MAGNETIC SCHR ODINGER OPERATORS AND LANDSCAPE FUNCTIONS JEREMY G. HOSKINS HADRIAN QUAN AND STEFAN STEINERBERGER

2025-05-02 0 0 2.15MB 13 页 10玖币
侵权投诉
MAGNETIC SCHR ¨
ODINGER OPERATORS
AND LANDSCAPE FUNCTIONS
JEREMY G. HOSKINS, HADRIAN QUAN, AND STEFAN STEINERBERGER
Abstract. We study localization properties of low-lying eigenfunctions of
magnetic Schr¨odinger operators
1
2(i∇ − A(x))2φ+V(x)φ=λφ,
where V: Ω R0is a given potential and A: Ω Rdinduces a mag-
netic field. We extend the Filoche-Mayboroda inequality and prove a refined
inequality in the magnetic setting which can predict the points where low-
energy eigenfunctions are localized. This result is new even in the case of
vanishing magnetic field. Numerical examples illustrate the results.
1. Introduction
1.1. Introduction. A fundamental problem in mathematics and physics is to un-
derstand the behavior of low-energy eigenstates of Schr¨odinger operators. Given
an open and bounded domain Ω Rd,and a potential V: Ω R0, this amounts
to characterizing solutions of the equation
φ(x) + V(x)φ(x) = λφ(x) in Ω
φ= 0 on
for small values of λ. Our primary focus in this paper is on the regime where V
oscillates rapidly at small scales. It is known that solutions of this equation may
become strongly localized [3]. Moreover, it is also understood that in this setting
the boundary conditions are not tremendously important. We mention that in the
case V= 0 it is possible to have fascinating localization phenomena that stem from
Neumann boundary conditions (see [13, 19, 30]) but these will not be discussed
here. Our main question is whether it is possible to predict efficiently, using only
V, where such highly-localized eigenfunctions will be concentrated.
1.2. Landscape. This question has received renewed attention in recent years after
the introduction of the landscape function by Filoche & Mayboroda [14]. They
define the landscape function as u:Ω:RRsatisfying
(∆ + V)u= 1 in
u= 0 on
and prove that uexerts pointwise control on all eigenfunctions (∆ + V)φ=λφ
|φ(x)| ≤ λu(x)kφkL(Ω).
2010 Mathematics Subject Classification. 35J10, 65N25 (primary), 82B44 (secondary).
Key words and phrases. Localization, Eigenfunction, Schr¨odinger Operator, Regularization.
S.S. is supported by the NSF (DMS-2123224) and the Alfred P. Sloan Foundation.
1
arXiv:2210.02646v1 [math.AP] 6 Oct 2022
2
The landscape function turns out to have remarkable predictive power: the first
few eigenfunctions tend to localize close to the maxima of u. Indeed, one can think
of 1/u as a suitable regularization of the potential Vwith the property that eigen-
functions localize in its local minima. This has inspired considerable subsequent
work, we refer to [4, 5, 6, 1, 2, 8, 9, 14, 15, 16, 18, 20, 21, 22, 24, 27].
1.3. Local Landscape. Another approach was proposed by the third author in
[32, 33]: there it was shown that, on sufficiently small scales, there exists a canonical
smoothing of the potential Vgiven by the Wiener integral
Vt(x) = E1
tZt
0
V(ωx(s))ds,
where ωx(s) is a Brownian motion started at xat time s. As is shown in [33],
this mollifier arises naturally when looking for local approximations of the identity
with good error estimates. Moreover, there is a probability-free way of computing
this object. Since Brownian motion is, in free space, distributed according to a
Gaussian and since expectation is linear, we can express VtVkt, where
kt(x) = 1
tZt
0
exp −kxk2/(4s)
(4πs)d/2ds
with an exponentially small error provided that tdist(x, c)2. Note that con-
volution with ktcan be performed rapidly using the Fast Fourier Transform. The
local landscape is, in a suitable sense, the canonical mollifier for small t(see [33])
and was empirically shown to have predictive power roughly comparable to that of
the landscape function [25] for larger values of t.
2. Magnetic Schr¨
odinger Operators
Our goal is to extend and unify these two approaches for magnetic Schr¨odinger
operators. Let Ω Rdbe a bounded domain with smooth boundary (mainly for
ease of exposition, this could be relaxed). We consider the eigenvalue problem
1
2(i∇ − A(x))2φ(x) + V(x)φ(x) = λφ(x)
where V: Ω R0,A: Ω Rdand φis subject to Dirichlet boundary condi-
tions. The main technical ingredient in our approach is a Feynman-Kac formula
for magnetic Schr¨odinger operators due to Broderix, Hundertmark & Leschke [7].
In particular, we will work with the regularity assumptions AC1(Rd,Rd) and
VC1(Rd,R0) which could be somewhat weakened (see [7] for a discussion).
In our setting of interest where Vis wildly oscillatory, one does not expect the
boundary conditions to have a significant impact on the behavior of eigenfunctions
inside the domain. The only prior work in this direction is due to Poggi [28] who
proposes a different extension of the landscape function leading to different results.
2.1. A Magnetic Filoche-Mayboroda inequality. We start by extending the
Filoche-Mayboroda Landscape function to the magnetic setting. We consider the
solution of the equation, henceforth referred to as the landscape function,
1
2∆ + Vu= 1 in
u= 0 on .
3
The factor 1/2 in front of the Laplacian is equivalent to the usual Laplacian up
to scaling Vand λby a factor of 2: it is the probabilistic normalization of the
Laplacian chosen to facilitate comparison with the result in [7].
Theorem (Filoche-Mayboroda [14]).For any eigenfunction ((1/2)∆+V)φ=λφ
subject to Dirichlet boundary conditions and all x
|φ(x)|
kφkLλ·u(x).
Note that the quantity on the left-hand side always assumes the value 1 at the
global extrema of any eigenfunction. Therefore, eigenfunctions corresponding to
eigenvalue λcan only localize at points xΩ where u(x)·λ1. Our first main
result shows that this inequality extends to magnetic Schr¨odinger operators.
Theorem 1. Let Rdbe a bounded domain with smooth boundary, let A
C1(Ω,Rd), let VC1(Ω,R0)and let φbe a solution of
1
2(i∇ − A(x))2φ(x) + V(x)φ(x) = λφ(x) (1)
subject to Dirichlet boundary conditions. Then, for all x,
|φ(x)|
kφkLλ·u(x).
An interesting aspect of the inequality is that the magnetic potential Adoes not
explicitly appear in the inequality (it appears implicitly in λ). Indeed, if Ais much
smoother than the potential V(in the sense of being virtually constant on the
scale of the eigenfunction), this is not particularly surprising since, intuitively, it
contributes only a phase modulation to the corresponding eigenfunction (see §3).
2.2. A Refined Localization Inequality. We also prove a refined localization
inequality which is stronger than Theorem 1 (and will imply Theorem 1).
Theorem 2. Under the same assumptions as Theorem 1, for all xand t > 0,
|ψ(x0)|
kψkLEωχ(ωx0, t) exp λ·tZt
0
V(ωx0(s))ds
where expectation is taken with respect to all Brownian motions ωx(s)started at
ωx(0) = xrunning for tunits of time and
χ(ω, t) = (1if 0st:ω(s)
0otherwise.
Note that when tis sufficiently small compared to the distance d(x0, ∂Ω) between
x0and the boundary of the domain (in particular, when td(x0, ∂Ω)2), then
virtually all Brownian motions will stay inside the domain for tunits of time: the
difference between having the term χ(ω, t) and omitting it is exponentially small.
One important fact is that Theorem 2 contains all the information contained in
Theorem 1 and more: this is made precise in the following Corollary.
Corollary. Under the same assumption as Theorem 1, we have
inf
t>0
Eωχ(ωx0, t) exp λ·tZt
0
V(ωx0(s))dsλ·u(x).
We refer to §3 where this is illustrated on concrete examples.
摘要:

