Quantum Graphs on Radially Symmetric Antitrees Noema Nicolussi - - PowerPoint PPT Presentation

quantum graphs on radially symmetric antitrees
SMART_READER_LITE
LIVE PREVIEW

Quantum Graphs on Radially Symmetric Antitrees Noema Nicolussi - - PowerPoint PPT Presentation

Quantum Graphs on Radially Symmetric Antitrees Noema Nicolussi University of Vienna (joint work with A. Kostenko) Differential Operators on Graphs and Waveguides, Graz 26 February 2019 Noema Nicolussi 26 February 2019 1 / 12 Definition Let


slide-1
SLIDE 1

Quantum Graphs on Radially Symmetric Antitrees

Noema Nicolussi University of Vienna (joint work with A. Kostenko)

Differential Operators on Graphs and Waveguides, Graz

26 February 2019

Noema Nicolussi 26 February 2019 1 / 12

slide-2
SLIDE 2

Definition

Let (sn)n be a sequence with s0 = 1 and sn ∈ N, n ≥ 1. The antitree for (sn)n is the (discrete) graph Ad = (V, E) obtained as follows: For every n ∈ N...

1

Put sn new vertices. Denote this vertex set by Sn.

2

Then connect every vertex in Sn with every vertex in Sn−1.

Ex.: sn := n + 1

S0 S1 S2 S3

Noema Nicolussi 26 February 2019 2 / 12

slide-3
SLIDE 3

Definition

Let (sn)n be a sequence with s0 = 1 and sn ∈ N, n ≥ 1. The antitree for (sn)n is the (discrete) graph Ad = (V, E) obtained as follows: For every n ∈ N...

1

Put sn new vertices. Denote this vertex set by Sn.

2

Then connect every vertex in Sn with every vertex in Sn−1.

Ex.: sn := n + 1

S0 S1 S2 S3

If every edge e ∈ E is assigned a finite edge length 0 < |e| < ∞, then A = (V, E, | · |) is called a metric antitree. ⇒ Quantum Graphs H (= Laplacians) on metric antitrees

Noema Nicolussi 26 February 2019 2 / 12

slide-4
SLIDE 4

Motivation: QG’s on different graph types?

Finite graphs: (= finitely many edges) σ(H) is purely discrete and the eigenvalues satisfy Weyl’s law.

Noema Nicolussi 26 February 2019 3 / 12

slide-5
SLIDE 5

Motivation: QG’s on different graph types?

Finite graphs: (= finitely many edges) σ(H) is purely discrete and the eigenvalues satisfy Weyl’s law. Infinite periodic graphs: σ(H) “usually” has band-gap structure (=union of closed intervals).

(Berkolaiko&Kuchment, “Introduction to Quantum Graphs”, 2013)

Noema Nicolussi 26 February 2019 3 / 12

slide-6
SLIDE 6

Motivation: QG’s on different graph types?

Finite graphs: (= finitely many edges) σ(H) is purely discrete and the eigenvalues satisfy Weyl’s law. Infinite periodic graphs: σ(H) “usually” has band-gap structure (=union of closed intervals).

(Berkolaiko&Kuchment, “Introduction to Quantum Graphs”, 2013)

Infinite (symmetric) Trees: Trees can be well analyzed. But: Their structure excludes (some) interesting phenomena!

(e.g. Solomyak 90’s; Breuer&Frank 08; Exner,Seifert&Stollmann 14)

Noema Nicolussi 26 February 2019 3 / 12

slide-7
SLIDE 7

Motivation: QG’s on different graph types?

Finite graphs: (= finitely many edges) σ(H) is purely discrete and the eigenvalues satisfy Weyl’s law. Infinite periodic graphs: σ(H) “usually” has band-gap structure (=union of closed intervals).

(Berkolaiko&Kuchment, “Introduction to Quantum Graphs”, 2013)

Infinite (symmetric) Trees: Trees can be well analyzed. But: Their structure excludes (some) interesting phenomena!

