Ground State Projector QMC in the valence-bond basis

Similar documents
Quantum Monte Carlo Simulations in the Valence Bond Basis

Stochastic series expansion (SSE) and ground-state projection

Quantum Monte Carlo Simulations in the Valence Bond Basis. Anders Sandvik, Boston University

Quantum spin systems - models and computational methods

Non-magnetic states. The Néel states are product states; φ N a. , E ij = 3J ij /4 2 The Néel states have higher energy (expectations; not eigenstates)

Quantum Monte Carlo simulations of deconfined quantum criticality at. the 2D Néel-VBS transition. Anders W. Sandvik, Boston University

Numerical diagonalization studies of quantum spin chains

Exact diagonalization methods

Quantum and classical annealing in spin glasses and quantum computing. Anders W Sandvik, Boston University

Solving the sign problem for a class of frustrated antiferromagnets

Simulations of Quantum Dimer Models

Mind the gap Solving optimization problems with a quantum computer

Quantum Annealing in spin glasses and quantum computing Anders W Sandvik, Boston University

The Quantum Adiabatic Algorithm

Quantum s=1/2 antiferromagnet on the Bethe lattice at percolation I. Low-energy states, DMRG, and diagnostics

Spin liquid phases in strongly correlated lattice models

LPTM. Quantum-Monte-Carlo Approach to the Thermodynamics of Highly Frustrated Spin-½ Antiferromagnets. Andreas Honecker 1

Global phase diagrams of two-dimensional quantum antiferromagnets. Subir Sachdev Harvard University

Quantum Spin-Metals in Weak Mott Insulators

Quantum Monte Carlo Simulations of the Half-filled Hubbard Model. Anders F. J. Gabrielsson

Dimerized & frustrated spin chains. Application to copper-germanate

The Quantum Heisenberg Ferromagnet

Classical Monte Carlo Simulations

Matrix-Product states: Properties and Extensions

2. Spin liquids and valence bond solids

Measuring Entanglement Entropy in Valence Bond Quantum Monte Carlo Simulations

Numerical Studies of Adiabatic Quantum Computation applied to Optimization and Graph Isomorphism

arxiv: v1 [cond-mat.str-el] 17 Jan 2011

News on tensor network algorithms

SPIN LIQUIDS AND FRUSTRATED MAGNETISM

Mind the gap Solving optimization problems with a quantum computer

Mind the gap Solving optimization problems with a quantum computer

Introduction to Quantum Monte Carlo

Coupled Cluster Method for Quantum Spin Systems

Andreas Kreisel. Institut für Theoretische Physik Johann Wolfgang Goethe Universität Frankfurt am Main. July,

Gapless Spin Liquids in Two Dimensions

arxiv:quant-ph/ v2 24 Dec 2003

Numerical Studies of the Quantum Adiabatic Algorithm

Spin liquids on ladders and in 2d

Spin liquids in frustrated magnets

J. Phys.: Condens. Matter 10 (1998) L159 L165. Printed in the UK PII: S (98)90604-X

Quasi-1d Antiferromagnets

Spin liquids on the triangular lattice

Second Lecture: Quantum Monte Carlo Techniques

with four spin interaction Author(s) Tsukamoto, M.; Harada, K.; Kawashim (2009). doi: / /150/

arxiv:cond-mat/ v1 [cond-mat.str-el] 24 Jan 2000

SSH Model. Alessandro David. November 3, 2016

Kitaev honeycomb lattice model: from A to B and beyond

/N

quasi-particle pictures from continuous unitary transformations

Heisenberg Antiferromagnet on a Triangular Lattice* ABSTRACT

7 Monte Carlo Simulations of Quantum Spin Models

Magnetic ordering of local moments

Spinons and triplons in spatially anisotropic triangular antiferromagnet

Paramagnetic phases of Kagome lattice quantum Ising models p.1/16

Z2 topological phase in quantum antiferromagnets. Masaki Oshikawa. ISSP, University of Tokyo

Detecting collective excitations of quantum spin liquids. Talk online: sachdev.physics.harvard.edu

Stochastic Series Expansion Quantum Monte Carlo

Critical Spin-liquid Phases in Spin-1/2 Triangular Antiferromagnets. In collaboration with: Olexei Motrunich & Jason Alicea

Quasi-1d Frustrated Antiferromagnets. Leon Balents, UCSB Masanori Kohno, NIMS, Tsukuba Oleg Starykh, U. Utah

