• Aucun résultat trouvé

Geometrical constraints to reduce complexity in quantum molecular systems

N/A
N/A
Protected

Academic year: 2021

Partager "Geometrical constraints to reduce complexity in quantum molecular systems"

Copied!
2
0
0

Texte intégral

(1)

HAL Id: hal-02531863

https://hal.archives-ouvertes.fr/hal-02531863

Submitted on 3 Apr 2020

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Geometrical constraints to reduce complexity in quantum molecular systems

Thomas Pérez, Patrick Cassam-Chenaï

To cite this version:

Thomas Pérez, Patrick Cassam-Chenaï. Geometrical constraints to reduce complexity in quantum molecular systems. UCA Complex days 2019, Mar 2019, Nice, France. �hal-02531863�

(2)

Geometrical constraints to reduce complexity in quantum molecular systems

Thomas Perez

1

, Patrick Cassam-Chenaï

1

1 Laboratoire J. A. Dieudonné, UMR 7351 UCA-CNRS, Nice (tperez@unice.fr, cassam@unice.fr)

P

RESENTATION OF THE PROBLEM

The energy of a molecule is not the sum of the energies of its atomic components. Our aim is to compute an accurate approximation of this quantity based on an electron pair model, that is to say by using an

antisymmetric product of two-electron wave functions, called “geminals”. In this model, the total wave function ‰e D g1 ^    ^ gn can also be described by a set of matrices Ck D ci;jk 

1i;j m , one for each

geminal gk D P

1i;j m

ci;jk 'i ^ 'j , where .'i /1im (resp. .'j /1j m) is a basis orbital of spin C 12 (resp. 12 ).

However, without further restrictions, such a model has a factorial computational complexity with the number of electrons determined by the calculation of geminal product overlaps and its applicability is therefore limited to small systems. We will introduce generalized orthogonality constraints between gem-inals to reduce the computational effort, without sacrificing the indistinguishability of the electrons.

G

ENERAL GEMINAL PRODUCT OVERLAP FORMULA

Theorem 1 : For k 2 f1; : : : ; ng, let gk and gk0 be geminals whose associated matrices are denoted respectively Ck

and Ck0 . The overlap between the wave functions ‰e D g1 ^    ^ gn and ‰e0 D g10 ^    ^ gn0 is given by the formula :

h‰ej‰e0 i D hg1 ^    ^ gnjg10 ^    ^ gn0 i D X 0Nn;0;:::;Nn;nn n P iD0 Nn;i D n P iD0 iNn;iDn . 1/Nn;0 X ;02Sn n Y i D1 Nn;i Q j D1 TNNn;i n;0;:::;Nn;i 1 .;  0/ iNn;i N n;i Š with : T Nn;i Nn;0;:::;Nn;i 1.; 0/ D tr h CŽ  i 1 P pD0 pNn;pC.j 1/iC1 C0  0 i 1 P pD0 pNn;pC.j 1/iC1    CŽ  i 1 P pD0 pNn;pCj i C0  0 i 1 P pD0 pNn;pCj i i :

Example .n D 3/ : We compute hg1 ^ g2 ^ g3jg10 ^ g20 ^ g30 i D A1 C A2 C A3 , as follow :

I 3=0+0+3 : N3;0 D 2; N3;1 D 0; N3;2 D 0; N3;3 D 1 A1 D P ;02S3 tr h C .1/Ž C0  0 .1/C Ž  .2/C 0 .2/0 C Ž  .3/C 0 .3/0 i 3 . I 3=0+1+2 : N3;0 D 1; N3;1 D 1; N3;2 D 1; N3;3 D 0 A2 D P ;02S3 tr h C .1/Ž C0  0 .1/ i tr h C .2/Ž C0  0 .2/C Ž  .3/C 0 .3/0 i 2 . I 3=1+1+1 : N3;0 D 0; N3;1 D 3; N3;2 D 0; N3;3 D 0 A3 D P ;02S3 tr h C .1/Ž C0  0 .1/ i tr h C .2/Ž C0  0 .2/ i tr h C .3/Ž C0  0 .3/ i 6 .

