Eigenstate Thermalization in Sachdev-Ye-Kitaev model & Schwarzian theory
by Pranjal Nayak
[basesd on 1901.xxxxx] with Julian Sonner and Manuel Vielma
University of Michigan, Ann Arbor January 23, 2019 Closed Quantum Systems
Quantum Mechanics is unitary!
(t) = (t, t ) (t ) |
How come we observe thermal physics?
How come we observe black hole formation? Plan of the talk
✓ Review of ETH
✓ Review of Sachdev-Ye-Kitaev model
✓ Numerical studies of ETH in SYK model
✓ ETH in the Schwarzian sector of the SYK model
✓ ETH in the Conformal sector of the SYK model
✓ A few thoughts on the bulk duals
✓ Summary and conclusion
!3 Eigenstate Thermalization Hypothesis (ETH) Quantum Thermalization
5 Quantum Thermalization
Classical thermalization: ergodicity & ⇐ chaos
5 Quantum Thermalization
Classical thermalization: ergodicity & ⇐ chaos
v.s.
Quantum thermalization: Eigenstate Thermalisation Hypothesis
S(E)/2 m n = (E) + e f(E,!)R h |O| i Omc mn mn
5 Quantum Thermalization
Classical thermalization: ergodicity & ⇐ chaos
v.s.
Quantum thermalization: Eigenstate Thermalisation Hypothesis
S(E)/2 m n = (E) + e f(E,!)R h |O| i Omc mn mn micro-canonical ensemble avg. energy of ensemble
5 Quantum Thermalization
Classical thermalization: ergodicity & ⇐ chaos
v.s.
Quantum thermalization: Eigenstate Thermalisation Hypothesis
S(E)/2 m n = (E) + e f(E,!)R h |O| i Omc mn mn micro-canonical entropy 1 in RMT random matrix
5 Quantum Thermalization
Classical thermalization: ergodicity & ⇐ chaos
v.s.
Quantum thermalization: Eigenstate Thermalisation Hypothesis
S(E)/2 m n = (E) + e f(E,!)R h |O| i Omc mn mn micro-canonical entropy 1 in RMT random matrix
A generic excited state will then thermalise by dephasing it(E E ) S = c⇤c e i j O E¯ + e h |O| i i j Oij !
!7 SYK model: a new Guinea Pig
1 J J4,5,7,8 10 1,2,4,6 7 4 9 6 3 J7,9,10,11 2 8 11 5 J2,3,5,6
[Sachdev-Ye ’93; Kitaev ’15]
!7 SYK model: a new Guinea Pig
1 J J4,5,7,8 10 1,2,4,6 7 4 9 6 3 J7,9,10,11 2 8 11 5 J2,3,5,6
[Sachdev-Ye ’93; ‣ a model of N Majorana fermions Kitaev ’15] ‣ with all-to-all couplings ‣ and quenched random couplings
!7 SYK model: a new Guinea Pig
1 J J4,5,7,8 10 1,2,4,6 7 4 9 6 3 J7,9,10,11 2 8 11 5 J2,3,5,6
[Sachdev-Ye ’93; ‣ a model of N Majorana fermions Kitaev ’15] ‣ with all-to-all couplings ‣ and quenched random couplings
H = − J ψ ψ ψ ψ ∑ i1i2i3i4 i1 i2 i3 i4 1≤i1 where, J is chosen from a Gaussian ensemble: 2 2 3! J ⟨Jijkl⟩ = 0 ⟨J ⟩ = ijkl N3 !7 Solvable Limit of SYK [Sachdev ’15; Parcollet, Georges ’00; Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] 8 Solvable Limit of SYK [Sachdev ’15; Parcollet, Georges ’00; Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] ‣ at large-N, it is often useful to use bi-local effective action 2 Scol J q 1 − = log Pf [∂ − Σ] + dτdτ′| G(τ, τ′) | − dτdτ′Σ (τ′, τ)G(τ, τ′) N τ 2q ∫ 2 ∫ 8 Solvable Limit of SYK [Sachdev ’15; Parcollet, Georges ’00; Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] ‣ at large-N, it is often useful to use bi-local effective action 2 Scol J q 1 − = log Pf [∂ − Σ] + dτdτ′| G(τ, τ′) | − dτdτ′Σ (τ′, τ)G(τ, τ′) N τ 2q ∫ 2 ∫ 1 is the Lagrange multiplier that imposes Σ(τ, τ′) G(τ, τ′) = ∑ ψi(τ) ψi(τ′) N i 8 Solvable Limit of SYK [Sachdev ’15; Parcollet, Georges ’00; Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] ‣ at large-N, it is often useful to use bi-local effective action 2 Scol J q 1 − = log Pf [∂ − Σ] + dτdτ′| G(τ, τ′) | − dτdτ′Σ (τ′, τ)G(τ, τ′) N τ 2q ∫ 2 ∫ 1 is the Lagrange multiplier that imposes Σ(τ, τ′) G(τ, τ′) = ∑ ψi(τ) ψi(τ′) N i ‣ In the large-N limit, the saddle point equations are given by, ∂τ G(τ, τ′) − [Σ * G](τ, τ′) = δ(τ − τ′) 2 q−1 J G (τ, τ′) = Σ(τ, τ′) 8 Solvable Limit of SYK [Sachdev ’15; Parcollet, Georges ’00; Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] ‣ at large-N, it is often useful to use bi-local effective action 2 Scol J q 1 − = log Pf [∂ − Σ] + dτdτ′| G(τ, τ′) | − dτdτ′Σ (τ′, τ)G(τ, τ′) N τ 2q ∫ 2 ∫ 1 is the Lagrange multiplier that imposes Σ(τ, τ′) G(τ, τ′) = ∑ ψi(τ) ψi(τ′) N i ‣ In the large-N limit, the saddle point equations are given by, ∂τ G(τ, τ′) − [Σ * G](τ, τ′) = δ(τ − τ′) 2 q−1 J G (τ, τ′) = Σ(τ, τ′) captures Melonic Physics! 8 Emergent Symmetry of SYK [Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] 9 Emergent Symmetry of SYK [Kitaev ’15; Polchinski, Rosenhaus ’16; Jevicki, Suzuki, Yoon ’16; Maldacena, Stanford ’16] ‣ In the deep IR limit, | τ | J ≫ 1 , the saddle point equations can be replaced by