(x)✦, and the potential-weighted map Md:ℋ1→L2(ℝ)M_d : \mathcal{H}^1 \to L^2(\mathbb{R}) is the bounded linear operator sending each u∈ℋ1u \in \mathcal{H}^1 to wd,uw_{d,u} in L2(ℝ)L^2(\mathbb{R}).✦ For potential decomposition data d=(b,V)d = (b, V) of a measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R}, the shifted form operator Ad:ℋ1→ℋ1A_d : \mathcal{H}^1 \to \mathcal{H}^1 is the bounded linear operator given by (4): Ad=D*D−Md*Md+(b+1)ι*ι.A_d = D^* D - M_d^* M_d + (b + 1)\,\iota^* \iota.(4)✦ The shifted form operator AdA_d is a continuous linear isomorphism of ℋ1\mathcal{H}^1 onto itself, with bounded inverse Ad−1:ℋ1→ℋ1A_d^{-1} : \mathcal{H}^1 \to \mathcal{H}^1.✦

For a measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} with potential decomposition data d=(b,V)d = (b, V), the variational solution operator Sd:L2(ℝ)→ℋ1S_d : L^2(\mathbb{R}) \to \mathcal{H}^1 is the bounded linear map Sd=Ad−1ι*S_d = A_d^{-1} \iota^*✦, and the shifted Schrödinger resolvent Rd:L2(ℝ)→L2(ℝ)R_d : L^2(\mathbb{R}) \to L^2(\mathbb{R}) is the bounded linear operator Rd=ι∘SdR_d = \iota \circ S_d.✦ Given an injective bounded linear operator R:E→ER : E \to E on a complex inner product space EE and a real scalar cc, the shifted inverse operator R−1−cIR^{-1} - cI (where II is the identity on EE) is the unbounded linear operator on EE with domain equal to the range ran⁡(R)\operatorname{ran}(R) of RR sending each Rf∈ran⁡(R)Rf \in \operatorname{ran}(R) to f−cRff - cRf.✦ For potential decomposition data d=(b,V)d = (b, V) of a measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R}, the Schrödinger operator HdH_d is the shifted inverse operator of RdR_d with shift b+1b + 1.✦ For a spectral exponent γ>1/2\gamma > 1/2 and an admissible potential WW at exponent γ\gamma, the Schrödinger operator Hγ,WH_{\gamma, W} is Hd(γ,W)H_{d(\gamma, W)}.✦

An unbounded linear operator AA on L2(ℝ)L^2(\mathbb{R}) with domain 𝒟(A)⊆L2(ℝ)\mathcal{D}(A) \subseteq L^2(\mathbb{R}) has a bounded resolvent at e∈ℝe \in \mathbb{R} if there exists a bounded linear operator R:L2(ℝ)→L2(ℝ)R : L^2(\mathbb{R}) \to L^2(\mathbb{R}) such that for all f,g∈L2(ℝ)f, g \in L^2(\mathbb{R}), one has f∈𝒟(A)f \in \mathcal{D}(A) and Af=ef+gAf = ef + g if and only if Rg=fRg = f.✦ For an unbounded linear operator AA on L2(ℝ)L^2(\mathbb{R}) with domain 𝒟(A)⊆L2(ℝ)\mathcal{D}(A) \subseteq L^2(\mathbb{R}), an element f∈L2(ℝ)f \in L^2(\mathbb{R}) is an operator eigenfunction of AA with eigenvalue e∈ℝe \in \mathbb{R} if f∈𝒟(A)f \in \mathcal{D}(A) and Af=efAf = ef; when such a nonzero ff exists, ee is an eigenvalue of AA.✦ Given an eigenvalue e∈ℝe \in \mathbb{R} of an unbounded linear operator AA on L2(ℝ)L^2(\mathbb{R}) with domain 𝒟(A)⊆L2(ℝ)\mathcal{D}(A) \subseteq L^2(\mathbb{R}), the operator eigenspace ℰ(A,e)\mathcal{E}(A, e) is the complex subspace {f∈𝒟(A):Af=ef}\{f \in \mathcal{D}(A) : Af = ef\} of L2(ℝ)L^2(\mathbb{R}).✦ A negative energy level ee is a real number satisfying e<0e < 0.✦

