=E

Computational Techniques for Quantum Lattice Models

Andreas Luchli, New states of quantum matter MPI fr Physik komplexer Systeme - Dresden http://www.pks.mpg.de/~aml aml@pks.mpg.de Lecture Notes at http:www.pks.mpg.de/~aml/
INSTANS Summer School 2010 - Benasque - 29/6/2010

Computational Techniques for Quantum Lattice Models

Exact Diagonalization (Lecture 1, Tuesday)

| |

DMRG et al. (Lecture 2, Wednesday)

Classical & Quantum Monte Carlo + ALPS (Lecture 3, Thursday)

 =E
First Lecture: Introduction to Exact Diagonalization
Andreas Luchli, New states of quantum matter MPI fr Physik komplexer Systeme - Dresden http://www.pks.mpg.de/~aml aml@pks.mpg.de Lecture Notes at http:www.pks.mpg.de/~aml/
INSTANS Summer School 2010 - Benasque - 29/6/2010

Outline
Main Idea

=E

Examples of typical applications Structure of an Exact Diagonalization package

Hilbertspace, Symmetries Hamiltonian Linear Algebra Observables

Applications

Exact Diagonalization: Main Idea

Solve the Schrdinger equation of a quantum many body system numerically

=E
Sparse matrix, but for quantum many body systems the vector space dimension grows exponentially! Some people will tell you thats all there is. But if you want to get a maximum of physical information out of a nite system there is a lot more to do and the reward is a powerful:

Quantum Mechanics Toolbox

Exact Diagonalization: Applications

Quantum Magnets: nature of novel phases, critical points in 1D, dynamical correlation functions in 1D & 2D Fermionic models (Hubbard/t-J): gaps, pairing properties, correlation exponents, etc Fractional Quantum Hall states: energy gaps, overlap with model states, entanglement spectra Quantum dimer models or other constrained models (anyon chains, ...) Full Conguration Interaction in Quantum Chemistry

Exact Diagonalization: Present Day Limits

low-lying eigenvalues, not full diagonalization
Spin S=1/2 models: 40 spins square lattice, 39 sites triangular, 42 sites Honeycomb lattice 64 spins or more in elevated magnetization sectors up to 1.5 billion(=109) basis states with symmetries, up to 4.5 billion without t-J models: 32 sites checkerboard with 2 holes 32 sites square lattice with 4 holes up to 2.8 billion basis states Fractional quantum hall effect different lling fractions , up to 16-20 electrons up to 3.5 billion basis states Hubbard models 20 sites square lattice at half lling, 21 sites triangular lattice 22-25 sites in ultracold atoms setting w.o. spatial symmetries up to 160 billion basis states

Structure of an Exact Diagonalization code

Ingredients
Hilbert space
Basis represention, Lookup techniques Symmetries

Hamiltonian Matrix
Sparse Matrix representation (memory/disk) Matrix recalculation on the y (matrix-free)

Linear Algebra : Eigensolver / Time propagation

LAPACK full diagonalization Lanczos type diagonalization (needs only |v = H|u operations) More exotic eigensolver techniques, real oder imaginary-time propagation,

Observables
Static quantities (multipoint correlation functions, correlation density matrices,...) Dynamic observables (spectral functions, density of states,...) Real-time evolution

Hilbert Space

Basis representation
States of the Hilbert space need to be represented in the computer. Choose a representation which makes it simple to act with the Hamiltonian or other operators on the states, and to localize a given state in the basis Simple example: ensemble of S=1/2 sites in binary coding

| [1 1 0 1]2 = 13
detection of up or down spin can be done with bit-test. transverse exchange S + S + S S + can be performed by an XOR operation:

[1 1 0 1]2 XOR [0 1 1 0]2 = [1 0 1 1]2

initial cong bit 1 at the two sites coupled nal cong

For S=1, one bit is obviously not sufcient. Use ternary representation or simply occupy two bits to label the 3 states.