(e.g. Solomyak 90’s; Breuer&Frank 08; Exner,Seifert&Stollmann 14)

Goal:

A model that can be fully analyzed - but still with “rich behavior”? Random walks on antitrees have diverse behavior! (Counter-examples to “Grigoryan’s completeness theorem for graphs”; Wojchiechowski 2011)

Noema Nicolussi 26 February 2019 3 / 12

slide-8
SLIDE 8

The Kirchhoff Laplacian H

Let A be a metric antitree and L2(A) =

e∈E L2(0, |e|) its L2-space.

Then consider the maximal operator Hmax :=

e∈E He, where

He = −d2/dx2

e,

dom(He) = H2((0, |e|)).

Noema Nicolussi 26 February 2019 4 / 12

slide-9
SLIDE 9

The Kirchhoff Laplacian H

Let A be a metric antitree and L2(A) =

e∈E L2(0, |e|) its L2-space.

Then consider the maximal operator Hmax :=

e∈E He, where

He = −d2/dx2

e,

dom(He) = H2((0, |e|)). Kirchhoff conditions: For every vertex v :

  • f is continuous at v
  • e edges at v f ′

e(v) = 0,

  • Noema Nicolussi

26 February 2019 4 / 12

slide-10
SLIDE 10

The Kirchhoff Laplacian H

Let A be a metric antitree and L2(A) =

e∈E L2(0, |e|) its L2-space.

Then consider the maximal operator Hmax :=

e∈E He, where

He = −d2/dx2

e,

dom(He) = H2((0, |e|)). Kirchhoff conditions: For every vertex v :

  • f is continuous at v
  • e edges at v f ′

e(v) = 0,

  • Definition:

Define the pre-minimal Laplacian as H0 := Hmax ↾ dom(H0) with domain dom(H0) = {f ∈ dom(Hmax)|f ∈ L2

comp(A), f satisfies KH conditions}.

Noema Nicolussi 26 February 2019 4 / 12

slide-11
SLIDE 11

The Kirchhoff Laplacian H

Let A be a metric antitree and L2(A) =

e∈E L2(0, |e|) its L2-space.

Then consider the maximal operator Hmax :=

e∈E He, where

He = −d2/dx2

e,

dom(He) = H2((0, |e|)). Kirchhoff conditions: For every vertex v :

  • f is continuous at v
  • e edges at v f ′

e(v) = 0,

  • Definition:

Define the pre-minimal Laplacian as H0 := Hmax ↾ dom(H0) with domain dom(H0) = {f ∈ dom(Hmax)|f ∈ L2

comp(A), f satisfies KH conditions}.

Define the Kirchhoff Laplacian H by taking closure, H := H0.

Noema Nicolussi 26 February 2019 4 / 12

slide-12
SLIDE 12

Idea: Antitrees are highly symmetrical ⇒ “dimension reduction”

Noema Nicolussi 26 February 2019 5 / 12

slide-13
SLIDE 13

Idea: Antitrees are highly symmetrical ⇒ “dimension reduction”

Additional assumption:

The antitree A is radially symmetric, i.e. for each n ≥ 0, edges connecting the vertex sets Sn and Sn+1 have the same length, say ℓn > 0.

Noema Nicolussi 26 February 2019 5 / 12

slide-14
SLIDE 14

Idea: Antitrees are highly symmetrical ⇒ “dimension reduction”

Additional assumption:

The antitree A is radially symmetric, i.e. for each n ≥ 0, edges connecting the vertex sets Sn and Sn+1 have the same length, say ℓn > 0.

Theorem (Kostenko–N.):

The “symmetric part” Hsym of H is equivalent to the Sturm-Liouville

  • perator defined on L2([0, L); µ) by (here, tn :=

j<n ℓj, L = n ℓn)

τf := − 1 µ(x) d dx µ(x) d dx f , µ(x) =

  • n≥0

snsn+11[tn,tn+1)(x), and Neumann BC (f ′(0) = 0) at x = 0.

Noema Nicolussi 26 February 2019 5 / 12

slide-15
SLIDE 15

Idea: Antitrees are highly symmetrical ⇒ “dimension reduction”

Additional assumption:

The antitree A is radially symmetric, i.e. for each n ≥ 0, edges connecting the vertex sets Sn and Sn+1 have the same length, say ℓn > 0.

