The Hilbert-glass transition: Figuring out excited states in strongly disordered systems David Pekker (U. Pitt) Gil Refael (Caltech) Vadim Oganesyan (CUNY) Ehud Altman (Weizmann) Eugene Demler (Harvard) Outline • Quantum criticality in the quantum Ising model + preview of punchline • Disordered Quantum Ising model and real-space RG • Extending to excited states – the RSRG-X method • The Hilbert-glass transition Standard model of Quantum criticality • Quantum Ising model: H J nz nz1 h nx z z z h h J z z x Standard model of Quantum criticality H J nz nz1 h nx • Quantum Ising model: z z z z z x h • Phase diagram T h J TQC ~ J h z Quantum critical regime Ferromagnet Paramagnet QCP J h Disordered Quantum Ising model • Quantum Ising model: z z zz • Phase diagram: zz zz x x H J h h n n n nn11 nn n z zz T z z x TQC ~ J h z Quantum critical regime TQC ~ Exp J h Ferromagnet Paramagnet QCP J h Surprise: Transition in all excited states H J n nz nz1 hn nx • Quantum Ising model: z • Phase diagram: z x Hilbert glass transition T typ ~ z z z z ... ... ... ... Or: ... ... PM [All eigenstates doubly degenerate] FM QCP J h Surprise: Transition in all excited states • Quantum Ising model: z x x x x J nnz h HH J n nnzz h J ' 1 n1 n n n n n n1 z x • Phase diagram: z x x Hilbert glass transition T typ ~ z x z x z z x ... ... ... ... Or: ... ... PM [All eigenstates doubly degenerate] FM QCP J h Surprise: Transition in all excited states • Dynamical quantum phase transition. • Temperature tuned, but with no Thermodynamic signatures. • Accessible example for an MBL like transition. • Phase diagram: Hilbert glass transition T typ ~ x-phase ... ... ... ... Or: ... ... PM FM QCP J h Hilbert glass phase Disarming disorder: Real space RG [Ma, Dasgupta, Hu (1979), Bhatt, Lee (1979), DS Fisher (1995)] • Isolate the strongest bond (or field) in the chain. J max1z 2z • Choose ground-state manifold. J max hleft , hright • neighboring fields: quantum fluctuations. E Domain-wall excitations J max 1 2 , 1 2 hleft1x hright 2x 1 Cluster ground state: 1 2 J max 1 2 , 1 2 heff 2 hlefthright J max Disarming disorder: Real space RG [Ma, Dasgupta, Hu (1979), Bhatt, Lee (1979), DS Fisher (1995)] • Isolate the strongest bond (or field) in the chain. • Choose ground-state manifold. • neighboring fields: quantum fluctuations. hmax 2x hmax J left , J right hmax E Anti-aligned: 2 J left1z 2z 1 Field aligned: J right 2z 3z 2 hmax 2 3 1 X 2 J eff J left J right hmax 3 RG sketch •Ferromagnetic phase: J h X •Paramagnetic phase: J h X X X X Universal coupling distributions and RG flow • Initially, h and J have some coupling distributions: (h) (J ) hmax h J max J Universal coupling distributions and RG flow • As renormalization proceeds, universal distributions emerge: RG (h) RG (J ) 1 1 J 1 g J 1 g h h max{J , h} h g h , g J flow with RG J • gh and gJ flow: RG-flow • These functions are attractors for all initial distributions. Paramagnet QCP 0 Ferromagnet ln J ln h g J gh What about excited states? Pekker, GR, VO, Altman, Demler; Huse, Nandkishore, VO, Sondhi, Pal (2013) E • Put domain walls in strongest bonds: J max1z 2z • neighboring fields: quantum fluctuations. J max hleft , hright Domain-wall excitations hleft1x J max 1 2 , 1 2 hright 2x 1 1 2 Excited eff h Cluster ground state: J max 1 2 , 1 2 No effect on coupling magnitude! 2 GS eff h hlefthright J max hlefthright J max What about excited states? Pekker, GR, VO, Altman, Demler; Huse, Nandkishore, VO, Sondhi, Pal (2013) hmax 2x • Make spins antialigned with strong fields: hmax J left , J right hmax E Anti-aligned: 2 J right 2z 3z J left1z 2z 1 Field aligned: 2 3 hmax X 2 1 J Excited eff 2 • No effect on coupling magnitude! J GS eff 3 J left J right hmax J left J right hmax RSRG-X Tree of states • At each RG step, choose ground state or excitation: [six sites with large disorder] RG sketch •Hilbert-glass phase: J h X •Paramagnetic phase: X X J h X X Excited state flow • Universal distribution