Basis representation
For t-J models at low doping it is useful to factorize hole positions and spin congurations on the occupied sites. For Hubbard models one can factorize the Hilbert space in up and down electron congurations. For constrained models - such as dimer models - the efcient generation of all basis states requires some thought.

One of the key challenges for a fast ED code is to nd the index of the new conguration in the list of all congurations (index f in Hi,f). Let us look at the example of S=1/2 spins at xed Sz

Basis lookup procedures (Lin tables)

One of the key problems for a fast ED code is to nd the index of the new conguration in the list of all congurations (index f in Hi,f).

[1 0 1 1]2 = 1110
But is 11 the index of this conguration in a list of all Sz=3/2 states ? no ! Use Lin tables to map from binary number to index in list of allowed states:
(generalization of this idea works for arbitrary number of additive quantum numbers)

Two tables with 2(N/2) [=sqrt(2N)] entries, one for MSBs and one for LSBs
[0 0] [0 1] [1 0] [1 1]
MSB

= = = =

X 0 1 2

[0 0] [0 1] [1 0] [1 1]
LSB

= = = =

X 0 1 0

Ind([0 1 1 1]) = 0 + 0 Ind([1 0 1 1]) = 1 + 0 Ind([1 1 0 1]) = 2 + 0 Ind([1 1 1 0]) = 2 + 1

= = = =

0 1 2 3

Basis lookup procedures (Lin tables)

Lookup can therefore be done with two direct memory reads. This is a time and memory efcient approach (at least in many interesting cases). An alternative procedure is to build a hash list [const access time] or to perform a binary search [log access time]. This becomes somewhat more involved when using spatial symmetries...

Symmetries
Consider a XXZ spin model on a lattice. What are the symmetries of the problem ?

H=
i,j

xy x x Ji,j (Si Sj

y y Si Sj )

z z z + Ji,j Si Sj

The Hamiltonian conserves total Sz, we can therefore work within a given Sz sector This easily implemented while constructing the basis, as we discussed before. The Hamiltonian is invariant under the space group, typically a few hundred elements. (in many cases = Translations x Pointgroup). Needs some technology to implement... At the Heisenberg point, the total spin is also conserved. It is however very difcult to combine the SU(2) symmetry with the lattice symmetries in a computationally useful way (non-sparse and computationally expensive matrices). At Sz=0 one can use the spin-ip (particle-hole) symmetry which distinguishes even and odd spin sectors at the Heisenberg point. Simple to implement.

change pa cally and experimentally. The exchange interactions in note = J 24 Mo72 Fe30 are quite small, J/kB 1.57 K , and this has The ma allowed for the experimental observation of a M = Ms /3 relevance t plateau at H Spatial Symmetries5.9 Tesla which has been explained tioned abo classically by Schrder et al.17 . In addition, this cluso deals with Spatial ter manifests a very broadreduction of Hilbert space symmetries are important for Inelastic Neutron Scattering the M = (INS) response as shown by Garlea et al.28 . On the dodecahed Symmetry resolved eigenstates teach us a lot about the physics at work, other hand, Mo72 V30 has a much stronger AFM exchange abatically 25,26 dispersion of excitations, symmetry thus is not well suited for the breaking tendencies, J/kB 250 K , and forth uud topological degeneracy, the eld-induced plateau. However, its ... observation of being well low-energy excitation spectrum can still be investigated 40 sites square lattice Icosidodecahedron (30 vertices) energy gap T PG =40 x 4 elements Ih:120 elements cial to the it must su can be ver low-tempe content at mentary p high order Refs. 9,10, der the cor The depen also found Our sec FIG. 1: (Color online) Schematic representation of the cuboc-