Theorem (Kostenko–N.):

The “symmetric part” Hsym of H is equivalent to the Sturm-Liouville

  • perator defined on L2([0, L); µ) by (here, tn :=

j<n ℓj, L = n ℓn)

τf := − 1 µ(x) d dx µ(x) d dx f , µ(x) =

  • n≥0

snsn+11[tn,tn+1)(x), and Neumann BC (f ′(0) = 0) at x = 0. Also, H decomposes as H = Hsym ⊕

  • n≥1

hn, where hn, n ≥ 1 are equivalent to regular, s.a. Sturm-Liouville operators.

Noema Nicolussi 26 February 2019 5 / 12

slide-16
SLIDE 16

Self-adjointness problem

Basic Questions: Is H self-adjoint? (For infinite graphs, H is not always self-adjoint! )

Noema Nicolussi 26 February 2019 6 / 12

slide-17
SLIDE 17

Self-adjointness problem

Basic Questions: Is H self-adjoint? (For infinite graphs, H is not always self-adjoint! )

If not, what are the deficiency indices n±(H) := dim ker(H∗ ± i)? ... and how do the self-adjoint extensions look like?

Noema Nicolussi 26 February 2019 6 / 12

slide-18
SLIDE 18

Self-adjointness problem

Basic Questions: Is H self-adjoint? (For infinite graphs, H is not always self-adjoint! )

If not, what are the deficiency indices n±(H) := dim ker(H∗ ± i)? ... and how do the self-adjoint extensions look like?

Theorem (Kostenko–N.):

Let A be a r.s. AT of volume vol(A) :=

e∈E |e| = n snsn+1ℓn.

Noema Nicolussi 26 February 2019 6 / 12

slide-19
SLIDE 19

Self-adjointness problem

Basic Questions: Is H self-adjoint? (For infinite graphs, H is not always self-adjoint! )

If not, what are the deficiency indices n±(H) := dim ker(H∗ ± i)? ... and how do the self-adjoint extensions look like?

Theorem (Kostenko–N.):

Let A be a r.s. AT of volume vol(A) :=

e∈E |e| = n snsn+1ℓn. Then

H is self-adjoint ⇐ ⇒ vol(A) = ∞. Moreover, if H is not self-adjoint, then n±(H) = 1.

Noema Nicolussi 26 February 2019 6 / 12

slide-20
SLIDE 20

Self-adjointness problem

Basic Questions: Is H self-adjoint? (For infinite graphs, H is not always self-adjoint! )

If not, what are the deficiency indices n±(H) := dim ker(H∗ ± i)? ... and how do the self-adjoint extensions look like?

Theorem (Kostenko–N.):

Let A be a r.s. AT of volume vol(A) :=

e∈E |e| = n snsn+1ℓn. Then

H is self-adjoint ⇐ ⇒ vol(A) = ∞. Moreover, if H is not self-adjoint, then n±(H) = 1. The symmetry assumption is crucial. We can construct non-symmetric, finite volume antitress with n±(H) = +∞!

Noema Nicolussi 26 February 2019 6 / 12

slide-21
SLIDE 21

The Finite Volume Case (= H is not self-adjoint)

Theorem (Kostenko–N.):

(i) Self-adjoint extensions form a one-parameter family Hθ, θ ∈ [0, π) given by boundary conditions at “infinity” cos(θ)f (∞) = sin(θ)f ′(∞), θ ∈ [0, π), (0.1) where f (∞) := lim|x|→L f (x) and f ′(∞) := limr→L

  • |x|=r f ′(x).

Noema Nicolussi 26 February 2019 7 / 12

slide-22
SLIDE 22

The Finite Volume Case (= H is not self-adjoint)

Theorem (Kostenko–N.):

(i) Self-adjoint extensions form a one-parameter family Hθ, θ ∈ [0, π) given by boundary conditions at “infinity” cos(θ)f (∞) = sin(θ)f ′(∞), θ ∈ [0, π), (0.1) where f (∞) := lim|x|→L f (x) and f ′(∞) := limr→L

  • |x|=r f ′(x).

(ii) The spectrum of Hθ is purely discrete for all θ and eigenvalues satisfy Weyl’s law N(λ; Hθ) = vol(A) π √ λ(1 + o(1)), λ → +∞. Here, N(λ; Hθ) = #(eigenvalues ≤ λ) is the ev. counting function.