S

CALING OF THE OVERLAP FORMULA FOR

H

m

LINEAR CHAINS

Hm molecules H6 H8 H10 H12 H14 H16

.n; number of partitions of n/ .3; 3/ .4; 5/ .5; 7/ .6; 11/ .7; 15/ .8; 22/

Number of terms in the sum (D .nŠ/2) 36 576 14 400 518 400 25 401 600 1 625 702 400

The computational cost rises too fast and makes the general model unpractical.

I

NTRODUCTION OF CONSTRAINTS FOR A SIMPLER ANSATZ

We will consider wave functions which are products of n singlet or triplet geminals (i.e. with symmetric or antisymmetric associated matrices) and we will impose to our geminals gk’s the so-called permutationally invariant 2-orthogonality constraints :

8i; j; k 2 f1; : : : ; ng distinct; (

hgi jgj i D 0

gk ê .gi ^ gj / D 0

i.e. in terms of matrices :

(

tr.CiŽCj / D 0

Ci CkŽCj C Cj CkŽCi D 0 : I The "maximal" linearly independent family of 2  2 matrices verifying these conditions is :

I2 D  1 0 0 1  I x D  0 1 1 0  I z D  1 0 0 1  and i y D  0 1 1 0  ; where x, y and z are the Pauli matrices.

E

XPRESSION OF THE OVERLAP FOR A SIMPLIFIED MODEL

Model : By considering integers h0 D 0 < h1 <    < hn, each matrix Ck (associated to gk) consists of a

1-orthogonal diagonal part of size hn  hn where only the coefficients of lines and columns hk 1 C 1; : : : ; hk

are nonzero, and of blocks jk Bkj of size 2  2 each proportionnal to I2, x, i y or z :

Ck D 0 B B B @ Dk 1k Bk1 : :: mk 0 Bkm0 1 C C C A with Dk D 0 B B B B B @

0

hk 1 k .hk 1C1/ : :: k hk

0

hn hk 1 C C C C C A :

The number of 2  2 blocks is m0 D m hn

2 2 N.

For all j , there is no more than one nonzero jk for each type of block (I2; x; i y; z).

Moreover, each Ck0 has the same matrix form as Ck (same types of blocks and zeros at the same places).

Theorem 2 : Let g1; : : : ; gn and g10 ; : : : ; gn0 be geminals verifying our constrained model conditions. We have :

hg1 ^    ^ gnjg10 ^    ^ gn0 i D X 0j1;:::;jnm0 distinct if nonzero gj1 1;g10     jn gn;gn0 ; with  j gu;gu0 D 8 ˆ ˆ < ˆ ˆ : hu P t Dhu 1C1 ut 0u t if j D 0 2 ju0uj otherwise

S

CALING OF THE CONSTRAINED MODEL FOR

H

m

LINEAR CHAINS

Hm molecules H6 H8 H10 H12 H14 H16

.n; hn; m0/ .3; 4; 1/ .4; 4; 2/ .5; 6; 2/ .6; 6; 3/ .7; 8; 3/ .8; 8; 4/

Number of terms in the sum 4 21 22 95 100 441

Références

Documents relatifs

Given a circuit C containing only majority gates, we can build in polynomial time an rMLP that is equivalent to C (as a Boolean function) and whose number of layers is equal to

In fact, the optimization problem is the following: given an initial data distribution, what is the best data partition, and the best mapping of this partition onto the processors,

In line with the present findings, we hypothesize that complex odorants (with a large variety of molecular features) will activate more olfactory receptors than low complexity

the above-mentioned gap between the lower bound and the upper bound of RL, guarantee that no learning method, given a generative model of the MDP, can be significantly more

Front-line therapy with rituximab added to the combination of cyclophosphamide, doxorubicin, vincristine and prednisone (Chop) significantly improves the outcome of

The convergence properties of Schwarz algorithms with continuously variable coefficients in each subdomain have already been studied using the technique of separation of variables

Note that this does not necessarily model all the things that would surprise the agent since there might still be things that the agent might conceive as possible with a low

In this paper, we will investigate whether the approximated optimal distributions of µ points regarding R2 based on different preference articulations [6] can be accurately