Quantum Phase Transitions in Low Dimensional Magnets

Emergent SU(4) symmetry and quantum spin-orbital liquid in 3 α-zrcl3

Winter School for Quantum Magnetism EPFL and MPI Stuttgart Magnetism in Strongly Correlated Systems Vladimir Hinkov

Disordered metals without quasiparticles, and charged black holes

From an Antiferromagnet to a Valence Bond Solid: Evidence for a First Order Phase Transition

Spinon magnetic resonance. Oleg Starykh, University of Utah

Scaling Theory. Roger Herrigel Advisor: Helmut Katzgraber

Frustration-free Ground States of Quantum Spin Systems 1

T S z 1,S z 2,..., S z N = S z N,S z 1,..., S z N 1

Quantum simulation with string-bond states: Joining PEPS and Monte Carlo

Ashvin Vishwanath UC Berkeley

Supersymmetry breaking and Nambu-Goldstone fermions in lattice models

Supplementary information I Hilbert Space, Dirac Notation, and Matrix Mechanics. EE270 Fall 2017

I. QUANTUM MONTE CARLO METHODS: INTRODUCTION AND BASICS

Luigi Paolasini

Nematicity and quantum paramagnetism in FeSe

Quantum-Classical Hybrid Monte Carlo Algorithm with Applications to AQC

Tensor network methods in condensed matter physics. ISSP, University of Tokyo, Tsuyoshi Okubo

Quantum disordering magnetic order in insulators, metals, and superconductors

Techniques for translationally invariant matrix product states

Valence Bonds in Random Quantum Magnets

H ψ = E ψ. Introduction to Exact Diagonalization. Andreas Läuchli, New states of quantum matter MPI für Physik komplexer Systeme - Dresden

VI.D Self Duality in the Two Dimensional Ising Model

The Mott Metal-Insulator Transition

Chiral Haldane-SPT phases of SU(N) quantum spin chains in the adjoint representation

Fully symmetric and non-fractionalized Mott insulators at fractional site-filling

Fermionic tensor networks

Hidden Symmetry and Quantum Phases in Spin 3/2 Cold Atomic Systems

Persistent spin current in a spin ring

Quantum Monte Carlo simulation of thin magnetic films

Quantum Phase Transitions

Quantum many-body systems and tensor networks: simulation methods and applications

Quantum phase transitions of insulators, superconductors and metals in two dimensions

Properties of monopole operators in 3d gauge theories

Quantum criticality of Fermi surfaces

LIBERATION ON THE WALLS IN GAUGE THEORIES AND ANTI-FERROMAGNETS

Symmetric Surfaces of Topological Superconductor

Quantum phases of antiferromagnets and the underdoped cuprates. Talk online: sachdev.physics.harvard.edu

Nonlinear Sigma Model(NLSM) and its Topological Terms

Time Evolving Block Decimation Algorithm

Transcription:

Quantum Monte Carlo Methods at Work for Novel Phases of Matter Trieste, Italy, Jan 23 - Feb 3, 2012 Ground State Projector QMC in the valence-bond basis Anders. Sandvik, Boston University Outline: The valence-bond basis Projector QMC with valence bonds Amplitude-product states J-Q chain: 1D valence-bond solid 1

Common bases for quantum spin systems Lattice of S=1/2 spins, e.g., Heisenberg antiferromagnet H = J S i S j = J Si z Sj z + (S + i S j + S i S+ j )/2 i,j i,j The most common basis is that of up and down spins = = S z = +1/2 = = S z = 1/2 One can also use eigenstates of two or more spins dimer singlet-triplet basis = ( )/ 2 = = ( + )/ 2 = The hamiltonian is more complicated in this basis 2

The valence bond basis for S=1/2 spins Valence-bonds between sublattice A, B sites Basis states; singlet products V r = N/2 b=1 (i rb, j rb ), r = 1,... (N/2)! (i, j) = ( i j i j )/ 2 The valence bond basis is overcomplete and non-orthogonal expansion of arbitrary singlet state is not unique = r f r V r (all fr positive for non-frustrated system) All valence bond states overlap with each other V l V r = 2 N N/2 N = number of loops in overlap graph Spin correlations from loop structure V l S i S j V r V l V r = 3 4 ( 1)x i x j +y i y j 0 (i,j in different loops) (i,j in same loop) A B More complicated matrix elements (e.g., dimer correlations) are also related to the loop structure K.S.D. Beach and A.W.S., Nucl. Phys. B 750, 142 (2006) V l V r V l V r 3