Noema Nicolussi 26 February 2019 7 / 12

slide-23
SLIDE 23

The Infinite Volume Case (= H is s.a.): Basic properties

Theorem (Kostenko–N.):

(i) The spectrum of H is purely discrete if and only if lim

n→∞

  • k≤n

sksk+1ℓk

  • k≥n

ℓk sksk+1 = 0.

Noema Nicolussi 26 February 2019 8 / 12

slide-24
SLIDE 24

The Infinite Volume Case (= H is s.a.): Basic properties

Theorem (Kostenko–N.):

(i) The spectrum of H is purely discrete if and only if lim

n→∞

  • k≤n

sksk+1ℓk

  • k≥n

ℓk sksk+1 = 0. (ii) H−1 is trace class (i.e.,

λ∈σ(H) |λ|−1 < ∞) if and only if

  • n≥1

snsn+1ℓ2

n < ∞

and

  • n≥0

ℓn snsn+1

n−1

  • k=0

sksk+1ℓk < ∞.

Noema Nicolussi 26 February 2019 8 / 12

slide-25
SLIDE 25

The Infinite Volume Case (= H is s.a.): Basic properties

Theorem (Kostenko–N.):

(i) The spectrum of H is purely discrete if and only if lim

n→∞

  • k≤n

sksk+1ℓk

  • k≥n

ℓk sksk+1 = 0. (ii) H−1 is trace class (i.e.,

λ∈σ(H) |λ|−1 < ∞) if and only if

  • n≥1

snsn+1ℓ2

n < ∞

and

  • n≥0

ℓn snsn+1

n−1

  • k=0

sksk+1ℓk < ∞. Idea: Apply results from spectral theory of Krein strings! Similarly: characterization of invertibility, spectral gap estimates,...

  • I. S. Kac & M. G. Krein, Criteria for the discreteness of the spectrum of a

singular string, Izv. VUZov, Matematika, no. 2 (3), 136–153 (1958).

  • I. S. Kac & M. G. Krein, On the spectral functions of the string, AMS, 1974.

Noema Nicolussi 26 February 2019 8 / 12

slide-26
SLIDE 26

The Infinite Volume Case (= H is s.a.): Spectral Types

σ(H) = σac(H) ∪ σpp(H) ∪ σsc(H) Recall the decomposition: H = Hsym ⊕

n≥1 hn

AT’s have “large point spectrum”: {k2π2/ℓ2

n; k, n ∈ N} ⊆ σpp(H)

Noema Nicolussi 26 February 2019 9 / 12

slide-27
SLIDE 27

The Infinite Volume Case (= H is s.a.): Spectral Types

σ(H) = σac(H) ∪ σpp(H) ∪ σsc(H) Recall the decomposition: H = Hsym ⊕

n≥1 hn

AT’s have “large point spectrum”: {k2π2/ℓ2

n; k, n ∈ N} ⊆ σpp(H)

... however, “typically” no absolutely continuous spectrum!

Theorem (Kostenko–N.):

Suppose the sets {ℓn}n≥0 and {sn+2

sn }n≥0 are finite and lim infn≥0 sn+2 sn

> 1. Then: σac(H) = ∅ if and only if {(ℓn, sn+2

sn )}n≥0 is eventually periodic.

Noema Nicolussi 26 February 2019 9 / 12

slide-28
SLIDE 28

The Infinite Volume Case (= H is s.a.): Spectral Types

σ(H) = σac(H) ∪ σpp(H) ∪ σsc(H) Recall the decomposition: H = Hsym ⊕

n≥1 hn

AT’s have “large point spectrum”: {k2π2/ℓ2

n; k, n ∈ N} ⊆ σpp(H)

... however, “typically” no absolutely continuous spectrum!

Theorem (Kostenko–N.):

Suppose the sets {ℓn}n≥0 and {sn+2

sn }n≥0 are finite and lim infn≥0 sn+2 sn

> 1. Then: σac(H) = ∅ if and only if {(ℓn, sn+2

sn )}n≥0 is eventually periodic.

  • C. Remling, The absolutely continuous spectrum of Jacobi matrices, Ann.
  • Math. 174, no. 1, 125–171 (2011).