change pa cally and experimentally. The exchange interactions in note = J 24 Mo72 Fe30 are quite small, J/kB 1.57 K , and this has The ma allowed for the experimental observation of a M = Ms /3 relevance t plateau at H Spatial Symmetries5.9 Tesla which has been explained tioned abo classically by Schrder et al.17 . In addition, this cluso deals with ter manifests a very broad visible, use graph theoretical tools the M = Symmetries are sometimes not easily Inelastic Neutron Scattering (INS)symmetry group [nauty,by Garlea et al.28 . On the to determine response as shown grape]. dodecahed other hand, Mo72 V30 has a much stronger AFM exchange abatically 25,26 In an ED code a spatial symmetry operation not site permutation operation. J/kB 250 K , and thus is is a well suited for the forth uud (could become more complicated eld-induced plateau. However, sites) with spin-orbit interactions and multiorbital its observation of the being well low-energy excitation spectrum can still be investigated 40 sites square lattice Icosidodecahedron (30 vertices) energy gap T PG =40 x 4 elements Ih:120 elements cial to the it must su can be ver low-tempe content at mentary p high order Refs. 9,10, der the cor The depen also found Our sec FIG. 1: (Color online) Schematic representation of the cuboc-

Spatial Symmetries: Building the basis

Build a list of all allowed states satisfying the diagonal constraints, like particle number, total Sz, ... for each state we apply all symmetry operations and keep the state as a representative if it has the smallest integer representation among all generated states in the orbit. Example: 4 site ring with cyclic translation T, Sz=3/2 sector

T 1 ([0 1 1 1]) [1 0 1 1] T 2 ([0 1 1 1]) [1 1 0 1]

3

T 0 ([0 1 1 1]) [0 1 1 1]

T ([0 1 1 1]) [1 1 1 0]

T 2 ([1 0 1 1]) [1 1 1 0]
T 3 ([1 0 1 1]) [0 1 1 1]

T ([1 0 1 1]) [1 1 0 1]
1

T 0 ([1 0 1 1]) [1 0 1 1]

...

keep state

discard state

Spatial Symmetries: Building the basis

For one-dimensional representations of the spatial symmetry group: 1 (g)|g(r) | = r Bloch state N |G| gG Norm of the state is given as:
N = (r)
gG,g(r)=r

The norm (and therefore the state itself) can vanish if it has a nontrivial stabilizer combined with a nontrivial representation . Example: 4 site S=1/2 ring with cyclic translations:

K=0
Sz=2 Sz=1

|0 1 1 1 , N = 1 |0 1 0 1 , N = 2 z=0 S |0 0 1 1 , N = 1

|1 1 1 1 , N = 2

K = /2
|0 1 1 1 , N = 1 |0 0 1 1 , N = 1

K=
|0 1 1 1 , N = 1 |0 1 0 1 , N = 2 |0 0 1 1 , N = 1

24=16 1+1 4+4 2 4

The Hamiltonian Matrix

The Hamiltonian Matrix

Now that we have a list of representatives and their norms, can we r calculate the matrix elements of the Hamiltonian ? s|H| =? Let us look at an elementary, non-branching term in the Hamiltonian:

h |r = h (r)|s

r We can now calculate the matrix element s|h | without double expanding the Bloch states:

Ns s|h | = r (g )h (r) Nr

key algorithmic problem: given a possibly non-representative |s , how do we | , as well as a symmetry element g nd the associated representative s s relating |s to | ?

The Hamiltonian Matrix

key algorithmic problem: given a possibly non-representative |s , how do we s nd the associated representative | , as well as a symmetry element g s relating |s to | ? Brute force: loop over all symmetry operations applied on |s and retain | and g . This is however often not efcient (many hundred symmetries). s Prepare a lookup list, relating each allowed conguration with the index of its representative, and also the associated group element linking the two. Gives fast implementation, but needs a list of the size of the non spatiallysymmetrized Hilbert space. For specic lattices and models (Hubbard models) clever tricks exist which factorize the symmetry group into a sublattice conserving subgroup times a sublattice exchange. They give | fast, then a hash or binary search is s s needed to locate | in the list of representatives in order to get its index.