Projector Monte Carlo in the valence-bond basis Liang, 1991; Sorella et al. (1998); AWS, Phys. Rev. Lett 95, 207203 (2005) (-H) n projects out the ground state from an arbitrary state ( H) n =( H) n i S=1/2 Heisenberg model c i i c 0 ( E 0 ) n 0 H = S i S j = H ij, H ij = ( 1 4 S i S j ) i,j i,j Project with string of bond operators n H i(p)j(p) r 0 (r = irrelevant) {H ij } p=1 (a,d) Action of bond operators (a,b) (c,b) (c,d) H ab...(a, b)...(c, d)... =...(a, b)...(c, d)... H bc...(a, b)...(c, d)... = 1...(c, b)...(a, d)... 2 A B A B (i, j) = ( i j i j )/ 2 Simple reconfiguration of bonds (or no change; diagonal) no minus signs for A B bond direction convetion sign problem does appear for frustrated systems 4

Sampling the wave function Simplified notation for operator strings n H i(p)j(p) = P k, k = 1,... Nb n {H ij } p=1 k 6-site chain Simplest trial wave function: a basis state V r P k V r = W kr V r (k) The weight Wkr of a path is given by the number of off-diagonal operations ( bond flips ) nflip W kr = 1 2 nflip n = n dia + n flip H ab...(a, b)...(c, d)... =...(a, b)...(c, d)... H bc...(a, b)...(c, d)... = 1...(c, b)...(a, d)... 2 Note: all paths contribute - no dead (W=0) paths Sampling: Trivial way: Replace m (m 2-4) operators at random P accept = 1 2 n new flip n old flip The state has to be re-propagated with the full operator string More efficient updating scheme exists (later...) 5

Calculating the energy Using a state which has equal overlap with all VB basis states e.g., the Neel state N N V r = ( 2) N/2 E 0 = N H 0 N 0 = k N HP k V r k N P k V r H acts on the projected state nf = number of bond flips nd = number of diagonal operations E 0 = n d + n f /2 H 6

General expectation values: A = 0 A 0 Strings of singlet projectors P k = n p=1 H ik (p)j k (p), k =1,...,N n b (N b = number of interaction bonds) We have to project bra and ket states P k V r = W kr V r (k) ( E 0 ) n c 0 0 k k V l P g = V l (g) W gl 0 c 0 ( g g 6-spin chain example: E 0 ) n A = g,k V l P g AP k V r g,k V l P g P k V r = g,k W glw kr V l (g) A V r (k) g,k W glw kr V l (g) V r (k) V l A V r Monte Carlo sampling of operator strings 7

Sampling an amplitude-product state A better trial state leads to faster n convergence bond-amplitude product state [Liang, Doucot, Anderson, 1990] 0 = k N/2 b=1 h(x rb, y rb ) V k Update state by reconfiguring two bonds a c b d P accept = h(x c, y c )h(x d, y d ) h(x a, y a )h(x b, y b ) If reconfiguration accepted calculate change in projection weight used for final accept/reject prob. S. Liang [PRB 42, 6555 (1990)] used parametrized state amplitudes determined parameters variationally improved state by projection V l A V r 8

Variational wave function (2D Heisenberg) All amplitudes h(x,y) can be optimized [J. Lou and A.W.S., PRB 2007, AWS and H.-G. Evertz, PRB 2010] variational energy error 50% smaller than previously best (<0.1%) spin correlations deviate by less than 1% from exact values amplitudes decay as 1/r 3 0.16 0.14 M s 2 0.12 0.10 0 20 40 60 80-0.669 E/N -0.670-0.671-0.672 variational projected -0.673-0.674 0 20 40 60 80 L 9

More efficient ground state QMC algorithm larger lattices Loop updates in the valence-bond basis AWS and H. G. Evertz, PRB 2010 Put the spins back in a way compatible with the valence bonds (a i,b i )=( i j i j )/ 2 and sample in a combined space of spins and bonds A Loop updates similar to those in finite-t methods (world-line and stochastic series expansion methods) good valence-bond trial wave functions can be used larger systems accessible sample spins, but measure using valence bonds (as before) 10

τ V hai = (graphs by Ying Tang) h m m ( H) A( H) 1 h 1 ( H)2m 2 i 2i m m V V power m should be large enough to obtain ground state V 11

