betaFeedback

WorksMathematics

Quantitative unique continuation for semiclassical Witten-Schrödinger type operator with low regular coefficients

Abstract

In this note we study eigenfunctions localization of the semiclassical selfadjoint Schrödinger operator with low regular coefficients on domains of R2\mathbb{R}^2. We show a unique continuation result on the local decay of the eigenfunction: local supremum on open sets can't vanish faster than global supremum on these same sets times an exponential term that decrease in the semiclassical limit h→0h\to 0 as exp⁡(−clog⁡(1/h)/h)\exp(-c\log(1/h)/h) in the worst case. We deduce from this result a control on the eigenfunctions of the semiclassical Schrödinger operator on a W2,∞W^{2,\infty} compact Riemann surface with Lipschitz positive metric and with Robin boundary condition. This last part leads us to study the Schrödinger operator with a Witten Laplacian.

Introduction

In this article, we study the localization properties of eigenfunctions uhu_h for semiclassical Witten-Schrödinger type operator PhP_h on R2\mathbb{R}^2. Specifically, we consider the eigenvalue equation:

Phuh:=(−h2Khdiv⁡(Kh∇)+Vh)uh=Ehuhon B(0,R)⊂R2,(1)P_hu_h := \left(-\frac{h^2}{K_h}\operatorname{div}(K_h\nabla)+V_h\right)u_h=E_hu_h \quad\text{on } B(0,R)\subset\mathbb{R}^2, \tag*{(1)}

where

  • Kh:=exp⁡(ϕh/h)K_h := \exp(\phi_h/h) with ϕh=ϕ+OW1,∞(h)∈W1,∞(B(0,R),R)\phi_h = \phi+ O_{W^{1,\infty}}(h) \in W^{1,\infty}(B(0,R),\mathbb{R}),

  • Vh:=V+OL∞(h)∈L∞(B(0,R),R)V_h := V + O_{L^\infty}(h) \in L^\infty(B(0,R),\mathbb{R}),

  • and Eh:=E+OR(h)∈RE_h := E + O_{\mathbb{R}}(h) \in\mathbb{R},

with hh the semiclassical parameter. If the function ϕh\phi_h is sufficiently regular, we can rewrite the operator PhP_h in the form −h2Δϕh/2+V~-h^2\Delta_{\phi_h/2}+\widetilde V, where Δϕh/2\Delta_{\phi_h/2} is the Witten Laplacian e−ϕh/2hΔeϕh/2he^{-\phi_h/2h}\Delta e^{\phi_h/2h}. This deformation of the usual Laplacian was introduce by Witten [21] to study the Morse theory and Supersymmetry. It also appeared in some study of quantum tunneling and metastable states as for example in the Kramers-Smoluchowski equation (see for instance [18]).

Our goal here is to understand how the weight KhK_h and the potential VhV_h affect the spatial concentration of uhu_h in the semiclassical limit h→0h \to0.

Our first result in this article is the following semiclassical adaptation of Theorem 2.2 from Logunov et al. [16] generalised to pondered Schrödinger operator:

Theorem 1. For hh small enough, let uh∈H1(B(0,R),R)u_h \in H^1(B(0,R),\mathbb{R}) be a solution on B(0,R)B(0,R) of the equation

Phuh:=(−h2Khdiv⁡(Kh∇)+Vh)uh=EhuhP_hu_h := \left(-\frac{h^2}{K_h}\operatorname{div}(K_h\nabla)+V_h\right)u_h = E_hu_h

in the sense of distributions: for all χ∈C0∞(B(0,R),R)\chi\in C^\infty_0(B(0,R),\mathbb{R}),

∫(h2∇uh⋅∇χ+Vhuhχ)Khdx=∫EhuhχKhdx.\int\left(h^2\nabla u_h \cdot\nabla\chi+ V_hu_h\chi\right)K_h\mathrm{d}x = \int E_hu_h\chi K_h\mathrm{d}x.