Hamiltonian Matrix Storage

Different possibilities exist: Store hamiltonian matrix elements in memory in a sparse matrix format Fast matrix vector multiplies, but obviously limited by available memory. Store hamiltonian matrix elements on disk in a sparse matrix format. In principle possible due to the vast disk space available, but I/O speed is much slower than main memory access times. Difcult to parallelize. Recalculate the hamiltonian matrix elements in each iterations on the y. Needed for the cutting edge simulations, where the whole memory is used by the Lanczos vectors. Can be parallelized on most architectures.

The Linear Algebra Backend

General facts

Linear Algebra: convergence of extremal (smallest or largest) eigenstates Fast Simple iterative algorithm Algorithm The most popular: Lanczos(only sparse MVM), low memory requirements
Developed by Cornelius Lanczos in the 1950s Belongs to the class of Krylov space methods Algorithm

Lanczos Algorithm (C. Lanczos,random |0 build a tridiagonal matrix with: Starting from 1950)
Three vector recursion

| = |

n = =

n+1 =

|n+1 = | /n+1 ,

0 H|n n |n1 , 1 0 n | , HN = | n |n , 0 | , || || = 0

1 1 2

0 2 2 .. .

............ 0 ..... 3 0 .. .. . . N2 0 N2 N1

0 0 0 N1 N1

.. 0 .......

Eigenvalues of HN converge rapidly towards eigenvalues of H. Once desired eigenvalue is converged, restart recursion and assemble the eigenvector.
very quick convergence for extremal eigenvalues !

Linear Algebra: Lanczos Algorithm

Once the ground state has converged, the vectors in the recursion tend to lose their orthogonality. As a consequence fake new eigenvalues show up in the approximate spectrum. These can be removed by heuristic techniques

Degeneracies of eigenvalues can not be resolved by construction. For this task one would need a band lanczos or the (Jacobi-)Davidson technique. However multiply degenerate eigenvalues are converged. Checkpointing is useful when performing large-scale simulations.

Full Diagonalization: Thermodynamics

Lapack / Householder complete diagonalization of the spectrum. Calculate partition function and all the thermodynamic quantities you want, often the only pedestrian method available for frustrated systems. Symmetries are also very important, because the computational requirements scale as O(D3), where D is the dimension of the block Hilbert space. Typical Ds for a workstation are a few 1000, up to a few 100000 on supercomputers.

F. Heidrich-Meisner, A. Honecker, T. Vekua, Phys. Rev. B 74, 020403(R) (2006).

Observables

Observables
In principle once can calculate any correlation function, since one has access to the full many body wave functions. When using spatial symmetries, the correlation functions need to be properly symmetrized too. Complicated correlation functions occur in frustrated systems:
Spin current correlations

Dimer-dimer correlations proportional to the square of the tomple S: its eective spectrum involves igenstates in each S sector, with eigenS(S + 1)/N .

displays such a tower of low lying levels m the other excitations. The symme(displayed in Tab. I) are those predicted ymmetry analysis; three soft modes at ) signal the full symmetry breaking of

N=40, =0.3

Frequency Dynamics
1 A| GA ( + i) = |A E0 + + i H

A = S (q), ck , . . .

Generate Krylov space of A| Use continued fraction used to invert (E0 + + i H)

Triangular Lattice Spin Dynamics dynamical spineld factor in zero structure Triangular lattice
20 15 10 5 0 20 15 10 5 0 N=36, 200 continued fraction iterations 20 20 26 / 2M/3 21 / M/3 15 15 10 5 0 20 15 10 5 0 10 5 0 20 27 15 10 5 0 30 / M

2 !/J

4 22 / K/2

2 !/J

2 !/J 28 / K

AML unpublished
0 2 !/J 4

2 !/J

2 !/J

Exact Diagonalization Real-Time Dynamics

It is expensive to obtain the full propagator

exp[itH]

Krylov methods exist to approximate the propagator for a given state |(0) One can get the time propagated state |(t) with only |v = H|u operations. Example: time evolution of a strongly correlated quantum systems after an abrupt change in the parameters in the Hamiltonian. Revivals and Relaxation.

0.9

(a) 1D Chain, Ui=2 J, Uf= 40 J

L=14, ED L=64, t-DMRG L=32, Uf=80 J 0 40 80 Frequency [J] 120

<b0b1>

0.6

0.3

0 0.9

(b) 2D Square, Ui=2 J, Uf=40 J

Nsites=16, ED Nsites=18, ED 0 40 80 Frequency [J] 120

<b0bx>

0.6

0.3

0.5

1.5
-1

Time t [J ]

C. Kollath, AML, E. Altman, PRL 2007

Exact Diagonalization: Applications

Quantum Magnets: nature of novel phases, critical points in 1D, dynamical correlation functions in 1D & 2D Fermionic models (Hubbard/t-J): gaps, pairing properties, correlation exponents, etc Fractional Quantum Hall states: energy gaps, overlap with model states, entanglement spectra Quantum dimer models or other constrained models (anyon chain..) Full Conguration Interaction in Quantum Chemistry

Tower of States spectroscopy

What are the nite size manifestations of a continuous symmetry breaking ? (eg in superuids/superconductors, magnetic order, spin nematic order) Order parameter is zero on a nite system ! (symmetric partition function) So usually one looks into order parameter correlations
(b)
z z

m d
x y x

Tower of States spectroscopy

What are the nite size manifestations of a continuous symmetry breaking ? (eg in superuids/superconductors, magnetic order, spin nematic order) Low-energy dynamics of the order parameter Theory: P.W. Anderson 1952, Numerical tool: Bernu, Lhuillier and others, 1992 Continuum

Energy

Dynamics of the free order parameter is visible in the nite size spectrum. Depends on the continuous symmetry group. U(1): (Sz)2 SU(2): S(S+1)

Magnons

Tower of States 1/N 1/L

S(S+1)

Symmetry properties of levels in the Tower states are crucial and constrain the nature of the broken symmetries.

Tower of States S=1 on triangular lattice

Bilinear-biquadratic S=1 model on the triangular lattice (model for NiGaS4).
H=
i,j

cos() Si Sj + sin() (Si Sj )

a)

SU(3) !!&

(b)
SU(3) !!" #!!"

!!& SU(3) !!"

AFQ

!!"

?
SO FM AFM FQ

#$%

FM AFM FQ

#$%

!!" "!!& SU(3)

"!!"

"#!!" SU(3) "!!&

"!!"

AML, F. Mila, K. Penc, PRL 06

Tower of States S=1 on triangular lattice: Antiferromagnetic phase

(b)
SU(3) !!"
SO AFM

SU(3) !!&

!!& SU(3) !!"

AFQ

#!!"

#$%

FM AFM FQ

#$%

=0 : coplanar magnetic order, 120 degree structure

"!!"

"#!!" SU(3) "!!&

"!!"

"!!& SU(3)

-5

! K other

36 35 34

Breaks translation symmetry. Tree site unit cell nontrivial momenta must appear in TOS non-collinear magnetic structure SU(2) is completely broken, number of levels in TOS increases with S Quantum number are identical to the S=1/2 case
AFQ, "=3$/8

-10

-15

33 32

-20 31

#$/2
42 -25

AFM, "=0
02 6 12 20 30 30

42 S(S+1) 6 12 20 02

30

42

Tower of States S=1 on triangular lattice: Ferroquadrupolar phase

(b)
(b)
SU(3) !!"
SO AFM

SU(3) !!&

!!& SU(3) !!"

AFQ

#!!"

=-/2 : ferroquadrupolar phase, nite quadrupolar moment, no spin order

x

#$%

FM AFM FQ

#$%

d
y x

"!!"

"#!!" SU(3) "!!&

"!!"

"!!& SU(3)

-60

-5

-65

36 No ! translation symmetry breaking. K only trivial momentum appears in TOS other 35 34 Ferroquadrupolar order parameter, only even S

-10

-70

-15 -75 -20

33

all directors are collinear SU(2) is broken 32 down to U(1), number of states in TOS is independent of S.
31

-80

FQ, "=#$/2
-85 02 6 12 20 30 -25

AFM, "=0
30 42 30

AFQ, "=3$/8
02 6 12 20 30 42

42 S(S+1) 6 12 20 02

Tower of States S=1 on triangular lattice: Antiferroquadrupolar phase

(b)
SU(3) !!"
SO AFM

SU(3) !!&

!!& SU(3) !!"

AFQ

(b)

#!!"

#$%

FM AFM FQ

#$%

=3/8 : antiferroquadrupolar phase, nite quadrupolar moment, no spin order, three sublattice structure.
x

d
y x

"!!"

"#!!" SU(3) "!!&

"!!"

"!!& SU(3)

36 35 34 33 32 31

Breaks translation symmetry. Tree site unit cell nontrivial momenta must appear in TOS Antiferroquadrupolar order parameter, complicated S dependence. Can be calculated using group theoretical methods.

"=0
42 30

AFQ, "=3$/8
02 6 12 20 30 42 S(S+1)

Exact Diagonalization: Applications

Quantum Magnets: nature of novel phases, critical points in 1D, dynamical correlation functions in 1D & 2D Fermionic models (Hubbard/t-J): gaps, pairing properties, correlation exponents, etc Fractional Quantum Hall states: energy gaps, overlap with model states, entanglement spectra Quantum dimer models or other constrained models (anyon chain..) Full Conguration Interaction in Quantum Chemistry

=1

A one-dimensional microscopic approach = 1/3

1 =1
= L2 Consider sample 1 = with lengths L1 ,1/3 1

= 1/3 Two dim electron gas 1 L1 , L 2

L1 , L 2 =1

choose
(Landau gauge)

Each electron state has xed area = 1/3 =1

1

= 1/3 1 L1 , L 2

(any shape is OK)

L1 , L 2

New States of Quantum Matter

 = 2/5, 4/11, 10/21, . . .

1 = 2m + 12 4 10 , , = = 1/3 . . . one-dimensional microscopic approach =1 2 4 10 5 11 121 = , , ... 5 11 21 e = e/(2m + 1) = 1 = 1/3 L e = e/(2m + 1) 1 , L2 1 = 1/2, 1/4, 3/8, . .. = 1/3 Each box is either empty 0 or lled 1 = 1/2, 1/4, 3/8, . . .

1 = = 1 2m + 1

= p/q L1 , L 2 = p/q x = p/q, q 001 q 1= p/q, 1 0 0 0 1 0 0 0 0 1 0

1 L ,L 1 2

k e

2 2 ik L y (xk L )2 /2 1

=1 (Landau gauge) = 1/3 1 L1 , L 2

A possible state at
(eg Coulomb V(r)=e /r)
2

1..0.....0..1 0..1.....1..0 k+m 1......1 k = 1/3 e = e/q

= Hamiltonianp/(2mp + 1) (ee-interaction) = p/(2mp + 1) 1..0.....0..1 0..1.....1..0

including

No kinetic energy!

Vk,m Vk,0

(all ee-terms that preserve position of CM)

k-m 1......1

(electrostatic repulsion)

= 1/3 e = e/q

Exact mapping of a single Landau level!

New States of Quantum Matter

y, as parameter . in the plane, with sh occurwe a consider aL disk The coecient , arising from a stem. The two phases ndary of length modeslarge compared to the L, localized near the boundary, is no ultraviolet divergent [2], but (where point may be charac(Topological) Entanglement Entropy ength. quantum is a universal additive tracing out m order; the In the ground state, by constant characte feature of the entanglement in the we c degrees of freedom the exterior of the disk,ground of infreedomreduced density matrices, and their entanglement entropiesentropy. in the topological entanglement ies Let ustwo at the look phases. arginal density This universal for the degrees ot be distinguished by operator quantity reects topolog of the entanglement that survive at arbit in the interior. The von therefore=canEbe entropy Neumann| studied usi Tr | tances, nsions a system with a System and log The quantum eld theory that captures the far-infrar of this density operator, a measure of der [1]. Environment S() = Tr[ log ] the medium, namely a topological quantu of a topologically orement of the interior and exterior variables, (TQFT) that describes the long-range A with remarkable conm For topologically ordered phases: interactions of the mediums massive qua particle excitations of tations. We nd tic variant of Perimeter/Area Law indistinrmore, in the innite = log D , S() = L (1) neracy depends on the + , osed surface on which D Total quantum dimension where D the Preskill quantum dim Topological entanglement entropy 1 is Kitaev &total PRL 06 medium, given by Levin & Wen PRL 06 ual properties emerge