τ V n oper(n) 7 1 1 01 67 6 1 01 32 5 1 0 2 4 0 0 3 1 1 01 76 2 1 0 0 4 1 1 1 01 76 V 0 0 1 2 3 4 5 1 0 2 use bit operation to flip operators 12

Convergence Trial state expanded in H-eigenstates 0 i = X c n ni n Projected state after m-th power m i = H m 0 i = X n Expectation value hai m = h0 A 0i +2h1 A 0i c 1 hai m = h0 A 0i + c exp c n E m n ni c 0 E1 E 0 m N e 0 e 0 = E 0 /M, =E 1 E 0 m +... Conclusion: m/n >> e0/δ in valence-bond basis Δ is the singlet-singlet gap M s 2 E/N 0.08 0 2 4 6 8 10 trial state also can have fixed momentum k=0 (e.g., ampl. product state) - only k=0 excited states (gap) 0.14 0.12 0.10-0.667-0.668-0.669 32 32 Heisenberg p=2 p=3 p=4 optimized (1) optimized (2) -0.670 0 1 2 3 4 m/n 13

14

VBS states from multi-spin interactions Sandvik, Phys. Rev. Lett. 98, 227202 (2007) The Heisenberg interaction is equivalent to a singlet-projector C ij = 1 4 S i S j s s C ij ij = ij, C ij tm ij =0 (m = 1, 0, 1) we can construct models with products of singlet projectors no frustration in the conventional sense (QMC can be used) correlated singlet projection reduces antiferromagnetic order/correlations H = J Q 2 i j i j k l J Q 3 i j i j k l m n J X hiji C ij Q 2 X hijkli C ij C kl including all translations - H is translationally invariant The J-Q chains have the same critical-vbs transition as the J1-J2 Heisenberg chain! - Heisenberg SSE and projector codes can be easily adapted to Q-terms 15

S=1/2 Heisenberg chain with frustrated interactions (J1-J2 chain) = J 1 = J 2 Different types of ground states, depending on the ratio g=j2/j1 (both >0) Antiferromagnetic quasi order (critical state) for g<0.2411... - exact solution - Bethe Ansatz - for J2=0 - bosonization (continuum field theory) approach gives further insights - spin-spin correlations decay as 1/r C(r) = S i S i+r ( 1) r ln1/2 (r/r 0 ) r - gapless spin excitations ( spinons, not spin waves!) VBS order for g>0.2411... the ground state is doubly-degenerate state - gap to spin excitations; exponentially decaying spin correlations C(r) = S i S i+r ( 1) r e r/ - singlet-product state is exact for g=1/2 (Majumdar-Gosh point) 0 critical 0.241... VBS g 16

VBS state in J-Q chains (more in tutorial) Y. Tang and AWS, Phys. Rev. Lett. 107, 157201 (2011) S. Sanyal, A. Banerjee, and K. Damle, arxiv:1107.1493 J Q 2 i j i j k l critical VBS J Q 3 0 (Q/J)c Q/J i j i j k l m n dimer operator: B i = ~ S i ~S i+1 In a symmetry-broken VBS: hb i i = a + ( 1) i In a finite system in which the symmetry is not broken: <Bi>=0 detect VBS with dimer correlation function D(r) = 1 N NX hb i B i+r i i=1 This is a 4-spin correlation function can be evaluated using the transition graphs (1- and 2-loop contributions) expression in the afternoon tutorial 17

Animation of the projected states - transition graph Animations by Ying Tang J =0 18

J/Q =0.5 19

J/Q =(J/Q) c 6 20

Estimator for the singlet-triplet gap The original VB basis spans the singlet space with one triplet bond, one can obtain the lowest triplet state (i, j) = ( i j i j )/ 2 [i, j] = ( i j + i j )/ 2 Under propagation, the triplet flips like a singlet but a diagonal operation on a triplet kills it H bc...[a, b]...(c, d)... = 1...(c, b)...[a, d]... 2 H ab...[a, b]...(c, d)... = 0 The initial triplet can be placed anywhere N/2 different triplet propagations Those that survive contribute to E1 Partial error cancellations in the gap = E 1 E 0 The ability to generate singlet and triplet states in the same run is a unique feature of VB projector Monte Carlo 21

Singlet-triplet matrix elements It is also possible to project one singlet and one triplet matrix elements between the lowest singlet and triplet states e.g., magnon weight in dynamic structure factor T (q) Sq z S(0) 22