Let R0<RR_0 < R and suppose that

sup⁡B(0,R0/2)∣uh∣≥e−Nhsup⁡B(0,R0)∣uh∣,\sup_{B(0,R_0/2)} |u_h| \ge e^{-N_h}\sup_{B(0,R_0)} |u_h|,

then for 0<r≤R0/40 < r \le R_0/4:

sup⁡B(0,r)∣uh∣≥(rR0)L(R0hϵ(h)+Nh+R0hosc⁡(ϕh))sup⁡B(0,R0)∣uh∣.(2)\sup_{B(0,r)} |u_h| \ge\left(\frac{r}{R_0}\right)^{L\left(\frac{R_0}{h}\epsilon(h)+N_h+\frac{R_0}{h}\operatorname{osc}(\phi_h)\right)}\sup_{B(0,R_0)} |u_h|. \tag*{(2)}

with LL an absolute constant, osc⁡(ϕh)=max⁡ϕh−min⁡ϕh\operatorname{osc}(\phi_h)=\max\phi_h-\min\phi_h and

ϵ(h):={Θlog⁡(R0Θh)if ∥∇ϕ∥∞>0,Θlog⁡(R0Θh)1/2else.\epsilon(h):= \begin{cases} \Theta\log\left(\frac{R_0\Theta}{h}\right) & \text{if }\|\nabla\phi\|_\infty>0,\\ \Theta\log\left(\frac{R_0\Theta}{h}\right)^{1/2} & \text{else.} \end{cases}

where Θ=max⁡{∥∇ϕ∥∞,∥V−E∥∞,1}\Theta=\max\{\|\nabla\phi\|_\infty,\sqrt{\|V-E\|_\infty},1\}.

If we forget the exponentials weights, we recover the usual semiclassical Schrödinger operator on R2\mathbb{R}^2. Thus this result gives constraints on speed limit of the Schrödinger eigenfunctions’ localization in a reasonable neighbourhood of the maximum. This works as a counterpart of the usual Agmon estimate that gives at least exponential decrease inside the classical forbidden region {V>E}\{V > E\} (see for instance [12], Chapter 7). This type of unique continuation is generally achieved thanks to Carleman inequality for smooth elliptic operator. In this case, Carleman inequality shows that eigenfunctions can’t localize faster than exponentially. With the low regularity of our parameters, Carleman inequality can’t be used any more. In fact, it still works in dimension 1 as shown by Dyatlov and Zworski [8] (see Theorem 2.32). In the same dimension Laurent and Léautaud [15] even show a unique continuation property for some class of potential in term of Agmon distance (see Theorem 1.4). In higher dimension, Klopp and Vogel [13] constructed a Carleman inequality for bounded complex potential (lemma 4). Their result imply a maximal rate of localization of e−Ch−4/3e^{-Ch^{-4/3}}, which is faster than the usual Carleman one. If we add some regularity to the potential, we can find better estimates as done by Vodev [20] who obtain a Carleman estimate for α\alpha-Hölder regular potentials in term of α\alpha (e−Ch−4/(α+3)e^{-Ch^{-4/(\alpha+3)}}, Theorem 3.1). We can also assume others properties for the potential as for instance Datchev [6] in dimension ≥3\ge3 who ask a radial decrease control on the potential and its radial derivative (Lemma 2.2) or Shapiro [19] in dimension 2 who ask a decrease control on the gradient of the potential (Lemma 3.2).

The proof of the Theorem 1 use the method developed by Logunov, Malinnikova, Nadirashvili and Nazarov [16] to study the Landis conjecture in dimension 2. As a reminder, the Landis conjecture (see Kondrat’ev and Landis [14]) deals with how quickly solutions u≠0u \ne0 to