The Claim

Theorem 1. The Sobolev graph satisfies (5): ΓH1 is closed in L2(ℝ)×L2(ℝ).\Gamma_{H^1} \text{ is closed in } L^2(\mathbb{R}) \times L^2(\mathbb{R}).(5)✦ For every measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} with potential decomposition data dd, the operator HdH_d satisfies (6), and for all f,g∈L2(ℝ)f, g \in L^2(\mathbb{R}) its graph satisfies (7): Hd is self-adjoint on L2(ℝ),H_d \text{ is self-adjoint on } L^2(\mathbb{R}),(6)✦ (f∈𝒟(Hd) and Hdf=g)⇔∃u∈ℋ1 such that ιu=f andh−W(Φv,Φu)=⟨ιv,g⟩L2for all v∈ℋ1.\begin{aligned} \bigl(f \in \mathcal{D}(H_d) \text{ and } H_d f = g\bigr) &\iff \exists\, u \in \mathcal{H}^1 \text{ such that } \iota u = f \text{ and}\\ &\qquad h_{-W}(\Phi v, \Phi u) = \langle \iota v, g \rangle_{L^2} \quad \text{for all } v \in \mathcal{H}^1. \end{aligned}(7)✦ For every spectral exponent γ>1/2\gamma > 1/2 and every admissible potential WW at exponent γ\gamma, the Schrödinger operator Hγ,WH_{\gamma, W} satisfies (8) and (9): Hγ,W is self-adjoint on L2(ℝ),H_{\gamma, W} \text{ is self-adjoint on } L^2(\mathbb{R}),(8)✦ Hγ,W is associated with the form h−W.H_{\gamma, W} \text{ is associated with the form } h_{-W}.(9)✦ When 1/2<γ<3/21/2 < \gamma < 3/2 and W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} is an admissible potential at exponent γ\gamma, the operator Hγ,WH_{\gamma, W} satisfies (10): for every e<0,Hγ,W does not have a bounded resolvent at e⇔e is an eigenvalue of Hγ,W,for every negative energy level e<0,ℰ(Hγ,W,e) is finite-dimensional over ℂ,for every a>0,{e≤−a:e is an eigenvalue of Hγ,W} is finite,for every e<0,there exists ε>0 such that t=e whenever t is an eigenvalueof Hγ,W with |t−e|<ε.\begin{aligned} &\text{for every } e < 0,\; H_{\gamma, W} \text{ does not have a bounded resolvent at } e\\ &\qquad\qquad\iff e \text{ is an eigenvalue of } H_{\gamma, W},\\ &\text{for every negative energy level } e < 0,\; \mathcal{E}(H_{\gamma, W}, e) \text{ is finite-dimensional over } \mathbb{C},\\ &\text{for every } a > 0,\; \{e \le -a : e \text{ is an eigenvalue of } H_{\gamma, W}\} \text{ is finite},\\ &\text{for every } e < 0,\; \text{there exists } \varepsilon > 0 \text{ such that } t = e \text{ whenever } t \text{ is an eigenvalue}\\ &\qquad\qquad\text{of } H_{\gamma, W} \text{ with } |t - e| < \varepsilon. \end{aligned}(10)✦

In the paper, the result reads: for p>1p > 1 (here p=γ+1/2p = \gamma + 1/2) and 0≤W∈Lp(ℝ)0 \le W \in L^p(\mathbb{R}), the form h−Wh_{-W} is closed and bounded below on H1(ℝ)H^1(\mathbb{R}) and every spectral projection of its associated self-adjoint operator H−WH_{-W} (here Hγ,WH_{\gamma, W}) onto (−∞,−α](-\infty, -\alpha] with α>0\alpha > 0 has finite rank, so that its negative spectrum consists of eigenvalues of finite multiplicity with possible accumulation only at zero, while on a bounded interval II the corresponding Dirichlet form on H01(I)H^1_0(I) is closed and bounded below with compact resolvent.

What Is Checked

The claims are stated over the Sobolev space H1(ℝ)H^1(\mathbb{R}) and its graph ΓH1⊆L2(ℝ)×L2(ℝ)\Gamma_{H^1} \subseteq L^2(\mathbb{R}) \times L^2(\mathbb{R}), together with the quadratic form h−Wh_{-W} and the shifted-resolvent operators HdH_d and Hγ,WH_{\gamma, W} (and their bounded-resolvent and spectral notions at real and negative energies) for measurable almost everywhere nonnegative potentials W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} with decomposition data d=(b,V)d = (b, V) and for potentials WW admissible at exponent γ\gamma with γ>1/2\gamma > 1/2 and γ∈(1/2,3/2)\gamma \in (1/2, 3/2). Given any measurable, almost everywhere nonnegative potential W:ℝ→ℝW : \mathbb{R} \to \mathbb{R} with potential decomposition data d=(b,V)d = (b, V), the potential-weighted map satisfies ∥Mdu∥L2≤b+1∥u∥ℋ1\|M_d u\|_{L^2} \le \sqrt{b + 1}\,\|u\|_{\mathcal{H}^1} for every u∈ℋ1u \in \mathcal{H}^1✦, and the shifted Schrödinger resolvent RdR_d is injective.✦