Noema Nicolussi 26 February 2019 9 / 12

slide-29
SLIDE 29

The Infinite Volume Case (= H is s.a.): Spectral Types

σ(H) = σac(H) ∪ σpp(H) ∪ σsc(H) Recall the decomposition: H = Hsym ⊕

n≥1 hn

AT’s have “large point spectrum”: {k2π2/ℓ2

n; k, n ∈ N} ⊆ σpp(H)

... however, “typically” no absolutely continuous spectrum!

Theorem (Kostenko–N.):

Assume that infn ℓn > 0. If sup

n ℓn = ∞

and lim inf

n≥0

sn+2 sn > 1, then σ(H) = R≥0 and σac(H) = ∅.

Noema Nicolussi 26 February 2019 10 / 12

slide-30
SLIDE 30

The Infinite Volume Case (= H is s.a.): Spectral Types II

But, for some antitrees σac(H) is very large!

Noema Nicolussi 26 February 2019 11 / 12

slide-31
SLIDE 31

The Infinite Volume Case (= H is s.a.): Spectral Types II

But, for some antitrees σac(H) is very large!

Theorem (Kostenko–N.):

  • n≥0

sn+2 sn − 1 2 < ∞ and inf

n≥0 ℓn > 0

= ⇒ σac(H) = R≥0.

Noema Nicolussi 26 February 2019 11 / 12

slide-32
SLIDE 32

The Infinite Volume Case (= H is s.a.): Spectral Types II

But, for some antitrees σac(H) is very large!

Theorem (Kostenko–N.):

  • n≥0

sn+2 sn − 1 2 < ∞ and inf

n≥0 ℓn > 0

= ⇒ σac(H) = R≥0. Example: Take sn = n + 1: sn+2 sn − 1 = n + 3 n + 1 − 1 = 2 n + 1

Noema Nicolussi 26 February 2019 11 / 12

slide-33
SLIDE 33

The Infinite Volume Case (= H is s.a.): Spectral Types II

But, for some antitrees σac(H) is very large!

Theorem (Kostenko–N.):

  • n≥0

sn+2 sn − 1 2 < ∞ and inf

n≥0 ℓn > 0

= ⇒ σac(H) = R≥0. Example: Take sn = n + 1: sn+2 sn − 1 = n + 3 n + 1 − 1 = 2 n + 1 Trace class arguments do not apply! In some sense, the condition means Hilbert–Schmidt perturbation!

Noema Nicolussi 26 February 2019 11 / 12

slide-34
SLIDE 34

The Infinite Volume Case (= H is s.a.): Spectral Types II

But, for some antitrees σac(H) is very large!

Theorem (Kostenko–N.):

  • n≥0

sn+2 sn − 1 2 < ∞ and inf

n≥0 ℓn > 0

= ⇒ σac(H) = R≥0. Example: Take sn = n + 1: sn+2 sn − 1 = n + 3 n + 1 − 1 = 2 n + 1 Trace class arguments do not apply! In some sense, the condition means Hilbert–Schmidt perturbation! Idea: Use “Szeg¨

  • ’s theorem” for Krein strings

Bessonov&Denisov, A spectral Szeg˝

  • theorem on the real line, preprint,

arXiv:1711.05671 (2017).

Noema Nicolussi 26 February 2019 11 / 12

slide-35
SLIDE 35

Example: Polynomial antitrees

Let q ∈ N and s > 0. Consider the polynomial antitree Aq,s given by sn = (n + 1)q, ℓn = (n + 1)−s.

Theorem (Kostenko–N.):

Let H be the Kirchhoff-Laplacian on Aq,s. Then H is self-adjoint ⇐ ⇒ s ≤ 2q + 1.

Noema Nicolussi 26 February 2019 12 / 12

slide-36
SLIDE 36

Example: Polynomial antitrees

Let q ∈ N and s > 0. Consider the polynomial antitree Aq,s given by sn = (n + 1)q, ℓn = (n + 1)−s.

Theorem (Kostenko–N.):

Let H be the Kirchhoff-Laplacian on Aq,s. Then H is self-adjoint ⇐ ⇒ s ≤ 2q + 1. Assume further that H is self-adjoint (s ≤ 2q + 1). Then: If s < 1, then σac(H) = R≥0.