Topological entanglement entropy for FQH states on the sphere

FQH model states are known to have topological order Can one extract based on the entanglement entropy ?
Haque, Zozulya & Schoutens, PRL 07
6 5
3 lA=4 2 0 0.05 0.1 lA=7
10
0

Sl (N ! ")

eigenvalue

A B

Sl (N)
A

4 3 2 1 0 0 -1 -2

0.15

10

!2

1/N

3
1/2

Sq. root of # orbitals in block, lA

FIG. 7: (Color creasing order fo block for N = 9,

Feasible, but tricky on the sphere. Complications due to varying length as a function of latitude

FIG. 6: (Color online.) Entanglement entropies in MooreRead state wavefunctions, extrapolated to the thermodynamic limit. Dashed line is a t to + c1 lA , with some points dropped. Inset plots SlA against 1/N for various xed lA .

also some error p tainty. Using th posed in the pre

How to do this on the torus

The torus can be tuned continuously by varying L1 and L2 (L1 L2 = 2 Ns).

L1

B
y

A
L2

x
y

We study orbital partitioning, which is expected to correspond to real space partitioning

L1

B
y

A
L2

Area law at constant L1

Degenerate wavefunctions averaging of entropies yields smooth curve !

1.4 1.2

SA(lA)

1 0.8 0.6 0
0.1

wave function 1 wave function 2 wave function 3 averaged S(lA)

y

L1

B
y

A
L2

L1=14

L1=10 6

(a) 9 12 15

L1=12

lA lA

L1=10 (b)
AML, Bergholtz & Haque, arXiv:1003.5656

Area law at constant L1

Increasing Ns (and thus L2) at constant L1 Saturation at large lA

function 1 function 2 function 3 ged S(lA)

L1=14

(a) 12 15

L1=12

L1=10

SA(lA)

B 2
L1=8
Ns=24 Ns=30

L1

B
y

A
L2

(b)

12

18

24

30

36

(c) 15 20 0 3 6 9 12 15 0

AML, Bergholtz & Haque, arXiv:1003.5656

lA

Entanglement entropy S(L1)

For large enough Ns, S(L1) converges for each L1
5 4
Ns=12 Ns=18 Ns=24 Ns=30 Ns=36 Ns=39

SA(L1)

3 2 1 0

(a)

=1/3, Laughlin

0.5

dSA/dL1

0.4 0.3 0.2 0.1 0 0

S(L1 ) = 2L1 2 + . . .

Boundary entropy density ()

(b)
2 4 6 8 10 12 14 16 18 20

L1

AML, Bergholtz & Haque, arXiv:1003.5656

Extracting the topological entanglement entropy

Use a running extraction, and monitor L1 convergence

SA(L1) - L1 dSA/dL1

-0.5 -1 -1.5 -2 0 (a) =1/3, Laughlin

-Ln(3)

-Ln(3)

S(L1 ) = 2L1 2 + . . .
(c)

-Ln(5)

=1/5, Lau

dSA/dL1

L1

-0.5 2 converges towards expected Ln(3) ! -1 -Ln(3) numerical determination for FQH states to date. Most accurate AML, -Ln(3) Bergholtz & Haque, arXiv:1003.5656 -1.5

FQH using DMRG ?