MAGNETICSCHRODINGEROPERATORSANDLANDSCAPEFUNCTIONSJEREMYG.HOSKINS,HADRIANQUAN,ANDSTEFANSTEINERBERGERAbstract.Westudylocalizationpropertiesoflow-lyingeigenfunctionsofmagneticSchrodingeroperators12(irA(x))2+V(x)=;whereV:!R0isagivenpotentialandA:!Rdinducesamag-netic eld.WeextendtheFiloche-Mayboro...

展开>> 收起<<
MAGNETIC SCHR ODINGER OPERATORS AND LANDSCAPE FUNCTIONS JEREMY G. HOSKINS HADRIAN QUAN AND STEFAN STEINERBERGER.pdf

共13页,预览3页

还剩页未读, 继续阅读

声明:本站为文档C2C交易模式,即用户上传的文档直接被用户下载,本站只是中间服务平台,本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私,请立即通知玖贝云文库,我们立即给予删除!
分类:图书资源 价格:10玖币 属性:13 页 大小:2.15MB 格式:PDF 时间:2025-05-02

开通VIP享超值会员特权

  • 多端同步记录
  • 高速下载文档
  • 免费文档工具
  • 分享文档赚钱
  • 每日登录抽奖
  • 优质衍生服务
/ 13
客服
关注