Noema Nicolussi 26 February 2019 12 / 12

slide-37
SLIDE 37

Example: Polynomial antitrees

Let q ∈ N and s > 0. Consider the polynomial antitree Aq,s given by sn = (n + 1)q, ℓn = (n + 1)−s.

Theorem (Kostenko–N.):

Let H be the Kirchhoff-Laplacian on Aq,s. Then H is self-adjoint ⇐ ⇒ s ≤ 2q + 1. Assume further that H is self-adjoint (s ≤ 2q + 1). Then: If s < 1, then σac(H) = R≥0. H is invertible if and only if s ≥ 1. The spectrum σ(H) is purely discrete if and only if s > 1.

Noema Nicolussi 26 February 2019 12 / 12

slide-38
SLIDE 38

Example: Polynomial antitrees

Let q ∈ N and s > 0. Consider the polynomial antitree Aq,s given by sn = (n + 1)q, ℓn = (n + 1)−s.

Theorem (Kostenko–N.):

Let H be the Kirchhoff-Laplacian on Aq,s. Then H is self-adjoint ⇐ ⇒ s ≤ 2q + 1. Assume further that H is self-adjoint (s ≤ 2q + 1). Then: If s < 1, then σac(H) = R≥0. H is invertible if and only if s ≥ 1. The spectrum σ(H) is purely discrete if and only if s > 1. H−1 is trace class if and only if s > q + 1

2.

Thank you for your attention!

  • A. Kostenko and N. Nicolussi, Quantum Graphs on Radially Symmetric

Antitrees, submitted, arXiv:1901.05404 (2019).

Noema Nicolussi 26 February 2019 12 / 12

slide-39
SLIDE 39

References I

  • G. Berkolaiko and P. Kuchment, Introduction to Quantum Graphs,
  • Amer. Math. Soc., Providence, RI, 2013.
  • R. V. Bessonov and S. A. Denisov, A spectral Szeg˝
  • theorem on the

real line, preprint, arXiv:1711.05671.

  • J. Breuer and R. Frank, Singular spectrum for radial trees, Rev. Math.
  • Phys. 21, no. 7, 929–945 (2009).
  • J. Breuer and M. Keller, Spectral analysis of certain spherically

homogeneous graphs, Oper. Matrices 7(4), 825–847 (2013).

  • R. Brooks, A relation between growth and the spectrum of the

Laplacian, Math. Z. 178, 501–508 (1981).

  • P. Exner, C. Seifert, and P. Stollmann, Absence of absolutely

continuous spectrum for the Kirchhoff Laplacian on radial trees, Ann. Henri Poincar´ e 15, no. 6, 1109–1121 (2014).

slide-40
SLIDE 40

References II

  • A. Grigor’yan, On stochastically complete manifolds (in Russian),
  • Dokl. Akad. Nauk SSSR 290, no. 3, 534–537 (1986), Engl. transl. :

Soviet Math. Dokl. 34, no.2, 310– 313 (1987).

  • I. S. Kac and M. G. Krein, Criteria for the discreteness of the

spectrum of a singular string, (Russian), Izv. Vysˇ

  • s. Uˇ
  • cebn. Zaved.

Matematika, no. 2 (3), 136–153 (1958).

  • I. S. Kac and M. G. Krein, On the spectral functions of the string,
  • Amer. Math. Soc. Transl. Ser. 2 103, 19–102 (1974).
  • M. Keller, Intrinsic metric on graphs: a survey, Math. Technol. of

Networks, 81–119 (2015).

  • C. Remling, The absolutely continuous spectrum of Jacobi matrices,
  • Ann. Math. 174, no. 1, 125–171 (2011).
slide-41
SLIDE 41

References III

  • M. Solomyak, On the spectrum of the Laplacian on regular metric

trees, Waves Random Media 14, S155–S171 (2004).

  • R. K. Wojciechowski, Stochastically incomplete manifolds and graphs,

in: D. Lenz et.al. (Eds.), “Random walks, boundaries and spectra”, 163–179, Progr. Probab. 64, Birkh¨ auser/Springer Basel AG, Basel, 2011.