(−Δ+V)u=0(3)(-\Delta+ V)u = 0 \tag*{(3)}

can decay at infinity when VV is bounded. Landis and Kondrat’ev conjectured that non-zero solutions cannot decay faster than exponentially. This was disproven by Meshkov [17] for complex potentials: he found a bounded potential and a solution that decay as e−C∣x∣4/3e^{-C|x|^{4/3}} in dimension 2. He also proved that is was impossible for a non zero eigenfunction to decrease faster than e−∣x∣ε+4/3e^{-|x|^{\varepsilon+4/3}} in any dimension higher than 1. Later, Bourgain and Kenig [3] obtained a quantitative estimate of the Meshkov result (Lemma 3.10). They showed that Schrödinger eigenfunctions on Rd\mathbb{R}^{d} could not decrease faster than e−C(∣x∣4/3log⁡∣x∣)−1e^{-C(|x|^{4/3}\log|x|)^{-1}}. Naturally, the question of whether the conjecture was true in the case of a real potential was raised. Some results in this regard have been obtained. First in dimension 2, Davey, Kenig and Wang [7] proved a weaker version of the Landis conjecture for second order elliptic operator with variable coefficients and with potential VV assumed to be non-negative. They obtained a decrease speed limit of e−C∣x∣log⁡(∣x∣)e^{-C|x|\log(|x|)}. Their result was improve by Logunov, Malinnikova, Nadirashvili and Nazarov [16] in case of the usual Schrödinger operator (not assumed to be non-negative), where they obtained a maximal decay of e−CRlog⁡(R)1/2e^{-CR\log(R)^{1/2}}. Their proof involves removing holes in the domain to increase the control of how functions oscillate by reducing the Poincaré constant. Then they use quasi-conformal mappings to transform the problem into one about harmonic functions on a domain with holes. To go deeper, Fernández-Bertolin, Stan and Roncal [9] did a complete survey of Landis’ conjecture. Unfortunately, the conjecture proved to be false in dimension ≥3\ge3. In recent work of Frank and Ivanisvili [10], a counterexemple decaying at speed e−C∣x∣4/3e^{-C|x|^{4/3}} was constructed thank to AI, thus recovering the result of Meskov. However, it seems that the conjecture is true if we consider non-negative operator as shown by Das and Pinchover [5]. Their result even show that the maximum rate of decline is slightly slower than Landis had predicted.

In a second time, we use the result of Theorem 1 to get information on the localization of Schrödinger’s eigenfunctions in a compact Riemannian setting. More precisely, let (M,g)(M,g) be a W2,∞W^{2,\infty} Riemann surface (W2,∞W^{2,\infty} means that the changing of charts are W2,∞∩C1W^{2,\infty}\cap C^{1}) equipped with a Lipschitz positive metric. We deduce from Theorem 1 the following result on closed (compact and without boundary) surface:

Theorem 2. Assume that MM is a closed Riemannian surface. For hh small enough, let be uh∈H1(M,R)u_h \in H^{1}(M,\mathbb{R}) weak solution of

−h2Khdiv⁡g(Kh∇g)uh+Vhuh=Ehuh on M,-\frac{h^{2}}{K_h}\operatorname{div}_{g}(K_h\nabla_{g})u_h + V_hu_h = E_hu_h\ \text{on }M,

where Kh=eϕh/hK_h=e^{\phi_h/h} with ϕh=ϕ+OW1,∞(h)∈W1,∞(M,R)\phi_h=\phi+O_{W^{1,\infty}}(h)\in W^{1,\infty}(M,\mathbb{R}), Vh=V+OL∞(h)∈L∞(M,R)V_h=V+O_{L^\infty}(h)\in L^\infty(M,\mathbb{R}) and Eh=E+OR(h)∈RE_h=E+O_{\mathbb{R}}(h)\in\mathbb{R}.

Then for any ball BrB_r of radius r<1/4r < 1/4:

