A new signature of quantum phase transitions from the numerical - - PowerPoint PPT Presentation

a new signature of quantum phase transitions from the
SMART_READER_LITE
LIVE PREVIEW

A new signature of quantum phase transitions from the numerical - - PowerPoint PPT Presentation

A new signature of quantum phase transitions from the numerical range talk at the conference Entropy 2018: From Physics to Information Sciences and Geometry University of Barcelona, Spain May 15th, 2018 speaker Stephan Weis Centre for


slide-1
SLIDE 1

A new signature of quantum phase transitions from the numerical range

talk at the conference

Entropy 2018: From Physics to Information Sciences and Geometry

University of Barcelona, Spain May 15th, 2018 speaker

Stephan Weis

Centre for Quantum Information and Communication Université libre de Bruxelles, Belgium joint work with

Ilya M. Spitkovsky

New York University Abu Dhabi, United Arab Emirates

slide-2
SLIDE 2

Overview

  • 1. Ground Energy
  • 2. Convex Geometry
  • 3. Numerical Range
  • 4. Results
  • 5. Conclusion
slide-3
SLIDE 3

Part I Ground Energy

slide-4
SLIDE 4

Quantum Phase Transitions

are characterized in terms of 1) long-range correlation in ground state 2) non-analytic ground energy 3) geometry of reduced density matrices

Zauner-Stauber et al. New J. Phys. 18 (2016), 113033 & Chen et al. Phys. Rev. A 93 (2016), 012309

4) strong variation / discontinuity of MaxEnt maps

Arrachea et al. Phys. Rev. A 45 (1992), 7104 & Chen et al. New J. Phys. 17 (2015), 083019

this talk clarifies in the finite-dimensional setting relationships between 2), 3), 4) and certain open mapping properties

slide-5
SLIDE 5

Differentiability of Ground Energy

  • ne-parameter Hamiltonian H(g) = H0 + g ⋅ H1,

g ∈ R angular representation A(θ) = cos(θ)H0 + sin(θ)H1, θ ∈ (−π

2, π 2)

ground energy λ(X) = minimal eigenvalue of X Observation λ ○ H is Ck/analytic at tan(θ) ⇐ ⇒ λ ○ A is Ck/analytic at θ focus on ground energy λ(θ) = λ ○ A(θ)

advantage: reduced density matrices and MaxEnt maps are easier described in angular coordinates

slide-6
SLIDE 6

Part II Convex Geometry

slide-7
SLIDE 7

Convex Geometry

K ⊂ C ≅ R2 compact convex support function ˜ hK ∶ C → R, ˜ hK(u) = minz∈K⟨z, u⟩ exposed face F(u) = argminz∈K⟨z, u⟩ ˜ hF(u)(v) = ˜ h′

K(u; v) = directional

derivative = lim

t↘0 1 t (˜

hK(u + tv) − ˜ hK(u)) end-points of exposed face, u ∈ S1 x±(u) = u˜ hK(u) ± u⊥˜ h′

K(u; ±u⊥) ∈ ∂K

using hK(θ)= ˜ hK(ei θ), x±(ei θ) = ei θ(hK(θ) ± i h′

K(θ; ±1)) ∈ ∂K

x± restricted to u ∈ S1 with x+(u) = x−(u) is the reverse Gauss map x±(u) = ∇˜ h(u) which parametrizes ∂K as an envelope

slide-8
SLIDE 8

Part III Numerical Range

slide-9
SLIDE 9

Numerical Range

density matrices Dn = {ρ ∈ Mn ∶ ρ ⪰ 0,tr(ρ) = 1} numerical range W = {⟨ψ∣H0 + iH1∣ψ⟩ ∶ ⟨ψ∣ψ⟩ = 1} = {trρ(H0 + iH1) ∶ ρ ∈ Dn}

W is the set of expected values of H0 and H1 (reduced density matrices)

Theorem 1 (Toeplitz) hW(θ) = λ ○ A(θ) = λ(θ)

  • Math. Z. 2 (1918), 187

von Neumann entropy S(ρ) = −trρlog(ρ), ρ ∈ Dn maximum-entropy inference map (MaxEnt map) ρ∗ ∶ W → Dn, z ↦ argmax{S(ρ) ∶ trρ(H0 + iH1),ρ ∈ Dn}

slide-10
SLIDE 10

Numerical Range — Diagonal Matrices

H0 = diag(E1

0,...,En 0 ) and H1 = diag(E1 1,...,En 1 )

W = conv{E1

0 + iE1 1,...,En 0 + iEn 1 }

A(θ) = diag(E1

0 cos(θ) + E1 1 sin(θ),...,En 0 cos(θ) + En 1 sin(θ))

  • W is a polytope
  • λ is piecewise harmonic
  • x+ and x− are piecewise

constant

  • flat boundary portions of W

≅ non-differentiable points of λ

  • ρ∗ is continuous
slide-11
SLIDE 11

Numerical Range — Non-Commutative

we assume dim(W) = 2 ⇐ [H0,H1] ≠ 0 analytic curves λ1(θ),...,λn(θ) and ONB’s ∣ψk(θ)⟩n

k=1 such that

A(θ) =

n

k=1

λk(θ)∣ψk(θ)⟩⟨ψk(θ)∣

Rellich, IMM-NYU 2, New York: New York University, 1954

  • λ is piecewise analytic
  • the maximal order of

differentiability of λ is even at non-analytic points analytic

  • max. order 0
  • max. order 2
slide-12
SLIDE 12

Numerical Range — Continuity of Inference

ρ∗ is analytic on the interior of W, W ○ ∋ z ↦ eµ1H1+µ2H2/tr(”), if z = x+(ei θ) then ρ∗(z) = maximally mixed state on span{∣ψk(θ)⟩ ∶ λk(θ) = λ(θ), λ′

k(θ) = λ′(θ;+1)}

  • the maps x+, x− ∶ S1 → ∂W cover all extreme points of W
  • ρ∗∣∂W may be discontinuous at extreme points of W because of C2 smooth

eigenvalue crossings with the ground energy λ

  • ρ∗∣F(u) is continuous on flat boundary portions F(u) ⊂ ∂W of W
  • for z ∈ ∂W: ρ∗∣∂W is continuous at z ⇐

⇒ ρ∗ is continuous at z

  • discontinuities of ρ∗ are irremovable because ρ∗(W) ⊂ ρ∗(W ○),

Wichmann JMP 4 (1963), 884

  • discontinuities of ρ∗∣∂W may be removable
slide-13
SLIDE 13

Numerical Range — Open Mappings

definitions

a map α ∶ X → Y between topological spaces is open at x ∈ X if α maps neighborhoods of x to neighborhoods of α(x) numerical range map f ∶ {∣ψ⟩ ∶ ⟨ψ∣ψ⟩ = 1} → W, ∣ψ⟩ ↦ ⟨ψ∣H0 + i H1∣ψ⟩ expected value map E ∶ Dn → W, ρ ↦ tr ρ(H0 + i H1) the inverse numerical range map f −1 is strongly (resp. weakly) continuous at z ∈ W if for all (resp. for at least one) ∣ψ⟩ ∈ f −1(z) the map f is open at ∣ψ⟩ Noticeable: Openness of linear maps on state spaces of C∗-algebras are studied since the 70’s (Lima, Vesterstrøm, O’Brian), with applications to quantum information theory: Shirokov, Izvestiya: Math. 76 (2012), 840

slide-14
SLIDE 14

Part IV Results

slide-15
SLIDE 15

Smoothness of λ, geometry of W, continuity of ρ∗

Let z = x+(ei θ) = x−(ei θ) be not a corner point: the statements in each column are equivalent λ is analytic locally at θ λ is C2k but not C2k+1 locally at θ, k ≥ 1 λ(θ) = λk(θ) = λl(θ) λ′(θ) = λ′

k(θ) = λ′ l(θ)

⇒ λk = λl ∃k ∶ λ = λk locally at θ / ∃ k ∶ λ = λk locally at θ ∂W is an analytic manifold locally at z ∂W is a C2k but not a C2k+1 manifold locally at z, k ≥ 1 ρ∗ is continuous at z ρ∗∣∂W has a removable discontinuity at z ρ∗∣∂W has an irremovable discontinuity at z Notice: S1 → ∂W, ei θ ↦ x±(θ) = ei θ(λ(θ) + i λ′(θ)) is only C2k−1 if λ is C2k !

slide-16
SLIDE 16

Open Mapping Conditions

Let z ∈ W be arbitrary: the statements in each column are equivalent ρ∗ is continuous at z ρ∗∣∂W has a removable discontinuity at z ρ∗∣∂W has an irremovable discontinuity at z f −1 is strongly continuous at z f −1 is weakly but not strongly continuous at z f −1 is not weakly continuous at z E is open at ρ∗(z) E is not open at ρ∗(z)

slide-17
SLIDE 17

Part V Conclusion

slide-18
SLIDE 18

Summary: Geometry and inference approach to the smoothness of the ground energy of a

  • ne-parameter Hamiltonian

References:

Weis and Knauf, Entropy distance: New quantum phenomena, JMP 53 (2012), 102206 Leake et al., Inverse continuity on the boundary of the numerical range, Linear and Multilinear Algebra 62 (2014), 1335 Rodman et al., Continuity of the maximum-entropy inference: Convex geometry and numerical ranges approach, JMP 57 (2016), 015204 Spitkovsky and Weis, A new signature of quantum phase transitions from the numerical range, arXiv:1703.00201 [math-ph]

slide-19
SLIDE 19

Thank you