S(L) ~ L1 m ~ exp[S] ~ exp[a L1], exponential effort in the physical width of the system (see DMRG talk tomorrow for more details on this) DMRG will not work easily for large 2D FQH samples fermionic MERA/PEPS in suitable gauge ?

B B BAB

Entanglement Spectra (Li & Haldane PRL 08)

The entanglement entropy is a single number ! Is there more one can extract from the reduced density matrix ? One can always write

=: exp[HEntanglement ]
S=
i

A B

Entanglement Spectrum:

HEntanglement | = |

i exp(i )

Assuming that the entanglement Hamiltonian and the physical Hamiltonian are similar (e.g. as in free fermionic systems), then one expects to see features related to the open boundary structure in the spectrum of the reduced density matrix FQH states have interesting edge physics, visible in entanglement spectrum ?

se the boundary of the Landau-level partitioning indenes an edge shared by region A and B. he intuitive picture, the quantum entanglement beMoore-Read state on the sphere (Li & Haldane, PRL A and B arises from correlated quasihole excitations 08) s the boundary along which the partitioning is carried
Entanglement spectrum has dispersive structure

A B
Degeneracy at large momenta follows CFT counting rule (edge theory of the Pfafan is U(1)+Majorana)
3 11

Wen, PRL 93

30

!(NA=5)

20 Entanglement Spectra of FQH States: Torus 6

3 2 1 0

3 2

2 1 1 1

2 2 1

-3 -2 -1 0 1 2 3 10 - - - - - - The torus offers several advantages, e.g. a partition has two contributing edges. ---"K

ik komplexer Systeme, N thnitzer Strae 38, D-01187 Dresden, Germany o , University of Oslo, P.O. Box 1048 Blindern, 0316 Oslo, Norway (Dated: November 20, 2009)

pectrum of quantum Hall states on the torus and show that it is arranged in d by modes of two separated chiral edges with unusual dispersion. Strikingly, ll torus circumferences, which allows for a microscopic identication of the m by perturbing the solvable thin torus limit.

40

50

!(NA=6)

---------- -A-B - - - - Chiral Luttinger liquid(s) in Entanglement Spectrum ? ---------- -0 -------------30 ------- ---------15 - - - - - - - - - - ---- - ---------------------- - 10 -----------of --- states on the torus - - Laughlin in the entanglement spectra- - - --------& - ---20 , Juha Suorsa , Emil J. Bergholtz - - - - - Haque Masudul - and ---5 -- -- --

B-A

10 0
B-B A-A

ondensed matter wed from the eld explosive growth interest, there are ts provide physigh more convenns. One such rare ordered states and ich the use of bieal exotic physics

60

B
x

-18

-12

-6

12

18

two-dimensional the only experises. These states ntion due to quantopological proptates is that their hiral luttinger liqrplay of two such ectra.

AML, E. Bergholtz, J. Suorsa & M. Haque = 1/3, Laughlin WF PRL 104 156404 (2010) Figure 2: (Color online) Entanglement spectrum for Ns = 36. Left panels s
the total system has no edge. Refs. [4, 8] used FQH states

Figure 1: (Color online) Torus setup for block entanglement computations. The lowest Landau level is spanned by orbitals which in Landau gauge are centered along the circles shown. The arrows indicate the chiralities of the virtual edges created by the block partitioning.

"KA

Exact Diagonalization Literature

N. Laorencie & D. Poilblanc,
Simulations of pure and doped low-dimensional spin-1/2 gapped systems

Lecture Notes in Physics 645, 227 (2004). R.M. Noack & S. Manmana,
Diagonalization- and Numerical Renormalization-Group-Based Methods for Interacting Quantum Systems,

AIP Conf. Proc. 789, 93 (2005). A. Weisse, H. Fehske

Exact Diagonalization Techniques

Lecture Notes in Physics 739, 529 (2008). A. Luchli

Numerical Simulations of Frustrated Systems

to appear in Highly Frustratred Magnetism, Eds. Lacroix Mendels, Mila, (2010).

Thank you !