functions independent of choice: RG (| h |) RG (| J |) 1 1 g h |h| 1 | J |1 g J max{J , h} h g h , g J flow with RG J RG-flow • Transition persists: X-phase HGT 0 Random-domain clusters Hilbert-glass ln J ln h g J g h / g J g h Order in the Hilbert glass vs. T=0 Ferromagnet • Symmetry-broken T=0 Ferromagnetic state: ... ... Order parameter: ... ... or m GS nz GS • Typical Hilbert-Glass excited state: ... ... Order parameter: mmEA z n • Temporal correlations: z z nn 2 m (0) (t ) z n z n Order in the Hilbert glass vs. T=0 Ferromagnet mEA z n T FM Hilbert Glass transition QCP 2 typ ~ ... ... ... ... J h HGT PM J h m GS nz GS ... ... QCP ... ... J h T-tuned Hilbert glass transition: hJJ’ model • Quantum Ising model+J’: H J n nz nz1 hn nx J n ' nx nx1 1 z x 2 z x 3 • J’>0 increases h for low-energy states. z x 4 z x • But: No thermodynamic signatures J ' J , h 5 1 T X-states • T (or energy-density) tuned transition Hilbert glass J h RSRG-X Tree of states 1 0 T Energy Color code: inverse T 1 0 T RG step • Sampling method: Branch changing Monte Carlo steps. RSRG-X results for the Hilbert glass transition Flows for different temperatures: Complete phase diagram: ~ hJ Thermal conductivity • No thermodynamic signatures – only dynamical signatures exist. • Only energy is conserved: Signatures in heat conductivity? Engineering • assume scaling form: ( ) ~ Dimension: 3 | c | Numerical tests Summary + odds and ends • New universality: -T-tuned dynamical quantum transition. - No thermodynamic signatures. • Developed the RSRG-X - access to excitations and thermal averaging of L~5000 chains. • Excited states entanglement entropy: - ‘area law’ in both phases - log(L) at the Hilbert glass transition (Follows from GR, Moore, 2004) • Other Hilbert glass like transitions? Edwards-Anderson order parameter Lifshitz localization – a subtle example • Tight-binding electrons on an irregular lattice. Jn • Density of states: H n J n1 n1 J n n1 (E ) Pure chain: Random J: Ek 2J cosk 1 (E) ~ 4J 2 E 2 (E) ~ 1 E ln 3 E Dyson singularity E Method of attack: Real space RG Ma, Dasgupta, Hu (1979), Bhatt, Lee (1979) • Reduced the largest bond • Eliminated two sites. • New Heisenberg chain resulting with new suppressed effective coupling. J max J left J right J 2J max J right J left 2 3 1 1 2 3 2 3 2 4 Universality of emerging distribution functions D.S. Fisher (1994) J 1 2 3 4 5 6 • Functional flow and universal coupling distributions: (J ) J max J 7 8 Universality of emerging distribution functions D.S. Fisher (1994) 1 2 3 5 4 6 7 8 • Functional flow and universal coupling distributions: RG (J ) J g 1 1 g~ | ln J max | 0 J max J Random singlet phase D.S. Fisher (1995) 1 2 3 4 5 6 • Low lying excitations: excited long-range singlets: (E) ~ 7 8 (E ) 1 E ln 3 E • Susceptibility: 1 1 (T ) ~ ~ T ln 2 T / T0 T Dyson singularity again! Engtanglement entropy in the Heisenberg model B • Pure chain: A B L EAB How many qubits in A determined by B EAB TrA A log2 A (QFT Central charge, c=1) 1 log 2 L 3 Holzhey, Larsen, Wilczek (1994). Vidal, Latorre, Rico, Kitaev (2002). • Random singlet phase: Every singlet connecting A to B → entanglement entropy 1. EL number of singlets entering region A. Engtanglement entropy in the Heisenberg model B • Pure chain: A B L EAB How many qubits in A determined by B (CFT Central charge, c=1) 1 log 2 L 3 Holzhey, Larsen, Wilczek (1994). Vidal, Latorre, Rico, Kitaev (2002). • Random singlet phase: Effective central charge 1 1 EL ln L ln 2 log 2 L 3 3 crandom 1 ln 2 1 GR, Moore (2004). For the experts: Does the effective c obey a c-theorem? No… Examples of enropy increasing transitions in random non-abelian anyon chains. Fidkowski, GR, Bonesteel, Moore (2008). Universality at the transition? Altman, Kafri, Polkovnikov, GR (2009) MG BG J g 1 RSG RG (J ) Insulator g=1 superfluid 1 g0 (~ J ) J J max Mechanical analogy J1 J2 Average effective Stiffness ~ spring constant Jn = 1 / J 1 (ave of inverse J) 0 when g=1.