Tensor Network States: Algorithms and Applications

Dec. 5, 2014 @ Beijing

Tensor Network in Chemistry:

Recent DMRG/TTNS Studies and
Perspectives for Catalysis Research
Naoki Nakatani
Catalysis Research Center, Hokkaido University, Japan

Catalysts
Without Catalyst

Ea
E'a

With Catalyst
Reaction Coordinate

Reaction rate is given byk

Catalyst promotes chemical reaction by lowering the

energy barrier
However, reaction mechanism becomes much
Whats
complicated
happene
d?

Ea
RT

Theoretical understanding plays an important role for designing new catalysts

What Catalyst is desired?

Difficult to use but very cheap

Widely used but very expensive

Designing a new iron catalyst: You can get much money!!

DMRG-related Activity in Chemistry

Why we need DMRG?
Providing a convenient tool to doing DMRG for Chemistry
Get starting practical applications

Single-reference Ansatz (post-HF approach)

virtual

Common molecules, Mean-field approximation (i.e. Hartree-Fock theory) works very well

1 electron configuration covers more than 99% of electronic

energy for the ground state
The rest can be treated by a small perturbation

occupied

Many-body wavefunction can be spanned by taking particle

excitations from a reference configuration

HF cia Eia HF cijab Eia E bj HF L

HF

dominant

Eia HF

dominant

small perturbation

Multi-reference Ansatz (CASSCF approach)

Secondary

Defining Active Space (AS), orbital

subspace which can describe
important contributions
Computational cost: O(N!) for AS
AS is limited up to 14~16 orbs.

Inactive

Active Space

Transition metal complex, Mean-field approximation doesnt work anymore

CASSCF

C n1L nk n1 L nk

n1L nk

Inside AS:
dominant
CASSCF

Outside AS:
small
perturbation

1,, k AS

Practical DMRG Calculations in Quantum Chemistry

DMRG can be applied to solve CASSCF wavefunction

C 1234 Tr A1
A2 A3 A4 where Ai : M M
Computational cost:
O(M3k3+M2k4)/sweep

1234

Memory cost: O(M2k2)

Disk storage: O(M2k3)

QC-DMRG code can approach ca 100 sites (orbitals) calculation

AS limitation is considerably eased

DMRG-CASSCF Implementations
Several implementations of DMRG-CASSCF are available
Self-involved packages:

Used by
only few
groups

ORZ package: DMRG-CASSCF/CASPT2 in T. Yanai group (obtain upon request)

ORCA package: DMRG-CASSCF/NEVPT2 in F. Neese group (free for academic
use)
Most users (including me) dont like to use a new code because reading the
manual is tough!!
Its so hard for small groups to implement everything from scratch
(e.g. integrals, SCF, geometry opt., relativistic effect, etc)
Integration to the conventional (i.e. well-established) package is important

DMRG on Molcas package

Its not free software, but they already have a lot of users
&rasscf
symmetry= 1
spin= 1
nactel= 16 0 0 Normal CASSCF input in Molcas
inactive= 49 0
ras2= 0 16
lumorb
ciroot= 1 1 1
Only 1 line is required to carry out DMRG-CASSCF calc.
dmrg= 100
All other settings are automatically determined for
novice users

Hopefully, we can get many DMRG/TNS people in near future!!

Benchmark: DMRG-CASSCF/cu4-CASPT2
NC dependency of polyene ground state (AS : full -valence)
1.E+05

1.00

only CASSCF

0.80
0.60

1.E+04

EDMRG0.40
ECAS / mEh

CASSCF/CASPT2

0.20

1.E+03
Wall time / sec.

DMRG-CASSCF
(M = 100)

0.00
-0.20

DMRG-CASSCF/cu4-CASPT2
(M = 100)

-0.40

1.E+02

cu4-CASPT2

-0.60
-0.80

1.E+01
6

-1.00
8

10

12

14

16

18

20

NC

10

12

14

16

18

20

NC

DMRG-CASSCF/CASPT2 shows polynomial

Energy agrees very well with CAS

scaling

Negative error in CASSCF might be a numerical

DMRG-CASSCF scales better for (14e, 14o) and

error

Benchmark: DMRG-CASSCF Geometry Optimization

[Fe2S2(CH3)4]2

Density Functional Theory (BP86)

HSFe-Fe = 3.003 / S-S = 3.514
HS-LS Gap = 0.86 eV
LS

Fe

Fe

Fe-Fe = 2.716 / S-S = 3.526

CASSCF(10 elec., 10 orbs.)

HSFe-Fe = 3.076 / S-S = 3.567

HS-LS Gap = 0.06 eV

LS
Fe-Fe = 3.045 / S-S = 3.575

CASSCF(22 elec., 16 orbs.) performed by DMRG-CASSCF

HSFe-Fe = 3.065 / S-S = 3.551

HS-LS Gap = 0.12 eV
LS

Fe-Fe = 3.011 / S-S = 3.557

Tree Tensor Network States for QC

Whats the entanglement structure of molecule?
Efficient QC-DMRG algorithm on Tree graph lattice
Illustrative calculations

TTNS: Note for Structure

Two different types of TTNS were proposed: Here, focused on TPS type TTNS

Layered TTNS

TPS-type TTNS

TTNS where each site

has physical index

tr A

n1L nk

n1

L Ani L Ank n1 L nk

Entanglement Structure of Molecule

H2O / cc-pVDZ
23

24

22

Plotted exchange interaction b/w two

orbitals
(main source of entanglement)
4

21

20

6
7

19
18

Hamiltonian is highly non-local

1

H hij a
a

ij lk ai a j ak al

i
j
2 ijkl
ij

8
9

17
10

16
15

14

13

12

11

Hard to map entanglement onto 1Dlattice!

Somehow to map onto 1D-lattice for DMRG

F Kij Rij2
Heuristic algorithm to minimize a cost function
ij

17

13 12

24

18 22

19 10

23 16 21 14

15 20 11

Exchange interactions to be closer as possible

Even though, there still be a lot of non-local interaction!!
Can any Tensor Network approach give the better computational scaling?

TTNS: Renormalization along with Tree Graph

Biparticity of tree graph (i.e. no cycles) is useful to keep Canonical Form of
wavefunction
MPS

O(M3k3+M2k4) per sweep

Having the same super-block structure

TTNS

O(M4k3+M2k5) per sweep

LEFT

RIGHT

TTNS is one of natural ND-generalizations of MPS

Non-local interaction can be considered
without much increase of computational
scaling
Many algorithms for MPS can be reused for

DMRG Algorithm on TTNS

DMRG on MPS

ai

a j

DMRG on TTNS

ai a j

a j
ai

O(M3k2) per site

O(M4k2) per site

ai a j

2-Site Algorithm
Half-Renormalization (HR): To reduce the variational space to be searched for each
step
M

n1

n2

M2 4M

4M

M2 4M
4

n1

n2
M

4M

Reduced computational cost without loss of

accuracy!

What shape of tree should be used?

Minimum Spanning Tree (MST)

Minimum Entangled(?) Tree (MET)

Mmax = 410

21

Mmax = 48
18

14

12

19
15

24

23

24

11
20

16

12

23
11

2
4

3
10

8
13

19
17

22

E = 76.243652 (M = 200)
CPU: 1210 sec.

22

20
18

16

21
15

6
10

14

13

17

E = 76.243491 (M = 200)
CPU: 827 sec.

N2 molecule: Case where TTNS works better

Large entanglement due to triple bond breaking
R = 1.1208

R = 1.4288

R = 1.9050

The same accuracy can be obtained

from TTNS with less than half # states
in MPS
Computational scaling is also better in
TTNS

Cr2 molecule: Case where MPS works better

Strong correlation from sextuple bond in Cr2

Number of states can be reduced in TTNS

but the total performance scales better in MPS

Densritic molecule: Case where TTNS works extremely fine

Stilbenoid dendrimer: Model for Light-Harvesting system in Photosystem

2-4 times faster in TTNS

g = 2 (110e, 110o)

Summary
DMRG on Molcas package

DMRG code (G. Chan) has integrated into conventional QC program so that novice
users can carry out DMRG calculation much easier
Large active space (~ 100 orbs.) can be taken into account
Geometry optimization, perturbation treatment, etc. can be used without further
coding
Hoping we can get a lot of DMRG users in Chemistry

For me, now Ive got the useful tool to do catalysis research

Tree Tensor Network States (TTNS)

Introduced and developed efficient DMRG algorithm on TTNS for Quantum Chemistry
Performance is highly depending on system
So far, it can introduce another freedom to chose a lattice where entanglement
structure of molecule is to be mapped