References

  1. [1]Ahlfors, L.V., 2006. Lectures on quasiconformal mappings. volume 38. American Mathematical Soc.DOI
  2. [2]Astala, K., Iwaniec, T., Martin, G., 2008. Elliptic partial differential equations and quasiconformal mappings in the plane (PMS-48). Princeton Mathematical Series, Princeton University Press, Princeton, NJ.
  3. [3]Bourgain, J., Kenig, C.E., 2005. On localization in the continuous anderson-bernoulli model in higher dimension. Inventiones mathematicae 161, 389–426. URL: http://dx.doi.org/10.1007/s00222-004-0435-7, doi:10.1007/s00222-004-0435-7.
  4. [4]Burq, N., Moyano, I., 2022. Propagation of smallness and control for heat equations. Journal of the European Mathematical Society 25, 1349–1377. URL: http://dx.doi.org/10.4171/JEMS/1213, doi:10.4171/jems/1213.
  5. [5]Das, U., Pinchover, Y., 2024. The landis conjecture via liouville comparison principle and criticality theory. URL: https://arxiv.org/abs/2405.11695, arXiv:2405.11695.
  6. [6]Datchev, K., 2014. Quantitative limiting absorption principle in the semiclassical limit. URL: http://dx.doi.org/10.1007/s00039-014-0273-8, doi:10.1007/s00039-014-0273-8.
  7. [7]Davey, B., Kenig, C., Wang, J.N., 2017. The landis conjecture for variable coefficient second-order elliptic pdes. Transactions of the American Mathematical Society 369, 8209–8237. URL: http://dx.doi.org/10.1090/tran/7073, doi:10.1090/tran/7073.
  8. [8]Dyatlov, S., Zworski, M., 2019. Mathematical Theory of Scattering Resonances. volume 200 of Graduate Studies in Mathematics. American Mathematical Society, Providence, Rhode Island. doi:10.1090/gsm/200.DOI
  9. [9]Fernández-Bertolin, A., Stan, D., Boncal, L., 2024. Landis’ conjecture: A survey. doi:10.48550/arXiv.2412.00788, arXiv:2412.00788.
  10. [10]Frank, R.L., Ivanisvili, P., 2026. Counterexamples to the landis conjecture in dimensions three and higher. URL: https://arxiv.org/abs/2608.00802, arXiv:2608.00802.
  11. [11]Gilbarg, D., Trudinger, N.S., 2001. Elliptic Partial Differential Equations of Second Order. Springer Berlin Heidelberg. URL: http://dx.doi.org/10.1007/978-3-642-61798-0, doi:10.1007/978-3-642-61798-0.
  12. [12]Helfer, B., 1988. Semi-Classical Analysis of the Schrödinger Operator and Applications. Springer Berlin Heidelberg. URL: http://dx.doi.org/10.1007/BFb0078115, doi:10.1007/bfb0078115.
  13. [13]Klopp, F., Vogel, M., 2019. Semiclassical resolvent estimates for bounded potentials. Pure and Applied Analysis 1, 1–25. doi:10.2140/paa.2019.1.1.DOI
  14. [14]Kondrat’ev, V.A., Landis, E.M., 1991. Qualitative Theory of Second Order Linear Partial Differential Equations. Springer-Verlag, Berlin. volume 32 of Partial Differential Equations III. pp. 87–192.
  15. [15]Laurent, C., Léautaud, M., 2023. Uniform observation of semiclassical Schrödinger eigenfunctions on an interval. Tunisian Journal of Mathematics 5, 125–170. doi:10.2140/tunis.2023.5.125.DOI
  16. [16]Logunov, A., Malinnikova, E., Nadirashvili, N., Nazarov, F., 2025. The landis conjecture on exponential decay. Inventiones mathematicae 241, 465–508. URL: http://dx.doi.org/10.1007/s00222-025-01340-1, doi:10.1007/s00222-025-01340-1.
  17. [17]Meshkov, V.Z., 1992. On the possible rate of decay at infinity of solutions of second order partial differential equations. Mathematics of the USSR-Sbornik 72, 343–361. URL: http://dx.doi.org/10.1070/SM1992v072n02ABEH001414, doi:10.1070/sm1992v072n02abeh001414.
  18. [18]Michel, L., Zworski, M., 2018. A semiclassical approach to the kramers–smoluchowski equation. SIAM Journal on Mathematical Analysis 50, 5362–5379. URL: https://doi.org/10.1137/17M1124826, doi:10.1137/17M1124826.
  19. [19]Shapiro, J., 2019. Semiclassical resolvent bounds in dimension two. URL: https://www.jstor.org/stable/26677715.arxiv.org/abs/1604.03852
  20. [20]Vodev, G., 2020. Semiclassical resolvent estimates for Hölder potentials. Pure and Applied Analysis 2, 841–860. doi:10.2140/paa.2020.2.841.DOI
  21. [21]Witten, E., 1982. Supersymmetry and Morse theory. Journal of Differential Geometry 17, 661 – 692. URL: https://doi.org/10.4310/jdg/1214437492, doi:10.4310/jdg/1214437492.

Paper details

vlnc:W_42gea43j
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents